Kleiner, Mendel-Electroacoustics-CRC Press (2013)

ElEctroacoustics
ElEctroacoustics
Mendel Kleiner
Boca Raton London New York
CRC Press is an imprint of the

Taylor & Francis Group, an informa business
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2013 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 20130208
International Standard Book Number-13: 978-1-4665-9943-7 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts
have been made to publish reliable data and information, but the author and publisher cannot assume
responsibility for the validity of all materials or the consequences of their use. The authors and publishers
have attempted to trace the copyright holders of all material reproduced in this publication and apologize to
copyright holders if permission to publish in this form has not been obtained. If any copyright material has
not been acknowledged please write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmit-
ted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented,
including photocopying, microfilming, and recording, or in any information storage or retrieval system,
without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.
com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood
Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and
registration for a variety of users. For organizations that have been granted a photocopy license by the CCC,
a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used
only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at

http://www.crcpress.com
For my parents, Nacha and Szlama, of blessed memory
Contents
Acknowledgments....................................................................................................xxi
Author................................................................................................................... xxiii
List of Symbols.......................................................................................................xxv
Chapter 1 Introduction........................................................................................... 1
1.1 Prerequisites............................................................................... 2
1.2 Other Books and References......................................................3
References............................................................................................. 3
Chapter 2 Introduction to Electroacoustic Systems...............................................5

2.1 Recording................................................................................... 5
2.2 Reproduction..............................................................................6
2.3 Linearity.....................................................................................7
Chapter 3 Sound and Its Properties..................................................................... 11

3.1 Sound Waves............................................................................ 11
3.1.1 Equation of Continuity................................................ 11
3.1.2 Equation of Motion..................................................... 11
3.1.3 Thermodynamic Properties......................................... 12
3.1.4 Wave Equation............................................................. 12
3.2 Plane Wave Solutions to the Wave Equation............................ 13
3.3 Frequency and Time Domains................................................. 13
3.3.1 jω-Method................................................................... 14
3.4 Impedance................................................................................ 15
3.4.1 Common Impedance Definitions................................ 15
3.4.2 Sound Field Impedances in a Plane Wave.................. 16
3.5 Solutions to the Wave Equation: Spherical Waves................... 17
3.5.1 Sound Field Impedance in a Spherical Wave.............. 18
3.6 Sound Intensity and Sound Power............................................20
3.7 Propagation Losses...................................................................20
3.8 Elementary Sound Sources....................................................... 22
3.8.1 Monopoles................................................................... 22
3.8.2 Power Radiated by a Monopole................................... 23
3.8.3 Dipoles.........................................................................24
3.8.4 Quadrupoles................................................................25
3.9 Reflection and Transmission at Boundaries.............................25
3.9.1 Perpendicular Sound Incidence...................................26
3.9.2 Reflection of Sound at an Angle.................................. 27
vii
viii Contents
3.9.3 Normal Reaction.........................................................28

3.9.4 Extended Reaction....................................................... 29
3.10 Huygens’ Principle................................................................... 30
3.11 Scattering.................................................................................. 30
3.12 Diffraction................................................................................ 31
3.13 Acoustic Reciprocity................................................................34
Review Questions................................................................................34
Problems.............................................................................................. 35
References........................................................................................... 37
Chapter 4 Waves in Membranes and Plates......................................................... 39

4.1 Introduction.............................................................................. 39
4.2 Wave Types in Infinite Media..................................................40
4.2.1 Longitudinal Waves.....................................................40
4.2.2 Transverse Waves........................................................ 41
4.3 Wave Types in Media of Limited Extension............................ 42
4.3.1 Quasi-Longitudinal Waves in Plates........................... 42
4.3.2 Out-of-Plane Vibration................................................ 43
4.3.3 Membrane Vibration................................................... 43
4.4 Transverse Waves in Thin Bars and Plates............................... 47
4.4.1 Shear Waves................................................................ 47
4.4.2 Bending Waves in Plates............................................. 48
4.5 Audibility of Resonance Characteristics.................................. 53
4.6 Sandwich Sheets....................................................................... 53
4.7 Vibration in Lossy Plates.......................................................... 55
4.7.1 Damping by Viscoelastic Layers................................. 56
4.7.2 Viscoelastic Materials................................................. 57
4.7.3 Free Layers.................................................................. 58
4.7.4 Constrained Layers...................................................... 59
4.7.5 Damping by Sand and Other Lossy Materials............ 59
Review Questions................................................................................60
Problems.............................................................................................. 61
References........................................................................................... 63
Chapter 5 Circuits and Circuit Components........................................................ 65

5.1 Introduction.............................................................................. 65
5.2 Linearity................................................................................... 65
5.3 Circuit Analysis Principles.......................................................66
References........................................................................................... 68
Chapter 6 Electromechanical Analogies.............................................................. 69

6.1 Continuous and Discretized Bodies......................................... 69
6.2 Mechanical Elements: Electrical Analogies............................. 69
6.2.1 Representation of Mechanical Components................ 69
Contents ix
6.3 Mechanical Impedance and Mobility....................................... 70

6.4 Mechanical Resistance............................................................. 71
6.5 Mechanical Compliance........................................................... 72
6.6 Mass.......................................................................................... 73
6.7 Levers....................................................................................... 74
6.8 Generators................................................................................ 76
6.9 Power Relationships................................................................. 77
Review Questions................................................................................ 77
Problems.............................................................................................. 78
References........................................................................................... 81
Chapter 7 Electroacoustical Analogies................................................................ 83

7.1 Acoustical Circuit Elements..................................................... 83
7.2 Waves in Tubes......................................................................... 83
7.3 Acoustic Impedance.................................................................84
7.4 Acoustic Capacitance............................................................... 85
7.4.1 Impedance Analogy.................................................... 85
7.4.2 Admittance Analogy................................................... 89
7.5 Acoustic Capacitance Due to a Stiff Diaphragm..................... 89
7.6 Acoustic Mass........................................................................... 89
7.6.1 Lined Tube...................................................................94
7.7 Length-End Corrections...........................................................96
7.7.1 Acoustic Resistance..................................................... 98
7.7.2 Resistance of Thick Sheets........................................ 100
7.7.3 Resistive Terminations.............................................. 101
7.8 Acoustic Transformers........................................................... 102
7.8.1 Abrupt Change of Cross-Sectional Area................... 102
7.8.2 Exponential Couplers................................................ 103
7.8.3 Quarter-Wave Transformer........................................ 105
7.9 Acoustic Generators............................................................... 106
7.10 Power Relationships............................................................... 107
7.11 Filters...................................................................................... 107
7.11.1 Low-Pass Filters........................................................ 108
7.11.2 Band-Pass and Band-Reject Filters........................... 109
7.11.3 High-Pass Filters....................................................... 113
7.12 Using Acoustical Analogies with Free Waves........................ 114
7.12.1 Plane Wave Incident on a Plane Sheet
Having Mass...........................................................115
Review Questions.............................................................................. 117
Problems............................................................................................ 117
References......................................................................................... 119
Chapter 8 Conversion between Analogies......................................................... 121

8.1 Impedance and Admittance Analogies.................................. 121
8.2 Conversion between Analogies.............................................. 121
x Contents
8.3 “Dot” Method......................................................................... 122

8.4 Transformation between Mechanical
and Acoustical Circuits.......................................................... 124
Problems............................................................................................ 125
References......................................................................................... 128
Chapter 9 Transducer Operating Principles....................................................... 129

9.1 Introduction............................................................................ 129
9.1.1 Reversible and Nonreversible Transducers............... 129
9.1.2 Direct and Indirect Conversion................................. 129
9.1.3 Active and Passive Transducers................................ 130
9.2 Transducer Operating Blocks................................................. 130
9.3 Conversion.............................................................................. 131
9.4 Electrodynamic Transducers.................................................. 132
9.5 Electromagnetic Transducer................................................... 133
9.5.1 Unbalanced Transducers........................................... 133
9.5.2 Balanced Transducers............................................... 137
9.6 Electroresistive Transducers................................................... 139
9.7 Capacitive Transducers........................................................... 142
9.7.1 Direct Use of Capacitance......................................... 143
9.7.2 Capacitive Microphones Using Static Electric
Charge with Externally Supplied Electrical Bias...... 144
9.7.2.1 Unbalanced Operation, Static
Conditions................................................144
9.7.2.2 Unbalanced Operation, Dynamic
Conditions.................................................. 147
9.7.2.3 Capacitive Microphones Using
Permanent Charge (Electret Bias)............. 149
9.7.3 Loudspeaker Operation............................................. 149
9.7.3.1 Unbalanced and Biased............................. 149
9.7.3.2 Unbalanced Electret Charge...................... 151
9.7.4 Transducer Electrical Analogy.................................. 151
9.7.5 Loudspeakers Using External Electric
Charge, Balanced...................................................... 151
9.8 Piezoelectric Transducers....................................................... 153
9.8.1 Piezoelectric Coupling in the Quartz Crystal........... 154
9.8.2 Electromechanical Relationships.............................. 154
9.8.3 Transducer Electrical Analogy.................................. 157
9.8.4 Piezoelectric Ceramics.............................................. 157
9.8.5 Piezoelectric Films.................................................... 158
9.9 Magnetostrictive Transducers................................................ 159
References......................................................................................... 162
Contents xi
Chapter 10 Radiation and Impedance.................................................................. 163

10.1 Introduction............................................................................ 163
10.2 Radiation of Sound and Power Loss....................................... 163
10.3 Sound Radiation Characterization......................................... 164
10.4 Radiation Ratio....................................................................... 165
10.5 Radiation Impedance.............................................................. 165
10.6 Vibrating Plane and Sound Field Intensity............................. 166
10.7 Power Radiated into an Infinitely Long Tube......................... 167
10.8 Impedance Matching.............................................................. 168
10.9 Fundamental Sources............................................................. 168
10.9.1 Monopoles................................................................. 168
10.9.2 Dipoles....................................................................... 172
10.9.3 Quadrupoles.............................................................. 172
10.9.4 Oscillating Sphere..................................................... 173
10.10 Pistons..................................................................................... 174
10.10.1 Free Circular Piston.................................................. 174
10.10.2 Circular Piston in a Baffle......................................... 175
10.10.3 Elliptical Piston in a Baffle....................................... 178
10.10.4 Rectangular Piston in a Baffle................................... 179
10.10.5 Circular Piston Radiator at the End of a Long Tube...... 181
10.11 Transverse Waves in Plates..................................................... 183
10.11.1 Bending Waves in an Infinite Sheet.......................... 184
10.11.2 Bending Waves in Damped Sheets............................ 185
10.11.3 Bending Waves in Finite Sheets................................ 186
10.11.4 Sound Radiation by Bending Wave Point
Excitation................................................................. 190
10.11.5 Sound Field Close to Nonradiating Bending
Wave Fields............................................................... 191
10.11.6 Radiation Factor versus Radiation Resistance.......... 191
10.12 Radiation Impedance as a Low-Pass Filter............................ 192
10.13 Finite Element and Boundary Element Methods................... 192
10.14 The Radiation Impedance of Circular Pistons....................... 193
Problems............................................................................................ 196
References......................................................................................... 197
Chapter 11 Sound Source and Acoustic Environment......................................... 199

11.1 Reflecting Surfaces and Radiation Impedance...................... 199
11.2 Single Rigid Plane Surface..................................................... 199
11.3 Multiple Surfaces....................................................................202
11.4 Power Output of Dipoles near Reflecting Surfaces................204
11.5 Room Modes...........................................................................206
11.6 Mutual Impedance..................................................................208
xii Contents

Problems............................................................................................ 211
References......................................................................................... 213
Chapter 12 Directivity......................................................................................... 215

12.1 Introduction............................................................................ 215
12.2 Directivity Functions and Directivity Plots............................ 215
12.3 Reciprocity............................................................................. 218
12.4 Monopole on a Rigid Baffle................................................... 218
12.5 Near-Field and Far-Field......................................................... 219
12.6 Near-Field of a Piston in a Baffle........................................... 220
12.7 Fresnel Zone of a Piston in a Baffle....................................... 220
12.8 Far-Field of a Piston in a Baffle..............................................224
12.9 Directivity and Directivity Index........................................... 226
12.10 Directivity and Frequency Response...................................... 227
12.11 Far-Field of a Piston at the End of a Long Tube..................... 229
12.12 Near-Field and Far-Field Frequency Response
of a Circular Piston................................................................. 229
12.13 Acoustic Center...................................................................... 230
12.14 Arrays..................................................................................... 231
12.14.1 Classifying Array Systems........................................ 232
12.14.2 Directional Properties of Array Transducers............ 232
12.15 Array Transfer Functions....................................................... 233
12.15.1 Array Factor, Wavelength, and
Inter-Element Distance.............................................. 233
12.16 Continuous Linear Arrays......................................................240
12.17 Polynomial Expansion of the Array Factor............................ 242
12.18 Wide Frequency Range Arrays.............................................. 243
12.18.1 Harmonically Nested Arrays....................................244
12.19 Signal-to-Noise Ratio in Receiver Arrays.............................. 245
12.20 Audible Artifacts of Large Arrays......................................... 245
12.21 Acoustic Lenses......................................................................246
Review Questions..............................................................................248
Problems............................................................................................248
References......................................................................................... 251
Chapter 13 Microphones and Sound Fields......................................................... 253

13.1 Introduction............................................................................ 253
13.2 Influence of the Microphone on the Sound Field................... 253
13.3 Pressure Sensing..................................................................... 256
13.4 Pressure-Gradient Sensing..................................................... 256
13.4.1 Principles................................................................... 256
13.4.2 Plane Waves............................................................... 257
13.4.3 Spherical Waves........................................................ 259
13.4.4 Proximity Effects...................................................... 259
Contents xiii
13.5 Two Ways to Achieve Directivity...........................................260

13.6 Common Microphone Directivity Patterns............................ 261
13.6.1 Gradient Order.......................................................... 261
13.6.2 Combinations of Gradient Order............................... 261
13.6.3 Unidirectional Microphones..................................... 263
13.7 Directivity Function and Directivity Index............................ 263
13.8 Representation of Scattering Using
Electroacoustical Circuits.......................................................264
Problems............................................................................................ 267
References......................................................................................... 268
Chapter 14 Microphones...................................................................................... 269

14.1 Introduction............................................................................ 269
14.2 Diaphragms and Membranes.................................................. 269
14.2.1 Membranes................................................................ 269
14.2.2 Diaphragms............................................................... 270
14.3 Microphone Analogies........................................................... 271
14.3.1 Pressure Microphones............................................... 271
14.3.2 Transducer Response Alternatives for the
Pressure Microphone................................................. 273
14.3.3 Pressure-Gradient Microphones................................ 273
Pressure-Gradient Microphone................................. 274
14.3.5 Combination of Pressure and
Pressure-Gradient Microphones................................ 275
14.4 Electrodynamic Transducers.................................................. 279
14.4.1 Moving Coil Microphones........................................ 279
14.4.2 Ribbon Microphones................................................. 283
14.5 Electromagnetic Microphones................................................ 285
14.6 Piezoelectric and Ferroelectric Transducers.......................... 286
14.7 Condenser Microphones......................................................... 287
14.7.1 Electroacoustic Analogies......................................... 288
14.7.2 Permanent Charge Condenser Microphones............. 289
14.8 Electrical Characteristics and Requirements......................... 289
14.8.1 Distortion................................................................... 292
14.8.2 Microphone Noise..................................................... 293
Problems............................................................................................ 295
References......................................................................................... 297
Chapter 15 Electrodynamic Loudspeaker Drivers.............................................. 299

15.1 Introduction............................................................................ 299
15.2 Moving-Coil Drivers.............................................................. 299
xiv Contents
15.3 Magnet Air Gap and Voice Coil............................................. 301

15.4 Diaphragms............................................................................ 303
15.4.1 Shape......................................................................... 303
15.4.2 Materials....................................................................306
15.4.3 Supports and Surrounds............................................307
15.5 Electroacoustic Analogies......................................................309
15.5.1 Mechanical System...................................................309
15.5.2 Acoustical System..................................................... 310
15.5.3 Transduction Mechanism.......................................... 311
15.5.4 Complete Circuit....................................................... 311
15.6 Frequency Response............................................................... 314
Problems............................................................................................ 315
References......................................................................................... 317
Chapter 16 Baffle and Box................................................................................... 319

16.1 Aerodynamic Short Circuit.................................................... 319
16.2 Infinite Baffles........................................................................ 321
16.2.1 Far-Field Sound Pressure.......................................... 321
16.3 Finite Baffles.......................................................................... 322
16.4 Closed-Box Enclosures........................................................... 323
16.4.1 Electroacoustic Analogies......................................... 323
16.4.2 Transfer Function...................................................... 325
16.4.3 Resonance Frequency................................................ 326
16.4.4 Q Factor and Frequency Response............................ 327
16.4.5 Front Radiation and Baffle Effect............................. 328
16.5 Practical Closed-Box Loudspeakers....................................... 329
16.5.1 Acoustic Suspension.................................................. 331
16.5.2 Internal Resonance and Modes................................. 332
16.6 Power and Efficiency.............................................................. 333
Problems............................................................................................ 335
References......................................................................................... 336
Chapter 17 Vented Box Loudspeakers................................................................. 337

17.1 Extended Low-Frequency Response...................................... 337
17.2 Loudspeaker as High-Pass Filter............................................ 337
17.3 Ported Box and Drone Cone Designs..................................... 338
17.4 Frequency Response Using Classical Filter Theory............... 341
17.5 Bandpass Designs................................................................... 342
17.6 External Filters.......................................................................344
17.7 Driver Cone Excursion...........................................................346
Problems............................................................................................ 347
References......................................................................................... 349
Contents xv
Chapter 18 Transmission Line Loudspeakers...................................................... 351

18.1 Introduction............................................................................ 351
18.2 Attenuation by Absorptive Fill and Lining............................ 352
18.3 Attenuation by Folds............................................................... 354
18.4 Circuit Analogies for Ducts.................................................... 355
18.4.1 Quarter-Wave Resonator........................................... 355
18.4.2 Discrete Component Analogies................................. 356
18.4.3 Anechoic Termination...............................................360
18.5 Special Considerations........................................................... 362
18.6 Dual Ported Transmission Line Loudspeakers...................... 362
Problems............................................................................................ 363
References.........................................................................................364
Chapter 19 Horns................................................................................................. 367

19.1 Introduction............................................................................ 367
19.2 Horn Equations....................................................................... 367
19.2.1 Horn Terminology..................................................... 367
19.2.2 Webster’s Horn Equation........................................... 368
19.2.3 Common Horn Expansion Functions........................ 369
19.3 Exponential Horn................................................................... 371
19.3.1 Wave Propagation and Cutoff................................... 371
19.3.2 Throat Impedance..................................................... 372
19.4 Conical Horns......................................................................... 373
19.5 Hyperbolic Horns................................................................... 374
19.6 Comparison of Horn Characteristics...................................... 374
19.7 Tractrix Horns........................................................................ 376
19.8 Finite Horns............................................................................ 377
19.9 Horn Directivity..................................................................... 380
19.9.1 Horns Using Combinations of Flares........................ 380
19.9.2 Multicell Horns......................................................... 380
19.9.3 Radial Horns............................................................. 382
19.9.4 Waveguide Horns...................................................... 382
19.9.5 Horn Arrays............................................................... 382
19.10 Horn and Driver...................................................................... 384
19.10.1 Low-Frequency Horns............................................... 384
19.11 Higher-Order Modes in Horns............................................... 385
19.12 Circuit Analogies for Horn Loudspeakers............................. 385
19.12.1 Efficiency................................................................... 386
19.12.2 Low Frequencies....................................................... 387
19.12.3 High Frequencies....................................................... 387
19.13 Stepped and Piecewise Linear Horns..................................... 388
19.14 Folded, Bent, and Coiled Horns............................................. 388
19.15 Horn Phase Plugs.................................................................... 389
19.16 Acoustic Center of Horns....................................................... 390
xvi Contents
19.17 Linear and Nonlinear Distortion............................................ 391

19.18 Horn-Shaped Connectors....................................................... 392
19.19 Horns and Room Acoustics.................................................... 392
19.20 Summary................................................................................ 393
Problems............................................................................................ 394
References......................................................................................... 394
Chapter 20 Gradient Loudspeakers..................................................................... 397

20.1 Introduction............................................................................ 397
20.1.1 Size and Multipole Approaches................................ 397
20.1.2 Gradient Loudspeaker Types.................................... 397
20.2 Use of Gradient Loudspeakers............................................... 398
20.2.1 Indoors: Low Frequencies......................................... 398
20.2.2 Indoors: Mid and High Frequencies.......................... 399
20.3 First-Order Gradient Sources................................................. 399
20.3.1 Bidirectional Array................................................... 399
20.3.2 Cardioid Directivity Arrays......................................400
20.3.3 “Acoustic Resistance” Box........................................400
20.4 Second-Order Gradient Sources.............................................404
Problems............................................................................................405
References.........................................................................................405
Chapter 21 Drivers Using Flexible Diaphragms..................................................407

21.1 Introduction............................................................................407
21.2 System Considerations............................................................407
21.3 Diaphragm Wave Fields.........................................................409
21.4 Diaphragm Sound Radiation.................................................. 412
21.4.1 Aerodynamic Cancellation........................................ 412
21.4.2 Modal Sound Radiation............................................ 412
21.5 Driving Point Impedance....................................................... 414
21.6 Electroacoustic Circuit Analogies.......................................... 416
21.6.1 Two Sliding Masses Driven by a Force..................... 416
21.6.2 Exciter Driving a Resonant Diaphragm.................... 417
21.7 Resonance and Sound Quality............................................... 418
Problems............................................................................................ 420
References......................................................................................... 421
Chapter 22 Multiway Loudspeakers.................................................................... 423

22.1 Introduction............................................................................ 423
22.1.1 Bandwidth................................................................. 423
22.1.2 Example..................................................................... 423
Contents xvii
22.2 Diaphragm Dimensions and Wavelength............................... 424

22.3 Loudspeaker Polarity, Phase, and Group Delay..................... 425
22.4 Placement of Drivers.............................................................. 426
22.4.1 Baffle Effect.............................................................. 427
22.4.2 Delay......................................................................... 428
22.4.3 Directivity................................................................. 428
22.4.4 Concentric Drivers.................................................... 429
22.5 Thermal and Linearity Aspects.............................................. 430
22.6 Loudspeaker and Listening Environment.............................. 430
22.7 Crossover Filters..................................................................... 431
22.7.1 System Considerations.............................................. 431
22.7.2 High- and Low-Impedance Active and Passive
Filters......................................................................... 432
22.7.3 Large-Signal Filters................................................... 433
22.7.4 Driver Electric Impedance and Zobel Networks...... 434
22.7.5 Small-Signal Filters................................................... 434
22.8 Summary................................................................................ 435
Problems............................................................................................ 436
References......................................................................................... 437
Chapter 23 Active Loudspeakers......................................................................... 439

23.1 Introduction............................................................................ 439
23.2 Loudspeaker Sound Field Characterization........................... 439
23.2.1 Transfer Function Measurement................................ 439
23.2.2 Low Frequencies.......................................................440
23.2.3 Direct Sound at Medium and High Frequencies.......440
23.2.4 Early Reflected and Reverberant Sound at
Medium and High Frequencies................................. 441
23.3 Analog Signal Processing....................................................... 442
23.3.1 Frequency Response Compensation.......................... 442
23.3.2 Time Delay Compensation........................................ 442
23.3.3 Current Amplifiers.................................................... 442
23.3.4 Electroacoustic Component Synthesis...................... 443
23.3.5 Diaphragm Motion Feedback.................................... 445
23.4 Digital Signal Processing.......................................................446
23.4.1 Transfer Functions.....................................................448
References......................................................................................... 450
Chapter 24 Headphones and Earphones.............................................................. 451

24.1 Introduction............................................................................ 451
24.2 Categorization........................................................................ 451
24.3 Design Considerations............................................................ 452
xviii Contents
24.4 Acoustic Environment............................................................ 455

24.5 Electrodynamic Headphones.................................................. 458
24.6 Electromagnetic Headphones................................................. 461
24.7 Piezoelectric Headphones....................................................... 463
24.8 Electrostatic Headphones....................................................... 463
Problems............................................................................................465
References.........................................................................................466
Chapter 25 High-Frequency Transducers............................................................ 467

25.1 Bandwidth and Power............................................................. 467
25.1.1 Bandwidth................................................................. 467
25.1.2 Transducer Choices................................................... 467
25.2 Semi-Resonant Capacitive Transducers.................................468
25.2.1 Acoustical Properties................................................469
25.2.2 Damping....................................................................469
25.2.3 Electrical Properties.................................................. 470
25.3 Piezoelectric Transducers....................................................... 471
25.3.1 Introduction............................................................... 471
25.3.2 Piezoceramic Bars..................................................... 471
25.3.3 Power Radiation........................................................ 473
25.3.4 Electromechanical Impedance Analogy................... 474
25.3.5 Q Factors................................................................... 476
25.3.6 Piezoceramic Disk Vibrators..................................... 476
25.4 Series and Parallel Resonance................................................ 477
25.5 Bandwidth and Ranging......................................................... 479
25.6 Piezoceramic Loudspeakers...................................................480
25.7 Piezoelectric Film Loudspeakers........................................... 481
25.7.1 Introduction............................................................... 481
25.7.2 Function..................................................................... 482
25.7.3 Electrical Properties.................................................. 483
25.7.4 Physical Configurations............................................. 483
25.8 Power Requirements of Piezoelectric
Loudspeakers and Transmitters��484
25.9 Parametric Loudspeakers for Audio....................................... 485
25.10 Ionophones.............................................................................. 487
References......................................................................................... 489
Appendix A: Introduction to Electric Components

and Classic Circuit Theory.................................................................................. 491
Appendix B: Filters and Filter Functions........................................................... 513
Contents xix
Appendix C: Magnetic Fields and Forces........................................................... 541

Appendix D: Time-Domain Approach to Directivity........................................ 553
Appendix E: Sound-Absorbing Materials.......................................................... 561
Appendix F: Resonance in Boxes and Rooms.................................................... 569
Appendix G: Level Definitions............................................................................. 583
Acknowledgments
I thank my wife Missan for her support and encouragement that made it possible
for me to write this book. Thanks also go to colleagues and students at the Division
of Applied Acoustics, Chalmers, for support and advice on the content in the early
versions of the manuscript, and to Samuel Kleiner for solving many mathematical
issues. I am particularly indebted to Neil A. Shaw for his detailed comments on the
manuscript, which made it possible to correct a number of omissions, errors, and
other shortcomings.
xxi
Author
Mendel Kleiner received his PhD in architectural acoustics in 1978 from Chalmers
University of Technology, Gothenburg, Sweden and is currently professor of
acoustics at Chalmers University of Technology and in charge of the Chalmers
Room Acoustics Group, where he teaches room acoustics, audio, electroacoustics,
and ultrasonics in the Chalmers Master Program on Sound and Vibration. He has
more than 50 publications and 110 papers, has presented many keynote lectures,
led courses at international conferences on acoustics and noise control, and
organized an international conference on acoustics. His main research areas include
electroacoustics and audio, computer simulation of room acoustics, electroacoustic
enhancement of room acoustics, room acoustics of auditoria, sound and vibration
measurement technology, product sound quality, and psychoacoustics. Dr. Kleiner
is a fellow of the Acoustical Society of America, serves on the Audio Engineering
Society’s Standards Committee on Acoustics, and was chair of its Technical
Committee on Acoustics and Sound Reinforcement for 15 years.
xxiii
List of Symbols
The SI [metric] system of units is used in this book. Units are written in brackets
as follows:
a radius, acceleration [m/s2]

A amplitude, equivalent sound absorption area [m2S], voltage amplification
factor
AF array factor
b width [m]
B bandwidth [Hz], magnetic flux density [T], susceptance
B′ bending stiffness per unit length for plates [Nm]
Bi Butterworth filter of order i
c speed of sound [m/s]
cg group velocity [m/s]
cph phase velocity [m/s]
cq crystal longitudinal wave velocity [m/s]
cB bending wave phase velocity [m/s]
cL longitudinal wave velocity [m/s]
cT transversal wave velocity [m/s]
CA acoustic compliance [m3/Pa]
CAB box air acoustic compliance [m3/Pa]
CAS diaphragm acoustic compliance [m3/Pa]
CASdc drone cone suspension acoustic compliance [m3/Pa]
CAW box wall acoustic compliance [m3/Pa]
CE electric capacitance [F]
CM mechanical compliance [m/N]
CMB box air mechanical compliance [m/N]
CMD diaphragm mechanical compliance [m/N]
CME electromechanical compliance [m/N]
CMF front box air mechanical compliance [m/N]
CP specific heat at constant pressure [J/K]
CV specific heat at constant volume [J/K]
Ci Chebyshev filter of order i
d distance [m]
dij piezoelectric strain coefficient [m/V]
D distance [m], damping, attenuation [dB], directivity function [dB], longitudi-
nal stiffness [Pa]
DI directivity index [dB]
e alternating voltage [V]
E modulus of elasticity, Young’s modulus [Pa], energy [J], voltage [V]
E 0 static voltage [V]
xxv
xxvi List of Symbols
f function, frequency [Hz]

f0 resonance frequency [Hz]
fC critical frequency [Hz], horn and duct cutoff frequency [Hz]
F force [N], directivity function
g function
G conductance [S], shear modulus [Pa], gain factor
h height [m], thickness [m], impulse response
H transfer function, magnetic field strength [A/m]
HD harmonic distortion [%]
I intensity [W/m2], moment of inertia [m4]
IMD intermodulation distortion [%]
IT intensity in duct or horn [W/m2]
I0 reference intensity = 1 · 10 -12 W/m2
j imaginary unit, j = √−1
ji spherical Bessel function of the first kind and order i
Ji Bessel function of order i
k wave number [m−1]
kB Boltzmann’s constant [J/K]
kE complex wave number for exponential horns [m−1]
kM spring stiffness constant [N/m]
kMS magnetostriction coupling coefficient [T−2]
kPE piezoelectric coupling coefficient
Kc coincidence number [m/s]
KEM electromechanical transformation factor
K ME mechanoelectrical transformation factor
l length [m]
ld duct length [m]
lh horn length [m]
lTH tractrix horn arm length [m]
L length, level [dB], longitudinal, lined, inductance [H]
LRi Linkwitz–Riley filter of order i
LEC voice coil inductance [H]
L I sound intensity level [dB]
Lp sound pressure level [dB]
Lu velocity level [dB]
LW sound power level [dB]
m molecular attenuation coefficient [1/m], expansion parameter [1/m]
m″ mass per unit area [kg/m2]
M moment [N m], molecular weight [kg], magnetomotive force [A]
MA acoustic mass [kg/m4]
MAdc drone cone acoustic mass [kg/m4]
MAD diaphragm acoustic mass [kg/m4]
MAP port acoustic mass [kg/m4]
MAR acoustic radiation mass [kg/m4]
MARd driver acoustic radiation mass [kg/m4]
MARp port acoustic radiation mass [kg/m4]
List of Symbols xxvii
MARv vent acoustic radiation mass [kg/m4]

MARW box wall acoustic radiation mass [kg/m4]
MAW box wall acoustic mass [kg/m4]
MAV vent acoustic mass [kg/m4]
MM mass [kg]
MMBM driver basket and magnet mass [kg]
MMM driver magnet mass [kg]
MMR mechanical radiation mass [kg]
Mm magnet magnetomotive force [A]
MMF magnetomotive force [A]
n modal density [1/Hz], refractive index, horn expansion rate
N number, number of modes, coil turns, transformer turns ratio
NH horn flare rate
p sound pressure [Pa]
p 0 reference sound pressure = 2 × 10 −5 Pa
P power [W], perimeter length [m]
PD dipole radiated power [W]
Pff free field radiated power [W]
PL perimeter length [m]
P0 static atmospheric pressure [Pa], reference power
PQ quadrupole radiated power [W]
q volume velocity [m3/s], electric charge [C], integer, crystal
Q volume flow [m3], Q factor (quality of resonance), quasi
QE electric charge [C]
QBi quasi-Butterworth filter of order i
r radius
r vector from origin to point x,y,z
r reflection coefficient
rA acoustic admittance [m5/Ns]
rL normalized resistance component of the input impedance at the surface of a
sound-absorptive material, rL = Re[Z2]/ρc
rM mechanical admittance [m/Ns]
rMB box mechanical admittance [m/Ns]
rMS suspension mechanical admittance [m/Ns]
R resistance [Ns/m] [resistive part of impedance], gas constant
R A acoustic resistance [Ns/m5]
R AB box interior acoustic resistance [Ns/m5]
R AL box leak acoustic resistance [Ns/m5]
R AP port acoustic resistance [Ns/m5]
R AR acoustic radiation resistance [Ns/m5]
R ARd driver acoustic radiation resistance [Ns/m5]
R ARp port acoustic radiation resistance [Ns/m5]
R ARv vent acoustic radiation resistance [Ns/m5]
R AS suspension acoustic resistance [Ns/m5]
R AT horn throat acoustic resistance [Ns/m5]
R AV vent acoustic resistance [Ns/m5]
xxviii List of Symbols
R AW box wall acoustic resistance [Ns/m5]

R E electric resistance [Ω]
REC voice coil electric resistance [Ω]
REF current sensing resistance for electric feedback [Ω]
REG generator electric output resistance [Ω]
REM microphone electric bridge resistance [Ω]
RF flow resistance of screens, etc. [Ns/m3]
R M mechanical resistance [Ns/m]
R MS suspension mechanical impedance [Ns/m]
RH relative humidity [%]
RMS root mean square
s radiation factor
s(t) time domain signal
sij mechanical stiffness coefficient [N/m]
S surface area [m2], signal (frequency domain)
S(ω) frequency domain signal
Sdc drone cone surface area [m2]
SD diaphragm surface area [m2]
SM horn mouth cross-section area [m2]
SS scattering cross-section area [m2]
ST tube (duct) cross-section area [m2], horn throat cross-section area [m2]
Sh1 StruveH function of order 1
t time [s], temperature [°C]
T absolute temperature [K], period [s], transversal, tension [N/m]
TH horn family parameter
T60 reverberation time [s]
u velocity, particle velocity [m/s]
U volume velocity [m3/s]
UB box volume velocity [m3/s]
UD diaphragm volume velocity [m3/s]
UP port volume velocity [m3/s]
v velocity [m/s]
V volume [m3], potential energy [J]
W width [m], magnetic energy
x coordinate
X reactance
X A acoustic reactance [m5/Ns]
X AR acoustic radiation reactance [m5/Ns]
X AT horn throat acoustic reactance [Ns/m5]
y coordinate
Y admittance [Ω]
YA acoustic admittance [m5/Ns]
YAR acoustic radiation admittance [m5/Ns]
YM mechanical admittance, mobility [m/Ns]
YMD driving point mechanical admittance, mobility [m/Ns]
List of Symbols xxix
YMR mechanical radiation admittance [m/Ns]

z coordinate
Z impedance
Z 0 characteristic impedance [Ns/m3]
Z A acoustic impedance [Ns/m5]
Z AD diaphragm acoustic impedance [Ns/m5]
Z AE ear canal acoustic input impedance [Ns/m5]
Z AM horn mouth acoustic impedance [Ns/m5]
Z AR acoustic radiation impedance [Ns/m5]
Z ARp port acoustic radiation impedance [Ns/m5]
Z ARv vent acoustic radiation impedance [Ns/m5]
Z AT horn throat acoustic input impedance [Ns/m5]
ZE electric impedance [Ω]
Z M mechanical impedance [Ns/m]
Z MB mechanical point impedance for bending waves [Ns/m]
Z MD mechanical driving point impedance [Ns/m]
Z MR mechanical radiation impedance [Ns/m]
Z MT horn throat mechanical impedance [Ns/m]
Z R sound field radiation impedance [Ns/m3]
ZS sound field impedance [Ns/m3]
GREEK LETTERS
Γ spatial distribution function
Λ magnetostriction constant [Ns/m4], mode number constant
Ξ specific flow resistance [Ns/m4]
Ω [V/A], angle, solid angle, resistance
Φ magnetic flux [Wb], piezoelectric conversion factor [N/V]
Ψ room eigenfunction
α absorption coefficient
β phase
γ help variable
δ damping constant [m−1]
ε permittivity of vacuum [F/m]
εr dielectric constant
ζ z-component of displacement [m]
η y-component of displacement [m], loss factor, efficiency
κ ratio between the specific heat at constant pressure and constant volume,
κ = CP/CV
λ wavelength [m]
λB bending wave wavelength [m]
λX bending wave wavelength at critical frequency [m]
μ integer, permeability [H/m]
μ0 permeability of free space [H/m]
μc core permeability [H/m]
xxx List of Symbols
μμ magnet permeability [H/m]

ν Poisson’s ratio
ξ extension [m], displacement [m]
ρ density [kg/m3]
σ tension [Pa], surface charge [C/m2], radiation ratio
τ shear stress [Pa], transmission factor
υ tractrix horn distance
Π angle
θ angle
ϕ angle
φ angle
χ structure factor for sound-absorbing materials
ω angular frequency (2πf ) [rad/s]
ω0 angular resonance frequency (2πf0) [rad/s]
GENERAL SYMBOLS
x̂ maximum value or peak value
x‾ average of x over time
〈x〉 average of x over space
x ̴ rms value of x
x underline indicates that x is a complex quantity
Δ difference
CERTAIN INDICES
′ per unit length, part 1
″ per unit area, part 2
* conjugate
+ positive direction
− negative direction
_ complex variable or function, time dependence ejωt
0 static, normal condition, resonance-, perpendicular to
b refraction-, bending-
c coincidence, critical
d diffuse
g limit-
i incident
m attenuation coefficient, average
r reflected
t transmitted
A acoustic, A-weighting
B bending wave, B-weighting
List of Symbols xxxi
C C-weighting
E electric
L longitudinal
M mechanical, receiver
O static
R radiation
S transmitter
T transversal
Uppercase letters are usually used to indicate amplitude [usually A, B] and num-
ber [usually N, M, Q].
Lowercase letters are used for length and distance [usually a, b, c, h, l, r, s, t] and
number [usually l, m, n, q].
1 Introduction
Electroacoustics is formed by the combination of knowledge from three areas, namely,

acoustics, mechanics, and electronics, as shown in Figure 1.1. Electroacoustics is
of interest for designers and users of almost any technical system that communi-
cates with humans. Professionals in electronics engineering, computer science, as
well as specialists in digital signal processing use and depend on audio and elec-
troacoustic engineering equipment and processes in their work. The interest in the
subject is self-evident for radio, television, and recording professionals. Equipment
for computer games and virtual environments of many kinds can be rendered more
efficiently and with better results, for enjoyment and presence, by using proper elec-
troacoustic engineering.
Electroacoustic devices such as microphones and loudspeakers are used in cars,
homes, churches, sports arenas, mobile phones, and many other places; they are key
parts in the modern communication society because they help transmit information.
Microphones and loudspeakers are always operated in a surrounding environment
about which the designer may know little. The final arbiter is the listener, whose
ability to hear features varies considerably between listeners, and, for a particular
listener, also with time, general condition, memory, etc. Since electroacoustic engi-
neering is multidisciplinary, any electroacoustic design will, of necessity, be a com-
promise between many factors. These factors make electroacoustics a fascinating
and challenging field in engineering.
The book discusses the key scientific and engineering principles that are neces-
sary to understand how these important transducers, as well as ultrasonic transduc-
ers, are designed. The compromises that are necessary in the design of practical
transducers are also introduced. The book is based on the theory necessary for
understanding how these transducers work, such as mechanical and acoustical anal-
ogies, conversion between analogies, transducers, radiation, and impedance. There
are also appendices on basic electric circuit and filter theory, room acoustics, and
sound absorbers. In contrast to older books, this book also presents a treatment of
arrays, acoustic center, as well as vented box and other loudspeaker enclosures.
The material presented is suitable for an advanced undergraduate or graduate
course on electroacoustics, technical acoustics, engineering acoustics, or commu-
nications acoustics being an outgrowth of the course on Electroacoustics that the
author has taught to graduate and advanced undergraduate students at Chalmers
University of Technology over many years. The book is designed to fit the needs of
graduate and advanced undergraduate students in electrical, mechanical, and com-
puter science departments, as well as transducer designers, acoustical consultants,
interested hobbyists, and laypersons. Relevant chapters also contain homework and
problems sections.
1
2 Electroacoustics
Fundamental
Oceanography
physical acoustics
Ultrasonic signal Sonar

processing engineering
Medicine Industry
Ultrasonics
ranging Sonics
Physiology Mechanics
Hearing Noise and

Electroacoustics
vibration
Psychology Architecture
Psychoacoustics Room acoustics
Electronics
Musical
Communications
instruments
Speech Music Art
FIGURE 1.1 Electroacoustics and its relationship to other areas. (Adapted from Lindsay, R.B.,
J. Acoust. Soc. Am., 39, 629, 1966.)
1.1 PREREQUISITES
Although the book contains sufficient basic material in electrical and commu-
nications engineering, the student will find it advantageous to have studied the
following material before starting to study the present book: Math courses includ-
ing material on the solution of ordinary differential equations, an introduction
to partial differential equations, and transform methods such as the jω method.
Engineering courses including basic (101) physics, mechanics, and electricity
would be advantageous as well. The book covers electroacoustic theory in an easy-
to-read style without resorting to matrix theory, which may not help in under-
standing the physics.
Because the main objective of the book is to teach engineering principles, stu-
dents will find the material useful in the broad range of applications they may come
across in their graduate research projects as well as later in their careers. Since the
book also contains material on how to measure and evaluate electroacoustic trans-
ducers, it will be of interest also to the buyer, quality control engineer, and evaluator
of electroacoustic transducers.
Introduction 3
1.2 OTHER BOOKS AND REFERENCES

The classical books on electroacoustics are those by Olson [1] and Beranek [2].
Both these books cover many areas besides electroacoustics. Olson discusses sound
reinforcement, audio systems, room acoustics, and much more, and Beranek also
includes noise control. The books by Hunt [3], Merhaut [4], and Rossi et al. [5] can
be regarded “pure electroacoustics.” This book is shorter and its scope is limited
to electroacoustics. An excellent recent book on electroacoustics is by Geddes [6].
The book by Gayford [7] is quite practical and shows many examples of designs.
In contrast to the previously mentioned books, the one by Kinsler and Frey [8]
also includes material on ultrasonic transducers, and theory on waves in mem-
branes, bars, and sheets. An often overlooked but excellent book is by Skudrzyk
on complex vibratory systems [9]. Gelshøj’s book [10] deals mainly with electro-
acoustic analogies as does Olson’s [11]. Fischer’s book is a short one [12]. Leach’s
book is directed toward students of electrical engineering and was probably the
first to introduce the use of computer modeling by SPICE software [13]. Along
with these books, there are many others covering the field from an audio view-
point. Such books are that of Borwick [14]. Colloms’ book treats the subject more
from an audio enthusiast’s viewpoint [15]. Benson’s book contains much interest-
ing material by various authors [16] and finally Lindsay’s book on the history of
acoustics contains material for anyone interested in transducers as well [17]. Of
course, the Journal and the Convention Proceedings of the Audio Engineering
Society, the IEEE Transactions on Audio and Electroacoustics, and the Journal
of the Acoustical Society of America are treasure troves for anyone interested in
electroacoustics [18].
REFERENCES
1. Olson, H. F., Acoustical Engineering, 3rd edn., D. van Nostrand, Princeton, NJ (1957),
Library of Congress Catalogue Card No. 57-8143.
2. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986) ISBN-13:
978-0883184943.
3. Hunt, F. V., Electroacoustics, American Institute of Physics, New York (1982) ISBN-13:
978-0883184011.
4. Merhaut, J., Theory of Electroacoustics, McGraw-Hill, New York (1979) ISBN-13:
978-0070414785.
5. Rossi, M., Acoustics and Electroacoustics, Artech House, Norwood, MA (1988) ISBN-
13: 978-0890062555.
6. Geddes, E., Audio Transducers, Gedlee, MI (2002) ISBN-13: 978-0972208505.
7. Gayford, M. L., Electroacoustics, Newnes Butterworth, London, U.K. (1970) ISBN-13:
978-0408000260.
8. Kinsler, L. E. et al., Fundamentals of Acoustics, 2nd edn., John Wiley & Sons, New York
(1962) ASIN: B000LC9DO6.
9. Skudrzyk, E., Simple and Complex Vibratory Systems, Pennsylvania State University
Press, University Park, PA (1968) ISBN-13: 978-0271731278.
10. Gehlshøj, B., Electromechanical and electroacoustical analogies. PhD thesis, Academy
of Technical Sciences, København (1947).
4 Electroacoustics
11. Olson, H. F., Solutions of Engineering Problems by Dynamical Analogies, 2nd edn.,
D. Van Nostrand, Princeton, NJ (1966) ASIN: B000NLS85K.
12. Fischer, F. A., Fundamentals of Electroacoustics, 1st English edn., Interscience
Publishers, Inc., New York (1955) ASIN: B0000CJA9N.
13. Marshall Leach, W., Introduction to Electroacoustics and Audio Amplifier Design,
3rd edn., Kendall Hunt Publishing, Dubuque, IA (2008) ISBN-13: 978-0757503757.
14. Borwick, J., Loudspeaker and Headphone Handbook, 3rd edn., Focal Press, Oxford,
U.K. (2001) ISBN-13: 978-0240515786.
15. Colloms, M., High Performance Loudspeakers, Wiley, New York (2005) ISBN-13:
978-0470094303.
16. Blair Benson, K., Audio Engineering Handbook, McGraw-Hill, New York (1988) ISBN-
13: 978-0070047778.
17. Lindsay, R. B., The story of acoustics, J. Acoust. Soc. Am., 39, 629–644 (1966).
18. www.aes.org, www.ieee.org, and acousticalsociety.org (sampled June 2012).
2 Introduction to
Electroacoustic Systems
This chapter introduces the concepts and ideas that will be used later in the book.
The main use of electroacoustics is in the design of systems for the pickup, recording/
transmission, and reproduction of sound in the audio range between 20 Hz and
20 kHz. However, just frequency response is not sufficient for quality; for faithful
audio signal reproduction, there should be inaudible level and time delay differences
between various parts of the audio spectrum. The sound also needs to be free of
noise and signal degradation, distortion, at all sound pressure levels of interest to
hearing. The spatial properties of sound reproduction are important as well; many
electroacoustic systems aim at retaining some of the spatial properties of the sound
field and its reverberation.
Figure 2.1 shows two electroacoustic systems. The top system (Figure 2.1a) shows
the recording of sound from sound sources such as musical instruments or voices.
The lower one (Figure 2.1b) shows the reproduction of recordings as well as signals
created synthetically by electronic devices. The signals need to go through many—
some possible sound quality degrading—system components. In addition, the s ignals
to microphone as well as the signals from loudspeakers are affected by the reverbera-
tion of the rooms in which they are used.
2.1 RECORDING
As indicated in Figure 2.1, the recording system chain is started by the microphone
and sound source. The purpose of the microphone is to convert the sound signal’s
time history into an equivalent electrical signal having the same time history. This
is done by an electromechanical conversion system, a transducer, in the microphone.
The microphone picks up the sound from the sound source. Usually, however,
the sound source is located in an environment that contributes noise and reflections.
Furthermore, the sound source has directional properties, and these typically vary
with frequency; for example, the sound source is likely to be less directional at low
frequencies than at high frequencies. Similarly, the microphone will also have direc-
tional properties depending on its design.
The sound pressure at the microphone forces movement in the microphone’s
mechanical system. As the mechanical system moves in the primarily magnetic or
electrostatic field of the transducer, an electrical signal will be generated that can be
recorded or measured. Because the signal is usually low in level, it is necessary to
amplify it to a suitable voltage level to reduce noise being added by cables and the
like. To avoid signal degradation, some modern microphones convert the signal to
5
6 Electroacoustics
Digital signal
A/D processing Data storage
Microphone Mixing -
preamplifier converter and/or
console Data reduction
(digitizer) Transmission
and coding
= microphone
= loudspeaker
(a)
Digital signal
Reception
D/A processing Power
and/or Decoder -
Data retrieval converter Room amplifier
correction
(b)
FIGURE 2.1 There are many systems involved in (a) recording and (b) reproducing sound.
The electroacoustic components are the microphones and loudspeakers.
a digital format before it is transmitted to the recorder. If the microphone does not
have flat frequency response, some correction can be done using the preamplifier
and its filters. Room reflections, however, can only be reduced by directional micro-
phones and, in some cases, by digital signal processing.
2.2 REPRODUCTION
Electroacoustic reproduction requires power amplifier which takes the low level
electrical signal and provides sufficient electrical power to overcome the electrical
losses in the electroacoustic transducer, typically a loudspeaker (“driver”) or vibrator
(“shaker”), and generate the desired sound at the receiver.
As in the case of the microphone, a loudspeaker’s transducer, the driver, relies
primarily on permanent magnetic or electrostatic fields for its function. The changing
electrical current provided by the amplifier results in mechanical vibration.
Since sound is generated by the acceleration of an air volume, the mechanical
vibration must be made to move a surface such as a membrane, diaphragm, or
piston that can move air. The larger the surface, the more air can be moved, but
the acceleration of the volume of air also depends on the vibration frequency. Low
frequencies generate less air movement than high frequencies, in part because at
these frequencies air escapes from the front side to the back side of the vibrating
surface. This flow makes it necessary to confine the back side of the surface, which
is the purpose of the loudspeaker box.
The sound waves generated can reach the receiver directly or by reflection over
various paths depending on the listening environment. To reduce the reflected sound,
many listen to recordings using headphones or earphones.
Introduction to Electroacoustic Systems 7
2.3 LINEARITY
A classical approach in electroacoustic engineering has been to assemble com-
plex electroacoustic systems from subsystems that are linear and unaffected by
the loading of one subsystem by another. For example, the sound pressure at the
microphone location should be unaffected by the presence of the microphone, and,
the output voltage of a power amplifier unaffected by the load of the loudspeaker
or vibrator.
Any practical system will however have some nonlinearity. Two types of sig-
nal degradation affect electroacoustic systems: linear and nonlinear distortions.
Frequency response unevenness is called linear distortion, in which various parts
of the spectrum are reproduced at different relative levels. This type of distortion
can often be overcome by frequency-dependent equalization, using analog or digital
signal processing. Nonlinear distortion however is characterized by the generation of
new frequencies in the recorded or reproduced signal, that is, frequencies that were
not present in the original signal. Such nonlinear distortion can only be compen-
sated for to some small degree, whether linear or digital signal processing is used.
Figures 2.2 and 2.3 show some of the characteristics of linear and nonlinear signal
degradation.
Research in psychoacoustics has shown that full linearity and information trans-
mission are not necessary for the reproduced signal to sound pleasant or faithful to
the listener. Hearing is characterized by masking phenomena, in which the percep-
tion of sounds is inhibited by other sounds, in both the time and frequency domains.
Both masking domains are of interest in electroacoustic engineering.
For example, the sound reflection properties of the room result in a very irregular
frequency response if measured over long time segments, as shown in Figure 2.4.
By designing highly directive loudspeaker systems, we can reduce these frequency
response irregularities.
Most voice and musical sounds have transients that are important to hearing.
The perceived direction to the sound source and the timbre of the reproduced
sound are to a large extent determined by the arrival of the first sound transient
from the source. It is much easier to detect the direction of a transient sound than
Level
Peak
Ripple Flat Rise Roll-off

Lref
Lref – 3dB
Low-frequency
High-frequency
cutoff Dip
Passband cutoff
fl fref fu
Frequency
FIGURE 2.2 Various terms used to describe the characteristics of the frequency response
of an electroacoustic system.
8 Electroacoustics
Marker for level and frequency
Level
Dropline to improve legibility
0 f1 2f1 3f 1 4f1 and so on Frequency

(a) (linear scale)
Fundamental Harmonic distortion

(both even and odd
harmonics shown here)
Intermodulation distortion
(sidebands)
Level
0 f1 f2 2f2 Frequency
(b) f2 – f1 f2 + f1 (linear scale)
Low-frequency High-frequency Harmonic to f2

component component
Difference tone distortion

Level
0 ˘f1 f1 f1 +˘f Frequency

(c) (linear scale)
Frequency modulation distortion

(sidebands)
Level
0 f1 f2 Frequency
(linear scale)
Low-frequency High-frequency
component component
(d)
FIGURE 2.3 Nonlinear distortion generates new spectral components. In electroacoustic

systems, the types of distortions that are typically considered are as follows: (a) harmonic
distortion (both odd and even harmonics shown in this example), (b) intermodulation distor-
tion, (c) difference tone distortion, and (d) frequency modulation distortion.
Introduction to Electroacoustic Systems 9
15
10
Relative SPL [dB] 5
0
–5
–10
–15
2000 2100 2200 2300 2400
Frequency [Hz]
FIGURE 2.4 An example of the variation of sound pressure level in the diffuse field of a
living room as a function of frequency.
that of a continuous sound. In large rooms that have much reverberation, the direct
sound will be very weak and it will be difficult to detect the direction to the source
by hearing.
The frequency response irregularities of reverberation caused by the interference
of waves arriving from many different directions are mostly resolved by binaural
hearing. The phase and time delay differences between the signals at the two ears are
used to analyze the signal. If the frequency response irregularities generated by the
room resonances are sufficiently close in frequency and reasonably damped, they are
to some extent forgiven by hearing. So, even the sound reproduction by transducers
based on resonant techniques can be enjoyed. The psychoacoustic limits for these
resonant systems are not well understood at this time. Cognition and emotion also
influence the analysis done by hearing.
The success of various systems that use irrecoverable signal compression shows
that many listeners are prepared to accept and hardly notice the strong signal deg-
radation inherent in these signals caused by the use of simple frequency and time
domain masking models.
Of course, for measurement purpose it is usually desired to have minimal signal
degradation so that the signal’s properties can be measured and analyzed correctly.
Since much is still unknown about hearing and its signal analysis, it is advanta-
geous to avoid signal degradation to the maximum extent, both for sound quality in
listening and accuracy in measurement.
3 Sound and Its Properties
3.1 SOUND WAVES

Sound waves in air are longitudinal waves, which means that the oscillation of the
wave motion is in the direction of propagation. Sound waves are characterized by
sound pressure, the excess pressure over atmospheric pressure, and particle velocity,
the average velocity of the gas molecules in a small volume. Note that the particle
velocity is a vector quantity.
The distance until the sound wave repeats itself is the wavelength. The wave-
length depends on both the velocity of sound in the medium and the frequency of
oscillation. The wave equation defines the relationship between the spatial and tem-
poral behavior of sound and describes the propagation of sound waves.
The wave equation is derived using the equation of motion, the equation of con-
tinuity, and an equation of state which describes the thermodynamic properties of
the medium [1,2].
3.1.1 Equation of Continuity

Figure 3.1 shows a section of a volume element that carries a sound wave. The equa-
tion of continuity expresses the relationship between change of density ρ and volume
V in that volume element having constant mass. A net increase in volume reduces
the density.
∂u x ∂ρ
ρ =− (3.1)
∂x ∂t
3.1.2 Equation of Motion

Newton’s equation of motion states that the force F needed to accelerate a mass m
at an acceleration of a is given by F = ma. This also applies to the mass of the gas in
the volume element under consideration in our case. The force needed to accelerate
the volume element depends on the difference in pressure between the planes x and
x + dx, respectively. Since the acceleration of the volume element is ∂ux/∂t, we obtain
the equation of motion in Cartesian coordinates as
∂p ∂u
= −ρ x (3.2)
∂x ∂t
11
12 Electroacoustics
ξ(x) ξ(x + dx)
P0 + p(x) P0 + p(x + dx)
x x + dx
FIGURE 3.1 The small volume element under consideration. P0 is the static pressure.
3.1.3 Thermodynamic Properties
If there is no heat transfer between the gas and its surroundings, the relationship
between sound pressure and instantaneous volume is described by Poisson’s equation
pV κ = constant (3.3)

where κ is the ratio of specific heats at constant pressure and volume, respectively,
that is, κ = Cp/Cv. Since the mass of the volume element is unchanged, the relation-
ship between changes in pressure and density is
∂p p
= κ (3.4)
∂ρ ρ
From physics, we know that the relationship between density and pressure in a gas is
p RT
= (3.5)
ρ M
where
M is the molecular weight of the gas or gas mixture
T is the temperature (in Kelvin)
R is the universal gas constant
so the thermodynamic properties are described by
∂p κRT
= (3.6)
∂ρ M
3.1.4 Wave Equation

Combining the three equations, we obtain the wave equation for plane waves as
expressed in sound pressure p:
∂2 p 1 ∂2 p
− = 0 (3.7)
∂x 2 c 2 ∂t 2
The wave equation may of course equally well be expressed in particle velocity, ux.
Sound and Its Properties 13
The quantity c depends on the thermodynamic properties of the gas in the

following way:
κRT
c= (3.8)
M
3.2 PLANE WAVE SOLUTIONS TO THE WAVE EQUATION

One way of finding solutions to a differential equation such as the wave equation is
to try various solutions to see if they satisfy the equation. If we use this approach in
investigating the possible solutions to the wave equation, a reasonable guess to math-
ematically describe waves is a general pair of functions such as
p ( x, t ) = f ( x − ct ) + g ( x + ct ) (3.9)

Here, f and g are arbitrary functions that have continuous derivatives of the first
and second order. Testing this solution, we find that it satisfies the wave equation.
Studying the properties of the solution, we see that the shapes of the functions
remain unchanged for variations of the variables x and t, but they are displaced in x
according to the value of t. The relationship between the time variable t and the space
variable x is x = tc. The shape of the wave moves with the speed c, which we call
the (propagation) speed of sound, obtained from Equation 3.8. For the air close to a
temperature of 20°C, the speed of sound is about c ≈ 331.6 + 0.6 t (m/s), where t is the
temperature of air expressed in °C.
According to the Fourier theorem, any physical wave shape may be obtained by
suitable superposition of sines and cosines. Since the shape of the wave does not
change as it moves, we understand from the Fourier theory that all frequency com-
ponents building up the shape of the wave propagate at the same velocity. Sound
propagation in gases is fairly independent of frequency, but frequency-dependent
phenomena, such as damping are discussed later. The term “nondispersive” is used
to describe the fact that the propagation velocity is frequency-independent. We will
later see that bending waves that may exist in loudspeaker diaphragms are dispersive.
3.3 FREQUENCY AND TIME DOMAINS

In studying the propagation of sound waves we have the possibilities of using time-
or frequency-domain representation. In room acoustics, because of the properties
of hearing, it is often intuitive to use the time-domain representation. In elec-
troacoustics, because of its reliance on the principles of electrical engineering, a
frequency-domain representation is often more intuitive. However, when systems
become nonlinear, the time-domain representation is often more practical in elec-
troacoustics. In this book, we use a “small signal” approach, and we will assume
that all properties are linear unless otherwise noted. Note that in using the Fourier
and Inverse Fourier transforms, we can switch signal representation between the
two domains.
14 Electroacoustics
3.3.1 jω-Method
In the rest of this book we will only study continuous sinusoidal sound and vibration
using a frequency-domain approach. It is practical to use the transform method called
the jω-method to describe time variation as well as various properties of sound and
media. The variables of sound pressure, particle velocity, etc., then become com-
plex variables, which greatly simplify the mathematics. The instantaneous sound
pressures and particle velocities are given by the real part of their complex quanti-
ties, which are designated by an underscore, for example, p in equations. Using the
inverse Fourier transform we can convert jω-method data to time data, for example,
to find the impulse response of a system.
Using the jω-method we can rewrite the wave equation in the following form:
∂2 p
+ k 2 p = 0 (3.10)
∂x 2
Here, k is the wave number, also called the propagation constant. The relationships
between the wave number, the speed of sound c, frequency f, and wavelength λ, are
given by
ω 2πf 2π (3.11)
k= = =
c c λ
Note that wave number times distance is equivalent to distance-relative wavelength

(except for a constant). Often it is more intuitive to consider distance in the form of
a number of wavelengths.
Using the jω method, we can write, as a special case, a possible solution to the
wave equation in the following form:
p ( x, ω, t ) = pe
ˆ j (ωt − kx ) (3.12)

This describes a one-dimensional sinusoidal wave propagating in the direction of
positive x. The magnitude of the sound pressure p(x,t) is denoted by p̂. Generally, we
do not explicitly write the time variation (at frequency ω) when we use this transform
method; instead, we write _p(x,k) since k is a function of ω. This provides a shorter
way of writing the equations. The general solution to the wave equation in the one-
dimensional case of course also has to feature a wave propagating in the direction of
negative x, and is written as
p ( x, k ) = pˆ + e − jkx + jα + pˆ − e jkx + jβ (3.13)

Here, α and β are arbitrary phase constants.
When working with sound power, it is more convenient to use the mean square
or the root-mean-square (RMS) value of the sinusoidal signals. The RMS value sRMS
of signal s(t) is defined as
1 t +T

sRMS =
T ∫
t
s 2 (t ) dt (3.14)
Here, T is the integration time which is a multiple of the period length for sinusoidal
signals. For signals that are not periodic such as random noise, the value of T needs
to be very large for a precision estimate of the signal’s RMS value, T→∞. In this text,
we use the ∼ (“tilde”) character on top of letters to show that we mean the RMS
value of a quantity. For a sinusoidal signal having a peak value ŝ, the RMS value is
sˆ
s = (3.15)
2
3.4 IMPEDANCE
In electrical engineering, the impedance is the complex ratio between voltage and
current. In acoustical and mechanical engineering as well, the ratio between two
interdependent quantities—such as force and linear velocity—at some point in a
system is often called impedance and denoted by Z. The complex impedance is
Z = R + jX, and has real (R) and imaginary parts ( jX) called resistance and reactance.
The inverse of impedance is usually called mobility in acoustics and mechanics.
In electrical engineering, the inverse of impedance is called admittance. We will
use both the terms in the book. The complex admittance and mobility are written
as Y = G + jB. The unit of admittance is siemens. The real part of Y is called con-
ductance, and the imaginary part is called susceptance, both measured in siemens.
Instead of using a real and an imaginary part, we can use magnitude and phase.
Impedance, admittance, and mobility are generally frequency-dependent, com-
plex quantities.
3.4.1 Common Impedance Definitions

The characteristic impedance, Z 0, is the ratio of sound pressure to particle velocity
in an infinite plane wave. The unit of characteristic impedance is Rayl (Ns/m3). The
characteristic impedance is usually resistive, if there are no propagation losses in the
medium, and equal to the product of the density of the medium and the propagation
velocity of sound in the medium. We usually write Z 0 = ρc.
Sound field impedance, ZS, is the ratio of sound pressure to particle velocity in the
reference direction at a chosen point in the medium. Sometimes, the term specific
acoustical impedance is used for this quantity. The unit of sound field impedance is
Rayl (Ns/m3).
Acoustical impedance, Z A, is the ratio of sound pressure to volume velocity U
in the normal direction to a reference surface in an acoustic system. Volume veloc-
ity is the product of the normal of the particle velocity and the surface area under
consideration. The particle velocity is then considered constant regarding amplitude
and phase over the surface. The unit of acoustical impedance is the acoustical ohm
(Ns/m5).
Mechanical impedance, Z M, is the ratio of force to velocity at a chosen point in
a mechanical system. The unit of mechanical impedance is the mechanical ohm
(Ns/m).
16 Electroacoustics
TABLE 3.1
Relationships between the Various Types
of Impedances
Multiply
ZM ZS ZA
by
to obtain
ZM 1 S S2
ZS 1/S 1 S
ZA 1/S2 1/S 1
The surface S is the surface over which the pressure is

acting or trough which the volume flow is moving

(perpendicularly).
The relationships between the various types of impedances are shown in Table 3.1.
The surface S in the table is the surface over which the pressure is acting or volume
flow is moving.
Similar definitions of interest used in mechanical engineering are the ratio
between force and displacement called dynamic stiffness and the ratio between force
and acceleration called apparent mass.
3.4.2 Sound Field Impedances in a Plane Wave

Using the equation of motion 3.2 for plane waves, we find that the sound field imped-
ance for a plane wave propagating in the direction of positive x is
p+
= ρc = Z 0 (3.16)
u+
We see that in this case the sound field impedance is the same as the characteristic
impedance of the medium. Obviously, the sound pressure and the particle velocity
are in phase.
In the same way, we find that the sound field impedance for a plane wave propa-
gating in the direction of negative x is
p−
= −ρc = − Z 0 (3.17)
u−
The minus sign in front of Z 0 in this equation is a result of our choice of always
regarding positive particle velocity of being in the direction of positive x, even for the
wave that propagates in the negative direction.
3.5 SOLUTIONS TO THE WAVE EQUATION: SPHERICAL WAVES

Using a derivation similar to the one used to obtain solutions to the wave equation in
rectangular coordinates, one can obtain the general solutions for the case of spheri-
cal symmetry as
A + − jkr + jα A − jkr + jβ
p (r , k ) = e + e (3.18)
r r
where A+ and A− are the wave amplitudes at 1 m distance, for the outward- and
inward-going waves, respectively, and r is the radius from the center of origin.
For the case of spherical symmetry, we are generally only interested in the wave
which moves outward, in the direction of positive r. Using the equation of motion for
a spherical wave
∂p(r , k ) ∂ u r (r , k )
= −ρ (3.19)
∂r ∂t
we can show that the radial particle velocity ur of such an expanding wave is given by
A+  1  − jkr
ur (r , k ) = 1+ e (3.20)
ρcr  jkr 

We note that we now have an additional term 1/jkr. This implies that there will be a
phase difference between _p and ur that will be dependent on the value of kr, that is,
on frequency and distance to origin.
The phase difference will be largest when kr ≪ 1, that is, when the frequency is
low and/or the distance to the center is less, and the radial particle velocity is
A+
ur (r , k ) ≈ e − jkr (3.21)
jρckr 2
We see that there is a phase difference of 90° between particle velocity and sound
pressure; the particle velocity lags the sound pressure. Another important observation
is that the amplitude of the particle velocity ur increases by 1/r 2 as the distance to the
origin is reduced.
When kr ≫ 1, that is, when the frequency is high and/or the distance to the center
is large, the radial particle velocity is
A + − jkr
ur (r , k ) ≈ e (3.22)
ρcr
18 Electroacoustics
We note that in this far region, particle velocity and sound pressure are in phase as in
a plane wave. The amplitude of both quantities also depends on distance in the same
way as in the far-field region.
To radiate an ideal spherically symmetrical wave, a radiator would have to feature
a spherical surface moving at the same radial velocity at all points. Ideally, such a
radiator would be infinitely small. One uses the terms “point source” and “mono-
pole” to label ideally small and radially radiating sound sources. A monopole source
is assumed to have a source strength that is independent of its acoustic environment,
and its vibration is constant. We can also say that it is a sound source that has high
internal impedance.
One way of approximating such radiators, under certain conditions, is to use
a small loudspeaker or an assembly of small loudspeakers. One might think that
it would be easy to simulate a spherical source by using a large number of small
loudspeakers mounted on a sphere. However, because of the finite difference in dis-
tance between the loudspeaker diaphragms as well as the fact that these usually
have a conical shape and are not a continuous part of the spherical surface, we find
that the radiation pattern is not ideally spherical at frequencies where the distance
between loudspeakers or the depth of the cones is greater than a small fraction of
the wavelength.
A loudspeaker usually consists of a box or enclosure and a driver. If we have a
loudspeaker box which is designed in such a way that only one side of the loudspeaker
driver’s diaphragm is facing the exterior of the box, and if the “dimensions” of the
loudspeaker box are much smaller than the wavelength of sound being generated—
for example, one third of the wavelength at the frequency of interest—then the sound
field radiated by the loudspeaker will be a fairly good approximation to a spherical
sound field at a sufficiently large distance. Close to the loudspeaker, the sound field
will not have the characteristics of that of a true, spherical sound source. The higher
the frequency, the less similar will be the characteristics.
A different way of studying the directivity properties of a radiator is to ana-
lyze its impulse response. If the surface of the spherical radiator could move
with infinite acceleration for a short time, the radiated sound pressure would be
a Dirac pulse, since sound pressure is proportional to the volume acceleration of
the sound source.
Only a truly spherical source can have this ideal “geometrical” impulse
response as discussed in Appendix D. The geometrical impulse response of a
dodecahedron-shaped loudspeaker, for example, such as the one shown in Figure
3.2, with ideal loudspeaker drivers, all in phase, over all its flat surfaces, will
have impulse response contributions due to edge diffraction. Correspondingly,
the frequency response of the loudspeaker will feature frequency-response
irregularities.
3.5.1 Sound Field Impedance in a Spherical Wave

The sound field impedance in a spherical wave varies according to the distance
to the origin. Using our previous Equations 3.18 and 3.20, we can show that the
FIGURE 3.2 A loudspeaker using a dodecahedron arrangement of drivers to approxi-

mate the omnidirectional sound radiation characteristics of a monopole. (Photo by Mendel
Kleiner.)
sound field impedance of a spherical, expanding wave varies with distance to the
origin r as
p (r , k ) jkr 1
Z S (r , k ) = = ρc = ρc (3.23)
ur (r , k ) 1 + jkr 1
+1
jkr

In the near-field region (kr ≪ 1), that is, for low frequencies and/or small distances
to the origin, the sound field impedance of the spherical wave is almost pure
reactance:
Z S (r , k ) ≈ jρckr = jωρr; kr 1 (3.24)

If the sound field impedance is primarily a reactance, as it is close to a small sound

source, the sound field is said to be a reactive near-field.
In the far-field region (kr ≫ 1), that is, for high frequencies and/or large distances
to the origin, the sound field impedance of the spherical wave will be almost real
and equal to the characteristic impedance of the medium. There, the sound field
impedance of an expanding spherical wave is the same as that for an infinite plane
wave.
Z S (r , k ) ≈ ρc = Z ; kr 1 (3.25)

20 Electroacoustics
3.6 SOUND INTENSITY AND SOUND POWER

Wave propagation is characterized by transport of energy in the direction of the
wave. The energy is transported at a velocity called the group velocity cg. If the sys-
tem is dispersive, that is, the group velocity varies with frequency, the group velocity
will be different from the phase velocity cph, the velocity needed to always see the
same phase in the wave. For sound waves in air, the difference between the group
velocity and phase velocity is usually negligible.
Sound intensity is a measure of the rate of energy transport per unit area in the
wave. The sound intensity is a vectorial quantity and is directed along the direction
of particle velocity.
1
I= Re  pu *  (3.26)
2
Here u* is the complex conjugate of u. We know that for an infinite plane wave in
the positive x-direction, sound pressure and particle velocity will be in phase, which
results in
p2
I =
pu = = ρcu2 (3.27)
ρc
We can calculate the sound power P being transported by a sound field by integrat-
ing the component of sound intensity normal to the surface, over the surface area in
question.

P=
∫ I dS (3.28)
For an infinite plane wave, the sound power being transmitted over a surface element
of area S is
P = IS =
puS = ρcu2 S (3.29)

Since the sound field impedances for a plane wave and a spherical wave under the
condition kr ≫ 1 are the same, Equation 3.29 can be used to calculate the sound
power in a spherical wave at far distances and/or high frequencies.
3.7 PROPAGATION LOSSES

As the sound wave moves away from a source, the amplitude and the intensity of
the wave are reduced not only because of the power spread over a larger area, called
“geometrical attenuation,” but also due to various loss mechanisms, which convert
the sound energy in the sound wave into heat. These losses are known as excess
absorption. The losses in sound propagation in air are due to imperfect adiabatic
processes, heat conduction, viscous losses, and relaxation phenomena. The first two
mechanisms are called “classical attenuation.” A derivation of the attenuation coef-

ficients can be found in Ref. [3].
In the propagating wave volumes of high pressure have higher temperature than
those at low pressure, the heat conduction in the wave leads to losses. The viscous
losses in a wave are of about the same order of magnitude as the losses due to heat
conduction. Shear deformation in the wave causes these losses. The relaxation phe-
nomena however are responsible for the major part of the losses at normal conditions.
When the gas is compressed, the gas molecules are given increased translational,
rotational, and vibratory energies. These energies cannot be returned to the sound
field unless the frequency is low, and thus lead to heat and losses. The attenuation
depends on both temperature and relative humidity. Losses due to viscosity in nar-
row tubes play a central role in engineering acoustic resistance devices.
The excess absorption is usually accounted for by simply introducing a damping
term in the expression for the intensity as a function of propagation distance x and
the attenuation coefficient m:
I ( x ) ∝ e − mx (3.30)
The attenuation due to heat conduction and viscous losses increases proportionally
to frequency squared. The relaxation causes an attenuation which peaks at certain
frequencies. The relaxation process involves excitation of the gas molecules to other
energy states, for example, from translational energy to vibrational and rotational
energies. The relaxation time depends on the time that the molecule remains in the
excited state. Maximum attenuation will occur when the period of oscillation in the
sound wave is about the same as the relaxation time.
Figure 3.3 shows curves, for some values of the relative humidity, of the
frequency-dependent attenuation of sound in air at normal temperature. Using these
101
100
Attenuation [dB/m]
10–1
10–2
10–3 RH 10% @ 20°C

RH 40% @ 20°C
RH 80% @ 20°C
10–4 Only heat conduction
and viscosity
20 50 100 200 500 1k 2k 5k 10 k 20 k 50 k 100 k 200 k

Frequency [Hz]
FIGURE 3.3 Attenuation in dB/m for sound propagation in air as a function of frequency
at a temperature of 20°C and with relative humidity, RH in %, as parameter [4]. The graph
can be used to calculate the attenuation coefficient m since the attenuation ΔL in dB over a
distance x is ΔL ≈ 4.3mx.
22 Electroacoustics
curves, it is possible to calculate corresponding values for the attenuation coeffi-

cient m. Air is a gas mixture, so there will be several relaxation processes.
The standard ISO 9613-1:1993 (E) can be used to calculate the attenuation of
sound in air for general combinations of temperature and relative humidity.
For the propagation of sound in air, the attenuation of the wave amplitude due to
losses in the air is small, except for certain combinations of temperature and frequency.
In the frequency range of interest in audio sound reproduction, the sound propagation
losses are generally only of interest when we deal with sound reinforcement over large
distances or with ultrasonic waves in air such as in some sonar engineering.
3.8 ELEMENTARY SOUND SOURCES

In our analysis of sound generation by objects and flows, we will find that many
sound sources can be represented by assemblies of elementary sound sources. The
simplest elementary source from which further sources can be built is the monopole.
Using monopoles, we can assemble other elementary sources as shown in Table 3.2.
3.8.1 Monopoles
We have already discussed monopoles when we studied simple spherical sound fields
in Section 3.5. Assume now that our oscillating monopole sphere has a mean radius a
and that there are no reflecting surfaces in its vicinity. Using the definition of acoustical
impedance we obtain the sound field impedance at the surface of the sphere at r = a as
2
ρc ( ka ) ρcka
Z S ( a, k ) = 2
−j 2
(3.31)
1 + ( ka ) 1 + ( ka )

2
Z S ( a, k ) ≈ ρc ( ka ) − jρcka; ka 1 (3.32)

We obtain the acoustical impedance seen by the vibrating surface by dividing the
sound field impedance by the surface area of the sphere.
ρω 2 ρ
Z A ( a,ω ) ≈ − jω ; ka 1 (3.33)
4πc 4πa
TABLE 3.2
Commonly Considered Elementary Sound Sources
Number of
Source Type Monopoles Phase
Monopole 1
Dipole 2 Sources out of phase
Quadrupole 4 Two sources in phase, two out of phase
We note that the real part of the acoustical impedance will be very small at low fre-
quencies and that it has a frequency dependence that is proportional to ω2.
We now want to find the sound pressure at some distance from the monopole
when the monopole has a certain volume velocity. Using the relationships given ear-
lier, we find that
U = 4πa 2 u (3.34)

e − jkr
p (r, k ) = jωUρ (3.35)
4πr
We will use this expression in deriving the sound pressure at some distance for vari-
ous cases of radiators. We should also note that the sound pressure at some distance
is proportional to the density of the gas and to the volume acceleration which fits in
with Newton’s second law, force equals mass times acceleration.
In our derivation of the wave equation we neglected terms due to nonlinearity
in our wave equation. We should include these in the analysis when the RMS
sound pressure in air becomes 102 Pa (about 140 dB) and higher. We do not want
the loudspeaker to be too small since the sound pressure increases as the distance
to the diaphragm becomes smaller. The nonlinearities of air put a lower bound
on the distortion that can be achieved with a loudspeaker. In horns and other
systems where sound levels are high, the distortion may become very large at
high frequencies.
3.8.2 Power Radiated by a Monopole

The acoustic power radiated by the monopole is also of fundamental interest. We
can find the radiated power by integration of the sound intensity. The sound inten-
sity in a spherical wave at far distance can be obtained using Equation 3.27 since
at far distance the sound pressure and the particle velocity are in phase just like
in a plane wave.
The radial sound intensity at a distance r from the origin of a spherical wave
depends on the radiated sound power P as
P
I= (3.36)
4πr 2
The intensity of sound in a spherical wave drops as 1/r2 and the sound pressure as 1/r
as a function of distance from the point source. We call this the distance law or the
geometrical sound attenuation of a spherical wave. The geometrical sound attenua-
tion is about −6 dB per distance doubling for a monopole.
The power radiated by the monopole can be shown to be
ρ 2 2 ρ4πa 4 2 2 (3.37)
P= ωU = ω u
4πc c
24 Electroacoustics
We note that the monopole is a poor radiator of sound since the radius appears as
a4 and the frequency as ω2 in the equation.
3.8.3 Dipoles
A dipole (or “doublet”) source is an elementary source which can be thought of as
consisting of two monopoles at a small distance b from one another as shown in
Figure 3.4. The dipole monopoles oscillate with the same frequency and with the
same volume velocity but out of phase. In the figure, this property is indicated by
plus and minus signs.
The sound pressure at point A at a distance r will be a superposition of the pres-
sures from each monopole, since the sources are assumed not to influence each oth-
er’s radiation. Because of both positive and negative interferences, the total sound
pressure will vary, not only with distance but also with angle θ.
Assume that the sources are much closer than the wavelength and sum up the
sound pressures. We then find that the ratio between the sound pressure from just the
“plus” monopole, _p0, and the sound pressure from the dipole, _pD, can be written as
pD (r, k, b, θ )  b
= −  jkb +  cos (θ ) (3.38)
p0 (r, k )  r

An important aspect of the sound field surrounding a dipole source is the presence
of what seems to be a transversally polarized wave. On the z-axis, the particle veloc-
ity is directed away from the dipole, but at all other locations there is also a particle
velocity component directed tangentially. In the z = 0 plane, the particle velocity
direction is perpendicular to the plane. A sound-reflecting object such as a sound-
reflecting disk that is placed in the z-plane will cause a sound pressure component on
each side of the disk and sense the force that is generated by the pressure difference
between the two sides of the disk.
The sound power radiated by the dipole will be different from that radiated by
the single monopole. The dipole radiates much less power than a single monopole
A
kb << 1 θ
r
+
b y
–
x
FIGURE 3.4 The coordinate system used in the analysis of the dipole. The two monopole
sources (+/−) are on the z-axis.
z z
kbx << 1
kby << 1
kby << 1
– +
bx y + – – + y
+ –
x x
by by
(a) (b)
FIGURE 3.5 Two examples of quadrupoles; in (a) all the monopole sources are in the z = 0
plane, in (b) all are on the y-axis.
having the volume velocity of one of the dipole elements. The dipole cancellation is
an important asset in many cases of noise control engineering but usually undesir-
able in audio engineering. The cancellation is sometimes called aerodynamic short
circuit or aerodynamic cancellation as discussed in Chapter 10.
3.8.4 Quadrupoles
Quadrupoles are somewhat more complicated elementary sources than dipoles and
can exist in many configurations. The characteristic of a quadrupole is that it consists
of two sources in phase and two sources out of phase in some pattern, but all sources
are always at a distance that is much smaller than the wavelength. One type of quad-
rupole, shown in Figure 3.5a, consists of four monopoles, two in phase and two out
of phase at distances bx and by in the z = 0 plane as shown in the figure.
We can find more complicated radiation pattern of any quadrupole by simple
superposition of the sound pressure due to each of its elementary sources. Quadrupole
sound power radiation is even weaker than that of dipoles and monopoles because of
the even stronger aerodynamic cancellation as discussed in Chapter 10.
3.9 REFLECTION AND TRANSMISSION AT BOUNDARIES

When a sound wave is incident on the boundary between two media, or on a bound-
ary where the characteristic impedance of a medium changes, part of the power in
the incident wave may be reflected and some transmitted. The ratios of reflected and
transmitted powers to the incident wave are determined by the characteristic imped-
ances, Z1 and Z2 and the speed of sound c1 and c2 on each side of the boundary as well
as on the angles of incidence and transmission relative to the boundary.
Because of the interference between the incident sound wave and the reflected
sound wave, one will be able to notice standing-wave phenomena as the waves add
up, in phase or out of phase. By a standing wave is meant the property of the sound
field that results in the envelope of the combination of the incident and reflecting
waves to be fixed, that is, we have a “standing” wave.
26 Electroacoustics
3.9.1 Perpendicular Sound Incidence

Assume a plane wave incident from the side of negative x-values on a plane
impedance boundary at x = 0. Since we have assumed linearity of the media, and
because of the coherence between the incident and reflected waves, the resulting
sound pressures and particle velocities are obtained by simple superposition of the
incident and reflected sound pressures, p̲i and p̲ir, and particle velocities, ui and ur,
respectively.
At the plane x = 0, the boundary conditions to be fulfilled are continuity of
pressure and continuity of particle velocity, that is,
p1 = p2 (3.39)

u1 = u2 (3.40)

Here, p1, p2 and u1, u2 are the sound pressures and particle velocities on the respective
sides of the boundary, x < 0 and x > 0. The sound pressure on the negative x-side of
the boundary is
p1 ( x, k ) = pˆ i e − jk1 x + pˆ r e jk1 x (3.41)

On the positive x-side of the boundary, it is
p2 ( x, k ) = pˆ t e − jk2 x (3.42)

Because of the definition of sound field impedance for a wave moving in the negative
x-direction, we obtain
u1 ( x, k ) = uˆ i e − jk1x + uˆ r e jk1x (3.43)

pˆ i − jk1 x pˆ
u1 ( x, k ) = e − r e jk1 x (3.44)
Z1 Z1
pˆ t − jk2 x
u2 ( x, k ) = e (3.45)
Z2
If p1, p2 and u1, u2 are eliminated from these equations, we further obtain
pˆ i + pˆ r
Z1 = Z 2 (3.46)
pˆ i − pˆ r
The relationships between the sound pressures, in the respective waves, are usually
expressed by a complex pressure transmission coefficient, t, and a complex pressure
reflection coefficient, r, as follows:
pˆ r Z 2 − Z1
r= = (3.47)
pˆ i Z 2 + Z1
pˆ t 2Z2
t = = (3.48)
pˆ i Z 2 + Z1
If the boundary has a complex impedance Z2, the characteristic impedance Z2 is

replaced by Z2 in the expression for the reflection coefficient.
If there are no losses at the boundaries, the incident, reflected, and transmitted
sound intensities (Ii, Ir, and It) must be related as
I i = I r + I t (3.49)
From the side of wave incidence, it seems as if the energy of the transmitted wave is
absorbed. The sound absorption coefficient, α, is defined as
Ir 2
α = 1− = 1 − r (3.50)
Ii
Quasi-plane waves can be generated in long, straight tubes if the tube walls are hard,
and the tubes wide enough, so that the influence of the viscosity of the medium can
be neglected. However, the tubes must not be so wide as to allow other modes of
sound propagation (see Appendix F). Such tubes can be used as acoustical compo-
nents to build various types of acoustical circuits, as further described in Chapter 7.
3.9.2 Reflection of Sound at an Angle

At oblique incidence (that is, when the wave is not incident perpendicular to the
plane boundary), the reflection and transmission coefficients will be dependent on
the angles of incidence and transmission. Two cases will be studied, “normal reac-
tion” and “Rayleigh type of reflection.”
By “normal reaction,” sometimes also called “local reaction,” it is meant that the
sound wave transmitted into medium 2 propagates at right angle to the boundary.
Normal reaction is observed when the sound, for example, due to a guiding structure,
is forced to propagate in a particular direction. This is approximately the case when
sound is transmitted into a porous absorber.
By “Rayleigh type of reflection,” sometimes also called “extended reaction,” it is
meant that the sound wave transmitted into medium 2 propagates at an angle to the
boundary, determined by Snell’s law.
c1 c2
= (3.51)
sin(ϕ1 ) sin(ϕ 2 )
28 Electroacoustics
y y
λ2
Wave moving
down the “tube”
2
1 x 1 x
1 1
λ1 λ1
λ2
Normal reaction Extended reaction
FIGURE 3.6 Oblique sound incidence on a plane impedance boundary.
Extended reaction typically occurs at boundaries between media in which free

wave propagation is possible, for example, between layers of air at different
temperatures.
We will now study the conditions for a plane wave incident at an oblique angle as
shown in Figure 3.6. Besides the requirement for continuity of sound pressure and
particle velocity, it is also necessary for the three sound waves on either side of the
boundary to propagate at the same phase velocity along the boundary. This means
that the angle of reflection must be the same as the angle of incidence.
The wave number components in the y-direction must be the same for all three
waves, since the wave numbers are based on the frequency of sound and the phase
velocity. The wave number, which we have studied so far, is just the absolute value
of the wave vector. In the present case, it is practical to regard the wave numbers
for waves traveling at oblique angles as components of wave vectors in the x- and
y-directions.
3.9.3 Normal Reaction
The obliquely incident sound wave shown in Figure 3.6 can be formulated as
p1 ( x, y, k ) = pˆ i e 1(
− jk x cos( ϕ1 )+ y sin( ϕ1 ))
(3.52)

Since the reflection leads to the wave vector component in the x-direction of the
reflected wave changing sign, one can write the resulting sound field in medium 1,
using superposition, as

( )
p1 ( x, y, k ) = pi + pr = pˆ i e − jk1 y sin( ϕ1) e − jk1 x cos( ϕ1 ) + re jk1 x cos( ϕ1 ) (3.53)
The particle velocity can now be obtained using the definition of sound field imped-
ance for a wave moving in the negative x-direction. We write
pˆ i
u1x ( x, y, k ) = ui + ur =
Z1
( )
cos(ϕ1 ) e − jk1 y sin( ϕ1) e − jk1 x cos( ϕ1) − re jk1 x cos( ϕ1) (3.54)

In the case of normal reaction studied now, the particle velocity in medium 2 is only
in the x-direction, since the sound wave transmitted into medium 2 propagates at
right angle to the boundary. The local complex sound field impedance on the side of
medium 2 at the boundary x = 0 is written as
p2
Z2 = (3.55)
u 2x
Consequently, the boundary conditions for the particle velocity will be fulfilled if
p1 ( 0, y )
Z2 = (3.56)
u1x ( 0, y )

This results in the pressure reflection coefficient, r, being
Z 2 cos( ϕ1) − Z1
r= (3.57)
Z 2 cos( ϕ1) + Z1
We note that the reflection coefficient is dependent not only on the characteristic
impedances of the two media but also on the angle of incidence. The actions of many
porous sound-absorbing materials can be modeled as local reactions. The acoustical
properties of sound-absorbing materials are discussed in Appendix E.
3.9.4 Extended Reaction

In this case the boundary condition regarding the propagation direction is given
by Snell’s law. Of course, the regular requirements for sound pressure and particle
velocity continuity across the boundary still apply.
Using the same approach as the previous one, we obtain the reflection coefficient
in the case of extended reaction as
Z 2 cos( ϕ1) − Z1 cos( ϕ 2)

r= (3.58)
Z 2 cos( ϕ1) + Z1 cos( ϕ 2)
We note that the propagation velocities of sound in the two media influence the pres-
sure reflection and transmission coefficients as given by Snell’s law.
2
 c sin( ϕ1) 
cos( ϕ 2) = 1 −  2  (3.59)
 c1

30 Electroacoustics
In some acoustically particularly interesting cases, for example, in some porous

sound absorbers, the speed of sound in medium 2 is considerably smaller than that
in medium 1, that is, c2 ≪ c1.
This means that cos(φ2) ≈ 1, that is, φ2 ≈ 0, and on the whole independent of the
angle of incidence φ1. The propagation in medium 2 is then approximately equal to
the case of normal reaction, and the propagation in the y-direction can be neglected.
When the angle of incidence φ1 becomes close to 90°, that is, for grazing inci-
dence, the pressure reflection coefficient r will become −1 if medium 2 has a finite
characteristic impedance. This means that all energy will be reflected and that the
phase of the reflected wave will be inverted. In many situations, this leads to the
phase inverted wave canceling sound which is not reflected, leading to substantial
cancellation of the sound pressure close to the sound-absorbing surface. An example
of this type of cancellation at large angles of incidence is when sound from a loud-
speaker driver propagates almost parallel to a sound-absorbing blanket on the face
of a loudspeaker box so that edge diffraction is reduced.
3.10 HUYGENS’ PRINCIPLE

Huygens’ principle states that each wavefront can be constructed from the previous
wavefront through the summation of the sound pressure due to small “wavelet” emit-
ting elements of the previous wavefront. The elements radiate unidirectionally into the
solid angle 2π. The new wavefront is the mathematical envelope of these “wavelets.”
Another way of phrasing Huygens’ principle is that every point on a plane vibrat-
ing surface may be considered as an outgoing wave. Because of the presence of a
mirror source in the baffle, following Equation 3.35, the pressure generated by an
element on a baffle having volume flow udS is
e − jkr
dp = jω ( udS ) ρ (3.60)
2πr
If the surface is an infinite baffle, there will be no interference from the back, and
hence no diffraction; so, the resulting sound pressure is found by integration of the
contributions from the vibrating area:
e − jkr

p(r, k ) = jωρ
2πr ∫ u dS (3.61)
This integral is called the Huygens–Rayleigh integral, and is exact if the radiating
surface is plane piston set in a rigid baffle. We will use this relationship in Chapter 12.
3.11 SCATTERING
Diffraction can be regarded as a special case of scattering. Usually, however, the
term scattering is used to describe the generalized reflection of sound by limited
size surfaces, objects, and surface unevenness. Backscattering is the term used for
the sound that is scattered back toward the sound source, while scattered sound
directed otherwise is called forward scattering. In ultrasonics, backscattering is of
most interest, while in electroacoustics and audio, forward scattering is of most
interest.
The idea of a scattering cross section is often used when considering the scatter-
ing by free objects. The scattering cross-section area SS (m2) is defined as the area
that a scattering object appears to have as it receives and scatters the sound power
from an incident plane wave.
Scattered wave power

SS = (3.62)
Incident wave intensity
Open resonators (Helmholtz resonator flasks and quarter-wavelength tubes, for

example) are effective in scattering sound at resonance. The air inside a Helmholtz
resonator, a guitar, or a violin is set into vibration by an incoming sound wave
that is likely to be quasi-plane. However, the energy which is reradiated by the
resonator will be almost omnidirectional since the resonator mouth must be small
compared to the wavelength to effectively function as a resonator. The scattered as
well as absorbed power will be the largest at the resonator’s resonance frequency.
When the resonator is critically tuned, the absorbed and scattered power will be
the same.
3.12 DIFFRACTION
A special case of scattering is diffraction. The term is typically used when the scat-
tering is by a simple edge or impedance discontinuity. Diffraction can be understood
from the Huygens principle discussed previously. Often in acoustics the only diffrac-
tion considered is that in the shadow zone behind an object, however—because of
the resolution of human hearing— also diffraction to the front of a barrier is some-
times of interest as regards the reproduction of sound by a loudspeaker.
Here, we limit ourselves to discussing the situation typical in using loudspeaker
drivers built into the surface of rectangular boxes. For a more general treatment, the
reader should consult Refs. [5–8]. Consider a very small monopole-like driver set on
the surface in a plane rigid rectangular loudspeaker box front. The box front is much
larger than the wavelength. In the region between the driver and the sharp 90° box
corner, waves progress as shown in Figure 3.7.
Usually, we are not interested in other diffracted sound field components than
those that interfere with the direct sound component from the driver within a short
time after the arrival of the direct sound. (Sound that reflects off walls, etc., is already
subject to scattering and absorption by the walls plus the frequency irregularities of
the loudspeaker as it radiates off its normal.)
There are many corner (edge) diffraction models. A good way of analyzing the
problem of corner diffraction can be shown to be to leave the traditional frequency-
domain formulations and instead use a time-domain formulation, as in Refs. [5–7].
In the time-domain formulation, the impulse response of the edge diffraction is com-
puted and compared to the direct sound from the loudspeaker. Two advantages of
32 Electroacoustics
Progressing waves
Loudspeaker
far away
Diffracted
wave front
Loudspeaker box corner
FIGURE 3.7 Sound waves from a very small loudspeaker in front of a rigid loudspeaker box.
Receiver is far away but

z on box front plane normal
(through driver)
rreceiver rdirect x1
Box corner
edge element
under consideration
x2
rsource Source point

on box front plane
Box wall is at right
angle to the box front plane y
FIGURE 3.8 One straight edge along the front of the loudspeaker box. The receiver is
assumed to be far away from the box but on the normal of the sound source on the loud-
speaker box front.
these time-domain models are that they are intuitive and that they do not make the
high-frequency assumption otherwise common.
Figure 3.8 shows the situation of most interest. The listening position is far away
from the loudspeaker but still on the normal of the driver.
The box front edges are subdivided into small Huygens’ source-like elements, and
each source element contributes a short pulse [6] in the impulse response. This is the
equivalent time-domain formulation of Equation 3.60. The arrival time of each such
pulse is found from the length of the path from the source over the edge element
(stretching from x1 to x2 along the edge) to the receiver. The total impulse response
is then found from the summation of the short pulse contributions of all the source
elements along the edge. The direct impulse contribution from the loudspeaker is
then added, and the total is transformed into the frequency domain by means of the
inverse Fourier transform to show the frequency response.
For simplicity, just consider one of the box front edges. The impulse response
contribution of each element on this edge will also depend on a directivity factor
Fβ that is a function of the edge element’s location along the x-axis [7]. The impulse
response contribution of the diffracted wave is found as follows:
x2
3c  rsource + rreceiver  Fβ
h (t ) = −
2 ∫ δ  τ −
x1
c 
 rsourcerreceiver
dx (3.63)

In the idealized case of the receiver being on the normal of the driver, and far from
the loudspeaker compared to the dimensions of the loudspeaker box front, the direc-
tivity factor Fβ will be about unity. Figure 3.9 shows an example of the calculated
impulse response for a small loudspeaker in a rectangular loudspeaker box. With
the direct sound positive as shown, the impulse response of the first-order diffracted
sound components is negative.
Direct sound
Diffracted sound
Sound pressure
Time [ms]
1 2
FIGURE 3.9 Calculated sound pressure at the receiver due to diffraction from the edges of
a loudspeaker box. The rectangular box has front baffle dimensions 0.4 by 0.64 m and depth
0.32 m. The monopole is placed on the front baffle 0.2 m from the top and side edges. The
receiver is 10 m away directly on the normal to the front baffle going through the monopole.
(Adapted from Vanderkooy, J., J. Audio Eng. Soc., 39, 923–933, 1991.)
34 Electroacoustics
The theory can also be used to calculate higher order diffraction. First, the first
order needs to be calculated for all the front box edges and added, then the sec-
ond order (due to the first order), and so on. For loudspeaker boxes, higher orders
and combinations of edge diffraction components are usually not as significant as
first-order diffraction components when the receiver can see the driver. The edge
does not create any second-order sources along itself. This interaction is included
in the integral. Software to calculate the impulse response due to diffraction may
be found in Ref. [8].
Since large loudspeakers are directional, and because the loudspeaker diaphragm
generates sound from all its area, there will be a low-pass filter action reducing the
diffracted sound. For these reasons the diffraction of such a loudspeaker will not be
as noticeable as with a small loudspeaker.
3.13 ACOUSTIC RECIPROCITY

The reciprocity theorem states that for sound waves that the sound pressure at a point
A excited by a volume source at a point B will be the same in phase and amplitude as
it would have been at B if A had been the location of the source. Reciprocity gives a
simple and attractive way of measuring the transfer function between two points by
interchanging the source and receiver locations. A microphone is likely to be much
easier to move than a loudspeaker, and interchanging their location can, for example,
give a possibility to find the optimum location of the loudspeaker.
REVIEW QUESTIONS
3.1 Which are the equations on which the wave equation for longitudinal waves in
air is based?
3.2 How can we see from the solutions to the wave equation for longitudinal waves
in air that there is no dispersion?
3.3 Discuss the interrelationships between the various impedance definitions used
in acoustics.
3.4 Show how the data in Figure 3.5 can be used to calculate the attenuation coef-
ficient m.
3.5 How is the radiated sound power related to the properties of a sound source?
3.6 Why do dipoles and quadrupoles differ from monopoles with regard to sound
power radiation?
3.7 What are the differences between the assumptions for the cases of local and
extended reaction for sound incident on a boundary between two media?
3.8 How is the reflection coefficient related to the sound absorption coefficient
of a surface?
3.9 Explain Huygens’ principle.
3.10 What is the difference between the terms scattering and diffraction as used in
audio and electroacoustics?
PROBLEMS
3.1 A spherical, sinusoidal wave travels away from a point source in air. The
frequency is 1 kHz. The acoustic pressure amplitude at 1 m from the point
source is 0.1 Pa. For air, ρ = 1.20 kg/m3, c = 343 m/s.
Tasks:
a. Plot the acoustic pressure amplitude and the particle velocity amplitude
as functions of the distance from the source.
b. Plot the phase angle between acoustic pressure and particle velocity as
a function of the distance from the source.
c. Graph the functions as functions of kr, that is, the product between wave
number and distance.
3.2 The surface of a sphere vibrates radially, and its volume velocity is frequency-
independent. The sphere has a radius of 0.05 m. The amplitude of the surface
vibration is 0.001 m at 100 Hz.
Tasks:
Calculate the radiated sound power as a function of frequency in the range
100 Hz–10 kHz.
3.3 The sound pressure in front of a hard surface varies with distance from the
surface due to interference between the incident and reflecting waves.
Task:
Calculate the sound pressure increase, relative to the free field pressure, that
one obtains when measuring the sound pressure level at a distance of 0.1 m in
front of a flat, hard surface having a reflection coefficient of 1. The sound wave
is perpendicularly incident on the surface.
3.4 Three small loudspeakers are connected to the same amplifier, such that their
radiation is correlated and in phase. The RMS value of the sound pressure at
1 m distance in a reflection-free environment was measured for each loud-
speaker radiating alone, giving the following set of results: p1,RMS = 0.63 Pa,
p2,RMS = 0.11 Pa, and p3,RMS = 0.20 Pa. Assume that the vibration of any of the
loudspeakers is unaffected by the others.
Tasks:
a. Determine the sound pressure level at the measuring point when the
distance to the three loudspeakers is the same.
b. As in (a), but assume that source 1 has been modified so that p1,RMS is
reduced to half its value.
c. As in (b), but assume that source 1 has been moved away from the mea-
surement point so that p1,RMS is reduced to half its value.
36 Electroacoustics
3.5 Sound is attenuated in air by various mechanisms as indicated in Figure 3.3.
Task:
Calculate the change in frequency response of the sound pressure at 10 and
100 m distance from a sound source compared to that at 1 m due to attenuation
caused by the properties of air at a temperature of 20°C and a relative humid-
ity of 40%. Consider the range 0.1–10 kHz.
3.6 Sound will be reflected by an impedance change as given by Equation 3.47.
Tasks:
a. Show how it would be possible to measure the mass per unit area of a
thin foil using high-frequency sound.
b. How can the foil be used as an acoustic filter?
3.7 Multiple impedance changes will cause repeated reflections that can be
analyzed for amplitudes and time delays.
Task:
A plane wave is generated by the piston in the system shown in the figure
below. Calculate the density of medium 2 using the time between reflections
and the amplitudes of the incident and reflected waves in mediums 1 and 2.
Incident and Transmitted and Transmitted
reflected waves reflected waves wave
Vibrating
piston
Medium 1, Z1 Medium 2 Medium 3, Z1
l1 l2
3.8 Two identical small loudspeakers are connected to the same amplifier so that
their radiation is correlated and in phase. The RMS value of the sound pressure at
1 m distance from either loudspeaker in a reflection-free environment is 0.63 Pa.
Assume that the vibration of any of the loudspeakers is unaffected by the other.
Tasks:
a. Determine the sound pressure at 10 m distance from the center of the
line between the two loudspeakers when they are placed 1 m from one
another assuming the frequency is 50 Hz.
b. As in (a), but assuming the frequency is 1 kHz.
c. Determine the particle velocity amplitude and direction at this distance
for the two cases.
3.9 Now assume the two loudspeakers in Problem 3.8 to be out of phase.
Task:
Determine the particle velocity amplitude and direction at this distance for
the two cases.
3.10 Two identical small loudspeakers are connected to the same amplifier so that
their radiation is correlated and in phase. The loudspeakers are placed next to
one another at 3 cm distance so that their diaphragms almost meet. The sound
pressure level at 1 m distance from either loudspeaker in a reflection-free envi-
ronment is 90 dB. Assume that the vibration of any of the loudspeakers is
unaffected by the other.
Task:
What will be the sound power output from the loudspeakers at 50 Hz?
3.11 Assume the two loudspeakers in problem 3.10 working 180° out of phase.
Task:
What will be the sound power output from the loudspeakers at 50 Hz?
REFERENCES
1. Beranek, L. L., Acoustics, McGraw-Hill, New York (1954).
2. Kinsler, L. E. and Frey, A. R., Fundamentals of Acoustics, 4th edn., John Wiley & Sons,
New York (1999) ISBN-13: 978-0471847892.
3. Cremer, L. et al., Principles and Applications of Room Acoustics, Vol. 2, Applied
Science Publishers, New York (1982) ISBN-13: 978-0853341147.
4. Kuttruff, H., Ultrasonics: Fundamentals and Applications, Springer, New York (1991)
ISBN-13: 978-1851665532.
5. Vanderkooy, J., A simple theory of cabinet edge diffraction, J. Audio Eng. Soc., 39,
923–933 (1991).
6. Svensson, U. P., Fred, R. I., and Vanderkooy, J., Analytic secondary source model of
edge diffraction impulse responses, J. Acoust. Soc. Am., 106, 2331–2344 (1999).
7. Svensson, U. P. and Wendlandt, K., The influence of a loudspeaker cabinet’s shape
on the radiated power, Proceedings of Baltic Acoustic 2000, published in Journal of
Vibroengineering, No. 3(4), pp. 189–192, Vilnius, Lithuania (2000).
8. Svensson, U. P., Edge Diffraction Toolbox. http://www.iet.ntnu.no/∼svensson/software/
index.html (sampled November 2011)
4 Waves in Membranes
and Plates
4.1 INTRODUCTION
Gases oppose volume changes resulting in longitudinal waves as a response to a
disturbance in gas volume. In contrast to sound propagation in gases, vibration
propagation in solids can involve many types of waves. Solid media not only
exhibit elasticity but also shear. Infinite solid media can carry both longitudinal
and transverse waves. Finite solid media can also carry many other wave types,
particularly bending, transverse, torsional, quasi-longitudinal, and different types
of surface waves. Some of the most important and common wave types will be
discussed in this chapter.
Usually we think of mechanical structures as being fairly rigid, moving as
solid bodies. This is only relevant as long as the body is small compared to the
waves that can be excited in the material of the body. Because of the elasticity and
mass of materials, there will be resonance and wave propagation in all mechanical
bodies. A characteristic phenomenon of sound propagation in solid structures is the
conversion between wave types at boundaries and edges. This wave conversion helps
distribute energy between the different modal systems.
Membranes and diaphragms are important elements in electroacoustics since they
are used to convert between force and pressure. The terms diaphragm and membrane
are used interchangeably in mechanical engineering. Diaphragm emphasizes the
function, whereas membrane emphasizes the material. In this book, we will use the
term membrane to mean a thin limp sheet under tension and the term diaphragm to
mean a rigid thin plate, piston, cone, or dome.
Membranes are used in condenser microphones since the electroacoustic design
is simple with a membrane that is under high mechanical tension. Another use of
membranes is in isodynamic and electrostatic loudspeakers where the dynamic force
on the membrane is evenly distributed and the moving mass must be kept small
because of the weak magnetic or electrostatic forces.
Diaphragms on the other hand are usually used, for example, in conventional
electrodynamic and piezoelectric loudspeakers. In these we often wish the motion
of the diaphragm to be that of a rigid piston. Since thin, low mass plates usually do
not have sufficient stiffness to be used as pistons, it is common practice to transform
a sheet to a cone, pyramid, or of similar shape. A dynamic force is then applied to
the apex along the axis of symmetry. Stiff plastic sheets are usually shaped into
loudspeaker driver cones by thermoforming.
Alternatively, cones can be made to have such high internal damping that wave
motion and associated resonances are effectively suppressed. The damping can be
39
40 Electroacoustics
FIGURE 4.1 Instantaneous displacement pattern of a conical loudspeaker diaphragm

showing breakup at one of the diaphragm’s circumferential bending wave resonances. 6 in.
diameter paper cone diaphragm at 1050 Hz. (Photo by Mendel Kleiner.)
achieved by external or internal viscoelastic layers or by shaping the cone from a

wave impregnated by a viscoelastic compound.
Figure 4.1 shows the measured instantaneous displacement pattern of a conical
loudspeaker diaphragm that has a bending wave resonance at about 1 kHz. The
modal behavior is clearly visible.
4.2 WAVE TYPES IN INFINITE MEDIA

4.2.1 Longitudinal Waves
In air, longitudinal waves are characterized by sound pressure and by particle motion
in the propagation direction of the wave, as shown in Figure 4.2a.
y (a)
Direction of propagation
x
z
(b)
FIGURE 4.2 Instantaneous displacement in (a) longitudinal and (b) quasi-longitudinal waves.
Waves in Membranes and Plates 41
In solids, the longitudinal wave is characterized by tension as well as by particle

motion in the direction of the wave. The wave equation for one-dimensional
longitudinal waves in solids is
∂ 2 ux
+ kL2 ux = 0 (4.1)
∂x 2
where
ux is the particle velocity in the x-direction
kL is the wave number (kL = ω /cL)
The wave propagates with a velocity cL , which depends on the elastic constants of
the solid as
D
cL = (4.2)
ρ

where
D is a constant representing the longitudinal stiffness of the material (expressed
in units of Pascal) and ρ its density
Young’s modulus, E, is related to the longitudinal stiffness, D, by
 2ν2 
E = D 1 −  (4.3)
 1− ν 
Here ν is Poisson’s ratio for the material, sometimes called the cross-contraction
number. For metals and rigid polymers ν is found by measurement to be about 0.3,
resulting in D ≈ 1.35E.
The form of the wave equation shows that waves propagate with the same speed,
irrespective of frequency. This is called nondispersive propagation.
4.2.2 Transverse Waves
The shear wave is a transverse wave that is characterized by shear forces and particle
velocity that is perpendicular to the propagation direction of the wave. The motional
pattern of shear waves is shown in Figure 4.3.
The wave equation for shear waves is written in the following way for the case of
wave propagation in the x-direction and particle velocity in the y-direction:
∂ 2 uy
+ kT2 uy = 0 (4.4)
∂x 2
where
uy is the particle velocity in the y-direction
kT is the wave number (kT = ω /cT)
42 Electroacoustics
x
z
FIGURE 4.3 Instantaneous wave motion displacement in a shear wave.
The wave propagates with a velocity cT, which depends on the elastic constants of
the solid as
G
cT = (4.5)
ρ

where
G is the shear modulus
ρ is the density of the solid
The shear modulus is given by
E
G= (4.6)
2 (1 + ν )

Assuming a Poisson’s ratio ν of 0.3, one obtains the ratio between the transverse and
longitudinal velocities as cT /cL ≈ 0.5.
In a plate that carries shear waves, the cross-contraction will cause quasi-
transverse waves. Because the cross-contraction is small, such a wave motion has
little coupling to the surrounding air.
4.3 WAVE TYPES IN MEDIA OF LIMITED EXTENSION

4.3.1 Quasi-Longitudinal Waves in Plates
Quasi-longitudinal waves in a plate are characterized by cross-contraction and have
the motional pattern as shown in Figure 4.2b. The quasi-longitudinal wave will
move at reduced speed because of the cross-contraction. The velocity cQL of a quasi-
longitudinal wave in a plate is given by an equation similar to Equation 4.1, but with
cL and D replaced by cQL and EQ, where
EQ
cQL = (4.7)
ρ

and
E
EQ = (4.8)
1 − ν2
The wave impedance of such a quasi-longitudinal wave is given by
ZQL = EQρ = ρcQL (4.9)

Note also that the quasi-longitudinal wave is characterized by poor sound radiation
capability since the out-of-plane movement resulting by the cross-contraction is
small.
Since quasi-longitudinal waves move at high speeds, the corresponding
wavelengths will be long. Typically, the wave speeds for metals are of the order of
5000 m/s, so even at 20 kHz the wavelength will be about 0.25 m, much larger than
the dimensions of typical audio transducers used for this frequency. The solid metal
and hard plastic parts of audio transducers can often be considered discrete masses
in their operating frequency range, but this should always be checked by testing.
4.3.2 Out-of-Plane Vibration
Three types of out-of-plane (transverse) vibrations are of special interest for
electroacoustic transducers since they may occur in the diaphragms of microphones
and loudspeakers. Such vibrations are transverse waves in membranes, shear waves
in sandwich sheets, and bending waves in bars and plates.
4.3.3 Membrane Vibration

In a membrane, the restoring force due to stiffness is small compared to that due
to membrane tension. The tension is produced by in-plane forces at the membrane
edges. At low frequencies, when the differential sound pressure Δp between the two
sides of the membrane is uniform over all points of the membrane, the displacement
ξ for a circular membrane will be [1]
a2  r2 
ξ = ∆p 1 − 2  (4.10)
4T  a 

where
T is the membrane tension along the edge of the membrane
r is the distance from the center of the membrane
a is the membrane radius
Typical examples of membranes in electroacoustics are those of condenser micro-

phones and electrostatic loudspeakers. Because membranes are often coupled
44 Electroacoustics
to the air in cavities, the resonance frequencies of their modes will be different
from free membranes; this applies, for example, to the membrane of condenser
microphones.
Membranes are usually avoided in voice coil electrodynamic loudspeaker
drivers since mostly rigid nonresonant pistonic diaphragm motion is desired.
In microphones, however, small size, simple manufacturing, and relatively low
impedance membrane are attractive from an engineering viewpoint. Devices
such as electrodynamic and piezoelectric microphones use rigid domes instead of
membranes.
Assume the membrane in the y = 0 plane so that the transverse vibration amplitude
in the y-direction is ξ. Membrane motion is described by the equation [2]
∂2 ξ ∂2 ξ
+ + km2 ξ = 0 (4.11)
∂x 2 ∂z 2
Here, the wave number km is
ω
km = (4.12)
cm
and the speed of the waves is
T
cm = (4.13)
m′′
Here, m″ is the mass per unit area of the membrane. For a rectangular membrane,
the solutions to the wave equation will be similar to those found for a rectangular
room.
Assume a rectangular membrane between x = 0, x = lx, z = 0, z = lz. The boundary
condition is ξ = 0 along these lines. The transverse vibration of free membrane waves
will be resonant with the standing wave mode shape
Ψ ( x, y ) = sin ( kmx x ) sin ( kmz z ) (4.14)

and the wave numbers kmx and kmz are
 q πx   q πz 
kmx =  x  and kmz =  z  (4.15)
 lx   lz 

where qx and qz are natural numbers 1,2, etc. The mode frequencies are given by the
condition
km2 = kmx
2
+ kmz
2
(4.16)

and the resonance frequencies are
2 2
c  qx   qz 
fq x , qz = m  l  +  l  (4.17)
2  x  z

Rectangular membranes stretched over a supporting frame are typically found in
push–pull electrostatic loudspeakers. The resonance frequencies of the membrane
are typically chosen to be below the frequency operating range of the loudspeaker.
The electrostatic loudspeaker electrodes have little acoustical influence except at
very high frequencies so the membrane motion of the first few modes is primarily
damped by the radiation of energy to the surroundings. Figure 4.4 shows two of the
lowest order modes for a rectangular membrane.
2
0
5
–2
0
–2
0 –5
(a) 2
2
0
5
–2
–3 0
–2
–5
–1
(b)
0
FIGURE 4.4 Instantaneous displacement of two low-order modes in a rectangular mem-

brane. (a) (qx,qz) = (1,1); (b) (qx,qz) = (0,1).
46 Electroacoustics
A circular membrane can have wave propagation radially and circularly so the
transverse motion ξ of the free membrane vibration can be written as
∧ ∧
ξ = ξ Ψ (r , ϕ ) = ξ Ψr (r ) Ψϕ ( ϕ ) (4.18)

where Ψr(r) and Ψφ(φ) are solutions to the wave equation for radial and circumferential
waves, respectively. The wave equation for the circular membrane is written in
circular coordinates r and φ as [2]
∂ 2 Ψϕ ( ϕ )
+ qϕ2 Ψϕ ( ϕ ) = 0
∂ϕ 2 (4.19)
∂ Ψr (r ) 1 ∂Ψr (r )  ω 2 qϕ 
2 2
+ +  2 − 2  Ψr ( r ) = 0
∂r 2 r ∂r c r 

Since the circular waves must be periodic in φ of the free circular transverse
membrane, their motion pattern must be of the form

Ψϕ ( ϕ ) = 
( )
sin qϕ ϕ
(4.20)

 ( )
cos qϕ ϕ
Here qm is an integer. Assume that the radius of the membrane is a. The boundary
condition for a circular membrane is ξ(a) = 0. The standing wave mode shape of the
membrane motion will be
 ωr 
( )
Ψ(r , ϕ) = cos qϕ ( ϕ + ϕ 0 ) J qϕ   (4.21)
 c 

Here, φ0 is an arbitrary angle. Since the membrane is fastened along its radius at
r = a, the allowed frequencies are those that make Jqφ (ωa/c) = 0. These frequencies
ωqrqϕ turn out to be [2]
1 T
ω1,0 ≈ 2.40 ω1,1 ≈ 1.59ω1,0 ω1,1 ≈ 2.14ω1,0
a m′′ (4.22)
ω 2,0 ≈ 2.30ω1,0 ω 2,1 ≈ 2.65ω1,0 ω1,1 ≈ 2.92ω1,0
So the resonance frequencies of the circular membrane’s modes are not harmonic.
Typically, a condenser microphone is designed such that the lowest order mode
will be at the low-pass cutoff frequency of the microphone, usually above the
audio frequency range. Because of the high resonance frequencies of the higher
order modes, they are outside the useful frequency range and are of little interest.
(a) (b)
(c)
FIGURE 4.5 Instantaneous displacement characteristics of three low-order modes in a

circular membrane. (a) (qr,qφ) = (1,0), (b) (qr,qφ) = (1,1), (c) (qr,qφ) = (2,0).
Because of the coupling of the membrane vibration to the air inside the micro-
phone capsule, the microphone membrane will behave like a drum and the mode
frequencies listed earlier will be changed [2]. The electrostatic charge will also
slightly alter the tension of the membrane. The damping of the wave motion will
typically be provided by viscosity of the air between the membrane and the static
electrode.
Figure 4.5 shows the shape of some modes that are likely to be excited in condenser
microphone membranes at very high frequencies.
4.4 TRANSVERSE WAVES IN THIN BARS AND PLATES

Transverse out-of-plane waves in plates are of particular acoustical interest since
they may couple well to the surrounding air, i.e., the wave motion in the plate can
radiate energy easily or an external sound field can easily excite vibration in the
plate. The out-of-plane waves are usually excited by a transducer, such as a loud-
speaker driver’s voice coil or an electrodynamic vibrator. Two transverse wave types
of large interest in electroacoustics are shear and bending waves.
4.4.1 Shear Waves
Shear waves are important in sandwich-type materials such as that shown in
Figure 4.6. In many of these materials, a core distance material is bounded by thin
skins. Sometimes, damping materials are integrated into the core. The shear waves
48 Electroacoustics
x
z
FIGURE 4.6 Example of a thin sandwich sheet and of an instantaneous deformation pattern
of a shear wave in the sheet. (Photo by Mendel Kleiner.)
in sandwich materials have a velocity that is determined by the combination of the

properties of the top and bottom skins and the core. The sandwich core may be
compressible, such as foam plastic cores, or incompressible such as honeycomb
cores. The latter cores have little bending stiffness. For a sandwich sheet having
an incompressible core, the wave motion will be in the form of shear waves over a
frequency interval, but for frequencies below and over this interval the wave motion
will be in the form of bending waves described in the next section.
The mode index, sheet geometry, and the speed of the transverse waves will
determine the modal pattern on the sheet and thus the radiation of sound. If the
distance between the antinodes (the points where the amplitude of the standing
wave pattern is a maximum) is small compared to the wavelength, the radiation
from the antinodes will cancel out from dipoles and quadrupoles, as described in
Chapter 3. Each mode will have its own particular radiation pattern and radiation
efficiency [3,4].
In audio engineering, sandwich sheets are used in several ways. One use is for
small, plane, very stiff diaphragms. These can be considered rigid and essentially
nonresonant at audio frequencies [5]. It is quite difficult to manufacture conical
sandwich plates although these turn out to have superior stiffness properties. Another
use is in large, resonant sheets that feed energy primarily into the reverberant field
of the room. In this case the aim is to design low surface mass, resonant diaphragms
that have high modal density in the audio range.
4.4.2 Bending Waves in Plates

The bending wave is the most important type of transverse wave in homogeneous
plates. Bending waves on a plate can be excited by a force along a line or at a point
or by an incident sound wave; such waves are called forced bending waves. The
wavelength of free bending waves is determined by the frequency and mechanical
x
z
FIGURE 4.7 Example of an instantaneous deformation pattern of a bending wave in a

thin plate.
properties of the plate, whereas that of forced waves depends on the pressure
distribution of the sound field around the plate.
Since bending waves are typical for thin flexible plates, their coupling to waves
in air is of considerable interest. The coupling is important both from the viewpoint
of reception and radiation. The radiation by bending waves will be further discussed
in Chapter 10.
The motion in bending waves is due to interaction between tension, shear, and
mass. Figure 4.7 shows the motional pattern of a bending wave. The bending wave
field variables are rotation, transverse particle velocity, bending moment, and shear
forces. The rotation is around the z-axis in the figure.
Because of the four bending field variables, the bending wave equation will have
a quite different form from the wave equations studied earlier. Expressed in particle
velocity uy at right angles to the plate, the equation is written as
∂ 4 uy ω 2 m′′
− uy = 0 (4.23)
∂x 4
B′
where
B′ is the bending stiffness per unit length
m″ the mass per unit area of the plate
The bending stiffness B′ for a plate is given by
Eh3
B′ = (4.24)

(
12 1 − υ2)
where E is Young’s modulus and h the thickness of the plate. The plate can be
considered thin as long as its thickness is less than one sixth of the bending wave’s
wavelength λB.
The more complex wave motion also results in different wave propagation
behavior. The solution is of the form
uy ( x, kB ) = Ae ± jkB x + Be ± kB x (4.25)

Here kB is the wave number for the bending wave. Insertion of this solution into
Equation 4.23 gives the wave number as
50 Electroacoustics
ω2 m′′
kB = 4 (4.26)
B′
Propagating waves are characterized by B = 0 and nonpropagating waves by A = 0

in equation 4.25. One finds nonpropagating waves close to discontinuities, such as
driving points, edges, and so on. Both waves are characterized by the same wave
number.
Since the wave number for bending waves is not proportional to frequency but
rather to the square root of frequency, the propagation velocity becomes frequency
dependent. Waves that have frequency-dependent phase velocity are called dispersive
waves.
The phase velocity is defined as the velocity at which the instantaneous phase of
the wave motion propagates, i.e.,
ω
cB = (4.27)
kB
Using these definitions, one can show that the phase velocity for the bending wave
in a plate as
ω ω2 B′
cB = = 4 (4.28)
kB m′′
The phase velocity of a bending wave in a thin sheet is usually much lower than the
quasi-longitudinal wave speed for the plate material.
The bending wave phase velocity is equal to that of sound waves in the surrounding
air at the so-called critical frequency, fc. The critical frequency is sometimes also
called the coincidence frequency. Using the condition cB = c, one finds the equation
for the critical frequency
c02 m′′
fc = (4.29)
2π B′
It is instructive to draw the bending wave phase velocity as a function of frequency

as is done in Figure 4.8, along with the wave number line for sound in air.
Slow bending waves have frequencies below fc, whereas fast bending waves have
frequency above fc. The bending stiffness is decisive for fc; the stiffer the plate the
lower the critical frequency.
It is convenient to use the coincidence number, Kc, defined as
K c = hfc (4.30)

to characterize materials. Table 4.1 shows the coincidence number for some
materials, which are commonly needed. Typically, the coincidence number is in the
10
Normalized phase velocity
Fast waves
3
c0
1
Slow waves
cB
0.3
0.1
0.02 0.05 0.1 0.2 0.5 1 2 5 10 20
Normalized frequency f/fc
FIGURE 4.8 Phase velocity cB for bending waves in a thin plate compared to the velocity
of sound in air c.
TABLE 4.1
Data for Some Common Materials in Electroacoustics
Density Young’s Modulus Coincidence
Material 103 kg/m3 109 N/m2 Number m/s
Aluminum 2.7 69 12
Steel (5% Nickel) 7.8 190 12
Glass 2.6 70 12
Paper 0.9 (2) 41
Polypropylene 0.9 3.4 32
Polyvinyldifluoride 2.1 0.3 163
Polystyrene 1.1 1.3 57
Polyvinylchloride 1.3 3.4 38
10–40 m/s range. Note that for common construction materials and thicknesses, the
critical frequency of the bending waves falls inside the audio frequency range. This
also applies to typical loudspeaker and microphone construction materials, such as
various plastics and papers.
Free bending waves propagate with the phase velocity cB given by the bending wave
equation (Equation 4.27). The wavelength of a free bending wave is consequently
determined by the material constants and the thickness of the plate in contrast to that
of a forced bending wave.
A finite plate where waves are reflected from the edges can be resonant. The
bending waves and the shear waves have different phase velocity behavior. The
modal density of the bending wave fields in a resonant plate will depend on their
phase velocity.
The modal density of a bending wave field is low and independent of frequency,
because of the frequency-dependent phase velocity of the bending waves.
52 Electroacoustics
The bending wave resonance frequencies of a rectangular plate, freely suspended at

its edges, are given by
π B′  q  2  q  2 
fq x , q y =  x  +  y   (4.31)
2 m′′  lx   ly  
 
where
lx and ly are the lengths of the sides of the plate
the mode indices qx and qy are natural numbers (i.e., 0, 1, 2, 3, …)
The mode shapes generated by bending waves on a freely supported rectangular

plate have a sinusoidal pattern. Equation 4.31 applies approximately to clamped
plates at high mode indices.
Figure 4.9 shows the mode shapes at three resonances of a circular plate that is
clamped at the edges. We see that the “effective” moving area is smaller than that for
the membrane modes shown in Figure 4.5. Such circular plates are typically found
as diaphragms in transmitters and receivers that need to be mechanically sturdy. The
drawback is lower sensitivity than that achieved by a membrane.
(1,0) (1,1)
(2,0)
FIGURE 4.9 Instantaneous displacement characteristics of low-order modes in a thin cir-

cular plate clamped at its edge.
At high frequencies, there are many modes and the average bending wave field
mode density dN/df in a plate is then approximately frequency independent:
dN S m′′
≈ (4.32)
df 2 B′
Here, S is the surface area of the plate that is assumed not to have an “extreme”
shape. In contrast, the modal density of rooms dN/df at high frequency is a function
of both frequency f and room volume V:
dN 4πV 2
≈ 3 f (4.33)
df c
Bending waves have low mechanical input impedance. The mechanical input
impedance Z MB when exciting an infinite plate at a point is
Z MB = 8 m′′B′ (4.34)
For a finite plate, the input impedance will oscillate between mass and compliance
behavior, just as the acoustic input impedance into a tube or a room. Because of
losses, the mechanical input impedance of a plate will tend toward the value given by
Equation 4.34, even for a finite plate, at least for high frequencies. Taking frequency
averages over suitably wide bands will further increase this tendency.
4.5 AUDIBILITY OF RESONANCE CHARACTERISTICS

Because the resonances of plates often have long decay times and high Q factors,
they tend to introduce “ringing” sounds in transducers as the resonances are excited
by the applied electric or acoustic signals. The audibility of such ringing depends on
the mode frequency and decay time as well as modal density.
Tests involving critical listening to resonant systems such as rooms and plates
show that there is a critical region in which insufficient modal density in the transfer
function leads to audible coloration [6]. Figure 4.10 shows the critical region and the
modal density of some rooms and plates.
The tests also showed that changing the geometry of the plate edges did not
noticeably change the coloration induced by the plate resonances. In a situation
where the signals that are listened to are being played back over several loudspeakers
or where the signals are feeding the reverberant field of the room, the coloration may
be masked by the coloration due to the room’s reverberant field.
4.6 SANDWICH SHEETS

In a sandwich sheet the core distance material is bounded by thin skins. The
high static stiffness of a sandwich sheet is due to the core that is designed to
only have shear movement, without bending or compression. The advantages of
54 Electroacoustics
Room volume 56 m3 Gold foil

10
reverberation plate
Modal density (modes/Hz)
5
Region of audible
2 coloration due to
insufficient modal
density in
1 reverberation
Room volume
0.5 7 m3
Steel sheet
0.2 reverberation
plate
0.1
0.1 0.2 0.5 1 2 5 10
Frequency (kHz)
FIGURE 4.10 Critical region for coloration due to insufficient modal density in some
rooms and plates found by listening tests. (Adapted from Kuhl, W., Eigentone density
and colouration of reverberant sound, Proceedings of the 6th International Congress on
Acoustics, #E-2-8, pp. E-69–E-72, 1968.)
sandwich sheets in loudspeaker driver design are their high flexural stiffness and
(comparatively) low surface mass.
The phase velocity of transverse waves in a sandwich sheet is shown in Figure 4.11.
In audio, the interest is often to radiate sound from the transverse shear movement of
the sheet. Typically, that can be from resonant modes in the region of constant phase
velocity as shown in the figure. To have a wide homogeneous frequency range, the
two crossover frequencies between shear and bending vibrations should be far apart.
The crossover frequencies ω1 and ω2 are
ω1 =
(
2G 2 1 − ν2 ); d h (4.35)
mS′′ Ed
ω2 =
(
6G 2 1 − ν2 ); d h (4.36)
mS′′ B′d
where
d is the thickness of each skin
h is the thickness of the core
E is Young’s modulus of the skins
B′ is the bending stiffness of each skin (free)
G is the shear modulus of the core
10
Normalized phase velocity
Fast waves cB2
3
cS
1
Slow waves
0.3 cB1
0.1
0.02 0.05 0.1 0.2 0.5 1 2 5 10 20
Normalized frequency
FIGURE 4.11 Behavior of the phase velocity of transverse waves in a sandwich sheet
having a honeycomb core. The phase velocity of transverse waves in a sandwich sheet
has a more complicated dispersive quality than those of either shear or bending waves.
Here, cB1 and cB2 are the phase velocities of the bending waves and cS the phase velocity
of the transverse waves. At frequencies below ω1 and above ω2, the waves propagate as
bending waves, in between these frequencies as shear waves. High sound radiation requires
the wavelengths on the sheet to be longer than the waves in the surrounding air, that is,
their phase velocity needs to be higher than the speed of sound in air c. (Adapted from
Kurtze, G., Physik und Technik der Lärmbekämpfung (in German), Verlag G. Braun,
Karlsruhe, Germany, 1964.)
The ratio between the crossover frequencies is
ω2 Bg′
= (4.37)
ω1 2 B′
Here, Bg′ is the combined bending stiffness for the sandwich sheet. If the ratio is
large, the shear wave field will have a modal density that is proportional to frequency,
because of the frequency-independent phase velocity.
4.7 VIBRATION IN LOSSY PLATES

Waves in plates are typically damped by internal damping, radiation, and near-
field flow resistance. Increased damping can be obtained using viscoelastic layers
attached to the surfaces of the plate or in between two plates.
The phase difference between extension ε and tension σ along a plate is determined
by Hooke’s law
σ = ε E = ε E0 (1 + j η) (4.38)

where E is the complex modulus of elasticity.
The damping of the waves is described by the complex part of the wave number,
which is proportional to the reciprocal of the wave velocity. The velocity of shear
56 Electroacoustics
waves is proportional to the square root of the modulus of elasticity, as shown by

Equations 4.5 and 4.6. The phase velocity of bending waves is proportional to the
fourth root of the modulus of elasticity as shown by Equations 4.24 and 4.28.
The damping properties are usually characterized by a loss factor η, which is
typically in the range from 10 −3 to 10 −2 for many construction materials. Provided
that the loss factor η is much smaller than unity, we obtain the complex wave number
for shear waves as
−1 2  η
k = k (1 + jη) ≈ k 1 − j  η 1 (4.39)

 2
and for bending waves as
−1 4  η
k B = kB (1 + jη) ≈ kB  1 − j  η 1 (4.40)

 4
The imaginary part of the wave number describes how quickly the wave decays
along the propagation path. For shear waves, one finds that the waves decrease with
distance x as
uy ∝ e − kηx 2 = e − πηx λ (4.41)

It is often practical to express the extra attenuation (on top of the geometrical
attenuation) as a level difference per unit length, just as in the case of sound in air.
For shear waves, the damping is given by
27.2 η
∆LS ≈
λS
[dB m ] (4.42)

For free bending waves, the damping is given by
13.6 η
∆LB ≈
λB
[dB m ] (4.43)

Using various forms of damping layers or additives one can usually increase the
losses considerably, but extra losses can also be added by air pumping, dry friction,
magnetic hysteresis, etc. In an “undamped” plate, such losses may be the dominant
loss mechanisms.
4.7.1 Damping by Viscoelastic Layers

As mentioned it is often desired to damp waves in loudspeaker diaphragms and
boxes to reduce sound radiation and vibration propagation. A simple way to increase
the damping is to add a viscoelastic layer to the plate, either externally or internally.
A different way is to couple the plate to another structure that has high damping.
A viscoelastic layer exhibits both viscosity and elasticity; chewing gum is a typical
example of a viscoelastic material; other examples are most polymers such as rubber
and plastics. Polymers have long molecular chains and allow many combinations of
elastic and viscous properties. Usually, the viscous properties are very temperature
dependent. The viscosity leads to energy conversion from vibration to heat.
Damping by viscoelastic layers depends on several factors: the temperature, the
type of excitation, the wave types involved, the coupling to other structures, and not
least the frequency of vibration since this determines the wavelengths of the different
waves. The addition of a viscoelastic layer also changes the mass per unit area and
the bending stiffness per unit length, and consequently it is not sufficient just to study
the loss factor.
It is advantageous to use the viscoelastic layers as close to the source of vibration
as possible, because there the application is only needed over a smaller area and also
because the bending waves usually dominate the wave-type spectrum close to the
source.
Finally, it must be emphasized that increasing the loss factor primarily affects
the resonant vibration. The forced vibrations are usually not affected, except by the
addition of mass or stiffness due to the layer.
4.7.2 Viscoelastic Materials
A feature of viscoelastic materials is that the loss factor is much higher and the shear
modulus much lower than that for typical construction materials. Most viscoelastic
materials exhibit large temperature and frequency dependences in Young’s modulus,
shear modulus, and loss factor. An example of the typical behavior is shown in
Figure 4.12.
Properties Properties
Viscoelastic
similar similar
properties
logη, log |G|, log Gve : re, log Gve : im
to rubber to glass
|G|
Gve : re
Gve : im
Increasing frequency
Increasing temperature
FIGURE 4.12 Typical properties of a viscoelastic material as a function of frequency and

temperature.
58 Electroacoustics
The shear modulus of a viscoelastic material has a real part Gve:re and an imaginary
part Gve:im. The parts have different frequency dependencies. The imaginary part
of the shear modulus increases quite rapidly in a specific range as temperature is
reduced or frequency increased. A rule of thumb is that a change in temperature by
5°C–7°C corresponds to a change in frequency by a factor of 10. The damping is
fairly broadband.
Viscoelastic materials are generally not suited as construction materials
because of their softness and temperature-dependent mechanical properties.
They are usually used in thin layers either on the plate, called “free” layers,
possibly together with various distance layers, or in between plates, called
constrained layers.
It is not necessary to add the viscoelastic layer to the entire plate. One can
use partial treatment by adding the viscoelastic material in strips or patches.
The optimum working frequency will then be adjusted by a factor proportional
to the ratio of the untreated and treated surface areas. This also applies to free
viscoelastic layers. One must note that it is possible to create new modal systems
in this way.
4.7.3 Free Layers
By attaching one or more viscoelastic layers to the surface or surfaces of a plate,
one obtains both extension and bending in the free layer. Generally, one obtains
the best use of the material if most of the total elastic energy is stored in the vis-
coelastic layer.
The loss factor for bending waves in a plate with a single-sided free, soft
viscoelastic layer is given by
hvehn2 Re  Eve 
ηB = ηve + η pl (4.44)
B′
where
ηB is the combined total loss factor
ηve is the loss factor of the viscoelastic layer
ηpl is the loss factor of the plate
Re[Eve] is the real part of the complex Young’s modulus of the viscoelastic
layer
hpl is the thickness of the plate
hve the thickness of the viscoelastic layer
hn is the distance between the neutral planes of the two layers. For soft layers,
hn = (hve + hpl)/2
B′ is the bending stiffness of the combined construction
Equation 4.41 shows that the viscoelastic layer should be thick and have as high a
Young’s modulus as possible. Often, Young’s modulus for viscoelastic layers is about
100 times smaller than that of the plate on which it is applied. Free viscoelastic layers
are therefore primarily used to damp thin plates. One also observes that the loss
factor of the combination has approximately the same frequency dependence as that
of the viscoelastic layer.
Application of a free viscoelastic layer is often used as an after-treatment to
resonant plates. While free layers can be used in this way, their use is at a disadvantage
because of the mass of the viscoelastic layer that needs to be thick. This limits the use
of free viscoelastic layers for loudspeaker diaphragm applications since loudspeaker
diaphragms must be lightweight.
Consider a free layer on both sides of a plate. If the layers are thick, it does not
matter if the layer is added by one half to each side. If the layer is thin, it will
contribute more to the combined loss factor if it is applied to one side only.
To increase the losses, one can add a flexible structure, known as a “spacing”
layer. An example is a honeycomb structure (such as the one shown in Figure 4.6
between the plate and the viscoelastic layer. This layer should ideally not store any
energy and should be very rigid to shear forces.
4.7.4 Constrained Layers

A constrained viscoelastic layer is sandwiched in between two plates so that it is in
contact with both plates. Its damping is achieved by its resistance to shear forces
between the two attached plates. The bending wave loss factor for such a sandwich
construction can be estimated using the principles outlined in Ref. [4].
A major advantage of a constrained viscoelastic layer is that the loss factor
curves show wide bandwidths and that the damping can become larger even with
a thin viscoelastic layer. If both plates have the same thickness and are of the same
material, the maximum combined loss factor is about one-third of the loss factor of
the viscoelastic layer.
Constrained layers are ideal for loudspeaker diaphragm applications since the
added mass of the viscoelastic layer can be made very small, while at the same time
the loss factor can be made high over wide temperature and frequency ranges.
4.7.5 Damping by Sand and Other Lossy Materials

Loudspeaker enclosure walls can often be well damped using sand, gravel, and
similar materials. Adding viscoelastic layer damping, using a damping compound
such as damping glue, to a plate does not necessarily lead to less sound radiation by
bending wave vibration in the plate, as is shown in Figure 10.18.
Because of the vibration generation mechanisms however, it may often be
advantageous to add mass to a structure affected by vibration-generating forces.
The added mass will increase the input impedance of the structure considerably
without any acoustic adverse effects, since the sand has such high damping that the
resonances will generally be unimportant. Heavy bracing of the structure, on the
other hand, may lead to increased input impedance but the structure’s resonances
60 Electroacoustics
FIGURE 4.13 Loudspeaker cabinet in which the outside walls are dual-sheet panels. The
airspace between the sheets is to be filled with sand. (Photo courtesy of Claudio Bonavolta.)
will remain and primarily only be shifted upward in frequency. Figure 4.13 shows a
loudspeaker cabinet where the outside walls are made from dual sheets so that the
airspace between the sheets can be filled with sand.
A more economical way to reduce loudspeaker box vibration is to fill the interior
of the box with an “open pore” stiffening structure. Along with stiffening the box
walls, such a structure also reduces the speed of sound inside the box and makes
the box interior appear larger in volume to the sound (see Chapter 13, Section on
acoustic lenses).
REVIEW QUESTIONS
4.1 What is the difference between membranes and plates?
4.2 Why do longitudinal and quasi-longitudinal waves propagate at different
velocities?
4.3 Describe the characteristics of bending wave motion.
4.4 Why is there a difference in mode density between waves in tensioned mem-
branes and bending waves in plates?
4.5 Describe the frequency characteristics of wave motion in sandwich sheets.

4.6 How will damping of bending waves in a plate depend on frequency?
4.7 Why are viscoelastic materials used for damping bending waves?
4.8 What is the difference between the damping action of internal and external
application of viscoelastic materials to plates?
PROBLEMS
4.1 Polycarbonate has a modulus of elasticity E = 2 · 1011 N/m, density
ρ = 7800 kg/m3, and Poisson’s ratio ν = 0.3.
Task:
Calculate the number of bending wave modes in the 1 kHz third-octave bands
for a 0.2 m2 large and 1 · 10 −3 m thick polycarbonate plate and draw the modal
density as a function of frequency.
4.2 A condenser microphone has a membrane that is stretched over a ring. The
ring has a diameter of 2 · 10 −2 m. The membrane is made of plastic having a
thickness h = 10 −5 m and density ρ = 1.4 · 103 kg/m3.
Task:
Calculate the radial tension necessary for the resonance frequency of the first
radial mode to be 10 kHz when the membrane is free. Disregard the influence
of the air load.
4.3 Using a room or a plate to act as a reverberator are two ways of obtaining
“artificial reverberation.” In the case of the plate, a bending wave field is set
up. The resulting sound (room) or vibration (plate) is sensed and amplified and
added to speech or music as “artificial reverberation.” If the respective modal
density is too low, the reverberation is perceived as “colored” and cannot be
used. The shaded area in Figure 4.10 shows in which modal density range the
reverberation is subjectively perceived as colored according to investigations
using music sound reproduction.
Tasks:
Determine which one of the possible alternatives, (a) or (b), can be used
satisfactorily.
a.
A steel plate having thickness =
h 0.5 mm and surface
dimensions = 1.2 · 1.8 m2.
b.
A gold plate having thickness h = 20 μm and surface
dimensions = 0.3 · 0.4 m2.
Material ρ (kg/m3) E (N/m2) ν

Steel 7,700 20 · 1010 0.3
Gold 19,000 8 · 1010 0.3
62 Electroacoustics
4.4 Electrostatic loudspeakers often have flexible rectangular membranes. One

can show that if the membrane is tensioned differently in the “short” and
“long” directions the wave equation can be written similar to Equation 4.11 as
∂2 y 2 ∂2 y
cS2 + cL 2 + ω2 y = 0 (4.45)
∂x 2 ∂z
Task:
Calculate the tension necessary in the “short” direction to have a first resonance
occur at 40 Hz when the membrane is 20 μm thick and is made of polyvinyl
dichloride that has E = 2.8 · 106 N/m and ρ = 1.4 · 103 kg/m3. Assume the wave
motion in the “long” direction to be negligible and the width of the membrane
to be 0.2 m.
4.5 A common way to measure the complex modulus of elasticity of a material is
to measure the transfer function of a thin bar of the material, sometimes called
the “Oberst method.” The bar is clamped at one end while the other end that
is free to move is provided with a thin nickel foil so that it can be excited by a
dynamic magnetic field. The field is generated by a coil placed near the free
end. A second foil and coil are placed at a different location along the bar, as
shown in the figure below.
Nickel foil Nickel foil

sensing point excitation point
Bar
lbar
The first three resonance frequencies f n of a clamped bar where the free length
of the bar is lbar are given by [8]
0.560 B′
f1 =
lbar
2
m′′
(4.46)
f2 = 6.27 f1
f3 = 17.6 f1

The damping factor η of the material can be estimated from the Q-value of
the resonances as described in Appendix F and Equation F.18. Polycarbonate
has a modulus of elasticity E = 2 · 1011 N/m, density ρ = 7.8 · 103 kg/m3 and
Poisson’s ratio ν = 0.3.
Tasks:
a. Calculate the resonance frequencies of a bar clamped at one end. The
bar made of polycarbonate plastic is 1 · 10−3 m thick and has a free length
of 1 · 10−1 m.
b. Calculate the loss factor if the width of the first resonance is 10% of the
center frequency.
c. What will be the reverberation time of the oscillation of the bar at this
frequency?
REFERENCES
1. Hawley, M. S. et al., The Western Electric 640AA capacitance microphone. Its history
and theory of operation, in Wong, G. S. K. and Embleton, T. F. W. (Eds.) AIP Handbook
of Condenser Microphones: Theory, Calibration and Measurements (Modern Acoustics
and Signal Processing), American Institute of Physics, New York (1994) ISBN-13: 978-
1563962844. pp. 8–34.
2. Morse, P. M., Vibration and Sound, 2nd edn., American Institute of Physics, New York
(1991) ISBN-13: 978-0883188767.
3. Vigran, T.E., Building Acoustics, Taylor & Francis Group, London, U.K. (2008)
ISBN-13: 978-0415428538.
4. Ver, I. L. and Beranek, L. L., Noise and Vibration Control Engineering: Principles and
Applications, Wiley, New York (2005) ISBN-13: 978-0471449423.
5. Taguchi, S. et al., Sandwich-construction loudspeaker diaphragm with foamed
high-polymer and carbon fiber, J. Audio Eng. Soc., 34(11), 895–903 (1986).
6. Kuhl, W., Eigentone density and colouration of reverberant sound, Proceedings of the
6th International Congress on Acoustics, Tokyo, #E-2-8 (1968), pp. E-69–E-72.
7. Kurtze, G., Physik und Technik der Lärmbekämpfung (in German), Verlag G. Braun,
Karlsruhe, Germany (1964) ASIN: B0000BKML2.
8. Kinsler, L. E. et al., Fundamentals of Acoustics, 2nd edn., John Wiley & Sons. New York,
(1962) ASIN: B000LC9DO64.
5 Circuits and Circuit
Components
5.1 INTRODUCTION
When we measure, record, or reproduce sound we wish to convert an electric signal’s
time history linearly into an equivalent sound or vibration with the same time
history or vice versa. For this we use electroacoustic transducers. An intermediate
step in most transducers is usually vibration; so electroacoustic devices may be
used to generate and sense vibration as well, although it is then better to call them
electromechanic transducers. Almost all electroacoustic transducers rely on an
implicit electromechanical transduction. Sound is typically generated or sensed by a
diaphragm connected to the sound or vibration transduction mechanism.
Examples of electroacoustic transducers are sensors for sensing sound and
vibration such as microphones (for audio, i.e., sound in the audible range), sonar
sensors (for ultrasound in air and water), and accelerometers (for vibration in solid
and flexible materials). Transducers that convert electric signals into equivalent
acoustic and vibratory signals are, for example, the almost ubiquitous electrodynamic
loudspeaker and earphone, force transducers (“shakers”), sonar transmitters, etc.
Most electroacoustic transducers based on the use of magnetic or electric fields
are reciprocal, that is, they may be used for conversion in both directions. There
are however also a class of devices used primarily in acoustic and vibration testing,
which are nonreciprocal. Examples of such transducers are some operating on optical
principles such as optical microphones, those relying on ultrasonic sensing such as
the ultrasound Doppler-based microphone, and those based on nonlinear mixing
(heterodyning) of ultrasound signals into audible sound such as the parametric
loudspeaker. Acoustic and vibration testing sometimes also uses thoroughly
nonreciprocal methods such as spark generators to create short pulses of sound and
hammer-like devices to generate short force pulses in vibratory systems.
5.2 LINEARITY
Typically, electroacoustic transducers are expected to be designed for minimum
nonlinear distortion in their conversion of electrical signals into sound or vibration,
and vice versa. As discussed in Chapter 2 and in Appendix A, distortion can be
both linear and nonlinear. Linear distortion is, for example, changes in impulse or
frequency response; nonlinear distortion is characterized by the addition of new
frequency components in the transducer output along with those input to the system.
65
66 Electroacoustics
Reciprocal transducers generally have some form of linear distortion. Usually the
sensing and generating frequency response characteristics differ by a general ±6 dB
per frequency doubling (octave) characteristic.
5.3 CIRCUIT ANALYSIS PRINCIPLES

The techniques used in analyzing and designing electric systems may be used to great
advantage in understanding the general operation of acoustic and vibratory systems.
While it is possible to analyze and design electroacoustic transducers by looking
at the electric, mechanical, and acoustic domains separately, transducers are
usually best analyzed using an approach where all relevant factors are converted
appropriately so that the system can be analyzed with all factors converted into only
a single domain, that is, for example, an equivalent electric network. Alternatively,
one can analyze the transducer in the acoustic or mechanical domains. The best
approach will vary from case to case, as will be seen later. The general idea is “to
make the behavior of mechanical systems as easy to visualize as the behavior of
electric devices without any knowledge of the underlying differential equations” [1].
Conventional computer software for the analysis of electric circuits can be
used to find the impulse response, transfer function, and frequency response of
the electroacoustic system, although the properties of the acoustic and mechanical
domains may be less linear than those of the electric domain. Some approximation is
usually necessary when representing acoustic, mechanical, and electrical components
outside their “original” domain. The fundamental principles of electrical circuits are
discussed in Appendix A.
Specialized software for electroacoustic analysis, such as AkAbak, is also available
which removes some of the electric “incongruities” caused by the electric properties
of primarily acoustic radiation components in the circuits [4]. Finite element method
(FEM) software may integrate mechanical, acoustic, electric, magnetic, and more
systems as in COMSOL’s Multiphysics software [5]. Such software can be used to
analyze not only the electroacoustic but also thermal and nonlinear properties of
the vibratory system and will allow very accurate modeling. The drawback of FEM
software is the time necessary for modeling and the lack of intuitive insight.
The approach in this book of representing the acoustic, mechanical, and electric
domains by a common circuit diagram uses the principle of analogies. The idea
of electromechanical and electroacoustical analogies is based on the similarity
of the differential equations used to describe mechanical, acoustic, and electric
phenomena. Strictly, the principle of analogies can be used in the analysis of each
of these domains separately, since the analogy generally involves representation of
mechanical and acoustical components by equivalent electrical symbols that connect
by the principles of electric networks.
Often, the analysis of an electroacoustic system starts by the conversion of the
acoustic and mechanic components (such as masses, springs, dampers, pistons,
volumes, etc.) into a circuit diagram that is fundamentally similar to an electric
circuit diagram. This acoustical or mechanical circuit diagram is then connected to
the electrical circuit by components such as transformers and gyrators using various
Circuits and Circuit Components 67
TABLE 5.1
Typical Symbols for Elements in Electrical Circuits and the
Rules of Manipulation
Constant-drop The quantity a is independent

generator of what is connected to the
a generator. The arrow points to
the positive terminal of the
generator.
Constant-flow The quantity b is independent
generator of what is connected to the
b generator. The arrow points to
the direction of positive flow.
a
Impedance-type
a = cb
b element
c
a
Resistance-type
a = cb
b element
c
a
Capacitance-type 1
a= b
b element jωc
c
a
Inductance-type
a = jωcb
b element
c
b c: 1 d a= c g
Transformation-type cb = d
a g
element a 2 g
=c
b d
b c :1 d a = cd
Gyration-type cb = g
a g element
a 2d
=c
b g
Ground This symbol is the one

(electrical) commonly used in electrical
engineering for circuit ground.
Ground This symbol is the one
(mechanical and commonly used in electrical
acoustical) engineering for chassis ground.
68 Electroacoustics
principles determined by the mode of electromechanical energy conversion applicable

to the particular transducer studied. Finally, the entire system is represented in
the chosen domain by a circuit which uses electrical (or similar) components and
appropriate symbols, and which can be analyzed by the electric network theory.
Gehlshøj [2] and Beranek [1] give the following four criteria for a successful
method to set up electrically based circuit representations of mechanical and acoustic
systems:
1. The method must permit the formation of schematic diagrams from visual
inspection of the devices.
2. They must be capable of such manipulation that will make possible the
combination of electrical, mechanical, and acoustical elements into one
schematic diagram.
3. They must preserve the identity of each element in combined circuits so that
one can recognize immediately a force, voltage, mass, inductance, and so on.
4. They must use the familiar symbols and rules of manipulation for electrical
circuits.
As is discussed in Appendix A, there are four generic quantities associated with

the circuit elements: the potential difference (sometimes called “the drop across
the element”), the flow through the element, the frequency response behavior of the
element, and the magnitude of the element. The seven conventional elements are
the ones shown in Table 5.1. It is important to realize that irrespective of whether
the circuit is electric, mechanical, or acoustic, the mathematical operations
associated with a particular symbol are invariant.
We will see that to make this thinking easy and intuitive we will choose symbols
for the mechanical and acoustical elements which are reminiscent of those for
the electrical circuit elements. While the discussion in this book only extends to
mechanical systems having linear movement, it is possible to successfully extend it
to mechanical systems also having rotational movement as shown in Ref. [3].
REFERENCES
978–0883184943.
2. Gehlshøj, B., Electromechanical and electroacoustical analogies, PhD thesis, Academy
of Technical Sciences, Copenhagen, Denmark (1947).
4. AkaBak software. Joerg Panzer R&D Team Software Development. (http://www.
randteam.de) Sampled October 2012.
5. COMSOL, Inc., 744 Cowper Street, Palo Alto, CA 94301, USA. (http://www.
comsol.com) Sampled October 2012.
6 Electromechanical
Analogies
6.1 CONTINUOUS AND DISCRETIZED BODIES

All real-world systems and transducers involve components where waves can
be excited, whether acoustic, mechanical, or electric. To simplify the analysis by
electrical network theory, it is generally desired to represent the components by
discrete equivalents so that they can be analyzed as if they were discrete objects
with lumped parameters. In practical terms, this means that wave motion (whether
acoustic, mechanical, or electromagnetic) is assumed to be at infinite speed. One
can also say that the discrete element has no internal motion due to an applied force.
When the delays caused by wave motion become important for the function of the
circuit, the components may be discretized appropriately, for example, by a series
network, etc. [1,2].
Consider as an example a thin solid aluminum rod. It is well-known that such
a rod can carry many types of waves, such as quasi-longitudinal, torsional, and
bending waves to name but a few.
For the rod to be modeled as a discrete mass lump, the difference in mechani-
cal impedance seen by the excitation force varies by no more than 5% from the
true modulus of the impedance at that frequency. (It is important to realize that
the phase difference may be relatively much larger.) In practice, this applies when
the length of the bar is no more than about one-sixteenth of the wavelength at the
frequency of interest.
As was discussed in Chapter 4, rods carry quasi-longitudinal waves that, for
example, in aluminum have the speed of about 5.3 · 103 m/s. The first half-wave
resonance frequency of the rod will be at about 25 kHz; so the rod can only be
considered a discrete mass lump up to about 1.5 kHz.
Bending waves have a frequency-dependent phase velocity that is much lower than
that of longitudinal waves; so, the wavelengths are shorter. This means that the size limit
for bodies to be counted as discrete masses is much smaller than that given by longitu-
dinal waves. At a frequency of 1 kHz, bending waves in a 1 mm thin sheet of aluminum
have a phase velocity of only 110 m/s which means that the wavelength is only 0.11 m.
6.2 MECHANICAL ELEMENTS: ELECTRICAL ANALOGIES

6.2.1 Representation of Mechanical Components
It is natural to look for ways of representing mechanical and acoustic systems by the
same principles that we use for electrical components since we then can use the intuitive
reasoning afforded by electrical circuits. To do so, we need to find analogies that can be
69
70 Electroacoustics
represented as resistors, capacitors, inductors, transformers, etc. In addition, we need

to devise graphical symbols (icons) for such analogies and a way of easily converting a
real-world mechanical construction into a workable analogy and vice versa.
While it seems straightforward to use force in analogy with voltage and velocity
in analogy with current, this is not always very practical since it is easy to measure
voltage and velocity but not current and force, since the latter requires the system to
be disjoined to install proper measurement transducers.
It is, as we shall see later, often more intuitive, and simpler, to compare voltage
to velocity and force to current. Such an analogy is called a mobility analogy
(or analogy of the second kind) while the previous approach is called an impedance
analogy (or analogy of the first kind).
6.3 MECHANICAL IMPEDANCE AND MOBILITY

The choice of representing mechanical elements by impedance or by admittance is
determined by the easiest conversion of the mechanical system into an equivalent
analog circuit that can be analyzed by means of electrical circuit theory. First, one
should note that the idea of impedance in mechanical systems is based on the relation
F/u, denoted by Z M, whereas admittance, usually called mobility, is based on the
ratio u/F and usually denoted by YM. (In some texts, M is used for mobility.) The
mechanical mobility is the inverse of mechanical impedance.
1
ZM =
YM
[ Ns/m ] (6.1)

It will turn out that it is easier to convert the mechanical system into an equivalent
analog electrical circuit by first converting the system into a mechanical circuit using
the mobility analogy since this approach more closely corresponds to the way we
study electrical circuits, namely that the voltage is easy to measure in an electric
circuit and that velocity in an analogous way is the easiest quantity to measure in a
mechanical system. Force will thus correspond to electrical current.
The resulting so-called mobility analog may then need to be converted into an
impedance analog using the duality discussed earlier. This conversion will usually
be necessary to couple the mechanical system to the electrical and acoustic systems
of a transducer, so that a simple system using only one of the physical domains
results. This will then allow calculation of, for example, radiated acoustic power as a
function of the voltage applied to a transducer.
The symbol for impedance and mobility is the same and is shown in Figure 6.1.
Z Y
(a) (b)
FIGURE 6.1 The symbols for impedance (Z) and mobility (Y) components are the same.
Electromechanical Analogies 71
6.4 MECHANICAL RESISTANCE

In analogy to electrical resistance, we may define mechanical resistance R M by
F
RM =
u
[ Ns/m ] (6.2)

A lossy mechanical element is generally called a damper or dashpot element.
A mechanical resistance will dissipate some of the mechanical energy in the circuit
as heat. The power lost is

P = RM u2 [ W ] (6.3)
An icon for mechanical resistance should preferably have some visual likeness to
real-world resistive components. One such component featuring this type of viscous
friction is the hydraulic or pneumatic damper. A suitable symbol is the one shown
in Figure 6.2a.
The inverse of mechanical resistance is called admittance, which is usually
denoted by rM in electroacoustics.
1
rM =
RM
[m/Ns] (6.4)

Mechanical admittance is used in mechanical mobility analogy circuits in the same
way as mechanical resistance is used in impedance analogy circuits. Note also that
the same icon is used for both mobility and impedance analogies (Figure 6.2b and c).
The operating characteristics of various forms of damping are shown in Figure 6.3.
We need to consider three types of damping, all of which exist in parallel in most
mechanical systems: viscous, hysteresis, and Coulomb damping.
Viscous damping, sometimes called linear damping, is the result of the damping
mechanism introduced by the viscosity of oil films and air films. The force needed to
move a body subject to viscous damping is directly proportional to its velocity, and
in this way viscous damping is analogous to the damping introduced by resistance
in electrical circuits.
Hysteresis damping is associated with the internal damping mechanisms in
bodies and is typical of elastic materials, such as plastics and rubber. In practice, it is
usually treated as linear damping but will be frequency-dependent.
RM RM rM
(a) (b) (c)
FIGURE 6.2 Symbols for mechanical elements having viscous damping: (a) The
“mechanical” symbol for a damper; (b) The symbol for mechanical resistance in impedance
analogy circuits; (c) The symbol for mechanical resistance in mobility analogy circuits.
72 Electroacoustics
Hysteresis
Viscous
Coulomb
FIGURE 6.3 Operating characteristics for various types of damping. (After Rasmussen, K.,
Analogier mellem mekaniske, akustiske og elektriske systemer, Polyteknisk forlag, København,
1981.)
Coulomb damping is the result of dry friction, and thus depends on the force of a
body to the normal of a surface. A typical example is the friction between bow and
string in playing musical instruments such as the violin and cello. Dry friction is
usually not continuous, although this is often assumed in analysis.
6.5 MECHANICAL COMPLIANCE

A mechanical spring is characterized by resistance to an extensional or compressional
force F due to the presence of stiffness in the mechanical element’s construction
material, which is formulated as
F = kM x (6.5)

Here, k M is the spring constant which quantifies the stiffness of the spring and
x is the resulting response. The inverse of the stiffness is called compliance and
denoted by CM.
1
CM =
kM
[m/N ] (6.6)

When working with impedances, we are interested in the ratio F/u rather than F/x.
Since the velocity u is the time-derivative of x we have
u = jωx (6.7)

So, we obtain the mechanical impedance and the mobility of a spring as
F 1
ZM = = (6.8)
u j ωC M
u
YM = = jωC M (6.9)
F
CM CM CM
(a) (b) (c)
FIGURE 6.4 Symbols for mechanical compliance elements, “springs”: (a) The “mechanical”
symbol for a spring; (b) for mechanical compliance in mobility analogy circuits; (c) for
mechanical compliance in impedance analogy circuits.
The icon for a mechanical spring is shown in Figure 6.4a. The icons shown in
Figure 6.4b and c are used to designate a spring for the mechanical mobility and
impedance analogies, respectively.
When modeling springs by analogies, it is important to realize that a coiled spring
is difficult to be represented as one lumped element. There are several reasons for
this, such as the difficulty of having the spring move in only one direction and the
long length of the metal wire wound into the coil.
Any real spring will have a mass associated with it. If we assume that the mass
and compliance distributions are distributed equally over the spring, one can show
that the effective mass of the spring is one third of the spring’s mass [2].
A reasonable approximation, from an impedance viewpoint, can be achieved by
subdividing the spring into a set of masses and springs which will better represent
the mechanical action of the distributed mass and spring action of the spring coil.
Bar-type springs will show similar behavior. Subdivision, such that each element
corresponds to a length smaller than one-sixteenth wavelength, is often used
(compared to acoustic elements in Chapter 7).
6.6 MASS
For a piece of material to be considered a discrete mass lump, the wave motion inside it
has to be so small as to be inconsequential. In transducers we usually want our masses
to move only in one direction of translation, that is, rectilinear movement, and have no
rotational movement. This is usually achieved quite successfully for electrodynamic
loudspeaker diaphragms because of the dual suspension mechanisms of the diaphragms.
Some loudspeakers exist in which sound radiation is due to bending wave motion; in such
transducers, there will also be rotary movement. Most electrodynamic microphones
and some electrodynamic loudspeakers have only a single suspension ring, and then
there may be “rocking” motion which is not possible to model using the analogies.
Mass is characterized by inertia; so a force is necessary to accelerate mass. We will
only consider translational movement, since this is the characteristic of moving parts
in most electroacoustic transducers. Since velocity is a result of acceleration, force has
to be applied to cause velocity. Because acceleration is the time-derivative of velocity,
Newton’s law of motion becomes
F = M M a = jωM M u (6.10)

74 Electroacoustics
MM MM MM
(a) (b) (c)
FIGURE 6.5 Symbols for mechanical mass elements: (a) The “mechanical” symbol for
a mass; (b) for mechanical mass in mobility analogy circuits; (c) for mechanical mass in
impedance analogy circuits.
We see that velocity and force are out of phase. The impedance and mobility of
the mass MM become
F
ZM = = jωM M (6.11)
u
u 1
YM = = (6.12)
F j ωM M
The icon for mass has been chosen so that it shows one fundamental property of
mass, namely that it always moves with reference to an inertial reference frame. The
common symbols are the ones shown in Figure 6.5. One attractive property of the
symbol in Figure 6.5b is that it is graphically similar to the symbol used for electrical
capacitors, which is advantageous in converting mechanical mobility circuits into
electrical circuits.
6.7 LEVERS
Levers are used to change the ratio between force and velocity in a mechanical
circuit. In this way their use has a close relationship to that of transformers in
electrical circuits. Note that while a lever can handle static displacements, electrical
transformers can only operate on alternating signals.
There are three cases of lever usage that are of interest, one- and two-armed
levers, and floating levers. Figure 6.6a and b show the first two cases.
The requirement for balance leads to the simple equation
F1l1 = F2l2 (6.13)

For small movements, we obtain from the geometry that
u1 u2
= (6.14)
l1 l2
F1
l1
F1
F2
l1 l2
l2
u1 u2 u1
u2
F2
(a) (b)
F3 F1
F2
l2 l1
u1
u2
u3
YM2 YM3
Mobilities constrained
to only move up and down
(c)
FIGURE 6.6 Examples of levers: (a) one-armed lever; (b) two-armed lever; (c) floating lever.
We can then form the impedance and mobility relationships as

2
F1  l2 
Z M1 = = Z M 2 (6.15)
u1  l1 

2
u l 
YM 1 = 1 =  1  YM 2 (6.16)
F1  l2 

We see from the previous equations that they are the same as those for an ideal
transformer with a transformation ratio of l1/l2 for the impedance type and l1/l2 for
the mobility type.
The floating lever, shown in Figure 6.6c, can, for small movements, be analyzed
using the method of superposition. Using the same principles of moment balance and
geometrical relationships and then applying superposition, we obtain the following
mobility relationship [3]:
2 2
u1  l1 + l2  l 
Y M1 = = Y M 3 +  1  Y M 2 (6.17)
F1  l2   l2 

The effective mass of a lever can be calculated using the same principle used for the
analysis of the effective mass of a spring. For a one-armed lever, one obtains masses
MM1 and MM2 associated with points 1 and 2 as
M M l1
M M1 = (6.18)
3 l1 + l2
76 Electroacoustics
(l1 + l2) : l2
F1 l1 : l2 F2 F1 F3
YM1 u1 u2 YM2 u3 YM3
(a)
YM1 u1
u1 l2 : l1 u2 l1 : l2 F2
ZM1 F1 F2 ZM2 u2 YM2
(b)
(c)
FIGURE 6.7 Lever circuit symbols: (a) the mobility analog for a simple lever; (b) the
impedance analog for a simple lever; (c) the mobility analog for a floating lever.
M M l2
MM 2 = (6.19)
3 l1 + l2
Here, MM is the total mass of the lever [3]. The impedance and mobility analogs for
the three lever cases are shown in Figure 6.7.
6.8 GENERATORS
There are two mechanical generators that are of interest, the constant force and the
constant velocity generators.
A simple constant force generator can be modeled as an electrodynamic
transducer mechanism. The “ideal” electrodynamic transducer nominally has zero
internal impedance.
The symbol chosen by Beranek [3] for constant force generators in the mechanical
domain is shown in Figure 6.8a. For analogies, the choice of symbol will depend on
the use of either the impedance or mobility analogy as shown in Figure 6.8b and c.
u1
F
F F
u2
(a) (b) (c)
FIGURE 6.8 Force generator symbols: (a) the “mechanical” symbol for a force generator;
(b) the symbol for a force generator in mobility analogy circuits; (c) the symbol for a force
generator in impedance analogy circuits.
u1
u u u
u2
(a) (b) (c)
FIGURE 6.9 Velocity generator symbols: (a) the “mechanical” symbol for a velocity
generator; (b) the symbol for a velocity generator in mobility analogy circuits; (c) the symbol
for a velocity generator in impedance analogy circuits.
A simple constant velocity generator can be modeled as a strong motor with a

shuttle mechanism, and its rotation is assumed not to be affected by the load.
The symbol chosen by Beranek [3] for constant velocity generators in the mechanical
domain is shown in Figure 6.9a. For analogies, the choice of symbol will depend on
the use of either the impedance or mobility analogy as shown in Figure 6.9b and c.
6.9 POWER RELATIONSHIPS

The power dissipated in a mechanical circuit depends on the “current” through
the resistive elements. Since we are interested in both impedance and mobility
analogies, we write the power loss in impedance analogy as
PM = RM u2 (6.20)
and for mobility analogy as
PM = rM F 2 (6.21)

Here, R M and rM are the resistance and admittance, respectively. Because we usually
tend to go first from mechanical reality to a mechanical mobility analogy, one should
calculate the mechanical power loss from the situation in that analogy and stay away
from calculation of the losses after various forms of conversion, particularly after
involving transducer mechanisms, etc.
REVIEW QUESTIONS
6.1 How are the mechanical equivalent circuit elements defined?
6.2 How are the mechanical equivalent circuit elements drawn and indicated in
the mechanical mobility and impedance analogies?
6.3 How would you go about checking if the mechanical mobility equivalent
circuit you have drawn is correct?
6.4 How does one practically accomplish mechanical components such as
compliances, resistances, and masses? Which are the limitations when one
wants to implement them in practice?
6.5 Explain the working principle of the two-lever systems.
78 Electroacoustics
PROBLEMS
6.1 The figure below shows a simple linear mass spring system.
Tasks:
a. Write the differential equation for force-balance in the system shown in

Figure P6.1.
b. Write the corresponding expression using the jω-method. Assume
one-dimensional, linear motion, and stationary conditions. The fric-
tion between the mass MM and the surface on which it glides can be
represented by a resistance R M.
c. Determine the mechanical mobility YM.
uM
CM
FM
MM
6.2 The mobility circuits in the figure below depict many mechanical systems.
Task:
Draw the mechanical systems (if possible) for the circuits shown.
F F
F F
6.3 The accelerometer is used to measure the vibration in mechanical systems. Its
operation is based on the use of a seismic mass MS as shown in the figure below.
The basic construction of an accelerometer is shown in the figure. The electrical
output voltage is proportional to the force on the quartz spring CQ connecting
the seismic mass and the mass of the accelerometer body MB. The accelerometer
body is usually attached to the vibrating surface using a thin layer of wax.
MQ
CS
MS
uF
Tasks:
a. Draw a mechanical mobility analogy for the system assuming the
vibrating surface on which the accelerometer is attached to be rigid.
b. Indicate where the force that generates the output voltage is located in
the figure.
c. Draw a mechanical mobility analogy for the system including the mobil-
ity of the vibrating surface on which the accelerometer is attached as
well as the compliance of the mounting wax (not shown in Figure P6.3).
d. What will be the difference in operation between the cases in (a) and (c).
80 Electroacoustics
6.4 The figure below shows a mass that is set in motion by two forces connecting
over spring mass systems. The mass is only free to do translatory motion in
one dimension.
uM
MM2
CM2 RM2 CM4 RM4
MM1 MM3
CM1 CM3
F1 F2
Task:
Draw the mechanical mobility circuit for the system assuming that the
components are free only to do translatory motion in one dimension.
6.5 The figure below shows a mass spring system that is acted upon by a force F.
The motion of the mass MM3 is damped by a second “dynamic” mass spring
system attached to MM3 but not to the rigid foundation. The components are
only free to do translatory motion in one dimension.
uM
MM3
F
CMD RMD
CM1 CM2
MMD
Tasks:
a. Draw the mechanical mobility circuit for the system assuming all
movements in one dimension only.
b. How should the “dynamic” system be tuned to damp the motion at the
resonance frequency of the main mass MM3?
REFERENCES
1. Gehlshøj, B., Electromechanical and Electroacoustical Analogies. PhD thesis. Academy
of Technical Sciences, København (1947).
3. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986)
ISBN-13: 978–0883184943.
4. Rasmussen, K., Analogier mellem mekaniske, akustiske og elektriske systemer,
Polyteknisk forlag, København (1981) (in Danish).
7 Electroacoustical
Analogies
7.1 ACOUSTICAL CIRCUIT ELEMENTS

When we model acoustical circuits, representing the physical world of pipes, orifices,
diaphragms, etc., we use acoustical components to simplify the understanding of
the operation of the circuits. The acoustical circuit elements are two-ports or four-
ports that can be characterized by the acoustical field variables, volume flow, and
sound pressure, at their ports. As with mechanical circuits there are a number of
common circuit elements. These can be represented in the physical world by their
corresponding acoustical components. For example, compliance in the acoustical
circuit will be represented in the physical world by a volume of a compressible fluid,
a gas, in a closed cavity.
The acoustical circuit elements are the acoustic compliance, mass, resistance,
transformer, and generator. Since acoustical components will use a gas or gas
mixture, such as air, their properties will be dependent on temperature and static
pressure. In this work, we assume that there is no static flow. Linearity may be a
problem when the dynamic, acoustical pressure becomes a sizable part of the static
pressure (about 105 Pa), for example, at sound pressures over 102 Pa in air.
7.2 WAVES IN TUBES

A prerequisite for the conventional use of electroacoustical circuit analogies is
the assumption of plane wave propagation in the components. However, the wave
propagation is seldom as plane waves; it would be better to say that the waves are
confined to tubes. The propagation at cross-sectional changes, at orifices, and bends
cannot be plane waves. For these cases, the analysis should ideally include higher
order modes of propagation or use a lumped components approach. The earlier form
of analytical analysis however quickly becomes mathematically unwieldy; so it is
necessary to resort to numerical approaches such as finite element analysis. If the
lumped components approach is used, one can add discrete components to the
circuits to represent, for example, the influence of tube bends. Given that the analysis
generally strives at understanding the major modes of operation of the acoustical
circuits, it is customary to assume that the propagation is as plane waves as long as
the dimensions of the acoustical components are small compared to wavelength.
A more detailed analysis can be found in Chapter 3 and in Appendix F.
For wave propagation inside a tube, the upper limit for the assumption of plane-
wave-only propagation will depend on the impedance ratio between the air and the
impedance of the tube wall. The cases of sound propagation in air in a tube having
83
84 Electroacoustics
rigid walls (such as a thick-walled metal tube) and that of sound propagation in
water in a tube having limp walls (such as a thin plastic tube) are different regarding
boundary conditions.
For plane wave sound propagation in air in a straight duct that has rigid walls and
a rectangular cross section (with dimensions of dx and dy, assume dx > dy), it has been
shown in Appendix F that there is an upper frequency limit f lim for “pure” plane wave
propagation:
c
flim < (7.1)
2d x
For a straight circular tube having rigid walls and an internal diameter d, the con-
ditions for plane wave propagation yield the value for the upper frequency limit f lim as
c
flim < (7.2)
1.71d
We see that the limit for plane wave propagation frequency condition is about the
same as that for the rectangular duct
λ (7.3)
d≈
2
The limiting frequency is sometimes called the cutoff frequency. Generally, the tube
will not be driven by a plane piston but rather by an uneven diaphragm such as a cone
or dome; it may be smaller than the tube cross section, be offset, flexible, noncircu-
lar, etc. Close to such an uneven surface there will be higher order modes excited in
the pipe (the plane wave mode is usually called the 0th order mode). The higher order
modes are evanescent (magnitude decaying exponentially with distance) as long as
the frequency is below the limiting frequency. They will cause uneven sound pres-
sure distribution across the cross section of the tube when there are aberrations from
ideal piston drive or load. Higher order modes can be excited at tube bends as well.
We also find that at the interconnection between tubes of different diameters,
there cannot be plane waves since the flow cannot instantaneously be redirected.
Such interconnections occur frequently in mufflers and other sound attenuators, in
microphones and loudspeakers, as well as in acoustic resonators.
In very thin tubes, the friction at the tube walls will result in damping of the wave.
The phase of the particle velocity does not change over a cross section in thin tubes
but the velocity profile across the tube will have a parabolic shape [1].
7.3 ACOUSTIC IMPEDANCE

The acoustic impedance Z A is formed as the ratio between sound pressure p and
volume velocity U over some surface in the system:
p
ZA =  Ns/m 5  (7.4)
U  

Electroacoustical Analogies 85
The volume velocity is formed as the product of particle velocity and the surface
area in question, if
• The particle velocity vector is parallel to the normal to the surface, and
• The particle velocity is the same at all points over the area
The acoustic admittance YA is formed in analogy as the ratio between volume veloc-
ity and sound pressure.
U
YA = m 5 /Ns (7.5)
p  

For us to be able to use electric filter theory in designing and understanding

acoustical circuits, we need the circuit elements of electrical engineering: capacitance,
inductance, resistance, transformer, as well as current and voltage generators. In the
following sections we will see how this can be achieved.
7.4 ACOUSTIC CAPACITANCE

7.4.1 Impedance Analogy
Acoustic compliance is associated with the compressibility of gases. A gas which
is enclosed in a cavity will be compressible, that is, an instantaneous reduction of
the cavity volume will lead to an increase in pressure. Since the pressure increase
is accompanied by a temperature increase, slow changes will be somewhat modi-
fied by the heat exchange with the cavity walls and by the relaxation processes in
the gas.
The acoustic compliance as an impedance element should correspond to
a capacitor in an electric circuit. A small electrical capacitor can be viewed as a
“lumped” capacitance element even at rather high frequencies because of the high
propagation velocity of electromagnetic waves. This is not the case for the acoustical
equivalent, because the velocity of sound in gases is comparatively low.
We want to have an acoustical circuit element that has acoustic impedance that
can be written analogously to that of an electrical capacitor, that is,
1
Z A (ω ) = (7.6)
jω C A
Here, CA is the acoustic capacitance or compliance. In this section we will

show that this is possible under some assumptions. A prerequisite for viewing the
impedance of the closed cavity as a single lumped compliance element is the absence
of modal behavior in the sound field inside the volume. In practice this means that
the maximum “unwrapped” dimension of the volume must be much smaller than the
wavelength of sound at the frequency of interest. The walls of the cavity are assumed
to be ideally rigid.
86 Electroacoustics
ZA,in
x = –lx x=0
FIGURE 7.1 A rigid lossless closed short tube used as an acoustic impedance component.
An extremely simplified approximation of any closed volume is a short straight

tube, having both length and cross section fulfilling the requirement mentioned in
the previous paragraph. Such a tube is shown in Figure 7.1. The tube is assumed
to have its end surface perpendicular to the tube axis. The wave injected into the
volume is assumed to come from a plane oscillating piston.
The piston will generate a positive-going wave in the x-direction in the tube. The
pressure of this wave can be written as
p( x, k ) = pˆ + e − jkx (7.7)

and its particle velocity as
pˆ + − jkx
u ( x, k ) = e (7.8)
ρc
The wall at the end of the tube will have a reflection coefficient r(ω) = 1 since the
impedance of the wall Zw is much higher than that of the gas.
Z w (ω) − ρc
r (ω) = (7.9)
Z w (ω) + ρc
The sound pressure and particle velocity in the tube, consisting of both the
forward- and backward-going waves, can then be written as

(
p( x, k ) = pˆ + e − jkx + r (ω)e jkx (7.10) )
u ( x, k ) = pˆ + e − jkx − r (ω)e jkx (7.11)
( )
ρc
The acoustic impedance into the tube at some distance from the tube end can then
be written as
Z A ( x, k ) =
(
ρc e − jkx + r (ω)e jkx ) (7.12)
S (e − jkx
− r (ω)e jkx
)

Here, S is the surface area of one side of the piston (and the cross-sectional area
of the tube). Assuming the length of the tube lx and r(ω) = 1 we can write the acoustic
impedance seen by the piston at x = −lx as
ρc
Z A ( −l x , k ) = cot ( klx ) (7.13)
jS
Since
ω 2πf 2π
k= = = (7.14)
c c λ
the impedance can also be written as
ρc  l 
Z A ( −l x , k ) = − j cot  2π x  = jX A (7.15)
S  λ
The reactance X A is shown graphically in Figure 7.2.

As long as the length of the tube is small compared to the wavelength of sound λ,
at frequency ω, the series expansion of the cotangent can be approximated by one or
two terms without large error:
3
1 klx ( klx ) ...
cot(klx ) =
klx
−
3
−
45
− (
klx < π (7.16) )

If we retain only the first term of the series expansion, we have the impedance
formulation
1
Z A (ω ) ≈ (7.17)
j ωC A
Reactance
2×107
1×107
kx
–6 –5 –4 –3 –2 –1
–1×107
–2×107
FIGURE 7.2 The behavior of the reactance X A(kx) of a lossless closed tube as function of kx.
88 Electroacoustics
CA
CA
(a) (b)
FIGURE 7.3 The symbol for an acoustic compliance: (a) impedance-type analogy; (b)
admittance-type analogy.
as desired. The acoustic compliance is then
Slx
CA = m 5 /N  (7.18)
ρc 2  

As long as the enclosed gas volume V has its largest dimensions much smaller
than, for example, 1/16 th of a wavelength, and there are no constrictions so that the
volume can be considered as one volume, its shape does not matter and the acoustic
compliance may be written simply as
V
CA = (7.19)
ρc 2
We draw the circuit element in the electroacoustical impedance analogy using the
same symbol as in electrical circuits, that is, as a capacitor as shown in Figure 7.3a.
The acoustic capacitor, representing the compliance of air enclosed in a cavity
using an acoustic impedance analogy, must always have one of its terminals at
circuit ground. The only way to have a capacitor without a ground connection in an
acoustical circuit is by implementing it as a mechanical compliance, for example, by
using a stiffly suspended diaphragm.
It is reasonable to expect that if we retain the first two terms of the series
expansion, we will have a better approximation than before. A comparison of the
circuit approximation given by this two-term approximation to the exact formulation
shows us that, as long as the length of the tube is less than 1/8 λ, the error in the
approximation of the impedance magnitude will be less than 5%.
Retaining the first two terms, the expression for Z A can be reformulated as
1 M
Z A (ω) ≈ + jω A (7.20)
j ωC A 3
Here, MA is the acoustic mass
ρlx
MA = (7.21)
S
MA/3
CA
FIGURE 7.4 A more accurate representation of the reactive part of the impedance of a
closed tube (MA = ρl/S).
Since we now need two circuit elements to represent this equation, the circuit
representing this more exact approximation to Z A is the one shown in Figure 7.4. As
noted earlier, the capacitor, in this impedance analogy approximation of Z A, must
have one of its terminals at circuit ground.
7.4.2 Admittance Analogy
Sometimes, we need to represent the acoustic compliance in the admittance analogy.
Because the admittance is the inverse of impedance, we have
1
YA = = jωC A (7.22)
ZA
We note that in the admittance analogy the compliance will act as an inductance,
that is, mechanically as a spring. We draw the circuit element in the electroacoustical
admittance analogy using the same symbol as in electrical circuit theory, as an
inductance, as in Figure 7.3b.
To obtain a better approximation, one can of course use more terms also in the
series expansion for the admittance analogy expression.
7.5 ACOUSTIC CAPACITANCE DUE TO A STIFF DIAPHRAGM

Since any acoustic compliance due to air enclosed in a cavity must have one of its
terminals at circuit ground, it is necessary to find a way to include a series capacitance
element into an acoustic impedance analogy circuit. This can be done by using a
stiffness-controlled diaphragm (ideally massless), as shown in Figure 7.5, well below
the lowest resonance frequency of the diaphragm.
7.6 ACOUSTIC MASS

The acoustic mass is a consequence of the inertia of a volume of gas. Assume that
some gas is confined by the walls of a tube open at the far end and driven by a piston
at the closed end. In contrast to the previous case of acoustic compliance, since the
90 Electroacoustics
CAS
Massless diaphragm CAS
ZA,in CA
Air volume V
(a) (b)
FIGURE 7.5 A stiffness controlled (massless) diaphragm can be used to introduce series
acoustic compliance into an acoustic impedance analogy circuit (a) and (b).
gas is free to move at the open end, the only obstruction to its movement will be the
mass of the gas, the walls of the cavity, and the acoustic impedance at the far end
of the cavity.
When a gas moves in parallel to the walls of a cavity, the flow conditions at the wall
may be laminar or turbulent. Laminar flow is characterized by the particle velocity
of the gas always being parallel to the tangent of the wall at each point along the
cavity. Because of the viscosity of the gas, the particle velocity will drop close to the
wall. The vibratory energy lost due to this “braking-effect” will be converted to heat.
As in the case of acoustic capacitance, since the pressure increase is accompanied
by a temperature increase, there will be a heat exchange with the cavity walls. If
the walls of the cavity are not ideally smooth, then there will be further losses. The
theory for losses in tubes due to viscosity in laminar flow is considered in Ref. [11].
For large surface wall roughness and high gas velocities, the flow will become
turbulent. Turbulent flow is characterized by the formation of eddies, the flow close
to the wall no longer being parallel to the wall. This is a nonlinear effect that leads to
comparatively high losses and is not further considered theoretically here.
The acoustic mass, as an impedance element, should correspond to inductance in
an electric circuit, that is, have a multiplying j-operator and a frequency-dependence,
which is directly proportional to frequency. An acoustic mass can be viewed as a
“lumped” mass element with similar limitations as for the acoustic capacitance just
discussed.
Note that in the case of converting acoustical components into acoustic impedance
circuit symbols we tend not to use an intermediate “acoustic icon” as in the analysis
of mechanical circuits. Because of the relative “intuitiveness” of acoustic impedance
circuits, the conversion is done directly into acoustical circuit symbols using idealized
tube, volume, and resistance icons.
We wish to have an acoustical circuit element that has acoustic impedance that can
be written similar to that of an inductance, that is,
Z A (ω) = jω M A (7.23)

Here, MA is the acoustic mass that has an inductance-like property. In this section,
we will show that this is possible under some assumptions.
A requirement for viewing the impedance of the open tube as a lumped mass
element is the absence of modal behavior in the sound field inside the tube. In practice,
this means that the maximum “unwrapped” dimension of the volume must be much
smaller than the wavelength of sound at the frequency of interest. For the gas to be
the only impedance-contributing element, the walls of the tube are assumed to be
ideally rigid, the air to be nonviscous, and no heat to be exchanged with the walls.
The coupling of energy to the air outside, at the far end of the tube, will be handled
by a special “end-correction” to be discussed later.
Again, the short tube is used as an extremely simplified approximation of any
open volume, this tube having both length and cross section fulfilling the requirement
mentioned in the previous paragraph. Such an open tube is shown in Figure 7.6.
The tube is assumed to have its far-end surface perpendicular to the direction of
propagation and to open onto an infinite baffle at both sides so that the two openings
are only connected by the tube. The wave injected into the tube is assumed to have
come from a moving piston, also perpendicular to the direction of propagation.
The piston can be assumed to generate a positive-going wave in the x-direction in
the tube. The pressure of this wave can be written as
p( x, k ) = pˆ + e − jkx (7.24)

and its particle velocity as
pˆ + − jkx
u ( x, k ) = e (7.25)
ρc
The opening at the end of the tube will have a reflection coefficient r(ω) with
a magnitude slightly smaller than unity because of the radiation of sound from
the opening. For this analysis, however, we again assume the impedance Zw to be
much smaller than the impedance of the air in the tube, which leads to a reflection
coefficient of
r (ω) ≈ −1 (7.26)
ZA,in
x = –lx x=0
FIGURE 7.6 An open short tube used as an acoustic impedance component.

92 Electroacoustics
The complete sound field in the tube, consisting of both the forward- and
backward-going waves is

(
p( x, k ) = pˆ + e − jkx − e jkx (7.27) )
pˆ + − jkx
u ( x, k ) =
ρc
(
e + e jkx (7.28) )

The acoustic impedance, at the piston, looking into the tube, can then be written as
Z A ( x, k ) =
(
ρc e − jkx − e jkx ) (7.29)
S (e − jkx
+e jkx
)

Here, S is the surface area of the piston (and the cross-sectional area of the tube).
Assuming the length of the tube lx we can write the acoustic impedance seen by the
piston as
jρc
Z A ( −l x , k ) = tan ( klx ) (7.30)
S
that is,
ρc  l 
Z A ( −l x , k ) = j tan  2π x  = jX A (7.31)
S  λ
The reactance X A for this case is shown graphically in Figure 7.7.
Reactance
2×107
1×107
kx
–6 –5 –4 –3 –2 –1
–1×107
–2×107
FIGURE 7.7 The reactance X A(kx) of an open-ended tube (S = 0.0001 m2) as function of kx.
Note that for this graph the load impedance at x = 0 has been set to zero.
As long as the length of the tube is small compared to the wavelength of sound λ,
at frequency ω, the series expansion of the tangent can be approximated by one or
two terms without large error. The series expansion of tan(klx) gives
tan ( klx ) = klx +

( klx )3 + 2 ( klx )5 ...  π
3 15  klx < 2  (7.32)

If we retain only the first term of the series expansion, we find the following
expression for the impedance:
Z A (ω) = jω M A (7.33)

as desired. The acoustic mass is
ρlx
MA =  kg/m 4  (7.34)
S  

This of course assumes that the load impedance at the end of the tube is zero;
there would otherwise be an extra term representing the load.
We draw the circuit element in the electroacoustical impedance analogy using
the same symbol as in electrical circuit theory, as an inductance, as shown in
Figure 7.8. The acoustic inductance, that is, mass, will need to be in series with the
load impedance ZL .
By retaining two terms of the series expansion, we will have a better approximation.
A comparison of the values given by the two-term approximation to the exact
formulation shows us that now, as long as the length of the tube is less than 1/8 λ,
the error in the approximation will be less than 5%. Retaining the first two terms,
the expression for Z A, assuming a zero impedance load ZL at the far end of the tube,
can be reformulated as
1
Z A (ω ) ≈ (7.35)
j ωC A 1
+
3 j ωM A

MA MA
(a) (b)
FIGURE 7.8 The symbol for an acoustic mass: (a) impedance-type analogy; (b) admittance-
type analogy.
94 Electroacoustics
MA
ZA
CA/3 ZL
FIGURE 7.9 A more accurate representation of the reactive part of the impedance of an
open-ended tube.
Here, the compliance CA is
Slx
CA = (7.36)
ρc 2
The parallel circuit of two elements needed to represent this more exact
approximation to Z A is shown in Figure 7.9.
As noted earlier, the capacitor, also in this circuit element approximation of Z A,
must have one of its terminals at circuit ground. A list showing the errors associated
with the various approximations of the impedance of an open-ended tube can be
found in Table 7.1.
7.6.1 Lined Tube
The acoustic analogy of a tube lined with a locally reacting porous sound absorber
will contain extra resistance and capacitance to reflect the impedance of the lining as
shown in Figure 7.10 [3]. This approach can also be used for the other approximations
shown in Table 7.1. The components R AL and CAL in Figure 7.10 are similar to the
components of a short closed tube discussed previously. For a circular tube of width
dL and length lx one finds the approximate components as shown in Equations 7.37
through 7.39 that apply to a thin lining that has the normalized resistive impedance
rL at its surface, a tube open area of S, and a lining perimeter P.
The impedance components of the short tube section can be approximated
by Ref. [3] to be
lx S
CA ≈
2ρc 2
(7.37)
ρlx
MA ≈
S
TABLE 7.1
Errors Associated with the Various Approximations of the Impedance
of Open-Ended and Closed-Ended Tubes
MA
l < λ/16 5%
MA
l < λ/8 5%
CA/3
MA/2 MA/2
l < λ/3 10%

CA
MA/4 MA/2 MA/4
l<λ 10%
CA/2 CA/2
MA/n 2MA/n 2MA/n MA/4
l < nλ/4 10%

l < nλ/8 3% CA/n CA/n CA/n CA/n
Source: Bauer, B.B., J. Acoust. Soc. Am., 38(5), 882, 1965.

96 Electroacoustics
MA/2 MA/2
RAL
CA
CAL
FIGURE 7.10 The impedance analogy for a tube internally lined with a thin locally reacting
porous absorber. (After Molloy, C.T., J. Acoust. Soc. Am., 21(4), 413, 1949.)
2ρcrL
RAL ≈ (7.38)
lx P
lx dL PL (7.39)
C AL ≈
2ρc 2
7.7 LENGTH-END CORRECTIONS

In our analysis we have so far assumed that the impedance load is ZL = 0 at the far
end of the tube. It is clear though that some of the air outside the tube opening at
the far end will move as well as the air inside the tube, as indicated by the flow lines
shown in Figure 7.11. This means that the effective length of the tube is larger than its
geometrical length. We compensate for this by introducing a length-end correction.
The length-end correction needs to be applied to any tube opening that opens into an
environment or cavity that is “much larger” than the tube. The usefulness of the end
correction is limited to those cases where the diameter of the tube is much smaller
than the wavelength.
FIGURE 7.11 Approximate flow lines at an orifice.

The end correction is associated with the radiation and near-field behavior of sound.
There will be some power loss from the far end of the tube, since the air is moving
and the movement of air will continue out to infinity if the tube opens into a free
half-space. This power loss can be represented by introducing a radiation impedance,
which contains both the effect we approximate by the end correction, that is, as a mass-
type reactance, and a resistance representing the power loss due to radiation. Since the
resistance is small, it is generally not included in the approximate end correction. We
will study the better representation of radiation impedance in Chapter 10.
Since the length-end correction depends on the geometry of the air movement at
the orifice, it will depend on the geometry that the opening is set in. In practice we
need to consider two cases: (1) the orifice in an infinite flat baffle and (2) at the end of
an infinitely long tube. The first case is of interest primarily for loudspeaker design,
and the second case for microphone design. The extra mass MA,end baffle added at the
end, and the length-end correction lcorr,baffle for a small circular hole in an infinite
baffle are given by
8ρ ρ
M A,end baffle = ≈ 0.270 (7.40)
3π a
2
a
8a
lcorr ,baffle = ≈ 0.849 a (7.41)
3π
The extra mass MA,end tube and the length-end correction lcorr,tube for a small circular
hole at the end of a semi-infinitely long tube are given by
ρ
M A,end tube ≈ 0.195 (7.42)
a
lcorr ,tube ≈ 0.613 a (7.43)

The representation of the approximate impedance into a tube having an open end can
then be drawn as shown in Figure 7.12.
When the tube is very short, for example, a hole in a thin sheet, it is necessary to
include end corrections for both sides of the hole. The same applies to the neck of a
Helmholtz resonator or to the duct of a ported box loudspeaker. In the latter cases,
it is then necessary to remove the end correction volume from the cavity or box
volume. While this gives a better approximation to the resonance frequency of the
resonator, a more exact analysis requires numerical methods, for example, the use of
finite element modeling.
For the case of a reasonably round, but noncircular hole, the dimensions of which
are much smaller than the wavelength of the sound, one can approximate the length-
end correction by assigning a radius aeff to it, the size of
S
aeff ≈ (7.44)
π
98 Electroacoustics
MA
ZA
CA/3 MA, end
FIGURE 7.12 An approximate representation of the reactive part of the impedance of an

open-ended tube including the end correction.
To find better approximations to the length-end correction for general hole

geometries and radiation cases, it is necessary to use numerical modeling such as
finite or boundary element modeling. Finally, it is important to remember that for air-
flow velocities at the tube opening of over about 7–10 m/s, there is risk for turbulence
which will introduce severe nonlinearity in the circuit.
7.7.1 Acoustic Resistance
Power loss in acoustical circuits can occur due to not only radiation but also vis-
cous and other losses. It is convenient to use the idea of volume flow through
acoustic resistance as the place where the losses occur. The attenuation due to air
absorption of sound energy could be included as a resistance, but since these losses
are extremely small over the small fractions of a wavelength which we study they
are generally neglected.
Acoustic resistance is needed to adjust the properties of microphones and loudspeakers
so that the desired frequency response and damping are obtained. Practical approaches
to incorporating acoustic resistance into acoustical circuits include the use of narrow
tubes and slits, cloths and weaves, and sintered materials. In all of these, the reason for
the acoustic resistance is the nonlaminar flow due to the influence of the viscosity of
air and the narrow channels due to narrow canals, densely packed particles or fibers.
Figure 7.13 shows examples of such a weave. The weave is also often used to symbolize
resistive elements in acoustical circuits as shown in Figure 7.14.
In a very narrow, shoebox-shaped slit such as the one shown in Figure 7.15, the
velocity profile will become parabolic resulting in a higher pressure difference being
needed to provide a certain volume flow. Since the flow is confined to the slit or
tube, it is also necessary to include the mass of the air inside the slit or tube into the
expression for the resulting impedance.
The acoustic impedance of such a slit having height h, width b, and length lx,
respectively, and where h < 0.003√f, is obtained as [4]
12 ηlx 6 ρlx
Z A (ω) = RA + jX A = + jω (7.45)
hb3
5 hb
FIGURE 7.13 A “Dutch” weave of metal wire is sometimes used to obtain a precise acoustic
resistance with little inductance.
RA RA rA
(a) (b) (c)
FIGURE 7.14 Symbols for acoustic resistance: (a) acoustic circuits; (b) acoustic impedance
analogy; (c) acoustic admittance analogy.
The dynamic viscosity of air, η, will depend on temperature, and at normal

conditions is about 1.86 · 10 −5 Ns/m2.
The acoustic impedance of a tube having radius a and length l, respectively, and
where a < 0.002√f, is obtained as [4]
8 ηl x 4 ρlx
Z A (ω) = RA + jX A = + jω (7.46)
πa 4
3 πa 2
In most cases however, such as when dealing with weaves, sintered or fibrous
materials, there will be many canals of different effective widths and lengths. In
these cases, the aforementioned equations are not directly applicable, and the acous-
tic resistance will need to be found by measurement.
Manufacturers of weaves for acoustic purposes will generally provide steady-
state (DC) flow resistance data. Note however that typically the impedance of the
weave will be different for AC flow and that the resistance will become nonlinear
100 Electroacoustics
lx
FIGURE 7.15 The slit geometry discussed in the text.
at high flow velocity. Alternatives to weaves are perforated plastic or metal foils,
wools—both natural and synthetic. Such materials can be used for general sound
absorption purposes, and are also useful, for example, in loudspeaker box design.
A disadvantage of thin, nonrigid materials is that the mechanical properties of the
weave or foil will become important when the resistance is high. Sintered metal or
ceramic resistors will be more rigid than resistors made from weaves, foils, and wool.
High-rigidity acoustic resistors can also be made from metal foams.
For sintered and fibrous materials, it will also be necessary to include a porosity
factor and a structure factor to describe the material. These materials can be modeled
in a general way using classic approaches but practical data for the acoustic resistance
can only be had by measurement. Variable acoustic resistors are available for special
purposes such as loudspeaker enclosures.
7.7.2 Resistance of Thick Sheets

Sometimes, it is necessary to use a plug of porous material such as natural or
synthetic wool, for example, glass or mineral wool, a sintered material, or a metal
foam. Figure 7.16 shows an example of the skeleton of such a metal foam. If a
material has a specific flow resistance of Ξ (Ns/m4), a plug of the material having the
cross section S and thickness h will have the static acoustic resistance R A
h
RA = Ξ  Ns/m 5  (7.47)
S
For audio frequencies, the reactance due to the mass of the air in the plug needs to
be added to find the total impedance of the plug.
Note that the propagation speed of waves in porous materials, such as fiberglass,
is lower than that for free waves, since propagation is no longer adiabatic. There is
heat transfer between the wave and the “skeleton” of the material.
FIGURE 7.16 A metallic foam structure.
7.7.3 Resistive Terminations
Sometimes, it is required to load the end of an acoustical circuit by a resistance.
Such a resistance can in practice be designed in two ways, either as an opening to
the outside, using a mesh or similar flow resistance, or as a resistive “stub.” The two
methods are illustrated in Figure 7.17.
The first method has several drawbacks, such as that sound from the outside may
leak into the circuit, that the resistance will also be featuring a small inductance
because of the end correction (radiation impedance), and that the resistance is subject
to contamination by dirt, water, etc., from the outside. Two advantages are that the
resistance will typically become more ideal toward low frequencies than at high
frequencies and that the value of the resistance can be chosen freely.
The method of using a closed tube with a resistive sound-absorptive termination
only offers a limited range of resistance, since the acoustic resistance is given by
ρc
RA = (7.48)
S
A wedge of suitable sound-absorptive material such as glass wool will effectively

absorb the wave traveling in the tube provided that the wedge is more than one quarter
wavelength long. We can use this technique to absorb the backward-traveling wave
from the rear of a loudspeaker to have a nonresonant reflection-free enclosure. Of
course we may still have a small opening at the end of the wedge; this is sometimes
used in the design of “transmission line” loudspeakers for various reasons. If the
dimensions of the tube are so large that there can be higher order modes along with
the plane wave mode, one possibility is to use subdividing walls inside the tube.
To generate high values of R A a different technique is necessary. Since narrow
tubes will feature high flow resistance because of the viscosity of air, we can use thin
In In
Flow Wedge made of sound-

resistance absorbing material
(a) (b)
FIGURE 7.17 Two types of resistive terminations of acoustical circuits: (a) simple flow
resistance open to “ground”; (b) resistive “stub” using a wedge of tapered sound-absorptive
porous material.
tubes (round or flat, for example) to generate “resistors” with large values of acoustic
resistance. These can be adjusted in length or used in parallel.
For very high values of resistance, we can use a tube with a wire inserted to
generate a controlled resistance slit; this technique is sometimes used in condenser
microphones to generate a well-controlled low-frequency limiting of the frequency
response of the microphone capsule.
7.8 ACOUSTIC TRANSFORMERS

Acoustic impedance transformation is more difficult to achieve than either
mechanical or electric impedance transformation. The acoustic transformer must,
by physical necessity, be quite large, and will only work well with simple sound
fields, such as those in tubes where the wavelength of sound is much larger than the
tube diameter.
7.8.1 Abrupt Change of Cross-Sectional Area

A simple change of cross section, such as that shown in Figure 7.18a, will not cause
any change in pressure or volume velocity and the acoustic impedance seen from
the left side immediately before the area change will be that of the right-hand tube.
p1 = p2 (7.49)

U1 = U 2 (7.50)

Expressed as mechanical impedance, there will however be a transformation. The

transformation ratios are indicated in Figure 7.18b and c.
One must remember though that there will be a reflected wave due to the discon-
tinuity. To avoid reflection, the two tubes must have the same acoustic impedance,
that is, Z A1 = Z A2. For tubes filled with air, Z A = ρc/S; so, for the case of unequal tube
cross-sectional areas, the reflection coefficient r(ω) will be given by
Z A2 − Z A1 S1 − S2
r (ω ) = = (7.51)
Z A2 + Z A1 S1 + S2
S1 ZA1 S2
U1 U2 (a)
u1 u2
ZA1 p2 p2 ZA2 ZM1 F1 F2 ZM2
1:1 S1 : S2
(b) (c)
FIGURE 7.18 A cross-sectional area discontinuity between two pipes: (a) S1 and S2 are the
cross-sectional areas of each pipe; (b) acoustic impedance analogy; (c) mechanical impedance
analogy.
7.8.2 Exponential Couplers

Exponential couplers may be used to change the relationship between sound pressure
and particle velocity at different places in acoustic systems as in the horns discussed
in Chapter 19. The simple quarter-wave tube impedance transformer can be thought of
as a simple but extreme coupler, only functioning at certain frequencies. Exponential
couplers are characterized by their taper, that is, expansion rate, and length. These
properties will determine the major acoustical characteristics of the coupler as an
impedance transformer. To obtain a change in impedance, it is necessary to allow for
a gradual change in exponential coupler’s cross section, as that shown in Figure 7.19a.
The figure shows an exponential coupler which is characterized by a gradual
change of cross-sectional area S(x) according to
S ( x ) = ST e mx (7.52)
x=0 x=l
S1 S(x) = S1emx S2
(a)
S2 : S1 S1 : S2
U1 U2 u1 u2
(b) (c)
FIGURE 7.19 An exponential coupler inserted between two pipes (cf. Figure 7.18): (a) S1 and
S2 are the cross-sectional areas of each pipe; (b) acoustic impedance analogy; (c) mechanical
impedance analogy. Note: Transformation ratios in (b) and (c) only valid at high frequencies.
Here, x = 0 is the starting point of the gradual expansion. The parameter m is
determined by choosing the coupler cutoff frequency fcutoff.
mc
fcutoff = (7.53)
4π
Below the cutoff frequency, the coupler will behave essentially like the abrupt
area change discussed previously; so there will be no impedance transformation.
The transformers are symbolically drawn as shown in Figure 7.19b and c. Because
of the reflections at the coupler’s end, the input impedance at x = 0 will not be smooth
over part of the frequency range as shown in Figure 7.20.
2.0
Normalized acoustic input impedance
Cutoff frequency = 200 Hz

Area ratio = 2
Real part
1.5
Imaginary part
1.0
0.5
0.0
100 200 500 1000 2000 5000 10000
Frequency (Hz)
2.0
Cutoff frequency = 200 Hz
Normalized acoustic input impedance
Area ratio = 4
Real part
1.5 Imaginary part
1.0
0.5
0.0
100 200 500 1000 2000 5000 10000
Frequency (Hz)
FIGURE 7.20 An exponential coupler inserted between two pipes can be used as an imped-
ance-matching device. Both couplers have a cutoff frequency of 200 Hz.
Since exponential horns are important as impedance transformation elements for

loudspeakers, they will be discussed in more depth in Chapter 19.
7.8.3 Quarter-Wave Transformer
A special type of impedance transformer is the “quarter-wave transformer” shown in
Figure 7.21a in which a tube of cross section S2 connects two tubes of unequal cross
sections S1 and S3.
The cross-sectional area S2 must be
S3 = S 1 S2 (7.54)

The quarter-wave transformer will only function at the frequencies to which it is

tuned. These are determined by the length of the transformer section ltransf, relative
to the wavelength of sound as
2n + 1
ltransf = λ n = 0,1, 2,... (7.55)
4
Continuous media can also be used for impedance transformation purposes;

although this is not practical for air, the principle is used for solid materials. The
bandwidth of the transformer is determined by the real part of the impedances that
are connected to either side.
X=0 X= λ
ltransf 4
S1 S3 = S1S2 S2
(a)
S2 : S1 S1 : S2
U1 U2 U1 U2
(b) (c)
FIGURE 7.21 A quarter-wave transformer inserted between two pipes: (a) S1 and S2 are the
cross-sectional areas of each pipe; (b) acoustic impedance analogy; (c) mechanical impedance
analogy.
7.9 ACOUSTIC GENERATORS

Acoustic generators are characterized by their ability to generate an oscillating
acoustical flow with associated sound pressure and particle velocity. Under most
circumstances, there will also be wave propagation which carries energy away from
the generator. Generators differ in their ability to move air against the pressure
created by the load impedance. In this way, acoustic generators behave like their
electric counterparts.
Electric generators having zero internal impedance are called voltage generators;
their acoustic impedance analogy counterparts are called constant pressure generators.
Analogously, electric generators having infinitely high internal impedance are called
constant current generators and their acoustic impedance analogy counterparts are
constant volume velocity generators. As with electric generators, it is difficult to
design ideal generators of either type to work over a wide frequency range. Also,
acoustic generators will have internal impedance.
No special symbols are typically used for generators in acoustical circuit
analogies since most generators will be based on the use of mechanical generators
in combination with a piston. Schematically, the generators can be drawn as shown
in Figure 7.22.
Examples of low internal impedance generators are aerodynamic sources, such as
those created by modulated air flows (of which sirens are crude examples). Modulated
air flows can also be generated by electrodynamic and piezoelectric flow-modulating
valves. One can achieve very powerful sound sources by modulated burning of, for
example, a butane and air mixture.
The plasma-type generator discussed in Chapter 25 is also a low-impedance source
[5]. Yet another type of low-impedance generator is the hot wire loudspeaker or ther-
mophone, based on the air expansion around thin electrically heated wires [6,7].
Conversely, high internal impedance sources are characterized by the acoustic
flow being generated by direct mechanical excitation by a vibrating piston, as in
the conventional electrodynamic loudspeaker. Another even more extreme example
is the electromagnetic telephone headphone receiver. The Pistonphone shown in
Figure 7.23 is an example of a very high internal impedance generator used for
microphone calibration purposes.
It is important to note however that if the front and back sides of the piston are
not isolated from one another, as for example, when a loudspeaker is mounted on a
U U p p
(a) (b) (c) (d)
FIGURE 7.22 Acoustic analogies for generators: (a) a volume velocity generator for
admittance analogy circuits; (b) a volume velocity generator for impedance analogy circuits;
(c) a pressure generator for admittance analogy circuits; (d) a pressure generator for impedance
analogy circuits.
Pistons moving Microphone facing

in opposite directions outer air cavity
Sealed inner air cavity
Outer air cavity
FIGURE 7.23 The Pistonphone principle: The two pistons oscillate in opposite directions to
minimize mechanical vibration in the handheld unit, modulating the volume of the enclosed
air. The oscillating pistons generate a well-defined sound pressure in the closed volume and
is used to calibrate microphones.
finite baffle or on a vented box, their joint behavior will be that of a dipole source,
when the piston extension is much smaller than the wavelength of sound generated.
7.10 POWER RELATIONSHIPS

The power dissipated in an acoustical circuit depends on the “current” through the
resistive elements. Since we are interested in both the impedance and admittance
analogies, the power loss for an impedance analogy can be written as
PA = RAU 2 (7.56)
For a mobility analogy, the power loss is written as
PA = rA p2 (7.57)

Here, R A and rA are the acoustic resistance and conductance, respectively.

In contrast to mechanical circuits, acoustical circuits are initially drawn in their
impedance form, and it is advantageous to calculate the acoustical power losses from
the situation in that analogy.
Again, it is advisable to stay away from calculation of the losses after various
forms of conversion, particularly after involving transducer mechanisms, etc.
7.11 FILTERS
We will often need various forms of acoustic filters to limit or extend the frequency
response of acoustic systems such as loudspeakers and microphones as well as in
various laboratory and experimental use. Often, we can regard our filter application
as a part of a system such as that shown in Figure 7.24. The filters are based on the
use of frequency-dependent reactance, usually used to generate resonance.
Sound source Transmission Transmission Radiation

having internal Acoustic impedance
line, length l1 line, length l2
impedance Zint filter load Zrad
and area S1 and area S2
FIGURE 7.24 A typical situation involving the use of a filter.
In the design of resonant filters it is necessary to carefully control the damping

of various circuits. This can be a problem in acoustical circuits since it is difficult to
isolate the component types into resistance, inductance, and capacitance, as well as
one can in dealing with electrical components. Practical acoustic resistances typically
have a considerable inductance component for example. Acoustic capacitances
typically have both parallel and series resistances associated with them. This will
make it difficult to obtain very narrow filters, such as those used in band-stop filters.
The temperature-dependence of the components makes it difficult to build stable
resonant circuits.
7.11.1 Low-Pass Filters
For low-pass filters we need either acoustic inductance in series in the circuit
or acoustic capacitance shunting the circuit as discussed in Appendix B and
shown in Figure 7.25. An example of the use of series inductance is the use
of thin foils and textiles, both woven and nonwoven. Textiles are primarily
used to insert flow resistance. As previously described, such materials typically
also feature inductance in series with the resistance. Foils are used as protection
against dust and humidity.
Loudspeaker and microphone grilles, needed to protect transducers from rough
handling, also act as low-pass filters. Tissues are occasionally used to limit the high-
frequency response of loudspeaker so as to give them a “sweeter” tonal character.
Various fabrics and foils used as protective covers for sound absorbers also act as
low-pass filters.
Figure 7.26 shows the circuits resulting from the use of foils and textiles for plane
waves incident perpendicularly on plane sheets of such materials. In this circuit,
RF is the flow resistance of the weave, and m″ is the mass per unit area. The source
and load impedances of the “free” wave is the impedance seen as one looks in either
direction.
Series inductance Shunt capacitance
FIGURE 7.25 Series and shunt low-pass filter systems.

ρc RF m˝
ρc ρc 2p pout ρc
Incoming Weave Outgoing

RF
m˝ wave wave
FIGURE 7.26 A plane wave passing through a weave or other mixed acoustic resistance and
inductance components, and the equivalent electrical circuit for the system.
The sound pressure in the outgoing wave can be found by simple voltage division.
The reflection coefficient r is found by using the source and load impedances
Z1 and Z2, respectively:
Z1 = ρc (7.58)

Z 2 = RF + jωm′′ + ρc (7.59)

It is important to remember that the effective inductance and resistance will vary
with the angle of incidence of the plane sound wave, Π, to the weave; their values
will become
RF ,eff = RF cos ϕ (7.60)

′′ = m′′ cos ϕ (7.61)

meff

Since the sound field is not a plane wave close to a loudspeaker, an exact
analysis will require the use of numerical techniques such as finite element method
modeling.
More complex low-pass filters often need to be implemented in acoustical circuits.
This can be done by combining the elements shown in Figure 7.25. By series of
combination of components as shown in Figure 7.27, one can obtain filters having
very sharp cutoff frequency response properties. For lower filter cutoff frequencies
the air mass in the tubes can be replaced by a higher density gas, or even U-shaped
tubes carrying fluids such as mercury can be used.
The characteristics of acoustical filters are determined the same way as those
of analog electrical filters; so it is in principle possible to design various filters.
Figure 7.28 shows how a low-pass filtering loudspeaker enclosure could be designed.
7.11.2 Band-Pass and Band-Reject Filters

Sometimes band-pass and band-stop filters are necessary to produce a desired
electroacoustical effect such as a frequency response dip at low frequencies to
compensate for a resonance or a slight peak at high frequencies to increase speech
intelligibility.
In Out
FIGURE 7.27 An idealized low-pass filter and its equivalent electrical circuit.
Compressed glass fiber

or other similar sound -
absorptive sheet
FIGURE 7.28 Implementation of a low-pass filter for a loudspeaker enclosure of the

labyrinth or transmission line type. Shaded area in picture is the sound-absorbing material
used to give mass and compliance network as shown in Figure 7.27. (After Olson, H.F.,
Acoustical Engineering, 3rd edn., D. van Nostrand, Princeton, NJ, 1957.)
CA MA
In Out
FIGURE 7.29 A simple band-pass pass filter and its associated acoustical parallel circuit
analogy.
The simplest type of acoustical band-pass filter is the resonant circuit using an
acoustic mass–compliance combination as shown in Figure 7.29. The damping of
the resonance will depend on the losses in the mass and compliance components
and is usually specified as a Q-factor as described in Appendix B. In practice, it is
extremely difficult to reach Q-factor values larger than 100 in acoustical circuits.
This is due to two effects, the losses due to viscosity at the walls influencing the
oscillating flow in the mass component and the heat transfer from the sound wave
to the walls of the compliance volume. Figure 7.30 shows a resonance circuit used
to select frequencies for a microphone used in a reverberation enhancement system.
The familiar noise of the sea shell at the seashore is due to its resonances enhancing
the sound that arrives to the sea shell.
It is also important to remember that at high flow velocities, typically 7–10 m/s,
turbulence will begin to be generated at the opening edges of the tube mass component.
This will result in added damping and a nonlinearity. This type of nonlinearity can
cause unpleasant swishing noise in the low-frequency sound of reproduction by ported
loudspeaker boxes since the port is a part of the resonant circuit.
Band-reject filters may be designed in a manner similar to band-pass filters. A simple
band-reject filter is shown in Figure 7.31. We recognize this filter as being simply a
Helmholtz resonator inserted into the circuit. We sometimes call this type of filter a
notch filter because of the graphical shape of its frequency response if the filter is sharply
tuned, that is, has a narrow relative bandwidth. In the same way as with the band-pass
filter, the Q-value of the resonator will depend on the losses in its mass and compliance
components, as well as on the real part of the impedance levels of the circuits connected.
Air cavity
Microphone
FIGURE 7.30 Using a microphone inside Helmholtz resonator, one can select small
frequency range for sound pickup.
In
MA
CA
Out
FIGURE 7.31 A simple frequency band-rejection filter and its associated acoustical series
circuit analogy.
In
stub
Further
resonances
Out
f1 f2
FIGURE 7.32 Close to one of its resonance frequencies, the stub will behave as a high-Q
tuned series resonance circuit.
Still another way to accomplish a notch filter is to attach one of the previously
discussed quarter-wave transformers to the circuit as indicated in Figure 7.32.
Such a piece of transmission line, sometimes called a “stub,” will feature very low
impedance at the point where it is attached since the input impedance of a quarter-
wave transformer is
Z A2 0
Z in = (7.62)
Z load
Here, Z A0 is the characteristic acoustic impedance of the transmission line and Zload is
the impedance of the load at the end of the transmission line. If the stub is blocked at
the end, Zload will be very large, and consequently the input impedance will be very
low. It is important to remember that the stub has many resonance frequencies, and
that above some frequency there will also be resonances due to the oblique modes
that can propagate in the stub. The theory described here is only valid for narrow
stubs assuming plane wave propagation.
We can also use the stub (including the wedge or other flow resistance) with an open
end, which will allow both gas flow and reasonably well-controlled acoustic imped-
ance. Another drawback of using stubs is that there will be many resonances, since
the action is based on a transmission line and not on discrete components (although
these of course will also exhibit wave-related phenomena at high frequencies).
We can also use multiple stubs to create band-reject and band-pass filters. This
is often done in various types of mufflers used to attenuate tones and other noise
0.25 m 0.25 m 0.25 m 0.25 m 0.25 m 0.25 m
0.25 m
80
60
Transmission loss (dB)
40
20
–20
20 50 100 200 500 1000 2000
Frequency (Hz)
FIGURE 7.33 Transmission loss incurred by three reactive filters of different complexities.
Cylindrical tubes having an area ratio = 10. Note how the λ/4 resonance of the stubs in each
expansion chamber introduces large insertion loss at approximately 0.7 kHz in the case on the
right. (After Davis, D. D. et al., Theoretical and experimental investigation of mufflers with
comments on engine-exhaust muffler design, National Advisory Committee for Aeronautics,
Report # NACA-TR-1192, 1954.)
coming from combustion engines and in air-handling equipments. Figure 7.33 shows
some filter designs and their associated transmission loss data. The transmission loss
is the ratio between the sound power in the output tube relative to the sound power
incident from the source. In dealing with these types of applications, it is important
to remember that gas flow through an acoustic filter will affect the operation of the
filter; for high-speed gas flows the performance will be very different from that
calculated for steady-state conditions.
7.11.3 High-Pass Filters
We also need high-pass filters since it is sometimes necessary to eliminate leakage
and flow, that is, DC components, in the system. While it is possible to use a mass-
type component, that is, open tube, to shunt the system to obtain a high-pass filter
action, this does not solve the problem of leakage. We also must remember that
it is not possible to have an acoustic compliance in series in the circuit unless the
compliance is implemented as a membrane, as discussed previously. The principles
of such filters are shown in Figure 7.34.
In Out In Out
Massless mechanical Mass shunt

compliance
(a) Series capacitance (b) Shunt inductance
FIGURE 7.34 Series (a) and shunt (b) high-pass filter systems and their acoustic impedance
analogies.
7.12 USING ACOUSTICAL ANALOGIES WITH FREE WAVES

It is convenient to use acoustical analogies also for the study of waves that propagate
freely without being confined to tubes and ducts. As an example, let us study the
sound field impedance conditions in a free plane wave.
The situation is shown in Figure 7.35. The plane wave is assumed to move in
the x-direction, starting at negative infinity in x, moving past x = 0 and on toward
positive infinity. Note that there is no particular surface area involved since we are
looking at a plane wave. The sound field impedance at x = 0 must be ρc, and the sound
pressure p. Note that in contrast to the situation in conventional acoustic impedance
analogies the current flowing is now particle velocity. Since the impedance seen by
the wave must be the same in both the negative and positive x-directions, the source
ρc −∞ ∞
2p p ρc
FIGURE 7.35 A plane wave passing by at x = 0. Note that the source needs to be assigned
double the pressure at x = 0.
pressure must be 2p. The circuit also fulfills the need for available intensity and
power since the intensity will be
p2
I = ρcu2 =  W/m 2  (7.63)
ρc  
and the power available over a surface S will be
P = ρcu2 S (7.64)

7.12.1 Plane Wave Incident on a Plane Sheet Having Mass

Now let us study the case of a plane wave that is incident perpendicular to a limp
plane sheet having a surface mass m″. We put the sheet at x and then obtain the
circuit shown in Figure 7.36. The impedance of the sheet is
Z = jωm′′ (7.65)

so the vibration velocity of the sheet is
2p
uw = (7.66)
2ρc + jωm′′
Note that the circuit neglects the reflection of the incident wave. The transmitted
wave however is correctly modeled. We can use this result to calculate, for example,
the absorption coefficient α0 of a (perfect) porous absorber covered by a thin foil.
The power absorbed will be the power transmitted to the side having positive x that
needs to be compared to the power available discussed in the previous section. One
can show that the absorption α0 coefficient is
1
α0 = 2
(7.67)
 ωm′′ 
1+ 
 2ρc 

x=0
–∞ ∞
ρc
jωm˝ u
2p p ρc
FIGURE 7.36 A plane wave incident on a limp sheet at x = 0 having surface mass m″.
The sheet acts as a filter reducing the level of the transmitted signal by 6 dB/octave
above the cutoff frequency. Such filters are sometimes used in front of loudspeaker
drive units to reduce their high-frequency response.
When the wave is incident at an angle Π on the limp sheet, the particle velocity
of the wave u will be higher than the velocity of the sheet um, and we will have
um = ucos(Π). The apparent surface mass of the sheet will drop to m″ cos(Π) since
the sheet will be easier for the wave to move, and its impedance is lower.
For parallel dual limp sheets the low-pass action is even more prominent.
Figure 7.37 shows the circuit that describes the situation. Table 3.1 shows that to
obtain the sound field impedance for the air trapped between the two sheets one must
multiply the acoustic impedance by the “acting surface.”
One can show that a sheet that has both stiffness and mass can be modeled in a
similar way [10]. The analogy for such a sheet is shown in Figure 7.38. The angle of
incidence of the plane wave on the sheet is φ.
The resonance frequency given by
ω 3 B′ sin 4 (ϕ)
jωm′′ + = 0 (7.68)
jcB 4
is identical to the critical frequency previously discussed in Chapter 4.
x=0
–∞ ∞
ρc jωm1˝ jωm2˝
uS2
2p p ρc2 ρc
jωd
FIGURE 7.37 A circuit model for a wave incident on a pair of limp sheets having surface
masses m1″ and m2″ separated by an airspace d long.
x=0 ω3B΄sin4(φ)cos(φ)
−∞ jcB4 ∞
u
ρc jωm˝cos(φ)
2p p ρc
FIGURE 7.38 The circuit representing a wave incident on a sheet having both bending
stiffness B′ and mass m″.
REVIEW QUESTIONS
7.1. How are the acoustical equivalent circuit elements defined?
7.2. How are the acoustical equivalent circuit elements drawn and indicated in the
acoustic admittance and impedance analogies?
7.3. How does one practically accomplish acoustical components such as compli-
ances, resistances, and masses? What are the limitations when one wants to
implement them in practice?
7.4 Derive the values for the acoustical components of the more exact circuits for
closed and open-ended tubes shown in Figures 7.4 and 7.9.
7.5 How does one calculate the power lost in an acoustic analogy?
PROBLEMS
7.1 The figure below shows a force generator connected to a piston having mass
MM and (single side) surface area Sp. The piston connects on one side to a
closed cavity having the acoustic compliance CA. The piston is only loaded on
the side toward the cavity.
Task:
Draw the impedance analogy circuit for the system using a transformer for the
conversion between the mechanical and acoustical sides.
Piston Air cavity
F
7.2 The figure below shows a force generator connected to a piston having
mass MMD, suspension compliance CMS, and a (single side) surface area Sp.
The piston connects on one side to an acoustical circuit.
Task:
Draw the complete circuit for the acoustic impedance analogy for the system.
RA1
MA1 MA2 CA1
F
MA3
MMD and CMS
RA2
7.3 The figure below shows an acoustic impedance analogy

Tasks:
a. Draw an acoustical circuit, which describes the system.
b. What would be the acoustical equivalent of connecting an acoustic
resistance in parallel over the MA4CA4-link to the right in the acoustical
circuit analogy shown in the figure?
MA5 MA6 MA7
MA1 MA2 MA3 MA4

U
CA1 CA2 CA3 CA4
7.4 The figure below shows a section through a pressure-gradient microphone

Tasks:
a. Define and name the various components.
b. Draw the circuit for the acoustic impedance analogy.
7.5 Consider the case of a plane wave in a tube as shown in the figure below. The
method of acoustic impedance analogies can be used to study the reflection of
a plane wave at the boundary between two different media. At the boundary
x = 0, the volume velocity is UG, the sound pressure is pG, and the impedances
of the (infinitely long) tube parts are ZAI for x < 0 and ZA2 for x > 0. Since the
sound pressure must be continuous across the boundary, the pressures must
relate as pG = p2.
ZA1 pi ZA2
pt
pr
x
x=0
Tasks:
a. Draw the acoustic impedance circuit with a sound pressure source p 0.
b. Where in the circuit can pG be found? Express pG using pi.
c. Determine the reflection coefficient r = pr/pi.
Check your results using known special cases, Z A2 = Z A1, Z A2 = ∞, Z A2 = 0 …
7.6 The figure below shows an engineering approximation to the main components
of the voice mechanism. Draw the circuit for the acoustic impedance analogy.
For an explanation of the concepts of radiation mass and resistance see
Chapter 10.
Nasal cavity
Nasal pharynx
Nostrils
Soft palate
Vocal tract Lips
Epiglottis Tongue
Laryngeal
pharynx
Vocal cords
Larynx
CAl is the acoustic compliance at the laryngeal pharynx.

CA2 is the acoustic compliance in the vocal tract.
CA3 is the acoustic compliance in the nasal cavity.
MAl is the acoustic mass at epiglottis.
MA2 is the acoustic mass at the nasal pharynx.
MA3 is the acoustical radiation mass at the lips.
MA4 is the acoustical radiation mass at the nostrils.
R AI is the acoustical radiation resistance at the lips.
R A2 is the acoustical radiation resistance at the nostrils.
REFERENCES
1. Rossi, M., Acoustics and Electroacoustics, Artech House, Norwood, MA (1988) ISBN-
13: 978–0890062555.
2. Bauer, B. B., Equivalent circuit of a tube or spring, J. Acoust. Soc. Am., 38(5), 882–882
(1965).
3. Molloy, C.T., The lined tube as an element of acoustical circuits, J. Acoust. Soc. Am.,
21(4), 413–418 (1949).
4. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986) ISBN-
13: 978–0883184943.
5. Deraedt, A., Electroacoustic transducer using corona effect, Proceedings of the Audio
Engineering Society Convention, Paris, Paper No. 3037 (1991).
6. Arnold, H. D. & Crandall, I. B., The thermophone as a precision source of sound.
Phys. Rev., 10(1), 22–38 (1917).
7. Xiao, L. et al., Flexible, stretchable, transparent carbon nanotube thin film loudspeakers,
Nano Lett., 8(12), 4539–4545 (2008).
8. Olson, H. F., Acoustical Engineering, 3rd edn., D. van Nostrand, Princeton, NJ (1957)
Library of Congress Catalogue Card No. 57–8143.
9. Davis, D. D. et al., Theoretical and experimental investigation of mufflers with comments
on engine-exhaust muffler design, National Advisory Committee for Aeronautics.
Langley Aeronautical Laboratory, Langley Field, VA. Report # NACA-TR-1192
(1954) CASI: 19930092208 (http://hdl.handle.net/2060/19930092208) sampled
December 2011.
10. Kurtze, G., Physik und Technik der Lärmbekämpfung (in German), Verlag G. Braun,
Karlsruhe, Germany (1964) ASIN: B0000BKML2.
11. Blackstock, D. T., Fundamentals of Physical Acoustics, Wiley, New York (2000)
ISBN-13: 978-0-47131979-5.
8 Conversion between
Analogies
8.1 IMPEDANCE AND ADMITTANCE ANALOGIES

As we have seen in previous chapters, it is possible to represent both mechanical and
acoustic systems by circuit components in an electric analogy. We have also seen
that there can be a case made for each of two types of analogies, the impedance and
admittance analogies. (These are sometimes also called analogies of the first and
second kinds.) It is also clear from our study that it is simpler to convert the working
of a mechanical system in the physical world into a mechanical admittance analogy
than into a mechanical impedance analogy. In the same way, we have noticed that
for acoustic systems it is easier to convert them into acoustical impedance systems
than into acoustical resistance systems. In some literature mechanical admittance is
called mobility.
As will become clear in our studies of integrated systems containing mechanical,
acoustical, and electrical systems, we will need some way of converting the various
mechanical, acoustical, electric, impedance, and admittance representations into
one another.
We will use transducers to couple the electrical side of a system to mechanical
or acoustical systems, and we will study such transducers in a special chapter later.
In this chapter we will study the coupling of acoustical and mechanical systems
and the conversion of impedance and admittance representations of systems into
one another.
8.2 CONVERSION BETWEEN ANALOGIES

Any method to convert impedance and admittance analogies into one another must
feature the following properties [1,2]:
• Elements in series in the circuit of one analogy correspond to elements in

parallel on the other.
• Resistance-type elements become conductance-type elements; capacitance-
type elements become inductance-type elements; and inductance-type
elements become capacitance-type elements.
• The sum of the drops across the series elements in a mesh of one analogy
corresponds to the sum of the currents at a node in the other analogy.
Since the systems have these properties we say that they are duals of one another. We
can approach the problem of converting one dual into the other either mathematically,
121
by setting up the appropriate equation systems for voltage drops around all meshes
and current sums at each node, similar to the conversion between electric Δ and Y
circuits in Appendix A. We then just have to decide on what property we want to
consider the driving field quantity and which we want to consider the resulting field
quantity. In electrical engineering, most people will consider the voltage difference
between nodes as the field quantity and the current in the mesh as the resulting
quantity, in analogy with the flow of water.
8.3 “DOT” METHOD

We will now study a graphical method, called the “dot” method of switching
between the approaches of each dual. The method is used in electrical engineering
and applies here as well. The following procedure is used [1]:
• Place a dot at the center of each mesh of the circuit and one dot outside all
meshes. Number these dots consecutively.
• Connect the dots with lines so that there is a line through each element and
so that no line passes through more than one element.
• Draw a new circuit such that each line connecting two dots now contains
an element that is the inverse of that in the original circuit. The inverse
of any element may be seen by comparing the corresponding columns for
admittance-type analogies with impedance-type analogies of Table 5.1.
An example of the approach follows. Figure 8.1 shows a simple mechanical system
for which we wish to draw the circuit for the mechanical impedance analogy.
The first step in converting the system to a mechanical impedance analogy is to
convert the mechanical system drawing to a mechanical admittance circuit drawing
using the mechanical symbols discussed in Chapter 5. The resulting circuit diagram
can be seen in Figure 8.2.
In Figure 8.3 we have redrawn the circuit using electrical symbols, but we are still
in the admittance analogy domain.
We now need to prepare the circuit for the “dot” method conversion process so
that we can derive the impedance dual of the circuit. Following the rules given in
the beginning of this section, we place “dots” inside each mesh, number the dots,
uM
rM1 rM2
MM1 MM2
CM1 CM2
rM3 rM4
FIGURE 8.1 This mechanical system consists of seven components and “ground.”
Conversion between Analogies 123
MM1 rM3 rM1 MM2 rM4 rM2
CM1 CM2
uM
FIGURE 8.2 Mechanical circuit for the system shown in Figure 8.1.
rM1 rM2
FM MM1 rM3 MM2 rM4
CM1 CM2
uM uM2
FIGURE 8.3 Mobility-type analogy for the system shown in Figure 8.1.
rM1 rM2
MM1 rM3 MM2 rM4
8 7
CM1 CM2
2 3 4 5 6
FIGURE 8.4 Introducing the “dots” and their interconnections.
and connect the dots so that there is a line through each element and so that no
line passes through more than one element. We then obtain the circuit, as shown
in Figure 8.4.
We can now draw the new circuit in which each line connecting two dots now
contains an element that is the inverse of that in the original circuit, that is, capacitance-
type elements become inductance-type elements and vice versa, and conductance-type
elements become resistance-type elements and vice versa.
Following the lines and the rules previously mentioned, we then obtain the
resulting dual circuit shown in Figure 8.5. We note that elements that were previously
in series have now joined in parallel and vice versa.
MM1 RM3 MM2 RM4 CM2

2 3 4 5 6 7
CM1 uM2
uM FM 8 RM2
RM1
FIGURE 8.5 The dual of the circuit shown in Figure 8.3.
While our ground connection in Figure 8.4 is at dot number 1, we do not necessarily
need to consider point 1 in Figure 8.5 as the ground connection. As we connect the
mechanical impedance analogy circuit to the acoustical impedance analogy circuit,
we may need to reconsider what we wish to designate as ground in the circuit.
8.4 TRANSFORMATION BETWEEN MECHANICAL

AND ACOUSTICAL CIRCUITS
We need a mechanical–acoustical transformer to couple the mechanical and acoustical
parts of a circuit. A typical example of a mechanical–acoustical transformer is the
plane piston, such as the diaphragm of a loudspeaker that couples the piston to the
surrounding air.
Let us assume that we have a situation shown in Figure 8.6 which shows a piston
sending a plane sound wave down an infinite tube. The piston has an area S and
oscillates at a velocity u along the axis of the tube.
To generate a sound pressure p in front of the piston we need a force _f , and the
volume velocity U will be uS.
F = Sp (8.1)

U
u= (8.2)
S
The acoustical impedance seen by the piston is
p ρc
= = RA (8.3)
U S
RA
FIGURE 8.6 A piston generating a plane sound wave that travels down an infinitely long
tube which has an acoustical impedance R A.
F 1:S p S:1
u U
u U F p
(a) (b)
FIGURE 8.7 Symbols for mechanical–acoustical piston transformers. The piston driving
area is S: (a) admittance analogy circuits; (b) impedance analogy circuits.
The circuit symbols are those shown in Figure 8.7.

We will apply this method extensively in our studies of transducers such as
loudspeakers and microphones.
REVIEW QUESTIONS
8.1 Which are the requirements for the conversion of impedance-type and
admittance-type analogies into one another?
8.2 What is meant by the “dot” method?
8.3 How is the “dot” method applied in practice in converting one type of analogy
into the other?
8.4 How do we represent the coupling of mechanical and acoustic systems using
a radiating surface by electrical components?
PROBLEMS
8.1 The figure below shows the mechanical system of a vibration-isolated com-
pressor MM1. Assume a force acting on this mass and that R M2 = R M3 and
CM2 = CM3. Assume a force acting on this mass.
Task:
a. How can the circuit be simplified if R M2 = R M3 and CM2 = CM3?
b. Draw the mechanical mobility and impedance circuit analogies for the system.
MM1
RM1 CM1
MM2
RM3 CM3 RM2 CM2

ZMF
8.2 The figure below shows a mechanical system that has the rightmost mass rig-
idly connected to a structure with an unknown internal impedance Z MF.
Task:
Draw the mechanical impedance circuit analogy for the system shown in
Figure P8.2. Assume one-dimensional and linear motion.
C2M
C1M ZMF
FM C1M
0.5MM 2MM 0.5MM
C2M C2M C2M
8.3 The mechanical system of a loudspeaker is shown in the figure below.

Loudspeaker diaphragm mass MMD, suspension compliance CMS, suspension
losses R MS, magnet mass MMD, basket compliance CMB, loudspeaker
enclosure mass MME , support base compliance CMSB, support base losses
R MSB, electrodynamic force FM.
Task:
Draw the mechanical mobility and impedance analogies for the system.
FM
MMM CMB
CMSB
MMD RMS MME
RMSB
CMS
8.4 A machine having a mass MM1 hangs in a spring from a rigid surface. An extra
mass MM2 is inserted as shown in the figure below to reduce the vibration level
of the machine.
Tasks:
a. Draw the mechanical impedance analogy.
b. Determine the value of the extra mass so that the vibration of MM1 is
minimized.
RM1 CM1 FM
MM1 uM1
RM3 CM3
uM2
MM2
RM2 CM2
8.5 The mechanical impedance analogy in the figure below shows a system that is
excited by two force sources.
MM2
RM2
CM3
CM2
uM
MM1 RM3
CM4
RM1
F1 F2
CM1
Tasks:
a. Convert the circuit into a mechanical mobility analogy.
b. Sketch the corresponding mechanical system.
8.6 The system in the figure below has both acoustical and mechanical com-
ponents. The pistonic masses MM2 and MM3 are free to glide on the walls of
the cavity. The cavity is rigidly supported by the foundation.
MM1
CM1
MM2 S1
Cavity CA MA ZAR
MM3 S2
CM2
Task:
Convert the circuit into a mechanical impedance analogy.
REFERENCES
978–0883184943.
2. Gehlshøj, B., Electromechanical and electroacoustical analogies, PhD thesis, Academy
of Technical Sciences, København, Denmark (1947).
9 Transducer Operating
Principles
9.1 INTRODUCTION
9.1.1 Reversible and Nonreversible Transducers
A characteristic of electroacoustic devices is that they are able to convert signals
between the electrical, mechanical, and acoustical domains with retained time
histories. The transducer may be considered a gateway between two domains, the
primary and secondary “sides.” For physical reasons, it is not possible to achieve
a perfect transduction since noise and linear, as well as nonlinear, distortion will
usually affect the signal during transduction. It is useful to think of transducers
as systems composed of a mechanical part, the “pure” transduction, and then an
electric part or the reverse.
Transducers that allow transduction in both directions between the domains are
called reciprocal or reversible transducers. For a transducer to be reversible, its action
needs to be based on the use of stored electric or magnetic energy as electrostatic or
electromagnetic fields.
Nonreversible transducers are based on the use of variable resistance; examples
of such transducers are carbon microphones, strain gauges, etc., or electromagnetic
waves such as light. One can argue that resistive components could be used in reverse;
for example, a resistive wire could be used for transduction from the mechanical to
electrical domain by variable resistance. In the reverse direction, the wire could be
made to move a mechanical or acoustic system if it was dynamically heated and
had low thermal capacity, i.e., short time constant, but this principle is seldom used
because of its inefficiency. Recent developments in nanowire technology indicate a
possibility for practical loudspeakers to be designed using thermal effects.
Another example of a nonreversible transducer for generating sound is the modu-
lated airflow loudspeaker in which air under pressure is released through a valve
capable of dynamic action, thus modulating the flow.
9.1.2 Direct and Indirect Conversion

Transducer can be grouped by properties such as, for example, sensors and transmit-
ters, active and passive transducers, or direct and indirect conversion transducers.
Direct conversion transducers are those where the output signal is obtained in
a single simple step of transduction such as in passive transducers, for example,
electrodynamic loudspeakers and microphones.
129
Indirect conversion transducers are those where the transduction is obtained by a

field variable modulating a secondary signal; examples of such transducers are sensor
systems using the primary domain variable to modulate a carrier signal followed by
subsequent demodulation of the carrier signal to generate a signal in the secondary
domain. This principle is sometimes used in microphones and gramophone cartridges.
9.1.3 Active and Passive Transducers

Passive transducers use signal energy from the primary side to feed the secondary
side. Most common electroacoustic transducers such as loudspeakers, earphones,
and microphones are passive devices.
Active transducers use outside power to generate the electrical output signal. An
example of this type of transducer is the resistance microphone where sound is sim-
ply modulating a resistance to change the current in an electrical circuit. Another
example is the ionophone loudspeaker, which uses an electrically generated and
modulated plasma to generate sound.
A bias-voltage-operated condenser microphone however is not considered
active in spite of it needing a bias voltage supply to set up the electric field inside
the transducer, necessary for the microphone to operate. On the other hand, a
condenser microphone based on the modulation by sound of an electric carrier
signal could be considered an example of an active device. Thus, the terminology
is slightly ambiguous.
9.2 TRANSDUCER OPERATING BLOCKS

A transducer has mechanical and electrical properties that are subject to physical
laws. Because of the similarities in the governing differential equations in the
mechanical and electrical domains, the transducer properties can be expressed by
analogies in which there are mechanical and electric components characterized
by capacitance, inductance, and resistance. It makes sense to subdivide the
transducer’s electromechanical properties into blocks as shown in Figure 9.1, where
the mechanical and electrical systems are connected to a “virtual transformer”
representing the actual transduction.
The acoustical side of the transducer will then be coupled to the mechanical side
as shown in Figure 9.2, and radiation can be represented by the acoustical block
Microphone or other sensor
Mechanical system Transduction system Electrical system
Loudspeaker or other transmitter
FIGURE 9.1 Conceptual building blocks of electromechanical coupling.

Transducer Operating Principles 131
Microphone or other sensor
Acoustical system Mechanoacoustical Mechanical system

system coupling
Loudspeaker or other transmitter
FIGURE 9.2 Conceptual building blocks of acoustomechanical coupling.
because of the volume velocity generated by the elementary sources on the trans-
ducer’s radiating surfaces. Analogously, sensing can be represented as the action of
sound pressure on the acoustic system.
9.3 CONVERSION
As discussed in the previous section we can conveniently represent the transducer
as consisting of a mechanical and an electrical system, separated by a virtual
transduction transformer.
In this view, the mechanical properties are represented by the force F and velocity
u at the mechanical terminals of the transducer, and the electrical properties are
represented by the voltage e and current i at the electric terminals.
Because of the coupling between the mechanical and electrical properties of the
transducer, the mechanical and electrical systems will be mirrored in one another.
For transducers, the relationships between the mechanical and electrical properties
can be written using two transformation factors K ME and KEM and impedances ZE and
Z M as a pair of characteristic equations
F = K ME i + Z M u
(9.1)
e = Z E i + K EM u
where K ME will depend on the characteristics of the transducer: electrodynamic,

electrostatic, piezoelectric, etc. Typically, KEM = K ME for a reciprocal transducer. If
the transducer is not reciprocal, one of the coefficients KEM or K ME will be zero and
thus prevent the transducer from working in one of the directions.
Since it will not be possible to operate the transducer without some influence
of either system, it is useful to investigate the resulting mechanical and electric
impedances when the electrical and mechanical system terminals are either left open
or short-circuited. We will do this as we study various common transducer types.
All transducers are associated with “generic” mechanical and electric impedance
characteristics. For example, transducers based on the interaction between charge
and electrostatic fields are associated with electric capacitance. Transducers based
on the use of moving charge in a magnetic field are associated with inductance.
9.4 ELECTRODYNAMIC TRANSDUCERS

Many transducers are based on the interaction between magnetic fields. The term
electrodynamic transducer is used for those having a static magnetic field and a
dynamic field surrounding a moving electrical conductor. A conductor that carries
electric current in magnetic field is subject to the Lorentz force that moves the
conductor as shown in Figure 9.3. The force can be used to move a diaphragm or
membrane as in a loudspeaker or headphone.
The force dF on the conductor piece dl by the current i in a magnetic field having
flux density B is given by the equation
dF = B i sin(α)dl (9.2)

Here, α is the angle between magnetic flux lines and the tangent to the conductor. In
most applications such as with cylindrical voice coils and magnets in electrodynamic
loudspeakers and headphones and microphones with linear ribbon and isodynamic
headphones and loudspeakers, the angle will be 90° so the total force is
l
∫
F = B i dl = Bl i (9.3)
0

The transformation factor between current and force is thus
K EM = Bl (9.4)
The force characteristic equation describes the force acting on the conductor
F = Bl i − Z M u (9.5)

where mechanical system’s impedance Z M includes the compliance necessary to keep
the conductor inside the magnetic field. The impedance Z M naturally also includes
the mass of the voice coil and possible losses in the mechanical system.
B
ZM
MMF F
x
= current i into page
FIGURE 9.3 A conductor carrying electric current in a magnetic field is subject to the
Lorentz force. Z M is the mechanical impedance affecting the motion of the wire.
Bl : 1 1
LE0
i ZM F
e u
FIGURE 9.4 An electrical analogy for the electrodynamic transducer. Note that the second-
ary side of the transformer is a mechanical mobility analogy.
The transducer can be used in reverse to sense the movement of the conductor.
When the conductor is moving inside a magnetic field having flux density B, there
is a voltage e induced in the conductor that tries to counteract the movement. The
current i through the conductor also generates a voltage between the ends of the
conductor due to its self-inductance LE0, which results in the characteristic equation
for the induced voltage along a length l
e = jωLE 0 i + Blu (9.6)

Setting the outside force F to zero and solving the characteristic equations 9.5 and
9.6 for the impedance ZET = e/i as seen from the electric terminals, we obtain
2 1
Z ET = jωLE 0 + ( Bl ) (9.7)
ZM
We notice that K EM = K ME = Bl. A simple electrical analogy that fulfills the
requirements of the characteristic equations 9.5, 9.6 (and of course 9.7) is shown
in Figure 9.4. Note that the secondary side of the transformer is a mechanical
mobility analogy [1].
9.5 ELECTROMAGNETIC TRANSDUCER

9.5.1 Unbalanced Transducers
Electromagnetic transducers are based on the modulation of the attraction force
between two poles of a magnetic system. They are also called moving armature
transducers or variable reluctance transducers. These transducers may be designed
to be single-sided, i.e., unbalanced, or balanced. Figure 9.5 shows the basic operating
principle of a single-sided electromagnetic transducer.
The magnetic circuit consists of a rigid half-toroidal high permeability core, a
permanent magnet, an air gap or slit, and a flexing or hinged armature. The current i
drives the magnetic circuit by a coil of N turns wound on the core. The flux contribu-
tions of the coil and magnet do not saturate the core so the small change induced by
i is quasi-linear. The cross-section area of the slit Sslit is the same as the cross-section
area Sm of the core, magnet, and armature. The slit has a width dslit if there would be
no magnetic flux. The slit is so small that flux leakage around it can be neglected.
With the magnet in the core, the slit is d−x.
x MMF = M + Ni
dslit
ZM
Rc
N turns
MMF
i
Rm
Sm
dm
FIGURE 9.5 A magnetic circuit driven by a permanent magnet M and a coil of N turns that
carries an electric current i. The circuit also consists of a rigid half-toroidal core, a slit, and a
flexing or hinged armature. The magnetic field tries to minimize the air gap opening dslit. The
stiffness of the mechanical impedance Z M counteracts the magnetic attraction force.
The flux in the magnetic circuit is due to the magnetomotive force MM induced by
the current in the coil and that of the magnet M.
M m = M + N i (9.8)

Since the reluctance of the air gap Rslit is much larger than those of the core Rc and
magnet Rm
dslit − x
Rslit =
µ 0 Sslit
dmagnet
Rm = (9.9)
µ mµ 0 Smagnet
dm
Rc =
µ cµ 0 Score
it will limit the flux and, since leakage is neglected, the flux density B in the slit will be
Φ
B= (9.10)
Sslit
where Φ is
MM M + Ni
Φ= ≈ µ0 Sslit (9.11)
Rm + Rslit dslit − x
As discussed in Appendix C the magnetic energy W in the slit is
B2
W = dslit Sslit (9.12)
2µ 0
The energy required to change the slit distance must be equal to the change in stored
energy due to change in slit volume
dx
Fm dx = W (9.13)
dslit
This gives the attraction force Fm as a function of the flux
Φ2
Fm = (9.14)
2µ 0 Sslit
which leads to
2
Fm = µ 0 Sslit
( M + N i ) (9.15)
2
2 ( dslit − x )

The dynamic force F is obtained after differentiation with respect to u and i as
MN M2
F ≈ µ 0 Sslit i + µ 0 Sslit u (9.16)
d slit
2
jω d slit
3

Consequently, the transformation factor KEM is
MN
K EM = µ 0 Sslit (9.17)
d slit
2

and the dynamic force can be written simpler as
M K EM
Fm = K EM i + u (9.18)
N jω d slit

After introducing the mechanical impedance load Z M in the circuit, using Equation 9.1,
and assuming any displacement x to be small compared to the slit opening dslit, one
finds that the characteristic force equation is
 MK EM 
Fm = K EM i −  Z M − u (9.19)
 jω dslit N 

where reactance due to the negative compliance CME because of the magnetic field
in the air gap is given by
1 MK EM (9.20)
− =−
jωC ME jωdslit N
For the transducer to be stable, the spring compliance in Z M must always be smaller
than the negative compliance caused by the magnetic field, or the air gap will shut
itself to minimize the stored energy in the system.
Now assume that the armature is moving with a displacement x. Because of the
changing slit opening, the flux will change and there will be a voltage induced in the
coil. This voltage is
∂Φ
e=N (9.21)
∂t
Differentiating Equation 9.9 with respect to the current i (assuming it small com-
pared to the static current I0) and the displacement x (assuming it small compared to
the slit opening dslit ), we obtain
N2 ∂ i N 2 I ∂x
e = µ 0 Sslit + µ 0 Sslit 2 0 (9.22)
dslit ∂t dslit ∂t
The inductance LE0 of the coil when the armature is kept still is
dΦ N 2 Sslit
LE 0 = N = µ0 (9.23)
di dslit
If this expression is used in Equation 9.22 along with the Equation 9.17 for the
transformation factor K EM we find that the second characteristic equation can be
written as
e = jωLE 0 i + K EM u (9.24)

since u = dx/dt. We note that the equations are very similar to those of the electrody-
namic transducer. Eliminating u from Equation 9.24 by using Equation 9.19 we find
that the electric input impedance ZET = e/i can be written as
2 1 2 1
Z ET = jωLE 0 + ( Bl ) = jωLE 0 + ( Bl ) (9.25)
1 1 1
ZM − +
jωC ME YM jω ( −C ME )

LE0 KEM : 1 –CME YM

i F
u
e
FIGURE 9.6 An electrical analogy for the electromagnetic transducer. Note that the sec-
ondary side of the transformer is a mechanical mobility analogy, and that there is a negative
compliance in the circuit due to the attraction force of the magnetic field in the air gap.
Note that the transformed secondary side impedance corresponds to that of a

parallel coupled circuit of two mechanical mobilities. A simple analogy that
fulfills the requirements by the characteristic equations 9.19, 9.24, and 9.25 is
shown in Figure 9.6.
9.5.2 Balanced Transducers
In the preceding derivation of the characteristic equations for the unbalanced
electromagnetic transducer, nonlinearities were neglected in the analysis. By designing
the electromagnetic transducer as a balanced system, usually called variable
reluctance, as shown in Figure 9.7, the nonlinearities may be reduced considerably.
The transducer can be considered as assembled of two unbalanced transducers
side by side sharing a common armature. The static magnetic flux Φ0 is provided
by the two permanent magnets. A fixed coil surrounding the armature is driven
by the electric current i from an external electric source and provides the dynamic
magnetic fluxes Φ1 and Φ2. Normally, the magnetic field path lengths dm, magneto-
motive forces M, reluctances Rm, and cross-section areas Sm are the same for the two
dm Rm Sm 0+ 1
dslit + x
N Sslit
M S Rslit, 1
M
N Rslit, 2
S Sslit
i
dslit – x
dm Rm Sm 0+ 1
FIGURE 9.7 An electromagnetic transducer that uses the balanced system shown here is
usually called a variable reluctance transducer. It can be considered as assembled of two
unbalanced transducers side by side sharing a common armature.
core sections to obtain the best balanced action. The air gap reluctances are much
larger than any other reluctances in the circuit.
The magnetic fluxes Φ1 and Φ2 are determined by the magnetomotive forces
provided by the magnets, M, and by the current i in the N turns of the coil.
2 MSslit
Φ0 = µ 0
2dslit
N i Sslit
Φ1 = µ 0 (9.26)
dslit − x
N i Sslit
Φ2 = µ 0
dslit + x
The forces on the two sides of the armature are then (see Equation 9.14)
Fm,1 =
( Φ0 + Φ1 )
2µ 0 Sslit
(9.27)
2
Fm,2 =
( Φ0 − Φ2 )
2µ 0 Sslit
Assume the motion of the armature being blocked, i.e., x = 0 so that the fluxes are
Φ1 = Φ2 = Φi. The net force Fm acting on the armature will then be
2 Φ 0 Φi 2 Φ 0 N
Fm = Fm,1 − Fm,2 = = i (9.28)
µ 0 Sslit dslit
The transformation factor is thus twice that of the single-sided transducer
2Φ0 N
K EM = (9.29)
dslit
One can show that the negative compliance is [2]
µ 0 Sslit dslit
C ME = (9.30)
2Φ0
and the coil inductance is
N 2 Sslit
LE 0 = 2µ 0 (9.31)
dslit
Obviously, the circuit analogy will be the same as that in Figure 9.7 except for the
values of K ME , CME , and LEO. Note that CME must always be smaller than CM in Z M
for the armature not to fall into one of the pole pieces due to random unbalance in
the system.
Another type of balanced electromagnetic transducers is the rocking armature
design. The advantage of the fully balanced rocking armature type design is that
the moving system can be made mass rather than compliance controlled. The unbal-
anced moving magnet design must be compliance controlled. However, the balanced
design also runs the risk of “pole over,” that is, a pole of the armature becomes per-
manently attached to a static pole.
9.6 ELECTRORESISTIVE TRANSDUCERS

Electroresistive transducers work by the principle of “modulated flow” and can be
made using both direct and indirect conversion. As mentioned initially, such trans-
ducers are generally considered active and unidirectional. They are typically used
as sensors since the electroresistive principle has low conversion efficiency. Much
outside power is needed but little signal power is generated.
Examples of active “resistive” transducers are those where an acoustical or
mechanical field is modulating the electric current supplied by an outside power
source, for example, carbon and piezoresistive microphones. The function of a
single “button” carbon microphone is based on dynamic compacting of elec-
trically conductive granules between two electrodes in a cavity, as shown in
Figure 9.8a. Figure 9.8b shows a double button design that reduces the nonlinear
distortion. The bandwidth is also enhanced due to the stretched diaphragm and
the control of the diaphragm resonances using the viscous damping provided by
the damping slits.
Since the granules are only loosely packed, variations in the degree of compac-
tion will cause the resistance of the circuit to vary and thus a modulated DC current.
The compaction of the granules is a nonlinear process, which means that the nonlin-
ear distortion of the carbon microphone is high consisting both of granulation noise
and conventional nonlinear distortion. The granulation noise, often called carbon
noise, is heard as signal “roughness.”
The electrical circuits for the microphones shown in Figure 9.8a and b are
shown in Figure 9.9a and b. The nonlinear distortion in the double button carbon
microphone is also reduced because of the removal of net DC current in the
transformer.
Silicon piezoresistive transducers do not suffer from these drawbacks. Silicon
piezoresistive transducers are manufactured by anisotropic etching and doping of the
semiconductor crystal to obtain a thin diaphragm that bends by sound pressure and
has monolithically integrated piezoresistors. The size and thickness (5–50 μm) of
the diaphragm is chosen to fit the desired working range. The basic design of such a
microphone is shown in Figure 9.8c. The silicon diaphragm is deformed by the sound
pressure and in turn deforms the piezoresistive sensors at the edge of the diaphragm.
The conductivity of the piezoresistor is influenced by the mechanical change of the
semiconductor crystal lattice. The diaphragm resistor is connected, using aluminum
Flexible edge Moving diaphragm

suspensions
Fixed electrode Carbon granule

(a) filled cavity (button)
Stretched diaphragm Acoustically open

electrode holder
Fixed electrode
Fixed electrode Damping slits
Carbon granule
(b) filled cavities (buttons)
Piezoresistor Metallization
Silicon oxide
Silicon
Silicon oxide
(c) Silicon diaphragm
FIGURE 9.8 Examples of microphones operating using a sound-modulated electric

resistance: (a) unbalanced using carbon granules, (b) balanced using carbon granules, and
(c) using piezoresistors.
E0
N:1
REM e
RE REM
(a)
N1:1 = N2:1 E0
REM RE + N:1
REM1 E0 A
–
e
e
REM2
(b) (c)
FIGURE 9.9 Three examples of simple electrical circuits for electroresistive microphones
and other electroresistive sensors: (a) single “button,” unbalanced, (b) double “button,”
balanced, and (c) balanced bridge circuit with amplification for silicon piezoresistive
micromachined microphone.
conductors, in a bridge circuit to compensate for temperature-induced drift of

resistivity as shown in Figure 9.9c with the help of a circuit integrated on the silicon
wafer. The circuit feeds the sensors with a temperature-controlled constant current
and incorporates a bridge amplifier. Because of their relatively low sensitivity,
piezoelectric microphones are only used to measure high sound levels. Since many
transducers can be incorporated on a single piece of silicon, the signal-to-noise ratio
can be improved by summing the transducer signals.
The electrical output e of the microphone is proportional to the compacting of
the carbon granules or the elongation of the piezoresistors and consequently to
diaphragm displacement, i.e.,
e = K EM ξM (9.32)

where
KEM is the transformation factor determined by the microphone design
ξM is the dynamic displacement of the diaphragm from its rest position
The electroacoustical analogy circuit of a electroresistive or piezoelectric trans-

ducer will be determined by its mechanical elements, as shown in Figure 9.10a,
since there is in principle no reaction from the electrical side except for heating
of the resistive material. Both microphones will have compliance and mass in the
mechanical circuit. The carbon granulate has mass and compliance. The granulate
must be contained by a compliant membrane or similar. The compliance and mass
of the silicon diaphragm will be determined by it geometry and material constants.
Typically the acoustical circuit will be limited to mass and compliance in series
with a resistance in the acoustical impedance analogy, as shown in Figure 9.10c.
Controlled flow transmitters may of course also be designed; one example is the
electropneumatic loudspeaker. In this type of loudspeaker, the flow of gas from a
pressurized reservoir is modulated using a needle valve coupled to a horn. Using an
ignited butane–oxygen gas mix, very high acoustic powers may be generated. Also,
sirens can be regarded as controlled flow transmitters although in this case the signal
generation is built into the device itself and it should therefore not be called a transducer.
FM FM ZAR MA
MM CM rM MM CM rM p CA
RA
(a) (b) (c)
FIGURE 9.10 (a) Mechanical circuit for an electroresistive microphone (CM includes air
cavity compliance). (b) The electromechanical mobility analogy. (c) The electroacoustical
impedance analogy including the radiation impedance as seen from the microphone
diaphragm.
9.7 CAPACITIVE TRANSDUCERS

There are many types of transducers based on the use of variable capacitance.
Generally, capacitive sensing transducers will be microphones and transmitting
transducers will be loudspeakers.
Microphones based on the use of capacitance are usually called condenser
microphones. They operate by the distance-induced change of capacitance between
two statically charged moving electrodes. Capacitor operation is nearly ideal for
microphones, since it offers low vibration sensitivity, high acoustic sensitivity,
sufficient linearity, and relatively low noise. Most condenser microphones use
an unbalanced mode of operation and an unbalanced electrical circuit. Typically,
these microphones are designed to have low mass, tautly tensioned diaphragms, so
that the microphone diaphragm works in a compliance-controlled mode over the
entire frequency range.
Loudspeakers based on the use of capacitance are usually called electrostatic
loudspeakers. They are based on the change of attraction force between two statically
charged electrodes when the voltage between the electrodes is changed. Virtually all
commercially available wide frequency range electrostatic loudspeakers for audio
use the balanced approach, i.e., push–pull operation, to accommodate the need for
larger diaphragm movement. Because of the small forces available using electrostatic
transducers, their use is limited to low-impedance applications.
Loudspeakers using unbalanced operation (“single-sided”) are primarily used
for specialized applications in very-high-frequency audio and ultrasonics where the
transduction nonlinearities are of less importance.
The principle of balanced and unbalanced transducers is shown in Figure 9.11.
Neglecting the influence of capacitor and membrane edge effects as well as the
holes of the static electrodes, the electric capacitance CE between two electrically
conductive plates at close distance is
εr ε 0 S
CE = (9.33)
d
where
εr and ε0 are dielectricity constants
S is the area facing each electrode
d is the distance between the electrodes
x x
Moving electrode diaphragm d0
d0 d0
(a) Static electrode (b)
FIGURE 9.11 Electrostatic transducers having movable diaphragms: (a) unbalanced

(“single-sided”) and (b) balanced.
As can be seen from Equation 9.33, the capacitance will vary inversely propor-
tional to the distance variation between the plates. For small distance variations
ξ about the static distance d0, the capacitance variation C∼ about the static capacitance
C0 will be
ξ
C~ ≈ C0 (9.34)
d0
The capacitance variation can be used in several ways to produce an electric output
signal proportional to the sound pressure.
Since the moving capacitor electrode diaphragm has mass and compliance, it is
convenient to symbolize its mechanical properties by a mechanical impedance Z M.
In its simplest form the impedance will behave as a damped mass-spring system. We
can see the transducer as a system composed of a mechanical side, a transduction
mechanism, and an electrical side.
9.7.1 Direct Use of Capacitance

Typically the direct use of the capacitive transducer involves the capacitance in a
resonant electrical circuit with an inductance L. The resonance frequency f will vary
along with the capacitance variation C∼ as
1
f = (9.35)
2π L (C0 + C∼ )

The capacitance variation can be used either directly to change the resonance
frequency of an oscillator or in an envelope detector circuit using a fixed oscillator
frequency. The resonance frequency must be much higher than any frequency that is
to be sensed. The envelope detection process can be thought of as similar to a sam-
pling procedure. Some alternatives are shown in Figure 9.12.
The first approach, shown in Figure 9.12a, requires demodulation of the frequency-
modulated signal, which can be achieved by simple circuits, for example, a pulse
counter or integrator. The main drawback of this approach is the random noise of the
oscillator, due to its limited Q-value, that will cause noise in the demodulated output
signal. It is difficult to achieve sufficient “pull” of the frequency by the capacitance
change using crystal resonance circuits so conventional inductance–capacitance
circuits need to be used.
In the second approach, shown in Figure 9.12b, a fixed-frequency low-noise
sine wave oscillator is used, for example, with some form of electronic bridge
circuit. A third possibility, shown in Figure 9.12c, is to use the condenser micro-
phone capsule as a component in an LC-circuit that has high Q and that has a
resonance frequency, which is somewhat offset relative to the sine wave genera-
tor’s frequency. This is an attractive method when the capacitance variations are
very small.
CE Low noise Pulse

e
oscillator generator
Frequency Integration
(a) modulator (pulse rate
demodulator)
CE
Low noise N :1
crystal
oscillator
e
Bridge circuit Rectification

(b) amplitude modulator and integration
(amplitude
demodulator)
Low noise CE
crystal e
oscillator
Frequency offset resonant circuit Rectification

amplitude modulator and integration
(amplitude
(c) demodulator)
FIGURE 9.12 Some alternatives for using condenser microphone capsules to generate an
audio output voltage. (a) Pulse rate modulation with an integrator pulse “counter” demodulator.
(b) Amplitude modulation using a capacitive bridge circuit and an amplitude demodulator.
(c) Frequency modulation with a high-Q resonant circuit and amplitude demodulator.
9.7.2 Capacitive Microphones Using Static Electric Charge

with Externally Supplied Electrical Bias
9.7.2.1 Unbalanced Operation, Static Conditions

A condenser microphone uses a moving electrode membrane close to a static
electrode as shown in Figure 9.11a to form a capacitor. In a traditional condenser
microphone, the electrodes are provided a static electric charge. Because of the
charge, any change of capacitance from its static value will result in a change in
Polarization
voltage supply
Bias resistor High input impedance amplifier
N:1
A
e Audio signal
Diaphragm output
DC blocking
capacitor
Back electrode
FIGURE 9.13 Basic elements of a traditional condenser microphone using high voltage bias.
voltage between the two electrodes since for a capacitance the following relationship
applies between capacitance CE , charge QE , and voltage E:
QE = ECE (9.36)
The electric charge can be supplied by a “bias” voltage in series with a “bias” resistor,
as shown in Figure 9.13.
The bias resistor R B needs to be large, so that the electrical time constant of the
microphone, the product of the resistance and the capacitance of the microphone
capsule, is large. The charge QE over the capacitor must be virtually constant
over more than one cycle of sound pressure variation. For audio frequency range
condenser microphones, the time constant is typically chosen to be larger than 1 s.
Fast variations in capacitance, such as those due to audio sound pressure variations,
will—because of the constant electric charge—lead to voltage variations that follow
the sound pressure variations.
The electrostatic force due to static charge tries to attract the electrodes, and it
is necessary to balance this attraction force by a static mechanical retaining force.
The microphone capacitance only behaves as a “pure” capacitance CE as long as
the electrodes are kept static. In its biased mode, the capacitance will have a value
different from CE since the attraction force of the bias charge will suck the mov-
ing electrode toward the static electrode until the attraction force is balanced by the
spring force of the compliance CMD holding back the movable electrode. The critical
distance x0 is that at which both forces are equally large but opposed.
For a microphone, the electrostatic force will be quite small compared to the
mechanical force since the diaphragm needs to be tautly tensioned to make its
fundamental resonance frequency to be at the upper end of the audio spectrum.
This is necessary to give the pressure-sensitive condenser microphone a fre-
quency-independent response, limited at the high frequency end by the diaphragm
diameter.
To find the critical distance, we start by studying the attraction force FE ,

which is [3]
CE E 2
FE = (9.37)
2d
If the movable electrode is moved a distance x toward the solid electrode, the retain-
ing spring force will be a compliance:
x
FC = − (9.38)
C MD
Assume that the diaphragm is moved a distance x toward the solid electrode (i.e., in
the negative x-direction). One finds [3] that static force balance FC = FE is
−2
−x εε S x
= E02 r 02  1 +  (9.39)
C MD 2d  d 
This equation can be solved numerically or graphically, as shown in Figure 9.14.

Because of the nonlinear attraction and linear mechanical retaining forces,
there is a maximum distance that one cannot allow the electrode to exceed since
beyond some limit x0 the movable electrode will hit the static electrode. This limit
is given by FC = FE , so, for stability, FC > FE , which results in the requirement for
stability being
− x0 1 (9.40)
≤
d 3
as shown in Figure 9.14. The negative compliance due to the electrical attraction
force must always be larger than the mechanical retaining force for stable operation.
14 Electrical attraction force

Mechanical retaining force
12
Region of stability
10
Force (a. u.)
6
Increasing bias voltage
4
2 Stationary
electrode
0
–1.0 –0.5 0.0 0.5 1.0
Diaphragm displacement (x/d)
FIGURE 9.14 Electrostatic and mechanical forces in the graphic solution of Equation 9.39.
9.7.2.2 Unbalanced Operation, Dynamic Conditions

Obviously, since the mechanical and electrostatic forces are interrelated, we expect
the compliance and capacitance to appear jointly in the transducer’s electrical circuit
analogy. For linearity, the microphone diaphragm will only make small excursions
ξ compared to the distance between the diaphragm and the back electrode d0 so
that ξ ≤ d0. A positive change of ξ as shown in Figure 9.11 decreases the capacitance.
Characteristic for the microphone is (1) that the voltage between the electrodes
can change both due to incoming electric charge (i.e., electric current) and change of
geometry (i.e., electrode distance) and (2) that the membrane movement is due both
to mechanical force (sound pressure) and to changes in electric charge (i.e., electric
current). The membrane velocity u is
u = jω ξ (9.41)

Since the electric charge Q 0 on the capacitor remains constant due to the bias circuit,
the audio voltage e′ between the electrodes must change as the membrane vibrates
with velocity u. Using Equation 9.36 and setting
Q0 Q0 Q  ξ 
E0 + e ′ = = ≈ 0  1 +  (9.42)
CE 0 + CEA  ξ  C E 0  d0 
CE 0  1 − 
 d0 
we obtain the movement-induced audio voltage as
Q0 ξ E
e′ ≈ = 0 u (9.43)
C E 0 d 0 j ωd 0
If there is a dynamic change of charge, i.e., an audio current i enters the microphone,
the voltage will change by e″ since the microphone is a capacitor:
1
e′′ = i (9.44)
j ωC E 0
The total audio voltage will be the sum of the two contributions
1 E
e= i + 0 u (9.45)
j ωC E 0 j ωd 0
The membrane velocity is due to the sum of forces, the mechanical force due to
the differential sound pressure acting on the membrane and an attraction force due
to the dynamic change of electric charge. The mechanical impedance determines
the velocity u resulting from these two forces. The mechanical force F′ is
F ′ = pS D (9.46)

The “dynamic” force F″ due to the dynamic change of electric charge q_ is obtained
from the equation for the attraction force between two charged electrodes
2
F + F ′′ =
(Q 0 +q ) ≈
Q02
+
Q0 q
(9.47)
0
2CE 0 d0 2CE 0 d0 CE 0 d0
Since current and charge are related as

i = jωq (9.48)

the force is
Q0 q E0 q E0
F ′′ ≈ = = i (9.49)
C E 0 d0 d0 j ωd 0
The equation of motion for the membrane then becomes

E0 1
pS = − i+ u (9.50)
j ωd 0 jωC MD
The two equations 9.45 and 9.50 are the characteristic equations of the single-sided
electrostatic transducer. The electrostatic transformation factor K ME is
E0
K ME = (9.51)
j ωd 0
Setting the current i to zero and combining Equations 9.43 and 9.51, one finds that
the transformation factor can also be
E0C MD
K ME = (9.52)
d0
It is now time to investigate how the properties of various electroacoustical analogies

relate to these equations. In a traditional condenser microphone, the electric imped-
ance seen by the microphone is extremely high so little current will be drawn from
the microphone. A simple analogy that fulfills the requirements by the characteristic
equations when i = 0 is the one shown in Figure 9.15. In this circuit the negative
mechanical compliance due to the electrostatic attraction has been merged with the
mechanical membrane compliance into one component.
E0CMD
1: CE0
u d0
pS CMD e
FIGURE 9.15 Electroacoustical analogy for the condenser microphone.

9.7.2.3 Capacitive Microphones Using Permanent Charge (Electret Bias)

A condenser microphone needs the energy stored in an electric field to function.
The method of charging the microphone by feeding the microphone electric charge
from a battery or other source described in the previous section is impractical from
several points of view. Primarily, it is the noise that comes from leakage current over
the isolator that poses a problem.
The charge can however be supplied to the microphone as a permanently
charged dielectric film, either as a membrane or as a layer on top of the static
electrode. The material used for electret films must have poor electric conductivity.
Monopolar electret films can be made by injecting charge into a material by bom-
barding it with ions or electrons; this is the type of electret most often used in
microphones. Another approach is that of thermal processing of a film having
electric dipoles at random. In this method, the film is subjected to a strong electric
field during the heating and cooling off processes. The dipoles can then be aligned
to form a heteropolar film. Static charge can also be injected during the heating
process leading to homopolar films; this process is used for film materials that do
not have dipoles [4,5].
When the film is to be used for a membrane, it is typically metallized ahead of its
polarization. A problem with electret films is that they cannot be under as much ten-
sion as metal foils, so the resonance frequency will not be as high, and consequently
the frequency response will not be extended toward higher frequencies as well as for
metal film condenser microphones having the same diameter membranes. When the
electret is attached to the static electrode, the membrane still can be made from a
metal foil and the mentioned frequency restrictions do not apply.
The electroacoustical analog circuit for the electret condenser microphone is the
same as that of the electrically biased microphone shown in Figure 9.15 [4]. Because
the capacitance of electret microphones is about the same as that of conventional
condenser microphones, they also need a closely mounted impedance converter
circuit. In most low-cost capsules, the impedance converter is located behind the
static electrode inside the microphone capsule.
9.7.3 Loudspeaker Operation
9.7.3.1 Unbalanced and Biased
Because of the attraction force between two charged electrodes, the transducer
discussed in the previous section can also be made to operate as a transmitter,
i.e., as a loudspeaker. Some major advantages of the electrostatic loudspeaker over
other loudspeaker types is that it is flat, thin, and can be given arbitrary shape since
the driving force is equally distributed over its surface so that diaphragm modes are
less excited.
Since the attraction force FE is independent of the polarity of the charge, an
alternating voltage applied to the transducer will result in a “rectified” force [3]. In
some transmitter operations such as when simply generating a stable high-frequency
tone using a resonant transducer, the nonlinearity may in fact be an advantage from
an electronic viewpoint.
Moving electrode diaphragm

Static electrode
FIGURE 9.16 Sell-transducer uses a diaphragm that lies on top of the undulations or
irregularities of the static electrode.
A type of single-sided electrostatic transducer common in ultrasonic applications

is the Sell-transducer shown in Figure 9.16. Here, the backplate is roughened or
micromachined to give air pockets. The compliance of the air in the pockets, together
with the membrane mass, forms a resonance in the desired frequency region. Because
the membrane lies directly on top of the solid electrode, the membrane foil needs to
have high isolation capability and good electrical breakdown properties. Note that
either the static electrode or membrane electrode may be connected to ground with
no change in operation. This choice is made from personal security considerations.
For linearity, the transducer needs to be supplied with a static polarizing voltage bias
E 0 on which the electrical signal e is overlaid. Naturally, this linearization is effective
only as long as E 0 ⋙ e.
With the bias voltage applied as shown in Figure 9.17, the force FE acting on the
transducer is given by Equation 9.37. The electric current IE into the transducer is
related to the dynamic force F″ by Equation 9.49. The ratio (E 0 /jωd0) is the electro-
static transformation factor studied previously. The use of the Sell transducer as a
loudspeaker for ultrasound is further discussed in Chapter 25.
In contrast to the condenser microphone, the audio frequency electrostatic loud-
speaker is usually designed so that the fundamental membrane resonance frequency
is at the low end of the desired audio range. Typically, the transducer resonances will
be well damped by the radiation impedance load. Since the resonance frequency will
be comparatively low, the mechanical compliance needs to be high, so there is a risk
that the negative electric compliance described previously will be so high that the
static conditions may become unstable and the membrane permanently drawn into
the static electrode.
Amplifier bias
voltage supply
Power amplifier having AC blocking

circuit with DC offset output inductance
A
Audio input
Membrane
Back electrode electrode
FIGURE 9.17 Simple circuit for a single-sided electrostatic loudspeaker (“singing condenser”).
9.7.3.2 Unbalanced Electret Charge

The static electric charge needed for the operation of the transducer can in this case be
provided to any of the two electrodes. Typically, the charge will be supplied by an elec-
tret film. The film can be used both as a membrane, after metallization, or as a layer
on the back electrode. The transducer can then be used in the same way as capacitive
transducers using outside electrical bias to submit the necessary static electric charge.
9.7.4 Transducer Electrical Analogy

The electric input impedance of the loudspeaker is characterized primarily by
electric capacitance since the coupling between the electrical and mechanical sides is
fairly low. The transformation factor K ME is given by Equation 9.52. The impedance
analogy of the single-sided transducer as a loudspeaker is shown in Figure 9.18. Note
that the mechanical compliance is altered because of the presence of the electrostatic
force so that the effective mechanical compliance becomes
CM
C ′M = (9.53)
1 − K ME
2
9.7.5 Loudspeakers Using External Electric Charge, Balanced

Since loudspeakers need to generate high volume velocity, it is necessary to design prac-
tical electrostatic loudspeakers using push–pull type (“balanced”) permanent charge
designs. The force F on an electric charge q in an electric field of strength E is given by
F = Eq (9.54)

The electric field is applied between the two static electrodes as shown in Figure
9.11b. The membrane is symmetrically positioned between the electrodes. The static
charge is applied to the membrane (moving electrode). By locally changing the
distribution of the charge over the diaphragm, or the distance between the static
electrodes, the force at each point of the diaphragm can be adjusted to give the loud-
speaker the desired radiation pattern. The charge needs to be static, which requires
that it must either be geometrically locked on or in the diaphragm, as, for example,
in an electret, or the diaphragm must have a very high surface resistivity so that the
E0CE0
1: CM MM RM
d0
ZTE CE0 ZRAD
FIGURE 9.18 Electric input impedance Z TE to a single-sided electrostatic loudspeaker.

Here CE0 and d 0 are the effective values with the polarization voltage applied.
Diaphragm bias Symbolic only

charge supply resistance distributed
evenly over diaphragm
Power amplifier Power amplifier
E/2 –E/2
A A
Audio input Back Back
electrode 1 electrode 2
Diaphragm center
electrode
–1
Phase inverter amplifier
FIGURE 9.19 Balanced drive electrical circuit for a balanced electrostatic push–pull
loudspeaker. The balanced drive could also be accomplished by using a transformer with a
center-tapped secondary winding.
charge distribution on the diaphragm is not altered during the period of the signal.
A possible electrical drive circuit for a balanced electrostatic loudspeaker is shown
in Figure 9.19.
The problem of static stability in the balanced transducer is similar to that of the
unbalanced transducer discussed earlier [3]. Because of the essentially symmetric
envelope of audio signals, the diaphragm needs to be centered in the air gap between
the two electrodes since this will minimize the risk of the diaphragm crashing and
distortion in the sound reproduced by the loudspeaker. The membrane can be cen-
tered either mechanically or electrically. Mechanically, the membrane may be ten-
sioned to make it remain in an average static position or suspended by “spacers” at
suitable intervals (this will be necessary if the electrodes are curved).
In analogy with Equation 9.39 one finds that the limiting distance x0 for the
balanced transducer is found from the solution to the equation
−2 −2
−x ε ε S  x  x 
= E02 r 02   1 +  − 1 −
 d   (9.55)
C MD 2d   d  

The analysis in Ref. [3] shows that the static limiting distance is the same as that
for the single-sided case (but of course symmetrical around x = 0). The region of
stability is shown in Figure 9.20 and is limited by the two points at which the slope
of the electrical force curve equals that of the mechanical force determined by the
membrane tension.
However, for dynamic conditions, that is, when the charge remains constant on the
membrane, there will not be any instability and the membrane can move well outside
the limit given by Equation 9.40. If the charge is allowed to move on the membrane, for
example, by using an external resistance and a conductive membrane, the analysis does
not hold and the charge may move locally to cause nonlinear distortion and arcing.
6 Electrical attraction force

Mechanical retaining force
4
2 Stationary
Force (a.u.)
electrode
0
Stationary
–2 electrode
4
Region of stability
6
–1.0 –0.5 0.0 0.5 1.0

Diaphragm displacement (x/d)
FIGURE 9.20 Electrostatic and mechanical forces acting on a push–pull electrostatic loud-
speaker that does not have constant charge operation.
E0CE0
1: CM MM RM
d0
ZE CE0 ZRAD
2
FIGURE 9.21 Electric input impedance ZE to a push–pull electrostatic loudspeaker. Here,

CE0 and d 0 are the membrane to static electrode values assuming symmetry since there is no
static displacement of the membrane (see Figure 9.11).
The electrical analogy of the push–pull type electrostatic loudspeaker is shown

in Figure 9.21. The electric input impedance between the two static electrodes is
similar to that of the single-sided loudspeaker with only small differences since in
this case there is no static displacement of the membrane because of the symmetry
of the operating mode.
9.8 PIEZOELECTRIC TRANSDUCERS

A piezoelectric effect is characterized by the property that a force applied to the
material results in electrical charge distributed on the surfaces of the material.
The polarity of the charge on a surface will depend on the direction of the force.
The effect is reciprocal, i.e., when electric charge is applied to the material, the
material deforms. This is called the reverse piezoelectric effect. Examples of
piezoelectric materials are quartz and other crystals such as Rochelle salt and
some lithium salts. The crystals need to be single crystals. Ceramics such as lead
zirconate titanate exhibit a much stronger piezoelectric effect. There are also
polymers such as polyvinylidene fluoride that have piezoelectric properties, which
are discussed in a later section. The strong coupling between mechanical stress
and electric charge makes materials having piezoelectric properties attractive for
electromechanical transduction. Due to the small mechanical losses in many of
the piezoelectric materials, they find applications in resonators of various types,
such as clock oscillators and electronic filters. The high internal impedance of
simple transducer designs using such materials makes them useful for mechanical
and hydraulic applications. The piezoelectric effect is fundamentally linear within
a wide range of deformation.
Wideband applications are generally reserved for the audio range where the
working frequency range can be kept below the fundamental resonance frequency
of the transducer. Since the effect is due to mechanical stress, piezoelectric trans-
ducers in principle have response to DC frequency, provided that the charge
developed does not leak from one area to another. Due to the relatively small
extensional changes (in spite of the strong coupling) that can be had, it is com-
mon to apply piezoelectric materials so that a lever effect, and larger movement,
is obtained.
9.8.1 Piezoelectric Coupling in the Quartz Crystal

The mechanical deformation can have the form of extension or shear of the
material. Many materials are characterized by concurrent deformation along sev-
eral axes. By suitable shaping of the material one can achieve movement primarily
along a single axis. The piezoelectric device can then be characterized mechani-
cally mainly by wave motion along one axis and for frequencies well below the
first resonance by compliance and mass. At acoustic frequencies, the device will be
characterized by capacitance and series and parallel resistances.
To understand the operating principle of the piezoelectric materials it is useful to
study the properties of a quartz crystal as an example. Figure 9.22a shows a sketch
of a quartz crystal, its associated crystal axes. The x-axis is sometimes called the
electrical axis and the y-axis the mechanical axis of the quartz crystal. When a rect-
angular piece of the crystal is cut as shown in Figure 9.22b, this is called an x-cut bar.
Generally, nine different constants and equations are needed to describe the
electromechanical action of piezoelectric materials. Some ways of cutting samples
from the crystal require less equations and constants to obtain a useful description
of the crystal’s piezoelectric properties. The x-cut longitudinal bar is such a cut.
The x-cut longitudinal bar is typically used in the range 5–500 kHz when large
amplitudes are needed. The derivation that follows is based on Ref. [7].
9.8.2 Electromechanical Relationships

Because of the coupling between the mechanical and electrical properties of quartz,
the force/velocity relationship will be coupled to the voltage current relationship of
the piezoelectric material. The bar can be acted upon by both mechanical forces and
electrical fields so both will determine its dynamic behavior.
Assume the piezoelectric bar shown in Figure 9.22b is fastened at one end to a
rigid surface at y = 0 and that a dynamic force Fy in the negative y-direction is applied
to the side at y = ly.
z
lx
lz
x
y
ly
y
(a) (b)
FIGURE 9.22 (a) Definition of quartz crystal axes and cutout needed for. (b) X-cut quartz
bar of piezoelectric material with the side y = 0 fastened to a rigid surface at y = 0 and that has
its sides at x = 0 and x = lx metallized. The bar has such piezoelectric properties that a voltage
applied to the metallized sides mainly causes a deformation of the bar along the y-axis. (After
Kinsler, L. E. et al., Fundamentals of Acoustics, John Wiley & Sons, New York, 1962.)
If the crystal was not piezoelectric, it would compress by a distance Δy as a result

of the applied force F, just as a simple spring:
∆y
F= (9.56)
CM
Here, CM is the mechanical compliance of the spring. It is however more practical

to use the following relation for the longitudinal strain ∂η/∂y when developing the
theory of wave propagation and resonance in the crystal:
∂η F y (9.57)
= − s22
∂y Sy

where
η is the longitudinal displacement in the positive y direction
s22 is the compliance coefficient
Fy is the compressional force on the ends of the crystal
Sy is the surface area of those ends, Sy = lxlz.
If, on the other hand, the bar was only a capacitor, the charge over it would be
q = eCE 0 (9.58)

The charge q is the product of the capacitance CE0, and the voltage e across the
capacitor. A thin flat capacitor with metallized surfaces lylz at x = 0 and x = lx has a
capacitance
ε 0 ε x l y lz
CE 0 = (9.59)
lx
where
ε0 is the dielectric constant of vacuum
εx is the dielectric constant of the bar material in the x-direction
One can also express the charge as a charge per unit area σx.
ε0ε′x
σx = ex (9.60)
lx
Now assume that a force is applied to the quartz bar over the ends at y = 0 and y = lx
and that a voltage ex is applied between the sides at x = 0 and x = lx. The surface charge
density σx will be due both to the capacitance of the bar and of possible deformation
along the y-axis, as given by
ex F
σ x = ε0ε x − d12 y (9.61)
lx Sy

The strain along the y-axis, ∂η/∂y, will be due both to the applied pressure on the bar
along the y-axis and on the electric field strength caused by application of the voltage
ex across the bar, between sides x = 0 and x = lx as given by
∂η Fy e
= − s22 + d12 x (9.62)
∂y Sy lx

The charge in a capacitor depends on the applied voltage and the capacitance. It is
more convenient for the discussion that follows to express the electromechanical
relationship in Equation 9.61 by the strain than by the stress Fy/Sy, so one obtains
e x d12 ∂η
σ x = ε0ε′x + (9.63)
lx s22 ∂y
Here, the equation has been simplified by replacing the dielectric constant εx in
Equation 9.61 by the clamped dielectric constant εx′ , i.e., the constant that would be
obtained if it was possible to prevent the bar from moving by rigidly clamping it also
at y = ly so that it cannot move in the y-direction.
 d122 
ε′x = ε x  1 −

(
ε 0 ε x s22 
)
= ε x 1 − kPE
2
(9.64)

The coefficient kPE is a measure of the piezoelectric material’s electromechanical

coupling
d122
kPE = (9.65)
ε 0 ε x s22

similar to the one used for the electrostatic transducer.
9.8.3 Transducer Electrical Analogy

The electric input impedance of the piezoelectric bar is primarily capacitive, but
with the effect of the mechanical resonances and antiresonances showing up in the
impedance curve as discussed in Chapter 25. If we assume that the frequency is
low, we need not take the wave motion of the bar into account, which leads to the
impedance analogy shown in Figure 9.23.
The piezoelectric transformation factor K ME is defined as
d12lx
K ME = (9.66)
s22
The transformation factor and its use are further discussed in Chapter 25.
9.8.4 Piezoelectric Ceramics

Piezoceramics are made up of small crystallite grains that are naturally random
oriented. By heating the ceramic above its Curie point temperature, applying a
strong electrostatic field, and letting the ceramic cool to below the Curie point, the
domains will be oriented in the field direction. The procedure is called poling. The
Curie temperature typically is in the range of 170°C–350°C. The polarization will
remain until as long as the material is not reheated. The electrodes are prepared for
electric connection by applying silver or graphite paint, or a similar material [7–9].
Common piezoceramics are barium titanate (BaTiO3), lead titanate (PbTiO3),
and lead zirconate titanate, “PZT” (PbZrO3). These man-made ferroelectric mate-
rials have many advantages over natural crystals. Physically the ceramics may be
d12 lx
1: CM
s22
u
CE0 e F
FIGURE 9.23 Analogies for a piezoelectric bar clamped at one end at frequencies below
the first length mode.
Transverse expansion
Electrodes
Thickness expansion
2
3
Face shear
Thickness shear 3 1
1 2 Side view Top view
FIGURE 9.24 Some deformation modes of piezoelectric materials.
made into bars, plates, discs, tubes, spheres, or other shapes and poled suitably.
Piezoceramics differ with regard to temperature stability, piezoelectric constants,
and dielectric and mechanical losses. The latter result in Q-values that range from
more than 1000 to less than 100.
Piezoelectric ceramics can be designed to be used in shear, thickness expansion,
or transverse expansion modes, as indicated in Figure 9.24. The piezoelectric
properties are therefore described by piezoelectric, elastic, and dielectric properties
and by various constants. For the piezoelectric charge and voltage constants dij and
gij, the first subscript refers to the direction of the applied field and the second to
the direction of the strain. The electromechanical coupling coefficient is indexed in
the same way. The direction over which the electric field is applied is usually given
index 3 or called the z-direction.
Thus, for example, the relation between applied charge and thickness change is
described by the d33 piezoelectric constant while the relation between applied voltage
and length change is described by the g31 piezoelectric constant.
As with crystals, the resonances of the material due to its shape and edge
conditions are going to be noticeable in the electric impedance characteristic. Since
the piezoelectric may also be printed thinly onto a carrier bar or disc made of a
material that has higher density and Young’s modulus, the resonance characteristic
of the carrier will, in this case, determine the mechanical behavior. Such discs are
used for resonant loudspeakers and beepers. For microphones, the piezoelectric
transducer is usually used below the frequency of the first resonance to obtain a
frequency-independent response.
9.8.5 Piezoelectric Films

The main reasons for the high Q resonant characteristics of piezoceramic bars and
disks are the large vibrating mass and high stiffness. Since the radiation impedance
into air is low, they are not suited for wide frequency range transducers because
of the poor impedance matching that leads to poor sound radiation and reception
properties unless used with a radiation-enhancing diaphragm. Ceramics are frag-
ile and do not lend themselves to the manufacture of thin elements. Matching lay-
ers can be used for impedance conversion but have limited frequency range. These
disadvantages can be addressed by using a piezoelectric film instead of a piezoc-
eramic with an attached diaphragm.
Piezoelectric film is typically made of a polarized fluoropolymer, polyvinyli-
dene fluoride (PVDF), or its copolymers, which can be made to be much more
piezoelectric than polyamide or polyvinyl chloride (PVC). PVDF pellets are melted
and extruded into sheets that are stretched at temperatures well below the melting
point so that thin foils are obtained. This forces the polymer molecules into parallel
crystal planes. The polymer is then exposed to very strong electric fields to polarize
the crystal planes and align their dipoles.
A conductive metal layer is then applied to each surface of the film without making
the edges conductive. The layers can be sputtered, printed, or painted. The metals
used for sputtering are typically nickel, aluminum, and gold in thicknesses of about
50–100 nm. Printing or painting is often done using silver paint. The metallized film
is then provided with electric leads for connectivity and covered on both sides by a
protective film.
Piezoelectric films are used primarily for headphone applications because of the
cost of the film. The voltages required in their operation are similar to those used
for electrostatic transducers with peaks of hundreds to thousands of volts, so for
both types there is the risk of dielectric breakdown and arcing. An advantage of
piezoelectric film transducers over electrostatic transducers is of course that there is
no risk of static or dynamic instability.
9.9 MAGNETOSTRICTIVE TRANSDUCERS

Typically, magnetostrictive transducers take the form of cylindrical magnetostric-
tive cores inside helical electrical coils as shown in Figure 9.25. Most will be
familiar with the hum from ballasts in fluorescent lighting or from transform-
ers in electrical equipment. This noise is due to magnetostriction, characteris-
tic of ferromagnetic materials such as iron, nickel, cobalt, and in alloys such as
Terfenol-D, which are characterized by a stronger magnetostrictive effect than
pure ferromagnetic metals. The effect causes shape change when the material
is subject to a magnetic field. The magnetization can be accomplished by static
magnetic fields, for example, by magnets and by alternating magnetic fields, by
alternating electric current, or by the combination of the two. The magnetostric-
tive effect may be positive or negative, i.e., may lead to an expansion or contrac-
tion of the material.
In contrast to piezoelectricity, the magnetostrictive effect is fundamentally
nonlinear since the shape change is the same irrespective of the direction of the
magnetic field, resulting in a doubling of the mechanical effect frequency com-
pared to that of the applied magnetic field. Because of this nonlinearity, magneto-
strictive materials must be biased by either a supplied static mechanical stress or
a magnetic field.
Sx
lx
FIGURE 9.25 Cylindrical core of magnetostrictive material having length lx. The bar is
surrounded by a coil carrying an electric current that induces a magnetic field along the
cylinder’s x-axis causing a length change ξ.
The static strain in a material as a function of the applied magnetic field depends
on the material properties and are described by a material constant kMS. As long as
the magnetic flux is well below saturation in the material, the strain induced by the
magnetic field B is
∂ξ
≈ kMS B2 (9.67)
∂x
For Cobalt, kMS ≈ 6 · 10 −6 T−2, while for Terfenol-D, kMS ≈ 2 · 10 −3 T−2. Figure 9.26
shows some typical measurement data [10].
Assuming a cylinder of magnetostrictive material as shown in Figure 9.25, the
strain equation, for small variations of the magnetic field around its static value, is
obtained as
∂ξ 1 Fx
=− + ΛBi (9.68)
∂x E Sx
where
E is Young’s modulus of the bar material
Fx is the applied force at the ends of the cylinder
Sx is the surface area of the cylinder ends
Bi is the applied magnetic field
Λ is the magnetostriction constant of the bar
The magnetostriction constant will depend on the external magnetic bias B 0
Λ = 2 EkMS B0 (9.69)
2.0
No prestress
16 MPa prestress
1.5 Strain ~B2.5
Strain × 0.001
1.0
0.5
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Magnetic flux density, B [T]
FIGURE 9.26 Examples of measured magnetic-field-induced strain in monolithic Terfenol.

Note the increased linearity and sensitivity caused by mechanical prestress of the material.
(From Engdahl, G., Ed., Handbook of Giant Magnetostrictive Materials, Academic Press,
New York, 1999.)
For the case of the magnetostrictive bar being used as a receiver, the change in flux
density may be due to an external applied magnetizing field with strength H or by a
change in strain and given by
∂ξ
Bi = µHi + µΛ (9.70)
∂x
Similar to Equation 9.57, one obtains that for Hi = 0, the relationship between force
and strain can be written as
∂ξ 1 Fx (9.71)
=−
∂x E′ S x
where the effective Young’s modulus E′ is

(
E ′ = E 1 − kME
2
)
(9.72)
Here, kME is the electromechanical coupling coefficient of the magnetostrictive

device
µΛ 2
kME = = 2kMS B0 µE (9.73)
E
similar to that for the electrostatic transducer. Magnetostrictive transducers have

about the same range of applications as those made from piezoelectric ceramics but
have the advantage that there is less need for high-voltage circuits. The cores are
typically made resonant to obtain maximum mechanical output. The mechanical
losses in magnetostrictive materials are usually quite low but there are also magnetic
hysteresis and eddy current losses in the material as well as resistance losses in the
magnetizing coil. In practical designs, the radiation load loss is made much larger
than these internal losses. The lever effect is also used in some varieties of these
transducers.
REVIEW QUESTIONS
9.1 Which are the dominant transducer mechanism types used in electroacoustic
transducers?
9.2 What is meant by direct and indirect conversion? Give examples.
9.3 Explain the principles of electrodynamic transduction.
9.4 How do electromagnetic transducers differ from electrodynamic transducers?
9.5 What is the working principle of piezoresistive transducers?
9.6 Why are balanced transducers preferred over unbalanced electrostatic
transducers?
9.7 Derive the conversion factor for a condenser microphone capsule.
9.8 Explain the principles of piezoelectric transducers.
REFERENCES
1. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986)
ISBN–13: 978–0883184943.
2. Merhaut, J., Theory of Electroacoustics, McGraw-Hill, New York (1979)
ISBN–13: 978–0070414785.
3. Hunt, F. V., Electroacoustics, American Institute of Physics, New York (1982) ISBN–13:
978–0883184011.
4. Sessler, G. M. and West, J. E., Condenser microphones with electret foil, J. Audio Eng.
Soc., 12(2), 129–131 (1964).
5. Wintle, H. J., Introduction to electrets, J. Acoust. Soc. Am., 53(6), 1578 (1973).
6. Rossi, M., Acoustics and Electroacoustics, Artech House, Norwood, MA (1988)
ISBN–13: 978–0890062555.
7. Kinsler, L. E. et al., Fundamentals of Acoustics, 2nd edn., John Wiley & Sons,
New York (1962) ASIN: B000LC9DO6.
8. Waanders, J. W., Piezoelectric Ceramics, Philips Components Marketing Division,
Eindhoven, the Netherlands (1991).
9. Gayford, M. L., Electroacoustics, Newnes Butterworth, London, U.K. (1970)
ISBN–13: 978–0408000260.
10. McKnight, G. P. and Carman, G. P., Large magnetostriction in Terfenol-D particulate
composites with preferred [112] orientation, in Proceedings of SPIE Smart Structures
and Materials 2001: Active Materials: Behavior and Mechanics, Vol. 4333 (2001),
pp. 178–183.
11. Engdahl, G. (Ed.), Handbook of Giant Magnetostrictive Materials, Academic Press,
New York (1999) ISBN–13: 978–0123885937.
10 Radiation and Impedance
10.1 INTRODUCTION
By sound radiation power is transmitted to the surrounding space but sound radiation
is not necessary to generate audible sound. Depending on the movement character-
istics of the loudspeaker driver’s diaphragm (the local amplitude and phase of vibra-
tion over its surface), the wavelength of sound in air, and the acoustic environment,
there may be either:
• sound power radiated to the far field (acoustic energy is transmitted away
from the diaphragm)
or
• a local reactive near-field sound pressure, that is, the diaphragm is moving
the air in its vicinity but without power radiation to the far field (close to a
plate carrying bending waves below the critical frequency)
From the viewpoint of audio, however, the latter can of course be considered as
sound generation as well since there is sound pressure in the near-field and this is
what we hear.
Generally, the driver’s diaphragm will be attached to a motor mechanism that
generates the force that causes the diaphragm to move. Depending on the driver
design, the force may be local or extended over most of the diaphragm area. Often,
the driver’s chassis will be set in motion by the reverse force and will vibrate as well.
Typically, the driver will be mounted in a baffle or box that has limited stiffness and
mass, and thus will also be set in motion by mechanical vibration transfer from the
driver. In addition, the sound pressure inside the box is likely to set the box walls in
motion with sound generation as a result.
10.2 RADIATION OF SOUND AND POWER LOSS

The radiation of sound can, from the viewpoint of electroacoustic analogies, be con-
veniently represented as power losses in resistance components in the circuit. Even
acoustical coupling between components in the loudspeaker can, under certain cir-
cumstances, be represented by components in the analog circuit. This is for example
the case when dealing with coupled radiators such as the loudspeaker(s)/openings in
loudspeaker boxes. The bass reflex vented box design is a typical example of a case
where the coupling may be important.
Once the sound has been radiated, the sound intensity at various angles at far dis-
tance will be determined by the geometry of the vibrating surface, that is, its shape
163
and dimensions. It is only in the far field of the sound source that the sound pressure
magnitude drops inversely to distance.
Radiated sound power will be relatively low as long as the vibrating surface is
much smaller than the wavelength, even when there is no aerodynamic short-circuit
present (such as with dipoles and quadrupoles, for example). For maximum sound
radiation, the radiator needs to be large compared to the wavelength of sound. The
radiation efficiency is often not a problem since in most cases the necessary electric
power to the transducer radiating sound is readily available.
Since the acoustic radiation load on the diaphragm is generally quite low, the
radiation efficiency of the loudspeaker will be small. Horns, resonant tubes, etc.,
may be used to match a loudspeaker diaphragm to the surrounding environment to
achieve higher efficiency. In the case of airborne sound we typically want to use the
horn to present a higher radiation load to the transducer than that obtained otherwise.
Nonlinearities can be important at high sound pressures. Because the air volume
is small in a loudspeaker box, at the throat of a horn, or inside a tube, the sound pres-
sure is likely to become high when the air is compressed by the driver diaphragm.
If the sound pressure becomes sufficiently high, the nonlinearities of the air will
become noticeable, for example, by a shift in static pressure, the so-called Rayleigh
sound radiation pressure. In a high-intensity beam of sound, the nonlinearity of
air will lead to high-pressure sound waves developing into shock waves (due to
heated air having a higher sound velocity). A further nonlinear phenomenon is that
of the air not moving linearly in a sound pressure high beam, causing a “spatial
rectification” and a mass flow of air in the direction of the beam, the Langevin
sound radiation pressure [1]. The discussion in the rest of this chapter assumes
linear air properties.
The presence of other coherently related sound sources will also lead to a change
in radiation load, which is discussed in Chapter 11. Image sources such as those seen
by the transducer due to reflection of sound by nearby surfaces will lead to changes
in radiation load, resulting in different sound radiation than that obtained with the
transducer mounted far away from any reflecting surfaces. The radiation load seen
by transducers mounted in an array varies between transducers located centrally
and those located at the end of the array. When the loudspeaker is radiating sound
into a cavity, the modes of the cavity will lead to a frequency-dependent radiation
load impedance.
10.3 SOUND RADIATION CHARACTERIZATION

We often want to be able to calculate the sound pressure next to vibrating surface
or the sound power that is radiated away from a vibrating surface. In Chapter 3 we
already studied how this can be done for a spherical sound source in “free” space.
However, any realistic sound source will have a different shape and vibration pattern
on its surface and we need a better method of characterizing sound radiation by the
vibrating surface of an object.
Typically, the vibration pattern of the surface will need to be known since sound
power radiation efficiency depends on the size and geometry of the surface, the
acceleration of the surface movement, if the surface is moving rigidly or if there are
Radiation and Impedance 165
waves moving on the surface of the diaphragm, in which case the characteristics of
the wave motion such as wave type, wavelength, and direction also become impor-
tant. The power radiation to the surrounding environment is usually characterized by
a radiation efficiency metric, the radiation ratio, or by a radiation impedance.
In this chapter we will only use the frequency domain approach to radiation. Using
the Fourier transform we can convert the transfer functions to impulse responses.
The direct calculation of the impulse response of various piston geometries is dis-
cussed in Appendix D.
10.4 RADIATION RATIO

The radiation ratio describes how much sound power is radiated by vibration of
a particular surface compared to that by vibration of a large rigid surface having
the same vibration velocity perpendicular to its surface. The radiation ratio σ is
defined as
P
σ= (10.1)
ρcS u2

where
P is the radiated power by one side of the surface
S is the radiating surface area
⟨ũ 2⟩ is the spatial average over the surface of the RMS vibration velocity
Note: Radiation ratio data is usually shown in its logarithmic form as 10 · log(σ).
If the radiation ratio is unity, the surface radiates as a large rigid plane piston that
is vibrating in the direction of its normal, equally at all points.
Roughly, one can say that we use the idea of the radiation ratio when we do not
know the specific vibration pattern of the surface, that is, when the surface is so
large and nonrigid that it carries transverse waves. Resonance patterns and edge
conditions considerably complicate the study of sound radiation from such surfaces.
10.5 RADIATION IMPEDANCE

An attractive way of studying the effects of the air on vibrating diaphragms is to use
the idea of radiation impedance in the analysis of acoustical circuits. In analogy with
electrical circuits, one can regard the radiation of acoustic power from an acoustic
system as a power loss in a resistive component. This is often a convenient approach
when working with electroacoustic systems. Because of the radiation conditions, the
radiation impedance usually is complex, that is, it has both real and reactive parts.
If written as an acoustic impedance Z AR we have
Z AR = RAR + jX AR (10.2)

but it can equally well be written as a sound field or mechanical impedance, depend-
ing on the preference.
The real part R AR is related to the sound power radiated while the imaginary
part X AR is related to the stored acoustic energy in the near-field of the source. In
contrast to conventional electrical resistance components, the real part of radiation
impedance generally has a frequency dependence. The R symbol is used since the
radiation resistance is often frequency-dependent.
Radiation impedance is typically used when studying sound radiation by plane
surfaces, which move with the same velocity and phase over their entire sound-
radiating area, although the theory is generally also applied to the various types of
diaphragms and membranes used for loudspeakers and microphones, which tend
to have conical and dome diaphragms. As long as the depth of the cone or dome is
much smaller than the wavelength under study, the theory for planar surfaces holds
reasonably well. The radiation impedance is usually considered separately for each
side of the loudspeaker diaphragm.
It is convenient to regard the air load on a tube of air as an impedance on a
virtual piston at the end of the air-filled tube as mentioned in Chapter 3. Even
though the flow lines of air of this case do not exactly correspond to those at a
physically hard piston, the approximation is generally sufficiently good for most
practical audio-engineering work.
10.6 VIBRATING PLANE AND SOUND FIELD INTENSITY

Assume a loudspeaker that has an infinitely large, perfectly flat, and ideally rigid dia-
phragm. Also assume the diaphragm to be driven by the transduction mechanism so
that the diaphragm vibrates perpendicularly to its surface with the same amplitude
and phase over its entire area.
Further, assume the loudspeaker diaphragm to be much larger than the wavelength
of sound radiated and that the diaphragm is located in the x = 0 plane. The plane
waves, for positive values of x, can be described by
ˆ − jkx (10.3)
p( x, k ) = Pe

Since the particle velocity u is given by the equation of motion
−1 ∂p(k, x )
u ( x, k ) = (10.4)
jωρ ∂x
the particle velocity for a plane wave moving in the positive x-direction is
Pˆ − jkx
u ( x, k ) = e (10.5)
ρc
The radiation impedance seen by each side of the diaphragm will be the sound
field impedance that in this case is the characteristic impedance of the medium
p
ZR = = ρc (10.6)
u
Since the outgoing waves are plane waves, the phase difference φ between the
particle velocity u and sound pressure _p is zero, and we find that for both x-directions
the intensity in the plane waves is
I = ρcu2 (10.7)

10.7 POWER RADIATED INTO AN INFINITELY LONG TUBE

Now assume that the flat piston radiates power into an infinitely long tube in
the positive x-direction and has the same cross section S as that of the tube. The
piston normal and its motion are along the tube axis. Since the tube is infinitely
long, there will be no sound reflected by the far end, and the radiation impedance
seen by the piston will be real. The radiation impedance can also be written as
mechanical or acoustical radiation impedance for use in the respective circuit
analogies as
Z MR = Z R S = ρcS (10.8)

ρc
Z AR = Z R /S = (10.9)
S
If the piston vibrates with a velocity u, the intensity I will be the same as for the
case of the very large diaphragm as given by Equation 10.7. The acoustic power
P emitted by the piston and carried down the tube by the wave will be
P = IS (10.10)
which can also be written as
P = IS = ρcSu2 = Re ( Z MR ) u2 (10.11)

or
ρc 2 2
P = IS = S u = Re ( Z AR )U 2 (10.12)
S
The tube cross section and the piston do not need to be circular for the wave in
the tube to be a plane wave, but can have any shape as long as the largest dimension
is smaller than one-half wavelength (see Chapter 3). Note however that if the piston
and tube do not have the same cross section, there will be evanescent higher order
modes excited close to the piston (discussed in Appendix F). In small tubes, the
viscosity of the air will affect the wave so that the particle velocity will be small
at the tube walls.
10.8 IMPEDANCE MATCHING

One of the reasons that the idea of radiation impedance is so attractive is that it allows
easy calculation of the sound radiation for a sound source for which the vibration
velocity is known. Assume the circuit in Figure 10.1 that shows an acoustical circuit
having a pressure source with an internal impedance and a load impedance. Both
impedances are assumed to be complex having the impedances Z Ai = R Ai + jX Ai and
Z AL = R AL + jX AL , respectively.
In analogy with electrical circuits, the power dissipated in the load circuit is
given by
P = U 2 Re {Z AL } = U 2 RAL (10.13)

One can show that the impedance condition that allows maximum power to be
obtained from the source is
Z Ai = Z AL
∗
→ RAi = RAL and X Ai = − X AL (10.14)

While this may seem attractive, it is important to observe that the same power is
dissipated in the internal resistance as in the external resistance, which may create
problems. In spite of the large power dissipated in the voice coil, the technique is
often used in horn loudspeaker systems. Most power radiation circuits instead strive
for high efficiency so that as much of the available power as possible is dissipated in
the load rather than in the internal resistance.
The power which is symbolically dissipated in the real part of the radiation
impedance is the power which is radiated by the vibrating surface. Integration of the
intensity in the sound field from the source, over a surface fully enclosing the sound
source, will give the same power value.
10.9 FUNDAMENTAL SOURCES

10.9.1 Monopoles
A monopole can be thought of as a sphere or spherical shell, very small compared to
the wavelength of sound, whose surface is oscillating radially, that is, “breathing”.
Monopoles, as noted in Chapter 3, can be used to build other more complex multipole
sources such as dipole and quadrupoles. Dipoles and quadrupoles are combinations
ZAi
U
p ZAL
FIGURE 10.1 An acoustical analogy circuit having a pressure generator with an internal
impedance Z Ai that is loaded by an external impedance Z AL .
of monopoles. The sound radiation by harmonic monopoles oscillating at the same

frequency will depend on the phase relationship between their oscillation velocities.
The monopole is of particular interest since the sound field radiated by closed box
loudspeakers, which have dimensions that are small compared to the wavelength of
sound, can be regarded approximately spherical at some distance. We can therefore
expect some similarity in the behavior of the radiation impedance of such loud-
speakers and that of a spherical radiator having the same radiating surface area.
The power radiation of a monopole can be determined from the wave equation
as described in Chapter 3. Solving the wave equation for the case of spherical
coordinates, one obtains the solution for outgoing waves as
A − jkr
p(r, k ) = e (10.15)
r
and we obtain the radial particle velocity as
A  1  − jkr
u (r , k ) =  1+ e (10.16)
ρcr  jkr 

Using our definitions in Equations 10.6 and 10.9, we can write the acoustic
radiation impedance at the outer surface of a spherical shell having a radius a as
ρc jka
Z AR (a, k ) = (10.17)
4πa 2 1 + jka
We can also rewrite the equation slightly to obtain
1
Z AR (a, k ) = (10.18)
4πa 2 4πa
+
ρc jωρ
This way of writing the radiation impedance shows that we can regard Z AR as
the impedance of two acoustical circuit components in parallel since the resulting
impedance Z due to two impedances Z1 and Z2 in parallel is
1 1 1
= + (10.19)
Z Z1 Z 2
We find the components of the acoustical parallel impedance as
ρc
RAR = (10.20)
4πa 2
u U
p ρc ρ
ZMR F 4πa2ρc 4πa3ρ ZAR
4πa2 4πa
(a) (b)
1 4πa2
F 4πa2ρc p ρc
ρ
YMR u 4πa3ρ YAR U
4πa
(c) (d)
FIGURE 10.2 Graphic circuit element representations of the radiation impedance Z and
mobility Y for monopoles, for a sphere with radius a. (a) mechanical impedance analogy,
(b) acoustical impedance analogy, (c) mechanical mobility analogy, (d) acoustical mobility
analogy.
and
ρ
jX AR = jω = jωM AR (10.21)
4πa
The mass load of the mechanical radiation impedance is equal to the mass of the
air, which is “statically” displaced by the sphere.
Figure 10.2 shows some different representations of the analogies that can be
used for the radiation impedance of the sphere. Since the radiation admittance YAR
is the inverse of the radiation impedance Z AR, the expression for acoustical radiation
mobility series circuit will be
4πa 2 4πa
Y AR(a, k ) = + (10.22)
ρc jωρ
While parallel circuits are useful in many applications, the exact original formu-
lations are usually not of interest for transducer design. When we need to quickly
analyze, for example, the impedance analogy of a transducer, it is more intuitive to
have a series-style circuit. The parallel component impedance circuits can be con-
verted to approximate series impedance circuits as follows:
The sound field impedance seen by the outside of the sphere is
ρc  1 
Z AR (a, k ) = jka  (10.23)
4πa 2
 1 + jka 

When the sphere is small compared to wavelength, that is, ka ≪ 1, we can do a
series expansion of the parenthesis. Retaining only the first two terms of the series
expansion, we obtain
ρc
Z AR (a, k ) ≈ jka (1 − jka ) (10.24)
4πa 2
and

Z AR (a, k ) ≈
ρc
4πa (
2 (
2
)
ka ) + jka (10.25)
We note that we obtain the expressions for the resistance and mass in
Equation 10.25 as
ρω 2
ℜAR (a, ω) ≈ (10.26)
4πc
ρ
MAR (a) ≈ (10.27)
4πa
For frequencies where ka ≪ 1, Z AR can be written as composed of a mass reactance

and a resistance in series
Z AR (a, ω) = ℜ AR + jX AR (10.28)

ρ 2 ρ
Z AR (a, ω) = ℜ AR + jX AR ≈ ω +j ω (10.29)
4πc 4πa
The Fraktur letter ℜ is used here to indicate that the resistance varies with
frequency in this approximation. The mechanical radiation impedance Z MR for the
sphere is obtained by multiplication of the acoustic impedance by the square of the
sphere surface area
( )
2
Z MR = S 2 Z AR = 4πa 2 Z AR (10.30)

and is
4πρa 4 2
Z MR (a, ω ) = ℜ MR + jX MR ≈ ω + j 4πρa 3ω (10.31)
c
The mechanical and acoustical radiation impedances can be represented by

equivalent electrical series circuits as shown in Figure 10.2 c and d. Figure 10.3
shows the series circuit representation of the normalized specific radiation
impedance for a sphere.
1
Normalized radiation impedance Z/ρc
0.3
0.1
0.03
0.01
0.003
0.001 Z= + jX
0.0003 X
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.3 Curves showing the normalized specific radiation impedance for a breathing
sphere of radius a. The ordinate can also be read as Z MR /ρc4πa2 or Z AR 4πa2/ρc.
10.9.2 Dipoles
Dipoles and their relative power radiation were discussed in Chapter 3. Dipoles
consist of two monopole sources out of phase, at some small distance much smaller
than the wavelength, as shown in Figure 3.4. Assume that each monopole by itself
would radiate a power P0 if in free field. Calculating the sound power radiated by the
dipole, PD, we find that the ratio between PD and P0 is
4π 2 b2
PD = P0 (10.32)
3 λ2
As we can see from Equation 10.32, the dipole radiates much less power than a
single monopole having the volume velocity of one of the dipole elements.
10.9.3 Quadrupoles
Quadrupoles were also introduced in Chapter 3 and consist of four monopole sources,
two in phase and two out of phase. There are several configurations possible, two
types of which were shown in Figure 3.5. One can show that the effective sound
power radiation for a quadrupole, PQ, is much smaller than that for either of the free
monopoles separately, P0. One can show that, for the type of quadrupole shown in
Figure 3.5a, the ratio of PQ to P0 is
16π 4 b 4
PQ = P0 (10.33)
15 λ 4
The sound power radiation for this quadrupole is a factor (kb)2/5 less than that of
the dipole studied earlier.
10.9.4 Oscillating Sphere
The sound radiation properties of vibrating loudspeaker diaphragm or other body
vibrating by translational movement can be roughly modeled for low frequencies
as a vibrating sphere having about the same volume as the dimensions of the body.
One can show that the radiation impedance of such a sphere of radius a is [2]
0.01901a 2ρ 0.2705ρ
Z AR (a, ω) = ℜ AR + jX AR ≈ ω 4 + jω (10.34)
c 3
a
The curves in Figure 10.4 show the normalized acoustical radiation impedance of
such an oscillating sphere.
1
0.3
0.1
0.03
0.01
0.003
0.001 Z= + jX
0.0003 X
0.0001
0.01 0.02 0.05 0. 1 0. 2 0. 5 1 2 5 10
ka
FIGURE 10.4 Curves showing the normalized radiation impedance level (dB) for a
transversally oscillating sphere of radius a. The ordinate can also be read as Z MR /ρc4πa2 or
Z AR 4πa2/ρc.
10.10 PISTONS
10.10.1 Free Circular Piston
The dipole case is of interest for some loudspeaker designs. A free electrodynamic
loudspeaker that is not mounted in an enclosure behaves as a dipole at low frequencies,
and is extremely inefficient at radiating acoustic power when the dimensions of the
loudspeaker are much smaller than the wavelength of sound. Some loudspeaker drivers
are used without a box, free or mounted in a free-standing baffle, to avoid coloration
of the sound due to internal box resonances or box wall vibrations. For such case, the
impedance behavior of the free plane circular piston radiator will be of interest.
We studied the properties of an oscillating sphere in the previous section. In the
case of a vibrating loudspeaker diaphragm it is convenient to write the impedance
as the sum of the impedances of the two sides. One then obtains the following low-
frequency approximations (ka < 0.5) to the combined acoustical radiation impedance
for both sides of a plane circular piston of radius a in free space [3]
16ρa 2 8ρ
Z AR (a, k ) = ℜ AR + jX AR ≈ ω 4 + jω 2 (10.35)
27π c
3 3
3π a
The curves in Figure 10.5 show the normalized acoustical radiation imped-
ance [2]. Comparing the results shown in this figure to those of the oscillating
1
0.3
0.1
0.03
0.01
0.003
0.001 Z= + jX
X
0.0003
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.5 Curves showing the normalized radiation impedance for both sides of a free
circular piston. Note that the asymptotic value is 2 since the impedance in this case is for the
total load. (After Beranek, L.L., Acoustics, American Institute of Physics, New York, 1986.)
sphere in Figure 10.4 we note that the impedance is higher. This is mainly due to
the surface area of the disk being only one-half of that of the sphere having the
same radius.
Compared to the baffle-mounted piston radiator, both the real and imaginary parts
of the radiation impedance for the dipole-style loudspeaker are lower. This is less
of a problem than one might think. For an electrodynamic driver, the absence of a
box typically leads to lower resonance frequency. The radiation impedance in a real
room leads to reasonable coupling for the low-frequency modes as well. Of course,
the driver’s radiating surface area needs to be large but this also helps in directing the
sound and reducing the effects of reverberation.
10.10.2 Circular Piston in a Baffle

Most commercial loudspeaker drivers have conical or semiconical diaphragms to
obtain higher stiffness so that mechanical resonances in the diaphragms can be
pushed to frequencies high enough not to cause problems in practical use. Some
loudspeaker drivers such as electrostatic and isodynamic ones (cf. Chapter 9) have
the driving force approximately equally distributed over their diaphragms, and in
this way excite the higher order mechanical resonances of the diaphragms less. To
obtain numerical impedance data for conical and semiconical loudspeaker drivers, it
is best to model their radiation using finite element or boundary element techniques.
These will not be discussed here.
For most practical situations, it is sufficient to use analytical impedance models
where the three-dimensional diaphragm has been replaced by a plane circular piston
having the same projected radius a and area S, as shown in Figure 10.6.
One can derive an expression for the radiation impedance acting on one of the
sides of the piston by calculating the total pressure on the piston and dividing by the
piston vibration velocity.
z
Observation point
r
r΄
θ
a y
σ
x
FIGURE 10.6 The coordinate system for calculation of the sound radiation by a circular
piston in an infinite baffle, both in the z = 0 plane.
The piston is assumed to be hard, so that the particle velocity at its surface is
that of the vibrational velocity u of the piston. The piston is further assumed to be
mounted in a plane surface that has infinite impedance. Initially, the back side of the
piston is neglected and there is no air flow between the front and back sides along
the perimeter of the piston.
Assume the piston centered on the z-plane as shown in Figure 10.6. We use
Huygen’s principle and regard the sound field generated to one side of the baffle by
the piston as that generated by a large number of small patches dS that act as small
sound sources at an infinitesimal distance from the top of a plane hard surface. The
sound pressure from one such small source on the surface at dS1, having a volume
velocity dU = udS1, at some other patch dS 0 at a distance r will be
ρω
p(r, ω) = j dU e − jkr (10.36)
2πr
We now integrate over the surface of the piston assuming this pressure contribu-
tion from each one of the small surface patches dS1. The result of the integration is
that the pressure at each point dS 0 on the piston surface is given by
ρω
p(r, ω) =
∫∫ j 2πr e
S piston
− jkr
udS1 (10.37)

Assuming the piston rigid so that it moves with the same velocity and phase over
its surface we obtain “Rayleigh’s integral”
ρω e − jkr
p(r, ω) = j
2π
u
∫∫
S piston
r
dS1 (10.38)

We now calculate the total reaction force on the piston by integrating the reaction
pressure over the piston surface elements dS 0, which will be
ρω e − jkr
F (ω ) =
∫∫
S piston
p dS0 = j
2π
u
∫∫
S piston
dS0
∫∫
S piston
r
dS1 (10.39)

Using the definition of mechanical radiation impedance Z MR = F/u we find the

equations for the mechanical and acoustic impedances as [2]
 J (2ωa /c)  Sh (2ωa /c) 

Z MR (a, ω) = ℜ MR + jX MR = πa 2ρc  1 − 1  +j 1 (10.40)
 2ωa /c  2ωa /c 

1
0.3
0.1
0.03
0.01
0.003
0.001 Z= + jX
0.0003 X
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.7 Curves showing the normalized radiation impedance for a circular piston
of radius a in an infinite plane rigid baffle. The ordinate can also be read as Z MR /ρc4πa2 or
Z AR 4πa2/ρc.
Z MR (a, ω )
Z AR (a, ω ) = (10.41)
(πa )
2
2

The functions J1 and Sh1 are Bessel and StruveH functions, for which series
expansions can be found in Ref. [4]. The normalized mechanical impedance given
by the equations is plotted in Figure 10.7. (If one assumes that the piston has the
same radiation conditions on both sides, the resulting radiation impedance will be
twice that given here.)
Circuit analogies that approximate these equations over the full frequency range
may be found in Ref. [2]. Usually we are only interested in the values of the radiation
impedance for ka ≪ 1, and the expression for Z AR can then be simplified to represent
a parallel circuit of resistance R AR and mass MAR with the following approximate
values for the resistance and mass components:
ρc
RAR ≈ 0.459 (10.42)
a2
ρ
M AR ≈ 0.270 (10.43)
a
We note that these values are slightly higher than those for the monopole. This is
due to the influence of the “self-reflection” in the piston/baffle combination.
Two useful approximations of the series representation of the radiation impedance
of a circular piston in a hard baffle (for ka < 0.5) are [2]
1.57a 4ρ
Z MR (a, ω) = ℜ MR + jX MR ≈ ω 2 + jω2.67a 3ρ (10.44)
c
0.159ρ 0.27ρ
c a
Expressions for the radiation impedances of many other types of radiating

surfaces may be found in Ref. [5].
10.10.3 Elliptical Piston in a Baffle

Many loudspeakers have diaphragms that are not circular but rather elliptical or
quasi-rectangular. The noncircular shape reduces the oscillations of the radia-
tion impedance curves at frequencies where the wavelength of sound is about the
loudspeaker diameter. Equations for the radiation impedance of elliptical piston
radiators may be found in Ref. [5]. Figures 10.8 and 10.9 show respectively the real
1
0.3
0.1
0.03
Real part
0.01
0.003
All curves a = 1
0.001 b=1
b = 0.5
0.0003 b = 0.25
b = 0.125
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.8 Curves showing the real part of the normalized radiation impedance for an
elliptical piston having axes 2a and 2b in an infinite plane rigid baffle for various ratios of a/b.
Series representation of the impedance Z = ℜ + jX.
1
0. 3
0. 1
Imaginary part
0.03
0.01
0.003
All curves a = 1
b=1
0.001
b = 0.5
b = 0.25
0.0003 b = 0.125
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.9 Curves showing the imaginary part of the normalized radiation impedance
for an elliptical piston having axes 2a and 2b in an infinite plane rigid baffle for various ratios
of a/b. Series representation of the impedance Z = ℜ + jX.
and imaginary parts of the radiation impedance for various ratios of the axes of an
elliptical piston radiator (long axis = 2b, short axis = 2a).
If the impedance is instead shown for constant surface area rather than for con-
stant axis length, the curves show more clearly the influence of the noncircular
shape. Figures 10.10 and 10.11 show the real and imaginary parts of the radiation
impedance of an elliptical piston radiator drawn for constant piston surface area.
10.10.4 Rectangular Piston in a Baffle

As discussed in the previous section, many loudspeaker diaphragms are noncircular.
A rectangular shape further stresses the tendencies already shown in the previous
section on elliptical pistons. Loudspeakers that have rectangular pistons or
membranes with large aspect ratios are often called “line sources.” Such loudspeakers
are typically based on rectangular electrodynamic “ribbon” or on electrostatic thin
membrane transducers. Both are characterized by membranes that have low mass
per unit area and that rely on radiation impedance load to remove the resonant
characteristics of their membranes. Because of the homogeneously distributed force,
resonant membrane modes are less excited and can often be neglected in the analysis.
Since the driving force is distributed equally over the membrane surface, the
radiation impedance concept is useful for studying their frequency-response
characteristics. This is easily done by regarding the membrane as a rectangular
piston set in a baffle. It is important however to note that in the case of resonant
1
0.3
0.1
Real part
0.03
0.01
0.003
a/b = 1
0.001 a/b = 2
a/b = 4
0.0003 a/b = 8
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
k √ab
FIGURE 10.10 Curves showing the real part of the normalized radiation impedance for
an elliptical piston having axes a and b in an infinite plane rigid baffle. Curves shown for
constant surface area. Series representation of the impedance Z = ℜ + jX.
1
0.3
0.1
Imaginary part
0.03
0.01
0.003
a/b = 1
0.001 a/b = 2
a/b = 4
0.0003 a/b = 8
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
k √ab
FIGURE 10.11 Curves showing the imaginary part of the normalized radiation impedance
for an elliptical piston having axes 2a and 2b in an infinite plane rigid baffle. Curves shown
for constant surface area. Series representation of the impedance Z = ℜ + jX.
3
Normalized radiation impedance Z = ρc: Real part
1
0.3
0.1
0.03
0.01
0.003
W/L = 1
0.001 W/L = 4
W/L = 0.25
0.0003
0.0001
0.01 0.03 0.1 0.3 1 3 10 30 100 300 1000
k W
FIGURE 10.12 Curves showing the real part of the normalized radiation impedance for one
side of a rectangular piston in an infinite rigid baffle for various width-to-length (W/L) ratios
drawn as functions of k·W.
modes in low mass, highly compliant membranes, the piston analysis fails and a
more rigid analysis is necessary [6].
The radiation impedance of the rectangular piston source is calculated the same
way as for the circular piston from the integral in Equation 10.38. Equations for the
calculation of strip-like pistons can be found in Ref. [5] but are unwieldy and will not
be reproduced here. Curves for the normalized acoustic radiation impedance seen by
one side of a piston having width W and length L mounted in an infinite rigid baffle
are shown in Figure 10.12. These curves were calculated using equations in this
reference. A somewhat different approach is given in Ref. [7]. For a ratio W/L = 1, the
curves are fairly similar in character to those of the circular piston.
The curves in Figure 10.13 show the real part of the normalized radiation
impedances for more extreme W/L ratios, similar to those found in electrostatic and
isodynamic loudspeakers drawn as functions of k√(WL).
10.10.5 Circular Piston Radiator at the End of a Long Tube

Most loudspeakers are boxes that are used as “free standing” units, that is, they
are not mounted into a wall or other baffle-like surfaces. Because the loudspeaker
drivers usually do not cover the entire front side of the box, it is necessary to consider
two frequency ranges, one “medium-frequency” range where the box dimensions
are so large compared to the driver’s radius that the box may be considered an
approximation to a baffle and one “low-frequency” range where the box can be
neglected and considered rather to be part of an infinitely long tube extending behind
the loudspeaker driver.
3
Normalized radiation impedance Z = ρc: Real part
1
0.3
0.1
0.03
0.01
0.003
W/L = 1/4
0.001 W/L = 1/40
W/L = 1/400
0.0003
0.0001
0.001 0.01 0.1 1 10 100 1,000 10,000 100,000
k √WL
FIGURE 10.13 Curves showing the real part of the normalized radiation impedance for one
side of a rectangular piston in an infinite plane rigid baffle for various width-to-length (W/L)
ratios drawn as functions of k√(WL).
Two useful approximations of the radiation impedances for a plane circular piston
of radius a at the end of a long tube (for ka < 0.5) are [2]
0.7854a 4ρ
Z MR (a, ω ) = ℜ MR + jX MR ≈ ω 2 + jω1.927a 3ρ (10.46)
c
0.0796ρ 0.1952ρ
c a
The radiation impedance acting on the outside of a plane circular piston at the
end of a long tube over a wider frequency range is shown in Figure 10.14. We note
that the apparent attached air mass is about half of that for the case of the piston in a
baffle, as is the real part of the radiation impedance.
These radiation impedance approximations are very useful when designing small
loudspeaker enclosures. The impedances toward the inside of the box are of course
given by the properties of the enclosed air, box walls, and whatever sound-absorbing
material that the box is filling with.
The radiation impedance for loudspeaker drivers’ mounted in real loudspeaker
boxes will be somewhere between the two extremes discussed here, the case of
the loudspeaker in an infinite baffle and that of the loudspeaker at the end of a
long tube. The radiation impedance for such conditions is best calculated using
boundary element modeling in which case the actual diaphragm shape can be taken
into account as well.
1
Normalized radiation impedance Z /ρc
0.3
0.1
0.03
0.01
0.003
0.001 Z= + jX
0.0003 X
0.0001
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.14 Curves showing the normalized mechanical radiation impedance for a
circular piston at the end of a long tube. (From Beranek, L.L., Acoustics, American Institute
of Physics, New York, 1986.)
10.11 TRANSVERSE WAVES IN PLATES

In the preceding sections, the piston was assumed to be rigid so radiation impedance
is a handy tool for calculating the radiated sound. At high frequencies most
mechanical structures will exhibit wave motion due to longitudinal, transverse,
and bending waves being excited in the structure by the driving force. Transverse
waves and bending waves are common wave types in loudspeaker diaphragms. The
coupling properties between bending waves and waves in air are important from the
viewpoint of both reception and radiation.
Two cases of sound-radiating surfaces are of special interest, those that carry
progressing waves and those that support standing waves. In either case the ratio
of the surface wave’s wavelength to the wavelength of waves in air (at the same
frequency) is an important property for sound radiation as is the geometry or the
radiating surface (dimensions, plane, curved, etc.). Standing waves can be considered
the modal behavior that results from two identical waves that move in the opposite
direction. Most loudspeaker diaphragms will need to be considered nonrigid above
some limiting frequency, and for higher frequencies the radiation will show peaks
and dips as a result of the vibration mode pattern set up in the surface.
Exact treatment of such sound radiation is seldom necessary since any practical
case is likely to be difficult or even impossible to model analytically. However, using
boundary and/or finite element modeling, useful numerical approximations can be
obtained for the radiation factor which is an important tool in determining the radia-
tion efficiency of vibration in these cases.
10.11.1 Bending Waves in an Infinite Sheet

The radiation ratio σ for an infinite sheet carrying a single frequency bending wave
with a linear wave front and constant amplitude is
1 1
 2
−   2
− 
  kB   2    ω c   2 
σ ( ω ) = Re  1 −     = Re  1 −     (10.48)
 k   ω 
   
   
We note from this expression that we have three different cases:
• Λb < λair low frequency, that is, ω < ωc

• λB = λair critical frequency, ω = ωc
• λB > λair high frequency, ω > ωc
At frequencies below the critical frequency there cannot be any radiation of sound
power to the far field. This can be understood intuitively by studying Figure 10.15.
Since the distance between the out-of-phase areas is smaller than one-half wave-
length of sound in the medium surrounding the sheet, destructive interference, also
called an aerodynamic short circuit, will prevent sound radiation.
Note however that close to the surface there will be audible sound pressure and
particle velocity. This sound field carries reactive power and is sometimes called a
reactive near-field or an evanescent field. The sound pressure in the near-field drops
very quickly with distance as discussed in a later section. It is important to remember
though that the flow velocities in the near-field may be very high so that the bending
wave vibration of a panel can be damped by a porous sound absorber placed nearby.
The radiation impedance of the surrounding air to each side of the sheet can be
shown to be [6]
p jωρ
Z R (ω ) = = (10.49)
uy kc2
−1
k2
Air movement pattern

y + +
– –
x
z
λB
Direction of wave
FIGURE 10.15 Air flow, at low frequencies f < fc, in the near-field at one side of a sheet
carrying a bending wave in the x-direction (λB < λair). The air being pushed out by positive
y-direction sheet movement is sucked in by adjacent negative y-direction sheet movement.
pmax pmax Normal
λair
y
x
z
λB
Direction of bending wave
FIGURE 10.16 Sound radiation by bending waves at one side of a sheet will be at an angle
φ to the sheet at high frequencies, i.e., f > fc. The wavelength of the bending wave is longer
than that of the longitudinal wave in air. The wavelength of the bending wave is determined
by the condition λ B sin(φ) = λ air.
so, using Equation 4.27, we find that for frequencies well below the critical frequency,
the radiation impedance to each side of the sheet is
B′
Z R ≈ jρcB = jρ ω 4 (10.50)
m″
where
B′ is the bending stiffness per unit length
m″ the mass per unit area of the plate as defined in Chapter 4.
This radiation impedance corresponds to that of a mass load of an air layer having a
thickness of λB /2π.
At frequencies above the critical frequency the sheet will radiate sound from both
of its sides. (Remember that the radiation ratio only describes radiation to one side.)
The waves will be radiated at an oblique angle φ as shown in Figure 10.16 given by
λ B sin(ϕ) = λ air (10.51)

Figure 10.17 shows the basic frequency dependence of the radiation ratio for
an unattenuated plane bending wave in an infinite plane sheet. If the sheet is
finite, the wave field in the medium surrounding the sheet will lose some of the
symmetry which is necessary for the cancellation of sound radiation at frequencies
below the critical frequency. The radiation will be primarily from the areas close
to the discontinuity.
10.11.2 Bending Waves in Damped Sheets

Another case of reduced symmetry is that of a sheet in which the bending wave
is damped, for example, due to internal losses. Again, the reduced symmetry
results in sound radiation below the critical frequency, as shown by the radiation
ratio curves shown in Figure 10.18. For these frequencies, increased damping is
accompanied by more radiation; so, damping a sheet does not necessarily lead to
less sound radiation [6].
10
0
10 log(σ) [dB]
–10
–20
–30
–40
0.031 0.063 0.125 0.25 0.5 1 2 4 8 16
Normalized frequency: ω/ωC
FIGURE 10.17 The radiation ratio, shown as 10 log(σ), as a function of frequency for an
undamped plane bending wave in an infinite sheet.
10
Approximate radiation ratios
0 for small loss factors η
and k/kB <<1
10 log(σ) [dB]
–10
–20 η = 0.1
η = 0.04
η = 0.01
–30
–40
0.031 0.063 0.125 0.25 0.5 1 2 4 8 16
FIGURE 10.18 The radiation ratio, shown as 10 log(σ), as a function of frequency for a
plane bending wave in a sheet for some cases of damping at frequencies below the critical
frequency. (From Cremer, L. et al., Structure-Borne Sound: Structural Vibrations and Sound
Radiation at Audio Frequencies, Springer, Berlin, Germany, 2005.)
10.11.3 Bending Waves in Finite Sheets

The larger the dimensions of the sheet compared to the wavelength of sound in
the medium, the smaller the sound radiation since the relative influence of the
sound radiation by edges and corners will diminish due to their relatively smaller
areas. The radiation ratios for some such cases are shown in Figure 10.19. The
data shown in this figure applies to the case of a strip of material, on two edges,
10
0
10 log(σ) [dB]
–10
b/λC = 3
b/λC = 8 Bending wave
–20 b/λC = 80 Radiation propagation
from
edges
–30
b
–40
0.031 0.063 0.125 0.25 0.5 1 2 4 8 16
FIGURE 10.19 The radiation ratio, shown as 10 log(σ), as a function of frequency for an
undamped plane bending wave in a finite sheet of width b for some values of b/λc. The wave-
length at the critical frequency of the sheet is λc. (From Cremer, L. et al., Structure-Borne
Sound: Structural Vibrations and Sound Radiation at Audio Frequencies, 3rd edn., Springer,
Berlin, Germany, 2005.)
carrying a (unattenuated) bending wave which radiates from the ends of the strip
(i.e., at the edges) [6].
Note that the symmetry leading to aerodynamic sound cancellation is also
disturbed when a sheet is stiffened, for example, by stiffeners or attachments as
discussed in a later section. This results in sound radiation at frequencies below
the critical frequency. Increasing the stiffness of a sheet does not necessarily
lead to lower sound radiation by bending waves, neither does fixing the sheet to
a point or line.
The vibration patterns of resonant sheets were discussed in Chapter 4. The radia-
tion from such sheets is best discussed using a multipole description of the sheet
vibration. To simplify the discussion, the vibrating sheet is assumed to be mounted
in an infinite rigid baffle. For a pure tone, the bending wave pattern on the sheet can
be thought of as composed by four plane bending waves that are incident pairwise
from opposite directions and cause the field in the sheet. Depending on the size of
the panel and the wavelength of bending waves and waves in air, there will be some
different cases. Figure 10.20 shows the displacement pattern of a plane sheet with
simply supported edges. We assume that the sheet is large compared to the wave-
length of sound in air.
At frequencies above the critical frequency the entire sheet is radiating sound.
This sheet resonance mode is called a surface mode.
At frequencies below the critical frequency, most of the sheet will suffer quad-
rupole cancellation. The radiation will be primarily from the corners of the sheet
and similar to that of four coherent monopoles. It can be estimated by the addition
of the sound pressure generated by these volume velocity sources in or out of phase
according to Equation 10.36. The volume velocity will be that averaged over the
z
Observation point
ly large
f << fC
ly y
lx large
lx
x
FIGURE 10.20 Displacement pattern for a corner mode on a sheet that is several air-
wavelengths long and wide. At frequencies below the critical frequency, the corners will
radiate like four coherent monopoles (of which two are out of phase).
corner source area. Note however that because of the presence of the rigid baffle, in
which the sheet is assumed to be set, the sound pressure and power will be twice that
of the free monopole.
Figure 10.21 shows the displacement pattern of a similar plane sheet with simply-
supported edges. We still assume that the sheet is large compared to the wavelength
of sound in air. Most of the sheet will suffer dipole cancellation. At frequencies
below the critical frequency, the radiation will be primarily from the two edges at
x = 0 and x = lx. Such a sheet is considered to have an “edge” mode.
For a sheet that is smaller in size than the wavelength of sound in air at the sheet’s
vibration frequency, the discussion is similar but it is now important to take the
interaction between the corner modes into account. This interaction will depend on
the mode pattern. The mode shown in Figure 10.22a has mode indices qx = 3 and
qy = 3, and has its corners out of phase, and will suffer quadrupole cancellation.
The mode in Figure 10.22b that has qx = 1 and qy = 5 will have edges along y = 0 and
y = ly that could be considered dipoles, but this essentially is still a case of quadrupole
cancellation since the sheet dimensions are assumed smaller than the air-wavelength.
A sheet that is so small that there are no bending wave modes (but there is some
bending) can be considered a piston source, and its radiated power is calculated from
Equation 10.36 taking into account the effective volume velocity which will be a
result of the vibration distribution over the surface.
It is clear from the preceding discussion that the sound radiation from a resonant
sheet at frequencies below the critical is likely to be quite complex. Taking into
account that most sound sources have signals that are not pure tones but rather
z
Observation point
ly large
f << fC
ly y
lx large
lx
x
FIGURE 10.21 Displacement pattern for an edge mode on a sheet that is several air-
wavelengths long and wide. At frequencies below the critical frequency, the edges will radiate
like two coherent dipoles.
z Observation point z Observation point
lx small lx small
ly small ly small
f << fC f << fC
ly ly
y y
lx lx
x x
(a) (b)
FIGURE 10.22 Displacement pattern for two modes at a frequency where the dimensions
of the sheet are much smaller than the wavelength in air.
approximately band-limited noise in the signal sense many modes are excited and
some form of averaging over the modes necessary to obtain a reasonable estimate of
the radiation factor. We will discuss this matter further in Chapter 21 where we study
loudspeaker drivers that use flexing diaphragms.
The discussion of resonant sheet radiation cancellation earlier applies to any
transversally moving sheet, not only one that carries bending waves. Figure 21.7
shows a design curve for approximating the average radiation ratio of a resonant
finite panel of perimeter length P and area S with simply-supported edges. The
reader is referred to Refs. [6,8]. Note however that in the low modal region it is
very difficult to calculate the radiation analytically. The damping of the modes
is primarily through internal losses and radiation. At slightly higher frequencies

(but still with edge lengths lx, ly ≪ λ), one can use Equation 10.37, Rayleigh’s
integral, if the vibrating surface is flush mounted into a rigid baffle:
ρω e − jkr
p(r, ω) = j
2π ∫∫
S piston
uz ( x, y, ω )
r
dS (10.52)

Here, uz(x,y,ω) is the vibration velocity in the z-direction of the surface at

points (x,y).
10.11.4 Sound Radiation by Bending Wave Point Excitation

We have noted that for sound radiation from bending waves, below the critical
frequency of the sheet, the wave pattern must be disturbed so that the aerodynamic
short-circuit action is prevented. This is the case, for example, when driving a flexible
sheet at a point or along a line at frequencies. It is also the case when a support point
on a sheet is kept still and there is an incident bending wave field. The support point
can be considered a driving point working to cancel the incoming bending wave
field. Figure 10.23 shows the principle for sound radiation from a flexural near-field
on an infinite, point-excited sheet. Circular areas away from the driving point cancel
aerodynamically, but close to the driving point there is no cancellation.
Assume that the wavelength at the critical frequency is λc and the RMS vibration
velocity is ũ at the point where the sheet is driven. One can show that the bending
wave mechanical input impedance at the driving point of an infinite sheet is given
by Equation 4.34.
Using that equation one can show that the sound power radiation from bending
waves in such a point-excited sheet is [6]
8ρc3 2
P= u (10.53)
π 3 fc2
If we compare this sound radiation to that of a rigid piston in a plane baffle, we

find that the radiation corresponds to that of a rigid piston having the radius
2
a= λ c (10.54)
π3
This radius corresponds to about a quarter wavelength at the critical frequency.
FIGURE 10.23 Sound radiation from a flexural near-field on a point-excited sheet.

10.11.5 Sound Field Close to Nonradiating Bending Wave Fields

In the previous discussion of undamped bending waves in sheets it was shown that
there is also radiation of sound power below the critical frequency for many reasons.
Since our hearing is pressure-sensitive, it is necessary to also study the near-field
sound pressure behavior for frequencies below the critical frequency.
Consider the flexing sheet shown in Figure 10.15. Here the frequency of vibration
is below the critical frequency, that is, kB ≫ k. The bending wave propagation is in
the x-direction and the transverse movement in the y-direction. The wave field in the
air in the x-direction must have the same phase velocity, that is, wave number com-
ponent, as that of the bending wave which is also moving in the x-direction, kB = k x.
Since the wave field in the air obeys
k y = k 2 − k x2 (10.55)

the sound pressure in the air surrounding the sheet for positive values of y can be
described by
kB2 − k 2 y
p ( x, y, k ) = pe
И − jkB x e − (10.56)

We see that there is no wave propagation away from the sheet below the critical
frequency. This type of sound field is called an evanescent field. The sound pressure
level drops off at about −8.7kB (dB/m) as we move away from the sheet.
10.11.6 Radiation Factor versus Radiation Resistance

It is clear from the preceding discussion that radiation impedance in the “classical”
electroacoustic sense is less suitable to calculate the sound radiation from resonant
sheets. The radiation factor indicates the radiation efficiency of the sheet, but is
usually applied to systems that are characterized by nonsinusoidal noise-like signals.
Although voice and music are band-limited signals, classical electroacoustics assumes
pure tone signals by using the symbolic jω-method.
One way of joining the methods to some extent is to use the radiated power
compared to the average of the mean-square value of the vibration over the sheet.
Using the idea of radiation resistance in this way, we obtain the definition of the
radiation factor
P = ρc u2 σS = ℜ MR u2 (10.57)

We can then obtain a form of mechanical radiation resistance ℜMR when using
the radiation factor:
ℜ MR = ρcσS (10.58)

10.12 RADIATION IMPEDANCE AS A LOW-PASS FILTER

The radiated sound power from a small piston was given in Equation 10.12. Many
electrodynamic loudspeakers can be roughly modeled as a small piston set in a baffle
or at the end of a long tube. In most of its operating range, the piston movement is
approximately governed by its mass and the applied force. The frequency-response
characteristics of the real part of the radiation impedance is then of considerable
interest. Investigating the properties of real part of the radiation impedance of
most small radiators that have “closed” backs, we find that it is proportional to
ω2 at frequencies when the diameter of the diaphragm is less than a wavelength.
For a frequency-independent force excitation, the loudspeaker will then have
a “flat” frequency response for those frequencies. Using Equation 10.40, we can
find the frequency response of such a loudspeaker. This power response is shown
in Figure 10.24 along with that of a first-order Butterworth low-pass filter (see
Appendix B). As we will see in later chapters, baffle, cone breakup, diffraction, and
many other effects cause larger deviations from the smooth frequency response than
do the piston functions of Equation 10.40.
10.13 FINITE ELEMENT AND BOUNDARY ELEMENT METHODS

Numerical methods form an important possibility to calculate the radiation impedance
of various loudspeaker configurations and vibration modes. In many practical cases
it will be necessary to use these methods since the geometrical shape of the radiator
and its mounting can also be taken into account. Nevertheless, the approach used
here should give the reader a better insight into the acoustics of the radiation.
10
Power response
Equation 10.40
0
Power frequency response
–10
–20
Power response
1st order Butterworth
–30
–40
0.1 0.2 0.5 1 2 5 10
ka
FIGURE 10.24 The cutoff of the power frequency response of a circular, piston-type loud-
speaker due to the change in radiation impedance characteristics at high frequencies com-
pared to the frequency response of a first order Butterworth lowpass filter. Constant driving
force and mass-controlled diaphragm are assumed.
10.14 THE RADIATION IMPEDANCE OF CIRCULAR PISTONS
Radiation Impedance and Mobility for One Side of a Plane Piston in Infinite Baffle
Impedance Analogy
Frequency Mechanical Specific Acoustic
Range Impedance Impedance Acoustic Impedance Analogous Circuit
ka < 0.5
Series ℜM = 1.57ω2a4ϱ/c ℜS = 0.5ω2a2ϱ/c ℜA = 0.159ω2ϱ/c
resistance, ℜ
Radiation and Impedance
ZR M1 ZR M1
Shunt RM1 = 4.53a2ϱc RS1 = 1.44ϱc RA1 = 0.459ϱc/a2 R1
resistance, R1
Mass, M1 MMI = 2.67a3ϱ MSI = 0.849aϱ MAI = 0.270ϱ/a
ka > 5 RM2 = πa2ϱc RM2 = ϱc RA2 = ϱc/πa2
Resistance, R2
ZR M2
Mobility Analogy
ka < 0.5
Series rm1 = 0.221/a2ϱc rs1 = 0.694/ϱc rA1 = 2.18a2/ϱc
admittance, r1
ZR M1 r1
ka > 5 rM2 = 1/πa2ϱc rM2 = 1/ϱc rA2 = πa2/ϱc
Admittance, r2
ZR r2
Source: Beranek, L.L., Acoustics, American Institute of Physics, New York, 1986.
193
194
Radiation Impedance and Mobility for the Outer Side of a Plane Piston at the End of a Long Tube
Impedance Analogy
ka < 0.5
Series ℜM = 0.785ω2a4ϱ/c ℜS = 0.247ω2a2ϱ/c ℜA = 0.0796ω2ϱ/c
resistance, ℜ
Shunt resistance, RM1 = 4.73a2ϱc RS1 = 1.51ϱc RA1 = 0.479ϱc/a2 ZR M1 ZR R1 M1
R1
ka > 5 RM2 = πa2ϱc RM2 = ϱc RA2 = ϱc/πa2
Resistance, R2
ZR R2
Mobility Analogy
ka < 0.5
Series rM1 = 0.212/a2ϱc rS1 = 0.665/ϱc rA1 = 2.09a2/ϱc
admittance, r1
ZR M1 r1
ka > 5 rM2 = 1/πa2ϱc rM2 = 1/ϱc rA2 = πa2/ϱc
Admittance, r2
ZR r2
Electroacoustics
Radiation Impedance and Mobility for Both Sides of a Plane Circular Disk in Free Space
Impedance Analogy
ka < 0.5
Series ℜM = 1.89ω4a6ϱ/C3 ℜS = 0.600ω4a4ϱ/C3 ℜA = 0.0190ω4a2ϱ/C3
resistance, ℜ
Radiation and Impedance
Mass, M1 MMI = 2.67a3ϱ MSI = 0.850aϱ MAI = 0.270ϱ/a ZR M1

ka > 5 RM2 = 2πa2ϱc RM2 = 2ϱc RA2 = 2ϱc/πa2
Resistance, R2
ZR R2
Mobility Analogy
ka < 0.5
Series rM1 = 0.0265ω2/ϱC3 rS1 = 0.0832ω2a2/ rA1 = 0.0261ω2a4/ϱC3
admittance, r1 ϱC3
ZR M1 r1
Mass, M1 MMI = 2. 67a3ϱ MSI = 0.850aϱ MAI = 0.270 ϱ/a
ka > 5 rM2 = 1/2πa2ϱc rM2 = 1/2ϱc rA2 = πa2/2 ϱc
Admittance, r2
ZR r2
195
REVIEW QUESTIONS
10.1 Show that the radiation impedance seen by a small vibrating sphere can be
expressed as a parallel circuit of mass and resistance.
10.2 Show that the circuit under certain circumstances can be usefully expressed as
a series circuit. What is special about this series circuit’s components?
10.3 Describe the differences in the frequency-response behavior of the radiation
impedance for these cases: (a) circular piston in a plane infinite baffle, (b)
circular piston covering the end of a circular tube, (c) free circular piston.
10.4 What are the main impedance characteristics for the sound radiation by a rigid
piston as a function of its surrounding?
10.5 How does radiation from (orthogonal) transverse waves in plates differ from
that of radiation by rigid pistons?
10.6 How is the loss factor of a plate related to the radiation impedance?
PROBLEMS
10.1 For a plane circular piston in an infinite baffle, the reactive part of the radia-
tion impedance can be seen as a mass adhering to the piston. This mass can
be seen as an air cylinder adhering to the piston.
Tasks:
a. How high in frequency is this approach valid?
b. Calculate the mass for a piston with diameter 0.2 m.
c. What is the height of this air cylinder?
d. Compare this value with the end correction used for the air column of a
Helmholtz resonator.
e. Calculate the end correction if the plane circular piston is placed at the
end of an infinite rigid cylinder of the same diameter as the piston.
10.2 A driver having a rigid piston of radius a can (theoretically) be placed in an
infinite baffle or at the end of a long tube. Assume that ka < 0.5 and that the
piston vibration is unaffected by the mounting.
Tasks:
a. What is the difference in sound power level for the two cases?
b. What is the difference in sound pressure level in the far field?
10.3 Assume that a driver diaphragm is mass-controlled and has a diameter of
0.2 m and a mass of 0.01 kg. The loudspeaker may be mounted in an infinite
baffle or at the end of a long tube.
Tasks:
a. What is the difference in sound power level for the two cases?
b. What is the difference in sound pressure level in the far field?
10.4 A driver having constant diaphragm velocity u is radiating a noise signal into
a room through a cavity and a duct (see figure below) causing a resonance at
some low frequency. Assume that the room and the duct are free from losses,
and that the room is very large. Do not consider modes in the cavity or room.
Cavity Duct Room
V = 1 m3 Diameter = 0.1 m
0.5 m
Task:
At which frequency will the sound transmission into the room be at its
maximum, and what will be the Q-value of this resonance?
10.5 In studies of commercial drivers one once found that there was a covariation
between nominal driver diameter d (m), maximum diaphragm displacement
xpeak, and maximum sound pressure level at 1 m distance as follows:
x peak = 2 ⋅ 10 −2 d [m ]
(10.59)
L p,max = 97 + 40 d [dB]

Task:
Derive an expression for the maximum sound pressure level of the direct
sound as a function of d and frequency when the diaphragm displacement is
the limiting factor. Assume the driver mounted at the end of a very long tube
that has diameter d and that d ≪ λ.
REFERENCES
1. Meyer, E. and Neumann, E.-G., Physical and Applied Acoustics, Academic Press,
New York (1972) ISBN-13: 978–0124931503.
2. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986). ISBN-13:
978–0883184943.
3. Skudrzyk, E. J., Foundations of Acoustics, Basic Mathematics & Basic Acoustics,
Springer, New York (1972) ISBN-13: 978–0387809885.
4. Abramowitz, M. and Stegun, I. A., Handbook of Mathematical Functions: With
Formulas, Graphs, and Mathematical Tables, Dover Publications, Mineola, NY (1965)
ISBN-13: 978–0486612720.
5. Mechel, F. P. et al., Formulas of Acoustics, 2nd edn., Springer, Berlin, Germany (2008).
ISBN-13: 978–3540768340.
6. Cremer, L. et al., Structure-Borne Sound: Structural Vibrations and Sound Radiation
at Audio Frequencies, 3rd edn., Springer, Berlin, Germany (2005) ISBN-13:
978–3540226963.
7. Bank, G. and Wright, J. R., Radiation impedance calculations for a rectangular piston,
J. Audio Eng. Soc., 38(5), 350–354, (1990).
8. Ver, I. L. and Beranek, L. L. (Eds.), Noise and Vibration Control Engineering: Principles
and Applications, Wiley, New York (2005) ISBN-13: 978–0471449423.
11 Sound Source and
Acoustic Environment
11.1 REFLECTING SURFACES AND RADIATION IMPEDANCE

Most electroacoustic transducers will be used in a sound-reflecting environment.
The influence of the room and particularly its surfaces on a sound source have long
been of interest to both engineering and science [1–7]. In the simplest theoretical
case the environment may be that of a single, plane, rigid, infinitely large reflecting
surface. When more surfaces are present or when the surfaces have limited size
or impedance, the mathematical theory of the influence of the environment on the
transducer becomes more complicated. It will then usually be difficult or impossible
to find an analytical solution to the problem, and one will have to resort to numerical
solution methods.
In this chapter, we will study the influence of radiation impedance and power
radiation of various arrangements of sound-reflecting surfaces of various cases
such as a large and rigid reflecting surface to that of a rectangular room. We will
also study the influence of neighboring radiators, such as when two loudspeakers
work in pair.
11.2 SINGLE RIGID PLANE SURFACE

A characteristic of the sound field next to a rigid reflecting plane surface is that the
particle velocity perpendicular to the surface is zero. One way of mathematically
modeling the sound field around a sound source in the presence of a large, plane, and
rigid reflecting surface is to fulfill the boundary condition by assuming the presence
of a mirrored monopole on the other side of the boundary, at the same distance from
the boundary as the real source. Such a source is often called a mirror source and is
conveniently assumed even if the surface is not ideally plane or rigid. In such cases,
there will also be sound absorption and sound scattering.
It is convenient to write the influence of the mirrored source on the acoustic
radiation impedance felt by the monopole source in the form
Z AR = Z AR
′ + Z AR
′′ (11.1)
where
Z′AR is the radiation impedance seen by the source in “free space”
Z″AR is the additional radiation impedance due to the presence of the reflecting
surface
199
A monopole source is assumed to have high internal impedance, that is, its vol-
ume velocity is independent of its acoustic environment. According to Chapter 3, the
free field sound pressure at a distance r from a monopole having a volume velocity
U is given by
e − jkr
p (r, k ) = jωUρ (11.2)
4πr
At low frequencies, ka ≪ 1, the monopole, having a radius a, senses a radiation

impedance in the free field according to Equation 11.3 (the same as Equation 10.29).
ρ 2 ρ ρω 2  j
′ (a, k ) = ℜ′AR + jX AR
Z AR ′ ≈ ω +j ω=  1 +  (11.3)
4πc 4πa 4πc  ka 
This impedance is sometimes called the self-impedance of the source. The self-
impedance may be expressed as mechanical or sound-field impedance as well.
As explained in Chapter 3, the boundary condition at the rigid surface is that the
particle velocity at a right angle to the boundary must be zero. This can be achieved
by assuming a mirrored source of the same volume velocity on the normal of the
surface. The source is at a distance d from a plane, rigid, infinitely large reflecting
surface. The mirrored source is the same distance from the surface as the “true”
source; so the distance between the sources is 2d. The particle velocity contributions
of both sources will add up to zero at the boundary. Figure 11.1 shows the real mono-
pole source next to a rigid surface and the resulting mirror image.
The sound pressure at the source is composed of two components, the sound
pressure p′ resulting from the radiation by the source itself and the sound pressure
p″ that results from the mirror source. Since we have assumed the source to be small
compared to a wavelength, the sound pressure contribution from the mirrored source
will be the same around the source. This sound pressure is
d d
s˝ s΄
FIGURE 11.1 A monopole s′ and its image source s″, both at a distance d from the reflecting
surface. Circle segments indicate instantaneous wave fronts of the direct and reflected waves.
Sound Source and Acoustic Environment 201
e − j 2 kd
p′′ (r, k ) = jωUρ (11.4)
4π ⋅ 2d
We can write this sound pressure in a way similar to that of Equation 10.29 as
ρω 2 e − j 2 kd
p′′AR (r, k ) = ⋅ U (11.5)
4πc − j 2kd
So, the impedance contribution Z″AR of this source is
ρω 2 e − j 2 kd ρω 2  sin(2kd ) cos(2kd ) 
′′ ( d, k ) = ℜ′′AR + jX AR
Z AR ′′ ≈ = +j (11.6)
4πc − j 2kd 4πc  2kd 2kd 
Figure 11.2 shows a plot of the ratio of real parts ℜ″AR /ℜ′AR as a function where
ℜ′′AR sin(2kd ) (11.7)

=
ℜ′AR 2kd
X AR
′′ a
= cos(2kd ) (11.8)
X AR
′ 2d
0.5
Ratio of real parts ˝/ ΄
–0.5
–1
0 1 2 3 4 5
kd
FIGURE 11.2 The behavior of the acoustic radiation resistance ratio ℜ″AR /ℜ′AR.
We note that the reactive part of Z AR does not change much when d is reduced
relative to wavelength since the ratio of the reactances X″/X′ is always less than
a/d ≪ 1, but that the real part is doubled. When the distance to the wall is very small,
that is, 1 ≫ kd ≫ ka, the radiation impedance sensed by the source will be twice
that in free space:
Z AR (a, k ) = 2 Z AR
′ (11.9)
The radiated power P can be determined from the product of the square of the
volume velocity and the sum of the real parts of the radiation impedances and
will be
ρω 2  sin(2kd )  2
P(d, ω) = ( ℜ′AR + ℜ′′AR )U 2 ≈ 1+ U (11.10)
4πc  2kd 
Typically, the volume velocity will be independent of the radiation load because
of the high internal impedance of the transducer. This is, for example, the case for
electrodynamic transducers such as conventional loudspeakers. The ratio of the
radiated power P1 will then be related to the power radiated by the monopole in
free field P0 as
P1 (d, k ) sin(2kd )
= 1+ (11.11)
P0 2kd
11.3 MULTIPLE SURFACES

Two other cases of interest are how the radiation impedance is affected by the pres-
ence of additional rigid planes such as at corners between two and three orthogonal
planes, as shown in Figure 11.3.
dxyz
dxy
dxy
(a) (b) (c)
FIGURE 11.3 Three different configurations of infinite, rigid reflecting planes: (a) plane—
the dx line is on the normal to the plane; (b) corner between two orthogonal planes—the
dxy line is on the diagonal between the two planes; and (c) corner between three orthogonal
planes— the dxyz line is on the space diagonal between the three planes.
S΄
S˝y
dxy
S˝xy and S˝yx S˝x
FIGURE 11.4 A monopole S′ and its image sources S″x, S″y, S″xy, and S″yx, all at a distance
d = dxy /√2 from the reflecting surfaces at a right angle corner. Circle segments indicate instan-
taneous wave fronts of the direct and reflected waves.
Figure 11.4 shows the geometry of the case of two orthogonal planes in more
detail. We study the sound pressure on the line of symmetry between the planes, but
the analysis can of course be done similarly for any other location. The impedance
behavior on the line of symmetry is simply the most extreme. In the same way as
in the previous section, we add up the sound pressure contributions from the mirror
image sources.
Because of the presence of the four extra mirror images (of which two coincide),
the real part impedance ratios on the diagonal dxy (see Figure 11.3) will be [2]
ℜ′′AR
ℜ′AR
= 2 j0 ( )
2 kd xy + j0 ( 2kd xy ) (11.12)

Here, j0(x) is the spherical Bessel function of the first kind and 0th order
sin ( x )
j0 ( x ) = (11.13)
x
The ratio of the radiated power will then be related to the power radiated by
the monopole in free field as shown in Figure 11.5. We see from the graph that as
the number of reflecting surfaces increase, the radiation impedance will increase at
close distance between the monopole and the surfaces.
Similarly, in the presence of the extra mirror images (of which several coincide) in
the case of a corner between three orthogonal planes, the real part impedance ratios
on the space diagonal dxyz (see Figure 11.4) will be [2]
12
(c)
9
Power level relative to source in free field [dB]
(b)
6
(a)
3
–3
–6
–9
–12
0 0.25 0.5 0.75 1
FIGURE 11.5 The power level of a monopole at a distance (a) dx /λ (b) dxv /λ (c) dxvz /λ from
a rigid plane, on the diagonal between two planes, and on the space diagonal between three
planes relative to free field conditions.
ℜ′′AR  2   2 
= 3 j0  kd xyz  + 3 j0  2 kd xyz  + j0 ( 2kd xyz ) (11.14)

ℜ′AR  3   3 
The sound power level of the source and mirror image source contributions
will vary with distance to the reflecting surfaces in the same way as the radiation
impedance. In the cases of dx = 0, dxy = 0, and dxyz = 0, the sound power levels
increase by 3, 6, and 9 dB relative to the free field value as shown in Figure 11.5.
This effect of corner placement of sound sources is sometimes used to enhance
the low-frequency output of a loudspeaker by placing the loudspeaker in a corner.
A corner between large walls will always enhance low-frequency sound radiation.
It is of course important to separate between the impedance generated by the mirror
images of a few planes and that generated by the infinite number of mirror images
of a room. The latter results in the modal behavior of the sound field, as discussed in
Appendix F and in Section 11.5.
11.4 POWER OUTPUT OF DIPOLES NEAR REFLECTING SURFACES

In electroacoustics, we often find devices that have bidirectional or even dipole
characteristics such as electrostatic and isodynamic loudspeakers. These are affected
similarly by the presence of reflecting planes. Two instructive cases are the dipole
3
(a)
Power level relative to dipole in free field [dB]
–3
–6 dx dx dx dx
(b) S˝ S΄
S˝ S΄
–9 Mirrored Dipole Mirrored Dipole
dipole dipole
(a) (b)
–12
0 0.25 0.5 0.75 1
x/λ
FIGURE 11.6 The radiated power of a dipole at a distance x from a rigid plane relative to
free field conditions for two different alignments of the dipole relative to the reflecting plane:
(a) dipole parallel to and (b) dipole perpendicular to the plane. When the dipole is perpen-
dicular to the plane the power output drops to zero when the dipole comes close to the plane.
parallel and at right angle to the wall. In both cases, the dipole forms a quadrupole
with its mirror image. Figure 11.6 shows the dipole and its mirrored configurations
for the two cases.
Case (a) is intuitive since the mirrored dipole is aligned parallel to the original
dipole and with its polarity in the same direction. For this case, the power output P
of the dipole follows:
P
= 1 + j0 ( d x ) + j2 ( d x ) (11.15)
Pff

Here, jn(x) is the spherical Bessel function of the first kind and nth order. For
the case of the dipole perpendicular to the plane, the power output P relative to
free field Pff is
P
= 1 − j0 ( d x ) + 2 j2 ( d x ) (11.16)
Pff

In some cases the loudspeakers will not be conventional loudspeaker enclosures,

which usually radiate like monopoles at low frequencies. Push–pull electrostatic
and ribbon-style loudspeakers are generally designed to work as dipoles, since the
diaphragm motion would otherwise be attenuated by the acoustic impedance of a

box. A dipole-type electrostatic loudspeaker will often be placed at almost right
angle to a nearby wall. The reason for this is clear from the inset graph in Figure
11.6. Because of the directionality of the dipole action, the sound power output of a
dipole will depend on how it is directed relative to the reflecting surface. The two
curves in Figure 11.6 indicate that a dipole-type loudspeaker placed with its back
directed toward a reflecting surface would have very little sound power output at low
frequencies since the dipole then becomes a quadrupole with its associated reduced
power radiation.
11.5 ROOM MODES

The acoustic input impedance to a room will be more complicated than that obtained
by just adding the influence of seven or more corners to the one studied previously.
The previous solution only applies, because we added the relevant number of mirror
images. Even in an imaginary six-walled rectangular room, the sound field is made
up of contributions from an infinite number of mirrored sources. The sound pressure
contributions from these sources result in the modal behavior of the sound field in a
room discussed in Appendix F.
The sound pressure from a monopole in a resonant room can be found using
Equation F.19. This equation gives us the sound pressure sensed by a monopole at
point x0;y0;z0 in the room as a result of the volume velocity injected at some other
point. To find the impedance at that point, the source and observation points are
assumed to be the same.
As discussed in Appendix F, for a rectangular room with small losses, one corner
at the origin and its sides along the positive coordinate axes, we would obtain the
complex sound impedance Z AR as [3]
Ψq2x, qy , qz ( x0 , y0 , z0 )
∞ ∞ ∞
jωρc 2
Z AR ( x0, y0, z0, ω ) =
V
 2 ∑∑∑
qz Λ q x , q y , qz  ω q x , q y , qz − ω
2
( )
+ 2 jω q x , q y , q z δ q x , q y , q z 

qx qy 
(11.17)
where
qx,qy,qz are triple nonnegative integers representing the mode number
V is the room volume
δqx,qy,qz is the damping coefficient associated with each mode
Ψqx,qy,qz and ωqx,qy,qz are the eigenfunction and eigenfrequency, respectively, for
each mode in the room
The constant Λqx,qy,qz depends on the mode number in the following way: if the mode
is created by waves moving perpendicular to one plane, that is, only one q is nonzero,
then Λ = 1/2, by waves moving in parallel to two planes, that is, only one q is zero,
then Λ = 1/4, and for waves moving in three dimensions, Λ = 1/8.
The problem inherent in this approach is to estimate the damping of the modes,
that is, to find the damping coefficients δqx,qy,qz as a function of the wall impedances.
A practical way of estimating the damping coefficients is to measure the half-widths

or reverberation times of the resonances as described in Appendix F.
In some cases, the acoustic impedance of the room may not be low compared to
that of the loudspeaker, and in those cases it will be necessary to correct the volume
velocity of the speaker by taking the acoustic load impedance of the room on the
loudspeaker into account. An expression for the acoustic impedance seen by a mono-
pole in a rectangular room with one nonrigid wall having a reflection factor r is [4]
 q πy   q πz   q πy 
cos2  y 0  cos2  z 0  cos  x 0  cos ( k x x0 )
 Ly   Lz   Lx 
∞ ∞ ∞
jρck
Z AR ( x, y, z, k ) =
V ∑∑∑
qx qy qz (
Λ qx , qy, qz kq2x , qy , qz − k 2 (1 − jη) )
⋅
⋅
(e jk x L x
− re − jkx Lx − (1 − r ) cos ( nx π ) ) (11.18)

(e jk x L x
− re − jk x L x
)

Here, k x is
2 2
 q π  q π
k x = k 2 (1 − jη) +  y  +  z  (11.19)
 L y   Lz 

and η, is the loss factor. The loss factor depends on the damping of the modes. The
relationship between loss factor and reverberation time is given by Equation F.18.
Equation 11.18 is useful since it allows us to calculate Z AR for a rectangular room
having an acoustically soft ceiling which is a common situation. Agreement with
more complex methods has been shown to be good for some trial cases [4].
If the damping is low, this equation is an alternative to Equation 11.17. It is rea-
sonable to assume that if all the six walls in the rectangular room are nonrigid and
the damping by each is low, the impedance seen by the source will be that of six
impedances in parallel, each impedance representing the influence on the radiation
impedance by the respective wall.
Usually a large number of modes are excited even at low frequencies, and by sum-
ming the pressure contributions over a reasonable number of modes by using the ear-
lier expressions, it is possible to analyze the interaction between the loudspeaker and
the sound field in a rectangular room for a pure tone or over some small bandwidth.
For more exact results the acoustic input impedance of the room has to be found by
a modal approach, either analytically or numerically using finite element modeling.
The expression for the output of a monopole in the case where the monopole is
positioned in a rectangular room under the assumption of high modal density and
narrow band noise signals is as follows [5]:
P
= 1 + j0 (2kx ) + j0 (2ky) + j0 (2kz) + j0 (2kξ1 ) + j0 (2kξ 2 ) + j0 (2kξ3 ) + j0 (2kr )
Pff

(11.20)
z
lz
x
lx
3.5 m 1.2 m 2.9 m
lx, ly, lz = 5.45, 4.85, 3.55 m ly

y
= slit absorbers
FIGURE 11.7 The room used for tests. Walls without absorbers were gloss painted
concrete.
where the effective distances ξ1, ξ2, and ξ3 are related to the coordinates of the source
relative to one of the room corner at the origin, at x, y, and z as
ξ12 = x 2 + y 2 ; ξ 22 = x 2 + z 2 ; ξ32 = y 2 + z 2 ; r 2 = x 2 + y 2 + z 2 (11.21)

Measurement of the radiation impedance can be done using an electrodynamic
loudspeaker that has high internal impedance due to a rigid, large mass diaphragm.
The vibration velocity of the diaphragm can be measured using an accelerometer,
and the sound pressure in front of the diaphragm using a microphone. Figure 11.7
shows a drawing of a chamber. The measured |Z AR | in the frequency range of
low modal density is shown in Figure 11.8. We note that the impedance tends
toward a mean value at high frequencies because of the large number of modes
in the room [6].
11.6 MUTUAL IMPEDANCE

Sometimes two sound sources will be coherent, that is, vibrating with the same fre-
quency. If the two sources are at close distance, the presence of the sound field of the
second source will affect the radiation impedance of the first source and vice versa.
An example is the ported loudspeaker box where the radiation from the port will
affect the radiation impedance seen by the loudspeaker driver. A second example is a
column loudspeaker. The column loudspeaker uses an array of loudspeaker drivers,
all supplied by the same source. The individual drivers may be adjusted for phase
and amplitude, but the presence of the neighboring drivers will affect the radiation
impedance seen by a driver. The radiation impedance will consequently be different
at the ends and at the middle of the array [7].
105
|Zspkr|
|Z| [Ns/m5]
104
|Zroom|
103
40 100 150
Frequency [Hz]
FIGURE 11.8 Modulus of the acoustic source impedance for loudspeaker and modulus of
acoustic radiation impedance into room at approximately x = y = z = 0. The reverberation time
of the modes was in the range 0.8–2.0 s.
The mutual impedance is defined as the force on one source due to the sound
pressure generated by the second source in analogy with the added radiation
impedance in Equation 11.1 due to the mirror image source.
F12
Z MR1
′′ ,2 =
u 2 (11.22)

Two cases are instructive: two sources operating at the same velocity or two
sources with different velocities. The first case is similar to that of the monopole
close to a rigid plane surface and reminds us of loudspeaker drivers mounted in
separate boxes in a two-element array, driven by the same amplifier. The second case
reminds us of a ported loudspeaker box where the radiation of the port influences the
impedance seen by the driver.
Typically we are interested in the impedances seen by pistons rather than by
monopoles. We take the piston in the rigid baffle shown in Figure 10.6 and add
another identical piston at some small distance d as shown in Figure 11.9. The self-
impedance of the pistons in Figure 11.9 will be the same as of that in Figure 10.6.
We use mechanical impedance in Equation 11.23 since it makes more sense as the
pressure varies over the pistons
 J ( 2 ωa / c )  Sh (2ωa /c) 
Z MR (a, ω ) = ℜ MR + jX MR = πa 2ρc  1 − 1  +j 1 (11.23)
  2 ωa / c  2ωa /c 

Let us start by investigating the case of only one piston. The sound pressure
generated by a vibrating piston is given by Equation 10.38. At low frequencies where
ka ≪ 1, the radiation impedance is dominated by the reactance and is approximately
z
Observation point
a a y
Piston 1 Piston 2
x
d
FIGURE 11.9 Two pistons in a baffle influencing one another by their near-fields.
1 2 8  8
Z MR (a, ω) = ℜ MR + jX MR ≈ πa 2ρc  ( ka ) + j ka ≈ j ρcka 3 (11.24)

2 3π  3
Using the definition of mechanical impedance, this gives us the sound pressure of
the piston at low frequencies as
8ka
p1′(a, k ) = jρc u1 (11.25)
3π
We will show in Chapter 12 that the sound pressure in the far field of a monopole
(or small loudspeaker) on a rigid plane is given by Equation 12.4, which at low fre-
quencies can be approximated as
e − jkr kπa 2 − jkr

p1′(r , k ) = jωU1ρ = jρc u1e (11.26)
2πr 2πr
Now let us study the mutual impedance between the pistons. When the pistons are
small, ka ≪ 1, and the distance d between them is also small, kd ≪ 1, but still a ≪ d,
we find using Equation 11.26 that the pressure at piston 1 due to the vibration u2 of
piston 2 is
kπa 2
′′ (d, k ) = jρc
p12 u2e − jkd (11.27)
2πd
If one rewrites the exponential into real and imaginary parts, one finds an
expression very similar to Equation 11.6 [7]:
( )
2
ρck 2 πa 2  u2 sin(kd + ϕ) u2 cos(kd + ϕ) 
′′ 12 ( d, k ) ≈ ℜ′′MR12 + jX MR
Z MR ′′ 12 =  u +j 
2π  1 kd u1 kd 

(11.28)
Here, we have assumed that the phase of u2 lags that of u1 by an angle ∏. Similar
equations can be derived for the case of drivers in small loudspeaker boxes, such
as pistons that are set at the end of long tubes. In practice, however, Equation 11.6
is sufficient to estimate the influence of the second driver on the power radiation
since the driver diaphragm diameters are small compared to the distance between
their centers.
It is clear from Equation 11.28 that the mutual impedance for small values of ka
has a mass-type character since its value is proportional to jk. Whether the mass is
added or subtracted depends on the phase of u2 relative to u1.
REVIEW QUESTIONS
11.1 How does the presence of nearby sound-reflecting planes influence the sound
radiation by a constant volume velocity sound source?
11.2 Describe the differences in the frequency response behavior of the radiation
impedance for these cases of loudspeakers:
a. circular piston covering the end of a circular tube close to reflecting plane
b. free circular piston close to reflecting plane
11.3 How is the impedance load on a piston affected by the presence of an identi-
cally vibrating piston at some distance?
11.4 Gradient loudspeakers can be considered combinations of omnidirectional
and bidirectional loudspeakers. How will the power output of such loudspeak-
ers be affected by the presence of reflecting planes?
11.5 Why is there a difference between the power output of a monopole source near
a corner between three planes only and the corner of a resonant room?
11.6 How is the sound radiation from a small sound source near a reflecting plane
compared with the sound radiation by two adjacent pistons?
PROBLEMS
11.1 A loudspeaker is mounted on a pedestal at the corner of a large room. The
sound from the loudspeaker turns out to have a strong peak at 100 Hz because
of resonance.
Task:
How far out from the corner should the loudspeaker be to reduce the peak as
much as possible?
11.2 Two loudspeaker drivers that are very small and that act as constant volume
velocity sources are placed 10 cm from one another. Assume the loudspeakers
to be at some distance from a rigid plane. The loudspeakers are connected to
the amplifier so that they radiate in anti phase. Assume that the midpoint of
the line between the loudspeakers is 0.3 m above the plane.
Task:
For an 88 Hz tone, what will be the difference in radiated power from the
loudspeakers when mounted parallel to the surface from that when mounted
normal to the surface? (Compare to problem 3.11.)
11.3 Two identical small loudspeaker drivers are flush-mounted in a rigid plane
and driven in parallel by the same amplifier. When only one loudspeaker is
mounted on the array, the sound pressure level on the normal to the plane in
the middle between the drivers at 10 m distance is 94 dB. Assume the drivers
to act as constant volume velocity sources. The distance between the driv-
ers is assumed to be large compared to their diaphragm diameters, but small
compared to 10 m. What will be the sound pressure level when both drivers
are radiating as a function of the distance between the drivers. The frequency
of the sound is 250 Hz.
11.4 A loudspeaker—that is very small and acts as constant volume velocity
source—is placed on the space diagonal between three orthogonal rigid walls
in a room that is nonresonant. The influence of the walls on the power out-
put of the loudspeaker can be inferred from Equation 11.14. The frequency
response of the power radiated by the loudspeaker Pff is given by
1
Pff ( ω ) = 2 (11.29)
2 2
  ω0   1  ω0 
1 −
    +
 ω  Q 2  ω 

Tasks:
a. Determine the frequency at which the power output is minimum if the
loudspeaker is placed 0.6 m from the corner.
b. The loudspeaker has a frequency response given by the equation given
earlier, so it is resonant at some low-frequency ω0. Adjust the frequency
of the resonance response peak so that it occurs at the frequency at which
the power response of the room is minimum. Determine the Q-value of
this peak if the response is to be the same as that at high frequencies.
c. For this Q-value, calculate the power frequency response of the loud-
speaker in the free field and at different corner distance to wavelength
ratios r/λ when the loudspeaker is placed on the space diagonal between
the walls.
11.5 The radiation impedance sensed by a monopole near a rigid plane wall is
given by
Z AR = Z AR
′ + Z AR
′′ (11.30a)
where Z′AR is the radiation impedance seen by the source in “free space” and
Z″AR is the additional radiation impedance due to the presence of the reflecting
rigid plane wall.
The radiation impedance sensed by a monopole in the free field is given by

Equation 11.3. Now assume the monopole to be placed in a resonant room
and only consider the lowest frequency mode (which in this case has wave
motion along the x-axis). With the monopole placed at xsource the sound
pressure at xrec is
 x   x 
cos  π source  cos  π rec 
2
ρc T60  lx   lx 
p( xsource , xrec , ω) ≈ U (11.30b)
13.8V ω T  ω ω 
1 − j 1,0,0 60  − 1,0,0 
13.8  ω1,0,0 ω 

where
U is the volume velocity of the monopole
ω1,0,0 is the resonance frequency of the mode
lx is the length of the room in the x-direction
T60 is the reverberation time of the room
V is the room volume
Task:
Calculate the power radiated by the loudspeaker in the room compared to that
radiated in free field assuming the loudspeaker placed at xsource = xrec = 1 m, that
the room volume V = 60 m3, lx = 6 m, and the reverberation time of the mode
T60 = 1 s.
REFERENCES
1. Klipsch, P. W., Corner speaker placement, J. Audio Eng. Soc. 7(3), 106–109 (July1959).
Reprinted in Loudspeakers, An Anthology, Part I by Audio Engineering Society (1980).
2. Waterhouse, R. V., Output of a sound source in a reverberation chamber and other
reflecting environments, J. Acoust. Soc. Am., 30(1) (1958).
3. Morse, P. M. and Ingard, K. U., Theoretical Acoustics, Princeton University Press,
Princeton, NJ (1987) ISBN-13: 978–0691024011.
4. Nilsson, E., Decay Processes in Rooms with Non-Diffuse Sound Fields, Report TVBA-
1004, Department of Engineering Acoustics, Lund University of Technology, Lund,
Sweden (1992).
5. Maling, G., Calculation of the acoustic power radiated by a monopole in a reverberation
chamber, J. Acoust. Soc. Am. 42(4), 859–865 (1967).
6. Kleiner, M. and Lahti, H., Computer prediction of low-frequency SPL variations in
rooms as a function of loudspeaker placement, Proceedings of the 94 Audio Engineering
Society Convention, Berlin, Paper 3577 (1993).
7. Jacobsen, O., Some aspects of the self and mutual radiation impedance concept with
respect to loudspeakers, J. Audio Eng. Soc., 24(2) (1976).
12 Directivity
12.1 INTRODUCTION
Directivity is the property used to describe the uneven angular intensity distribution
for a transmitter and the uneven angular sensitivity for a receiver. Directivity can be
achieved in two ways: by suitable geometrical properties, distribution of sources/
receivers, and by change of vibration (or sensitivity to vibration) phase and amplitude.
The directivity can be studied both with frequency-domain and time-domain
analyses. We will use the frequency-domain approach in this chapter. The time-
domain approach is discussed in Appendix D.
A transmitter may consist of several radiators such as loudspeaker drivers, and
a receiver of several sensors such as microphones. We usually call such aggregates,
where several units work at the same time, arrays, and we will also study the properties
of such arrays. The directivity of an array will be a function of the directivity of
each individual sensor but also of the arrangement of transducers in the array and
the acoustic interaction between the transducers, for example, due to mutual radiation
impedance or scattering. We will not consider the latter two factors. Note however
that a loudspeaker that has several drive units where each unit is intended for its
specific frequency range is usually not called an array.
We will also study the differences between sound in the near-field and in the far-
field of a transducer and how these terms are used. Finally, we will consider the idea
of an acoustic center.
12.2 DIRECTIVITY FUNCTIONS AND DIRECTIVITY PLOTS

The directivity function F(θ,φ) is a way of describing the directional properties of
a transducer. Except for the theoretical monopole or ideal pressure sensor, every
transducer will have an associated directivity function. The directivity function for
any practical transducer will be frequency dependent.
The coordinate system in which the directivity function is defined is shown in
Figure 12.1. In this book, the z-direction is usually defined as the transducer’s main
axis of radiation or reception.
For a sending transducer, the directivity function F(θ,φ) is the relationship between
the sound pressure in the far-field at angles (θ,φ) compared to the sound pressure in a
reference direction, usually (θ,φ) = 0,0 for the same distance to the acoustic center of
the transducer. The directivity function (θ,φ) for a receiver is the relationship between
the receiver’s electric output voltage in the far-field at angles (θ,φ) compared to the
output for incoming sound at a reference direction, usually (θ,φ) = 0,0, and for the same
distance to its acoustic center. In practice, it is also difficult to measure the phase; so
F(θ,φ) = |F(θ,φ)| is generally measured and used. Usually the directivity function is
normalized relative to the direction of maximum sensitivity so that 0 ≤ F(θ,φ) ≤ 1.
215
z z
Observation
Incident plane point
sound wave Radiated
sound
r ∞
Diaphragm θ Diaphragm θ
y y
x x
FIGURE 12.1 Definition of angles for the directivity function of a transducer. Left figure
shows receiving case, right shows transmitting case.
It is difficult to measure the directivity function accurately. For example, the location
of the transducer’s acoustic center may be frequency dependent. Usually, an anechoic
chamber is necessary although modern time-gating impulse response measurement
techniques may remove that need above a certain frequency. For a large transducer, it
may be difficult to be at a sufficiently large distance and for a small transducer it may
be difficult to achieve sufficient sensitivity. In both cases, the signal, in directions of
low sensitivity, may be contaminated by mechanical, acoustical, and electric noise.
The acoustic “noise” may be composed of both ambient noise and reflected sound
from the surroundings.
Directivity plots are used to graphically display the directional properties of the
transducer. The directivity plots showing D(θ,φ) are usually based on

( )
D (θ, ϕ ) = 20 log F (θ, ϕ ) (12.1)
Three types of directivity plots are shown in Figure 12.2. Figure 12.2a shows the
traditional polar plot, Figure 12.2b the alternative linear plot, and Figure 12.2c a
“directivity balloon” of the type often used to show the directivity characteristics of
complex transducers such as array loudspeaker systems.
The directivity function is used to describe the properties of narrow-band
transmitters and receivers such as those often encountered in sonar applications.
In audio, the frequency response of the transducer in various directions is of more
interest since the frequency response allows a better understanding of the relative
directional characteristics of the transducer. The frequency response in the relevant
directions relative to the integrated response over all angles is often a useful indicator
of the transducer’s directional properties.
We studied the properties of small transducers in Chapters 3 and 11 and noted that
small transducers have monopole characteristics if only one side of the transducer is
facing the exterior sound field. Examples of such transducers are small closed box
Directivity 217
0°
–10 dB
–20 dB
–30 dB 0 dB
–40 dB
–90° 90°
(a) 180°
0
–10
20 Log [F(θ)]
–20
–30
–40
–50
–180 –90 0 90 180
(b) Angle θ [°] cylindrical symmetry assumed
(c)
FIGURE 12.2 Three graphical representations of the directivity characteristics of the

same transducer. (a) shows a traditional polar plot, (b) an alternative “linear” plot, and
(c) a “directivity balloon”.
loudspeaker units and most pressure microphones. As long as their dimensions are
much smaller than a wavelength, that is, kd ≪ 1, they will be virtually omnidirectional.
Since most microphones are small, we usually do not have to consider their influence
on the sound field by scattering or diffraction of incoming low-frequency sound.
Large transducers, such as loudspeakers, have more complicated directivity
characteristics that cannot be described by simple functions and they may be
difficult to measure. At sufficiently high frequencies, any transducer will need

to be considered large.
12.3 RECIPROCITY
By directivity, it is meant that the sound intensity at a far distance from a transmitting
transducer such as a loudspeaker varies with the viewpoint. The directivity of a
transmitting receiver varies with the viewpoint in the same way for an incoming
plane sound wave. This is called reciprocity and is a general property for acoustic
systems. The electroacoustic reciprocity theorem, rephrased from Ref. [1] is stated
as: If a monopole of volume velocity U1 at a point A produces a sound pressure
p2 at point B, then a monopole of volume velocity U2 at a point B produces a sound
pressure p2 at point A such that
U1 U 2 (12.2)
=
p2 p1

The principle of reciprocity as discussed in Chapter 3 may be used to great advan-

tage in electroacoustics, particularly in measurement and calibration.
12.4 MONOPOLE ON A RIGID BAFFLE

In Chapter 11, we studied the sound radiation for the case of a monopole in free
space close to rigid surfaces. In Chapter 3, we found that the sound pressure at a
distance r from a monopole in free space having a volume velocity U is
e − jkr
p (r,ω ) = jωUρ (12.3)
4πr
The monopole is near a rigid plane, so the sound at some distance will be the
sum of the sound from the monopole and its mirror image. Because of linearity, we
can again use the principle of superposition of pressure to determine the total sound
pressure at the observation point. Since each monopole is near a hard surface, the
principle of doubling of pressure due to mirror images also applies. The monopole
radiation is omnidirectional in a 4π solid angle. Because of the assumed infinitesimal
distance between the monopole and its mirror image in the rigid baffle, the radiation
of the pair will be omnidirectional although only in a 2π solid angle:
e − jkr
p (r,ω ) = jωUρ (12.4)
2πr
This means that the sound intensity is four times that of the monopole in free
space. We remember that the real part of the radiation impedance seen by the mono-
pole only doubled by the presence of the rigid surface; so, the additional doubling
of intensity comes from the reduction of the solid angle of radiation from 4π to 2π.
Directivity 219
12.5 NEAR-FIELD AND FAR-FIELD

The output of a monopole was considered already in Chapter 3 as were combinations
of simple sources such as dipoles and quadrupoles. In Chapter 10, we considered
the radiation impedance of more complex sources but we did not study the particle
velocity and sound pressure close to the vibrating surface for these cases.
Let us assume that a vibrating surface patch is a part of the plane of an otherwise
infinite rigid plane baffle in the z = 0 plane as shown in Figure 12.3. The surface
element at point rP is moving with velocity uP(rP) in the z-direction. We use the
shorthand r to describe the coordinates of the points.
The sound pressure at any point in the field away from the baffle, that is, z > 0, can
be calculated by superposition of the sound pressure generated by the distribution of
monopoles over the piston. Each monopole has a volume velocity
dU P = uP dS (12.5)

Following the same basic reasoning as in Chapter 10, Equations 10.36 through
10.38, we find that the sound pressure from a patch vibrating with velocity uP(rP) to
the normal of the plane is given by the Rayleigh integral
ρω uP (rP )e − jk r − rP
p(r0 , ω) = j
2π ∫
S patch
r − rP
dS1 (12.6)

Here, |r − rP| is the distance between the patch and the observation point. We
define three zones depending on their distance to the vibrating patch: (1) reactive
near-field zone, (2) Fresnel zone, and (3) Fraunhofer zone.
In the reactive near-field zone, the coupling between the field and the vibrating
surface is strong. The particle velocity is determined by the vectors generated by
the pressure gradient. We noted the behavior of sound in the reactive near-field zone
when we discussed the evanescent field from the sheet carrying bending waves.
z
Observation point
r
r = r – rp
rp
x
FIGURE 12.3 The (white) vibrating surface patch set in a (gray) rigid plane at z = 0.
A sheet characterized by flow resistance suitably placed in this field would cause
power loss and contribute to the damping of waves in the sheet. A reasonable esti-
mate for the depth of this field is about λ/2. This corresponds to the depth of the
region for which the radiation impedance is affected by a neighboring source.
In the Fresnel zone, the particle velocity has lost some of its importance and the
intensity vectors mainly point in the z-direction. In this zone, the sound pressure is
still varying considerably since the sound pressure contributions from different points
on the patch vary widely in phase so that quasi-plane wave fronts are not obtained.
In the Fraunhofer zone, usually called the far-field, the sound pressure contributions
are mainly in phase. Characteristic for the Fraunhofer zone is that sound pressure
drops with distance according to the geometrical distance law, that is, p(r) < 1/r, and
that sound pressure and particle velocity are in phase. This means that the sound
intensity is given by
p2 (r )
I r (r ) = (12.7)
ρc
In the far-field, the particle velocity vectors are directed radially away from the
origin as are of course the intensity vectors. Another characteristic of the far-field
is that the angular distribution of the intensity in independent of the distance to the
origin. One way of defining the onset of the far-field is to define it as starting at the
point where the difference between the actual rates of pressure drop is within 95% of
that given by the geometrical distance law.
Figure 12.4 shows an outline of the basic behavior of the sound pressure level as a
function of distance for some simple coherently radiating sound sources (equal phase
and amplitude over their surfaces). It is seen that the onset of Fraunhofer region is
determined by the largest dimension of the sources. We will look at the sound field
in more detail later in this chapter.
12.6 NEAR-FIELD OF A PISTON IN A BAFFLE

We now assume that we have a plane vibrating circular piston of radius a set at the
center of a plane and rigid baffle as shown in Figure 12.5.
The sound pressure over the baffle and piston is calculated using Equation 12.6.
Figure 12.6 shows the magnitude of the sound pressure in front of the piston in the
x = 0 plane for four ratios of wavelength to piston radius.
The sound pressure close to the surface of the piston has a complicated structure
for frequencies over that where the piston has a size comparable to that of the
wavelength. Figure 12.7 shows isobars near a piston for a case where ka = 10, that is,
the piston radius is about 1.5 λ.
12.7 FRESNEL ZONE OF A PISTON IN A BAFFLE

The difference between the near-field and Fresnel zones becomes more pronounced
as we see from the graphs in Figure 12.6 where the sound pressure varies widely
Directivity 221
Point source –6 dB/dd

r
Line source log (r)
I
–3 dB/dd
r
a
–6 dB/dd
Plane source
2a2/λ log (r)
a
r –3 dB/dd
b –6 dB/dd
b<a
2b2/λ 2a2/λ log (r)
FIGURE 12.4 Attenuation of sound intensity as a function of distance as shown for some
types of sources. Distances to crossover points are approximate values.
z
Observation point
r
ro = r – rp
θ
a y
rp
FIGURE 12.5 The vibrating circular piston of radius a set in a plane rigid baffle at z = 0.
70
60
10
80 50
10 40
70
30
60
–5 5
–5 5
0
0
(a) 5 0 (b) 5
0
60
10 40
40 10
20
20
0
–5 5 –5 5
0 0
(c) 5 0 (d) 5
0
FIGURE 12.6 Sound pressure level in front of a vibrating piston. (a): ka = 1, (b): ka = 10, (c):
ka = 32, and (d): ka = 100. We note the development of the directivity and of the Fraunhofer
region.
1.0
0.8
0.6
0.4
0.2
ka = 10
0.0
–1.0 –0.5 0.0 0.5 1.0
FIGURE 12.7 Isobars close to an oscillating piston for ka = 10. The white patch was outside
the calculation range.
Directivity 223
ka = 8π
RMS sound pressure (normalized to ρcu)
1.5
0.5
0
0 2 4 6 8 10
z/a
FIGURE 12.8 The solid line shows the rms sound pressure as a function of z on the center
normal of a circular disk at ka = 8π. The dashed line shows the geometrical attenuation sound
pressure curve that gives the same far-field pressure.
depending on the ratio between wavelength and piston radius. The sound pressure on
the normal of the circular piston can be shown to be [1]

p( z, ω) =
πa 2 (
ρcU − jkz
e − e − jk z2 + a2
) (12.8)
The RMS sound pressure as a function of z on the normal of the circular disk
is shown in Figure 12.8. The dashed line shows the geometrical attenuation sound
pressure curve that gives the same far-field pressure.
We note that the difference between the sound pressure curves increases as we
move to lower values of z/a. Figure 12.9 shows the difference in percentage.
It is important to note that when we measure the sound pressure from a transducer
(or use a directional microphone), there is a minimum distance beyond which the
sound pressure drops by distance according to the geometrical distance attenuation
(that is 6 dB per distance doubling). This shortest far-field distance r(λ) for ka > 1 is
often given as
2d 2
rmin (λ ) ≈ (12.9)
λ
Here, d is the piston diameter and λ is the wavelength of sound. The error between
the exact solution for the on-axis RMS pressure and the approximation is then less
than about 3% as shown by Figure 12.9.
10
ka = 8π
Difference between sound pressures in percentage
0
4 8 12 16 20
z/a
FIGURE 12.9 The difference in percentage between the curves shown in Figure 12.8.
12.8 FAR-FIELD OF A PISTON IN A BAFFLE

We will now study the far-field radiation characteristics of the circular piston shown
in Figure 12.5. We will call the radius to the surface point under study, rP. The
distance from the observation point r to a point rP on the piston is then |r − rP|. If we
use the angles defined in Figure 12.5, we find that
r − rP = r 2 + rP2 − 2rrP sin(θ) cos(ϕ) (12.10)

We now integrate the sound pressure generated by the vibrating piston to obtain
the graphs in Figures 12.6 and 12.7.
Numerical integration is slow and does not give immediate insight into the
importance of various variables. Simple analytical expressions for the sound pressure
on the axis of symmetry are of interest for the two common cases of circular and
rectangular pistons. For points far away from the piston, the magnitude of the
denominator in the integral in Equation 12.6 varies negligibly with the location of
the point on the piston surface. The phase however may vary widely. We find that we
can rewrite the distance as
r − rP ≈ r − rP sin(θ) cos(ϕ) + ... (12.11)

Directivity 225
We can find an analytical expression, using a series expansion, for points at a

far distance
a 2π
e − jkr
∫
0
∫
p(ω, r, θ) ≈ jωρ rP drP e − jkrP sin(θ )cos( ϕ ) dϕ
0
2πr
uP (12.12)

After some mathematics one finds that the integration over the surface of the
piston results in
 2 J1 ( ka sin(θ))  e − jkr
p(ω, r, θ) = jωρ 
ka sin(θ )

2 π r
(
πa 2 uP (12.13) )
 
Here, J1(x) is a Bessel function. Because of the cylindrical symmetry there is no

variation due to the angle ∏.
Replacing πa2uP in this expression by UP we can see that the expression for the
total sound pressure at the observation point r is similar to that of a monopole on a
hard surface in Equation 12.7 that has the same volume velocity. The difference is a
directivity term that gives us the directivity function F(θ) as
2J1 ( ka sin(θ))
F (θ, k ) = (12.14)
ka sin(θ)
We can find the directivity function in the same way for the rectangular piston
β α
L
y
W
rp ro = r – rp
FIGURE 12.10 The vibrating rectangular piston with sides W and L with its center at
x = y = 0, symmetrically set in a plane rigid baffle at z = 0.
The directivity function can be shown to be
 kL   kW 
sin  sin(α ) sin  sin(β)
 2   2 
F (α, β, L, W , k ) = (12.15)
kL kW
sin(α ) sin(β)
2 2
We have already noted that the smaller the piston the more omnidirectional
its sound radiation is, and that for small sound sources the geometrical distance
attenuation is −6 dB per distance doubling. While a point source will radiate
spherically symmetrical waves, a long but narrow slit-like rectangular source will
radiate similar to a line source, that is, radiate cylindrical waves. In contrast to
spherical waves, the geometrical distance attenuation is −3 dB per distance doubling
for continuous waves. Notice however that a cylindrical sound source cannot have a
delta-function impulse response, that is only possible with an ideal spherical source.
An interesting psychoacoustic effect when listening to cylindrical waves is
that the sound appears to always come from the point of the line source at a close
distance. Another effect, due to room acoustics, is that for a vertical line source the
ceiling and floor reflections contribute less to the coloration of the sound that we hear
since a line source loudspeaker excites the reverberant field less than does a spherical
sound source for the same on-axis direct sound. In domestic environments, the line
source is usually vertical to the floor and ceiling surfaces, and the images of the line
source will extend beyond these. This results in a quasi-infinite line source that is
then mirrored in the side walls of the room.
12.9 DIRECTIVITY AND DIRECTIVITY INDEX

The directivity concept as commonly used only has relevance for distances that are
in the far-field, that is, in the Fraunhofer region. It is clear from Figure 12.6 that the
idea of directivity is only applicable as a descriptor for “far” distances. A loudspeaker
that has a narrow main lobe and is intended primarily for listeners at far distances is
sometimes referred to as a “long throw” loudspeaker.
The gain factor G describes the intensity of the sound in the direction of maximum
radiation relative to the intensity at the same distance from a monopole having the
same radiated sound power. The gain factor is given by the ratio
4π
G= 2π π
(12.16)
∫ ∫ F(θ, ϕ)
2
sin(θ)dθdϕ
0 0
It is however more common and practical to use the directivity index, DI,
expressed in dB units as a metric for the directionality of the piston’s radiation rather
than the gain factor. The directivity index is defined as
DI = 10 log(G ) (12.17)

Directivity 227
0
ka = 6
–3
–5 –3 dB width
20 Log[|F(θ)|] [dB]
–10
Level of 1st side lobe
–15
–20
–25
–30
–90 –45 0 45 90
Angle θ[°] off-axis of circular piston
FIGURE 12.11 Some terms used to characterize directivity characteristics.
Two other useful metrics are the −3 dB point for the main lobe and the relative
level of the first side lobe. These are indicated in Figure 12.11.
A circular piston has a −3 dB main lobe width of about 30 λ/a degrees and has
the level of the first side lobe at about −17.6 dB relative to the level of the main lobe.
A rectangular piston has a −3 dB main lobe width of about 51 λ/W (and 51 λ/L)
degrees and has the first side lobe level that is about −13.2 dB relative to the level of
the main lobe.
12.10 DIRECTIVITY AND FREQUENCY RESPONSE

Directivity means that the power radiated is concentrated toward (typically) one
direction. Many electrodynamic transducers radiate about the same power for
constant input voltage, irrespective of frequency; so, changes in the gain factor
will lead to deviations from a linear frequency response also for this reason. Some
examples of the directivity characteristics of a circular piston in a rigid baffle are
Figure 12.13, on the other hand, shows the directivity index as a function of ka
for the same piston.
0°
Ka = 2.0 Ka = 4.0 Ka = 8.0 Ka = 16.0
–10 dB –10 dB
–20 dB –20 dB
–30 dB 0 dB –30 dB 0 dB
–40 dB –40 dB
–90° 90° –90° 90°
FIGURE 12.12 Examples of the directivity characteristics of a circular piston in a rigid

baffle for four values of ka.
20
Directivity index [dB]
10
0
0.1 0.2 0.5 1 2 5 10
ka
FIGURE 12.13 The directivity index as a function of ka for a circular piston in a rigid baffle.
0
relative to on axis response [dB]
–10
Frequency response to sides
–20
θ = 15°
–30
θ = 30°
θ = 45°
θ = 60°
–40
0.1 0.2 0.5 1 2 5 10
ka
FIGURE 12.14 Frequency response at four off-axis angles for a circular rigid piston that has
a radius of 0.1 m and flat frequency response on axis.
It is also instructive to study the frequency response of the side radiation by the
piston if the on-axis response has been normalized to be frequency independent. The
frequency response for four angles away from the normal is shown in Figure 12.14.
We note that the fall-off rate is very high for angles larger than 30°. For a rigid
circular piston driver having an effective diameter of 0.2 m, the response to the
sides is already down by 3 dB at about 1.6 kHz. Clearly compensating for the
frequency response nonlinearity due to the directivity will be difficult in practice,
particularly when dealing with multiway loudspeaker systems. Side-wall reflections
Directivity 229
in the listening room are likely have very uneven frequency response, even if the
on-axis frequency response is equalized to be “flat”.
12.11 FAR-FIELD OF A PISTON AT THE END OF A LONG TUBE

The directivity of a piston at the end of a semi-infinite cylindrical tube is of interest
for at least two reasons. First, most laboratory microphones for measurement of
sound pressure are tubular with the diaphragm at one end of the cylinder. As long as
the microphone has a length that is several times the wavelength, the directivity will
be unaffected from an engineering viewpoint. Because of reciprocity, the directivity
will be essentially the same as that of a piston at the end of the tube. Second, many
loudspeakers come in small boxes that have little baffle effect. Their directivity can
also be roughly approximated by that of a piston at the end of a tube.
The curves in Figure 12.15 show results that were calculated from a theoretical
model of the radiation from a plane piston at the end of a semi-infinite tube of
negligible thickness [2–5]. Because of the limitations of the theory used to calculate
these directivity functions, the curves are only shown for frequencies up to ka = 3.83.
We note that for small values of ka the radiation is almost omnidirectional but that
directivity rapidly increases as ka increases. Except for the radiation to the back, the
directivity for large values of ka is similar to those of the piston in a baffle.
For ka values larger than 1, the best method to find the theoretical directivity
is by numerical modeling of the entire transducer and housing (that is, driver and
box) using boundary or finite element modeling. This approach will also yield the
influence of the diffraction of the box and the nonplanar loudspeaker diaphragm.
12.12 NEAR-FIELD AND FAR-FIELD FREQUENCY

RESPONSE OF A CIRCULAR PISTON
We are sometimes faced with the dilemma of having to measure the anechoic
frequency response of a loudspeaker without access to a satisfactory nonreflecting
environment. It is however possible to measure the frequency response, close to a
0° 0°
ka = 1 ka = 2 ka = 3.0 ka = 3.83
–10 dB –10 dB
–20 dB –20 dB
–30 dB 0 dB –30 dB 0 dB
–40 dB –40 dB
–90° 90° –90° 90°
180° 180°
FIGURE 12.15 Directivity patterns for a plane piston at the end of a semi-infinite cylindri-
cal tube for four values of ka.
piston and then compensate by a known quantity. For ka ≤ 1, the ratio between the
sound pressure immediately in front of the center of the piston having radius a and
the sound pressure in the far-field at a distance r can be shown to be [6,7].
p′ff ( ω, r ) a − jωr / c (12.18)

≈ e
pnf′ ( ω, a )center 2r

By using this correction term, we can conveniently measure in the near-field (for
ka ≤ 1) and then just multiply the complex sound pressure with the ratio given by
Equation 12.18.
12.13 ACOUSTIC CENTER

The idea of the acoustic center finds its use primarily when several transducers are to
be coupled together such as in multiway or array loudspeaker systems.
Correct knowledge of the location of the acoustic center of transducers is also
necessary when transducers are to be used at close range to one another, and their
distance to one another is critical as, for example, in some types of microphone
calibration. The location of the acoustic center of a transducer typically varies
with frequency as in horns and loudspeakers. The location may even vary in an
important way for microphones. Finally, it should be emphasized that the location
of the acoustic center is independent of whether the transducer is used for sending
or receiving.
There are several ways in which the acoustic center may be defined [7–10]. Three
ways are commonly referred to:
1. Extrapolation of the location of the acoustic center from measurement of

the geometrical attenuation (distance law) for the RMS sound pressure in
the far-field.
2. Extrapolation of the location of the acoustic center from measurement of the
curvature of the wave front in the far-field using polar response amplitude
or RMS sound pressure measurement; this is sometimes called the “wave
front” method.
3. Extrapolation of the location of the acoustic center from measurement of
(spherical) wave front phase or time of arrival data in the far-field; this is
sometimes called the “time alignment” method.
Even for the simple case of a pulsating sphere, the definitions give different results.
Definitions 1 and 2 position the acoustic center at the center of the sphere, whereas
definition 3 puts it at the surface of the sphere for high values of ka. For a circular
piston in a baffle, all three methods give different results [7].
Usually, the definition according to (2) is used since it gives data that can be used
for arrays. Figure 12.16 shows the difference between the methods. It is seen that
delay and center of curvature are two effects that affect any measurement of acoustic
center using the time-alignment method. First, the center should be determined
Directivity 231
(a) (b)
FIGURE 12.16 Definitions of the acoustic center by (a) time delay (b) wave front.
using the wave front method; then the time delay inherent in the radiator (usually a
transmission line or a horn) should be compensated for. One can in this way separate
acoustic center from acoustic origin, as shown in Figure 12.16.
Following the data presented in Ref. [9], the mouth is the physical location from
where the sound radiates, the effective aperture being the size of the horn mouth.
12.14 ARRAYS
For multiway loudspeakers, the area of listening is usually reasonably well-defined
and the interest in the acoustic center is mostly to time-align the sound contributions
from the low-, mid-, and high-frequency loudspeakers.
The need for a clearly defined acoustic center is most obvious when working
with arrays. Arrays are sets of spatially separated transducers used for directional
transmission and reception of signals. Using arrays, we can enhance the signal
quality, for example, by making the signal stronger, that is, increase transmission
or reception gain to overcome electronic system noise, or by increasing the signal-
to-noise ratio by rejecting noise coming from directions other than the desired
signal’s direction.
By array pattern synthesis we mean the synthesis of directivity patterns using
controlled interference between received or transmitted signals. Since array
applications may involve both transmission and reception of signals, it is important
to note that the array directivity does not depend on the use, but only on the spatial
distribution of the transducers (i.e., geometry of the array) and on transducer
directivity. The directivity is achieved by constructive and destructive interference
between the signals transmitted or received. Since the interference depends on
phase, which in turn depends on distance and wavelength, the array directivity will
be frequency dependent.
We will look at arrays in a way similar to that which we have applied to pistons,
considering them composed of small sound sources. The main difference will be that
we will look at systems that are characterized by large array length to wavelength
ratios, and that have shading, that is, the phase and amplitude varying over the length
of the array. Our analysis of bending wave radiation from large sheets was done
using this method as well. The simple one-dimensional array theory presented here
can be applied to 2D and 3D arrays as well, as long as the sources/receivers are very
small compared to wavelength.
12.14.1 Classifying Array Systems

In sonar engineering, one differs between monostatic and bistatic applications. In
a monostatic application sending and receiving will be done with the same array,
while in a bistatic application separate arrays will be used for the transmitting and
receiving functions. These possibilities are usually not available in audio engineering
where we are usually interested in arrays of microphones to follow moving sound
sources and loudspeaker arrays to direct the sound to the audience.
Receiving arrays can be subdivided into active and passive arrays, depending on
whether signal processing techniques are used or not. Another fundamental subdivision
is into discrete and continuous arrays, depending on whether the transducers are
picking up signals at discrete points or over some area. This can also be the case for
the transmitters. A third way of subdividing arrays is into local and distributed arrays,
depending on how the transducers are grouped in single or multiple arrays.
Both transmitter and receiver arrays function because of the coherent summation
of sound pressure while at the same time (hopefully) summing noise components
noncoherently. The coherent summation will give stronger signal than noncoherent
summation, which results in an improved signal-to-noise ratio. The use of coherent
summation is sometimes called delay and sum beamforming. Such beamforming can
be accomplished by both passive and active signal processing techniques. Passive
techniques can use both acoustic and/or electric delay and sum beamforming.
Depending on the design of the summation, one can have both local and focused
arrays. Local arrays are arrays in which the individual array elements are much
closer to one another than to the sound source of interest. Focused arrays are usually
used at close range to the pickup point.
Local arrays can be further subdivided into broadside and end-fire arrays.
Broadside arrays are characterized by nominally having their main lobe away from
the direction of the array while end-fire arrays are characterized by their main lobe
being along the direction of the array.
When the various transducers have different weights in the summation, we speak
about shaded (sometimes also called windowed) arrays. In transmitting arrays, such as
loudspeaker columns, the end transducers will see different radiation impedance than
the center ones; so, there will be shading because of this effect. There is also the matter
of the size of the individual transducers. Small receiving transducers such as micro-
phones and hydrophones, at reasonable distance, will not shadow one another, but in
large arrays, influence of scattering by the many transducers may become problematic.
12.14.2 Directional Properties of Array Transducers

It is important to note that transmitters and receivers used in arrays can have any
directional characteristics, although simple theory often assumes that they are all
Directivity 233
omnidirectional. Of course, the total array directivity function will depend on the
transducer and array directivity functions. Far away from the array, the combined
directivity will be that of the product of the array and transducer directivities.
Cardioid microphones are very useful in audio arrays since cardioid microphones
can be made very small and still contribute to considerable signal-to-noise ratio
improvements over omnidirectional microphones.
For arrays to be effective they must use many transducer elements and be large
compared to the wavelength. However, simple array theory is based on the assumption
that the sound source is positioned far away from the array so that incoming waves
only differ by a frequency-dependent phase factor.
12.15 ARRAY TRANSFER FUNCTIONS

We will only discuss receiving arrays using transducers that are small compared to
wavelength here, because these are simpler to design than sending arrays which often
need transducers that are fairly large compared to wavelength. An added problem in
transmitter arrays is that the radiation impedance for an element is affected by other
elements.
For the receiving array, we are interested in the electric output signal of the
array system consisting of microphones and amplifiers. The electric signal from the
transducers will depend on their sensitivities and directional characteristics.
The sound pressure at each microphone will depend on the source volume velocity,
directional characteristics, and its distance to the microphones. The transducer
can be inherently directional (or made to be directional by a combination of small
“elementary” transducers at each array point) or omnidirectional, but we will only
consider omnidirectional transducers.
The electric output signal of each transducer channel can be written as
eout = H SP HT H geoU (12.19)

where
U is the volume velocity of the source
HSP is the transfer function of the signal processing system
HT is the transfer function of the microphone (which will include microphone
sensitivity due to the angle of arrival of sound, i.e., directivity)
Hgeo is the geometrical transfer function of the omnidirectional source
e − jkr
H geo (ω, r ) = jωρ (12.20)
4πr
12.15.1 Array Factor, Wavelength, and Inter-Element Distance

Let us first study the directivity of arrays to be used over a narrow frequency band. By
a narrow band is typically meant of bandwidth of about 10% of the center frequency.
The derivation here follows that in Ref. [11].
Point source in the far-field
z
r
rn
θ
Microphone n
Ω
y
d
x
FIGURE 12.17 A linear, local array of small microphones along the y-axis.
Assume that we have an array assembled of N small microphones at equal inter-

element distance d along the y-axis in Figure 12.17. The sound pressure at nth
microphone is pn. Using the symbols in the figure, the angle Ω to the line array is
cos ( Ω ) = sin (θ ) sin ( ϕ ) (12.21)

The array’s spatial distribution can then be written as
N −1
Γ( x, y, z) = δ( x ) δ( z ) ∑ A δ ( y − n d) (12.22)
n=0
n

The sound pressure at the point of observation will be dependent on both the
mentioned transfer functions of the individual microphones and the associated
signal processing, and the properties of the array. We call the geometrical reception
properties of the array the complex array factor (AF). The array factor serves the
same purpose for the array as the directivity function F that we use to describe the
directional properties of microphones. For arrays that use microphones that have
identical directivity, the array factor and the microphone directivity are multiplied to
obtain the effective array factor.
In the most common case, the linear array illustrated in Figure 12.17, the array
factor will depend on the wave number k, the number of microphones N, the distance
between the transducers d (here assumed to be constant but could vary), a phase
factor describing the phase differences between each microphone, the relative ampli-
tude of each microphone An, and the angle Ω against the array axle, here the y-axis.
The sound pressure at the microphones will depend on the sound source. Assume
that the sound source is omnidirectional, has a volume velocity U, and is at such far
distance r that the sound pressure at each one of the microphones can be written as
e − jkrn
pn (ω, rn ) = jωρ U (12.23)
4πrn
Directivity 235
We now look at the sound pressure at each one of the microphones in the array.
Far from the sound source, the incoming wave field can be considered plane over
the extension of the array. The amplitude will be approximately the same at all
microphones but there will be a phase difference in the signal received by each
microphone. We are interested in the effects of this phase-shift but we consider the
amplitudes about equal:
e − jkrn
pn (ω, rn ) = jωρ U (12.24)
4πr
Let us assume that the signals from the microphones are all summed up
electrically. The summed output signal e will then be
e (ω, rn ) = H eaU
∫∫∫ M ( x, y, z) e
V
− jkrn
dxdydz (12.25)

where M(x,y,z) is the microphone distribution.
For convenience, we have written Hea instead of HSPH T. The integral is the array’s
directivity function AF usually called the array factor. It is used similar to the con-
ventional directivity function discussed previously. We note that the array factor is
the Fourier transform of the spatial distribution of the microphones. We note that
the distance contribution for each microphone on the array relative to the origin is
rn ≈ r − nd cos ( Ω ) (12.26)

The linear array in Figure 12.15 will have

N −1
AF =
∫[A en
j k y cos Ω
∑ δ ( y − nd )] dy (12.27)
n= 0

that we can be written as
N −1
AF = e − j k r ∑Ae
n=0
n
j k n d cos (Ω )
(12.28)

Usually we do not include the geometrical phase component e−jkr in the array factor
since the absolute phase is usually not important. We then obtain the expression for
the sum of the electric output of the microphones as
N −1
e − jkr
e (r, Ω, ω ) = jωρ
4πr
U ∑H
n= 0
ea ,n Ane j k n d (Ω) (12.29)
cos

Here, the part before the summation sign can be thought of as the pressure
magnitude at each microphone, the geometrical transfer function multiplied by U,
Im
A3 3 kd cos(Ω)
Ω
ngle
o ra A2
o rf
h as 2 kd cos(Ω)
ngp
lti
su
Re
A1
A0 kd cos(Ω)
Re
FIGURE 12.18 Graphic determination of the array factor, AF, for a four-element array.
or simply a strength factor. Figure 12.18 shows how we can calculate AF by a phasor
diagram if we assume Hea,1, Hea,2, Hea,3, and Hea,4 all unity. We notice that the array
factor is symmetrical around the array axis, in this case the y-axis.
The transfer functions of the microphone channels are typically adjusted for the
desired weights |An| and to differ by the same “progressive” time or phase difference
α, so that
H ea,n = An e jnα
(12.30)

We include these weights and the phase difference, so that we can write
N −1
AF = ∑Ae
n=0
n
j n ( k d cos (Ω ) + α )
(12.31)

It is now practical to introduce a new phase variable β(d,Ω, ω), such that
β(d, Ω, ω) = kd cos(Ω) + α (12.32)

With this variable, we can write AF as a sum that is periodic in 2π
N −1
AF (d, Ω, ω) = ∑Ae
n=0
n
j nβ
(12.33)

Only a part of the period is “visible,” that is, can be seen in the directivity plot for
a certain frequency. How much of that is visible depends on the variable β, that is, on
k, d, Ω, and α. Figure 12.19 shows an example of how the visible range is determined
Directivity 237
|AF (β)|
β
–2 π 2π
kd
Ω
kd cos(Ω)
kd cos(Ω) + α
FIGURE 12.19 An example of the dependency of the array factor on k, d, Ω, and α. Note
the periodicity of AF as a function of β.
from these variables. The choice of visibility range will depend on signal-to-noise
considerations and the need for gain in the main lobe.
For a linear array of microphones having equal gain settings A0 we find
N −1
AF (d, Ω, ω) = A0 ∑e
n=0
j nβ
(12.34)

The sum of such a geometric progression is given by
N −1
e N j β −1
∑n=0
ej nβ =
e j β −1
(12.35)

This results in an array factor
Nβ
 j Nβ
j −j
Nβ
  Nβ 
e  e 2 −e
2 2
 ( N −1) β sin 
  j  2 
AF (d, Ω, ω) = A0 e = A0 e 2 (12.36)
j
β
 jβ −j 
β
 β
e e − e 
2 2 2 sin  2 
 

The exponential
( N −1) β
j
e 2 (12.37)
is a phase factor that is due to our choosing one end of the array to be at the origin.
If we would have had an odd number N of microphones with the center of the array
at the origin, we would have found this expression for AF:
1  Nβ 
( N −1)
sin 
 2 
2
AF (d, Ω, ω) = A0
1
∑ e j nβ
= A0
 β
sin  
(12.38)
− ( N −1)
2  2

We recognize this form of AF as similar to that we found when we studied the
directivity of an assembly of monopoles. Figure 12.20 shows the |AF| polar diagram
for a uniform three-element array that has kd = π and α = −π/2.
By choosing d and α, we can direct and shape the main lobe for a frequency ω
where the wave number is k. We note that for α = 0, we obtain maximum sensitivity
at right angle to the array; this is called a “true” broadside array.
|AF (β)|
0.8
0.6
0.4
0.2
β
0 π π 3π
–2π – 3π –π –π 2π
2 2 2 2
)|
(Ω
F
|A
Ω
kd cos(Ω)
kd cos(Ω) + α
α = −π/2
FIGURE 12.20 The modulus of the array factor |AF(β)| for a uniform three-source
array having a negative inter-element progressive phase-shift of −π/2 and an inter-element
distance d of one half wavelength. The method of constructing the corresponding polar plot
of |AF(Ω)| is shown. Note that the polar plot curve has its own scale, the main lobe maximum
is at |AF(0)| = 1 and that the arrow points at |AF(45°)| ≈ 0.85.
Directivity 239
Directivity
Grating lobe 0 dB
–10
–20
–30
–40
–50
–60 y
FIGURE 12.21 Two directivity plots for a seven-element end-fire array. The upper, solid
curve shows the array’s directivity with excessive inter-element distance causing a grating
lobe at Ω ≈ 120° while the lower, dashed curve shows the array’s directivity after adjustment
(reduction) of the inter-element distance to eliminate the grating lobe.
If the microphones are too close, or the frequency too low, so that kd is small,
there is only one main lobe. If kd on the other hand is large there will be one or
more side lobes. If there are additional “main” lobes such as the main lobe, these are
called grating lobes. Figure 12.21 shows how an inappropriate choice of excessive
intermicrophone distance has resulted in one of the maxima of the |AF| function
being turned into a grating lobe.
The side lobes can be reduced or removed by suitable windowing of the AF in the
form of adjustment of the magnitude or phase of the individual array elements. This
can be achieved by acoustical or electronic means, and is applicable to both sending
and receiving arrays.
The presence of grating lobes will often degrade the signal-to-noise ratio if there
is much noise in their directions. In a diffuse noise sound field the presence of grating
lobes will limit the increase in signal-to-noise ratio that can be obtained by the array.
By using the proper transfer functions Hea,n it is possible to progressively reduce the
gain for array microphones at high frequencies, thus making the array size the same
relative to wavelength always. This technique can be used to prevent the formation of
grating lobes. We can think of this as a frequency-dependent windowing of the array
sensitivity. It is sometimes called harmonic nesting.
We have seen that if there is a progressive phase-shift α between the transducer
channels, the main lobe will be turned to a new angle Ω m. This means that one can
move the main lobe (and the side lobes) to the desired direction in space by using
a phase-shift or time delay unit that shifts the phase of each source in the array by
a multiple of the desired phase-shift. This is only possible however in narrowband
systems where the bandwidth is small compared to the carrier frequency. This is
usually the case in many ultrasonic sonar systems.
In audio systems that need to function over a wide frequency range it is instead the
progressive time delay that needs to be increased in multiples of some desired value.
Note however that the array needs to have the same effective length relative to the
wavelength of sound at all frequencies to obtain the same array factor at all frequencies.
12.16 CONTINUOUS LINEAR ARRAYS

Assume that the inter-microphone distance is reduced toward 0 while at the same
time the number of sources is increased so that the array length L is the same. We
call the coordinate along the array axis y, which corresponds to nd. We also write An
as A(y). Let Ω be the angle between the array-axis and the line from the observation
point to the origin as mentioned previously. We then obtain the array factor as

AF =
∫ A( y) e j ( k cos( Ω ) + α ) y
dy (12.39)
where α is the phase-shift per unit length along the x-axis.

Now assume that we have a linear array characterized by the array sensitivity
A −L L
<y<
 0 2 2
A( y) =  (12.40)

0 at all other points
This AF is similar to the directivity function that we applied when studying the
rectangular piston. We find that
sin (β1 )
AF ( L, Ω, α, ω ) = A0 L (12.41)
β1
where
1
β1 ( L, Ω, α, ω ) = (k cos (Ω) + α ) L (12.42)
2
An example of the array function for a uniform continuous line array is shown in
Figure 12.22. We note that if there is a gradual progressive phase-shift, the main lobe
will appear at an angle Ωmax
α
cos ( Ωmax ) = − (12.43)
k
Directivity 241
AF (β)
1
0.8
0.6
0.4
0.2 β
– 3π –π π π π 3π 2π
–2π –
2 2 2 2
FIGURE 12.22 Examples of the array factor for a uniform continuous line array (solid line)
and for an array having cosine squared tapered amplitude (dashed line).
Using a window function we can reduce the level of the side lobes. We already
noted that the array function is the Fourier transform of the spatial distribution of
transducers; so the same techniques of windowing that are used in signal analysis
can be used to reduce side lobes in the array function. If, for example, the continuous
array previously discussed is windowed to have a tapered sensitivity
2 A0 cos2  πy  − L < y < L

  L 2 2
A( y) =  (12.44)

 0 at all other points

the side lobe levels will be reduced as shown in Figure 12.22 since the AF will
then be [12]
π 2 sin (β1 )
AF (ω, L, Ω, α ) = A0 L (12.45)
2π 2β1 − 2β13
The two corresponding polar diagrams for these cases are shown in Figure 12.23.
90°
kL = 5
–10 dB
–20 dB
–30 dB 0 dB
– 40 dB
180° 0°
FIGURE 12.23 Examples of (half) polar diagrams for a two broadside arrays. Angle is Ω.
The solid line on the right shows one half of the polar diagram for a uniform continuous line
array while the dashed line on the left shows one half of the polar diagram for an array having
a cosine squared tapered amplitude.
12.17 POLYNOMIAL EXPANSION OF THE ARRAY FACTOR

The array amplitude An can also be used in a different way to adjust the directivity.
Using the shorthand
γ = e jβ = e jkd cos(θ) (12.46)

the array factor can sometimes be written in polynomial form as
N −1
AF ( γ ) = A0 + A1 γ + A2 γ + ... AN −1 γ N −1 =
2
∑A γ
n=0
n
n
(12.47)

This polynomial can now be factored, and the roots of the polynomial, γn, will
then correspond to those angles at which the array factor is zero, provided of course
that these parts of the array factor are in the visible range. This factorized polynomial
can now be used to synthesize an array having the desired directivity. By multiplying
the factors and returning to the polynomial form we will obtain the weighting values
for the respective sources.
The process is best illustrated by an example. The binomial array is attractive
since it does not have any side lobes. Assume initially that we have only two sources
(N = 2).
AF ( γ ) = ∑A γ
n=0
n
n
= A0 + A1 γ = A0 (1 + γ ) (12.48)

A simple rewrite gives us
j  j −j 
β β β β
j  β
A0 (1+ γ ) = A0e 2  e 2 + e 2  = 2 A0e 2 cos   (12.49)
   2

If we now want the main lobe to retain its main shape (in principle) but want it
sharper (higher directivity), we can introduce another factor in the following way by
extending the array by another element (N = 3):
AF( γ ) = (1 + γ ) (1 + γ ) = 1 + 2 γ + γ 2 . (12.50)

We see that the weightings will be A0 = 1, A1 = 2, and A2 = 1. Figure 12.24 shows
some cases for further higher polynomials having d = λ/2.
Figure 12.25 shows the corresponding directivity plots. Comparing these to earlier
plots of uniform arrays we note that the absence of side lobes is paid for by a larger
main lobe. The choice between an array having such larger main lobes and an array
having a narrower main lobe, but also side lobes, must be decided by taking into
account the assumed noise characteristics (including reverberation) of the background.
Directivity 243
Binomial arrays
AF (β) N =2 N =4
1 N =3 N =5
0.8
0.6
0.4
0.2 β
– 3π π 3π
–2π –π –π π 2π
2 2 2 2
FIGURE 12.24 Examples of the array factor for some binomial arrays.
90°
Binomial arrays
N=2
N=3 –10 dB
N=4
N=5 –20 dB
–30 dB 0 dB
– 40 dB
180° 0°
FIGURE 12.25 Examples of the directivity for the broadside binomial arrays shown in
Figure 12.24.
12.18 WIDE FREQUENCY RANGE ARRAYS

We have noted that the lobe pattern is critically dependent on the length, inter-element
distance, and operating frequency of the array. To maintain the same lobe pattern the
array must be made progressively smaller as operating frequency increases.
If the lobe pattern is not an issue, time-delayed summation is sufficient in order
for the main lobe to remain at constant angle relative to the array. In acoustics, the
time delay can be achieved either passively, by using for example plastic tubes or
similar devices to carry a portion of the wave to a summation point, or electronically
for example by using digital signal processing in some form. The use of time-
delay techniques however does not prevent the formation of grating lobes at high
frequencies unless combined with suitable windowing. When grating lobes occur
they will be accompanied by a reduced directivity index for the array.
The successful design of active arrays depends to a large extent on how transducers
can be positioned. The greatest benefit will be had when working with signals having
high inherent signal-to-noise ratios. The use of delays is a simple form of active
summation as shown in Figure 12.26.
In active beamforming arrays, using digital signal processing, it is not adequate
to talk about end-fire or broadside arrays. The beamforming may also be adaptive.
M0 M1 Mi Mn Microphones
G0 G1 Gi Gn Amplifiers
τ0 τ1 τi τn Delay units
+ + + Signal out
FIGURE 12.26 The principle for simple delay and sum receiving arrays.
The resulting directivity may be similar to that obtained by traditional techniques but
not necessarily so. The receiver beamformer can be made to optimize the reduction
of noise, interference, or reverberation separately.
12.18.1 Harmonically Nested Arrays

If the effective width of the array is made successively smaller by suitable window-
ing, a technique known as harmonically nested arrays, it is possible to achieve more
uniform characteristics. The improvement at low frequencies is at the expense of the
directivity index at high frequencies. Figure 12.27 shows an example of the layout of
a harmonically nested array.
A major disadvantage of the harmonically nested array is that the array requires
many transducers. Demonstrations of large distributed arrays containing thousands
of transducers have been made. By using digital signal processing and adaptive
techniques, it is possible to simultaneously track many sound sources. These
techniques are well-known in the fields of radar, sonar, and geophysics. Their
application in audio engineering is made difficult because of the wide bandwidth
that is of interest. For speech, a bandwidth of about a decade (300 Hz –3 kHz) is
necessary for sound quality and intelligibility. Identical AF properties over the full
audio bandwidth will be difficult to cover by any practical array.
HF behavior
MF behavior
LF behavior
Physical 0
–10
B]
microphone
–20
Level [d
array
to a from
50
[°]
Microphone signal D/A-conversion –30

rray
–40
0
y
nor le awa
Computer software implementation 500

of bandpass filters and crossover Fre 1000 –50
mal
que
Ang
ncy 1500
(a) Output [Hz 2000
(b) ]
FIGURE 12.27 Basic principle (a) and simulation results (b) for a harmonically nested
broadside array.
Directivity 245
For more homogeneous wide-band array properties, one can use other types of
array designs such as Bessel and semi log-periodic arrays [6,13]. The latter design
uses a hybrid between linear and logarithmic microphone spacing and includes the
filtering of all channels by finite impulse response linear phase filters.
Many sound reinforcement arrays are used out-of-doors, and the low-frequency
sound may be particularly annoying to the neighboring environment. Using gradient
loudspeakers that have bidirectional or cardioid polar patterns in the array may
reduce such problems.
12.19 SIGNAL-TO-NOISE RATIO IN RECEIVER ARRAYS

Arrays are also advantageous from the viewpoint of signal-to-noise-ratio as receiver
systems. If the noise added by each transducer of the array (for example, micro-
phones) is the same, and can be considered uncorrelated, the total signal-to-noise-
ratio of the array will be improved significantly. This is of course in addition to the
signal-to-noise improvement provided by the directivity of the array. For a point
in the main lobe direction of a uniform in-phase array, the signal-to-noise level is
improved by ΔL.
∆L = 10 log ( N receivers ) [ dB] (12.51)

12.20 AUDIBLE ARTIFACTS OF LARGE ARRAYS

The analysis of arrays presented this far has assumed that the time history of the array
signal has little importance; our interest has been focused on the frequency response
behavior. Human hearing however is binaural and its analysis system has a fine time-
frequency resolution. Let us consider two extreme array positions to illustrate the issues
involved. First, consider a broadside array in the median plane; in this case the x = 0
plane, placed in front of the listener as shown in Figure 12.28. The array is assumed
to be a few meters away from the listener and all array elements are assumed to be
radiating in phase and with the same amplitude; there is no shading. The magnitude of
the angle δ is assumed to be about 20° or more for either end of the array.
In this case the listener is limited by coloration effects to determine the length
of the array. Listening to a signal such as wide band noise or music, the sound
subjectively seems to emanate from the part of the array that is in the horizontal plane
irrespective of the relative position of the array, that is, the position is independent of
the values of zupper and zlower.
For a similar situation, but now with the array in the z = 0 plane and parallel to the
x-axis in the figure, the situation is entirely different. Now the arrival times of the
sound contributions from the various transducers in the array play a role since our
binaural hearing can decipher the arrival times. As long as the listener is positioned
immediately in front of the array center, the effect heard will be that of coloration
due to the delay of the sound from the two sides of the array. When the listener is
listening off-center to the array, the arrivals of the last contribution from the end of
the array will be distinctly heard.
Frontal plane Median plane

z = zupper
r Array
parallel to
the z-axis
δ at y = l y
x y
Horizontal plane
z=–zlower
FIGURE 12.28 A wide frequency range large array in the median plane of the listener.
The simplest array of this type is of course the stereo loudspeaker pair when
driven by the same monophonic signal. When the listener is on the centerline of
the array, the listener will hear a “phantom” sound source located between the two
loudspeakers. The effect breaks down when the listener moves off-center and the
apparent sound source moves to the closest loudspeaker. This is due to the precedence
effect.
12.21 ACOUSTIC LENSES

By acoustic lenses we mean devices that by a change of the refraction index (in
other words propagation speed) we can focus or defocus sound waves in analogy
to the way ordinary lenses, for example, glasses, refract optical waves. Acoustic
prisms may be useful for directing sound away from a target area. In contrast to
conventional lenses for light, acoustic lenses are relatively small compared to the
wavelengths of sound, even where ultrasonic waves are concerned. Optic lenses only
need to work over a frequency range of an octave (visible light from 400 to 800 nm),
whereas an acoustic lens to be used for audio frequencies needs to work over several
octaves. In ultrasonics however the relative bandwidth is often even smaller than in
optics because of the resonant nature of many ultrasonic electroacoustic transducers
such as those used in medical sonar, etc. As in optics, the refractive index n is used
to characterize the lens “material.” The refractive index is the ratio between the wave
speed in the medium and that in the lens, n ≥ 1.
The idea of shaded arrays was discussed previously and implies the use of
different sensitivity along the array in a suitable fashion. In a loudspeaker column,
this can be achieved electrically or acoustically. To achieve the necessary amplitude
shading acoustically, a piece of porous material such as glass wool could be placed
in front of the individual loudspeaker. The glass wool however is characterized by
Directivity 247
a reduced speed of sound; so the construction will also feature “lens action.” The
change in propagation speed can be accomplished in various ways.
One obvious way is to make the lens of a material or gas that has a sound
propagation speed that differs from that of the surrounding medium. Lenses for
use in ultrasonic engineering may be manufactured by using materials that have
approximately the same impedance as the surrounding medium but different
longitudinal wave speed. The use of acrylic lenses is common to refract sound
waves in water and other fluids. For airborne sound, the situation is similar, but
homogeneous lenses would need to be made using lightweight, thin foil “balloons”
filled with a gas or gas mixture that has a specific impedance close to that of air
but with a different speed of sound.
A second method is to use a mechanical system such as an array of rigid elements
to achieve the desired delay, such as path length devices or obstacle arrays. Lenses
made by obstacle arrays are somewhat more complicated to manufacture than path
length devices.
Two types of path length devices and two types of obstacle arrays are shown
in Figure 12.29. The path length devices are assumed to have canal cross sections
that are much smaller than the wavelength so that the waves in the canals are plane
waves. The refractive indexes and delays achieved by these devices are shown in the
figure. The obstacle devices can be considered delay lines composed of tubes with
periodic constrictions [14]. The obstacle arrays are assumed to be so densely packed
that the distances between the obstacles are much smaller than a wavelength.
Path length devices Obstacle arrays
t0
n=
tpath Serpentine plates n = 1 + N 2 πa3 Spheres
3
(a) (b)
t0
n= n = 1 + N 8 a3 Circular disks
tpath Slant plates 3
(c) (d)
FIGURE 12.29 Four mechanical arrays of rigid elements used to achieve the desired
refraction index n. Figures (a) and (c) show two path length devices. The path time is t0 for
the wave directly and tpath is the time it takes for the wave to pass the delay device. Figures
(b) and (d) show two obstacle arrays. The number of spheres or (thin) disks per volume unit is
N, and the radius of the spheres or disks is a. (From Kock, W.E. and Harvey, F.K., J. Acoust.
Soc. Am., 21(5), 471, 1949.)
A serious problem in using acoustic lenses is of course the limited focal depth. The
focal depth can be calculated in analogy with that of optical lenses. The chromatic
aberrations will be large for a wide frequency range lens. Note that reciprocity also
applies to lenses.
REVIEW QUESTIONS
12.1 Why do we have different format plots to illustrate transducer directivity?
12.2 What is the difference between far-field and near-field?
12.3 What is meant by the reactive near-field, Fresnel, and Fraunhofer zones?
12.4 Why can the directivity only be determined from the far-field?
12.5 Define reciprocity.
12.6 What is the difference between gain factor and directivity index?
12.7 How does one calculate the directivity index from knowledge of the directivity
function?
12.8 What will be the frequency response of rigid circular pistons?
12.9 How is the array function defined?
12.10 What is the relationship of the array function to the Fourier transform?
12.11 How are broadside and end-fire arrays different?
12.12 What is meant by visible angle?
12.13 What is the difference between side lobes and grating lobes?
12.14 How can side lobes and grating lobes be minimized?
12.15 How can one achieve common array properties over a wide frequency range?
12.16 How is the signal-to-noise ratio influenced by arranging the transducers as an
array?
12.17 What are the design principles of acoustic lenses?
12.18 What is meant by the acoustic center of a loudspeaker?
12.19 Can a microphone have an acoustic center?
PROBLEMS
12.1 When measuring the directivity of a loudspeaker it is important to have the
microphone in the loudspeaker’s far-field where the sound pressure level of
the loudspeaker’s sound drops by 6 dB per distance doubling. In this case the
loudspeaker is to be considered a rigid circular piston of diameter 0.4 m set in a
rigid baffle. With the diaphragm volume velocity U the far-field approximation
for the sound pressure is
jρck
p approx = U (12.52a)
2πr
The exact expression for the sound pressure on the axis of symmetry of a rigid
piston of diameter a set in a baffle is
ρc  − jkz z2 + a2  U (12.52b)
p exact = e − e − jk
πa 2  
Directivity 249
Task:
Determine numerically how far away from the loudspeaker one should measure
for the error (i.e., the difference between the far-field and near-field expressions)
to be smaller than 1 dB at 5 kHz? (A numerical solution is sufficient)
12.2 A circular pistonic loudspeaker driver has concentric decoupling compliances
that subdivide it into rings as shown in the figure below.
Concentric compliant rings
Baffle Baffle
Voice coil force
The innermost piston ring is driven at its center; so all frequencies will be
reproduced by this part of the diaphragm. One can as a first approximation
assume that the compliances act as ideal mechanical filters so that outside
rings will be fully decoupled over the cutoff frequency. The outside rings can
then be considered parts of the infinite baffle in which the driver is assumed
to be mounted.
Task:
In a particular loudspeaker that had a piston diameter of 2·10 −1 m the decoupling
was chosen so that the rings had the diameters 5·10 −2, 8·10 −2, 13·10 −2 m. Calculate
the decoupling frequencies so that the directivity index does not exceed 2 dB
and plot the directivity index over the frequency range 20 Hz–20 kHz.
12.3 A microphone’s directivity can be calculated by summing the force of the
incoming sound wave over the diaphragm surface. The figure below shows a
microphone that has a quadratic diaphragm with sides a long. The sound wave
_p is incident at an angle θ to the diaphragm normal and parallel to one of the
membrane sides. Assume that the effects of diffraction can be neglected.
Normal to
diaphragm Incoming wave
ˆ −jkx
p = pe
θ
a
a
Microphone
diaphragm
Microphone
body
Tasks:
a. What will be the magnitude of the force acting on the diaphragm as a
function of frequency and angle of incidence?
b. For which value of θ will the effect of oblique incidence result in −1 dB
level difference compared to θ = 0?
12.4 Arrival time analysis is important in the use of wide frequency range arrays.
Three microphones are placed in the z = 0 plane at various (x,y) coordinates.
In an experiment, the arrival times τi of sound was as follows: Point 1 (0,0.5)
τ1 = 7.08 × 10 −3 s, point 2 (0,0) τ2 = 6.04 × 10 −3 s, and point 3 (0.5,0) τ3 = 5 × 10 −3 s.
Tasks:
a. Determine the direction in the xy plane from which the sound arrived.
b. Design time delays for the microphones so that the array is focused on a
point (4,5) in the xy plane.
12.5 A linear three-element summing ultrasonic microphone array is used to detect
arriving plane waves as shown in the figure below.
Arriving plane waves
Microphones
Ω
λ/2 λ/2
Tasks:
a. Determine the far-field directivity.
b. Determine the difference in sensitivity level between the Ω = 0° and Ω = 90°
directions.
12.6 The figure below shows the directivity of a summing linear microphone array
that has an inter-microphone distance of 8·10 −2 m. The angle Ω is the angle
away from the line interconnecting the microphones.
0
20 log [|AF (Ω)|] [dB]
–5
–10
–15
–20
0 10 20 30 40 50 60 70 80 90
Angle Ω (°)
Directivity 251
Task:
Determine the number of microphones in the array!
12.7 A sound system loudspeaker column can be considered a linear array of
small sound sources. Such columns are often mounted vertically and an angle
offset is obtained electronically using delay lines. In this case the inter-driver
distance is 12·10 −2 m.
Tasks:
a. Design a four-element array that has its main lobe 45° away from the line
interconnecting the loudspeakers.
b. Will there be any grating lobes at the upper frequency limit of 3.5 kHz?
12.8 The figure below shows two identical bidirectional microphones that are
placed at right angles in the z = 0 plane. A toroidal microphone pattern can
be achieved by the combination of two such microphones using a 90° phase-
shifting network (sometimes called a Hilbert transformer).
z
–
– + y
+
x
Task:
Show that the combination results in a toroidal directivity function around the
z-axis.
REFERENCES
1. Kinsler, L. E. et al., Fundamentals of Acoustics, 2nd edn., John Wiley & Sons, New York,
(1962) ASIN: B000LC9DO6.
2. Levine, H. and Schwinger, J., On the radiation of sound from an unflanged circular pipe,
Phys. Rev., 73(4), 383–406 (1948).
3. Ando, Y., Experimental study of the pressure directivity and the acoustic centre of the
circular pipe horn loudspeaker, Acustica, 20(6), 366–369 (1968).
4. Ando, Y., On the sound radiation from semi-infinite circular pipe of certain wall
thickness, Acustica, 22(4), 219–225 (1969).
5. Yong, S., Comparing Theory and Measurements of Woodwind-Like Instrument
Acoustic Radiation, Department of Music Research, Schulich School of Music, McGill
University, Montreal, Quebec, Canada (2009).
6. Keele, D., Effective performance of Bessel arrays. J. Audio Eng. Soc., 38(10), 723–748
(1990).
7. Jacobssen, F., A note on the concept of acoustic center, J. Acoust. Soc. Am., 115(4),
1468–1473 (2004).
8. Vanderkooy, J., The acoustic center—A new concept for loudspeakers at low frequencies,
Proceedings of the 121 Audio Engineering Society Convention, San Francisco, Paper
6912 (2006).
9. Ureda, M. S., On the movement of a horn’s acoustic center, Proceedings of the 106
Audio Engineering Society Convention, Munich, Paper 4986 (1999).
10. James, T. W., Effective acoustic center redefined, J. Acoust. Soc. Am., 62(2), 468–469
(1977).
11. Weeks, W. L., Antenna Engineering, McGraw-Hill, New York (1978) ASIN: B0000EG3RW.
12. Jasik, J. (Ed.), Antenna Engineering Handbook, McGraw-Hill, New York (1961) ASIN:
B001K58MHW.
13. van der Wal, M. et al., Design of logarithmically spaced constant-directivity transducer
arrays, J. Audio Eng. Soc., 44(5), 497–507 (1996).
14. Kock, W. E. and Harvey, F. K., Refracting sound waves, J. Acoust. Soc. Am., 21(5),
471–481 (1949).
13 Microphones and
Sound Fields
13.1 INTRODUCTION
The ideal microphone should convert the acoustic waveform of a sound pressure or
sound pressure gradient to an identical electric waveform. Physical and engineering
limitations make this impossible, for example, the presence of microphone will
change the sound pressure of the sound field.
The electrical transduction will have limitations that will affect the microphone’s
output signal. Some of these may be eliminated using signal processing, but at the
frequency band extremes there will always be differences between the sound pres-
sure time signal and the electrical time signal. A practical microphone will also
have nonlinearities and add noise, as will be discussed later. Because any practical
microphone is also a mechanical device, microphones will be sensitive to vibration,
entering through the housing, which will produce an electrical output as well.
Many microphones also contain electronic signal amplification and other
signal enhancement circuits. These will also contribute to the nonlinearities and
the background noise of the microphone. Some microphones are based on signal
conversion principles other than direct conversion, for example, modulation and
demodulation of high-frequency carrier signals. The carrier and conversion process
can add noise and distortion. Digital microphones have built-in analog-to-digital
converters that—while minimizing analog transmission errors—at the same time
can add all the errors and noise present in any analog-to-digital converter system.
13.2 INFLUENCE OF THE MICROPHONE ON THE SOUND FIELD

Microphones are usually small; a typical commercial electret microphone has a
cylindrical shape with length and diameter of a few millimeters. In spite of this, over
frequencies of a several kilohertz, even such small microphones will show direc-
tivity. Microphones can now be integrated directly on silicon and be as small as
allowed for by signal-to-noise considerations. Then, the directivity will depend on
the size of the entire assembly.
For a measurement microphone to have negligible influence on the sound field,
it must have small volume and surface area; cylindrical and spherical shapes are
preferred since they have symmetrical properties. At the same time it should have
low noise, but in the case of a measurement microphone any noise may many times
be overcome by repeated averaging. Most measurement condenser microphones
will have a microphone capsule that has a flat membrane stretched over the end of
253
a cylinder. The capsule is generally provided with a grille to protect the sensitive
diaphragm from mechanical damage. In use, the capsule is attached to a microphone
preamplifier, as described later, forming a complete microphone.
In the case of measurement microphones, the amplifier to which the capsule is
attached is typically shaped as a cylinder, usually with the same diameter as the
capsule, having a diameter in the range of 1–2 cm and a length of 5–10 cm. Such
a measurement microphone is acoustically similar to a long rigid cylinder with a
flat top.
Studio microphones in contrast must be large enough to be conveniently handled.
Often such microphones have a diameter of 2–5 cm and a length of about 5–15 cm
when the microphone body is included, even though the microphone capsule may be
much smaller.
Various grilles and other protectors may change the frequency and directional
response of the microphone capsule. One usually avoids putting the microphone
diaphragm in a cavity. The sound pressure at the bottom of a narrow cylindrical
cavity that has a rigidly closed end will vary with the depth of the cavity. Over
frequencies at which the width of the cavity is larger than about λ/2, the direction of
the incident sound will be important for the sound pressure at the bottom of the cavity.
One advantage of the condenser microphone over many other types of microphones
is that it lends itself to be manufactured without a cavity since the membrane can be
stretched over the “true” tube end, as sketched in Figure 13.1.
Figure 13.2 shows the level difference between the theoretical top center SPL
and the free-field SPL at the center of the end of a rigid cylinder for an incident
plane wave, as drawn in Figure 12.1a for various angles θ. We note that the sound
pressure at the center of the top of the cylinder varies by more than ±10 dB in the
high-frequency region because of the scattering of sound due to the microphone’s
presence in the field. We also note that the excess pressure due to scattering can be
minimized by having the sound incident parallel to the microphone diaphragm.
The pressure on the diaphragm varies considerably from point to point once
the cylinder becomes large compared to wavelength. Pressure cancellation for off-
axis incoming waves will further reduce the microphone’s sensitivity. The fact that
the microphone diaphragm is a membrane rather than a piston reduces this effect.
Membrane
Tension
adjustment
ring
Capsule Isolator
housing
Electrode and connector
FIGURE 13.1 A cylindrical measurement condenser microphone capsule.

Microphones and Sound Fields 255
10
Center relative to free field pressure (dB) θ = 0°
θ = 45°
5 θ = 90°
θ = 135°
–5
–10
–15
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5
Ratio cylinder diameter/wavelength
FIGURE 13.2 Theoretical values for the sound pressure at the center of the end of a cylinder
relative to the free-field sound pressure. (After Muller, G.G. et al., J. Acoust. Soc. Am.,
10(1), 6, 1938.)
Narrow sensing tube
FIGURE 13.3 A probe-tube microphone for sound pressure sensing.
Typically, the effective membrane area is about one-half of the geometrical membrane
area [2]. At high frequencies, there will be modes in the cavity that will be excited by
the motion of the membrane [3].
In some situations the microphone is flush mounted in a surface, for example,
when measuring pressure fluctuations in pipes and on surfaces such as fuselages. If
one uses a microphone capsule intended for flat sound pressure response there will
be little need for frequency-response correction as long as the microphone is small
compared to the wavelength.
One way of minimizing the effect of the microphone’s presence on the sound field
is to use a probe-tube microphone, such as the one shown in Figure 13.3. The tube
leading to the microphone leads to an air cavity where the microphone senses the
pressure. The tube forms a resonant circuit together with the compliance of the air of
the cavity. The viscosity of the air in the tube damps the resonances. Because of the
impedance of the microphone diaphragm there may be a slight increase in effective
cavity volume.
Using a second microphone for calibration it is fairly easy to linearize the fre-
quency response of the probe microphone. The frequency-response correction can
be achieved by acoustic, electric, or signal processing method or by a combination
of methods.
13.3 PRESSURE SENSING

A typical sound pressure sensing microphone will be based on the design principle
shown in Figure 13.1 and uses a membrane that moves due to the pressure difference
between its front and back. The back side faces a compliant air spring. The resonant
properties of membranes were discussed in Chapter 4. The membrane can often be
idealized as a mass/spring combination for frequencies up to and slightly above the
first resonance frequency of the membrane. The compliance and viscosity of the air
in the cavity will change the resonant conditions to some extent.
The net force F acting on the diaphragm will be proportional to the differential
sound pressure on the diaphragm p multiplied by the effective diaphragm area SD.
F = pS D (13.1)

The force can be used to generate a frequency-dependent movement of the diaphragm

as desired, depending on the transduction principle used, or the available frequency-
response equalization by acoustic, electric, or signal processing means.
13.4 PRESSURE-GRADIENT SENSING

13.4.1 Principles
The pressure gradient is the difference in sound pressure between two points in a
sound wave divided by the distance between the points. The pressure gradient can be
sensed either by using two microphones or by using a microphone capsule specially
adapted for the purpose. Pressure-gradient microphones are needed to determine
particle velocity, for example, in measuring sound intensity. The pressure-gradient
microphone is sometimes referred to as a “velocity” microphone although strictly
particle velocity sensing is different from pressure-gradient sensing [4]. In contrast to
array microphones, the directional characteristics of pressure-gradient microphones
will be largely independent of frequency.
Figure 13.4a shows a microphone capsule that allows access of the wave to both
sides of a single diaphragm or membrane. Usually this type of microphone uses
electrodynamic conversion with a diaphragm that has mainly mass-controlled
movement within its working range. If the mass is low, the diaphragm impedance will
be low as well. Drawbacks of this type of microphone are its extreme sensitivity to
mechanical damage, wind, and static position as well as the difficulties in calibration
because of its mechanical design, as will be discussed later. Once the size of the
≈ ∆l
≈ ∆l
(a) (b) Spacer
FIGURE 13.4 Examples of two types of pressure-gradient microphones: (a) based on direct
access of the incoming wave to both sides of a diaphragm; (b) based on the use of a matched
microphone pair. Note the spacer used to ensure a fixed distance between the microphones.
diaphragm becomes an appreciable part of a wavelength, the directivity due to the

membrane geometry will play a role as well in determining the microphone output.
One can also obtain pressure-gradient microphone by the use of two identical
conventional pressure-sensing microphones, such as those indicated in Figure 13.4b,
and phase-inverting one of the microphone outputs. The sum will give the pressure-
gradient characteristic after frequency-response equalization.
The practical bandwidth of this solution to the measurement of the pressure
gradient is limited by the ability to match the capsules to have the same characteristics,
since even small differences, particularly in phase response, will cause large errors
at low frequencies. Another drawback in using such microphone pairs is that the
signal-to-noise ratio will be insufficient for audio use.
In audio use, for example, studio microphones typically need to have a low
background noise level. It is difficult to manufacture microphones that have a noise
level as low as the threshold of hearing, Even if each microphone in a pressure-
gradient pair has such a low level, the effect of the subtraction of pressures results in
a reduced signal-to-noise ratio since the capsules’ self-noise signals are uncorrelated.
Conventional pressure-sensing microphones use a membrane that is stiff over the
audio range and has high impedance. The diaphragm movement due to incident sound
then becomes less, which limits the signal-to-noise ratio of the single microphone
and particularly of the microphone pair since the gradient-proportional signal is
obtained by subtraction of the single microphone signals from one another.
13.4.2 Plane Waves

Assume that the microphone has a very thin free diaphragm that is at
angle θ relative to the z-axis. The sound wave is incident in the negative z-direction
Incident
sound wave
Microphone y
diaphragm
x
FIGURE 13.5 Definition of the coordinate system for the discussion of pressure-gradient
microphone characteristics. The microphone diaphragm is at an angle θ relative to the z-axis.
The center of the microphone diaphragm is at the origin. The effective distance between the
two sides is Δl.
as shown in Figure 13.5. The area of one side of the diaphragm is SD, and the effective
distance between the two sides is Δl.
The incident plane wave field is
p ( z,ω ) = pe
ˆ jkz (13.2)

The pressure difference between the sides of the diaphragm is determined by angle,
effective distance Δl, and average pressure differential Δp/Δz. We can then obtain the
net force acting on the diaphragm as
∂p
F ( ∆l, S D , θ ) = ∆lS D cos (θ ) (13.3)
∂z
This can be rewritten to
F ( ∆l, S D , θ, ω ) = jkS D ∆l cos (θ ) pe

ˆ jkz (13.4)

The RMS-value of the force will then be
∆lS cos (θ )
F ( ∆l, S D , θ, ω ) = ωp (13.5)
c
Due to the |cos(θ)| term the microphone will have symmetrical directivity (except for
phase) around θ; so, the terms“bi-directional” and“figure-of-eight” are often used for
this microphone type. We also realize that the force on the diaphragm is proportional
to the particle velocity in the sound field which explains its fourth name—“velocity”
microphone. Direct sensing of the particle velocity is difficult [4].
13.4.3 Spherical Waves
Let us study the case when the incident sound wave is due to a spherically symmetric
sound field. The pressure in the sound field is then
A − jkr
p(r, ω) = e (13.6)
r
Using the expression for the equation of motion for spherical waves we obtain
∂p A  1 + jkr  − jkr
=− e (13.7)
∂r r  r 
Let us assume as before that the distance between front and back of the microphone
is small compared to wavelength. Using Equation 13.3 and substituting r for z we
find the net force acting on the diaphragm as
A  1 + jkr 
F ( ∆l, S D , θ, ω ) ≈ S ∆l cos(θ) e − jkr (13.8)
r  r 
The influence of kr is better shown if the equation is written as follows. The RMS-
value for the force as a function of distance is
1 + k 2r 2 S ∆l
F ( ∆l, S D , θ, ω ) ≈ p(r, ω) ω cos(θ) (13.9)
kr c
We see that the output of the microphone is proportional to pressure and frequency for
large values of kr, but that the output increases relative to pressure for small values of kr.
13.4.4 Proximity Effects

It is important to remember that the sound fields close to most sound sources are
characterized by a reactive near-field. It is only in the far-field (Fraunhofer region)
of a sound source that the sound pressure and particle velocity are in phase and
proportional. Close to large vibrating surfaces such as sheets carrying bending
waves below the critical frequency, the locations of maximum particle velocity and
pressure will be different. We have also seen in Chapter 12 that the reactive near-field
of a piston can extend far when its dimensions are large compared to wavelength.
Close to a monopole, the RMS particle velocity u~ is proportional to 1/r 2 while
the RMS sound pressure p~ is only proportional to 1/r. This proximity effect can
sometimes be used to advantage, for example, to increase the signal-to-noise and/or
direct-to-reverberation ratios. The level difference as a function of wave number k
and source distance r (use Equation 13.9) is given by
 1 + k 2r 2 
∆L = 20 log   (13.10)
 kr 

50
r = 0.05 m
40 r = 0.2 m
r=1 m
30
∆L [dB]
20
10
0
20 50 100 200 500 1k 2k 5k 10k 20k
Frequency [Hz]
FIGURE 13.6 Output level difference between a pressure gradient and a pressure
microphone as a function of frequency with source-to-microphone distance as parameter.
The microphones are assumed adjusted for flat frequency response in the far-field.
Figure 13.6 shows ΔL as a function of frequency for some distances of the microphone
to a monopole source. We see that the increase in level can be considerable and
that the frequency response will depend on the distance and frequency. To obtain
a “flat” frequency response, a first order Butterworth high-pass filter is needed for
equalization (see Appendix B). Note that the filter cutoff frequency must be set
according to the distance from the sound source.
In some applications such as in recording the sound field close to a sound source,
the pressure-gradient microphone increases the signal-to-(ambient)-noise ratio. It will
also increase the direct-to-reverberant sound ratio since both noise and reverberation
are likely to arrive from “far away” and thus will be almost plane waves. Because of
this property, the pressure-gradient microphone is often frequency-response equalized
and then used for “close-talking” applications such as “lip” microphones. It is then
important to design the microphone so that the distance to the lips is kept constant.
13.5 TWO WAYS TO ACHIEVE DIRECTIVITY

We have studied how directivity can be obtained in two different ways.
The first type of microphone is directive by having long or wide sound pickup
areas. Examples of such microphones are line-array microphones and microphones
which use parabolic reflectors. A variant of this type of microphone uses sound
pickup at multiple points; examples of such microphones are passive and active
array microphones using the principles described in Chapter 12. Array microphones
rely on the length, area, or volume size of the array to give the desired directivity.
Since all microphones have finite size, all microphones will become directional at
sufficiently high frequencies.
The second type of microphone is directive by using multipole techniques.

By adding the outputs from coincident microphones (i.e., microphones which
are essentially at the same point in space) that have different gradient orders and
angles, one can achieve many types of directivity functions. The pressure-gradient
microphone has a very sharp null, for example, in its directivity characteristic, which
is that of a dipole. Quadrupole and other multipole patterns can also be realized, both
passively and actively. As with the pressure-gradient microphone, the directional
characteristics of “multipole” microphone combination can be equalized to be
essentially frequency-independent as long as the microphone dimensions are small
compared to wavelength. This feature is often used in microphones for music and
voice recording. The advantage of multipole techniques is that the microphone can be
small. The disadvantage is lower sensitivity and associated high background noise.
13.6 COMMON MICROPHONE DIRECTIVITY PATTERNS

13.6.1 Gradient Order
The suppression of signals off the z-axis in pressure-gradient microphones
(assuming the coordinate system shown in Figure 13.5) is achieved by the cosine-
directional amplification factor and by the fact that the pressure gradient in a
spherical sound field (close to the source) is inversely proportional to the radial
distance to the origin squared in contrast to the pressure which only varies inverse
to the radial distance.
Typically higher order gradient microphones will be designed using combinations
of lower order gradient microphones, for example, by the combination of two gradient
microphones at some small distance along the z-axis. The resulting microphone
will have directional characteristics that are proportional to cos2(θ). Combining
more gradient microphones along the z-axis—at distances much smaller than the
wavelength—one can achieve other bidirectional characteristics [5].
Equation 13.11 defines a class of bidirectional directivity functions typical for
“gradient microphones.” The number n specifies the “gradient order.” A first order
gradient characteristic has n = 1. The zeroth order characteristic is omnidirectional.
F (θ) = cosn (θ) (13.11)
Such higher order gradient microphones are seldom used, however since the desired
noise suppression can often be better achieved by signal processing means.
13.6.2 Combinations of Gradient Order

By adding the outputs from coincident microphones (i.e., microphones which are
essentially at the same point in space) and which have different gradient orders and
angles, one can achieve various directivities. Unidirectional characteristics are often
desired and these can also be of various gradient orders.
Some common audio microphone directivities are shown in Table 13.1. These and
other directivities particular to microphones will be further discussed in Chapter 14.
262
TABLE 13.1
Some Important Audio Microphone Directivity Characteristics
Second Order Second Order
Characteristic Bidirectional Hypo-Cardioid Cardioid Hyper-Cardioid Super-Cardioid Bidirectional Cardioid
0° 0° 0° 0° 0° 0° 0°
20 dB
Polar response 20 dB 20 dB 20 dB
20 dB 20 dB 20 dB
pattern
F(θ) cos(θ) (7+3 cos(θ))/10 (1+cos(θ))/2 (1+3 cos(θ))/4 (1+2 cos(θ))/3 cos2(θ) (cos(θ)+cos2(θ))/2
−3 dB pickup arc 90° 89° 131° 105° 115° 60° 77°
Random −4.77 −2.84 −4.77 −6.02 −5.70 −6.99 −8.70
energy efficiency
(dB) (=−DI)
Distance factor 1.73 1.39 1.73 2 1.93 2.24 2.74
Nulls ±90° ±180° ±109° ±120° ±90° ±90° & ±180°
Electroacoustics
13.6.3 Unidirectional Microphones
The unidirectional characteristic has nominally zero sensitivity in the direction
opposite to the direction of maximum sensitivity. This property is achieved by
addition of the omnidirectional and bidirectional characteristics in the proportion
1:1. Using other proportions one can adjust the attenuation of sound from back and
thus the front/back ratio. There are several unidirectional directivity functions. The
most common one is the “cardioid” pattern given by the function
1 + cos(θ)
F(θ) = (13.12)
2
A common characteristic is also the super-cardioid directivity pattern, given by
1 + 2 cos(θ)
F(θ) = (13.13)
3
A similar type is the “hyper-cardioid” given by the directivity function
1 + 3 cos(θ)
F(θ) = (13.14)
4
This should not be confused with the “hypo-cardioid” that has the directivity function
7 + 3 cos(θ)
F(θ) = (13.15)
10
It is important to note that the patterns will be affected by the nonideal acoustical
circuit components, discussed in Chapter 14, as well as by scattering by the
microphone body. It is difficult to achieve good unidirectional characteristics at the
audio range frequency extremes.
Because the unidirectionality is achieved by combination of pressure and pressure-
gradient characteristics, such microphones will be subject to similar proximity
effects as those found for pressure-gradient microphones earlier in this chapter.
13.7 DIRECTIVITY FUNCTION AND DIRECTIVITY INDEX

The properties of directivity function and directivity plots were studied in Chapter
12. The gain factor, G, defined by Equation 12.16 is a power-related metric, which
describes the ratio between the microphone power output for sound incident in its
most sensitive direction and the mean power output over all angles. A level-related
metric of the microphone’s ability to reduce room noise and reverberation is the
directivity index, DI, defined by Equation 12.17.
A metric that is sometimes more intuitive than the directivity index is the
random energy efficiency, which gives the attenuation in dB of the microphone to
a diffuse sound field (e.g., random incidence noise or reverberation) relative to an

omnidirectional microphone having the same sensitivity in the nominal 0°-direction.
The random energy efficiency is by definition equal to the negative dB number of
the directivity index.
The distance factor listed in Table 13.1 shows the increase in source-to-microphone
distance that can be used for the same direct-to-reverberant sound ratio.
13.8 REPRESENTATION OF SCATTERING USING

ELECTROACOUSTICAL CIRCUITS
Since the microphone presents an obstacle to the incoming sound field, the sound
pressure at the diaphragm will be different from that in the free wave due to
scattering as discussed in Chapter 12. (Note that the following analysis does not
take into account the angle between the microphone axis and the arriving sound
field, but assumes that the sound field direction is always in the axial direction of the
microphone.) The discussion here follows that in [8].
Start by considering the internal impedance in the plane wave field discussed in
Chapter 7, particularly Figure 7.36. To have a sound pressure p at the observation
point, it is necessary to assume that the wave coming from −∞ has an internal
impedance of ρc and that when the wave progresses toward +∞, it sees the same
impedance value.
With a microphone the situation is more complex than for the infinite sheet
introduced for sound isolation purposes in Chapter 7, because the microphone has
a finite and small size. To study this situation, now insert an open cylindrical tube
having a cross-sectional area S = πa2 in such a way that the tube axis is parallel
to the incoming sound wave as shown in Figure 13.7a. Since the internal acoustic
impedance of the tube is Z A = ρc/S, the wave will enter the tube, and the analog
circuit for the system is that shown in Figure 13.7b.
Now consider a wave that comes from inside the tube and travels in the negative
x-direction as indicated in Figure 13.8. When the wave meets the open end it will
generate a sound field outside the tube. The radiation impedance Z AR into that
sound field is approximately that which would be seen by a piston at the end of
the tube.
Wave fronts of progressing wave ρc

RA =
S
−∞ ∞
U
2p1 pD ρc
RA =
S
−∞ x +∞
x=0 x=0
(a) (b)
FIGURE 13.7 (a) A plane wave enters a thin wall hollow cylinder that has an opening at x = 0.
(b) Acoustic impedance analogy. (From Leach, W.M., J. Audio Eng. Soc., 38(7/8), 566, 1990.)
Wave fronts of wave

inside and outside tube ρc
R A=
S
−∞ ∞
U
ZAR pD 2p2
−∞ x +∞
x=0 x=0
(a) (b)
FIGURE 13.8 (a) A plane wave in a hollow cylinder radiates from the tube opening at x = 0.
(b) Acoustic impedance analogy. (From Leach, W.M., J. Audio Eng. Soc., 38(7/8), 566, 1990.)
U
ρ
MA1 = 0.1952 —
a
RA1 CA1 a 3
CA1 = 5.43 ——
ρc2
ZAR p
MA1 ρc
RA2 RA1 = 0.1607 ——
a2
ρc
RA2 = 031.8 ——
a2
FIGURE 13.9 An improved circuit for modeling the acoustic radiation impedance seen by
a piston at the end of a long cylinder. (From Leach, W.M., J. Audio Eng. Soc., 38(7/8), 566,
1990.)
ρc
RA = ——
ZAR S
ZAR
H(ω)p1 pD 2p2 H(ω) = 1+
RA
FIGURE 13.10 A circuit that functions for waves coming from either source.
(From Leach, W.M., J. Audio Eng. Soc., 38(7/8), 566, 1990.)
For a narrow tube that has radius a, the radiation impedance Z AR for ka ≪ 1
is given by Equation 10.47. A somewhat better approximation covering the whole
frequency range (but still does not take the wiggles of the impedance curve around
ka = 1 into account) is shown in Figure 13.9 [6].
The circuit shown in Figure 13.10 will function for waves coming from either
source [8].
1 + jb1ω – b2ω2
HT (ω) =
1 + jc1ω − c2ω2
CA1
MA1
RA1 b1 = RA1||RA2CA1 +
(RA1 + RA2)||RA2
RA2
MA1
2RA1CA1MA1
b2 =
RA1 + RA2
HT (ω)p1 pD
MA1
c1 = RA1||RA2CA2 +
(RA1 + RA2)
RA1CA1MA1
c2 =
RA1+RA2
FIGURE 13.11 An example of a circuit that modifies the pressure p1 in the frontal
incoming plane wave to the pressure pD at the microphone diaphragm. The circuit is useful
for frequencies where the wavelength is longer than the diameter of the microphone.
(From Leach, W.M., J. Audio Eng. Soc., 38(7/8), 566, 1990.)
RA1 + RA2
C*A RA1
H1(ω) = 1+
MA1 RA2
+
H1(ω)p1 R*A pD RA2(RA1 + RA2)
R*A =
RA1
+ CA1RA21
p1 C*A=
(RA1 + RA2)2
(b)
FIGURE 13.12 A second example of a circuit that modifies the pressure p1 in the frontal
incoming plane wave to the pressure pD at the microphone diaphragm by the same factor as in
Figure 13.11. (From Leach, W. M., J. Audio Eng. Soc., 38(7/8), 566, 1990.)
The pressure increase at the diaphragm due to scattering of an axially incident

plane wave on the cylinder (up to the first pressure maximum, compare with
Figure 13.2) can be accounted for by simple additions to the impedance circuit. There
are several circuit alternatives possible. Either one can account for the directivity
pressure increase by introducing a transfer function to the pressure generator as
shown in Figure 13.11 or one can use a circuit that involves two sources as shown in
Figure 13.12. In both cases, the pressure increase due to the transfer function HT (ω)
is that of an average over all angles of incidence [8].
Of course, the circuits here can be modified to use the low-frequency ka ≪ 1
approximations to the radiation impedance Z AR that were given in Chapter 10 for a
piston at the end of a long cylinder [7].
REVIEW QUESTIONS
13.1 What could be the advantage of pressure-gradient sensing over pressure
sensing?
13.2 Explain the difference between particle velocity sensing and pressure-gradient
sensing.
13.3 Discuss the advantages and disadvantages of using pressure-gradient
microphones close to a sound source.
13.4 How can one design microphones that have unidirectional directivity?
13.5 Why will unidirectional microphones be subject to proximity effects?
13.6 What is the relationship between the directivity index and the random energy
efficiency factor?
13.7 When will room reverberation be attenuated by the random energy efficiency
factor?
13.8 Describe the differences between the various cardioid microphone
characteristics.
PROBLEMS
13.1 The proximity effect is the result of the increase of pressure gradient close to
a sound source.
Task:
Derive an expression that shows the influence of the proximity effect on the
output of an ideal cardioid microphone as a function of wave number and
distance.
13.2 Close to a large sound-absorbing plane, the sound field will not be diffuse, that
is, arrive from all spatial directions. Due to the sound absorption there will
only be sound from a solid angle of 2π.
Task:
Calculate the random energy efficiency of a cardioid microphone for this case.
13.3 Alternative ways of representing the acoustic impedance load on the
microphone and the scattering by the microphone body are discussed in
Section 13.8.
Task:
Compare the output level increase due to scattering by the microphone body
calculated using impedance components to that obtained by physical theory
shown in Figure 13.2. Assume a microphone diameter of 25 · 10 −3 m.
13.4 The 15 · 10 −2 m long probe tube of a microphone has an internal diameter
2 · 10 −3 m and leads to a chamber that has a volume of 20 · 10 −9 m3.
Tasks:
a. What will be the resonance frequencies of this system?
b. Assume that the microphone diaphragm and back cavity have equivalent
effective volume of 1 · 10–7 m3. How will the presence of this extra
compliance affect the resonances?
REFERENCES
1. Muller, G. G. et al., The diffraction produced by cylindrical and cubical obstacles and by
circular and square plates, J. Acoust. Soc. Am., 10(1), 6–13 (1938).
2. Hawley, M. S. et al., The Western Electric 640AA capacitance microphone. Its history
and theory of operation, in Wong, G. S. K. and Embleton, T. F. W. (Eds.) AIP Handbook
of Condenser Microphones: Theory, Calibration and Measurements (Modern Acoustics
and Signal Processing), American Institute of Physics, New York (1994) ISBN-13:
978–1563962844.
3. Morse, P., Vibration and Sound, American Institute of Physics, New York (1991) ISBN-
13: 978–0883188767.
4. Schultz, T. J., On the distinction between velocity-sensitive and pressure gradient-
sensitive microphones, J. Acoust. Soc. Am., 28(3), 498 (1956).
978–0883184943.
7. Bauer, B. B., On the equivalent circuit of a plane wave confronting an acoustical device,
J. Acoust. Soc. Am., 42(5), 1095–1097 (1967).
8. Leach, W. M., On the electroacoustic-analogous circuit for a plane wave incident on the
diaphragm of a free-field pressure microphone, J. Audio Eng. Soc., 38(7/8), 566–568
(1990).
14 Microphones
14.1 INTRODUCTION
In this chapter, we will study the electroacoustic properties of microphones in
their linear operating range. Electroacoustic and electromechanical analogies will
be used to illustrate the working principles of various microphone transduction
techniques. These microphones are reciprocal transducers and the analysis here
will be used to advantage in our studies of loudspeakers. A classic review of
microphone types is found in Ref. [1] and a more modern one in Ref. [2]. Some
specialized microphones are based on heat or ultrasonic sensing techniques, thus
not using sound field sensing by electroacoustic means and will not be considered.
Refs. [3,4] discuss many different types of microphones and their electroacoustic
analogies.
The microphone’s diaphragm or membrane converts the acoustic pressure
variations to mechanical vibrations that can be transduced into an electric signal
as indicated in Figure 14.1. In this text, we will consider thin uniformly stretched
membranes where the movement due to the sound pressure is determined by its
mass and elasticity and membranes in which the movement is determined by the
material’s bending stiffness, mass, and elasticity. In electroacoustics, a diaphragm is
usually considered a more piston-like construction where the nominally rigid piston
is suspended by a compliant ring. Microphones that use capacitive transduction
usually have membranes whereas those that use electrodynamic transduction have
diaphragms.
14.2 DIAPHRAGMS AND MEMBRANES

14.2.1 Membranes
A tensioned circular membrane is supposed to move in phase over its entire surface.
At low frequencies, the sound pressure will be virtually identical over all points and
the membrane’s static displacement is given by Equation 4.10.
The membrane however, like a string, features resonances at which it can vibrate
in various modes. The resonances will be damped by radiation, internal losses in the
membrane, and acoustic resistances. The displacement patterns of some modes in
a circular membrane are shown in Figure 4.5. Figure 4.9 shows the corresponding
modes for a circular, stiff diaphragm clamped at its edge. The modes are formed
both radially and circularly as shown in the figures. For the membrane in Figure 4.5,
the ω1,0 mode having qr = 1 (radial) and qm = 0 (circular) is the lowest frequency mode.
Close to the resonance frequencies of the respective modes, the frequency response
of the mode movement corresponds to that of a simple linear mass–spring system
269
Membrane
(distributed mass and compliance)
Transduction
sensor
Cavity
(a)
Diaphragm (mass)
Flexible edge suspension

(compliance)
Transduction
sensor
Cavity
(b)
FIGURE 14.1 Typical designs: (a) a membrane having distributed mass and compliance and
(b) a diaphragm having a mass suspended by a compliant edge.
close to resonance. Inexpensive electret microphones and most studio condenser

microphones have plastic membranes that are sputtered with a thin metal layer, but
classical professional condenser microphones for acoustic measurement use nickel
membranes that exhibit stiffness as well. Microphones have been made that use
embossed metal membrane to increase stiffness.
Even if the incoming acoustic wave is plane and normal to the membrane, the
wave field in front of the membrane will be distorted, which helps excite the various
modes. The higher-order modes also have higher resonance frequencies. Generally,
one avoids using the microphone at frequencies where these modes become
important for the membrane motion. Condenser microphones are regularly designed
to be used at and somewhat above the first radial mode frequency. One can extend
the frequency range of condenser microphones considerably by carefully adjusting
the diffraction/reflection due to the microphone housing and its shape to the effects
due to the membrane resonances [3,7].
The ribbon microphone uses a “linear” membrane that in principle vibrates in
one direction only and that has its first resonance at the lower end of the audio range.
The transversal modes of the ribbon are similar to those of a string. Unless carefully
damped, the ribbon has many audible resonances.
14.2.2 Diaphragms
Microphone capsules using compliance suspended diaphragms also exhibit many
resonant modes and associated vibratory behavior. Besides the modes of the
diaphragm itself, the diaphragm’s mass together with the suspension’s compliance
at low frequencies forms a single-degree-of-freedom vibratory system that has a
resonance usually somewhere in the middle of the audio range. Because of the wires
needed to connect the electrical leads to the voice coil, which is mounted on the
Microphones 271
MAD CAS RAS ZAD

UD UD
UD
p1 p2 p1 p2 p1 p2
(a) (b) (c)
FIGURE 14.2 The microphone’s mechanical system neglecting the air load. (a) Simplification
of diaphragm mechanics. (b) The acoustic impedance analogy. (c) The simplified acoustic
impedance analogy.
diaphragm and the associated mass imbalance, there are also “rocking modes” where
the (circular) diaphragm typically rocks from side to side along a diameter axis.
14.3 MICROPHONE ANALOGIES

Simply stated, most microphone diaphragms or membranes may be thought of
as simple linear single-degree-of-freedom resonant vibratory systems as shown
in Figure 14.2. The diaphragm is usually considered behaving as a compliantly
suspended stiff mass. Below its first resonance frequency, a membrane can be
considered a compliantly suspended mass. This vibratory mass of the membrane
will only be about half of its real mass.
The force FD will be p̲ SD, where SD is the effective surface area of the microphone
diaphragm (which will also include part of the suspension ring) or membrane,
and p̲ the pressure differential between the front and the back. Since microphone
capsules are acted upon by the incoming sound pressure and contain many acoustic
components, we prefer to represent the diaphragm’s mechanical properties by an
acoustic impedance analogy as shown in the figure. In these analogies, the current
flowing in the circuit is the volume velocity UD driven by the pressure difference
created by the pressures on the sides of the diaphragm, p̲ 1 and p̲ 2. The mass–spring–
damper system forms a series circuit in the impedance analogy and a parallel circuit
in the mobility analogy.
14.3.1 Pressure Microphones

A sketch of the acoustomechanical layout of the pressure-sensitive microphone
capsule is shown in Figure 14.3. The volume at the back side of the microphone
capsule is isolated against the sound pressure but is pressure equalized so that quasi-
static air pressure changes will not affect the function of the transducer as an acoustic
(audio frequency range) sensor. The diaphragm rests on the air cushion in the closed
cavity. One could also build a microphone where the diaphragm would rest on some
flexible material having a low Young’s modulus.
If the microphone capsule is much smaller than any wavelength of interest,
the pressure p̲ F will be the same over the capsule. The pressure p̲ of the incident
p ZAR ZAD
Closed UD
cavity
pF UD pD
p pF pC CAB
Diaphragm ZAD Capillary tube

(a) (b)
FIGURE 14.3 The basic pressure-sensitive microphone. The diaphragm is weakly coupled
to the transduction mechanism that is not included in the schematic: (a) Physical construction.
(b) Acoustic impedance analogy. The impedance Z AD will depend on the transduction
principle used. The air of the closed cavity forms a soft compliance CAB. In this circuit, Z AD
also includes the electrical side on diaphragm motion. The capillary tube is not included in
(b) because of its high impedance.
sound field is sensed by the microphone through the radiation impedance Z AR at

the diaphragm as shown in Figure 13.1 so the acoustomechanical circuit for the
microphone capsule becomes that shown in Figure 14.3b.
1
Z AD +
jωC AB
pF = p (14.1)
1
Z AR + Z AD +
jωC AB

Generally, however, the radiation load Z AR is very small compared to the other
components of the circuit and may be neglected so that pF = p. The quantities of
interest are either the volume velocity or the displacement of the diaphragm or
membrane. If there is no electrical load, the volume velocity will be
p
UD = (14.2)
1
Z AR + Z AD +
jωC AB

We will see that the analogy of the closed box loudspeaker is almost the same,
less the electric side.
As with the loudspeaker box, there is a need for at least a small leak so that the
static pressure inside the cavity is equalized to the ambient pressure, since otherwise
the diaphragm might be displaced relative to its position of rest. This is done using
a capillary tube as indicated in Figure 14.3. Because of the small dimensions of a
microphone, diaphragm displacement would be a larger problem in microphones
than in loudspeakers. If we introduce a small leak into the capsule, we obtain the
basic mechanical capsule design and circuit shown in Figure 14.4. Since the mass of
air inside the capillary tube is very small, the reactance ωMleak of the tube is small
compared to the resistance Rleak.
Microphones 273
Closed
cavity ZAD Rleak
UD UB Mleak
pF UD
pB pD pB
pF pC CAB
Diaphragm ZAD
Capillary
(a) tube (b)
FIGURE 14.4 (a) A static-pressure-equalized pressure-sensitive microphone and (b) its

acoustic impedance analog. The acoustic resistance of the leak is designed so that the product
RleakCAB has a value that is much larger than the period of oscillation at the lowest frequencies
of interest. Usually, the product RleakCAB is about several seconds. The reactance ωMleak of the
tube is very small compared to Rleak so it has been neglected here. The small influence of the
radiation impedance is also not represented in the analogy.
The pressures at the front and the back of the capsule will be almost identical at
low frequencies. Since the acoustic resistance of the leak is very large, and has a time
constant of several seconds (in order not to affect the audio response), we can neglect
the leak in our audio range analysis of the microphone. It is important to note though
that, even if the cutoff frequency of the low-pass filter formed by the leak resistance
and the capsule air volume’s compliance is below the audio range, there will still be
a phase shift introduced into the response of the microphone. By having a second
phase shift network at the back of the microphone capsule, it is possible to reduce the
influence of this phase shift on the microphone’s response.
14.3.2 Transducer Response Alternatives for the Pressure Microphone

The sensing of the diaphragm movement is typically done using electrodynamic,
capacitive, or piezoelectric transducers. Since the movement of the diaphragm will
depend on the acoustomechanical properties of the resonant circuit, essentially
consisting of the diaphragm mass, suspension compliance, and losses, the alternatives
for achieving a frequency-independent electric response to the sound pressure are
three: mass, compliance, or resistance control of the diaphragm movement.
Mass control movement needs a sensor that has a response proportional to
diaphragm acceleration, resistance control needs a transducer response proportional
to diaphragm velocity, and compliance control needs a transducer response
proportional to diaphragm displacement. Of course, with electronic equalization,
these rules need not strictly be adhered to, but compromises are likely to influence
the microphone’s signal-to-noise ratio.
14.3.3 Pressure-Gradient Microphones

Pressure-gradient microphones may be designed using two pressure microphones
at some distance and subtracting the electric power output of one from the other.
This method however usually leads to a microphone with reduced dynamic range,
∆l
ZAR ZAD UD
pF pB
UD pF pB
Diaphragm ZAD
(a) (b)
FIGURE 14.5 (a) A pressure-gradient-sensitive microphone and (b) its acoustic impedance
analog. The acoustic path length Δl is frequency dependent. Here Z AR is the combined
radiation impedance for both sides.
because of the small differences in pressure acting on the microphones at a small

distance. It also requires matched microphones and microphone electronics to avoid
errors because of amplitude and phase differences in their transduction processes.
Because of these factors, most pressure-gradient microphones are constructed to
allow the pressure gradient of the sound field to act directly on the diaphragm. In
this way, the differences due to electronics, etc., are eliminated, but it is still difficult
to design a pressure-gradient microphone to have completely symmetrical pressure-
gradient sensing around its axis.
The acoustomechanical analogy of the pressure-gradient microphone is shown in
Figure 14.5. Note that the effective path length is frequency dependent.

Pressure-Gradient Microphone
The force acting on the diaphragm will in this case be proportional to the pressure
gradient, which in turn is proportional to the product of particle velocity and
frequency. Since the particle velocity is proportional to sound pressure in a plane
wave, the equation of motion in gases is
∂p ω
= − jωρ u = − j p (14.3)
∂x c
Assuming the angle of incidence θ = 0, the equation of motion for the diaphragm
becomes
∂p ω  1 
∆l = − j p∆l =  jωM AD + RAS + U D (14.4)
∂x c  j ω C AS 

Microphones 275
After rearrangement, we obtain
c  R 1 
p= − M AD − AS + 2 U D (14.5)
∆l  jω ω C AS 

It is clear from this equation that for a flat frequency response, a microphone capsule
using a diaphragm driving a velocity-sensitive transducer will need to be mass con-
trolled. A transducer sensitive to diaphragm displacement will need to be resistance
controlled since
uD = jωx D (14.6)

This leads to
cS  1 
p=− jωM AD + RAS + xD (14.7)
∆l  jωC AS 

The electrodynamic transducer is the transducer of choice for passive pressure-

gradient microphones. Other designs using, for example, dual capsules and electrical
networks to sense the pressure gradient will necessarily require more complicated
electrical designs. Dual capsules however allow the choice of microphone
characteristic at the microphone by simple electrical switching and thus affords
larger flexibility in the recording venue.
14.3.5 Combination of Pressure and Pressure-Gradient Microphones

In some recording situations, we wish to have a microphone that in some sense has a
“unidirectional” directivity function such as
1 + A cos(θ)
F (θ) = (14.8)
1+ A
The definition of the angle θ is shown in Figure 13.5. The desired unidirectional
characteristic can be obtained if one combines the directional characteristic of a
pressure microphone with that of a pressure-gradient microphone. The value of A
in Equation 14.8 will determine the particular shape of the unidirectional response.
Various microphone directivities were shown in Table 13.1. Figure 14.6 shows
common cardioid family microphone directivity characteristics.
There are several ways of obtaining a unidirectional characteristic, for
example, by using a combination of two capsules or to use an acoustical design
allowing acoustic use of the front and back sound pressures. Several directional
characteristics of the unidirectional type can be obtained by both methods. There
are considerable practical problems in combining two microphone capsules using
different transducers. Not only will the microphone capsules need to be spaced
Hypo cardioid, A = 3/7 Cardioid, A = 1 Super cardioid, A = 2 Hypercardioid, A = 3

0° 0°
–10 dB –10 dB
–20 dB –20 dB
–30 dB 0 dB –30 dB 0 dB
–40 dB –40 dB
–90° 90° –90° 90°
180° 180°
FIGURE 14.6 Some different cardioid microphone characteristics and their dependence
on the parameter A in Equation 14.8. Angle shown is θ in Figure 13.5. Since the patterns are
symmetrical around the z-axis, only one half of each pattern is shown.
apart, the phase characteristics are likely to be very different particularly at

the audio range extremes so the microphone combination would only function
properly over a fairly narrow frequency range. For these reasons, it is generally
preferred to achieve the unidirectional characteristic by acoustical means. It will
in any case be difficult to obtain the same directivity characteristic over the entire
audio frequency range.
Matched microphone pairs are sold for sound intensity measurement that requires
summation and subtraction of sound pressure. These are very expensive. Another
problem is that, at low frequencies when the front and back pressures are very similar,
the resulting output signal will be low, reducing the effective signal-to-noise ratio
because of the noise of the electronics.
The standard method using an electroacoustic design is to subject the diaphragm
to sound on both sides. The sound reaching the back side is delayed (compared to a
bidirectional microphone) by using a phase-shifting network. Such a design is shown
in Figure 14.7. The circuit is driven by two pressures, the pressure at the front of the
capsule driving the diaphragm directly and the pressure at the rear of the capsule that
drives the diaphragm through an acoustic resistance/capacitance circuit.
The figure shows the acoustomechanical construction of a unidirectional
microphone and its electroacoustic analogy as an acoustic impedance analogy. The
microphone diaphragm has the acoustic impedance Z AD, the port has the acoustic
resistance R AP (its reactance ωMAP is neglected since it is small), and the cavity
volume has the acoustic compliance CAB.
Assume that the microphone is in a sound field that is a plane incident wave p̂e−jkx
and that the microphone is small compared to the wavelength of sound. The pressure
p̲ F in front of the microphone diaphragm at x = xF is then
ˆ − jωxF / c (14.9)
pF (ω, xF ) = pe

Microphones 277
∆l
ZAD UD UP (MAP) RAP

pF UD pB
UB
pF CAB pB
Diaphragm ZAD Acoustic

resistance
Cavity
(a) (b)
FIGURE 14.7 (a) Combination pressure and pressure-gradient-sensitive microphone and

(b) its acoustic impedance analog. The time constant product R ABCAB is much shorter than that
in the microphone shown in Figure 14.8. The acoustic path length Δl is frequency dependent.
Here xF is the location of the microphone diaphragm and that the pressure pr at the
rear microphone entrance at the rear port at xB = xF + Δl is
∂p
pB (ω, x B ) = pˆ (1 + ∆l )e − jωxB / c (14.10)
∂x
We need to calculate the current, i.e., volume velocity UD, through the impedance
ZAD to find the vibration velocity of the diaphragm uD. This impedance represents the
acoustic mass, compliance, and resistance of the diaphragm. The current determines
the electric output of the microphone. If the current can be made both frequency
independent and proportional to (1 + cos(θ)), we have designed a unidirectional
microphone capsule that can use a velocity-sensitive transducer. The capsule should
then ideally provide a flat frequency response when placed in a plane wave field
irrespective of the frequency of sound. The output voltage of an electrodynamic
transducer will be
Bl
e= U D (14.11)
SD
To obtain the volume velocity UD through the diaphragm impedance Z AD, we

need to analyze the two loops of the circuit. Expressing the pressures p̲F in the front
and p̲B in the back of the diaphragm as functions of the volume velocities “through”
the diaphragm UD and through the rear port UP, we find that
1
pF = U D Z AD +
jωC AB
(U D + U P ) (14.12)

and
1
pB = U P RAP +
jωC AB
(U D + U P ) (14.13)

It is the vibration velocity uD that is of interest because it generates the electric

output of the microphone. By eliminating UP, we obtain the pressure over the
diaphragm impedance Z AD as
 pF − pB 
Z AD  pF RAP +
 jωC AB 
pD = U D Z AD = (14.14)
RAP + Z AD
Z AD RAP +
jωC AB

and since the difference between the pressure in the front and the back of the
microphone capsule is
pF − pB ≈ pF jk∆l cos (θ ) (14.15)

we have
 k∆l cos (θ ) 
Z AD  RAP +
 ωC AB 
p D = pF
R + Z AD (14.16)
Z AD RAP + AP
jωC AB

Since
pD
UD = (14.17)
Z AD
we obtain the expression for the diaphragm velocity UD without any electrical load
on the microphone’s output terminals as
 ∆l cos (θ ) 
 1+
pF  cRAPC AB 
uD = (14.18)
S D 1 + 1 / Z AD + 1 / RAP
jωC AB

Because of the electrodynamic transducer characteristic, given by Equation 14.11,

we obtain the microphone’s electrical sensitivity after normalization as
1 + A cos (θ )
e (θ ) ∝ pF (14.19)
1+ A
Equation 14.19 is similar to Equation 14.8. Figure 14.6 shows some of directivity
characteristics that can be obtained by changing the value of A in Equation 14.19.
Microphones 279
14.4 ELECTRODYNAMIC TRANSDUCERS

The electrodynamic transducers operate using the attracting and repelling forces of
magnetic fields. Electrodynamic and magnetic microphone transducers operate at
low impedances and are generally resistive devices that can operate with rather long
cables without appreciable reduction in sensitivity or linearity of frequency response.
The transduction ratio depends on the magnetic field’s flux density and the length of
the electric wire in the field. The transducer’s electromechanical analogy was shown
in Figure 9.4. Two types of electrodynamic microphones are common, those that use
a moving coil and those that use a moving ribbon.
14.4.1 Moving Coil Microphones

The basic principle of the pressure-sensitive electrodynamic microphone is shown in
Figure 14.8. Figure 14.9 shows the development of its acoustic impedance analogy.
Since the diaphragm of this microphone needs to have its mechanical motion
resistance controlled, there will need to be a compromise between sensitivity and
bandwidth. A wide bandwidth along with reasonable sensitivity typically requires
the use of various resonant acoustic networks.
With the acoustic, mechanical, and electric impedances expanded into their
components, we find the analog circuit shown in Figure 14.10.
Figure 14.11 shows an improved moving coil microphone with some additional
acoustical circuit components to enhance its frequency response. We will see how
these components enhance the frequency response at the passband edges of the
microphones frequency response shown in Figure 14.17. The microphone has a grill
protecting the diaphragm, then the diaphragm proper with its moving coil and mag-
netic field gap, and the volume behind the diaphragm. The back volume is vented to
the outside using a tube.
The new, more complex acoustical analogy is shown in Figures 14.12 through
14.14. Note that there is little interaction between the front and back sides for sound
pressure and radiation impedance since the tube and the diaphragm are small and
the diaphragm is primarily active above the tube resonance frequency.
Flexible edge UD ZAD

Diaphragm p
suspension F
p
B
Voice
coil Airgaps
CAB
Magnet Cavity
Soft iron
structure
(a) (b)
FIGURE 14.8 The basic electrodynamic microphone (a) and its electroacoustic analog (b).
The voice coil has the electric impedance ZEL , not indicated in the figure. The impedance Z AD
includes the mass, compliance, and losses of all moving mechanical parts.
zAB p SD:1
B
UD
SD:1 ZEC
zARF F zMD 1:Bl
p UD e eout ZEL
F uD
(a)
zARF zAB ZEC

SD:1 F zMD 1:Bl
p e eout ZEL
F UD uD
(b)
zARF ZEC
zAB zAD SD:Bl
p UD e eout ZEL
F
(c)
zEL
zEC
ZARF ZAD Bl:SD
p ZAB
F
(d)
FIGURE 14.9 The mobility analogy (a) of the basic moving coil microphone shown in
Figure 14.8, redrawn to join the front and back sides of the diaphragm in (b), redrawn to join
the acoustical and mechanical sides into one acoustic mobility analogy (c), and finally in (d)
the acoustic impedance analogy.
Microphones 281
LEC SD2
(Bl)2
CAS (Bl)2
ZARF RAS MAD UD REC SD2
pD CAB
(Bl)2
ZELSD2
FIGURE 14.10 The acoustic impedance analogy for the microphone shown in Figure 14.14
with the various acoustical, mechanical, and electrical components all transferred to the
acoustic side.
Radiation impedance, ZARF
Grill, MAG RAG Diaphragm, suspension,

pF and coil, MAD CAS RAS
Air cavity, CAB1

Air cavity, CAB3 Air cavity, CAB2
Resistance, RAH Air cavity, CAB4
Airgaps,
MAVC1RAVC1 Air cavity, CAB5
MAVC2 RAVC2
Tube, MATRAT pB
Radiation impedance, ZARB
FIGURE 14.11 An improved moving coil microphone having additional acoustical circuit
components.
ZARF MAG RAG

UD
pF CAB1
FIGURE 14.12 The acoustic impedance analogy for the front side circuit of the microphone
RAVC1 MAVC1 RAH RAT MAT ZARB

UD
MAVC2 RAVC2
CAB2 CAB4 CAB5 p
B
CAB3
FIGURE 14.13 The acoustic impedance analogy for the circuit to the back of the diaphragm
in the microphone shown in Figure 14.11.
FD LEC RE
rMS CMS MMD 1:Bl i
e0 ZEL
UD
FIGURE 14.14 The electrical and mechanical circuits of the diaphragm of the microphone
The acoustic impedance analogy, for the front circuit of the microphone shown
in Figure 14.11, consists of the grille and the driving sound field pressure as shown
in Figure 14.12.
Similarly, the acoustic impedance analogy, for the back circuit of the microphone,
consists of the circuit components and the driving sound field pressure as shown in
Figure 14.13.
The diaphragm mass and compliance are in this case best thought of as joined
with the transducer transformer as shown in Figure 14.14. Note that since the
circuits in Figures 14.13 and 14.14 are acoustic impedance circuits, it is necessary
to introduce a conversion from mechanical to acoustic impedance analogy as shown
in Figure 14.15. The current in this circuit is the velocity uD of the circuits shown
previously in Figures 14.12 and 14.13. Combining the three circuits, Figures 14.12,
14.13, and 14.15 into one, we obtain the complete analogy shown in Figure 14.16.
LEC
(Bl)2
(Bl)2 e0
RMS CMS MMD
UD REC Bl
FD (Bl)2
ZEL
FIGURE 14.15 The mechanical impedance analogy for the circuit shown in Figure 14.14.
Microphones 283
p
D
(Bl)2
RAS CAS e0
MAD ZEL SD2
BlSD
MARF + MAG RAG UD

(Bl)2
REC SD2
CAB1
LEC SD2
(Bl)2
RAH
MARB + MAT RAT MAVC2 RAVC1 MAVC1
p p CAB5 CAB4 CAB2

F B
RAVC2
CAB3
FIGURE 14.16 The complete acoustomechanical impedance analogy for the electro-
dynamic microphone in Figure 14.11. The radiation impedances Z ARF and Z ARB have been
replaced by their respective radiation masses MARF and MARB since the radiation resistances
are negligible.
Figure 14.17 shows the typical frequency response of the microphone described.
Note how the medium-frequency-range response is kept flat, thanks to diaphragm
damping, and how the resonances at the audio frequency band extremes extend the
response.
14.4.2 Ribbon Microphones
The simplest pressure-gradient microphone is the ribbon microphone shown in
Figure 14.18. It consists of a limp metal ribbon that has been stiffened by creasing
and suspended between the poles of a permanent magnet. We note that it is an
electrodynamic microphone, where the ribbon is used both as a differential pressure
sensor and as an electrical circuit component.
The acoustomechanical analogy for the ribbon microphone is shown in
Figure 14.19. Because of the very weak coupling between the electrical and
acoustical sides of the ribbon microphone, the electrical side is eliminated in the
acoustic impedance analogy circuit shown in Figure 14.20. The electrical side can
be introduced in the same way as in the pressure microphone studied earlier, i.e., in
the branch containing the ribbons mass MAD, compliance CAD, and losses R AD. Note
however that the ribbon has very low mass so that the damping introduced by the
20
With resonant circuits
w/o resonant circuits
10
Frequency response [dB]
–10
–20
–30
0.02 0.05 0.1 0.2 0.5 1k 2k 5k 10 k 20 k
Frequency [kHz]
FIGURE 14.17 The curves show (solid line) an example of a theoretical frequency response
curve for an electrodynamic microphone. (dashed line) As above but with extra resonant
networks to improve effective frequency range ( ±3 dB limits).
Ribbon
holders Thin, pleated
and metal foil ribbon
electricals
terminal
Soft iron
Magnet
FIGURE 14.18 Front view of a ribbon microphone.
branch containing the impedances of the slits, the mass MAS, and the resistances R AS
are important components. In addition, the ribbon will be somewhat damped by the
radiation impedance Z AR.
If the ribbon is insufficiently damped, one will note the resonances that are a
result of its limp, but string-like character. The fundamental resonance is usual
placed at around 30 Hz, so there will be many higher modes in the audio frequency
range. These modes need to be damped to avoid their resonances to affect the sound
quality of the microphone.
Microphones 285
1:SD
YARF p
UC F
REC LEC MMD CMS rMS 1:SD

Bl:1 FC
REL e UC UC YARF p
B
FIGURE 14.19 The mechanical mobility analogy for the gradient microphone.
MAS RAS
ZAR UD MAD RAD CAD
pF pB
FIGURE 14.20 The acoustic impedance analogy circuit for the gradient-sensing ribbon
microphone.
14.5 ELECTROMAGNETIC MICROPHONES

The operating principles of electromagnetic transducers were discussed in Chapter 9.
The electromagnetic microphone is a classical design in which the magnetic flux,
provided by a permanent magnet, is modulated by an air gap [6]. The air gap opening
is in turn modulated by the sound pressure acting on a diaphragm. The movement
of the diaphragm can be used directly using the diaphragm as a part of the magnetic
circuit or the diaphragm can act on a hinged “balanced armature.” The two principles
are shown in Figure 14.21. The direct method shown in Figure 14.21a will result in
insensitive microphones since the static magnetic force requires a stiff diaphragm to
prevent the diaphragm from being sucked in to the pole piece. The balance armature
principle shown in Figure 14.21b can be used to provide sensitive microphones and
lends itself to miniaturization. In the balanced armature microphone, the armature
carries only the differential flux. The output voltage generated is proportional to
armature velocity.
Cobalt-iron microphone
diaphragm Drive coils
Air cavity
Soft iron poles Magnet

(a)
Microphone diaphragm
and drive rod
Suspension N
Air cavity S Center
Sensing coil pole piece
N Magnets
S
Flexing balanced
(b) armature
FIGURE 14.21 Two types of moving armature microphones. The microphone in (a) uses
an unbalanced magnetic circuit and the one in (b) a balanced magnetic circuit requiring less
stiffness that results in higher sensitivity and less nonlinear distortion.
14.6 PIEZOELECTRIC AND FERROELECTRIC TRANSDUCERS

Piezoelectric and ferroelectric transducers use materials that will convert movement,
such as compression/expansion or bending, into an electric voltage at the terminals
of the transducer or vice versa. Quartz and lithium sulfate are examples of naturally
piezoelectric materials. A ferroelectric material that is common in electroacoustic
transducers is lead-titanium-zirconate.
Since these materials have very high Young’s modulus, it is necessary to shape
the materials so that the mechanical input impedance sensed by the diaphragm is
as low as practical. The diaphragm is part of an acoustomechanical impedance
converter.
Both piezoelectric and capacitive transducers are characterized by capacitive
properties that require an amplifier to isolate the capsule so that the cable or other
load will not destroy the linear frequency characteristic or affect the signal level.
Because of the high input impedance of the amplifier used with these transducers,
they can be usefully modeled using a simple transformer circuit, as shown in
Figure 14.22.
The electromechanical analogy of the transducer is shown in Figure 14.23 [7].
The electric capacitance CE is typically a few nanofarad. The transformation ratio
K ME will depend on the physical properties and size of the bar.
Microphones 287
Flexible edge Diaphragm

suspension
Piezoceramic
bender Cavity
FIGURE 14.22 A schematic drawing of a piezoelectric (or ferroelectric) transducer using a

bender bar-type element.
MM CE
uD KME :1
FD CM e0
FIGURE 14.23 The electromechanical impedance analogy for the bender bar below and
around the fundamental resonance frequency, when the bar is free to move at one end only.
The effective mass MM and compliance CM of the bar can be measured with the electrical
circuit open. The diaphragms mass and compliance are assumed to be included in the bar
components.
Because of the (usually) very high-impedance electrical load on the capsule, the
electrical side of the transducer can be modeled as a voltage generator in series with
a capacitance. The mechanical side will be a mechanical device such as a disk or bar
operating below the first mechanical resonance of the device. One end of the transducer
is kept fixed and the other attached to the diaphragm that needs to be stiff to drive the
large mechanical impedance of the bar or disk without deformation. The mechanical
impedance of the bar or disk at the attachment point will be that of a mechanical
compliance at frequencies below the resonance region. The damping of the diaphragm
will be important for the frequency response of the microphone capsule.
14.7 CONDENSER MICROPHONES

Permanently charged and bias-charged capacitive transducers were discussed in
Chapter 9. The condenser microphone usually consists of a microphone capsule
that acts as a sound-pressure-dependent capacitor and some associated electronics
to convert the capacitance change to a voltage or current. The capacitor has at
least one electrode that is formed as a diaphragm whose movement follows the
pressure of the sound field at the microphone. Some condenser microphones use two
capacitors to form the microphone capsule; such capsules can be used to achieve
variable directivity. Some microphones use several capsules to obtain simultaneous
recordings of the sound field. A comprehensive review of condenser microphones
may be found in Ref. [8].
Condenser microphones sense the sound-pressure-caused displacement of the
diaphragm. The basic condenser microphone is pressure sensing, but pressure
gradient and combinations of pressure and pressure-gradient characteristics that give
various forms of cardioid directivity characteristics are very common. One of the
attractive properties of the transducer is the ease by which it can be engineered to
give the desired directivity. In some sonar applications and sensing for ultrasonics,
the diaphragm surface area needs to be large to achieve the desired directivity in the
ultrasonic frequency range.
For large a electric signal output, the static capacitance needs to be large and that
leads to the distance between the two electrodes being very small. The surface area
of the diaphragm needs to be large enough to have a large dynamic capacitance, but
not so large as to affect the microphone directivity.
The static electrode (backplate or back electrode) needs to be specially designed.
The distance between the diaphragm and the backplate is small so the viscosity of the
air in this volume will help control the diaphragm resonance. Because of the small
distance between the electrodes, the air between the electrodes will be compressed or
moved sideways along the diaphragm surface. The flow conditions are complex, and
some of the flow needs to be diverted into the holes in the backplate. The diaphragm
will flex and be slightly displaced due to the static bias voltage that attracts it to the
backplate. The sound pressure on the outside of the microphone capsule will also
contribute to the flexing, which will give the diaphragm section curve an almost
parabolical shape [8]. By giving the backplate a slight concave shape, it is possible to
reduce the distortion generated by these effects.
14.7.1 Electroacoustic Analogies

From the viewpoint of circuit analogies, we can represent the moving parts of the
microphone capsule by a mass–compliance–resistance combination. Figure 14.24
shows the basic electroacoustic components of a condenser microphone capsule. The
diaphragm is tensioned so that the first resonance of the transducer’s diaphragm is
usually at or above the audio frequency range. This leads to a compliance-controlled
motional system. The components MAS and R AS due to the holes in the solid backplate
are used to control the diaphragm resonance characteristics. This allows the
frequency response to be adjusted to the desired sound field-sensing characteristics.
The effective mass of the diaphragm is often not included in the electromechanical
circuit, as for example is the case for the circuit shown in Figure 9.15, since the
capsule’s diaphragm is usually compliance-controlled over the full audio range.
Metallized
diaphragm
MAD and CAD
Fixed back CA1

electrode
RAS and MAS Air cavity, CA2
Leak or port to adjust

gradient characteristics
RAP and MAP
FIGURE 14.24 The acoustical components of the condenser microphone.

Microphones 289
ZARF MAD CAD U MAS RAS MAP RAP Z ARB

D
p CA1 p
F
CA2 B
FIGURE 14.25 The acoustic impedance analogy for the condenser microphone capsule
shown in Figure 14.24. Compare to Figure 14.7.
Figure 9.15 shows the circuit that is applicable to a single-sided transducer such as
that of a conventional bias-charged condenser microphone.
For a bias-charged transducer, the transduction coefficient (in the first
approximation) depends on the bias voltage E, the mechanical compliance of
the diaphragm CMD, and the distance between the moving and static electrode x0.
The electric capacitance of the transducer CE0 depends on the same variables. The
mechanical compliance depends on the bias voltage as do the losses due to the
viscosity of air trapped between the moving and static electrode. The compliance CMS
is measured with the electric terminals open, and the capacitance CE0 is measured
without any force applied to the diaphragm.
Figure 14.25 shows the acoustic impedance analogy for the condenser microphone
without the electrical side included. This acoustomechanical circuit applies to all
the three common types of condenser microphones mentioned. When we use the
microphone with a traditional analog preamplifier, we sense the displacement of the
diaphragm. This requires the diaphragm to be taut, with its first resonance at the upper
limit of the audio frequency range, 10–20 kHz. Higher-frequency resonances, which
are a result of the “drum skin”-like behavior of the diaphragm, are usually not of
interest since they usually fall far out of the audio frequency range.
To include the electrical side, the transduction circuit shown in Figure 9.15 is
inserted replacing the components MAD and CAD in Figure 14.25. The resulting circuit
is the one shown in Figure 14.26.
14.7.2 Permanent Charge Condenser Microphones

Permanently charged (electret) transducers use a plastic film that has been treated
so that a polarized charge is present on the film. The polarized film can be used
directly as a diaphragm or attached to the backplate and using conventional metal
film diaphragm. The latter approach allows better frequency response and stability.
The stability of teflon foil-based electret microphones is very high. The acoustome-
chanical circuit is the same as that for biased capacitive transducers.
14.8 ELECTRICAL CHARACTERISTICS AND REQUIREMENTS

A microphone will generally be a part of some sound measurement or recording
system as shown in Figure 14.27, that includes the transducer proper, various
ECAD
Z ARf MAD UD 1: CE0
x0SD
p CAD
F e0 e
CA1
RAS
CA2 MAS
RAP
p
B
Z ARb MAP
FIGURE 14.26 The complete acoustic impedance analogy for the condenser microphone
capsule shown in Figure 14.25 with the electrical part of the circuit included.
Microphones Mi Mn M0 M1 Mi Mn Microphones
A/D inside microphone casing
Analog/digital converters
Analog/digital converters
Amplifiers Gi Gn Gi Gn Amplifiers
Attenuators
Amplifiers Gi Gn Analog/digital
converters
Level indicators VU VU
Mixer or multi-track recorder Host computer or recording device
FIGURE 14.27 Microphones form the initial part of the recording chain: (a) traditional
analog electronics, (b) “digital” microphones with built-in A/D converters, identification,
etc., and (c) analog microphones driving a digital recorder/measurement device.
electronics (switches, amplifiers, analog/digital converter, radio transmitter including

an antenna or a cable/connector), as well as a microphone body or case.
The analog electric signal from the microphone capsule will in most cases need to
be amplified, processed, and/or digitized to be useful. For most analog microphones,
the output signal will be available as a voltage source that has some internal
impedance as shown in Figure 14.28. The voltage may be available in a single-ended
mode or in a balanced mode. The internal impedance is usually made very low to
Microphones 291
+e +et
0 0
(a) (b)
+e +et
0 0
–e –et
(c) (d)
FIGURE 14.28 The output signal of an analog microphone is typically available as a

voltage source in series with an internal impedance (not shown here): (a) single-ended output,
(b) single-ended output with output transformer blocking DC current component, (c) balanced
output, (b) balanced output with output transformer blocking DC current component. Dots on
right hand of transformer in (d) indicate winding polarity.
avoid capacitive shunting of the signal at high frequencies for microphones that are
designed to be used with long cables. The open-circuit voltage levels of microphones
are typically about 10 mV for a sound pressure of 1 Pa.
Condenser microphone capsules often operate using an external bias voltage as
shown in Figure 9.13. These microphone capsules are used with an impedance con-
verter to make it possible to connect the microphone capsule to an amplifier. The
cable capacitance would otherwise reduce the output voltage of the capsule. The
impedance converter must have a very high input impedance so that the electrical
load on the microphone capacitance is small even at low frequencies. The output
impedance and current capability determine the maximum cable length.
Traditionally, microphones have been designed to have a “flat,” i.e., frequency-
independent, response. With the availability of microelectronics microphones
capsules may instead be designed to be used with active electronics and/or digital
signal processing, shaping the frequency response as needed as well as affecting the
signal in other desired ways to improve the microphone characteristics. A combination
of acoustical and electronic equalization could be used to optimize directivity or
noise characteristics.
Some microphone capsules are designed to be used with a remote exterior
amplifier connected by wire, and others may be designed to have built-in amplifiers
and be connected by wire, radio, or by digital transmission. Those that have built-in
amplifiers will work into electrical load impedances that are typically very high.
This is typical of condenser microphone capsules and microphones. Because the
amplifier is close to the microphone, there will be a small (but not negligible) load on
the electric output terminals of the capsule. Capsules that need to work with a remote
amplifier need to be able to work with the electrical load of both the cable and the
20
With RC = 2 . 10–5 (s)
w/o capacitive load
10
Frequency response (dB)
–10
–20
–30
0.02 0.05 0.1 0.2 0.5 1k 2k 5k 10 k 20 k
Frequency (kHz)
FIGURE 14.29 The electrical load by cable capacitance will reduce both the available
output voltage and the frequency response of the output signal.
amplifier. Short coaxial cables are typically characterized by electric capacitance

of about 100 pF/m. The capacitance in cables will limit the high-frequency output
capability and the frequency response of an analog microphone, as shown in
Figure 14.29 unless an isolation amplifier at the microphone, having a low output
impedance and high-drive current capability, is used.
14.8.1 Distortion
Microphones will have an upper limiting sound level due to nonlinearities in the
transducer itself or in one of its auxiliary components, such as a built-in transformer
or a connected active electronic device as well as mechanical limits. It is common
to divide the transducer errors into two groups, linear errors (linear distortion),
usually frequency response deviations, and nonlinear errors (nonlinear distortion),
such as electronic or mechanical overload. Nonlinear distortion adds new frequency
components to the electric output signal.
Obviously, the electrical wave form rendered by a microphone having linear,
but no nonlinear distortion, may be different from that of the acting acoustic field
component for which the microphone was designed, but the microphone will not
generate any new frequency components. Linear distortion is due to frequency
response deviations, i.e., filter effects, which cause magnitude and phase aberrations
of the various frequency components of the acoustic signal.
Nonlinear distortion is generally specified as the percentage of harmonic
components in the electric output signal relative to the desired pure tone signal. The
test frequency is typically 1 kHz. The sound pressure levels used for testing distortion
characteristics are in the range of 100–160 dB, depending on the intended use of the
microphone. Harmonic distortion in the range of 1% is considered an upper limit.
Microphones 293
The distortion will typically be higher in a sensitive, low-noise microphone since

the diaphragm and transduction mechanism have been optimized for low sound
pressures [5].
14.8.2 Microphone Noise

There is both electrical and acoustical intrinsic noise in any microphone output
signal, as shown in Figure 14.30 [9]. Since all microphones are based on electrical
principles, they will also generate electric thermal noise.
The noise is generated by the resistance in the microphone, in the case of an
electrodynamic microphone from the resistance of the voice coil RE. The resistance-
generated noise voltage eTEN is
eTEN = 4kBTRE ∆f (14.20)

Here kB is Boltzmann’s constant, RE is the electric resistance (Ω), T the absolute
temperature (K), and Δf the system bandwidth (Hz). Since many electrodynamic
Ambient Thermal Thermal Shot effect Magnetically Thermal Quantization

acoustic acoustic electric noise, induced electric noise
noise noise and thermal hum noise Jitter
thermal electric and other
mechanical noise, noise
noise and other
mechanisms
Microphone
Air Air mechanical Amplifier
system
Studio Microphone cable Mixing desk A/D
converter
Analog signal Digital signal
Analog microphone connected over cable to a mixing console with subsequent A/D conversion
Ambient Thermal Thermal Shot effect Quantization Digital

acoustic acoustic electric noise, noise format
noise noise and thermal Jitter conversion
thermal electric
mechanical noise,
noise and other
mechanisms
Microphone
Air Air mechanical Amplifier
system
Studio Microphone A/D Algorithmic
converter
Analog signal Digital signal
Digital microphone connected over cable to a digital signal processor
FIGURE 14.30 There are many sources of noise in any microphone system.
microphones have 200 Ω internal resistance, the generated noise voltage will be
about 0.25 μV over the audio range [9].
The acoustic noise, ẽTAN, is generated by the molecules of the air impinging on
the diaphragm. By testing the microphone in vacuum, it is possible to separate
between acoustic and electric noise components in the microphone’s output signal.
The electrical self-noise is dominant in most microphones. The acoustic noise is
given by
eTAN = 4kBTRAR ∆f (14.21)

Since the noise-generating mechanisms are uncorrelated, the voltages will add as
eTN = eTEN
2
+ eTAN
2
(14.22)

Typically, the noise from the microphone is required to be lower than that in
the environment where the microphone is being used so that sufficient signal-
to-noise ratio can be obtained. The microphone’s noise level is often specified
as an equivalent sound level in dBA. Because the microphone and the ambient
noise generally have very different spectra as shown in Figure 14.31, this may be
difficult to attain. The ambient noise in a room typically has a roll-off of about
−5 dB per octave, whereas the microphone electrical self-noise at medium and high
frequencies typically increases by +3 dB per octave. Since room and microphone
noise sound different, electrical noise generated by a microphone may be audible
over the room noise even if the noises have the same A-weighted sound level.
One way of overcoming this problem is to design the microphone so that there
is an acoustic pre-emphasis. This frequency response nonlinearity can then be
suitably removed by active or passive analog electronic circuits or by digital signal
processing.
80
Sound pressure level (dB)
60 Typical electrodynamic
Typical room
microphone
noise spectrum
noise spectrum
40
RC 25
20 MA
F
0
31 63 125 250 500 1k 2k 4k 8k 16 k
Octave band center frequencies (Hz)
FIGURE 14.31 Examples of typical noise spectra for ambient and electronic white noise,
respectively. Ambient noise is typically due to traffic and heating and ventilation equipment.
Microphones 295
Condenser microphones also generate noise due to the amplification circuit

connected close to the microphone capsule. Biased condenser microphones also
suffer from leakage currents between the diaphragm and the backplate that cause a
special type of noise.
It is also important to note that any microphone will be sensitive to vibration
because of the inertia of the diaphragm. This is a larger problem in electrodynamic
microphones than in electrostatic microphones due to the considerable diaphragm
mass of the electrodynamic microphone.
REVIEW QUESTIONS
14.1 Draw the analog acoustic impedance circuit for the following three microphone
types when designed for omnidirectional directivity:
a. Condenser microphone
b. Ribbon microphone
c. Electrodynamic microphone
14.2 Draw the analog acoustic impedance circuit for the ribbon microphone when
designed for bidirectional directivity.
14.3 Draw the analog acoustic impedance circuit for the following two microphone
types when designed for unidirectional directivity:
a. Condenser microphone
b. Electrodynamic microphone
14.4 Design a ribbon microphone for omnidirectional directivity. What must
be changed compared to a ribbon microphone designed for bidirectional
directivity?
14.5 Design a ribbon microphone for cardioid directivity. What must be changed
compared to a ribbon microphone designed for bidirectional directivity?
14.6 Why is a ribbon microphone simplest designed using a mass-controlled
membrane impedance?
14.7 Why are extra resonant circuits necessary to increase the bandwidth of an
electrodynamic microphone?
14.8 Why are condenser microphones usually designed with compliance-controlled
membranes?
14.9 Derive the directivity characteristic of a microphone that has a vented back
chamber.
PROBLEMS
14.1 A 1″ diameter condenser microphone was used for a sound pressure level
measurement. After the measurement, the microphone was calibrated, and it
was noticed that the bias voltage had dropped from 200 to 130 V.
Task:
Estimate the error that the reduced bias voltage could have caused in octave
bands.
14.2 A typical basic condenser microphone capsule is shown in Figure 14.24. The
sensitivity of the microphone is given by Equation 9.43. Such a capsule is
usually used in an electronic circuit of the type shown in Figure 9.13.
Task:
Sketch the pressure frequency response of the microphone using the numerical
data below taking into account the electrical components. The diaphragm
area is SD = 1.56 · 10 −4 m2 and the distance between the diaphragm and back
electrode is d0 = 2 · 10 −5 m.
M AD = 35.4 kg/m 4

C AD = 0.9 ⋅10 −12 m 5 /N

C A1, M AP , RAS , M AS = 0

C A2 = 5.88 ⋅10 −12 m 5 /N

RAP = 9.02 ⋅109 Ns/m 5

CE = 6·10 –11 F

RE ,bias = 265 MΩ

E0 = 200 V

14.3 A typical 1″ diameter condenser microphone has a sensitivity of 50 mV/Pa

when used with a 200 V bias voltage is.
Task:
If a ½″ microphone has a similar construction, the same ratio CAD /CA2, and
also the same mechanical diaphragm compliance (as the 1″ microphone) but
the polarization voltage is increased to 400 V, what sensitivity (in mV/Pa)
would this ½″ microphone have?
14.4 A 1″ diameter condenser microphone is to be used for measurement of low-
frequency sound in the 1–10 Hz infrasound range. The pressure equalization
hole is blocked however. The electric capacitance of the capsule CE is 60 pF
and the amplifier input resistance REL is 700 MΩ in parallel with a capacitance
CEL of 3 pF.
Task:
Determine the corrections necessary for the measured values in this frequency
range.
Microphones 297
14.5 A condenser microphone that has the following data is to be used with a tube
as a probe microphone.
M MD = 1.2 ⋅10 −6 kg

f0 =15 kHz

S D = 2 ⋅10 –4 m 2

Task:
What is the equivalent volume of the microphone membrane?
14.6 A cylindrical electrodynamic microphone is placed in a plane wave sound field
so that the waves impinge perpendicularly on the diaphragm that is on the end
of the cylinder. The microphone has a diameter of 2.5 · 10 −2 m, an open-circuit
sensitivity of −60 dB relative to 1 V/Pa, and a purely resistive 200 Ω internal
impedance. What is the power efficiency of the microphone; that is, how much
of the incident sound power is maximally converted to electric power?
14.7 A certain full frequency range ribbon microphone has an aluminum ribbon
that is 2 · 10 −2 m long, 5 · 10 −3 m wide, and 2 · 10 −5 m thick. The ribbon is
mounted so that the effective front–back distance is 1 · 10 −2 m. The magnet
gap has a flux density of 1 T. Neglect acoustic leakage between the edges of
the ribbon and the magnet poles.
Task:
Determine the maximum RMS output voltage of the microphone when in a
plane wave at a frequency of 1 kHz and a sound pressure level of 94 dB!
14.8 An electrodynamic loudspeaker that is mounted in the wall of a small closed
box is used as a microphone. The loudspeaker driver has the following data:
SD = 3 · 10 −2 m2, RE = 7 Ω, Bl = 5 N/A, MMD = 2 · 10 −2 kg, and the resonance
frequency of the loudspeaker mounted in the box f0 = 45 Hz. The Q value of
the loudspeaker measured with the terminals short-circuited is Q = 0.7.
Task:
Determine the frequency response and sensitivity of the loudspeaker when
used as microphone. Neglect voice coil inductance and assume the loudspeaker
terminals open (infinite electrical load impedance). Only consider frequencies
where the loudspeaker is much smaller than the wavelength.
REFERENCES
1. Bauer, B. B., A century of microphones, Proc. IRE, 50, 719–729 (1962).
2. Rayburn, R. A. and Eargle, J., Eargle’s The Microphone Book: From Mono to Streo
to Surround—A Guide to Microphone Design and Application, 2nd edn., Focal Press,
London, U.K. (2004) ISBN-13: 978–0240519616.
3. Olson, H. F., Acoustical Engineering, Van Nostrand Reinhold, New York (1972).
4. Olson, H. F., The quest for directional microphones at RCA, J. Audio Eng. Soc., 28(11),
776–786 (1980).
5. Ono, K. et al., Development of a super-wide-range microphone, Proceedings of the
120th Audio Engineering Society Convention, Paris, Paper 6637 (2006).
6. Gayford, M. (Ed.), Microphone Engineering Handbook, Focal Press, London,
U.K. (1994) ISBN-13: 978–0750611992.
7. Beranek, L. L., Acoustics, American Institute of Physics, New York (1986) ISBN-
13: 978–0883184943.
8. Wong, G. and Embleton, T., AIP Handbook of Condenser Microphones, AIP Press,
New York (1994) ISBN-13: 978–1563962844.
9. Olson, H. F., Microphone thermal agitation noise, J. Acoust. Soc. Am., 51(2A), 425–432
(1972).
15 Electrodynamic
Loudspeaker Drivers
15.1 INTRODUCTION
A loudspeaker generally consists of a loudspeaker driver (or “motor”) and a loud-
speaker enclosure. The term loudspeaker is popularly used to describe both the
driver and the complete loudspeaker including the enclosure. The ideal loudspeaker
would convert the electric signal applied to its terminals into an acoustic signal at
listening distance while retaining the waveform.
An electromechanical transduction mechanism moves the diaphragm or membrane
that in turn moves the surrounding air. There are many types of transduction
principles that can be used to generate the force. The most common designs use
electrodynamic, piezoelectric, or electrostatic transduction. Loudspeaker drivers
have also been constructed using modulated air pressure, modulated ionization of
air, etc., but such drivers are not used for audio but rather for various industrial
purposes. Piezoelectric drivers are becoming common for telephones and other
mobile equipment because of their high efficiency and low mass.
Most drivers use diaphragms that are designed to work as rigid or flexible pistons,
but drivers using membranes or flexing sheets are also becoming common. This
chapter will discuss drivers that have piston-type diaphragms. Drivers that use
flexing sheets will be discussed in Chapter 21.
Besides the sound radiation by the diaphragm, there is usually unwanted sound
radiated by vibration of the enclosure’s sides. These are excited both mechanically
by the driver and acoustically by the sound field inside the box. The reader is directed
to Chapter 10 and Ref. [1] for a discussion of these topics.
In-depth discussions of the many issues involved in loudspeaker transducer and
enclosure design and use may be found in Refs. [2–7] and in the many contributions
to be found in the pages of the Journal of the Audio Engineering Society, IEEE
Transactions on Audio and Electroacoustics, and others. A comprehensive review
of driver nonlinearities can be found in Ref. [6].
15.2 MOVING-COIL DRIVERS

The electrodynamic moving-coil driver functions as a linear motor that moves air
back and forth to generate sound. The various mechanical elements are shown in the
299
Flexible edge suspension Flexible

wiring
Diaphragm
Flexible center suspension
Dust cap
Basket
Electric terminals
Soft iron structure
Voice coil
Airgaps
Vent hole
“Shorting” ring or cap
Magnet
Soft iron structure
Acoustic resistance
and dust protection
FIGURE 15.1 A cross-sectional sketch of a typical moving-coil low-frequency driver.
sketch of a typical low- or mid-frequency driver in Figure 15.1. For our purposes,
three main system components are of interest:
• The magnet assembly and voice coil, forming the transduction mechanism
• The voice coil, diaphragm (cone or dome), and suspension, forming the
mechanical system
• The diaphragm, dust cap, vent hole, dust filter, surround, and enclosure,
forming the acoustical system
Figure 15.2 shows a high-frequency driver but with a dome-shaped diaphragm (dome
tweeter). The dome may be convex or concave. Concave domes are slightly more
omnidirectional than convex domes (although convex domes look more omnidirec-
tional) [8]. Horns often use drivers that have concave diaphragms together with a
special adapter called a phase plug to improve the coupling of air vibration to the
horn. The principal drivers are discussed in Chapter 19.
Diaphragm Diaphragm
Sound Flexible edge Flexible edge
absorptive suspension suspension
plug
Voice coil
Magnet
Airgaps Airgaps
Soft iron structure
(a) (b)
FIGURE 15.2 A cross-sectional sketch of a moving-coil high-frequency driver. (a) Convex

dome diaphragm with sound absorptive plug inside, (b) concave dome diaphragm.
Electrodynamic Loudspeaker Drivers 301
Low-frequency drivers are often called “woofers,” high-frequency drivers

are usually called “tweeters,” and the intermediate-range drivers are just called
“midrange” drivers.
15.3 MAGNET AIR GAP AND VOICE COIL

The driver’s magnetic fields interact to produce the force that moves the voice coil
relative to the magnet. An AC current through the voice coil creates a dynamic mag-
netic field. The static field is provided by a magnet joined with the chassis of the
driver. Usually, the mass of the magnet and chassis is so large compared to that of
the voice coil that they will not have any noticeable movement.
The voice coil is the center of the loudspeaker driver and is usually cylindrical.
A stiff voice coil former is attached to the diaphragm and carries the voice coil wire
windings. It is important that the voice coil former and leads are attached in such a way
to the diaphragm that they do not excite unwanted bending waves in the diaphragm.
The goal is to have a voice coil that has low mass and low electric resistance. The
voice coil is wound on a former for stability. The former must also contribute to the
cooling of the windings, so metal voice coil formers are preferred for high-power
applications. In high-power, low-frequency drivers, the voice coil is often made of
aluminum or copper and is cooled by incorporating special perforations in the voice
coil former to increase turbulence and cooling by air. A common problem is that the
voice coil still overheats. This makes the voice coil lacquer or glue soften and lose its
grip on the voice coil wire. The loose voice coil windings cause rub and “buzz,” and
ultimately cause the driver fail.
The voice coil winding must have a height suited to the design of the magnetic field
in the air gap. Some examples are shown in Figure 15.3. In principle, it is possible
to compensate somewhat for the nonlinearities in the magnetic field and diaphragm
suspension by special voice coil designs. In many cases the voice coil has a height that is
slightly less than that of the air gap as in Figure 15.3b. Low-frequency drivers sometimes
have long voice coils as shown in Figure 15.3c, since these drivers are less sensitive to
high voice coil mass, and the voice coils need to be able to make large movements to
generate sufficient volume velocity. High-frequency loudspeakers have voice coils that
do not require much movement so the short coil shown in Figure 15.3a can be used that
increases the force to mass ratio of the moving system thus increasing efficiency.
The voice coil resistance is determined by windings of the coil wire, usually made
of specially formed copper, aluminum, or occasionally, silver wire. The wire cross
section is often hexagonal or rectangular so that the electric resistance of the voice
coil can be close wound and the size of the coil. The resistance must be chosen so
that the power losses in the cables leading to the loudspeaker are small compared
(a) (b) (c)
FIGURE 15.3 Examples of (a) short, (b) medium, and (c) long voice coil designs.
30
Voice coil resistance
Impedance magnitude (Ohm)

25 Driver free
Driver in closed box
20
15
10
0
20 50 100 200
Frequency (Hz)
FIGURE 15.4 Typical behavior of the modulus of the electric input impedance of an elec-
trodynamic loudspeaker with the low-frequency driver mounted in the box and with the driver
free away from the box. The stiffness of the enclosed air in the box forces the resonance
frequency upward.
to the losses in the voice coil winding, which encourages the use of high-resistance
circuits. At the same time, a high voice coil resistance is a disadvantage from the
viewpoint of amplifier design, particularly in automotive applications. A DC resis-
tance in the range of 3–7 Ω is typical.
The mounted voice coil will have an electrical impedance that reflects the voice
coil resistance and the inductance of the coil, the magnet, and mechanical systems.
Figure 15.4 shows the typical voice coil impedance modulus as a function of fre-
quency. The inductance will cause a rise of the impedance at high frequencies. At
low frequencies, the movement of the mechanical system will cause a strong back
electromotoric force in the voice coil that results in one or several impedance peaks.
The frequencies and levels of these peaks can be used when adjusting the acoustical
design of the loudspeaker enclosure.
The magnet system contributes to the voice coil inductance and losses. A
“shorting” ring made of copper is sometimes attached to the center pole piece of the
magnet to reduce and linearize the inductance by reducing the eddy currents that are
generated in the pole pieces by the voice coil’s magnetic field.
The static magnetic field in the air gap of the driver is provided by a magnet, or occa-
sionally by a DC current-driven coil, and carried by a soft iron magnetic circuit. Figure
15.5 shows some magnetic circuit designs. The desired magnetic field is determined in
part by the mechanical system, the electrical system, and the loudspeaker box. All of
these factors influence the loudspeaker sensitivity, the resonance frequency, the damp-
ing at resonance, and the suitability of the loudspeaker for a particular loudspeaker box
construction. The magnetic circuit of high-quality drivers must be designed so that the
field in the air gap can be kept within small tolerances in production.
To have sufficient magnetic flux density in the air gap, the gap has to be quite
narrow. Typically, the voice coil is cooled by the magnet structure, and a narrow air
gap is necessary to be able to cool the voice coil properly by allowing the transfer to
Ferrite Metal
magnet magnet
Vent hole Soft iron
Soft iron structure
structure
Dust filter and
acoustic resistance
(a) (b)
FIGURE 15.5 Two different magnet systems for electrodynamic loudspeakers: (a) Ferrite
magnet typical for low-frequency drivers and (b) metal magnet (e.g., neodymium) typical for
high-frequency drivers.
the air gap sides. The heat is then dissipated to the surrounding air by the magnetic
system and loudspeaker chassis. Ferrite magnets provide less cooling than do metal
magnets. If the air gap is too narrow, the unavoidable sideways movement of the
voice coil will lead to contact between the air gap walls and the voice coil resulting in
noise and the eventual failure of the voice coil. This is a common problem in micro-
speakers for mobile phones. A wide air gap will result in a weak magnetic field, low
loudspeaker sensitivity, and poor cooling. By adding a magnetic liquid cooling agent
consisting of a viscous suspension of ferrite particles, one can have a somewhat wider
air gap and still maintain a reasonable flux density and promote better cooling. This
is a common approach used in midrange and high-frequency drivers. The method is
usually unsuitable for low-frequency drivers because of the viscosity of the fluid. A
disadvantage is that the magnetic fluid may have the tendency to dry out.
The magnetic system is not symmetric in the axial direction so the field in the air
gap is not the same on the inside and outside of the system. With a short voice coil,
the nonlinearity is of little importance in high-frequency drivers. In low- and mid-
frequency drivers, the asymmetry is of large importance. Some of the nonlinearity of
the magnetic field in the air gap can be eliminated by using two air gaps and oppos-
ing voice coils driving the same diaphragm; such drivers are called dual voice coil
drives. Dual coil drivers often also have less mass and stray magnetic fields. Figure
15.6 shows the basic design of such a dual voice coil driver.
15.4 DIAPHRAGMS
15.4.1 Shape
The moving assembly consists of the diaphragm, the voice coil, and the outer and
inner, or spider, surrounds that ensure linear movement by the voice coil in the air
gap. The vibration and radiation properties of the diaphragm are determined by the
following factors:
• Diaphragm material: Thickness, density, Young’s modulus

• Diaphragm profile: Flat, straight cone, hyperbolic cone, corrugated circular
sections, etc.
Diaphragm
Voice coil 1
Magnet
Soft iron
Shorting ring
Voice coil 2
Basket and
supporting structure
FIGURE 15.6 The basic design of a “dual drive” driver combining the action of two voice
coils to reduce nonlinear distortion due to magnetic field asymmetry.
• Diaphragm reinforcements and inhomogeneities

• Voice coil: placement mass and attachment
• Surround suspension: mass, stiffness, losses, and geometry
• Air load on both sides of the diaphragm
Some drivers have diaphragms that behave as rigid pistons over most of the desired
frequency range, so its entire surface will move in phase. These drivers tend to
become very directional when the wavelength of sound is comparable to or larger
than the diaphragm radius as shown in Figure 12.15. Since it is difficult to combine
high stiffness and low mass in a flat sheet, most drivers have conical or hemispheri-
cal diaphragms to increase the diaphragm stiffness. The properties of the diaphragm
system influence the sound reproduction properties so much that many commercial
drivers are similar except for the diaphragm.
There are high-quality drivers in which the diaphragm is designed to vibrate as
a stiff but still flexing sheet. By controlling the diaphragm bending modes and their
damping, one can achieve a fairly uniform frequency response and directivity, par-
ticularly if one is only concerned with the response in the reverberant sound field of
a room. Such drivers usually have flat circular or rectangular diaphragms.
Cones using damped flexible diaphragm materials have long been popular in the
construction of high-quality drivers. With such cones, one can achieve nonresonant
behavior, and the effective cone area will shrink as the frequency is increased. Some
common cone shapes and profiles are shown in Figure 15.7. The corrugated cone
shown uses radial compliance decoupling of vibration to adapt the radiating area to
wavelength. This simplifies the use of the loudspeaker in multi-loudspeaker systems,
where different loudspeakers are used for different frequency ranges. To control the
directivity properly, one wants the loudspeakers to have about the same effective
area in the crossover frequency region.
The use of elliptical cones makes it possible to increase the cone area in some appli-
cations where a circular cone would not use the available area optimally. The elliptical
Straight cone
Hyperbolic cone
Corrugated cone
Some possible sections
FIGURE 15.7 Three examples of typical diaphragm profiles.
shape can also be used to give the loudspeaker different directivity characteristics in the
horizontal and vertical planes. The depth of the cone or half sphere causes delay and
considerable interference between the radiation from the inner, center, and outer parts
of the cone resulting in frequency response dips at quite low frequencies. A conical dia-
phragm will show similar but smaller directivity than the piston. The theory presented
in Chapter 12 can be used to approximately calculate the directivity of shallow cones
and other diaphragm shapes. (Note that the surrounds also radiate sound.)
For deep cone shapes, and when the exact motion and directivity pattern is desired,
it is advantageous to use finite element or boundary element modeling to design the
diaphragm and the associated directivity pattern [10]. Finite element modeling can
also be used to model possible buckling and other nonlinearities. If the diaphragm
is weak, large movements may cause it to buckle under the air load of the box. This
leads to considerable nonlinear distortion.
The choice between convex or concave diaphragms is determined to some part by
the maximum allowable thickness of the loudspeaker. Concave diaphragms give less
directional radiation than do convex ones, but convex diaphragms can sometimes be
used to house and hide the magnet in shallow loudspeakers as shown in Figure 15.8.
Concave, dome-shaped diaphragms have long been popular. The choice of domes
over cones is mainly an esthetic matter, although looking at Figure 15.9, one can note
that the directivity of concave domes is smoother than that of convex domes.
Most diaphragms display resonant modal behavior at high frequencies. Domes
have to be appropriately damped in the same way as conical diaphragms. Examples
FIGURE 15.8 Typical car door loudspeaker using an inverted voice coil and magnet system
that reduces depth.
ka = 3 0° ka = 3 0°
ka = 2 30° ka = 2 30°
ka = 1 ka = 1
60° 60°
90° 90°
0 –2.5 dB –6 dB –12 dB 0 0.25 0.5 0.75 1 0 –2.5 dB –6 dB –12 dB 0 0.25 0.5 0.75 1
Convex dome Concave dome
FIGURE 15.9 Calculated directivity patterns, |F(θ)|, for the sound radiation by vibrating
convex and concave dome diaphragms for various values of the wave number k and dome
radius a. Drivers assumed mounted in infinite rigid baffles. For the case shown, the dome
radius and the dome height/depth ratios were constant. (From Suzuki, H. and Tichy, J.,
J. Acoust. Soc. Am., 69(1), 41, 1981.)
30 Hz 1.2 kHz 9.5 kHz

Ideal pistonic motion Rocking mode Axisymmetric mode
FIGURE 15.10 Laser scans of the vibration displacement in a 25 mm diameter convex

soft dome diaphragm showing pistonic motion and mode shapes for two of its bending wave
resonances. (From Bank, G. and Hathaway, G.T., J. Audio Eng. Soc., 29(5), 314, 1981.)
of mode shapes at some resonances of a loudspeaker dome diaphragm are shown

in the laser vibrometer scans in Figure 15.10. Low-frequency woofer motion can be
studied using a stroboscope.
The mechanical compensation of improper loudspeaker diaphragm behavior is
economically very attractive and can take many forms such as:
• Progressive vibration isolation of the cone

• Use of small, concentric extra cones in combination with vibration isolation
of the main cone
• Use of soft, flexible cones with high damping
All of these actions have the goal of reducing the vibrating and radiating area at
medium and high frequencies.
15.4.2 Materials
Most simple loudspeakers use conical diaphragms made of thermoformed plastic
from sheet stock, or from paper pulp that has been dried to conical shape. Such loud-
speaker cones have a poor rigidity resulting in resonance and associated irregular
frequency response.
Over some frequency, the poorly damped bending wave motion of the cone will
lead to resonant behavior due to the reflection of bending waves from the edges
where the diaphragm meets the suspension. This leads to uncontrolled resonances
and modal behavior, called “breakup.” Many diaphragm constructions are also char-
acterized by noncircular modes excited by asymmetrically mounted wire leads, etc.
(see Figure 4.1). The resonant behavior can be minimized by using various damping
compounds on the diaphragm, by using diaphragms made of materials with high
internal damping, or by using constrained damping layers.
Thin lightweight cones cannot be stiff unless they are shaped as fairly deep cones.
The cone shape was originally due to the cones being made from rolled paper. Modern
paper cones are usually made by shaping and drying paper pulp with additives such
as animal hair, glue, softeners, etc. This technology makes it possible to design cones
in various geometries, and one can have curved or flexible sections. The main advan-
tage of using cones made of paper and felt mixtures is that the damping can be made
much higher than that in homogeneous plastic cones. Because of the high Young’s
modulus of paper, one can make sandwich constructions of two sheets of paper with a
constrained viscoelastic layer to provide required damping of resonances.
A somewhat different technique is applied when making diaphragms of woven
materials. Traditionally, resin impregnated cloth was used to cover the center of the
cone so that dirt would not enter the voice coil gap. Dome diaphragms have long
been made of impregnated cloth since it was possible to make soft domes in that way.
Coarse, woven Kevlar cloth, impregnated using soft plastics, has been used success-
fully in diaphragm design. The coarse cloth can be made to use the viscous damping
properties of the plastic in a good way.
An advantage of plastics is that one can form the material by injection molding
or thermoforming processes, which are much faster and simpler than the paper pulp
process. Since plastics are more homogeneous between batches, one can have more
similar properties and higher production yield. Typical plastics are polystyrene or
neoprene mixtures. A disadvantage of these plastics is their low internal damping
that cannot be increased much by added external damping compounds. A better
plastic for loudspeaker diaphragms is talc or mica-filled polypropylene. The addition
of solid compounds helps in keeping the damping high and constant over a wide
temperature range. Conventional plastic-based viscoelastic materials function only
over a quite narrow temperature range and are not suitable, for loudspeakers for
automotive use where the temperature range is typically −40°C to 80°C.
Diaphragms made of sandwich-type materials can also be used for loudspeakers
and are discussed in Chapter 21. A typical sandwich panel is made of two outer
layers of stiff material such as plastic with an intermediary honeycomb structure.
The honeycomb structure has low bending stiffness. Such materials have long been
used in ship and aircraft design to build stiff, lightweight walls and structures.
15.4.3 Supports and Surrounds

The diaphragms of most moving-coil low- and mid-frequency range drivers need to
make large displacements to radiate sufficient low-frequency energy. The maximum
diaphragm displacement of driver designs is typically 5–10 mm. Horn loading and
acoustically resonant circuits can be used to reduce the need for long diaphragm
travel. High-frequency driver diaphragms move much less than those of low- and
mid-frequency range drivers since far-field sound pressure is proportional to dia-
phragm area and acceleration.
For conventional drivers using conical diaphragms, there will be outer and inner
suspensions (“surround” and “spider”), as shown in Figure 15.2. Unless the dia-
phragm is supported by two concentric suspensions, it is normally not possible to
design a low-frequency driver in such a way as to avoid the vibrating voice coil
being displaced off-center in the magnetic field. In the worst case, the voice coil
will contact the walls of the magnet in the air gap. The suspensions also protect the
diaphragm in shipping and handling.
Typically, short-throw designs using only a single surround are found in high-
frequency drivers, mobile phone loudspeakers, and earphone drivers. Such drivers
are characterized by low-mass dome-type diaphragms and are designed to have reso-
nances in the medium- to high-frequency range. This makes it possible to use single-
ring surround suspensions that are stiff enough to keep the voice coil from being
catastrophically displaced in the magnetic field.
The compliance characteristics of the suspensions must be both linear and
nonlinear. In the linear operating range of the suspension, it must have the correct
compliance to give the desired resonance frequency and not show any hysteresis.
The suspensions must also protect the voice coil from making excursions that would
bring it outside the magnet air gap; this is usually the task of the inner suspension. It
is difficult to design loudspeaker suspensions with just the right amount of nonlinear-
ity so that the voice coil movement will always be linear in the main operating range.
The inner suspension is usually made of impregnated cloth so that the air trapped
between the inner suspension and the magnet assembly is free to escape. The inner
suspension must have high radial static stiffness so that the voice coil remains cen-
tered in the air gap. In some experimental designs, the inner suspension has been
made of metal rods, which makes it possible to achieve higher linearity and less
hysteresis. Some disadvantages of this type of approach are that dirt can enter the
air gap and that the rods need to be provided with viscoelastic damping to remove
bending wave resonances in the rods themselves, as well as cost.
An important task of the outer suspension is to provide sufficient damping to the
cone rim so that resonances due to radial and circular bending wave propagation in
the cone are minimized.
In guitar loudspeaker drivers, the outer rim is usually made as corrugations in
the cone. This type of surround has comparatively high stiffness and low losses that
allow severe bending wave reflections at the rim. It is also a common suspension
design in single suspension earphone and treble drivers.
High-quality low-frequency and midrange drivers using cone diaphragms can
have the type of outer suspension shown in Figure 15.11, often called “half-roll”
surround. Half-roll surrounds allow much larger and more linear movement than
corrugated surrounds. Half-roll surrounds are usually made of impregnated cloth,
foam plastic, or rubber. The choice of material is critical for the durability of the
loudspeaker. Surround “rot” is a common cause of driver failure and can be caused
by ozone and exposure to strong light.
Nitrile rubber roll surround Anechoic

termination for
bending waves
Basket
Diaphragm
FIGURE 15.11 An example of a “half-roll surround”-type suspension designed to act as a

vibration absorber.
An advantage of the half-roll surround is that it can be combined with vibra-

tion damping at the outer rim of the cone. A suitable flat section of the surround is
allowed to cover the outer end of the cone. Its thickness can be tapered or it can be
shaped into triangles or other shapes to better absorb the incoming bending waves.
This section, if designed properly, can be made to act as an anechoic termination for
the radial bending waves in the cone. This will help prevent formation of resonances
due to reflection of bending waves at the rim. Finite and boundary element modeling
are superior tools for diaphragm design that allows the study of the combined electri-
cal, thermal, and mechanical behavior and acoustical radiation of the driver [7,11].
15.5 ELECTROACOUSTIC ANALOGIES

In this section, we will derive a simplified electroacoustic circuit representing
the loudspeaker driver (where some of the effects of the enclosure are included)
that allows for intuitive electrical analysis of the circuit. We will use acoustic and
mechanical impedance and mobility analogies.
The electroacoustic analysis of driver systems can also be done using dedicated
electroacoustic software, but for rapid results and an engineering understanding of the
factors involved, electroacoustic analogies are extremely useful and convenient [2,10].
15.5.1 Mechanical System

The mechanical system consists of a mass-spring system, allowing translational
movement only, where the diaphragm and voice coil are the mass, and the spider
and outer surround are the compliance, as shown in Figure 15.2. Because of the way
the compliances move, some of their mass is added to the mass of the system. It is
common practice to assume that half of the mass of these parts can be added to the
mass of the diaphragm and voice coil, resulting in a total moving mass of MMD and
compliance CMS. Many loudspeakers also have a compliant air cushion behind the
voice coil; this will be dealt with when we discuss the acoustic system. The spider
and surround generally have quite high mechanical losses, which we represent by a
mechanical resistance R MS.
The single-degree-of-freedom system is drawn as shown in 15.12a and can be
modeled with the symbols used in electroacoustic mobility analogies, as shown in
Figure 15.12b. The active force in the circuit, F, will come from the transduction
Electrodynamic Vibration F
Diaphragm force F velocity u
mass MMD
u MMD CMS rMS

Compliance Compliance
suspension CMS losses rMS
Basket (ground)
(a) (b)
FIGURE 15.12 A single-degree-of-freedom mechanical system. (a) Mechanical elements,

(b) electroacoustic mobility analogy.
mechanism, and some force will also be necessary to move the air on both sides of
the diaphragm.
15.5.2 Acoustical System
The acoustical system will be studied in more detail later, but we assume for our pur-
poses here that the loudspeaker is mounted in a baffle radiating acoustic power to the
spaces on both sides of the baffle. The radiation impedances seen by the diaphragm
are Z ARf and Z ARb, and these need to be converted to mechanical mobility-type com-
ponents so that they can be added to the circuit shown in Figure 15.12b.
This is done in two steps. First, the acoustical radiation impedance components
are converted to mechanical radiation impedance using
Z MR = Z AR S D2 (15.1)
This relationship can be symbolically drawn using a transformer as shown in

Figure 15.13.
We now convert from acoustical to mechanical mobility using
1
Y MR = (15.2)
Z MR
which results in the circuit shown in Figure 15.14.
F 1:SD p u SD:1 U
u U F p
(a) (b)
FIGURE 15.13 Symbols for mechanical–acoustical piston transformers. Piston area is SD.
(a) Mobility analogy circuits, (b) impedance analogy circuits.
u MMD CMS rMS Y MRf Y MRb
FIGURE 15.14 The mechanical mobility analogy for the loudspeaker driver radiating with-
out a baffle, with the (symmetrical) radiation represented by two identical radiation mobilities
YMRf and YMRb.
This circuit can now be conveniently added on to the circuit shown in Figure
15.12b, resulting in the complete mechanoacoustical circuit shown in Figure 15.16.
15.5.3 Transduction Mechanism
The magnet and voice coil system generates a force on the voice coil when the voice
coil carries an electric current. For a current i, the force F acting on the voice coil
will depend on the magnetic field strength B and the coil wire length l as
F = Bl i (15.3)

This conversion is graphically represented as a transformer, as shown in Figure 15.15.
15.5.4 Complete Circuit

Once we attach the electrical circuit, we obtain a mixed electrical and mechanical
circuit as shown in Figure 15.16, where we have converted the circuit using mechan-
ical elements from Figure 15.14 to a circuit with electrical symbols replacing the
mechanical mobility symbols. Note that we are still considering the mechanical
side as a mechanical mobility analogy even though we use electrical symbols for
the mechanical elements. Because we have both front- and back-side radiation, we
must have two transformers, one for each side of the diaphragm converting the
diaphragm velocity to the appropriate volume velocities in the front- and back-side
acoustic systems.
i Bl:1 F
e u
FIGURE 15.15 The transformer symbol used to symbolize the conversion between electrical
and mechanical sides of an electrodynamic transducer.
i REG REC LEC F MMD CMS rMS

Bl:1 1:SD 1:SD
eG e uD Y ARb Y
ARf
(a)
i REG REC LEC F

Bl:1
e uD MMD CMS rMS Y Y

MRf MRb
(b)
FIGURE 15.16 The electroacoustic analogy for the electrodynamic loudspeaker imple-
mented as a mechanical mobility analogy. (a) Basic analogy showing all three domains. (b)
After conversion of the acoustic mobilities to mechanical mobilities.
Bli F
REG + REC LEC

eG
(Bl)2 (Bl)2 uD MMD CMS rMS Y MRf Y
Bl MRb
FIGURE 15.17 The electroacoustic analogy for the electrodynamic driver implemented as
a mechanical mobility analogy but with the transformer removed.
We note that the electrical side contains the circuit elements REG, REC, and LEC.
The resistance REG represents the amplifier output impedance (which of course is
often complex) and the cable resistance. The voice coil circuit elements are the resis-
tance REC and the inductance LEC. Other circuit elements may be added as needed. By
using the transformer action, we can now redraw the circuit to be an all-mechanical
mobility circuit as shown in Figure 15.17.
One of our aims in using electroacoustic analogies is to find very simplified cir-
cuits that are easily analyzed intuitively. As stated initially, we want to obtain an
acoustical or mechanical mobility circuit for the loudspeaker driver. Intuitively,
series circuits are often the easiest to analyze. Using the “dot method,” we can con-
veniently switch between series and parallel circuits and between mobility—and
impedance—style circuits.
Since we wish to end up with an impedance-style series circuit, it is best to
convert the circuit in Figure 15.17 to an all-parallel circuit. This is done using
Thevenin’s and Norton’s circuit theorems discussed in Appendix A. (Note that the
Norton–Thevenin transition does not conserve the total power developed in the
circuit and thus cannot be used for determining the efficiency.) Using the theorems,
we can switch between sources represented as parallel and series circuits, as shown
in Figure 15.18.
REG + REC
i e Bl
G (Bl)2
uD
LEC REG + REC + j ωLEC
eG REG + REC LEC
(Bl)2 (Bl)2 uD
Bl (Bl)2
(a) (b)
FIGURE 15.18 Conversion of the electrical circuit from (a) a voltage-source representation
to (b) a current-source representation.
Electrical F Mechanical Acoustical
REG + REC
e Bl (Bl)2
G Y
MMD CMS rMS Y MRf MRb
REG + REC + j ωLEC LEC
(Bl)2
FIGURE 15.19 The circuit in Figure 15.17 converted to a mechanical mobility parallel
circuit.
LEC
(Bl)2
MMD CMS RMS Z MRf
uD
e Bl
G Z
(Bl)2 F MRb
REG + REC + j ωLEC
REG + REC
FIGURE 15.20 The mechanical impedance representation of the electrodynamic driver.
Changing the source from a voltage-style source to a current-style source in

Figure 15.18, we obtain the parallel mechanical mobility analogy circuit shown in
Figure 15.19.
The next step is to convert this mechanical mobility analogy into a mechanical
impedance analogy; we now obtain the series circuit shown in Figure 15.20.
Finally, we convert this mechanical impedance analogy, shown in Figure 15.21,
to an acoustical impedance analogy as shown later by simply dividing the imped-
ances of all components by S D2 . The acoustic compliance CAS of the suspension is an
SD2LEC
(Bl)2
MAD CAS RAS Z ARf
UD
e Bl
G
(Bl)2 p Z
(REG + REC + j ωLEC)SD ARb
(REG + REC)SD 2
FIGURE 15.21 The acoustic impedance representation of the electrodynamic loudspeaker

as mounted in an infinite baffle.
important loudspeaker property and is sometimes given indirectly in data sheets as

the corresponding compliance equivalent volume VAS.
15.6 FREQUENCY RESPONSE

The acoustical load on a small sound source, the sound field impedance which is the
radiation load, is primarily reactive and has a mass-type character for distances far
from the origin where kr ≪ 1, as shown by Equation 3.24. For distances such that
kr ≫ 1, the sound field impedance is primarily resistive. Since the radiation load by
the air influences the mechanical system (it adds to the mass and resistance in the
equation of motion), and since the radiated power depends both on the vibrational
velocity of the loudspeaker diaphragm and the real part of the acoustical load on the
diaphragm, the loudspeaker driver must be designed with this in mind.
Most high-quality electrodynamic moving-coil drivers are designed to operate
mainly in the region where the diaphragm radius, a, is much smaller than the wave-
length so that ka ≪ 1, i.e., the radiation load Z AR is primarily reactive and mass like.
The resistive component of the series representation of Z AR increases proportionally
to frequency squared and is written R AR here to emphasize this property. As long as
the diaphragm is small compared to wavelength, the mass load due to radiation will
be constant. The mass component of the radiation load may be as large as the mass
of the loudspeaker diaphragm itself. This puts a limit to how nonrigid the diaphragm
can be.
It was shown in Chapter 12 that in the ka ≪ 1 frequency region, the pressure in
the far field of a vibrating rigid piston is proportional to piston acceleration. This
means that the vibrational velocity must diminish proportionally to the frequency for
the sound pressure at the listening point to be frequency independent.
For constant force—such as is generated by the electrodynamic system—the
vibrational velocity is determined by the vibrating mass, which is the sum of the
diaphragm mass and the mass due to the radiation load. Mass-controlled operat-
ing conditions can only be achieved if the resonance frequency of the mechanical
system is well below the lowest frequency at which the loudspeaker is designed to
operate.
At high frequencies, the sound pressure will drop because the radiation load no
longer compensates for the decrease in vibration velocity with increasing frequency.
The directivity however will start increasing already for frequencies at which the
diameter of the diaphragm is smaller than the wavelength, with an increase in sound
intensity on the symmetry axis of the loudspeaker.
REVIEW QUESTIONS
15.1 In Figure 15.1, what is the purpose of the vent hole in the magnet since there
is already air behind the dust cap?
15.2 Discuss the magnet system design of electrodynamic moving-coil drivers!
15.3 Which factors influence the voice coil impedance of the moving-coil driver?
15.4 What are the advantages and disadvantages of a dual voice coil design?
15.5 Why are cones used instead of flat sheets as loudspeaker driver diaphragms?
15.6 What is the purpose of cone corrugations?
15.7 How can one increase the vibration damping in diaphragms?
15.8 Draw the mechanical impedance and the electric impedance analogies for
an electrodynamic loudspeaker driver, taking the electrical circuit and the
forward and rear radiation impedances into account.
PROBLEMS
15.1 In an experiment to determine the salient low-frequency “lumped parameter”
electromechanical properties of a certain loudspeaker driver, the following
procedure was used:
a. The Bl force factor was determined by placing the driver on a table
so that the cone could move vertically and its position determined.
A 2·10 −2 kg mass was then attached to the cone and the cone was sup-
plied a 3·10 −2. A DC current to bring it back to its original position.
Task:
Determine the Bl force factor! How will gravitation influence the measurement?
b. Next, the resonance frequency of the driver was determined with and
without the 2·10 −2 kg mass. It was found to be 51 Hz with the mass and
67 Hz without the mass attached.
Task:
Determine MMD and CMS! Did gravitation influence the measurement? How
did the radiation impedance influence the measurement if the cone had a
diameter of 0.2 m?
c. Finally, the driver was again suspended vertically so that the diaphragm
could move freely and the voice coil impedance measured using a 10 kΩ
resistor in series. At very low frequencies, the voltage was 3.2 mV AC
when the applied voltage to the system was 10 V AC. The Q-value of the
impedance at resonance was measured as 2.9.
Task:
Determine the mechanical resistance R MS of the driver suspension and the
voice coil resistance R E! How did the radiation impedance influence the mea-
surement of R MS?
15.2 One way of linearizing a moving-coil driver is to use a dual voice coil design
such as that shown in Figure 15.6. A related approach is to use two identical
conventional drivers mounted against one another as shown later and electri-
cally connected in parallel so that they move in phase as shown below. This is
sometimes called an isobaric connection. Assume the front and back radiation
impedances as Z ARf and Z ARb
Bl,SD,MAD,RAS,CAS,REC Z
ARf
Infinite baffle
Bl,SD,MAD,RAS,CAS,REC
Z
ARb
Task:
Draw the joined acoustic impedance analogy of the system.
15.3 The same drivers used in Problem 15.2 are now connected next to one another
as shown below and electrically connected in parallel again so that they move
in phase.
Z Z
ARf ARf
Infinite baffle
Bl,SD,MAD,RAS,CAS,REC Bl,SD,MAD,RAS,CAS,REC
Z Z
ARb ARb
Task:
Draw the joined acoustic impedance analogy of the system.
15.4 The reverberation time of a resonant system is T60 defined in Appendix F.
Consider a freely suspended moving-coil driver that has the following data:
f0 = 40 Hz, QMS = 3.5, REC = 6.5 Ω, Bl = 7.0 N/A, MMD = 4·10 −2 kg, and cone radius
a = 0.2 m.
Task:
Determine the reverberation time of the freely suspended moving-coil
driver’s cone and suspension system with and without its electric terminals
short-circuited.
15.5 The Norton–Thevenin transition does not conserve the total power devel-
oped in circuits and thus cannot be used for determining the efficiency of
the driver.
Task:
Derive an expression for sound power radiated from the front side of a driver
mounted in an infinite rigid baffle and the efficiency of the driver. Assume the
radiation resistance much smaller than the other resistance in the impedance
analogy circuit. The voice coil inductance and generator impedance can be
neglected.
REFERENCES
1. Vér, I. L. and Beranek, L. L. (Eds.), Noise and Vibration Control Engineering: Principles
and Applications, 2nd edn., Wiley, New York (2005) ISBN-13: 978-0471449423.
978-0883184943.
3. Olson, H. F., Acoustical Engineering, D. Van Nostrand, Princeton, NJ (1957) reprinted
by Professional Audio Journals (1991) ASIN: B0006EX7E6.
4. Borwick, J. (Ed.), Loudspeaker and Headphone Handbook, 3rd edn., Focal Press,
Oxford, U.K. (2001) ISBN-13: 978-0240515786.
5. Colloms, M., High Performance Loudspeakers, 6th edn., Wiley-Blackwell, New York
(2005) ISBN-13: 978-0470094303.
6. Klippel, W., Tutorial: Loudspeaker nonlinearities—Causes, parameters, symptoms,
J. Audio Eng. Soc., 54(10), 907–939 (2006).
7. Frankort, F. J. M., Vibration patterns and radiation behavior of loudspeaker cones,
J. Audio Eng. Soc., 26(9), 609–622 (1978).
8. Suzuki, H. and Tichy, J., Sound radiation from convex and concave domes in an infinite
baffle, J. Acoust. Soc. Am., 69(1), 41–49 (1981).
9. Bank, G. and Hathaway, G. T., A three-dimensional interferometric vibrational mode
display, J. Audio Eng. Soc., 29(5), 314–319 (1981).
10. Leach, W. M., Introduction to Electroacoustics and Audio Amplifier Design, 3rd edn.,
Kedall Hunt Publishing, Dubuque, IA (2008) ISBN-13: 978-0757503757.
11. Panzer, J., Radiation impedance of cones at high frequencies, Proceedings of the
112th Audio Engineering Society Convention, Munich, Germany, Preprint 5520 (2002)
Available at http://www.randteam.de/ (sampled January 2011).
16 Baffle and Box
16.1 AERODYNAMIC SHORT CIRCUIT

Any transducer using a diaphragm or membrane to generate sound is essentially
a bidirectional radiator, since the air is excited by both sides of the vibrating
surface. Since the driver dimensions are typically smaller than the wavelength,
sound will diffract around the driver forming an aerodynamic short circuit. The
radiation will be similar to that of a dipole. We saw in Chapter 3 that the dipole
is a very inefficient radiator. A short-circuited diaphragm needs to make very
large excursions, particularly at low frequencies, to generate sufficient sound
pressure for home, studio, and cinema sound reproduction. Because of practical
limitations, the large excursion is unavoidably associated with large nonlinear
distortion.
The driver short circuit can be prevented using a baffle or some enclosure to
separate the sound from the two diaphragm sides. Figure 16.1 shows an overview of
commonly used loudspeaker enclosures. The four most common designs are:
• Closed back
Included in this group are closed-box (acoustic suspension) and closed
transmission line (anechoic termination) designs. There are two types
of closed-box designs: those where the box interior is so small that there
are no interior acoustic box resonances (acoustic suspension) within the
working range of the loudspeaker and those that are so large that there are
resonances.
• Parasite resonator
Included in this group are ported box (bass-reflex), distributed port, and
drone cone designs.
• Acoustical filter
This type of enclosure is not shown in the figure but is a closed-box design
using a passive acoustical filter controlling the front sound radiation. These
designs are often called bandpass designs.
• Acoustic transformer
Included in this group are half- and quarter-wave resonators, open
transmission line, and various horn designs.
This chapter will discuss baffle and closed back designs. Parasite resonator and
acoustical filter designs will be discussed in Chapter 17. Acoustic transformer
and horn designs are treated in Chapters 18 and 19. Gradient loudspeakers are the
319
Bandpass Closed box Closed box Bass reflex Distributed

box with air leak box port
Acoustic suspension Acoustic Single small Air mass in bass Many small
Small box, compliance and mass designs
and low-pass filter suspension hole reflex opening holes
Closed box Drone cone Drone cone Acoustic vent Acoustic

resistance
Spurious Heavy diaphragm Heavy Porous sound Porous sound

movement small movements diaphragm absorber absorber or cloth
by box sides
Closed box Infinite baffle Baffle Open back
Acoustic
suspension
Half-wave Bass reflex Quarter-wave

resonator box resonator
Traveling wave or semi-resonant designs
Inverted parabolic horn Voigt horn Parabolic horn
Transmission line, open end Exponential and

other flare horns
Sound absorptive material

Closed box
Transmission line, closed end
Sound absorptive material

inside Acoustic
suspension
FIGURE 16.1 Some designs used in the prevention of aerodynamic sound cancellation
in electrodynamic loudspeaker drivers. (After Tyrland, S., Construction of a monitor loud-
speaker, Thesis (in Swedish), Report 74-35, Department of Applied Acoustics, Chalmers
University of Technology, Göteborg, Sweden, 1974.)
topic of Chapter 20; these are a mix of open and closed back designs. Loudspeaker
systems that use designs where the driver is radiating directly such as baffle,
closed, and ported box designs, are sometimes called direct radiator systems in
contrast to, horn systems where the driver is radiating through an acoustical filter
of some sort.
Baffle and Box 321
Most loudspeakers will feature one or more loudspeaker driver units that are
mounted in the wall of a box. The listening room itself may often be regarded as
a box as well. The loudspeaker will drive the sound field in both the “outside” and
“inside” boxes. Both sound fields are likely to feature resonances that will affect the
driver at low frequencies. Since the loudspeaker box and the room will generally be
of very different size, their acoustical properties may be discussed separately.
16.2 INFINITE BAFFLES

If the loudspeaker is mounted in a wall separating two large rooms, the wall can be
considered as an approximation to an infinite baffle. Sometimes a closet or similar
can be used as an approximation to a second large room. When the back room is
suitably damped then—for frequencies well above the lowest resonance frequencies
of the back room—the loudspeaker effectively sees the back room as an infinite
volume. If the back room is large and suitably damped, the rear radiated sound
reflected back to the loudspeaker will be so weak that it will not be heard through
the loudspeaker. It must also be remembered that the loudspeaker has poor sound
isolation. From the viewpoint of sound isolation, it is effectively a hole in the wall.
16.2.1 Far-Field Sound Pressure

In Chapter 12, it was shown that the far-field sound pressure of a vibrating piston in
an infinite rigid baffle is proportional to piston acceleration and surface area as given
by Equation 12.13. A directivity term is needed to be taken into account the effect of
sound power concentration as the piston dimensions become close to or larger than
the wavelength. We noted that the sound pressure at far distance from a monopole on
the baffle was twice that of the monopole in free space. This is due to the increased
radiation impedance because of the presence of a mirror image of the source behind
the baffle and the smaller solid angle of radiation.
A simplified approach for obtaining the same result for low frequencies is based
on of the power dissipated in ℜMRf, the real part the front radiation impedance
Z MRf [2]. The ℜ symbol is used to show that the resistance is frequency dependent.
The acoustical analogy of the loudspeaker in a baffle is shown in Figure 16.2.
We know that the power P dissipated in a resistance R M by a current u~ in a
mechanical impedance analogy can be written as follows:
P = RM u2 (16.1)
In contrast to our previous approach, we do not include the angle dependence

of p. Instead, we assume that the loudspeaker is radiating its power equally in all
directions into one half-space, which is a good assumption for ka ≪ 1. Thus, the
sound intensity I at a distance r is
P
I= (16.2)
2πr 2
LEC
(Bl)2
MMD CMS RMS MMRf
uD MRf
(Bl)2 MMRb
eG Bl REG +REC
F
REG + REC +jωLEC
MRb
FIGURE 16.2 The electroacoustic analogy for the electrodynamic loudspeaker when
mounted in an infinite baffle. MMRf, ℜMRf, MMRb, ℜMRb are the radiation impedance compo-
nents for the two sides of the baffle.
Since the particle velocity and the sound pressure are in phase in the far field, as in
a spherical wave, we have
p2
I= (16.3)
ρc
Inserting the value for u calculated from the electroacoustic analog circuit, we obtain
the pressure at some distance as
1 1 ωρ
p(r , ω ) = e Bl S piston ⋅ ⋅ (16.4)
r REG + REC ℜ2 + ωM 2

MRf ( MRf )
To add the directivity, one can now multiply the rms pressure by the directivity term
derived in Chapter 12 written as a function of ω
 ωa 
2J1  sin (θ )
 c  (16.5)
F (θ, ω ) =
ωa
sin (θ )
c

16.3 FINITE BAFFLES

The main purpose of the baffle is to prevent the aerodynamic short circuit, i.e., to
increase the path that the wave radiated from the back side of the piston has to travel
before it interferes with the direct sound from the front surface of the piston. For a
finite baffle to be effective, it must have dimensions about as large as the wavelength
at the lowest frequency one wishes to reproduce. An example of the frequency
response of a driver in square baffle having sides 1.2 m is shown in Figure 16.3 for
two driver positions on the baffle.
Baffle and Box 323

0
–5
–10
–15
–20
20 50 100 200 500 1 k 2 k 5k 10 k 20 k
Frequency [Hz]
FIGURE 16.3 Two examples of far-field on-axis frequency response curves for a loudspeaker
mounted on center (solid line) and asymmetrically (dashed line) in a square baffle having
1.2 m long sides. (After Olson, H.F., Acoustical Engineering, D. Van Nostrand, Princeton, NJ,
1957, re-printed by Professional Audio Journals, 1991, ASIN: B0006EX7E6.)
The diffracted wave can be attenuated to some extent by providing a sound-

absorptive layer on the rear side of the baffle. A sound-absorptive layer on the front
side of the baffle would somewhat reduce the sound pressure at distance because of
less pressure doubling at the surface due to the acoustically softer surface.
The baffle must have sufficient mass and rigidity so that it is not excited by the
loudspeaker. This requires baffles that are quite heavy and braced. Baffle resonance
will lead to reduced sound insulation between the front and rear side of the baffle.
Baffle sound radiation can interfere with the direct sound from the loudspeaker.
An advantage of the baffle approach over a box enclosure is the absence of
internal acoustical resonance. This is a particularly useful advantage when using
loudspeakers that have low-mass-membrane-type diaphragms and thus very poor
sound insulation. Electrostatic loudspeaker panels use such low-mass flexible
membranes and are often mounted in baffles for this reason.
The positioning in rooms of loudspeakers mounted in baffles requires particular
attention. In contrast to conventional loudspeaker boxes, which usually obtain the
strongest low-frequency coupling to the sound field of the room when mounted in
corners, baffle loudspeakers are best mounted away from a wall. Because the baffle
loudspeaker is effectively two coherent sources out of phase, the alignment of these
sources relative to the reflecting surface is important as indicated by the power
radiation curves for dipoles shown in Figure 16.4. For better low-frequency sound
power generation, the finite baffle loudspeakers should be positioned so that their
dipole action is parallel to the close reflecting wall.
16.4 CLOSED-BOX ENCLOSURES

16.4.1 Electroacoustic Analogies
The closed-box design is the most basic of all loudspeaker enclosures and usually
the starting point for the design of more complex enclosures. The sound field in a
x +
2
–
P/P0 1.5
x
0.5 + –
0 0.25 0.5 0.75

x/λ
FIGURE 16.4 The power radiated by a dipole close to a rigid plane depends on its direction
relative to the plane. This graph shows the power P radiated by a dipole for two different
directional alignments of the dipole, relative to the free field power radiated P0. The center of
dipole is at a distance x from the reflecting plane. (After Waterhouse, R.V., J. Acoust. Soc. Am.,
30(1), 4, 1958.)
loudspeaker box will feature modes the same way as the listening room. The mode
theory of rooms is described in Appendix F. Since most loudspeaker boxes have much
smaller volumes than rooms in which they are placed, their resonant modes will be
higher in frequency. Note that the sound field will be resonant for any shape of room or
box. Loudspeaker box designs such as closed box, ported box, and other vented designs
are usually intended to operate well below the frequency of first mode of the box.
As is discussed in Appendix F and shown in Chapter 7, the characteristic acoustic
impedance component of a rigid closed box at very low frequencies is its acoustic
compliance, CAB. More exact models may be found in Ref. [5] that shows comparisons
to finite-element models.
Because the sound absorption of the enclosure walls is due to porosity, vibrational
losses, air pumping at joints, leaks, etc., as well as the inclusion of sound-absorptive
materials in the box, there will be one or more resistance terms as well. The values
of these resistance components are, understandably, difficult to determine early in
the design process.
The mechanical impedance analogy for a baffled loudspeaker was shown in
Figure 16.2 that included the impedance components due to front- and back-side
radiation. After replacing the impedance components for the back-side radiation in
the baffle by the impedance components for the closed box, we obtain the box shown
in Figure 16.5.
The corresponding electroacoustic analogy is shown in Figure 16.6. As shown in
this series circuit, the two radiation components will be series coupled. In the analog
circuit, the acoustical properties of the box are represented by two components,
Baffle and Box 325
Bl
SD
MAD
Bl = Force factor
RAS
CAS SD = Diaphragm area
REC MAD = Diaphragm acoustic mass
ZAR
RAS = Suspension acoustic resistance
REC = Voice coil electric resistance
CAS = Suspension acoustic compliance
CAB CAB = Box volume acoustic compliance
ZAR = Radiation impedance = AR + jωMAR
FIGURE 16.5 The components of a closed-box loudspeaker.
(Bl)2
(REG +REC)SD2 MAD CAS RAS MARd ARd
UD
MAW
RAL
CAW
eGBl RAB RAW
(REG + REC)SD MARL
CAB
MARW
ARL
ARW
FIGURE 16.6 The acoustical impedance analogy for a closed-box loudspeaker showing the
impedance components due to diaphragm radiation and box load. Dotted components due to
leakage and box wall resonances are usually not considered in basic design.
CAB due to the compliance and R AB due to the losses inside the box. The losses due
to leaks in the box walls and in the driver are represented by R AL , which we will
usually not consider. The presence of important box wall vibration may require
the inclusion of additional lossy resonant circuits in the analogy represented by the
dotted components in Figure 16.6. These resonant circuit and leakage elements may
also have radiation impedance components. Unless the loudspeaker enclosure is
very large, these are usually negligible. The voice coil inductance LEC has also been
eliminated from the circuit.
16.4.2 Transfer Function
The transfer function of the loudspeaker is determined by product of the radiation
transfer function Hup and the electroacoustic transfer function Heu studied earlier.
The frequency response is 20 log(|H(ω)|).
Since closed-box loudspeakers are often quite small compared to the wavelength
of interest in their low-frequency operating range, they will behave as monopoles
at very low frequencies. For this case, the transfer function of the closed-box
loudspeaker Hep will be approximately given by
p(r, ω) ρBl e − jkr

H (r , ω ) = = H f (ω ) (16.6)
e (ω ) RE 4πr
Here Hf (ω) is the dimensionless filter function determining the frequency response
of the loudspeaker. With suitable damping provided by the resistance components
in the circuit, it will provide “flat” frequency response as discussed in Appendix B.
Since the acoustic compliance must always have one of its terminals at circuit
ground, the box loss resistance R AB is positioned “on top” of CAB. The radiation
impedance Z AR consists of a resistance ℜARd and a mass MARd and will also be
“floating” above ground. These components will be very small for most loudspeaker
drivers. Usually, the total resistance ΣR A is given by the sum
B 2l 2

∑ RA =
RE S D2
+ RAS + ℜ ARd + RAB (16.7)
In this equation, the dominant terms are usually due to the magnetic damping (note
that REG was set to zero) and losses in the loudspeaker driver’s suspension R AS.
The electroacoustic transfer function is determined mainly by the mass terms in
the frequency region somewhat above the “box” resonance frequency. For constant
~
input voltage to the voice coil, the volume velocity U C in the circuit in Figure 16.6
will vary as 1/ω. The real part of the radiation impedance however is proportional to
ω2 for frequencies where ka < 1, so the result is a relatively flat frequency response
over the resonance frequency up to this limit.
16.4.3 Resonance Frequency
The “free” driver resonance frequency will not be the same as that of the driver in
an infinite baffle since the baffle causes a larger mass load than that of the dipole
condition of the free driver.
The circuit reactance determines the resonance frequency of the loudspeaker.
The sum of the reactance terms is
1 1

∑X A = ωM AD −
ωC AS
+ ωM ARd −
ωC AB
(16.8)
Note that since the compliances are in series, they add as capacitance. The resonance
frequency ω0 of the loudspeaker in free field
1
ω0 = (16.9)
M ADC AS

Baffle and Box 327
will be changed to
1 (16.10)
ω 0,boxed =
( M AD + M ARd ) CC AS+CCAB
AS AB

in a baffle.
Here MARd is the “equivalent mass” of the radiation impedance load on the dia-
phragm. This mass load is around a few grams for a piston having a radius of 10 cm.
This is sufficient to alter the resonance frequency since the typical mass for a loud-
speaker driver diaphragm of such size is about 20 g. The result is that the resonance
frequency of the driver will be increased when mounted in the box. As we have
noted before, a large diaphragm mass leads to poor loudspeaker efficiency. Acoustic
suspension-type enclosures discussed later require the diaphragm to be very stiff so
that it, as it vibrates, will not buckle or break up under the acoustical load of the air
in the box and the applied force.
At frequencies somewhat below the resonance frequency, the frequency response
increase will be proportional to ω2, i.e., the frequency response will increase by
+12 dB per octave.
The resonance frequency of the loudspeaker mounted in the box can be
decreased by
• Using a loudspeaker having higher compliance

• Using a loudspeaker having higher diaphragm mass
• Using larger enclosure volume
• Filling the enclosure with some porous material that lowers the speed of
sound inside the enclosure (see Appendix E)
• Using an amplifier circuit that adds virtual mass to the driver (see Chapter 23)
Finally, it should be mentioned that the compliance of the loudspeaker can be

made negative (and stabilized by a servo system). In this way, it is possible to lower
the resonance frequency considerably. In this way, the negative compliance can
compensate for the compliance of the air in the loudspeaker enclosure, particularly
for very small air volumes [6].
16.4.4 Q Factor and Frequency Response

The Q factor of the resonant system in Figure 16.6 is given by the ratio
ω 0,boxed ( M AD + M ARd )
QB = (16.11)
RAT
Because the resonance frequency of the driver is raised due to the presence of the
box compliance, the system’s Q factor QB will be higher than that of the “free”
10
0
–3
–10
–20
Q=2
Q = 0.707
–30 Q = 0.25
0.1 0.2 0.5 1 2 5 10

ω/ω0
FIGURE 16.7 Second-order Butterworth high-pass filter frequency response for three
values of QB. The resonance frequency is ω0. At the resonance frequency, the relative level is
20 log(QB). Note that the −3 dB cutoff frequency depends on the damping.
driver QD. The frequency response function of the small closed-box loudspeaker will
have a second-order high-pass filter character given by
ω2
H (ω ) = (16.12)
ω0
ω 2 − ω 20 − j ω
QB
Figure 16.7 shows the response for three values of QB. The solid line shows the
Butterworth maximally flat characteristic. With a QB -value of 1/√2, the circuit
will have a second-order “maximally flat” Butterworth filter frequency response
as discussed in Appendix B. An overdamped response will have insufficient low-
frequency response whereas the underdamped response may sound boomy. When
QB = 1, there will be a slight overshot in the frequency response in the frequency
region immediately above ω = ω0. At the resonance frequency, the relative level is
20 log(QB). The properties of the listening room should be taken into account when
deciding on the amount of damping needed.
16.4.5 Front Radiation and Baffle Effect

The previous analysis assumed that the loudspeaker and its driver are very small
compared to wavelength. The driver directivity starts to become considerable when
values of ka ≈ 1 and larger as shown in Figure 12.13. If the baffle effect coincides
with the driver directivity, this may result in a very large frequency response effect.
A practical example of the occurrence of this can be seen in Figure 22.1.
Baffle and Box 329
Real part of
Main Baffle effect radiation
+ resonance fully developed impedance
Far field sound level relative to
constant
driver in infinite baffle
0 Infinite baffle
–6 dB Closed box
Directivity increase
Start of as diaphragm
baffle effect dimensions become
large compared to
wavelength
Frequency
FIGURE 16.8 Typical far-field frequency response of the driver showing the baffle effect
for two mounting conditions. At high frequencies, there will be a frequency response drop
due to voice coil inductance as well.
In practice, the loudspeaker box will usually act as a baffle for the driver so
that the frequency response will show the baffle effect roughly outlined in Figure
16.8. If the box front is made very large, the response will be similar to that of
the driver mounted in an infinite baffle; that is, the level will be 6 dB higher
than that for the small, free loudspeaker box. The frequency response rise, also
called the baffle effect, can be compensated by a passive or active shelving filter
as discussed in Appendix B. The baffle effect can also be considered a result of
diffraction at the edges of the loudspeaker box, mainly from the surface in which
the driver is placed.
The baffle effect can be considerable when excessive symmetry and sharp
corners are present. This is shown in Figure 16.9c. A common method to avoid the
diffraction by box fronts is to provide the front with a sound-absorptive layer. Since
the sound-absorptive layer cancels the wave in the region close above it, the wave
field along the layer will be damped before it reaches the edge thus diminishing the
edge diffraction. Alternatively, the box front edges may be rounded or truncated as
shown in Figure 16.9 j or 1. We also note that all boxes lead to a baffle shelving effect
in the frequency response of the loudspeaker, but that the response of a loudspeaker
set in a sphere has little oscillation due to diffraction.
16.5 PRACTICAL CLOSED-BOX LOUDSPEAKERS

The ideal loudspeaker box has rigid walls and is an immovable support for the driver
and provides a known, well-defined acoustic impedance for the driver at low fre-
quencies. In addition, its interior does not reflect any high-frequency sound and it
exterior does not cause any edge diffraction.
Real loudspeaker boxes can only approximate these desirable properties. The
loudspeaker box has limited mechanical impedance; it can be set in vibration not
only by the sound inside the box but also by the driver itself. At some frequencies,
the walls of the loudspeaker may be almost transparent to sound.
15 15
10 10
Response in dB
Response in dB
5 5
0 0
–5 –5
–10 –10
100 200 300 400 600 8001000 2000 3000 4000 100 200 300 400 600 800 1000 2000 3000 4000
(a) Frequency in Hz (e) Frequency in Hz
15 15
10 10
Response in dB
Response in dB
5 5
0 0
–5 –5
–10 –10
100 200 300 400 600 800 1000 2000 3000 4000 100 200 300 400 600 800 1000 2000 3000 4000
(f)
(b) Frequency in Hz Frequency in Hz
15 15
10 10
Response in dB
Response in dB
5 5
0 0
–5 –5
–10 –10
100 200 300 400 600 8001000 2000 3000 4000 100 200 300 400 600 8001000 2000 3000 4000
(c) Frequency in Hz (g) Frequency in Hz
15 15
10 10
Response in dB
Response in dB
5 5
0 0
–5 –5
–10 –10
100 200 300 400 600 800 1000 2000 3000 4000 100 200 300 400 600 800 1000 2000 3000 4000
(d) Frequency in Hz (h) Frequency in Hz
FIGURE 16.9 On-axis frequency response curves for loudspeaker boxes of different shapes
due to diffraction. The small omnidirectional driver used for these measurements had a
diameter of 22 mm. Sizes: (a) Sphere 2 ft diameter; (b) half sphere 2 ft diameter; (c) and (d)
2 ft diameter, 2 ft length; (e) 2 ft sides; (f) base 2 ft diameter, 1 ft height; (g) 2 ft diameter,
2 ft height; (h) square 2 ft base, 1 ft height. (From Olson, H.F., J. Audio Eng. Soc., 17(1), 22,
January 1969.)
Baffle and Box 331
15 25
10 20
Response in dB
Response in dB
5 15
0 10
–5 –5
–10 –10
100 200 300 400 600 8001000 2000 30004000 100 200 300 400 600 8001000 2000 30004000
(i) Frequency in Hz
(k) Frequency in Hz
15 15
10 10
Response in dB
5 Response in dB 5
0 0
–5 –5
–10 –10
100 200 300 400 600 8001000 2000 30004000 100 200 300 400 600 8001000 2000 30004000
(j) Frequency in Hz (l) Frequency in Hz
FIGURE 16.9 (continued) On-axis frequency response curves for loudspeaker boxes
of different shapes due to diffraction. The small omnidirectional driver used for these
measurements had a diameter of 22 mm. Sizes: (i) square 2 ft base, 2 ft height; ( j) truncated
pyramid 6″ height, edges of truncated sides 1ft, length of edges of parallelepiped 1 and 2 ft;
(k) lengths of edges 2 and 3 ft; (l) lengths of edges 1, 2, and 3 ft, lengths of truncated surface
1 and 2.5 ft. (From Olson, H.F., J. Audio Eng. Soc., 17(1), 22, January 1969.)
There are several approaches to closed-box design. The choice of approach depends
on the type of loudspeaker used, box volume, and the desired frequency response in
the region of the low-frequency cutoff. Since the mass moves against the compliance of
both the air trapped in the box and the loudspeaker suspension, the resonance frequency
of the driver mounted in a box is going to be higher than its free resonance frequency.
The original approach was to let the “mounted” resonance frequency only be
minimally higher than the free resonance frequency. This is a reasonable choice
for a loudspeaker that has a relatively stiff suspension. The typical applications are
loudspeakers mounted in quasi-baffle-type enclosures, i.e., very large enclosures.
This approach requires loudspeaker drivers having strong magnets or mechanical
resistance high enough to give reasonable damping. A different approach is to
make the box so small that its compliance is much less than that of the loudspeaker.
The combined compliance will then be about that of the closed-box volume. This
approach is called “acoustic suspension” design.
16.5.1 Acoustic Suspension
The acoustic suspension design results in a considerable increase in the mounted
resonance frequency compared to the free resonance frequency of the driver.
However, the addition of the compliance of the trapped air often results in less
nonlinear distortion. The reason is that the enclosed air spring may be more linear
than the mechanical diaphragm suspension. Acoustic suspension designs need
loudspeakers having rigid cones and large-diameter voice coils. Sometimes a small
capillary air leak is necessary for the diaphragm to remain statically in its neutral
position when atmospheric pressure changes if the box and driver are very airtight.
The closed-box design always results in an increased resonance frequency, unless
the box is filled by a suitable sound-absorptive material as discussed previously.
The resonance frequency increase in turn leads to an increase in the QT value of
the mechanical system. The result is a more resonant character of the high-pass
properties of the transfer function of the loudspeaker system.
16.5.2 Internal Resonance and Modes

Large loudspeaker boxes have interiors that for mid and high frequencies need to
be considered as small chambers or rooms. Wave motion needs to be considered for
volumes that have dimensions larger than ½ wavelength. Exact theoretical treatment
of these volume interiors is difficult except for a few special cases such as an idealized
rectangular room with losses that is discussed in Appendix F. Numerical techniques
such as finite-element modeling should be applied for practical cases.
High-frequency geometrical acoustics is usually of little interest in the design of
loudspeaker box since high-frequency sound is usually sufficiently absorbed by the
wadding that most boxes are filled with or that is applied to the box side interiors
to increase modal damping. Sometimes one finds that it is useful to use a special
sound-absorptive wedge behind the back of the driver if the box back wall reflection
of sound is a problem.
The influence of sound absorption in a room on eigenfrequencies and
eigenfunctions is discussed in Ref. [8] and in Appendix F. In the simplest approach,
a thin sound-absorptive lining is applied to the walls inside the box to remove the
effect of the “room resonances” of the box volume. However, many loudspeaker
boxes are fully filled with sound-absorptive material. The presence of a suitable
sound-absorptive material will usually increase the compliance considerably. This
is due to the nonadiabatic behavior of sound as it moves through a medium that
has heat exchange with a secondary structure, in this case the fibers of the sound-
absorptive material. The heat exchange can lower the speed of sound to about 80%
of its value in free air.
A consequence of the reduced propagation speed is an apparent increase in box
volume; if the volume is entirely filled by the material, the increase in apparent
volume can be up to about 50%. In addition, the porous filling can efficiently damp
the modes. Occasionally the porous filling can move due to the viscous drag by the
particle velocity. This may reduce the resonance frequencies further.
Sometimes manufacturers publish the driver compliance as an equivalent
acoustical volume. This is a useful information—if the driver is used in a closed box
having that volume—the resulting driver resonance frequency will be about 1.4 times
that of the unmounted driver. This volume is typically the minimum recommended
for a closed-box loudspeaker design. However the QT value of the circuit must be
adjusted for the new resonance frequency.
Baffle and Box 333
16.6 POWER AND EFFICIENCY

We sometimes need to calculate the power losses in circuits using electroacoustic
analogies. One reason for this could be to estimate the loudspeaker conversion
efficiency from electric to acoustic power. It is important to remember that it is
always best to calculate the power lost using the nonintegrated circuit as shown in
Figure 16.10. Table 16.1 shows the expressions used in determining the power.
Since most of the power supplied to the loudspeaker will be lost as heat in the
voice coil, the Bl force factor, the voice coil resistance REC, and the diaphragm mass
of the loudspeaker MMD are important factors in determining the amount of power
radiated by the loudspeaker.
The power lost in the voice coil resistance is simple to calculate for the region
that is above the fundamental resonance frequency of the loudspeaker but below the
range where the voice coil inductance starts to increase the impedance. It is
e2
Pheat ≈ (16.13)
RE
if the Bl force factor is so small that the mechanical–acoustical system has little
influence on the impedance on the electrical side.
Sometimes one can assume that the power lost in the mechanical circuit
admittance rMD is negligible. At low frequencies above the resonance frequency of
RE LEC MMD CMS rMS MAR

i Bl : 1 F 1 : SD p
eG e uC uC U rAR
FIGURE 16.10 In power loss calculations, it is best to use this type of nonintegrated circuit
representation, calculating the losses in each domain separately. (Note the mobility type
analogy on the acoustical side.)
TABLE 16.1
Power Loss Calculations in the
Nonelectrical Domains
Domain Mechanical Acoustical
Mobility rM F 2 rA p 2
Impedance RM u2 RAU 2

the driver, i.e., in the mass-controlled region, the power radiated by the loudspeaker
to one side of the baffle (at ka ≪ 1) is approximately
2
 eBl
1 
Pacoustic ≈ ℜ MR (16.14)
 REC M MD + 2 Z MR 

Here we have assumed that the electrical circuit elements consist of only REC. The
expression for the real part of the mechanical radiation impedance ℜMR for a circular
piston in a baffle for ka ≪ 1 is
ρS D2 ω 2
ℜ MR (a, ω) ≈ (16.15)
4πc
Writing the efficiency as
Pacoustic
η≈ (16.16a)
Pheat
we obtain the efficiency as
η≈
( Bl )2 SD2 ρ (16.16b)
REC M MD
2
2πc
Since the diaphragm mass MMD will be roughly proportional to its thickness h and
surface area SD, the efficiency will be proportional only to h, Bl, and REC as follows:
η∼
( Bl )2 (16.16c)
REC h2

The diaphragm thickness is, in practice, about the same for many types of loudspeak-
ers, due to the requirements of manufacturing processes, durability, and transport-
ability. As a result, Bl and REC are the dominant parameters. A low REC is difficult
to combine with a high Bl. In practice, most electrodynamic loudspeakers have an
efficiency in the range of 1%–5%.
πa 4ρ
Pout =
4c
(ωu)2 (16.17)

REVIEW QUESTIONS
16.1 Derive the analog electrical circuit for an electrodynamic driver mounted in an
infinite baffle and in a closed box assuming equal radiation from both sides.
16.2 Derive an expression for the radiated power as a function of input voltage for
an electrodynamic loudspeaker mounted in an infinite rigid baffle.
16.3 Derive an expression for the far-field sound pressure as a function of input
voltage for an electrodynamic loudspeaker mounted in an infinite rigid baffle.
Baffle and Box 335
16.4 Why is the frequency response of an electrodynamic driver “flat” over part of
its frequency range?
16.5 Derive an expression for the efficiency of an electrodynamic driver in an
infinite baffle in the working frequency range and an expression for the Q
factor at the resonance frequency.
PROBLEMS
16.1 A loudspeaker manufacturer gives the following data for a 10″ moving-coil driver:
f0 = 67 Hz, QD = 0.38, CMS = 1.8 · 10−4 m/N, VAS = 1.5 · 10−2 m3, MMD = 31 · 10−3 kg,
and effective radiating area SD = 3.5·10−2 m2. During the design of a suitable closed
box for the driver, it becomes clear that there is an error in the data sheet.
Tasks:
a. Assuming that the resonance frequency and radiating area are correct, find
the error and calculate the correct value in the driver specification.
b. For this driver, what is the box volume that will give QB = 0.707?
16.2 The resonance frequency f0 of a moving-coil driver was measured for three
different conditions: (1) freely suspended ( f0 = 120 Hz), (2) freely suspended
but with an extra mass MM,add of 10 −2 kg attached to the cone ( f0 = 100 Hz), and
(3) mounted in a closed box having a volume of 1 × 10 −3 m3 ( f0 = 150 Hz).
Tasks:
a. Calculate the mass of the cone MMD and the mechanical compliance of the
suspension CMS.
b. Calculate the effective radiating area SD.
16.3 A loudspeaker manufacturer gives the following data for a 10″ moving-coil driver:
Bl = 4.3 Tm, REC = 5.5 Ω, CMS = 9.36·10 −4 m/N, RMS = 1.84 kg/s, MMD = 15·10 −3 kg,
and effective radiating area SD = 1.2·10 −2 m2.
Tasks:
a. Calculate QB for the driver mounted in a closed box that has volume
V = 10−2 m3.
b. Calculate the box volume needed for QB = 0.707 for the driver mounted in a
closed box.
16.4 A closed-box loudspeaker system uses a pair of isobaric connected drivers. The
10″ moving-coil drivers have Bl = 4.3 Tm, REC = 5.5 Ω, CMS = 9.36·10 −4 m/N,
R MS = 1.84 kg/s, MMD = 15·10 −3 kg, and effective radiating area SD = 1.2·10 −2 m2.
Tasks:
Calculate the resonance frequency for the drivers mounted in a closed box that
has volume V = 10 −2 m3
16.5 In many cases, the maximum low-frequency output of a loudspeaker will
be limited by the cone area and cone maximum displacement. A study of
commercial voice coil drivers showed that the peak displacement x̂ (m) was
related to driver cone diameter d (m) as
x̂ = 2 ⋅ 10 −2 d (16.18a)
The study also revealed that because of voice coil heating, the maximum
sound pressure level Lp,max at 1 m distance was related to d as
L p,max = 97 + 40d [dB] (16.18b)

Tasks:
a. Derive an expression for the maximum sound pressure level Lp at 1 m
distance as a function of cone diameter d and frequency f when the cone
displacement x̂ is the limiting variable. Assume the driver in a small
enclosure and radiating omnidirectionally.
b. Calculate for two driver diameters, 0.1 and 0.2 m, the maximum sound
pressure level Lp at 1 m distance in the frequency range 40–200 Hz taking
into account both the voice coil heating effect and the maximum cone
displacement.
16.6 Two identical drivers are mounted close to one other in a loudspeaker enclosure
and work acoustically in phase. The drivers are electrically connected in
parallel. Neglect the outside acoustical driver-to-driver coupling since the
internal coupling is strong because of the confined air volume.
Task:
Draw the mechanical impedance analogy for the system. What will be the new
resonance frequency?
REFERENCES
1. Tyrland, S., Design of a monitor loudspeaker, Thesis (in Swedish), Report 74–35,
Department of Applied Acoustics, Chalmers University of Technology, Göteborg,
Sweden (1974).
978-0883184943.
4. Waterhouse, R. V., Output of a sound source in a reverberation chamber and other
reflecting environments, J. Acoust. Soc. Am., 30(1), 4–13 (1958).
5. Backman, J., Improvement of one-dimensional loudspeaker models, Proceedings of the
123rd Audio Engineering Society Convention, New York, Paper 7253 (2007).
6. Matzuk, T., Improvement of low-frequency response in small loudspeaker systems
by means of the stabilized negative-spring principle, J. Acoust. Soc. Am., 49(5A),
1362–1367 (1971).
7. Olson, H. F., Direct radiator loudspeaker enclosures, J. Audio Eng. Soc., 17(1), 22–29
(January 1969).
8. Kuttruff, H., Room Acoustics, Spon Press, London, U.K. (2009) ISBN-13:
978-0415480215.
17 Vented Box
Loudspeakers
17.1 EXTENDED LOW-FREQUENCY RESPONSE

It is often desired to reduce the high-pass cutoff frequency from that possible with
a closed box mounted driver. This can be done using electrical, mechanical, or
acoustical means.
Acoustically, a lower cutoff frequency can be achieved with a larger box compli-
ance that requires a larger box or having a steeper cutoff of the frequency response
below the cutoff frequency with retained box size which requires the use of extra
acoustic circuit components. The most usual approach is to limit the box size and
to provide the acoustic circuit with extra resonant circuits. The resonant circuits can
also be used to limit the high-frequency response giving the frequency response of
the loudspeaker a band-pass filter characteristic.
Lowering the cutoff frequency mechanically requires larger moving mass in the
diaphragm voice coil assembly. If the diaphragm area remains unchanged, this leads
to reduced efficiency and sensitivity. With a larger diaphragm area, the directivity of
the driver increases as does the risk of diaphragm buckling at large displacements.
Electrically, the high-pass filter response can be changed by “preprocessing” the
signal fed to the driver by a suitable transfer function, usually a resonant circuit, by
using negative feedback with a sensor attached to the diaphragm, or by using an
electronic circuit that, because of the impedance seen by the driver, adds virtual
mass to the driver’s mechanical system.
17.2 LOUDSPEAKER AS HIGH-PASS FILTER

It is convenient to study the frequency response of loudspeakers as if they were
electronic filters and to use terms and ideas from electrical filter theory. In filter
theory, a filter having the frequency response of the loudspeaker, in the frequency
region we are studying here, would be called a high-pass filter. Such filters are
characterized in classical filter theory by the following properties:
• Cutoff frequency. This is the frequency where the response has dropped or
reached −3 dB relative to that in the region of asymptotically flat frequency
response, i.e., the filter passband.
• Filter order. The filter order multiplied by ±6 gives the dB rate of increase or
decrease in the filter in the region(s) outside the flat frequency response region.
• Ripple. This is the oscillatory deviation from flat frequency response in the
nominally flat frequency response region.
337
Each “filter family” will have its own transfer function characteristics. Some
classical filter families commonly used in electronic engineering and audio filter
applications are:
• Butterworth filters. These are characterized by “maximum flat” frequency

response in the filter passband.
• Bessel filters. These are characterized by constant group delay, i.e., linear
phase response in the filter passband (constant delay).
• Chebyshev filters. These have ripple in their passband response but give
somewhat higher response in the passband until the cutoff frequency is
reached.
It was pointed out in Chapter 16 that the acoustic suspension or closed box loudspeaker
has a second-order high-pass filter frequency response, that is, a drop of 12 dB
below the cutoff frequency f−3dB. The choice of filter characteristic is, in this simple
case, done by choosing the appropriate damping. We note from Figure 16.6 that the
diaphragm mass, the voice coil resistance, and the Bl product all must contribute
to the frequency response and damping characteristic. The damping naturally also
affects the transient response of the loudspeaker. Figure 16.7 shows some amplitude
response curves for different values of resonance damping.
17.3 PORTED BOX AND DRONE CONE DESIGNS

A simple and common method to acoustically enhance sound output in some
frequency region is to use an extra resonance generated by a “parasitic” radiating
element, similar to the way frequency response is extended in the voice coil
microphones studied in Chapter 14. This is the working principle of the so-called
ported enclosure (sometimes also called vented box or bass reflex enclosure). The
basic principle is that shown in Figure 17.1.
Alternatively, the function of the mass of the port air can be replaced by that of a
“drone cone.” The mass MAP or MAdc is driven by the transducer via the compliance
CAB, much like the mass at the far end of a spring that is shaken at the close end. The
masses MAD and MAP (or MAdc) will be part of a coupled system.
Figure 17.2 shows the acoustical part of the circuit representing the diaphragm,
box volume, and port. The addition of the extra resonant circuit is apparent in the
analog circuit shown in Figure 17.2 as a branch parallel to the box acoustic imped-
ance R AB + CAB. The R AB resistance is due to losses such as those due to possible
porous absorption and box wall flexure. Most boxes will also have some leakage
(besides that of the port) that is represented by a resistance R AL . Experimental results
show that inclusion of R AL is often sufficient to represent the losses associated with
practical loudspeaker boxes [8].
The port branch consists of the tube air mass MAP in series with the resistance due
to friction in the tube R AP, and in series with mass MARp and resistance ℜARp due to the
radiation impedance load on the outside of the tube. Radiation by the diaphragm is
represented as the radiation resistance ℜARd and ℜARp. The corresponding circuit for
the case of the drone cone system is shown in Figure 17.3. Note that the drone cone
Vented Box Loudspeakers 339
Bl Bl
REC REC
SD SD
MAD MAD
RAS RAS
CAS ZARd RAB CAS ZARd
RAB
CAB
CAB
SP
MAP RAP Sdc
MAdc RAdc ZARdc
ZARp
CASdc
(a) lP (b)
FIGURE 17.1 Parasitic radiators can be used to extend the low-frequency cutoff of a
loudspeaker system: (a) the ported enclosure uses the air column in the tube port as a parasitic
resonant radiator to enhance low-frequency sound output. (b) The drawing on the right shows
the analogous drone cone approach. Note that drone cone needs suspension. Compare to
closed box in Figure 16.5.
MAD CAS RAS MARd MARp

ARd UD UP ARp
(Bl)2 UB UL
REC SD2 RAB MAP
eGBl
RAL
REC SD
CAB
RAP
FIGURE 17.2 The acoustic impedance analogy for the ported box loudspeaker.
MAD CAS RAS MARd MARp

ARd UD UP ARp
(Bl)2 UB UL
MAdc
REC SD 2 RAB
eG Bl
RAL CASdc
REC SD
CAB
RAdc
FIGURE 17.3 The acoustic impedance analogy for the “drone cone” loudspeaker.
compliance CASdc is in series with the mass and not connected to earth in the same
way as the driver’s components. As long as the compliance CASdc is much larger than
that of the air in the box CAB, it will have little influence on the function of the circuit.
A simplified equivalent circuit of the ported box loudspeaker is shown as an
acoustic impedance analogy in Figure 17.4. The resistance in box volume circuit
CAS
MAD RAS MARd
UD UP
(Bl)2 UB
MAP
eG Bl RECSD2
CAB
REC SD
MARp
FIGURE 17.4 Simplified acoustical impedance analogy for ported box at very low
frequencies.
has been removed since this simplifies the analysis. One particular complexity
has been neglected in the circuit, namely the coupling between the two oscillating
masses on the outside of the enclosure. This coupling will affect the resonance
frequencies and the impedances slightly. The coupling will be more pronounced
when the port and the diaphragm are very close together. This is usually not the
case for conventional loudspeakers.
The calculation of the sound pressure is straightforward using superposition of
pressures since the interaction between the two radiators has been neglected. The
sound pressure contributions p̲ P and p̲ D due to the volume velocities UP and UD
generated by the port air and loudspeaker diaphragm vibration are
e − jkrD
pD (rD ,ω ) = j ω ρU D (17.1)
4πrD
e − jkrP
pP (rP ,ω ) = j ω ρU P (17.2)
4πrP
In practice, the distances from the diaphragm and port to the listening position
are going to be almost the same rP ≈ rD ≈ r at the wavelengths of interest for low-
frequency sound reproduction systems. Because the sound pressure is proportional
to volume acceleration, the shape and area of the port or drone cone do not matter as
long as they are small compared to wavelength.
Since we assume the system linear, we can apply superposition. Because of the
reference direction of Up, from the rear of the loudspeaker diaphragm, its pressure
contribution is added using a minus sign.
e − jkr e − jkr
psum (r,ω ) = j ω ρ (U D − U P ) = j ω ρU B (17.3)
4πr 4πr
All that remains now is to calculate the volume velocities of interest, and we will
obtain an expression for the sound pressure in the far field of the loudspeaker (in an
anechoic environment). We also note that since it is the volume velocities that are
of interest, there is little acoustical difference between the port and drone cone
approaches as long as CASdc is large.
It is important to note that the approach here as well as in many references applies
to small-signal conditions; that is, all components are assumed to behave linearly.
17.4 FREQUENCY RESPONSE USING CLASSICAL FILTER THEORY

As mentioned previously, the analysis can use classical filter theory in which the
transfer functions of the high-pass filters are described as polynomials. Such a filter
will have the general polynomial form
ωn
H (ω ) = s (17.4)
∑ω n
+ c n −1ω n −1 + ... + c 0

The frequency response of the filter is of interest to us, i.e., 20·log|H(ω)|. When the
modulus of the transfer function is of the form shown in Equation 17.5, the filter is
said to be a Butterworth high-pass filter of order n.
ωn
H (ω ) = (17.5)
ω 2 n + ω c2 n

The cutoff frequency of a filter ωc is defined as the frequency at which the modulus
of the transfer function has dropped by −3 dB relative to its asymptotic value at
high frequencies. Above the cutoff frequency, the Butterworth filters have transfer
functions that have a quite frequency-independent (“maximum flat”) response. The
asymptotic high-end value will be |H(ω)| ≈ 1 without ripple while the slopes of the
transfer functions will be 6·n dB/octave, well below the cutoff frequency.
Other target filters that are often used, such as Bessel and Chebyshev filters, are
characterized by other types of polynomials. The reader is referred to a standard
textbook for such designs [8] and to Refs. [1,2]. Refs. [3–9] give an extended treatment.
The Bessel filter characteristic gives the best transient response in the sense
that there is no overshoot on transients. Because of the resonances of the listening
room, the Bessel filter frequency response is probably unnecessary. The Chebyshev
frequency response characteristic has a fair amount of ringing due to the resonances
needed to give the filter its sharp frequency cutoff characteristics.
In many cases, the loudspeaker driver is specified beforehand, because of supply,
cost, and convenience. For a given loudspeaker, adjustment of the loudspeaker
enclosure to obtain a second-order Butterworth response pushes the cutoff frequency
up, often by about half an octave above the free-field resonance frequency, although
this can sometimes be compensated by filling the box interior with a sound-absorptive
material as discussed in Chapter 16.
We often need to extend to extend the low-frequency range of the loudspeaker,
so that the low-frequency cutoff is at a lower frequency than the free-field reso-
nance frequency of the driver. This can be achieved using more complex enclosures.
This is achieved at a price however since such enclosures will usually give more
delay at low frequencies as a result of the resonances that need to be introduced as
0
–3
–5
–10
B4
–15 QB3
C4
–20
–25
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE 17.5 Frequency response characteristics for three different ported box alignments
for the same driver obtained by changing the box volume. The alignment numbers refer to
3 (QB3), 5 (B4), and 8 (C4) in Table 17.1.
shown in Appendix B. Such resonances and the delays they cause may be audible
and unacceptable if larger than about 1 ms at high frequencies.
Design of ported box low-frequency loudspeakers is in practice often limited to
adjusting the circuit components so that the polynomials can be obtained, either
with a given loudspeaker driver or with a given box volume. Figure 17.5 shows some
frequency responses that are possible using a loudspeaker driver having a resonance
frequency f S and compliance CAS with different box volumes having compliances
CAB. As shown, a flat, less resonant high-pass response is obtained at the penalty of
an increased high-pass cutoff frequency.
Table 17.1 shows the parameter adjustments that are needed to obtain various
high-pass response transfer functions. The table uses variables derived using the
components shown in the circuit in Figure 17.4. A more extensive table and further
design equations may be found in Ref. [1].
Midrange and high-frequency loudspeakers sometimes benefit from the addition
of a port to their enclosures. The reason is that the port also acts as a low-pass filter for
sound arriving at the loudspeaker. The diaphragm of a midrange or high-frequency
driver may be forced to move by the low-frequency sound pressure from a nearby
bass loudspeaker. This low frequency will generate intermodulation distortion in the
sound from the loudspeaker. By short-circuiting this sound using a port to the back
of the midrange or high-frequency driver enclosure, the distortion can be minimized.
17.5 BANDPASS DESIGNS

Enclosures featuring the bandpass design are characterized by the addition of
one or two acoustical low-pass filters to the loudspeaker enclosure. There are two
types of bandpass designs; both are essentially modifications of a ported enclosure.
TABLE 17.1
Design Adjustments for Different Vented Box Designs
No. Type Ripple (dB) ω−3/ωS ω−3/ωb α = CAS/CAB QT
1 QB3 0 2.68 1.34 10.48 0.180
2 QB3 0 2.28 1.32 7.48 0.209
3 QB3 0 1.77 1.25 4.46 0.259
4 QB3 0 1.45 1.18 2.95 0.303
5 B4 0 1.000 1.000 1.414 0.383
6 C4 0 0.867 0.935 1.055 0.415
7 C4 0.2 0.729 0.879 0.729 0.466
8 C4 0.9 0.641 0.847 0.559 0.518
9 C4 1.8 0.600 0.838 0.485 0.557
Source: Adapted from Thiele, A.N., J. Audio Eng. Soc., 19(5), 382, 1971.
Note: The designs are denoted quasi-Butterworth third order (QB3), Butterworth
fourth order (B4), and Chebyshev fourth order (C4). In the table, ω−3 is the
−3 dB cutoff frequency, ωs is the free driver resonance frequency, and ωb is the
port and box resonance frequency. QT is the Q of the driver circuit.
1
ωs =
(MAD + MARd)CAS
1
ωb =
(MAP + MARp)CAB
ωs(MAD + MARd) (MAD + MARd) 1

QT = =
RAT CAS (Bl)2
RAS + ARd +
RECS2D
In the first design, the front side of the loudspeaker is sealed off to become a closed
box system. Such a system is shown in Figure 17.6a. The second approach is to fit
a low-pass filter to the front side of the loudspeaker as shown in Figure 17.6b. The
acoustical impedance analogy of the bandpass ported box design in Figure 17.6b
is shown in Figure 17.7.
The main advantage of the approach is the low-pass filter action of the port(s).
The acoustic mass of a port reduces the amount of nonlinear distortion radiated
from the loudspeaker. This is particularly essential in sub-woofer loudspeakers,
i.e., loudspeakers for the frequency range below 100 Hz. The reason is the sensitivity
of hearing to distortion in this frequency range due to the reduced sensitivity of
hearing to the fundamental, as indicated by the equal-loudness contours studied
earlier. High particle velocities at the port because of resonance may cause
turbulence so the port cross-sectional area should be made large enough to keep the
velocity to below about 5 m/s.
ZAR ZAR2 ZAR1
SP SP1
MAP SP2 MAP1
MAP2
SD SD
MAD MAD
RAS RAS
CAS CAS
CAB2 CAB1 CAB2 CAB1
(a) (b)
FIGURE 17.6 Two different bandpass enclosures: (a) with closed box design and (b) with
ported box design.
eG Bl (Bl)2
ZAR2 RESD RESD2 CAS ZAR1
UP2 MAD RAS
UC UP1
UB2 UB1
MAP2 CAB2 CAB1 MAP1
FIGURE 17.7 The acoustic impedance analogy for a bandpass ported box design shown in
Figure 17.6b.
It is also of course possible to design higher-order acoustical low-pass filters that

use more box volume and ports, having even steeper low-pass filter action.
17.6 EXTERNAL FILTERS

Since the high-pass transfer function can only be of the fourth order for a passive
ported box loudspeaker system, any increase in transfer function order must be by
an external device, analog or digital [9–15]. Although one can argue that such a
loudspeaker is an active loudspeaker and should be included in Chapter 23, it is
more practical to cover this type here. Passive and active circuits for second-order
electrical filters can be found in Appendix 2.
The transfer function for a high-pass response higher than the order allowed by
the box can be written as a product between an external transfer function and the
transfer function of the loudspeaker. Let us assume that the closed box is tuned
to a second-order Butterworth characteristic B2 with the transfer function HB2(ω)
and that we wish to have a fourth-order B4 characteristic that has the transfer
function HB4(ω). Then, to have a fourth-order response, we would need to multiply

HB2(ω) by HEXT (ω) as shown
ω2 ω2
H B 4 (ω) = H B 2 ( ω ) H EXT ( ω ) = ⋅ (17.6)
ω 2 + j1.4142ω + 1 ω 2 + a1ω + a0
By suitable choice of the constants a1 and a 0, we can make this transfer function the
same as that of a ported enclosure tuned to a Butterworth fourth-order characteristic
ω4
H B4 (ω) = (17.7)
ω 4 − j 2.6131ω 3 − 3.4142ω 2 + j 2.6131ω + 1
or any other fourth-order characteristic we wish. The frequency responses of the

three transfer functions HB4(ω), HB2(ω), and HEXT (ω) are shown in Figure 17.8. We
note that HEXT (ω) must give a slight rise to the response around the cutoff frequency
to give the correct frequency response.
A natural question to ask is of course why one should want to use such an approach
instead of just tuning the box and driver for a B4 characteristic. One reason could be
that the signal fed to the loudspeaker has much signal below the cutoff frequency. We
should remember that the cancellation of low frequencies afforded by the ported box
B4 tuning is due to dipole cancellation. Below the cutoff frequency, the driver will be
free to do very large excursions since only some damping and the compliance limits
the diaphragm movement. One drawback of using the external filter is that more
electric power will be needed for the same sound pressure. Another more serious
drawback of using external filters is that the tolerances in driver manufacture are
0
–3
–5
–10
B2
–15 B4
EXT
–20
–25
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE 17.8 The frequency responses for the three filter functions, HB2(ω), HB4(ω), and
HEXT (ω).
comparatively large, the higher the order of the desired filter function, the smaller
the component tolerances must be. It is also important to allow for the aging of
mechanical components such as loudspeaker and drone cone compliances.
For HP filters of an order higher than four, the use of an external filter is necessary
in any case. If a fourth-order “box” filter characteristic is chosen, then that choice
can be made so that the box is much smaller than that required by a conventional
B4 tuning. The correct B4 tuning is then achieved by using an appropriate compen-
sating second-order HP filter.
17.7 DRIVER CONE EXCURSION

Since a flat frequency response in the far field requires constant frequency-independent
volume acceleration, the excursion necessary for the closed box loudspeaker dia-
phragm must increase toward low frequencies to compensate for the ω2 characteristic
of the acceleration. The magnitude of cone excursion is an important aspect in the
design since the nonlinear distortion due to magnet field and compliance nonline
arities is likely to increase along with increasing diaphragm excursion magnitude.
The excursion xMD of the diaphragm is the integral of the diaphragm velocity uMD
that can be found from the mechanical circuit analogy.
Figure 17.9 shows the normalized frequency responses for three types of alignments
using the same diaphragm mass and tuned to the same −3 dB cutoff frequency. We
see that at frequencies close to the cutoff frequency, the excursion magnitude is
smaller for the B4 and B6 alignments than it is for the closed box B2 alignment.
Note that the B6 alignment is achieved with an external second-order HP filter as
described in the previous section.
0
Diaphragm excursion [dB]
B2 –3
–5 B4
B6
–10
–15
–20
–25
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE 17.9 Frequency response curves for the diaphragm excursion normalized to the
same diaphragm mass and cutoff frequency for the second-order (B2), fourth-order (B4)
alignments, and sixth-order (B6) alignment that has been achieved with an external second-
order HP filter.
REVIEW QUESTIONS
17.1 Derive the acoustic impedance circuit analogy for an electrodynamic
loudspeaker having a ported box design. Draw the circuit so that the compliance
of the box volume is attached to ground in the style of Figure 17.7.
17.2 Derive the acoustic impedance circuit analogy for an electrodynamic
loudspeaker driver mounted in a bandpass-type box. Draw the circuit so
that the compliance of the box volume is attached to ground in the style of
Figure 17.7.
17.3 Why is it that we can think of loudspeaker box alignment in terms of filter
functions?
17.4 Discuss the filter characteristics of some common filter functions used in
loudspeaker box alignment.
17.5 Why does not the shape of the port opening influence the frequency response?
17.6 Discuss the advantages and disadvantages of using external filters for
loudspeaker box alignment.
PROBLEMS
17.1 Two identical voice coil drivers electrically connected in parallel and with
the same polarity share a closed box with volume VB. Assume that the box is
designed for a B2 alignment.
Tasks:
a. Assume one of the drivers removed and its mounting hole in the box
blocked. Calculate the new QT value.
b. Assume one of the drivers electrically disconnected (terminals open) but let
the driver remain mounted in the box. What will be the new response?
c. Assume one of the drivers electrically disconnected (terminals short-
circuited) but let the driver remain mounted in the box. What will be the
new response?
17.2 Three types of loudspeaker enclosures are shown below. Assume that the box
dimensions can be considered small compared to the wavelength.
ZAR
REC
Bl
SD1 SP
MAD1 ZAR1 CAB ZAR1
RAS MAP
CAS1 Bl
SD Two identical
MAD Bl
SD2 RAS drivers REC
MAD2 CAS electrically SD
RAS ZAR2 REC connected MAD
CAS2 out-of-phase RAS
ZAR2 CAS
CAB CAB2 CAB1 and
in parallel
Drone cone Bandpass Isobarik
Task:
Use the acoustic impedance analogies in Problems 15.3 and 15.4 and
adjust them to model the driver and enclosures. How is the radiated power
calculated?
17.4 The figure below shows a small bandpass loudspeaker enclosure that has a
leak in one of its volumes with an acoustic resistance R AL . Assume that the box
dimensions can be considered small compared to the wavelength.
ZAR2 ZAR1
SP1
SP2 MAP1
MAP2
SD
MAD
RAL
RAS
CAS
REC
Bl
CAB2 CAB1
Tasks:
a. Draw the acoustic impedance analogy for the system.

b. How is the radiated output power affected by the presence of the leak?
17.4 The figure below shows a small bandpass loudspeaker enclosure. The
loudspeaker is to function as a sub-woofer for the reproduction of very low-
frequency sound. The port of the lowpass filter is facing a car compartment
that here can be considered a compliance CA with a small leak R AL .
ZARL RAL
CAB2 CAB1
SP1
MAP1 RACC
CACC
SD
MAD
RAS
CAS
REC
Bl
Tasks:
a. Draw the acoustic impedance analogy for the system.

b. Show the pressure that corresponds the sound pressure in the compartment.
c. Estimate the highest frequency at which the assumption of the car
compartment as a compliance is valid.
17.5 A voice coil driver has these data: f0 = 40 Hz, Qs = 1.1, REC = 6.5 Ω, Bl = 7.0
N/A, SD = 3 · 10 −2 m2, and MMD = 2 · 10 −2 kg.
Task:
Design (if possible) boxes corresponding to alignment numbers 3 (QB3),
5 (B4), and 8 (C4) in Table 17.1 and Figure 17.5.
REFERENCES
1. Thiele, A. N., Loudspeakers in vented boxes: Part l, J. Audio Eng. Soc., 19(5), 382–392
(1971).
2. Thiele, A.N., Loudspeakers in vented boxes: Part 2, J. Audio Eng. Soc., 19(6), 471–483
(1971).
3. Small, R. H., Vented-box loudspeaker systems—Part 1: Small-signal analysis, J. Audio
Eng. Soc., 21(5), 363–372 (1973).
4. Small, R. H., Vented-box loudspeaker systems—Part 2: Large-signal analysis, J. Audio
Eng. Soc., 21(6), 438–444 (1973).
5. Small, R. H., Vented-box loudspeaker systems—Part 3: Synthesis, J. Audio Eng. Soc.,
21(7), 549–554 (1973).
6. Small, R. H., Vented-box loudspeaker systems—Part 4: Appendices, J. Audio Eng. Soc.,
21(8), 635–639 (1973).
7. Borwick, J. (Ed.), Loudspeaker and Headphone Handbook, Focal Press, Oxford,
U.K. (2001) ISBN 0-240-51578-1.
8. Zverev, A. I., Handbook of Filter Synthesis, Wiley-Interscience, New York (2005) ISBN-
13: 978-0471749424.
9. Leach, W. M., Active equalization of closed-box loudspeaker systems, J. Audio Eng.
Soc., 29(6), 405–407 (1981).
10. Leach, W. M., A generalized active equalizer for closed-box loudspeaker systems,
J. Audio Eng. Soc., 38(3), 142–146 (1990).
11. Benson, J. E., An introduction to the design of filtered loudspeaker systems, J. Audio
Eng. Soc., 23(7), 536–545 (1975).
12. Benson, J. E., Synthesis of high-pass filtered loudspeaker systems: Part 1—isolated
filters driving second order (closed-box) systems, J. Audio Eng. Soc., 27(7/8), 548–561
(1979).
13. Benson, J. E., Synthesis of high-pass filtered loudspeaker systems: Part 1(a)—A
supplementary note on QB2, SC3, and SC4 alignments, J. Audio Eng. Soc., 27(9),
667–672 (1979).
14. Benson, J. E., Synthesis of high-pass filtered loudspeaker systems: Part 2-isolated filters
driving fourth order (reflex) systems, J. Audio Eng. Soc., 27(10), 769–779 (1979).
15. Zaustinsky, E., Are equalized closed-boxes preferable to vented boxes? Proceedings of
the 81st Audio Engineering Society Convention, Los Angeles, Paper #2415 (1986).
18 Transmission Line
Loudspeakers
18.1 INTRODUCTION
Impedance matching by tubes and ducts was discussed in Chapter 7. Ducts are
common in electroacoustics; they can be found in various types of delay lines and
transmission line loudspeaker cabinets. The term transmission line is usually used
for those loudspeaker enclosures where the duct is at least a quarter of a wavelength
long at the lowest operating frequency of the loudspeaker system. Many loudspeaker
designs use transmission lines that behave like open or closed tubes or ducts, with
the driver mounted at one end generating a plane wave through the duct. Horns may
be considered tapered ducts.
Figure 18.1 shows the two common types of transmission line speaker enclosures,
those that are effectively anechoic terminations for the waves radiating from the back
side of the driver’s diaphragm and those that use a “rear” opening, a mouth, to radi-
ate sound. The latter can roughly be thought of as quarter-wave transformers that
convert the low impedance seen by the duct opening to a high load impedance on the
back of the driver. By tapering the duct and filling it with an acoustic damping mate-
rial, the line characteristics can be tailored for the desired resonance frequency and
damping. The transmission line enclosure can then function as an anechoic termina-
tion over most of the frequency range while at the same time acting as a quarter-wave
transformer in a narrow frequency range.
One advantage of the transmission line enclosure over the closed box or vented
enclosure is its form factor. The transmission line can be made thin and long in
contrast to the quite compact enclosures required by the other types. In many cases,
the enclosure will be some form of hybrid between the ported box and transmission
line. The cross section area will be large close at the driver and then form a straight
or tapered duct as one moves away from the driver.
Transmission lines are also used for microphones; the probe tube microphone is
one example. The human ear canal acts as a probe tube that protects the mechani-
cally sensitive eardrum from mechanical damage. In microphone probe tubes, how-
ever, the internal diameter is so small that viscous losses at the probe walls introduce
considerable damping to the resonances of the tube.
Waves in a duct can bounce between the walls of the duct just as waves do in a
room. The duct can be regarded as a room having an extreme shape. Appendix F
discusses the theory of sound propagation in straight unlined ducts with rigid walls.
It is shown that a straight duct will carry only plane sound waves as long as the fre-
quency of the sound is below the cutoff frequency of the duct. The cutoff frequency
for rectangular ducts is the frequency at which the largest duct dimension is one
351
Possibly closed end with vent
ZARd
Bl
REC
SD
MAD
(a) Possibly vented ld RAS
CAS
ZARd
Wedge made of sound-absorbing material

(b) Possibly air-space to enhance low frequency sound-absorption
FIGURE 18.1 Two types of transmission line loudspeakers: (a) with open far end and
(b) closed far end but absorptive filling to simulate infinitely long tube. In both cases, the
length of the tube needs to be about λ/4. There may also in this case be a vent at the far end
of the line.
half wavelength or wider as given by Equation F.27. The circular duct will have its
cutoff frequency when the duct diameter is about 58% of the wavelength, as given
by Equation F.34.
18.2 ATTENUATION BY ABSORPTIVE FILL AND LINING

Attenuation of waves in ducts can be due to a lossy medium that fills the duct, to
duct walls that have finite impedance, or simply to the duct having varying cross
section. One must differentiate between wide ducts, i.e., ducts that have internal
dimensions that are much larger than the wavelength and those ducts that have at
least one dimension smaller than one-half wavelength of the sound.
In most ducts, the sound waves will travel down the duct as a combination of
a plane wave and higher-order modes. The higher-order modes are generated by
nonplanar sound sources such as drivers that have diaphragms that are not flat pis-
tons, that do not fill the entire duct cross section, or are asymmetrically mounted.
The driver or exciter may also not vibrate with the same phase and amplitude over
their surface, for example, because of modal vibration in the diaphragm. Since
higher-order modes travel farther for a certain duct length, they will be attenuated
quicker. Many porous wall absorbers also have higher attenuation for waves that
have oblique incidence.
Ducts filled with porous sound-absorbing material such as glass and mineral wool
will attenuate the sound. For low frequencies below the duct cutoff frequency, the
attenuation will be roughly that of plane waves as described in Appendix E and
Transmission Line Loudspeakers 353
depend primarily on the flow resistance of the material. Glass and mineral wool can
be considered to have a rigid structure.
By using a flexible porous filling such as natural (fatty) sheep’s wool, the tube can
often be shortened. The reason is that the mass and friction of the wool in combina-
tion with the elasticity of wool forms a distributed resonant system that can reduce
the transmission line length requirement. This is particularly useful for woofer appli-
cations. Plastic fiber wools have similar properties, but the fatty surface and friction
of the wool fibers make them superior in this application.
For ducts lined with sound-absorptive material, the attenuation will depend on the
sound-absorption characteristics of the material. A simplified theory for the attenua-
tion of low-frequency “plane” waves by sound absorbent lining in ducts can be found
in Ref. [1]. More exact theories may be found in Refs. [2,3]. Note that “true” plane
waves cannot exist in lined ducts. The attenuation for plane waves ΔL [dB/m] will
be approximately
P
∆L ≈ 1.1
S
α0 [dB/m ] (18.1)

where
P is the acoustically lined width of the duct
S is the cross section of the open duct area
α is the sound-absorption coefficient for the material for perpendicular incidence
This equation was derived for the case of low frequencies in a duct where one wall is
covered with a porous absorber as shown in Figure 18.2 and can be considered very
conservative. The attenuation for oblique waves will be much higher.
The attenuation can often be twice the D value, so the formula should only be
considered to show the influence of the P/S ratio and the sound absorption. The
attenuation will be largest in the frequency range where the open height is about
one half wavelength. The attenuation is larger when the sound-absorbing material is
applied in designs with intervening walls so that there will be no wave propagation
along the duct inside the absorber.
The formula is not valid for frequencies where the duct dimensions are large
compared to wavelength. The attenuation will then be small since sound will travel
similar to light rays through the open area of the duct. A bend or even a small
deviation from a straight canal will increase the attenuation.
Sound absorbent lining
h h
FIGURE 18.2 A duct lined on one side only. The duct area is S = P·h.
18.3 ATTENUATION BY FOLDS

Many types of loudspeakers feature ducts with bends or folds. These include trans-
mission line, quarter-wave resonator, and “folded” horn loudspeakers. Attenuation
or reflection of waves may be desired for the function of the box/duct. The attenu-
ation by the presence of bends varies widely depending on the angle of the fold, its
geometry, and the amount and position of absorbent lining at and close to the fold.
In a duct that has bends or folds, these will cause reflections primarily at mid and
high frequencies. Any fold will introduce a change in impedance even for the plane
wave mode. Rapid changes in duct cross section should be avoided so that sound is
not reflected back and the transmitted wave attenuated.
Sometimes losses due to duct folds and bends are not desirable; this is the case, for
example, in folded horns. By beveling the duct corners, one can have high-frequency
sound mirror through the duct. Folds may also be equipped with special corner
fillers to reduce the attenuation.
At low frequencies, the fold will act as a local compliance and the attenuation
will then be insignificant. As the fold starts to become one half wavelength wide,
the attenuation will become large as indicated in Figure 18.3a, even if the duct
walls close to the fold are unlined. Attenuations of the order of 5–10 dB may be
found for 90° folds (usually called “elbows” in duct literature) at these frequencies.
Because of the presence of higher-order modes in the duct, the attenuation will start
to become large even for waves having longer wavelengths. If high attenuation is
desired, the lining should be at least 5–6 duct widths on either side of the elbow as
shown in Figure 18.3b.
Many loudspeaker boxes may be considered short ducts with an elbow immedi-
ately behind the driver. Unless treated, the back wall may reflect sound back toward
the driver at high frequencies. Fitting a sound-absorbent lining to the box interior
λ/2
Incident wave h
Length from bend

Unlined duct
Sound absorbent lining
(a) (b)
FIGURE 18.3 Lined and unlined ducts. (a) When the duct width is one half wavelength, it is
difficult for the incident plane wave to drive the wave after the elbow because of cancellation.
(b) For large attenuation of both plane and higher-order mode waves, the lining should extend
several open duct width h on either side of the elbow.
back wall both reduces high-frequency reflection of sound back through the driver
and increases the acoustical losses in the box over a wide frequency range.
18.4 CIRCUIT ANALOGIES FOR DUCTS

The possibility of representing sound propagation in tubes and ducts by acoustical
analogies was discussed in Chapter 7, and Table 7.1 shows some examples of discrete
lumped element approximations that can be used for this purpose.
18.4.1 Quarter-Wave Resonator
Figure 18.4 shows a section through a conical horn. In the conical horn or duct, the
cross section area varies to the square of the length z from the apex.
The acoustic impedance Z A1 that is seen into the “throat” at z1 of such a lossless
conical horn is given by Olson [4] as


ρc
sin ( kld ) + jZ A2
(
sin k (ld − θ2 )) 

ρc  S2 sin ( kθ2 )  (18.2)
Z A1 ( z1, z2 , k ) =
S1 
 Z A2
(
sin k (ld + θ1 − θ2 ) )−j
(
ρc sin k (ld + θ1 ) ) 

 sin ( kθ1 ) sin ( kθ2 ) S2 sin ( kθ1 ) 

where
kθ1 = arctan(kz1)
kθ2 = arctan(kz2)
Z A2 is the load impedance on the end of the duct or “mouth” at z2
For the cylinder, the equation simplifies to [4]
 ρc 
ρc  Z A2 cos ( kld ) + j S sin ( kld ) 
Z A1 (ld , k ) = (18.3)
S  ρc cos kl + jZ sin kl 

S
( d ) A2 ( )
d 


ZA1
ZA2
ld = z2 - z1
z
z=0 z = z1 z = z2
FIGURE 18.4 The geometry for the conical horn for Equation 18.2.
For lossy ducts, the complex wave number k must be used in analogy to the treatment
of sound-absorbing materials in Appendix E. The wave number will become com-
plex if the duct walls are not rigid or if the tube contains sound-absorptive material.
It is important to remember that the wave speed in such materials is 15%–20% lower
than that in free air.
18.4.2 Discrete Component Analogies

While one can use the impedance relations in Equations 18.2 and 18.3, it is often
more intuitive to use discrete component analogies in the analog circuits.
Consider the closed box loudspeaker acoustic impedance analogy shown in
Figure 16.6. There the acoustic load of the air inside the box on the back of the
driver appears as the two components CAB and R AB. In the ported box loudspeaker
shown in Figure 17.7, the acoustic load seen by the back of the driver has been
replaced by a parallel circuit. In this circuit, the diaphragm’s front-side sound
radiation is represented by energy losses due to the volume velocity current
through the mass M ARd and resistance ℜARd and the radiation from the port open-
ing by the energy losses due to the volume velocity current through M ARp and
resistance ℜARp.
In using a transmission line and acoustic labyrinth loudspeaker box designs, we
replace the discrete component parallel circuit analogy by a more complex load, but
the basic principle is the same: We wish to have useful sound radiation both from
the front of the driver diaphragm and from a port, in this case the transmission line
opening. The physical difference between the ported, transmission line, and acoustic
labyrinth designs lies in the small volume of air inside the ported box that results in
that box interior resonance do not need to be considered.
For tubes and ducts that have length larger than λ/16, there are a number of
options available for representing the acoustic impedance as shown in Table 7.1.
The choice depends mainly on the accuracy desired. Striving for extreme accu-
racy is usually unnecessary since practical tubes and ducts will have flexible walls
and, particularly if filled by sound-absorptive material, considerable and difficult to
calculate interior losses.
The analogies shown in Table 7.1 apply only to tubes and ducts that have constant
cross section area. Most transmission line loudspeakers use folded ducts and are
designed similarly to the one shown in Figure 18.5. Here the cross section area varies
along the length of the transmission line.
Consider the straight but tapered duct shown in Figure 18.6. The duct entrance
is at z = 0 and the duct is terminated at z = ld by an open port. The width of the duct
w is constant, but its height h(z) varies along the length axis z, so it is a form of
sector horn.
We use the open tube analogy for the acoustic impedance in Table 7.1. Since
the duct is terminated by an impedance that is smaller than the duct characteristic
impedance at the far end, the circuit impedance analogy must represent that end
by an inductance, which is a mass component. (The impedance analogy for a tube
closed at the far end will be terminated by a capacitance, which is a compliance com-
ponent. In electrical terms, this corresponds to the analogy consisting of Π-links.)
Wedge made of high specific flow

resistance porous sound-
absorbing material ZARd
Bl
Filling made of low specific flow REC
resistance porous sound- SD
absorbing material MAD
RAS
CAS
ZARp
FIGURE 18.5 A transmission line loudspeaker that uses a tapered and folded duct partially
filled with sound-absorptive material.
h(z)
as a function of position z
Tapered duct height h(z)
z
0 ld/8 ld/4 3ld/8 ld/2 ld
MA1 MA2 MA3 MA4
CA1 CA2 CA3 CA4
FIGURE 18.6 An analogy for a tapered duct.
For the open duct, the load impedance is the radiation impedance that is primarily
inductive as well.
The duct width is w and the height at coordinate z is h(z). The choice of compo-
nents in the T-links will now depend on the duct cross section area and the duct cross
section area is w·h(z) as follows:
Following Figure 18.6, the length of all four sections is ld /4, so with Equation 7.34,
the first section at A is obtained as
ρld ρ ld
M A1 = = (18.4)
4S  ld 
4w h  
 8

The second inductance at section B is obtained as
ρ ld
M A2 = (18.5)
 3l 
4w h  d 
 8 

The compliance for section A is obtained using Equation 7.36 as
l 
w h  d  ld
Sld  8
C A1 = = (18.6)
4ρc 2 4ρc 2
For section B, the compliance is
 3l 
w h  d  ld
 8 
C A2 = (18.7)
4ρc 2
The remaining link components for sections C and D are calculated in the same way.
Transmission line loudspeakers usually use electrodynamic voice coil drivers so we
will use that analogy. By inserting the four links into Figure 17.7, replacing the box
and port components, the full acoustic impedance analogy for the transmission line
loudspeaker is obtained as shown in Figure 18.7.
The duct taper can now be varied to obtain the desired characteristics. Since the
transmission line usually has some kind of absorbent filling, its effect on the length
can be taken into account by using a lower speed of sound in the equations for MA1,
MA2, MA3, MA4, CA1, and CA2, typically about 280 m/s for glass and mineral wool.
However, a better approximation for the analogy of a T-link with absorptive material
is that shown in Figure 7.10 for thin absorptive layer on the duct walls or the one in
Figure 18.8. This circuit is similar to that shown in Figure 7.10 but with additional
losses included due to the sound-absorptive filling.
eGBl
RESD MAD CAS RAS MA1 MA2 MA3 MA4
UC UP
MARd (Bl)2 MARp

RESD2
CA1 CA2 CA3 CA4
ARd ARp
FIGURE 18.7 The acoustic impedance analogy for the transmission line loudspeaker using
a discrete component representation of the duct.
MA1 RA,M1 RA,M2 MA2
RA,C1a
RA,C1b CA1
FIGURE 18.8 An improved acoustic impedance analogy for a transmission line section
filled with a sound-absorbing material such as glass wool or natural wool.
The component values in the circuits 18.7 and 18.8 will need to be adjusted to fit
with measurement results of the duct’s resonance frequencies and their damping.
It is often advantageous to slant the side walls of transmission line loudspeakers.
By changing the cross section area along the length of the duct, one is creating a sort
of horn/transmission line hybrid. The reason for this is that one can often adjust the
load of the transmission line on the driver and lower its resonance frequency while
at the same time keeping output high from the duct opening as shown in Figure 18.9.
0.22 m
0.45 m 0
–5
(a) –10
–15
–20
0.22 m 0.22 m 0
–5
(b) –10
–15
–20
0.22 m 0
0.45 m –5
–10
–15
2.00 m –20
20 50 100 200
Mouth Throat
(c)
FIGURE 18.9 A transmission line enclosure can be tuned somewhat by adjusting the taper
of the transmission line. The measured frequency response curves on the right show the
sound pressure level at the transmission line mouth. Shrinking the mouth size as in (a) leads
to the fundamental tube resonance occurring at a lower frequency (with higher Q) effectively
decreasing the cutoff frequency of the loudspeaker system as compared to (b) and (c). (After
Tyrland, S., Construction of a Monitor Loudspeaker (in Swedish), Report 74–35, Department
of Applied Acoustics, Chalmers University of Technology, Göteborg, Sweden, 1974.)
In summary, the transmission line, quarter-wave resonator duct loudspeaker offers

several advantages over the resistance terminated duct:
• Low-frequency sound is radiated from the transmission line mouth.

• The transmission line mouth sound output is essentially in phase with that
from the front side of the driver diaphragm.
• The transmission line load on the driver may be adjusted so that the
resonance frequency of the driver is reduced by the impedance load
of the tapered “inverted horn.” This is its primary advantage over the
conventional ported box loudspeaker.
18.4.3 Anechoic Termination
The transmission line enclosure with an anechoic termination was shown in Figure
18.1b. Such a loudspeaker box offers the driver an acoustic resistance back-side load
instead of the acoustic compliance of the closed box loudspeaker. The idea is not
only to provide a suitable acoustic resistance to trim the response of the loudspeaker
at resonance but also to avoid reflected high-frequency sound that can bounce back
from the enclosure and travel through the diaphragm to the listener.
The termination will consist of quarter-wavelength-long sound-absorbing wedge
made of plastic foam, a sound-absorbing fibrous material (glass wool, acrylic wool,
sheep’s wool, etc.), or even just an acoustically resistive mesh. Figure 18.10 shows
0.3 –10
0.2 –14
|r| Lp,ref – Lp,inc
0.1 –20
0 –26
50 100 200 500 1k 2k
d = λ/4 Frequency [Hz]
Wedge Airspace
|prefl|=|r pin| 0.20 m
d = 1.02 m 0.15 m 0.27 m
FIGURE 18.10 Measured plane wave sound reflection for a glass wool wedge used as the
anechoic termination of a transmission line. The reflection coefficient magnitude |r| and the
relative sound pressure level of reflected wave to incoming wave L p,ref − L p,inc are shown as
functions of frequency. Wedge bottom has square cross section. (After Beranek, L.L. and
Sleeper, H.P., J. Acoust. Soc. Am., 18(1), 140, 1946.)
the reflection factor magnitude for a wedge with an airspace behind it. The wedge
has a pyramidal shape with a square cross section bottom. By leaving an airspace
on the side of the wedge and at its back, the low-frequency range of the wedge can
be extended.
A duct filled with sound-absorbing material along its length will attenuate the
wave that travels in the duct. Shaping the material to a wedge minimizes the reflected
wave. The filled length of the duct needs to be approximately one quarter wavelength
at the lowest operating frequency to be effective.
Thus with a suitable resistive duct termination, the acoustic load on the back
of the driver’s diaphragm is almost fully sound-absorptive, for frequencies above
the cutoff frequency of the termination. Above the cutoff frequency given by the
properties of the wedge, the acoustic impedance of the duct termination is the duct’s
characteristic impedance given by Equation 18.8
ρc
RAB = (18.8)
ST
Here ST is the area of the tube interior cross section. At low frequencies, the load
impedance will depend on whether the far duct end is open or closed. With a closed
end, the low-frequency impedance will become a compliance and with an open end
it will be a mass load as discussed previously. Note that having a pure resistive ter-
mination R AB does not change the free-field resonance frequency of the driver but it
can introduce considerable damping to the system. Figure 18.11 shows the acoustical
analogy of the system.
By adjusting the load resistance (essentially by choosing the right duct cross sec-
tion area) one can adjust the high-pass filter characteristics to those of, for example,
a second-order Butterworth filter.
The design offers the advantage of absence of coloration due to interior enclosure
resonances and reflected high-frequency sound but has the drawback of no additional
radiation of low-frequency sound in the same way as in ported box loudspeakers.
Also the duct length may need to be long at low frequencies so that the duct may
have to be folded. This will cause unwanted reflections as mentioned previously.
eGBl
RESD CAS
MAD RAS UC
MARd (Bl)2
RESD2
RAB
ARd
FIGURE 18.11 The acoustic impedance analogy for the resistance terminated driver.
In practice, a low-frequency driver will act somewhat as a wall and attenuate the
reflected sound depending on the mass per unit area of its diaphragm.
In mid- and high-frequency-range loudspeakers, the duct end is sometimes
left open so that the sound from the low-frequency loudspeaker will cancel at
the mid- or high-frequency-range driver’s diaphragm to reduce intermodulation
distortion as mentioned in Chapter 17 for ported box enclosures. This is some-
times the case in the loudspeaker design along the lines of a tapered duct shown
in Figure 18.9a.
18.5 SPECIAL CONSIDERATIONS

Sometimes the driver is mounted in one of the duct walls. This could impair the
function of the duct box in several ways. There could be high-frequency reflection
from the opposing inner duct wall causing a reflected wave that passes through
the driver diaphragm and causes interference with the diaphragm front radiation.
This type of mounting also is more effective in generating oblique modes that
could be reflected by bends in the duct. By splitting the duct into many parallel
ducts, for example, by using separating walls or tubes, the oblique modes may be
pushed so high up in frequency that they will not interfere with the operation of
the loudspeaker system.
18.6 DUAL PORTED TRANSMISSION LINE LOUDSPEAKERS

In systems that need to work only within a fairly limited frequency range, one can
use dual transmission line boxes in analogy with the dual ported bandpass boxes
that were discussed in Chapter 17. The circuit analogy shown in Figure 18.12
will be very reminiscent of that shown in Figure 17.7. The tube lengths need to
be adjusted relative to one another to obtain a sufficiently flat frequency response
in the passband. The Q factors of the peaks will depend on the tube diameters.
Since hearing has long time constants at low frequencies, the Q factors need not
be very low. The fact that the sound at different frequencies will be radiated from
different locations is also no problem because of the poor localization ability
in hearing.
eGBl
RESD MAD CAS RAS
UC
UP2 UP1
(Bl)2
ZAR2 RESD2 ZAR1
FIGURE 18.12 The acoustic impedance analogy for the dual ported transmission line
loudspeaker.
d1 d2
(a)
d1
d2
(b)
FIGURE 18.13 (a) A low-frequency dual ported transmission line loudspeaker. (b) A folded
transmission line loudspeaker for enhanced efficiency at a single frequency. The bend is short
compared to the wavelength.
Figure 18.13a shows a possible design of such a dual ported transmission line
loudspeaker. The lengths d1 and d2 are chosen to obtain a wide frequency range with
well-damped resonances.
Another way of using the dual port system is that shown in Figure 18.13b. The
simultaneous use of a d1 = ¼ λ section and a d2 = ¾ λ section tuned to the same
frequency results a high acoustic load on the driver at the resonance frequency.
The volume velocities at the ports are in phase, and there is considerable coupling
between the openings. The advantage of this system is that the electroacoustic
efficiency can be made very high at the operating frequency.
REVIEW QUESTIONS
18.1 Transmission lines may be open or closed ducts. What are the main differ-
ences from the viewpoint of the driver?
18.2 What effect does the quarter-wave transformer have on impedances?
18.3 The rear radiation impedance will change if the (open) transmission line is
tapered. The primary effect is that the quarter-wave resonance frequency is
changed. Why and in what way?
18.4 Why is it desirable to use porous sound-absorbing material in a transmission
line loudspeaker duct?
18.5 How will duct sound-absorbent lining affect the sound transmission through
the duct?
18.6 How will duct folds and bends affect the sound transmission through the duct?
PROBLEMS
18.1 The analog lumped element acoustic impedance circuit for an open-end trans-
mission line loudspeaker was shown in this chapter. Assume a straight duct
that has a cross section area SD = 10 −2 m2 and a length of 2 m. The duct is
terminated by an acoustic resistance that is three times the duct characteristic
impedance.
Tasks:
a. Use a four-element lumped analogy to calculate the input impedance of the
duct f in the frequency range below which the duct is one wavelength long.
b. Compare the results in (a) to those obtained using Equation 18.3.
18.2 A tapered rectangular duct has a length of 2 m and its cross section area
SD varies from 4·10 −2 m2 at its entrance to 1·10 −2 m2 at its end. Its width is
constant w = 0.2 m.
Task:
Calculate the acoustic impedance seen from the entrance if the end is loaded
by the radiation impedance of a piston at the end of a long tube. Only study
frequencies below those where the tube is λ long.
18.3 A straight circular tube that has a cross section area SD = 1·10 −2 m2 and a
length lD = 2 m. It is terminated at the close end by a short-circuited voice-coil
driver mounted in a box that has an internal volume VB = 0.1 m3. The driver
has the following data: Bl = 8 N/A, REC = 7 Ω, CMS = 1·10 −3 m/N, R MS = 2 kg/s,
MMD = 20·10 −3 kg, and SD = 1·10 −2 m2. The far end of the tube is loaded by the
radiation impedance of a piston.
Task:
Calculate approximately the two lowest resonance frequencies of the system
(a graphic solution is sufficient).
18.4 A voice coil driver with the following data is mounted at the end of the duct:
Bl = 8 N/A, REC = 7 Ω, CMS = 1·10 −3 m/N, R MS = 2 kg/s, MMD = 20·10 −3 kg, and
SD = 1·10 −2 m2. The driver is mounted at one end of a straight duct that has a
cross section area SD = 1·10 −2 m2 and a length of 2 m. The far end of the duct is
terminated by the radiation impedance of a piston in a long tube.
Tasks:
a. Calculate the shift in resonance frequency of the driver that results from
the duct’s acoustic load.
b. Calculate the power radiated by the driver cone at the lowest resonance
frequency if the loudspeaker is fed a voltage ~ e = 1 V if the loudspeaker
is radiating into “free space.” Neglect the influence of the external field
on the driver.
c. Calculate the peak displacement x̂ at the open end of the duct at the
lowest resonance frequency. Also calculate the driver’s peak cone
displacement.
REFERENCES
1. Beranek, L. L. (Ed.), Noise and Vibration Control, Institute of Noise Control Engineering,
Washington, DC (1988) ISBN: 0-962007209.
2. Ingard, U., Noise Reduction Analysis, Jones and Bartlett Publishers, Boston, MA (2009)
ISBN-13: 978–1934015315.
3. Mechel, F. P. et al., Formulas of Acoustics, 2nd edn., Springer, Berlin, Germany (2008),
ISBN-13: 978-3540768340.
5. Tyrland, S., Construction of a Monitor Loudspeaker (in Swedish), Report 74–35,
Department of Applied Acoustics, Chalmers University of Technology, Göteborg,
Sweden (1974).
6. Beranek, L. L. and Sleeper, H. P., The design and construction of anechoic sound
chambers, J. Acoust. Soc. Am., 18(1), 140–150 (1946).
19 Horns
19.1 INTRODUCTION
Horns are used in electroacoustic systems for many reasons. Some advantages of
using horns in loudspeaker systems are:
• Enhancement of electroacoustic conversion efficiency

• Avoidance of sound coloration due to the acoustic resonance of loudspeaker
boxes
• Better control over the directivity
• Reduction of transducer-generated nonlinear distortion
Horns can be considered a class of transmission line loudspeakers (or vice versa).
Whereas in the conventional transmission line loudspeakers discussed in the previ-
ous chapter, there were the practical alternatives of resistance or quarter-wave trans-
former loading, the first is inefficient and mainly used for mid- and high-frequency
applications, the latter is used only over a narrow frequency range, usually in the
low-frequency range where the long decay time of the resonant duct can be accepted
from the viewpoint of hearing and room acoustics.
Transmission lines are used for microphone probes, but horn coupling is often
used for more wide band and less resonant properties. Additionally, passive horns are
sometimes used for medical, speech, and hearing purposes. Horns are ubiquitous in
music although most horns used in trumpets, trombones, and other brass instruments
are better regarded as resonant transmission line devices that have a short horn sec-
tion attached for impedance matching purposes.
“Reversed” resonant horns are often used in ultrasonic and other non-audio elec-
troacoustic systems and are then called concentrators. Examples include the transfer
of energy from an often large exciter to the small work area needed for milling and
welding purposes and for power matching and maximum energy transfer in medical
and industrial systems.
19.2 HORN EQUATIONS

19.2.1 Horn Terminology
When a wave propagating in a tube meets with an abrupt change in acoustic
impedance, part of its energy will be reflected back, as discussed in Chapter 7.
The interference between the forward and backward propagating waves results in
a change in acoustic impedance from the characteristic impedance of an infinite
tube Z AT = Z 0/S. The simple quarter- and half-wave tube impedance transformers can
be thought of as extreme horns that only function at certain frequencies. Horns are
367
Mouth (bell)
Neck
Throat
ZAT S (x)
ST
x Parabolic
Conical
SM Exponential
Length = l Hyperbolic
(a) x = xT x = xM (b)
FIGURE 19.1 (a) Horn terminology: the throat area is ST, the mouth area is SM. The cross
section area of the horn at coordinate x is S(x). The acoustic impedance at the throat looking
in the positive direction of the horn is Z AT. (b) Common cylindrical horn profiles.
however characterized by a certain taper, i.e., expansion function and length that
give the horn duct or tube a continuously, smoothly varying cross section.
Horns can be regarded as couplers used to change the relationship between sound
pressure and particle velocity at different places in acoustic systems. These properties
will determine the major acoustical characteristics of the horn as an impedance
transformer. Typically, the horn is used to adapt the low acoustic impedance of an
air load to the high acoustic impedance of a mechanical system.
A finite horn will start expanding at the “throat” and end at the “mouth” as
shown in Figure 19.1. Generally, in audio engineering, the horn transducer/driver
will be attached to the throat and the sound radiation of the horn will be from the
mouth. The flare, the geometry of the mouth, the phase, and direction of the particle
velocity at the mouth will determine the radiation properties and the directivity of
the horn. In practice, however, one often resorts to the assumption of the radiation
impedance seen by the mouth being that of a piston in a baffle or at the end of a tube,
as appropriate. This is obviously unsatisfactory for large curved horn mouths since
there the assumption of plane waves propagating in the horn is clearly false, and a
better approach is to use numerical modeling, for example, boundary-element (BE)
modeling of both the horn and its environment.
The taper of the horn is specified by the cross section area expansion rate. The
cross section area S(z), usually determined by some equation, is a function of distance
from the throat of the horn along its axis. Since low-frequency-range horns tend to
become long and bulky, it is common to fold, coil, or bend these horns. Since this
may considerably alter the horn’s acoustic properties in the same way as discussed
for transmission line loudspeakers in Chapter 18, the linear theory presented next
should only be regarded as a guideline to horn properties. In practice, straight horns
are used only in some musical instruments and in some loudspeakers for medium,
high, and ultrasonic frequencies.
19.2.2 Webster’s Horn Equation

In Webster’s simple horn theory, the horn is assumed to be straight, i.e., the cross
section surface S(z) is only a function of values along the x-axis. The horn is assumed
Horns 369
to carry only plane waves, although there may be waves propagating in both
directions along the horn axis. The sound pressures in the horn are also assumed
to be low enough for propagation to be linear. The horn is often assumed to have a
circular cross section in a plane perpendicular to the horn axis.
Webster’s horn equation 19.1 is a slightly modified wave equation used to model
the sound propagation in tubes having a monotonically, slowly changing cross
section S(z). The derivation of Webster’s horn equation may be found in Refs. [1–5].
∂2 p ∂p
∂z 2
+
∂
( ( ))
∂z
ln S ( z )
∂z
+ k 2 p = 0 (19.1)
Many horn profiles have been suggested in the literature [6–10]. Most of these are
based on the use of Webster’s horn equation. In recent literature, also other types have
been suggested such as tractrix horn, which differs by its assumption of spherical
waves close to the mouth of the horn, where the rapid flare makes the assumption of
plane wave propagation incorrect [9,10]. Folded horns have been studied extensively
as well [11,12].
Of course, numerical analysis of horn properties using finite- or boundary-
element analysis is an ideal way of studying horn properties but is outside the scope
of this book [13,14]. Using this methodology, the electroacoustic transducer can be
included in the analysis as well.
19.2.3 Common Horn Expansion Functions

Three aspects of horn performance are critically determined by the choice of horn
expansion function:
• Transfer function (far-field pressure for constant input volume velocity)

• Acoustic impedance at the throat
• Nonlinear distortion
The directivity, phase distortion, and resonance characteristics can be regarded as

secondary properties and are determined by the length of the horn and the shape of the
bell opening. Some traditional common horn cross section area functions are given in
the following. The expansion profiles of some of these horns are shown in Figure 19.1b.
Parabolic [3]
S ( z ) = ST z (19.2)

Conical [3]
S ( z ) = ST z 2 (19.3)

Exponential [1]
S ( z ) = ST e mz (19.4)

Hyperbolic [6]
( )
2
S ( z ) = ST cosh ( N H z ) + TH sinh ( N H z ) (19.5)

In these equations, S(z) is the cross section area along the horn, ST is the throat
area, m is the horn expansion rate (flare rate) for exponential horns, N is the flare
rate for hyperbolic horns. For hyperbolic horns, T H is the horn family parameter
[6]. Hyperbolic horns have T H < 1, the catenoidal horn T H = 0, and the exponential
horn T H = 1.
The start coordinate is z = 0 for exponential horns, and some value z0 for horns
based on other expansion functions such as conical and hyperbolic horns. These
common horn shapes are shown in Figure 19.1b.
The reason for choosing exponential horns for audio is usually their good
combination of features such as their bandwidth, relatively low nonlinear distortion,
and suitable throat impedance characteristics (Table 19.1).
For many purposes, though, other horns may be better. Conical and parabolic
horns and their properties are discussed in Ref. [3] and will not be considered here
since they are seldom used in audio applications, although conical are sometimes
used because of their noncritical design. Stepped, Gaussian, and Bezier curve horns
are common in solid horns for ultrasonic welding applications; these should be more
appropriately considered as power concentrators than proper electroacoustic horns.
Finally, it is worth noting that the impedance conversion can also be achieved using
layers of progressively changing acoustic media.
Usually, the complex input impedance ZAT at the throat is of interest for transducer
matching, electroacoustic efficiency, and far-field sound pressure. In the plane wave
horn approximation, the sound waves at the mouth are assumed to be plane. The
radiation characteristics of the mouth are assumed to be those of a piston set in a long
tube or large baffle. The radiated power will then be the product of the real part of the
throat acoustic impedance and the mean square value of the volume velocity at the
throat. At low frequencies, the volume velocity at the throat will be approximately
that of the transducer diaphragm driving the horn at its throat. The far-field sound
TABLE 19.1
Positive and Negative Aspects of Several Commonly Used Horn
Profiles in Electroacoustics
Horn Profile Positive Negative
Parabolic Easy to design and fabricate Poor impedance conversion
Conical Easy to design and fabricate Poor impedance conversion
Exponential Good wide band impedance conversion Some nonlinearity
Hyperbolic Very good and high impedance conversion Relatively nonlinear
Stepped High impedance conversion Depends on step resolution
Horns 371
pressure is then calculated from the equivalent piston’s geometric properties and
the total sound power radiated. Finding the complex throat impedance is of great
importance to be able to couple the horn to a suitable transducer and to predict the
resulting sound pressure in the far field.
19.3 EXPONENTIAL HORN

19.3.1 Wave Propagation and Cutoff
Many horns are characterized by an exponential taper since the exponential horn
usually gives a good mix of horn advantages and disadvantages. Because of its
attractive impedance properties, the exponential horn is often used in practice as the
starting point of horn design for both audio and ultrasonic horns.
Applying the cross section function in expression 19.4 to Webster’s horn equation,
we find the horn equation for the wave propagation in an exponential horn:
∂2 p ∂p
+m + k 2 p = 0 (19.6)
∂z 2
∂z
The solution to the horn equation representing waves progressing in the positive
z-direction will be
p( z, k E ) = A+ e
− jk E z
(19.7)

Here kE is the wave number that depends on frequency and flare rate. Inserting this
solution into the horn equation, we obtain the equation for kE as a function of the
expansion constant m and the frequency (since k = ω/c):
k E2 + jmk E − k 2 = 0 (19.8)

This quadratic equation has the solution
m 4k 2 − m 2
kE = +j (19.9)
2 2
The sound pressure of waves propagating in the positive z-direction along the horn
will be described by the expression
m 4 k 2 − m2
− z − jz
p( z, k ) = A+ e 2 e 2
(19.10)

For frequencies so high that
4k 2 > m 2 (19.11)
it will be possible for waves to propagate and transport power along the horn. The
frequency above which sound wave propagation is possible is called the cutoff
frequency fc:
mc (19.12)
fc =
4π
For lower frequencies, there will only be a reactive near-field close to the driving
point where the transducer is mounted. Note that the theory does not apply at zero
frequency. It is of course possible to blow a static flow into the horn since an infinite
horn has infinite volume.
19.3.2 Throat Impedance
Using the equation of motion for one-dimensional waves
j ∂p (19.13)
u ( z, ω) =
ωρ ∂z
we can calculate the volume velocity U(z) over each cross section. Inserting the
expression for p̲ (z), we obtain
 4ω 2 − ( mc )
2 
S m
U ( z, ω ) = +j  p ( z, ω ) (19.14)
jωρ  2 2c 
 

This leads to the following expression for the complex acoustic impedance Z AT:
Z AT ( ω, m ) = ℜ AT ( ω, m ) + jX AT ( ω, m ) (19.15)

The acoustic impedance at the throat, looking in the positive x-direction, will be
Z AT ( ω, m ) =
ρc
1−
( mc ) + j ρc2m (19.16)
ST 4ω 2 2ωST
The expression for Z AT can be rewritten as
ρc  ω 
2
ω 
Z AT ( ω, ω c ) = ℜ AT + jX AT = 1 −  c  + j c  (19.17)
ST   ω ω
 
For frequencies below fc, the impedance Z AT becomes reactive but with a positive
sign. The frequency dependence of the real and imaginary parts of the radiation
impedance of an infinite exponential horn is shown graphically in Figure 19.2.
Studying Equation 19.16, we note that for frequencies above the cutoff frequency
ωc = 2πfc, the reactance behaves as a negative compliance X AT = 1/jωCAT. This property
Horns 373
1.0
ZAT ST
Re
ρc
Im
0.8
Normalized throat impedance
0.6
0.4
0.2
0.0
0.1 0.2 0.5 1 2 5 10
Normalized frequency ω/ωc
FIGURE 19.2 The real and imaginary parts of the normalized acoustic impedance at the
throat of an infinite exponential horn.
is often used in the design of horns having a box or other closed volume facing the rear
of the driver’s diaphragm. It may be advantageous to balance the negative compliance
of the horn to the positive compliance of the enclosed air behind the transducer to
increase output at low frequencies. This is called reactance cancellation or annulling
and is often used for low-frequency horns that use conventional electrodynamic
loudspeakers as drivers.
19.4 CONICAL HORNS

The conical horn is simple to manufacture since it does not have double curvature.
The wave equation for the conical horn is
∂2 p 2 ∂p
+ + k 2 p = 0 (19.18)
∂z 2 z ∂z
and the complex acoustic impedance Z AT at the throat of an infinite conical horn is
ρc k 2 zT2 ρc kzT
Z AT ( ω, zT ) = +j (19.19)
ST 1 + k 2 zT2 ST 1 + k 2 zT2
Here zT is the distance of the horn throat from the cone’s apex and ST the throat area.
We note that Z AT is composed of smoothly varying functions, that there is no cutoff
frequency, and the horn has a real impedance that extends down to zero frequency.
Compared to the exponential horn, the real part of Z AT is much smaller over a wider
frequency range as shown in Figure 19.3. This is a result of the horn “neck” being
wide compared to that of an exponential horn.
1.2
AT ST
ρc
1
Hyp
Normalized throat resistance
Exp
0.8 Con
Par
0.6
0.4
0.2
0
100 200 500 1k 2k 5k 10 k
Frequency [Hz]
FIGURE 19.3 Comparison of the real part of the throat impedance of some infinite horns.
Here the hyperbolic horn has TH = 0.65.
19.5 HYPERBOLIC HORNS

When analyzing the impedance properties of the infinite conical horn, we noted
that the real part of its acoustic impedance was fairly low, much smaller than that
of the exponential horn. It also was much more frequency dependent, which means that
the frequency response of the horn will not be flat. The hyperbolic horn can be used to
overcome these issues and also those of the exponential horn. A characteristic of the
infinite hyperbolic horn is its fairly flat (ideal) acoustic impedance. The wave equation
of the hyperbolic horn may be derived using Equation 19.1. The complex acoustic
impedance Z AT at the throat for frequencies above the cutoff frequency ωc is [3]
1 T
1−
ρc k 2 zT2 ρc kzT
Z AT ( ω, zT ) = +j (19.20)
ST 1−T2 ST 1−T2
1− 2 2 1− 2 2
k zT k zT
Here T is the horn family parameter, zT is the position of the horn throat, and ST is the
throat area. For T < 1, the hyperbolic horn has a throat resistance larger than that of
the exponential horn. The cutoff frequency is given by
kzT =1 (19.21)

Figure 19.4 shows that the behavior of the acoustic impedance close to the cutoff f requency
depends on the choice of T. For a maximally flat response, T should be about 0.6.
19.6 COMPARISON OF HORN CHARACTERISTICS

For the infinite hyperbolic and exponential horns, there is a cutoff frequency below
which the impedance is purely reactive and there can be no propagating wave in
Horns 375
2
TH = 0.3
1.8
AT ST
TH = 0.5
ρc
TH = 0.65
1.6 TH = 1
Normalized throat resistance 1.4 TH = 2
TH = 4
1.2
1
0.8
0.6
0.4
0.2
0
1 1.5 2 3 4 5 6 7 8 9 10
Normalized frequency ω/ωc
FIGURE 19.4 Frequency dependence of the normalized throat resistance of hyperbolic

horns for various values of the horn family parameter TH. Hyperbolic horns have TH < 1 and
the exponential horn TH = 1.
the horn. Compared to an exponential horn, the ℜAT of the hyperbolic horn is more
constant over a wide frequency range as shown in Figure 19.5. This advantage leads
however to more nonlinear distortion and dispersion. These drawbacks are common
to all horns that have a long neck and short mouth.
It is also instructive to study the relative level of the high-frequency normalized real
part of the radiation impedance as shown in Figure 19.6. We note that with the throat’s
acoustic resistance ℜAT expressed in level L = 10·log(ℜAT), the linearity of frequency
response of the (infinite) horns is quite good, certainly much more linear than that of
1.2
XAT ST
ρc
1
Hyp
Exp
0.8 Con
Par
0.6
0.4
0.2
0
100 200 500 1k 2k 5k 10 k
Frequency [Hz]
FIGURE 19.5 Comparison of the imaginary part of the throat impedance of some infinite
horns. Here the hyperbolic horn has TH = 0.65.
0
Normalized throat resistance level
–5
ST
ρc
AT
–10
10 Log
Hyp
–15 Exp
Con
Par
–20
–25
100 200 500 1k 2k 5k 10 k
Frequency [Hz]
FIGURE 19.6 Comparison of the level of normalized throat resistance for some infinite
horns. Here the hyperbolic horn has TH = 0.65.
other horn types and many other loudspeaker target designs. Note that the reactance
of the horn will also influence the response, but this was not included here. Also the
addition of the driver and electronics will affect the characteristics as shown later.
19.7 TRACTRIX HORNS

The theory presented so far has assumed that the flare rate does not change quickly in
the mouth region of the horn. An alternative horn type, the tractrix horn, based on the
tractrix curve needs an alternative theory. The tractrix curve following the coordinates
zTH and yTH in the zy-plane is based on the pair of parametric equations [10]
 υ
zTH = lT tanh   − υ
 lTH 
(19.22)
 υ
yTH = lT sech  
 lTH 

The mouth of the horn is at z = 0. The arm has a constant length lTH and takes its start
at z = –υ. The slope of the tractrix curve is generated by the end of the arm. Moving the
start point in the positive z direction then generates the tractrix horn curve as shown in
Figure 19.7. Because the horn flare changes along the length of the horn, being expo-
nential at the throat but infinite at the mouth, Webster’s horn equation does not hold.
Depending on the choice of lTH value, the tractrix horn can be shaped to closely
follow that of an exponential horn or to have a more pronounced bell. Two important
characteristics of tractrix horns are that the flare rate of the neck is quite high and
that the flare rate at the mouth is usually much higher than that in other horns. This
can be expected to lead to interesting properties different from those of other horns.
The higher flare rate in the neck region means that the throat impedance will be
Horns 377
lTH
z
0
lTH
υ
FIGURE 19.7 One half section of an axially symmetric tractrix horn. The tractrix curve is
generated by moving a tangent line of length lTH with its one end at υ on the z-axis. The z-axis
is the axis of rotation. Wavefronts assumed spherical.
smaller than that of the exponential horn in the frequency region immediately above
the cutoff frequency. The high flare rate at the horn mouth leads to less reflections
and a smoother characteristic of the throat impedance than that of a “cut” expo-
nential horn. Since there is less dispersion because of the smaller compression, the
resonances will be placed more evenly in frequency. Finally, the directivity of the
tractrix is between that of a piston and that of a spherical source [11].
19.8 FINITE HORNS

At frequencies above its cutoff frequency, the infinite horn will present an essentially
resistive load on the loudspeaker diaphragm. All real horns however need to be finite.
In finite horns as well as in finite tubes and ducts, there will be sound reflected from
the ends. The effect of the reflections will show up as resonance peaks in the throat
impedance characteristic.
The Q factors of these resonances will depend on the impedances of the bounding
surfaces. Normally, the horn is rigid and the driver has a fairly high impedance. The
Q factor at any resonance will thus depend mainly on the acoustic impedance seen
at the mouth and thus on the size of the mouth compared to wavelength. When the
mouth size is large compared to wavelength, the radiation impedance will be almost
real, and there will be little reflected sound and no resonance. The frequency at
which resonances stop being of importance is thus determined by the horns’ mouth
geometry. Using the equations for the sound pressures and particle velocities of for-
ward and backward propagating plane waves in the horn, one can find the acoustic
throat impedance Z AT for the finite horn. The throat impedance of a finite-length
conical section was already given in Equation 18.2.
For a finite exponential horn of length l, the reflection coefficient is found from
the impedance mismatch between the local horn impedance ρc/S(l) at z = l and the
acoustic load impedance at the mouth Z AM. The acoustical throat impedance Z AT is [3]
ρc  S M Z AM cos ( bl + θ ) + jρc sin ( bl ) 

Z AT = (19.23)
ST  jSM Z AM sin ( bl ) + ρc cos ( bl − θ ) 

Here
a = m/2
b = k 2 − a2 (19.24)
θ = arctan(a /b)
Figure 19.8 shows the impedance characteristics of a finite cylindrical symmetry

exponential horn with horn mouth size as parameter. The data shown are particu-
larly interesting because they show how the resonances in an exponential become
less pronounced as the mouth diameter is increased. It is only when the horn mouth
circumference becomes larger than the wavelength that the impedance characteristic
becomes reasonably smooth. This is to be expected from the behavior of the radia-
tion impedance for a piston set in an infinite baffle, because in case (d) the condition
is ka = 1 at 100 Hz. As the horn mouth diameter increases, so must the horn length to
keep the throat area constant. The increased length results in a reduction of the horn
resonance frequencies.
Although the fluctuations in the real part of the exponential horn’s throat
impedance shown in Figure 19.8 may seem large, they are quite small expressed
in decibel form. Even the impedance oscillations at medium horn mouth diameters
can be considered quite small when compared to the low-frequency response
in a small listening room. If one assumes that a cutoff frequency of 40 Hz is
sufficient, then 0.3 λc = 2.65 m diameter. This may be quite acceptable if the horn
is mounted so that its mouth is in a corner of three surfaces, since then, because of
the mirroring, the horn will have an equivalent surface area four times that shown
and the effective diameter is only 0.67 m. It is also important to keep in mind that
many low-frequency horns are used in small rooms that have high acoustic input
impedance at low frequencies, which will also affect the throat impedance. When
studying the electrical input impedance of horns driven by electrodynamic drivers,
one can often see the influence of room resonances as peaks in the electrical
impedance magnitude curve.
There are many ways in which the resonances can be controlled. The most
obvious is to make the horn mouth so large (and horn so long) that the ripple in
the impedance characteristic becomes acceptably low. One can also make the horn
mouth serrated or suitably modulated so that the reflections become more diffuse.
A third possibility would be to use resonators or openings along the side of the
horn [16]. Nonlinear effects make the latter alternative less attractive; at least the
resonators cannot be placed close to the horn throat. A fourth alternative is to use a
filter ahead of the horn mouth [17].
For hyperbolic horns that have TH in the range 0.5–1, the throat impedances
are similar to those of exponential horns [23]. The flare rate of a hyperbolic horn is
given by NH
2πfC
NH = k = (19.25)

f = fC
c
Horns 379
Horn mouth diameter

in λC
0.15 0.10
0.30 0.25 0.20
8
4
AT ST
ρc
0
2000
1500
1000
Frequency [Hz] 500
Horn mouth diameter

in λC 0.15 0.10
0.30 0.25 0.20
4
Normalized throat reactance
2
XAT ST
ρc
–2
–4
2000
1500
1000
Frequency [Hz] 500
FIGURE 19.8 Examples of the normalized throat impedance of a finite exponential horn
for which the horn mouth diameter has been varied. The horn throat diameter and cutoff
frequency, 327 Hz, are constant. The wavelength at the cutoff frequency is λc. The horn is
assumed terminated by an impedance corresponding to that of a piston in an infinite baffle
and to radiate into a solid angle of 2π.
The length lh is determined by the flare rate NH and the ratio D2 of mouth and throat
areas SM and ST.
SM
D=
ST
lh =
1
loge 
(
 D + D2 − 1 − T 2
H )  (19.26)
NH  1 + TH 
 

The equation for the throat impedances of finite hyperbolic horns for various TH
values can be found in Refs. [19.25].
19.9 HORN DIRECTIVITY

Achieving better directivity control has long been an aim of horn designers.
Horns used for sound reinforcement purposes need to have frequency-independent
directivity and must be arrayable. There are many types of arrays: two examples are
clusters and line arrays. It is important to note that the directivity of horns does not
follow the intuitive visual appearance, neither does the acoustic center.
Since it is quite feasible to model horns using BE and FE software [15], the
approaches outlined here can be regarded as simple approximations. Combinations
of numerical methods need to be used since modeling the far-field sound field would
require prohibitively large FE and BE models. An advantage of FE modeling is that
for the case of low-frequency horns, the horn and the room may be modeled together
for optimum results.
19.9.1 Horns Using Combinations of Flares

By making the horn a combination of different horn profiles as shown in Figure 19.9,
the directivity can be better controlled. This horn uses conical sections to provide the
desired characteristics. One usually strives for constant directivity in both the vertical
and horizontal directions. This is generally not possible, and at low frequencies, any
horn will become more omnidirectional. Since horns using a combination of different
horn profiles often combine three or more different flares, it cannot be helped, but
there will be breaks in the flare. These breaks cause diffraction, reflection, and
resonance.
19.9.2 Multicell Horns

Many types of horns have been suggested to deal with low-frequency driver loading,
resonance character, and directivity. The earliest approach to directivity issues
were multicell horns such as the one shown in Figure 19.10. These use a number of
symmetrical, wide directivity, usually, exponential horns, where the mouth openings
of the cells are combined into a spherical array “cap”, driven by a single driver.
Horns 381
Top view
Front view
Side view
FIGURE 19.9 The “manta-ray” group of horns are designed using conical sections to
provide more frequency-independent directivity properties. (After Henricksen, C.A., and
Ureda, M.S., J. Audio Eng. Soc., 26(9), 629, 1978.)
FIGURE 19.10 A 2 × 4 multicell horn with a curved front.
By eliminating the possibility for higher-order modes in the horn and providing
a number of essential point sources on a spherical cap, multicell horns address
important issues in horn design.
The throat impedance will be that of the individual throat impedances in parallel
and the individual horn profiles are all part of the “full” horn. Depending on the horn
profile used for the “full” horn, the low-frequency loading near the cutoff frequency
can be larger. Because of the assembly of sources on the spherical cap section, the
directivity is almost omnidirectional at low frequencies, and at high frequencies the
radiation is determined by the solid angles determining the complete loudspeaker
sector. At frequencies where there is negative interference between the cell horns in
the middle and those in the periphery, there will be a dip in response as shown in
Figure 19.11. The depth of the dip will of course depend also on the directivity of the
150
6 cells
130
110 5 cells
Beam width [°]
90
4 cells
70
3 cells
50
2 cells
30
100 200 500 1k 2k 5k 10 k
Frequency [Hz]
FIGURE 19.11 Beam width between −6 dB points of a set of multicell horns. Each horn
has a 25° arc and mouth width of 0.2 m. (Adapted from Beranek, L.L., Acoustics, American
Institute of Physics, New York, 1986.)
individual cells. Other drawbacks of multicell horns are cost and weight due to the
complexity of construction.
19.9.3 Radial Horns
Radial horns can be considered a type of cylindrical symmetry sector horn. Radial horns
are designed to have a small “vertical” angle and a large “horizontal” angle and are often
used in mid- and high-frequency-range horns for home and studio use. The horn profile
is shaped along the vertical angle sector. Radial horns may contain vanes placed radially
along the horizontal direction for pattern control. When high dispersion along one plane
is necessary, the vertical angle can be made very small. Such horns are sometimes called
diffraction horns. An example of a 80° radial horn is shown in Figure 19.12.
19.9.4 Waveguide Horns

A considerable improvement in horn directivity can be obtained by using the so-called
waveguide approach to horn design. The waveguide horn can be considered a version
of the multicell horn but with path length equalization. Path length devices such as
acoustic lenses were discussed in Chapter 12. Waveguide horns can be considered
horns with a built-in acoustic lens. The objective of the path length equalization is to
obtain a linear array of cell horn openings. Since the middle paths are shorter than
the outside paths of a typical horn, the middle paths need added length to give a
smooth isophase wavefront at the horn mouth.
19.9.5 Horn Arrays
The multicellular horn is only used for mid- and high-frequency sound reproduction
and can be considered a spherical cap array. Larger horns for low frequencies need to be
Horns 383
Top view
Front view
Side view
FIGURE 19.12 A radial horn.
Wavefront “quiet spots” Wavefront curvature error
Wavefront
Horn
profile
FIGURE 19.13 An array of three horns. Conventional horns will not produce a linear and
equal magnitude wavefront even when fed by identical drivers. A real array composed of
individual loudspeakers can be considered as a line source with an overlaid negative array to
compensate for quiet spots in the real array.
arrayable. In arraying horns, it is important to note that the directivity of the array does not
exhibit any sharp interference nulls. Some designs are better than others in this respect.
The wavefront location for each horn is critical since according to Huygen’s principle,
the wavefront sources generate the next wavefront. If the progressive wavefronts of the
array built by the individual horns are to be smooth and reach the intended area without
interference dips, then the wavefronts at the horn mouths should be approximately plane
as indicated in Figure 19.13. The wavefronts however cannot be plane as discussed
earlier. There will always be some wavefront curvature, even if the horn is narrow.
A conservative requirement for the phase error is that it should be less than λ/8 at
any mouth point and frequency in the operating range; as an example, this leads to a
wavefront phase linearity requirement of ±5 mm at 10 kHz.
In addition, the horn walls and mounting gear take up some space so the array
will have “quiet spots” as indicated in Figure 19.13. The effect of these spots can be
Height [m]
20 J-shaped horn array
10
g slope
Seatin
0
0 10 20 30 40 50 60 70 80 90 100
Distance [m]
FIGURE 19.14 A J-shaped variable curvature array used to cover a wide area. The array is
shaped (and shaded) to provide constant intensity over the seating area.
estimated by assuming the array a perfect limited length line source and then add-
ing to that sound pressure the pressure from an additional array that has the same
strength per unit length but which is limited to the blind spots and is out-of-phase
with the line source. To avoid aliasing in the calculation of the resulting sound field
using spatial sampling, the array grid should have sampling distances smaller than
λ/2; however, λ/6 is recommended.
It is important to realize that because of the size of wide band arrays used in
sound reinforcement, the far-field limit may be very far away. Consider an array
such as the J-shaped array used in Figure 19.14. A typical estimate of the distance
to the start of the far-field region is l2/λ where l is the length of the array. For a 5 m
long array, this leads to a far-field limit of about 150 m at 2 kHz, so many listeners
will be inside the near-field region of the array with its associated interference dips.
For large distances, it is also necessary to take the attenuation by air humidity and
turbulence into account as well as the effects of wind and temperature gradients.
Due to the uncertainties of these effects, it is usually not necessary to calculate the
resulting sound field very exactly.
It is clearly necessary to use an optimization technique to determine the ampli-
tude and delay shading necessary to give the desired coverage over the listening area.
19.10 HORN AND DRIVER

Either or both sides of the driver diaphragm can be used to drive horn(s).
Loudspeakers that use both sides of the diaphragm are sometimes called compound
horn loudspeakers.
19.10.1 Low-Frequency Horns
The low-frequency horn shown in Figure 19.15 has the horn attached to the front of
the loudspeaker. Horns for mid and high frequencies usually use a diaphragm that
sees a closed box at its back. The compliance CAB of the air in this box is often used
to compensate the negative compliance of the horn throat CAT at frequencies over the
cutoff. Compound horns sometimes use mid-frequency-range horn loading on the
front of the diaphragm and a ported box loading on the back.
Horns 385
FIGURE 19.15 The “Klipschorn” is a low-frequency horn that makes use of a room corner
to continue the horn flare. (From Klipsch, P.W., J. Acoust. Soc. Am., 13(2), 137, 1941.)
19.11 HIGHER-ORDER MODES IN HORNS

Any horn is a form of duct, and as in any duct, there can be other modes than the
plane wave mode. These higher-order modes can be excited by at least four mecha-
nisms: driver asymmetry (mounting and movement), horn asymmetry, horn bends
and curvature, and mouth or load asymmetry. The higher-order modes of course
mostly occur at high frequencies, since the horn mouth region needs to be large
to avoid sound to reflect. The phase and magnitude properties of these modes will
affect the sound radiation pattern of the horn.
In any case, the modes will be difficult to diagnose directly unless measured
using some kind of sound pressure sensing probe or miniature microphone. It is
easier to see the effect of such modes in the far field of the horn, for example, in the
polar diagram.
When horns are designed using numerical approaches, such as the BE or FE
methods, the higher-order modes can be easily noted, diagnosed, and eliminated.
Since the modes are a result of the horn being too wide, one simply needs to
subdivide the horn into a sum of smaller horns, for example, by inserting partitioning
walls similar to a multicell horn or by converting to an equivalent-sector horn. The
approach will usually be the same whether the horn has a rectangular or circular
cross section.
19.12 CIRCUIT ANALOGIES FOR HORN LOUDSPEAKERS

As was seen from the previous sections, much of the work on horn loudspeakers has
been focused on achieving a frequency-independent throat impedance and, by way
i RE LEC Bl : 1 FC MMD CMS CMB rMS rMB CMF ST : S D FT
eG e uC uT zMT
FIGURE 19.16 The mechanical mobility analogy for a front-loaded horn loudspeaker
having a closed boxed rear acoustic load.
of this, a flat frequency response. The reactive part of the throat impedance is also of
interest as it will affect the response of the loudspeaker and certainly the electrical
input impedance of the driver.
Figure 19.16 shows the analog mechanical mobility circuit for a horn loudspeaker
using a closed box for the back radiation of the diaphragm. The compliance of the air
behind the driver diaphragm has compliance CMB. It is usually not possible to avoid
this undesired small air volume between the driver cone and the horn throat. The
compliance CMF of this air volume acts as a shunt and reduces the volume velocity
available to drive the horn.
For a straight horn, the horn throat mobility MMT is resistive at high frequencies,
irrespective of horn shape function. For these frequencies, the circuit can be simpli-
fied as shown in the electrical impedance analogy in Figure 19.17. In practice, rMB
will have a very small value because the cavity behind the loudspeaker can be made
very loss free.
19.12.1 Efficiency
Obviously, for the highest possible efficiency, we want the current i1 to be small so
that the circuit will consist of only the electrical impedance RE and the transferred
horn mobility rMTB2l2. If the latter is equal to RE, there will be maximum power
transfer (i.e., 50% efficiency), and if it is higher than RE, the efficiency will also be
higher but at the same time there will be less sound power radiation. For the latter
condition, the horn throat area should be as small as possible, i.e., the horn should be
long, which is impractical and contributes to nonlinear distortion.
i RE
i1 iT
eG B2l2rMB B2l2rMT
FIGURE 19.17 A highly simplified electrical impedance analogy circuit for the horn
loudspeaker, shown in Figure 19.16, at high frequencies (ST = SD).
Horns 387
CMSCMB
i RE Bl : 1 FC CMB + CMS ST : SD FT
eG e uC uT zMT
FIGURE 19.18 A highly simplified mechanical mobility analogy of the horn loudspeaker at
frequencies slightly above the cutoff frequency.
19.12.2 Low Frequencies
Figure 19.18 shows a mechanical mobility analogy of the horn loudspeaker at
low frequencies slightly above the cutoff frequency of the horn. At these low
frequencies, it is possible to balance the combined compliance of the box volume
CMB and the driver surround CMS against the negative compliance of the throat
input impedance, making the mechano-acoustical circuit essentially resistance
controlled.
19.12.3 High Frequencies
At very high frequencies, close to the upper cutoff frequency of the horn loudspeaker,
the air volume formed between the diaphragm and the horn throat will provide a
decoupling inductance in the mechanical analogy shown in Figure 19.19. Depending
on the value of the various circuit components shown in the circuit, the air space
in front of the horn can be shown to provide a suitable resonance and sharper
high-frequency cutoff or can be a great disadvantage and limiting factor in horn
efficiency and high-frequency performance. In mid- and high-frequency-range horns,
a special insert is used to equalize the path length to the horn throat from all points
of the driver diaphragm. Such “compression” horn drivers often use a dome-shaped
concave diaphragm such as that shown in Figure 15.2.
i RE LEC B2l2CMF
MMD
eG B2l2rMT
B2l2
FIGURE 19.19 At very high frequencies, close to the upper cutoff frequency of the horn
loudspeaker, the air volume formed between the diaphragm and the horn throat will provide a
decoupling inductance that can be combined with the throat impedance negative compliance
to form a low-pass filter (ST = SD assumed).
19.13 STEPPED AND PIECEWISE LINEAR HORNS

The horns studied so far have been characterized by a smooth expansion function.
Intuitively, however, one understands that the horn can be made up of short coni-
cal sections. One can achieve horn-like properties by using a stepped horn profile
in the spirit of those shown in Figure 19.20 instead of a continuous profile. This
approach to horn design is quite common in electromagnetic engineering, particu-
larly for transition couplers. The finer the steps, the more broadband will the horn
be [19]. It is also possible to gradually or stepwise change the properties of the gas
that fills the horn.
Piecewise linear techniques are often used for low frequency horns such as the
model shown in Figure 19.15.
19.14 FOLDED, BENT, AND COILED HORNS

The length of a horn is essentially determined by the cutoff frequency and throat
impedance that are chosen for the horn design. Mid- and high-frequency-range
horns are usually sufficiently short to be used as straight horn designs for audio
and sound reinforcement purposes. Low-frequency horns, particularly those with a
cutoff frequency below a few hundred hertz, will often need to be designed curved
or folded.
An example of a folded horn is shown in Figure 19.15. When the wavelength of
sound is long compared to the cross section of the fold, there will be relatively little
attenuation due to the fold, but when the fold width/length becomes on the order of
half a wavelength as shown in Figure 18.3a, there will be considerable attenuation,
as discussed in Chapter 18. The high-frequency losses at the elbows can be reduced
by beveling the corners by inserting a 45° corner reflector. This is seldom necessary
since the audio bandwidth required of a low-frequency horn seldom is larger than
1:10, for example, 40 Hz to 400 Hz. Many custom-built low-frequency horns are
similar to radial horns, radiating into or out of a corner.
FIGURE 19.20 Principle of a stepped profile horn. (Based on Ragan, G. L., Microwave
Transmission Circuits, Dover, New York, 1965.)
Horns 389
FIGURE 19.21 Section of a reentrant horn for public address systems.
A common type of folded horn is the axial symmetry reentrant horn shown in
Figure 19.21. Because of the high-frequency attenuation of the bends and the low-
frequency response cutoff due to limited mouth size, reentrant horns sound quite
resonant. They are mainly used for the speech frequency range usually 400 Hz to
4 kHz for alarm and public address systems.
19.15 HORN PHASE PLUGS

Optimal coupling of the driver to the horn mouth is an essential part of horn design.
As outlined in Section 19.10, the bandwidth and efficiency of the horn depends on
the loudspeaker driver and horn combination. An important application of horns is in
various couplers. In most horn applications, high efficiency is the goal.
To obtain high horn transduction efficiency, it is usually necessary to use an
electrodynamic driver that has diaphragm area 2–10 times larger than the throat
area. Any air chamber between the horn driver diaphragm and the throat will act as
a high-frequency shunt because it will be easier to compress the air in the chamber
than to send it out through the constriction at the throat. Obviously, there must
be some form of air passages that lead the air from the diaphragm to the throat.
In high-frequency horns, these passages lead through a “phase plug” as shown in
Figure 19.22, using an “inverted dome” driver.
Of course, nothing prevents the use of phase plugs in low-frequency horns, but
the air chamber between the diaphragm and the horn throat serves as a part of an
acoustic low-pass filter. This “front” air chamber should not be confused with the
back air chamber that is used for nulling the negative compliance of the throat. The
horn loudspeaker requires the driver to have a free air resonance frequency at least an
octave above the cutoff frequency of the horn. Drivers used for low-frequency horns
ideally should have very low-mass, highly stiff diaphragms since the diaphragm mass
together with the air chamber volume is the limiting factor in the high-frequency
response.
Phase plug
FIGURE 19.22 The construction of a compression driver and its associated phase plug.
The horns discussed in this Chapter have used electrodynamic drivers; however,
many other types used are primarily piezoelectric transducers.
19.16 ACOUSTIC CENTER OF HORNS

The acoustic center may be defined on wavefront or time delay considerations as
discussed in Chapter 12. One definition of acoustic center is that it is the point from
which spherical acoustic waves seem to be diverging from the source, when observed
at far distance. At low frequencies, the horn mouth is usually quite small compared
to wavelength and the sound waves radiated have their center of curvature at the
horn mouth as if the mouth is a point source. At high frequencies, the waves seem
to emanate from some place far into the horn, eventually at the throat. The acoustic
source has not moved however only the apparent center of the radiated wave field. For
arraying purposes, the acoustic center is at the mouth since this is what determines
the radiated sound field as can be observed from, for example, interference nulls in
the array sound field [20].
Figure 19.23 shows an example of measured wavefronts close to a horn mouth.
The measured data show the low-frequency wavefront (which extends the farthest to
the side) has a different curvature than the center part of the wavefront, which has
relatively more high-frequency content.
120 mm
Horn mouth Horn axis

40 mm
200 mm
FIGURE 19.23 Wavefronts outside a short, high-frequency horn, showing that the acoustic
center is different for high and low frequencies. (From Nomoto, I. et al., J. Audio Eng. Soc.,
24(1), 9, 1976.)
Horns 391
19.17 LINEAR AND NONLINEAR DISTORTION

Horns are characterized by both linear and nonlinear distortion. Linear distortion as
uneven frequency response was covered in Section 19.10. It is important to remember
that there are also other forms of linear distortion caused by phase anomalies.
The group delay of the horn is not linear as discussed previously, which leads to
dispersion. This dispersion is usually not a problem for low-frequency loudspeakers
because of the relatively poor time resolution of hearing at low frequencies.
There are three main causes for nonlinear distortion in horn loudspeakers:
• Diaphragm compliance characteristic

• Magnetic-flux density and voice-coil characteristics
• Air nonlinearity
The diaphragm compliance, flux, and voice-coil characteristics are usually a small
problem thanks to the small voice-coil excursion in horn loudspeaker systems.
Because of the gas nonlinearity described by Poisson’s equation 3.3, there will be
an excess pressure (compared to the static ambient air pressure) in the air chamber
at the throat entrance as shown by Figure 19.24. This excess pressure is often called
the Rayleigh sound radiation pressure. This pressure will of course be dynamically
varying with the audio program material being played and will force the voice coil
away from its ideal center position. A suitable leak between the back and front air
chambers will eliminate this displacement. This displacement is a common problem
in small closed box loudspeaker systems.
A different air nonlinearity problem that increases with increasing frequency is due
to sound traveling faster at high than at low temperature. The pressure maxima will
move faster than the pressure minima of the wave, which leads to the creation of an
N-wave if the propagation takes place over a long distance. The distorted wave form
is of course associated with the generation of overtones in the sound in the throat,
because the sound pressure is very high in the constricted throat part of the horn and in
p(t) Resulting pressure change
Rayleigh sound
radiation pressure t
P0 Average
pvκ = constant pressure
with signal
v
v(t) Forced volume change
FIGURE 19.24 Rayleigh pressure generated radiation pressure.

the compression driver proper. The more wavelengths that are inside the high pressure
zone, the greater the distortion.
Horns having a hyperbolic cross section will feature less frequency dependence
and better low-frequency response than exponential horns, as shown in Figure 19.5.
This property is obtained at a price, however. The nonlinear distortion generated
by the horn is a function of the intensity of the wave as a function of path length
along the horn axis. The hyperbolic horn’s cross section area is narrow over a longer
distance than in exponential horns so the intensity will be higher over a longer path.
Equation 19.27 gives the second harmonic distortion percentage HD2 (%) in an
infinite exponential horn at standard conditions. We note that HD2 is a function of
intensity IT and the ratio of the signal frequency f to the cutoff frequency fc [2]. Since
the distortion is mainly generated at the narrow part of the horn close to the throat,
the equation will be also relevant for finite exponential horns.
f
HD2 = 1.73 ⋅ 10 −2 IT (19.27)
fc
In high-frequency compression drivers, the nonlinear distortion is primarily a result

of air compression. The amount of nonlinear distortion is critically dependent on
the driver diaphragm area and the volume between the diaphragm and the phase
plug. The larger the diaphragm area, the lower will the sound pressure be for a
certain intensity. For this reason, even horn drivers have fairly large voice coil and
diaphragm diameters, 10 cm being typical for high-quality drivers.
To control nonlinear distortion in a horn-loaded multi-speaker system one must
choose the crossover frequencies carefully. This is not as large a problem as one
might think but limits the frequency range of any single horn to a ratio of 1:10. The
mean spectrum density level in a music signal on the average drops off by 6 dB per
octave beyond about 500 Hz for classical symphonic music, but is fairly constant up
to that frequency. For popular music, the shape of the mean spectrum density level is
such that it drops by about 6 dB beyond about 50 Hz for popular music [22]. It makes
sense to avoid having the low-frequency horn reproduce the high-energy parts of
the spectrum around 500 Hz. The mid-range horn cutoff and crossover frequencies
would then be best located in the 200–400 Hz frequency region.
19.18 HORN-SHAPED CONNECTORS

The use of exponential horn duct transformations as impedance transformers was
discussed in Chapter 7. However, as with the other horns studied in this chapter,
the characteristics will depend on the flare types and rates. Any gradual area
transformation will act as an acoustic transformer over some frequency. Because
ducts have limited width, there will also be sound transmission below the cutoff
including the possible steady-state flow.
19.19 HORNS AND ROOM ACOUSTICS

An often neglected advantage of using horns in indoor environments is the large
sound absorption provided by the horn. As an approximation, one can regard the
Horns 393
horn mouth impedance as a “ρc-surface.” Because of linearity and reciprocity, any

transducer will also act as a sound absorber because of its losses.
Clearly, a conventional ported box enclosure will act as a Helmholtz resonator
(or rather like two coupled resonators) and a closed box enclosure as a Helmholtz
resonator. The absorption will depend on the tuning and the losses but also on the
electric load on the loudspeaker’s electric terminals (that are usually shorted by way
of the amplifier).
A low-frequency horn will correspondingly act as a much larger sound absorber
when it is placed at an “optimum” position in the corner of a room. Because of the
much tighter coupling between the driver and the room provided by the horn, one
can usually see the influence of the room modes in the frequency range served by
the horn. Because of the damping of the modes due to the horn, the low-frequency
sound reproduction may be much improved compared to that of a pair of small
conventional electrodynamic loudspeakers in the room, even if these have a lower
“cutoff” frequency than the horn.
19.20 SUMMARY
Some advantages of using a horn as a coupler to an electroacoustic transducer are
that it allows good control of the wavefront properties, the directivity characteristics
can be better controlled, and high acoustic power output can be obtained over a wide
frequency range using a single transducer. On the negative side, in reality all horns
will be characterized by resonant, dispersive, and nonlinear wave propagation. Many
of the reasons to use horns have been removed by the availability of inexpensive
power amplification and drivers. Remember that commercial electrodynamic drivers
may have large inter-unit variations in response, because of magnet properties and
unavoidable resonances in the diaphragm. From many viewpoints, possibly also
ecological, a horn is an attractive way to efficiently generate high sound power.
REVIEW QUESTIONS
19.1 In what way is a horn loudspeaker different from a transmission line
loudspeaker?
19.2 Which assumptions form the basis of Webster’s horn equation?
19.3 Which are the most common horn types for loudspeakers and what
distinguishes their properties?
19.4 What is meant by the cutoff frequency for a horn?
19.5 How do the throat impedances of finite and infinite horns differ?
19.6 Why is it advantageous to let the box compliance cancel the throat compliance
in horns that have a driver with a closed box back?
19.7 Which properties are necessary for a horn to be suitable for arraying?
19.8 Which factors need to be controlled in folding a horn?
19.9 What is the function of the phase plug in a mid- or high-frequency horn?
19.10 Where is the acoustic center of the sound field radiated by a horn?
19.11 Explain the reasons for the differences between the nonlinear characteristics
(distortion) resulting from various horn geometries.
PROBLEMS
19.1 A compound horn loudspeaker uses a voice-coil driver that is loaded by horns
on both sides of the diaphragm.
Task:
Draw the acoustic impedance analogy for a compound horn that drives a
low-frequency and a mid-frequency horn.
19.2 An electrodynamic loudspeaker driver is coupled to an exponential horn so that
a small cavity results between driver diaphragm and horn throat. The working
range of the horn is the frequency range in which resistance is dominating the
impedance of the impedance analogy of the driver and horn circuit.
Task:
a. Derive expressions for the radiated power and electroacoustic efficiency.
b. How should the throat area of the horn be chosen to maximize the radi-
ated power?
c. Determine the filter falloff rates at the frequency band limits.
19.3 A voice-coil driver that is to be used for driving a low-frequency exponential
horn has the following data: RE = 6.4 Ω, LE = 0.13 mH, Bl = 7.2 N/A,
MMD = 11·10 −3 kg, f0 = 65 Hz, QM = 5.1, and SD = 2.2 · 10 −2 m2. QM is the Q-value
of the mechanical system. The minimum volume between diaphragm and
horn throat is VDT = 8 · 10 −4 m3. The box that closes off the back of the driver
cone can be chosen freely.
Tasks:
a. Design an exponential horn having a lower cutoff frequency of 50 Hz,
using the driver.
b. What is the upper −3 dB frequency of this horn?
c. What will be the maximum cone displacement at f = 3fc when a voltage
of 10 V is applied to the driver terminals?
19.4 A voice-coil compression driver that is to be used for driving a mid-
frequency-range hyperbolic sector horn has the following data: R E = 8.5 Ω,
Bl = 18 N/A, QM = 10.3, MMD = 3.5 · 10 −3 kg, f 0 = 0 . 55 kHz, aT = 1.25 · 10 −2 m 2,
and SD = 8 · 10 −3 m 2. QM is the Q-value of the mechanical system.
Task:
a. Calculate the dimensions of a hyperbolic horn having a lower cutoff
frequency of 300 Hz, using this driver.
b. What is the efficiency in the mid-frequency range?
REFERENCES
1. Webster, A. G., Acoustic impedance and the theory of horns and of the phonograph,
Proc. Natl. Acad. Sci. U S A, 7(5), 275–282 (1919).
978–0883184943.
Horns 395

4. Mawardi, O. K., Generalized solutions of Webster’s horn theory, J. Acoust. Soc. Am.,
21(4), 323–330 (1949).
5. Eisner, E., Complete solutions of the “Webster” horn equation, J. Acoust. Soc. Am.,
41(4B), 1126–1146 (1967).
6. Salmon, V., A new family of horns, J. Acoust. Soc. Am., 17(3), 212–218 (1946).
7. Salmon, V., Generalized plane wave horn theory, J. Acoust. Soc. Am., 17(3), 199–211
(1946).
8. Thiessen, G. J., Resonance characteristics of a finite catenoidal horn, J. Acoust. Soc.
Am., 22(5), 558–562 (1950).
9. Molloy, C. T., Response peaks in finite horns, J. Acoust. Soc. Am., 22(5), 551–557
(1950).
10. Lambert, R. F., Acoustical studies of the tractrix horn. I, J. Acoust. Soc. Am., 26(6),
1024–1028 (1954).
11. Jensen, A. O. and Lambert, R. F., Acoustical studies of the tractrix horn. II, J. Acoust.
Soc. Am., 26(6), 1029–1033 (1954).
12. Klipsch, P. W., A low frequency horn of small dimensions, J. Acoust. Soc. Am., 13(2),
137–144 (1941).
13. Carlisle, R. W., Method of improving acoustic transmission in folded horns, J. Acoust.
Soc. Am., 30(7), 687–687 (1958).
14. Geddes, E., Audio Transducers, GedLee, Northville, MI (2002) ISBN-13:
978–0972208505.
15. Post, J. T. and Hixson, E. L., A Modeling and Measurement Study of Acoustic Horns,
Thesis, Electroacoustics Research Laboratory, Department of Electrical and Computer
Engineering the University of Texas at Austin, Austin, TX (May 1994).
16. Olson, H. F., Sound concentrator for microphones, J. Acoust. Soc. Am., 1(3A), 410–417
(1930).
17. Stewart, G. W., Acoustics: A Text on Theory and Applications, D. Van Nostrand Company,
Inc, New York (1931) reprint by Stewart Press (2007) ISBN-13: 978–1406750119.
18. Henricksen, C. A. and Ureda, M. S., The manta-ray horns, J. Audio Eng. Soc., 26(9),
629–634 (1978).
19. Ragan, G. L., Microwave Transmission Circuits, Dover, New York (1965) ISBN-13:
978–0486614083.
20. Ureda, M. S., On the movement of a horn’s acoustic center, Proceedings of 106th Audio
Engineering Society Convention, Munich, Paper# 4986 (1999).
21. Nomoto, I. et al., A technique for observing loudspeaker wave-front propagation,
J. Audio Eng. Soc., 24(1), 9–13 (1976).
22. Borch, D. Z. and Sundberg, J., Spectral distribution of solo voice and accompaniment
in pop music, Royal Institute of Technology, Stockholm, Sweden, Report TMH-QPSR,
Vol. 43, pp. 31–35 (2002).
23. Hayes, C. D., Acoustic spectrum shaping utilizing finite hyperbolic horn theory, Jet
Propulsion Laboratory, Pasadena, CA, N.A.S.A. Technical Report 32–1141 (1967).
20 Gradient Loudspeakers
20.1 INTRODUCTION
20.1.1 Size and Multipole Approaches
Loudspeakers can be very directional at high frequencies when the diaphragm begins
to be similar in size to the wavelength as noted in Chapter 10. At low frequencies,
except for large horn and direct radiator loudspeaker systems, loudspeaker directiv-
ity from closed and ported box systems in this working range is wide.
In some situations directivity is desirable at low frequencies. Transducer directivity
can be obtained by size (geometry) and/or by multipole (gradient) techniques.
Multipole techniques were shown to function well for microphones due to their
small size. Although not mentioned in Chapter 13, a further reason for the success
in applying multipole techniques to microphones lies in their relative homogeneity
in electroacoustic properties, achieved in a first-order gradient (e.g., bidirectional
and cardioid) since the sound pressure can be fed to the back side of the diaphragm
through an acoustic delay network. This is similar to using two identical microphones
spaced a small distance along with some signal processing.
The gradient microphone was discussed in Chapter 13, and its directivity
characteristics were shown to be constant over a much wider frequency range than
those of arrays that rely on size for directivity. Reciprocity implies that similar
directivity properties must apply to a multipole loudspeaker as to a multipole
microphone. At least two microphones are necessary to achieve second- and higher-
order gradient directivity characteristics. Similarly, two or more loudspeaker
drivers are necessary to achieve second- and higher-order gradient directivity
characteristics, but with more drivers there will be more severe driver matching
problems.
20.1.2 Gradient Loudspeaker Types

A gradient loudspeaker is sometimes defined as a loudspeaker that has two or more
loudspeakers spaced a small distance apart and fed signals with a difference in phase
[1]. By this definition, a gradient loudspeaker would only be a special type of broad-
band endfire array similar to the other arrays discussed in Chapter 12.
A bidirectional electrostatic loudspeaker should then not be called a gradient
loudspeaker although a similar, bidirectional ribbon microphone is usually called
a gradient microphone. The definition in Ref. [1] is unnecessarily restrictive. Many
electrostatic loudspeakers have dipole sound radiation, i.e., first-order gradient
397
bidirectional, characteristics at low frequencies since they are usually not used
with a loudspeaker box to contain their rear radiation.
Such loudspeakers can be shown to provide good sound reproduction in rooms
as long as sound level requirements at low frequencies are not too formidable.
Electrodynamic drivers are used for gradient loudspeakers since such drivers can
move large volumes of air. The first problem to consider is that since the far-field
sound pressure is proportional to loudspeaker-generated volume acceleration,
loudspeakers must be fairly large and can shadow one another. Also, in contrast to
the case of microphones, commercial electrodynamic loudspeakers suffer material
and production variance so the problems found with microphones occur even
more so with electrodynamic loudspeakers. At low frequencies, the problems are
a result of the difficulties in manufacturing loudspeaker drivers that have similar
compliance and magnetic properties. At high frequencies, diaphragm bending
wave resonances and the coupling between the voice coil and diaphragm become
problematic and result in large variations in sound pressure at frequencies of a
few kilohertz for a 20 cm diameter paper or plastic diaphragm loudspeaker. The
frequency range over which higher-order gradient loudspeakers can be used to
achieve a certain directivity characteristic is much more limited than that of
microphones.
20.2 USE OF GRADIENT LOUDSPEAKERS

Gradient loudspeakers can find use in both indoor and outdoor applications. Outdoors,
a unidirectional loudspeaker directivity characteristic can reduce neighborhood
noise contamination from the use of a loudspeaker system.
20.2.1 Indoors: Low Frequencies

Indoors, at frequencies below the lowest room mode, the net effect of gradient
loudspeakers compared to omnidirectional loudspeakers is reduced sound pressure
level (SPL), since the sound pressure in the room is proportional to the total volume
acceleration of the loudspeaker which is lower for the gradient loudspeaker than that
of the equivalent omnidirectional “monopole”-type loudspeakers. This drawback
may be compensated to some extent with loudspeaker drivers that are capable of
large diaphragm excursions and by signal processing.
The interplay between loudspeaker and room in the sparse mode frequency range,
which is below about 200 Hz in a typical room of 50 m3 volume, will depend on the
type of gradient operation. The Schroeder frequency f S given by Equation F.20 is the
frequency above which the room modes seen in the steady-state frequency response
curves start to become “inseparable”.
A bidirectional gradient loudspeaker will couple to the room modes differently
from a monopole loudspeaker. A cardioid-type gradient loudspeaker will couple well
to the room modes irrespectively of its placement since it combines monopole with
gradient operation.
Gradient Loudspeakers 399
20.2.2 Indoors: Mid and High Frequencies

It is important to consider that hearing acts as a directional sensing short-time
Fourier analyzer, so the arrival times and directions of the early reflections have
considerable importance for frequencies much higher than the Schroeder frequency.
The steady-state frequency response curve of the room is about 20 dB wide above
the Schroeder frequency because of interference between simultaneously excited
modes. The Schroeder frequency is mainly of interest for reverberation time and
steady-state SPL measurements to determine above which frequency the modes are
so dense in frequency that the sound field of the room can be considered diffuse.
Due to the properties of hearing and the signal characteristics of speech and
music, as well as because of cognitive factors, it is still useful to be able to use
the nulls afforded by gradient loudspeakers to minimize the audible effect of
reflections by room surfaces. The fairly constant directivity characteristics of
gradient loudspeakers, such as bidirectional electrostatic loudspeakers, are desirable
particularly in the mid-frequency range.
20.3 FIRST-ORDER GRADIENT SOURCES

The sound radiation from a monopole (sometimes called a zero-order gradient
source) was studied in Chapter 3, and the sound pressure from a radially vibrating
sphere generating the volume acceleration ωU was given in Equation 3.35. By
combining these sources, we can produce arrays with special features, such as one-
and two-dimensional arrays, that have polar patterns with a first- and higher-order
bidirectional or unidirectional characteristic. Often there is a need to use linear
arrays (i.e., “column” loudspeakers) of gradient sources to compensate for their
relative inefficiency in low-frequency sound radiation.
20.3.1 Bidirectional Array
If we sum the sound pressure contributions from two out-of-phase omnidirectional
sources at a small distance D from one another using Equation 3.35, we find that the
far-field sound pressure is
e − jkr  kD 
p (r, d, k ) = ωρU sin  cos ( Ω ) (20.1)
2πr  2 
Here Ω is the angle away from the line connecting the sound sources as shown in
Figure 12.17. To have a constant bidirectional directivity pattern, the wave number
and the distance between the sources must be such that the product D/λ < 0.5. When
D/λ becomes much larger, the directivity pattern will start to feature many lobes
as discussed in Chapter 12. The response of the bidirectional array in the Ω = 0 and
π directions as a function of D/λ and kD is shown in Figure 20.1.
For flat frequency response in the range D/λ < 0.5, it is necessary to pre-filter the
signal to the loudspeaker with a −6 dB per octave low-pass or shelving filter.
D/λ
0.125 0.25 0.5 1 2
0
–3
Relative response [dB]
–6
A point in the far-field
r
–12
Ω
+ – z
D
–18
π/4 π/2 π 2π 4π
kD
FIGURE 20.1 Frequency response of a dipole sound source consisting of two equal
magnitude out-of-phase constant volume velocity sources (Ω = 0).
20.3.2 Cardioid Directivity Arrays

Usually, a unidirectional directivity is more desirable than a bidirectional one. An
array having a first-order unidirectional directivity characteristic (usually referred to
as a cardioid directivity) can be obtained by using two out-of-phase sound sources at
a distance D/2 and having a small time delay τ = d/2c in the feed to one of the sources
as shown in the insert in Figure 20.2. The sound pressure for such an array is given
by Equation 20.2.
e − jkr  kd kD 
p (r, d, D, k ) = ωρU sin  + cos ( Ω ) (20.2)
2πr  4 4 
Figure 20.2 shows the sound pressure frequency response of the array for Ω = 0 and
d = D. When d = D, the directivity pattern has the familiar cardioid shape discussed
in Chapter 12. As in the case of the bidirectional loudspeaker, it is necessary to use
pre-filtering to compensate since the frequency response for Ω = 0 increases by 6 dB
per octave up to about D/λ < 0.5, if constant volume velocity sources are used.
The cardioid pattern loudspeaker will have the same directivity index as the cardioid
microphone, so an important characteristic of the cardioid loudspeaker is that it will
feed 4.8 dB less energy into the reverberant field than a conventional omnidirectional
loudspeaker. This results in much higher clarity of sound in a reverberant environment.
20.3.3 “Acoustic Resistance” Box

In practice, it is often desirable to avoid the use of two drivers because of cost
and space. It will also be difficult to find two identical drivers although a digital
D/λ
0.125 0.25 0.5 1 2
0
–3
–6
r
D/2
–12 Ω
+ – z
τ=d/2c
–18
π/4 π/2 π 2π 4π
kD
FIGURE 20.2 Frequency response of a dipole sound source consisting of two equal
magnitude constant volume out-of-phase velocity sources with a small time delay τ when
Ω = 0 and d = D.

∆l
r
Ω
Vent radiation
impedance, ZARv Air cavity, CAB RAB z
Diaphragm radiation
impedance, ZARd
Vent, MAV RAV Diaphragm, suspension,
and coil, MAD CAS RAS
FIGURE 20.3 The principle of the cardioid pattern acoustic resistance loudspeaker box.
self-calibrating system could possibly be used to equalize the drivers. Using a

loudspeaker box with a single driver, applying the principle of acoustic resistance
but with the resistance placed on the back of the box—similar to that of a cardioid
microphone—achieves the same goal [2,3]. The principle of the cardioid pattern
acoustic resistance loudspeaker box is shown in Figure 20.3.
It is seen from this figure that the loudspeaker working principle follows that of
the cardioid microphone shown in Figure 14.7 and the ported loudspeaker box shown
in Figure 17.1a. Figure 20.4 shows the acoustical part of the circuit representing the
diaphragm, box volume, and acoustic resistance.
Care must be taken concerning the choice of material for making the acousti-
cal resistance. The high pressure generated by the combination of particle velocity
and resistance may cause the acoustic resistance material to deform and the resis-
tance to become nonlinear. The deformation may also cause increased reactance.
eG Bl
REC SD MAD CAS RAS MARv
UD UV ARv
UB
(Bl)2
MARd REC SD2 RAB MAV
CAB RAV
ARd
FIGURE 20.4 The acoustic impedance analogy for the acoustic resistance box loudspeaker.
Resistance implemented by the use of materials that have small pores, such as metal
gauze, plastic foam, tightly woven fabrics, is reported to cause serious distortion at
high particle velocity when used in gradient loudspeakers. Good results are reported
to have been obtained with glass wool and coarse wadding [3]. It is important that
the resistance can be trimmed to the correct value.
The addition of the resistance and inductance representing the added back part
of the complete circuit is shown in Figure 20.4 as a branch in parallel to the box
acoustic impedance due to R AB and CAB with the mass MAV and resistance R AV due
to the vent impedance load on the back of the box. Radiation is represented by
the two radiation impedance components MARv and ℜARv. Usually, MARv ≫ MAV and
ℜARv ≪ R AV. The size of the box Δl must of course be such that Δl/λ < 0.5 and Δl ≪ r.
The simplified equivalent circuit is shown as an acoustic impedance analogy in
Figure 20.5. The inside box volume resistance circuit has been removed from the cir-
cuit as this simplifies the analysis. The coupling between the two oscillating masses
on the outside of the enclosure due to radiation impedance has been neglected. The
coupling is usually not of importance for conventional loudspeakers.
eG Bl
REC SD CAS RAS RAV
MAD
UD UV
(Bl)2 UB
REC SD2
CAB MAV
FIGURE 20.5 The acoustical impedance analogy for the acoustic resistance box loudspeaker
at very low frequencies.
The calculation of the sound pressure is straightforward using superposition of

pressures since the interaction between the two radiators has been neglected. The
sound pressure contributions p̲ v and p̲ d due to the volume velocity Uv and Ud generated
in the port air and loudspeaker diaphragm vibration are at far distance
e − jkrD
pD ( ω, rD ) = j ω ρU D (20.3)
4πrD
e ( D
− j kr + k∆l cos (Ω ))
e − jkrV
pV ( ω, ∆l, Ω, rV ) = j ω ρUV ≈ j ω ρ UV (20.4)
4πrV 4πrD
In practice, the distances from the diaphragm and port to the listening position are
going to be almost the same rV ≈ r⋅ D + kΔlcos(Ω) at the wavelengths of interest for
low-frequency sound reproduction systems. Since we assume the system is linear, we
can apply superposition. Because of the reference direction for U V, which is driven
from inside the box, i.e., from the rear of a loudspeaker diaphragm, the pressures
are added using a minus sign as for the ported box loudspeaker. By calculating
the volume velocities, one finds the equation for the far-field sound pressure of the
loudspeaker.
) e4πr (20.5)
− jkr
(
psum ( ω, ∆l, Ω, r ) ≈ j ω ρ U D − jUV k∆l cos ( Ω )

Typically, the unidirectional cardioid directivity pattern is desired and is obtained in
analogy with the theory for the cardioid microphone when
∆l = cRAV C AB (20.6)
The unidirectional loudspeaker should be designed so that the low-, mid-, and high-
frequency properties complement one another. As was noted initially, the directivity
characteristics start to deform for Δl/λ values larger than about 0.5, and at high
values there will be various side lobes if the loudspeaker front- and back-side
radiation is omnidirectional. Ideally, the driver should have a diaphragm size such
that its frontal radiation starts to become directional when the “array” stops being
unidirectional [3].
As with the ported loudspeaker box, it is necessary to limit the diaphragm
excursion below the cutoff frequency of the loudspeaker using an analog or digital
electronic filter so that the loudspeaker driver is not damaged by spurious low-
frequency signals, and so damage and clipping does not occur due to suspension and
voice coil displacement limitations. Another reason to limit the very low frequencies
is the Doppler distortion generated by high frequencies riding on top of the large low-
frequency velocity movements of the diaphragm. Because of the flow resistance R AV,
the situation is not as critical as for the ported loudspeaker box. A more in-depth
discussion of the acoustic resistance loudspeaker box and its loudspeaker driver
requirements as well as its use in rooms can be found in Refs. [2–4]. In using the
cardioid loudspeaker outdoors, it is advantageous to stack a number of units so that
one obtains a column system since this further reduces the low-frequency excursion
of the individual diaphragms.
Finally, it should be pointed out that drivers for a gradient loudspeaker must
have very rigid diaphragms that withstand buckling since the box volume in these
loudspeakers tends to be minimized and the compliance of the box air volume
small.
20.4 SECOND-ORDER GRADIENT SOURCES

Because of the added cancellation effects as gradient order increases, the practical
applications for higher-order loudspeakers are limited to those where extended low-
frequency response is not required such as in speech reinforcement situations in very
reverberant environments.
We can design higher-order gradient sources using larger arrays [1] as shown in
the insert in Figure 20.6. The insert shows an array built around two dipole sources at
a distance D/2 and connected so that one of the dipole sources is fed a signal delayed
by τ = d/2c seconds. All inter-speaker distances are small compared to the distance r
in the far-field of the observation point. The sound pressure for such an array is given
by Equation 20.7 and is shown in Figure 20.6 for Ω = 0, d = D, and Δ = D/4.
e − jkr  k∆   kd kD 
p (r, d, D, ∆, k ) = jωρU sin  cos ( Ω ) sin  + cos ( Ω ) (20.7)
πr  2   4 4 

D/λ
0.125 0.25 0.5 1 2
0
–3
–6
–12
–18
–24 A point in the far-field

D/2
–30 r
∆/2 ∆/2
–36 Ω
– + + – z
–42
τ = d/2c
–48
π/4 π/2 π 2π 4π
kD
FIGURE 20.6 Frequency response of a second-order unidirectional sound source consist-

ing of our equal magnitude constant volume in- and out-of-phase velocity sources with a
small time delay when Ω = 0, d = D, and Δ = D/4.
Here Δ is the distance between the sources in the pairs that make up the dipoles.
We see that the directivity characteristic of such an array will be a cosine pattern
as expected from the dipoles and multiplied by a cardioid pattern as expected from
the time delay of one of the pairs. We also note the frequency response rise of about
+12 dB per octave that results in even more pre-filtering of the signal being necessary
than for the first-order cardioid.
REVIEW QUESTIONS
20.1 What are the advantages of using gradient-type loudspeakers in rooms?
20.2 Why would cardioid loudspeakers be advantageous in outdoor sound
reinforcement applications?
20.3 Explain how the choice of electronic time delay in the feed circuit of a gradient
loudspeaker affects its directivity pattern.
20.4 Which factors do you think cause the breakdown of the cardioid directivity
pattern for a practical loudspeaker according to Figure 20.3?
PROBLEMS
20.1 Design a cardioid-type gradient loudspeaker using an electrodynamic driver
that has MMD = 2.1 · 10 −2 kg, CMS = 8.2 · 10 −4 m/N, R MS = 7 . 0 · 10 −1 Ns/m, Bl = 9.2
N/A, SD = 2.3 · 10 −2 m2, LEC = 2.5 · 10 −4 H, REC = 6.9 Ω.
20.2 For an applied voltage of 1 V rms, determine the driver diaphragm’s vibration
amplitude as a function of frequency for a cardioid loudspeaker having a
driver with the characteristics in Problem 20.1.
20.3 Determine the sensitivity of the response and directivity characteristics of a
cardioid loudspeaker to variation in the vent resistance R AV. Use the data in
Problem 20.1 and assume that the vent resistance can vary by ±10%.
REFERENCES
1. Olson, H. F., Gradient loudspeakers, J. Audio Eng. Soc., 21(2), 86–93 (1973).
2. Iding, W. H., Unidirectionally radiating loudspeakers, Proceedings of the 2nd Central
European Audio Engineering Society Convention, Munich, Germany, Paper M08 (1972).
3. Holmes, T. J., The “acoustic resistance box”—A fresh look at an old principle, J. Audio
Eng. Soc., 34(12), 981–989 (1986).
4. Backman, J., Theory of acoustical resistance enclosures, Proceedings of the 106th Audio
Engineering Society Convention, Munich, Germany, Paper 4979 (1999).
21 Drivers Using Flexible
Diaphragms
21.1 INTRODUCTION
Loudspeaker drivers that use nonrigid flexing diaphragms—called flex drivers in
this book—use flexible diaphragms in contrast to the nominally rigid diaphragms of
conventional electrodynamic voice coil drivers discussed previously.
Most drivers are based on the motion of a rigid conical piston. True piston motion
will be possible at least at the low-frequency portion of the working range of a
particular driver. At mid and high frequencies, most conventional voice coil driver
cones will show resonant and modal behavior. This is due to both the “point” excita-
tion by the voice coil and the limited wave speed in the diaphragm that results from
its distributed mass and stiffness.
There will be resonances and a diffuse vibration field in the diaphragm similar
to that of the sound field in a room whatever the diaphragm shape. The higher the
frequency, the more diffuse the vibration field.
The designer of traditional drivers strives to make them behave as a nonresonant
flat bandpass filter when mounted in a baffle or box as described in Chapters 15
through 19 [1,2]. At high frequencies the diaphragm is usually not rigid: waves in
the diaphragm propagate away from the voice coil and will be reflected at the edges
and by diaphragm asymmetries. This leads to the resonant bending wave fields in the
diaphragm, circumferential and/or radial.
The mid- and high-frequency bending wave resonances in flexible plates will
have low modal density since the two-dimensional vibration fields are a result of
bending waves. Metal cone diaphragms will have particularly audible resonances
due to the relatively low material damping in metals, which results in the resonances
having long decay times.
A designer of traditional loudspeakers, interested in good sound reproduction,
will try not to use a driver in this range but rather add mid- and high-frequency
drivers to provide continued relatively nonresonant sound reproduction [3].
21.2 SYSTEM CONSIDERATIONS

Various types of drive mechanisms may be used for drivers designed around the
use of flexible diaphragms. In most cases, the driving force or moment is generated
electrodynamically although piezoelectricity has also been used. Figure 21.1 shows
sketches of two types of voice coil-type flex drivers. Note that the magnet assembly
with its voice coil is often called an exciter in the case of a flex driver.
407
Sandwich plate Suspension
Voice coil
Magnet
(a) Soft iron structure
Wave absorbing wedge Voice coil Bending sheet

attachment point
Voice coil
Magnet
Soft iron structure
(b)
FIGURE 21.1 Some possible diaphragm designs for drivers using flexing diaphragms: (a)
honeycomb sandwich-type resonant diaphragm, (b) edge-damped flexible diaphragm.
Drivers based on the use of flexing diaphragms may be designed to use the diaphragm
in different ways such as the following:
• Direct radiation due to propagating bending waves ((1) driving point

radiation and (2) quasi plane wave radiation)
• Direct radiation due to resonant bending wave fields (usually mixed driving
point radiation and resonant plane wave radiation)
• Direct radiation due to resonant transversal wave fields
The radiation is complicated and depends on the shape of the diaphragm as discussed
in Chapter 10. The flexing diaphragm can support many wave types. The important
waves from the viewpoint of sound radiation are bending waves and out-of-plane
transverse waves. The driving point impedance for bending waves is resonant
tending toward resistiveness with increasing frequency.
Doing away with the heavy magnet, cone, basket, which are the bulk of conventional
electrodynamic loudspeakers, seems like a good thing provided one can succeed in
designing a similar-quality new loudspeaker that has low mass, thickness, and bulk
as well as freedom of installation geometry. Note that the resonant diaphragms of
flex drivers can have widely varying shapes since they do not have to be designed
with rigidity in mind. The diaphragm geometry is rather chosen so the resonant
modes of vibration have a suitable frequency distribution, damping, geometrical
distribution of vibration, and sound radiation.
Figure 21.2 shows a photo of a cut through the diaphragm of a flex loudspeaker
that uses a resonant honeycomb sandwich plate.
Drivers Using Flexible Diaphragms 409
FIGURE 21.2 Honeycomb diaphragm and flex driver in the background. (Sample courtesy
of NXT/HiWave Technologies PLC Cambridge United Kingdom, photo by Mendel Kleiner.)
Compliance
Voice Sandwich plate
Magnet
Soft iron
structure
FIGURE 21.3 Flex driver where the magnet is suspended by the diaphragm.
Since many flex drivers have quite stiff and sturdy honeycomb sandwich-type
diaphragms, it is attractive to make them “self-supporting,” i.e., attaching the
driver directly to the diaphragm without support other than the compliance shown
in Figure 21.3. This requires very low driver mass so that the diaphragm does not
buckle and limit the force factor of the electrodynamic driver.
Very successful electrodynamic piston driver designs have been manufactured
using radially enforced honeycomb materials. The use of this type of material
lends itself to many types of loudspeaker diaphragms, not only resonant ones [4,5].
A photo of a cut-up of such a non-flexing diaphragm is shown in Figure 21.4.
21.3 DIAPHRAGM WAVE FIELDS

The flexing diaphragm can support many wave types. As noted in section 21.2,
the important waves from the viewpoint of sound radiation are bending waves
and out-of-plane transverse waves. The main characteristics of the diaphragm
besides its geometry are its mass per unit area and bending stiffness as discussed
in Chapter 4.
FIGURE 21.4 A driver diaphragm cut-up to show the interior honeycomb reinforcing
structure. (Sample courtesy of Technics, Panasonic Corporation of North America,
Chesapeake, VA; photo by Mendel Kleiner.)
The diaphragm vibrational wave field generated by the transducer will be composed
of a direct and a reverberant field. For sufficiently wide frequency range sound, there
will be a diffuse vibration field in the diaphragm similar to that of the sound field in a
room. This is analogous to the way a sound field in a room excited by a monopole is
composed, as discussed in Appendix F. It is reasonable to assume that decay time of the
diaphragm resonances should be much shorter than of the environment in which the
driver is used. Listening to loudspeakers in reverberant rooms and anechoic chambers
confirms this hypothesis. The resonances must be damped by some means.
At the driving point of a thin plate (h < λB /6, see Chapter 4), there will be a bending
wave near field in the plate as shown in Figure 21.5. One can show that the sound
radiation by this near field corresponds to that of a rigid piston in an infinite baffle
where the piston has a diameter d ≈ 0.25 λB [7].
The phase velocity of bending waves in a homogeneous diaphragm is given by
Equation 4.28 so the wavelength of the bending waves is given by
B′
λB = 4 (21.1)
ω 2 m′′
Since the bending wave wavelength varies with frequency, the sound radiation by
the near field at the plate driving point corresponds to that of a driver diaphragm that
has a size that varies with frequency.
As with any resonant bending wave system, the radiation will be quite frequency
dependent as was discussed in Chapter 10. One usually tries not to use the driver for
frequencies above the critical frequency of the diaphragm. The critical frequency for
Relative vibration level (dB)

0
–5
–10
0 1 2 3 4 5
kBr
FIGURE 21.5 The velocity-level characteristics of the bending wave field to the side of
the point where the driver is attached to the diaphragm on an infinite stiff plate. (After
Cremer, L. et al., Structure-Borne Sound: Structural Vibrations and Sound Radiation at
Audio Frequencies, 3rd edn., Springer, New York, 2005.)
thin plastic sheets can be quite high: for a 1 mm thick polypropylene diaphragm, it
is about 30 kHz.
The phase velocity of transverse waves in a sandwich diaphragm is similar to that
of bending waves, except for the frequency region in which the bending stiffness
moves from being high to low, due to the bending stiffness being determined by
the spaced sub-plates at low frequencies, and the sum of the sub-plate’s bending
stiffnesses at high frequencies, as shown in Figure 4.11.
This affects the resonance characteristics of the diaphragm. The bending wave
mode frequencies for a thin rectangular homogeneous plate, freely suspended at
its edges, are given by Equation 4.31. The bending wave field’s modal density, per
Equation 4.32, is constant with frequency whereas the modal density of the sandwich
plate increases with frequency in the transition region, adapting to the modal densities
of the low- and high-frequency bending wave fields. The average bending wave
modal density dN/df in a homogeneous thin plate is given by Ref. [7].
dN 2πS D
≈ (21.2)
df 3.6cL h
Here SD is the plate area, cL the longitudinal wave speed in the material, h the
thickness of the plate, and h < λB /6. The exact frequencies of the modes will also be
affected by the impedance load of the exciter, but these are not usually of any great
interest since the exact frequencies are not important for the driver’s operation and
the resulting sound quality.
The energy that is injected into diaphragm motion is dissipated by acoustical and
mechanical losses. The acoustical losses are due to the radiation of sound power from the
diaphragm. For resonant flex drivers, the vibration primarily generates a local reactive
sound field with particle motion but little sound power radiation. The diaphragm motion
will push air aside in one region and suck air in at some other region.
The mechanical losses occur at the driving point, in the diaphragm itself due to
nonideal flexing, and at the edges not only due to energy transfer to the loudspeaker
frame or basket but also due to wave conversion into other types of wave fields due to
the discontinuity of the edge. The internal losses in diaphragm materials vary widely
as discussed in Chapter 15.
21.4 DIAPHRAGM SOUND RADIATION

At the driving point of the diaphragm, there will be a bending wave near field in the
diaphragm as shown in Figure 21.5. The driving point near field will be overlaid on
the resonant field. The relative power of the sound radiation by the near field at the
driving point and that of the resonant modal wave field will depend on the surface
area and geometry of the diaphragm, its damping, and its edge conditions. There
will always be radiation from the near-field area around the driving point, and as
mentioned in the previous section, the sound radiation by this near field corresponds
to that of a rigid piston with diameter d ≈ 0.25 λB in an infinite baffle [7–9].
21.4.1 Aerodynamic Cancellation
Since the front- and back-side radiation of the flexing diaphragm is out of phase,
there will be aerodynamic cancellation. If the loudspeaker is to be mounted on
or close to a wall, one must consider that a dipole perpendicular and close to a
sound-reflecting surface effectively works as a quadrupole with considerable sound
cancellation effects as studied in Chapter 11. Since one might want to mount the
flex driver loudspeakers on a wall, the thinness of the construction being one of the
attractive properties of flex driver loudspeakers, one must remove most of the rear-
side radiation, for example, by enclosing the diaphragm back by a box. In this case,
the air inside the box will also affect the modal distribution. The cancellation can
also be reduced or removed by mounting the driver in a baffle or by using sound-
absorptive material at the back of the diaphragm.
21.4.2 Modal Sound Radiation

The resonant vibrational field on the plate is composed of symmetric and asymmetric
modes. The vibration pattern and sound radiation of each of these modes will be
different as described in Chapter 10. Examples of a symmetric and an asymmetric
diaphragm mode on a diaphragm mounted on a closed box are shown in Figure 21.6.
There will be air pumping inside the box because of the diaphragm vibration.
Symmetric modes will be more influenced by the air pumping than asymmetric
modes. If the air volume of the box is large and well damped as in a conventional
loudspeaker, it will have little influence except on the symmetric modes that have the
lowest frequencies. Loudspeakers that use flexing diaphragms however tend to have
shallow boxes to use the design possibilities offered by the small magnet systems
of the exciters. In such loudspeakers, the addition of sound-absorptive material
inside the box will greatly influence the coupling and the damping of the modes.
Symmetric Asymmetric
FIGURE 21.6 Symmetric and asymmetric modes of a diaphragm attached to a

loudspeaker box.
The position of the material will be particularly important for its effectiveness in
damping asymmetric modes [10,11].
Note that several modes can be excited at one frequency because of the finite
amplitude of the modes due to the damping. The diaphragm will generally be of
the size of the wavelength of sound in the mid-frequency range. For frequencies
below this frequency range, the radiation will be that of the sum of volume velocity
generated by the modes at any particular frequency so the frequency response will
be irregular. In this low-frequency range, the sound radiation of the diaphragm will
be omnidirectional whatever the distribution of the modes [9,11]. For frequencies
at which the diaphragm is large compared to the wavelength of sound in air, the
sound radiation will be composed of the contribution from corner and edge
modes as shown in the figures in Chapter 10. The pattern will be that of a one- or
two-dimensional array, and if the modal density is high, the sound radiation will be
fairly omnidirectional, especially if averaged over a third-octave band.
In the frequency range below the critical frequency, the average radiation ratio
for a bending wave field can be approximated as shown in Figure 21.7. For most
diaphragm materials that have high critical frequency fc, the power radiation will
essentially be dependent only on the mean square vibration velocity over the mid- and
1
Radiation ratio (log scale)
e
/octav
1.8 dB
λ2c
S
ve
cta
/o
dB
+6
c2 P2 cλc 3c fc
−
–1 100
2Sfc 8S P2 P 4
Frequency [Hz] (log scale)
FIGURE 21.7 Design curve for approximating the average radiation ratio of a resonant
finite panel of perimeter length P and area S with simply supported edges. Here λc is the
wavelength of sound at the critical frequency. (After Beranek, L.L., ed., Noise and Vibration
Control, Institute of Noise Control Engineers, New York, 1988.)
high-frequency audio range. Since the force from the exciter is approximately
frequency independent and the driving point impedance roughly resistive, the
radiated power response will be frequency independent. The frequency response
however at any one point in the room will vary widely due to the excitation and
radiation conditions.
Of course, one can design the diaphragm so that only parts of it operate in any
frequency range. Such decoupling is usually done using concentric mass rings to
form radial lowpass filters. The reason to do this is of course to avoid the serious
sound coloration due to the interference inherent in the radiation by the corner and
edge modes. This trivial approach is similar to that used in old-style paper cone
loudspeakers with compliant decoupling rings [2,13]. Alternatively, the sandwich
material can be made to have continuously radially changing properties.
21.5 DRIVING POINT IMPEDANCE

The driving point impedance is defined as the ratio between the complex force and
velocity (in the direction of the force) in a system. We have already studied the
mechanical driving point impedance for mass and compliance. Other values for
mechanical driving point impedances may be found in Ref. [7]. For drivers, it is
primarily the driving point impedances of conical diaphragms and plane plates that
are of interest.
Since the diaphragm radiates sound, there will be an impedance element added to
the driving point impedance representing the radiation mass load and power losses.
At low frequencies, the radiation impedance will be dominated by the reactive part
that has mass character as studied in Chapter 4.
When we studied conventional electrodynamic drivers in Chapter 16, we assumed
the basket and magnet as having infinite mass so that the force from the voice coil
was only moving the conical diaphragm at its center. This is a good approximation
as long as the basket and diaphragm are rigid. At some frequency though, typically
over a few kilohertz for a low-frequency range driver, the resonant character of the
diaphragm becomes noticeable as the frequency response of the loudspeaker starts
to become ragged, showing a multitude of peaks and dips as in Figure 21.8. This is
due to the standing wave field patterns set up by the waves in the diaphragm.
The driving point impedance of the finite flexing disk will depend on the way
that it is suspended. Any practical mass must be held in static position by some
arrangement; for example, it can be freely suspended or attached to some frame
using a compliant element such as a compliant surround. This applies both to drivers
that have conventional cone diaphragms and to those that have flexing diaphragms.
For a freely suspended mass, gravitation forms the restoring force.
In contrast to the conventional electrodynamic driver, the basket is often
eliminated in flex drivers. In some of these drivers, the sandwich-type diaphragm is
so statically stiff that the magnet structure can be mounted on the diaphragm or very
close to it. The electrodynamic driver, because of the impedance conditions, can be
very light. The driver is then often suspended compliantly on the diaphragm at one
point and the voice coil rigidly at another in a concentric fashion.
Mechanical
impedance Magnet and basket
modulus mass
(log scale)
Diaphragm
break-up resonances
Diaphragm
and voice coil Suspension
mass
compliance
f0 fmax
Frequency (log scale)
FIGURE 21.8 Modulus of cone driving point impedance (level vs. log frequency) for a
conventional electrodynamic driver.
Other flex driver designs use a circular bending plate as shown in Figure 21.1b.
The idea is in this case to have a nonresonant circular plate similar to an infinite plate.
The driving point impedance will then be dependent on the effectiveness of the
anechoic termination and the internal damping in the plate.
The driving point impedance for bending waves on an infinite plate is real and
frequency independent, as shown by Equation 4.34. For the finite diaphragm, the
impedance will depend on the modes, their frequencies, damping, and the position
of the driving point relative the diaphragm modes. As discussed previously, there
will be damping of the modes due to sound radiation but generally the damping is
due to internal losses in the diaphragm and losses at the diaphragm edges. A simply
supported edge will yield smaller losses than a clamped edge attached to a frame. In
the latter case, the bending motion at the edge will depend on the bending stiffness
Mechanical
impedance Magnet
modulus mass
(log scale)
8√m˝B΄
Diaphragm
resonances
Diaphragm
mass
Frequency (log scale)
FIGURE 21.9 Typical modulus of the driving point impedance for a transducer having a
freely suspended flexing diaphragm (level vs. log frequency).
of the edge support. Figure 21.9 shows the characteristic driving point impedance of
a finite plate. Since the driving point impedance now is much lower than in the case
of the rigid cone, shown in Figure 21.8, the mass of the driver’s magnet system can
be made smaller than that for a conventional driver.
Some flex drivers use two or more exciters driving the diaphragm to obtain higher
sound power output and a more even excitation of the vibrational modes. There are
also designs that use a more traditional approach and attach the diaphragm and the
magnet structure using a form of chassis similar to the basket of a conventional
electrodynamic driver.
21.6 ELECTROACOUSTIC CIRCUIT ANALOGIES

21.6.1 Two Sliding Masses Driven by a Force
The electroacoustic analogy of the flex driver is similar to that of a freely suspended
electrodynamic driver, but it is instructive to start by studying some elementary cases
of how forces act on masses. We will start by studying two masses MM1 and MM2,
which are acted upon by a transduction force as shown in the mechanical mobility
analogy in Figure 21.10.
The masses can be considered as the combined driver basket and magnet mass,
MMBM, and the diaphragm mass MMD. In a conventional electrodynamic driver,
MMBM ≫ MMD, so most of the force is used to move the diaphragm. Usually the basket
is attached to the loudspeaker box that—in addition—is coupled to the floor and
thus to the “ground” of the inertial system. Because ground has almost infinite mass,
almost all the force will be used toward generating diaphragm velocity irrespective
of the impedance of the diaphragm.
Now assume that MMBM is small and that the diaphragm is attached to the basket
by way of some compliant structure. The structure acts as a spring so the diaphragm
is decoupled from MMBM over some frequency as shown in Figure 21.11.
uM1 FM uM2
MM1 MM2
FM
FM
uM2 uM1 uM2 MM2

MM1
MM1 MM2
FIGURE 21.10 Two masses free to slide on a flat friction-free plane and the associated
mechanical mobility analogy.
uM F uD
MMBM YMD
CMS
F
F
MMBM CMS YMD

CMS
MMBM YMD
FIGURE 21.11 The basket and magnet mass MMBM and the generalized impedance YMD
shown on top, below the associated mechanical mobility analogy.
The low-frequency cutoff will be determined by the compliance–mass

relationship, since below the system resonance frequency most of the force will be
directed toward the driver shorted by the compliance CMS.
21.6.2 Exciter Driving a Resonant Diaphragm

From the viewpoint of the electrodynamic transduction mechanism, the voice coil
mass is a part of the load joined in series with the driving point impedance of the
diaphragm. If the resonant diaphragm is very large, its impedance is almost resistive
because of the internal and radiation losses, since there will be little energy reflected
back from the boundaries. In this case, the low-frequency cutoff will also depend on
the MMBM to CMS relationship.
The mobilities of the voice coil mass MMC and the driving point mobility YMD
appear in parallel as shown in Figure 21.12. They are functionally one mechanical
component. Note that the magnet mass MMM is included in the circuits because it is
attached to the flexing sandwich plate through the lossy compliance CMS and RMS [14].
The conversion from mobility to impedance analogy follows the same principles
as those that lead to the circuit in Figure 16.6. The resulting circuit is shown in
Figure 21.13. Note that the voice coil mass MMC has a very small reactance value
compared to the driving point impedance Z MD = 1/YMD except at high frequencies
where the presence of the voice coil mass will lead to a drop in frequency response.
The magnet mass MMM has however a very high value, which means that most of the
current over the highpass cutoff frequency will go toward driving the diaphragm.
If the compliance CMS holding the magnet attached to the diaphragm is sufficiently
large, the circuit will simply be that of the force driving the diaphragm impedance.
uM F uD
MMM YMD
CMS
MMC
RMS
CMS
CMS
RMS
RMS
MMM YMD MMC
MMM YMD MMC
FIGURE 21.12 The magnet mass and the generalized impedance of a flex driver without
basket shown on top, below the associated mechanical mobility analogy.
LEC
(Bl)2
CMS RMS MMC
uD
eG Bl (Bl)2
F MMM ZMD
REC + REC + jωLEC REG + REC
FIGURE 21.13 The mechanical impedance analogy for the flex driver.
21.7 RESONANCE AND SOUND QUALITY

Sound quality of loudspeakers is determined by many factors: linear and nonlinear
distortions being the most important, it also depends on the placement of the
loudspeaker and the listener in the listening environment and the sound reflections
added by the environment. Ref. [15] contains a thorough discussion on the effects of
room acoustics on sound reproduction.
A good loudspeaker usually has
• Linear frequency response (directly or by using a simple equalizer or filter)

• Linearity, i.e., no perceptible nonlinear distortion
• Directivity characteristics that suit the application and room

• Suitable sensitivity and power-handling capability
It should also be inexpensive to manufacture and have low mass so that it can be
manufactured and shipped inexpensively.
Hearing plays an important role in analyzing sounds and that subjectively good
sound quality can be obtained also with loudspeaker having linear distortion such as
flex drivers based on the use of a resonant diaphragm. Hearing acts as an advanced
two-channel short-time spectral analyzer to resolve the time-frequency behavior of
the sound.
Linear frequency response and low nonlinear distortion are important sales
arguments. It is reasonable to assume that a loudspeaker that has linear frequency
response is better from the viewpoint of fidelity than one that has a ragged frequency
response. However, we know that the frequency response of a room to a sound source
is very ragged once the direct sound is weak compared to the reverberant sound.
The binaural unmasking capability of hearing lets the precedence effect determine
the subjective response to the complex sound field comprised of the direct sound
from the loudspeaker and the room reflections. Closing one ear canal opening with
a finger lets us listen to the “true” pressure response of the loudspeaker and room
combination. Convolving single channel audio by a room impulse response reveals
the coloration that the room filter adds due to its temporal reflection and spectral
resonance characteristics.
Flat frequency response is not necessary for good sound reproduction but a ragged
response is acceptable only under special circumstances. An important question
to be answered by future research is what properties of the loudspeaker’s transfer
function or nonlinearity cannot be accepted. The colored sound quality of metal
cone loudspeakers compared to paper and polypropylene cones indicates that both
diaphragm damping and mode structure are important to sound quality.
The sound from flex driver loudspeakers will be composed of both sound from the
direct field, where the exciter is attached, and sound from the resonant field. Neither
of these is characterized by a flat sound pressure frequency response. This means that
the sound (except for its spatial properties) will be similar in quality to that passing
through a multi-resonant filter or reverberator. Investigations into the sound quality of
reverberators is therefore of interest. The coloration due to inadequate modal density in
electronic reverberators is well investigated [16]. Investigations into the coloration due
to insufficient modal density of plate and room reverberators resulted in the conditions
shown in Figure 4.10 [17]. It is difficult to find objectively measured data for the modal
density of commercial loudspeaker such as distributed mode loudspeakers, balanced
mode radiators, and others. The frequency response of the loudspeaker does not
indicate the modal density and decay characteristics and neither does the frequency
response of a room. Clearly, the decay time of any resonant loudspeaker system must
be much lower than that of the room in which the loudspeaker is used [1].
The critical frequencies and modal densities of some diaphragm materials and
typical thicknesses are shown in Table 21.1. It is seen that unless the diaphragm
is very large, the modal density will be very low, much lower than that deemed
acceptable in the experiments reported in Ref. [17]. One can of course argue that the
TABLE 21.1
Coincidence Number and the Critical Frequency for Some Common
Diaphragm Materials and Typical Thicknesses
Young’s Coincidence Typical Critical Modal
Density Modulus Number Thickness Frequency Density
Material (kg/m3) (Pa) (m/s) (m) (Hz) (Hz−1)
Paper 5.0 · 102 2.0 · 109 31 1 · 10−3 31 · 103 1.6 · 10−2
Aluminum 2.7 · 103 6.9 · 1010 12 3 · 10−4 41 · 103 2.2 · 10−2
Polypropylene 9.1 · 103 1.6 · 109 46 1 · 10−3 46 · 103 2.4 · 10−2
For comparison purposes the surface area has been assumed as 0.01 m2 for all materials.
influence of the lower modal density of resonant flex drivers is offset by the shorter
decay time of the modes. This remains to be shown. Note that modal spacing in
frequency is the inverse of modal density, i.e., a modal density of 0.02 modes/Hz
corresponds to modes being about 50 Hz apart.
REVIEW QUESTIONS
21.1 Why can the magnet assembly be supported by the diaphragm in the flex
driver shown in Figure 21.5?
21.2 Why is there sound radiation from the driving point in bending wave appli-
cations? What would happen if a diaphragm point would be held still in an
incoming bending wave field?
21.3 How is the near-field radiation by a point-excited plate compared to that of a
rigid diaphragm?
21.4 What precautions must be taken when using a flex driver close to a sound-
reflecting plane to ensure efficient sound radiation?
21.5 How does the driving point impedance of an infinite plate differ from that of
a finite one?
PROBLEMS
21.1 A driver that has a rigid cone diaphragm has the following data:
MMD = 2.1 · 10 −2 kg, CMC = 8.2 · 10 −3 m/N, R MS = 7.0 · 10 −1 Ns/m, Bl = 9.2 N/A,
SD = 2.3 · 10 −2 m2, LEC = 2.5 · 10 −4 H, REC = 6.9 Ω. The driver faces a closed box
having a volume VB = 6.0 · 10 −2 m3.
Tasks:
a. Calculate the driving point impedance level as a function of frequency.
b. Compare this driving point impedance to that of a polypropylene sand-
wich plate that has a mass twice that of a single 5 · 10 −4 m thick plate and
a bending stiffness 10 times that of a single 5 · 10 −4 m thick plate. Use
data from Table 21.1.
21.2 A polypropylene sandwich plate made of the material mentioned in Problem

21.1 has dimensions 0.3 m by 0.49 m.
Tasks:
a. Calculate the 10 lowest resonance frequencies of the freely suspended
plate.
b. What will be the modal density of the plate?
c. How does this modal density compare to the coloration criteria shown
in Figure 4.10?
d. What is the minimum plate size to avoid coloration according to these
criteria?
21.3 A very large polypropylene sandwich plate of the type described in Problem
21.1 is excited by an electrodynamic self-supported transducer.
Task:
If the magnet mass is 2 · 10 −2 kg, what will be the lower cutoff frequency
of the system?
REFERENCES
1. Beranek, L. L., Acoustics, American Institute of Physics, New York (rev. ed. 1986)
ISBN-13: 978-0883184943.
2. Olson, H. F., Acoustical Engineering, D. van Nostrand, New York (1957).
3. Colloms, M., High Performance Loudspeakers, Wiley, London, U.K. (2005) ISBN-13:
978-0470094303.
4. Sakamoto, N. et al., Loudspeaker with honeycomb disk diaphragm, J. Audio Eng. Soc.,
29(10), 711–719 (1981).
5. Nakazono, J. et al., Coaxial flat-plane loudspeaker with polymer-graphite honeycomb
sandwich plate diaphragm, Proceedings of the 66th Audio Engineering Society
Convention, Los Angeles, Paper # 1662 (May 1980).
6. Tanaka, J. et al., New loudspeaker diaphragms using fiber reinforced plastics, Proceedings
of the 75th Audio Engineering Society Convention, Paris, Paper # 2061 (March 1984).
7. Cremer, L. et al., Structure-Borne Sound: Structural Vibrations and Sound Radiation at
Audio Frequencies, 3rd edn., Springer, New York (2005) ISBN-13: 978-3540226963.
8. Panzer, J. and Harris, N., Distributed-mode loudspeaker radiation simulation,
Proceedings of the 105th Audio Engineering Society Convention, San Francisco, Paper
# 4783 (September 1998).
9. Angus, J. A. S., Distributed mode loudspeaker polar patterns, Proceedings of the 107th
Audio Engineering Society Convention, New York, Paper # 5065 (September 1999).
10. Cremer, L., Principles and Applications of Room Acoustics, Vol. 2, Applied Science,
New York (1982) ISBN-13: 978-0853341147.
11. Azima, H. and Panzer, J., Distributed-mode loudspeakers (DML) in small enclosures,
Proceedings of the 106th Audio Engineering Society Convention, Munich, Paper # 4987
(May 1999).
12. Beranek, L. L. (Ed.), Noise and Vibration Control, Institute of Noise Control Engineers,
New York (rev. ed. 1988) ISBN-13: 978-0962207204.
13. Frankort, F. J. M., Vibration patterns and radiation behavior of loudspeaker cones,
J. Audio Eng. Soc., 26(9), 609–622, September 1978.
14. Roberts, M., Exciter design for distributed-mode loudspeakers, Proceedings of the
104th Audio Engineering Society Convention, Amsterdam, Paper # 4743 (May 1998).
15. Toole, F., Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers
and Rooms, Focal Press, London, U.K. (2008) ISBN-13: 978-0240520094.
16. Schroeder, M., Natural sounding artificial reverberation, J. Audio Eng. Soc., 10(3)
219–223, July 1962.
17. Kuhl, W., Eigentone density and colouration of reverberant sound, Proceedings of the
6th International Congress on Acoustics, Tokyo, Japan, Paper # E-2–8 (1968).
22 Multiway Loudspeakers
22.1 INTRODUCTION
22.1.1 Bandwidth
Generally, some form of “flat” frequency response is desired for a loudspeaker.
However, single direct radiator drivers that cover the full audio range will have
both the frequency response and the directivity response vary in an unacceptable
way, since the ratio of wavelength to loudspeaker dimensions undergoes a large
change from low to high frequencies. In addition, a single driver would also
exhibit serious nonlinear effects because of the need for large excursions at low
and medium frequencies and the Doppler nonlinear distortion that is a caused
by high-frequency (HF) vibration riding on top of low-frequency (LF) diaphragm
vibration. Mechanical limitations also prevent pistonic motion over the full audio
range for a practical-sized driver.
The piston frequency range of conventional electrodynamic loudspeaker
drivers using conical diaphragms is usually only 3–4 octaves wide. Since the
audio range is about 9 octaves wide multiple drivers, each designed for a portion
of the desired frequency range, need to be used unless a wide range design is
used, such as a loudspeaker back-loaded by a horn. Many loudspeaker systems
for high-quality audio are two- or three-driver designs. Each driver is assigned a
working frequency range with some small overlap and mounted in an optimized
enclosure type. The division of energy into different frequency ranges appropriate
for each driver is done by electronic filters, usually called crossover filters or
networks.
22.1.2 Example
Figure 22.1 shows the frequency response of a once popular near-field monitor for
control room use. For such a loudspeaker, the off-axis sound is of little interest in
the design, since in control room environments, early reflections and reverberation
will be negligible at the listening position.
Near-field monitors are designed so that the direct sound from the loudspeaker
will have flat frequency response with minimal distortion. The frequency response
curves in Figure 22.1 however show dips, peaks, troughs, and bumps. Elimination of
the dips is not as necessary as controlling the frequency response peaks, troughs, and
humps [1]. We note that the solid line is characterized by the following:
• Highpass filter cutoff at around 150 Hz. Rise rate is +12 dB per octave
below around 125 Hz because of the closed box-type enclosure. For a
small loudspeaker, the cutoff frequency cannot be decreased by increasing
423
5
0
–5
–10
–15
In-phase
Out-of-phase
–20
20 50 100 200 500 1 k 2 k 5k 10 k 20 k
Frequency [Hz]
FIGURE 22.1 An example of the on-axis frequency responses of the individual drivers of a
two-way near-field monitor loudspeaker system (thin dotted lines) and the responses for cor-
rect and incorrect (out-of-phase) interconnection (anechoic measurements).
diaphragm suspension compliance since the total compliance will be

limited by the enclosed air of the enclosure. The cutoff frequency could
be decreased by adding mass to the diaphragm, but the loudspeaker would
then become less efficient. The use of an electronic filter to extend the LF
response would require large driver diaphragm excursion, which is difficult
to combine with low nonlinear distortion.
• Frequency response peak at about 1.8 kHz with a rise starting at around
0.5 kHz. This rise is due to a combination of diffraction- and directivity-
induced effects and could be controlled with a shelving filter as described
in Appendix B.
• Some cone resonances starting at around 1 kHz. The resonance frequencies
will be different for each driver sample. The modal behavior of the
diaphragm at the resonances will also result in a frequency response that is
very measurement point dependent. For these two reasons, such peaks and
dips are difficult to control for an industrially manufactured loudspeaker.
• Peaks and dips in the high midrange at about 5 and 7 kHz. These have
low Q factors and could be controlled using two cut (attenuation) filters to
reduce the peaks.
• Lowpass filter roll-off starting at around 12 kHz. This roll-off is fairly easy
to control using a small signal electronic filters, but the use of large signal,
passive filters would require a reduction of loudspeaker sensitivity.
22.2 DIAPHRAGM DIMENSIONS AND WAVELENGTH

The directivity of driver sound radiation was discussed in Chapter 12 and shown to
depend on diaphragm size, motion, and environment. A free driver is additionally
subject to aerodynamic cancellation.
Multiway Loudspeakers 425
To avoid such cancellation, the drivers are mounted in a loudspeaker enclosure,

on a baffle, or to a horn. The box or baffle prevents sound from the back of the driver
from combining with the sound from the front of the driver so the loudspeaker
radiation becomes hemispherical. This results in the “baffle effect” that is seen
as a slow rise in frequency response resulting from the loudspeaker box front
dimensions becoming large compared to wavelength. The loudspeaker box front
will act as a baffle at frequencies where its dimensions start to become of the order
of the wavelength.
For drive units that are mounted in small boxes, the conditions will be similar
to a driver mounted at the end of a long tube. Baffle and box effects depend on the
geometry of the box as shown in Figure 16.9. It is difficult to separate the baffle effect
from the directivity effect of the driver when the driver dimensions are similar to
the box front dimensions. The baffle effect will depend on the exact mounting of the
driver as will be discussed later in this chapter.
Sometimes a full-range baffle effect is desired from loudspeakers in small
enclosures. By mounting the enclosure recessed a bit into the wall, its front and
driver are then flush with the wall, so the baffle effect will apply across the whole
frequency range.
Another possibility in loudspeaker design is a multiway loudspeaker where all
drivers radiate in omnidirectionally. This is likely to be unsatisfactory because
of the interference between the direct sound and between the direct sound from
all the drivers and reflected sound. The interference generates peaks and dips in
the frequency response both for on-axis and off-axis sound. The sound pressure
contributions by the drivers are additive so there will always be interference between
the drivers. The drivers cannot be at the same place physically, and the crossover
filters give only limited attenuation close to the crossover frequency. The reflected
sound must be considered for small rooms unless it is attenuated by 20 dB or more
compared to the direct sound, for example, by sound absorptive or diffusive treatment
of the relevant room surfaces.
Some loudspeakers are very directional, for example, large loudspeaker arrays,
horns, or wide frequency-range ribbon or electrostatic drivers. This limits the
energy supplied to early reflections and reverberation, and these loudspeakers can
use simpler crossover filters, although frequency response equalization may still
be necessary.
22.3 LOUDSPEAKER POLARITY, PHASE, AND GROUP DELAY

Keeping track of loudspeaker polarity is important in stereo and other multichannel
systems. When a voltage is applied to the driver terminals, its diaphragm will move.
If there is an outward movement, the terminal connected to the non-grounded output
of the amplifier is said to be the plus terminal. The polarity is important when com-
bining several drivers in a loudspeaker and when loudspeakers are used in stereo or
multichannel setups.
The transfer function phase response should be smooth and linear so that the
group delay is the same at all frequencies for the complete loudspeaker system.
It is often difficult to measure group delay of loudspeakers with accuracy better
than about 0.1 ms, corresponding to a signal delay path length of about 3 cm. The
relationship between phase and group delay is discussed in Appendix B. In addition
to the signal delay caused by the crossover filters, the geometry of the loudspeaker
units and their mounting on the loudspeaker box, and its shape, there is also a
transmission delay from the points of sound generation to the point of measurement
or listening. The absolute delay is seldom of any great importance except where
synchronization with visual information is desired as with speech, music, and
sound effects.
Crossover filters are used to split the energy between drivers in different frequency
ranges and affect the phase over a large frequency range. In most cases, the phase
changes quickly close to the cutoff frequency of a filter with corresponding large
frequency-dependent delays. In loudspeakers, the phase anomalies may lead to
audible group delay aberrations which may change with a listener’s location since
the drivers cannot be mounted at the same place.
22.4 PLACEMENT OF DRIVERS

With two or more drivers in a multiway loudspeaker unit, the question arises: What
is the optimum strategy for placing the drivers on the baffle or enclosure front?
Should they, for example, be placed laterally or vertically as shown in Figure 22.2?
Since most audio listening will use stereo or multichannel loudspeaker systems and
associated recordings, both precise localization of phantom sources and frequency
response are of importance. The precedence effect causes the direct sound to be
the most important for the formation of phantom sources. For lateral placements,
movement of the acoustic center laterally as a result of a crossover from the lowpass
to the highpass loudspeaker driver results in the phantom source being located
differently for high and low frequencies. This is most noticeable on transient sounds
(a) (b) (c) (d) (e)
FIGURE 22.2 Drawings (a)-(e) show some ways of placing drivers on a loudspeaker box
front. Dark grey circles show low frequency drivers, light grey circles show mid-frequency
driver, and medium grey circles show high-frequency drivers.
such as handclaps, castanets, drums, etc. Consequently, placing the drivers on a

vertical line makes most sense, but for stereo listening, close, mirrored placement
of the different drivers on the left and right side enclosures is also possible with
acceptable results.
22.4.1 Baffle Effect
The baffle effect is the rise in loudspeaker output power with frequency as a result
of the increased radiation impedance that results when the baffle or box front
becomes large compared to wavelength. The loudspeaker configurations shown in
Figure 22.2 differ in baffle effect. The more full circle Fresnel zones that fit on the
box front, the larger will be the baffle effect. The closer the loudspeakers are to
an edge, the smaller will be the baffle effect. This results, for example, in that the
baffle effect will be the larger for the (e) than the (d) mounting. For the (a) mounting,
the effect will be larger than that for the (b) mounting, since the small HF driver
in that case “sees” more of the box front. In Figure 16.9, the baffle effect is clearly
seen in all box configurations for a single loudspeaker. Configurations (e) and (k) in
that figure shows the large frequency response rise that is a result of the symmetric
placement. Configurations (i) and (l) in Figure 16.9 clearly show how a reduction of
the effective acoustical size of the baffle by chamfering the box front reduces the
baffle effect.
The baffle effect will also be minimized by using vertically displaced drivers that
are located at tube ends as shown in Figure 22.3. Note however that this requires
the loudspeaker tubes to have about the same diameter as the driver diaphragms.
Even so, straight tubes such as the HF tube in the figure should be tapered to
reduce reflections from the external far end of the tube. The waves inside the tubes
Adjusted to same
effective source distance
taking crossover delays
HF into account
LF
Height adjusted to
listener’s head location
FIGURE 22.3 Placing drivers at the ends of tubes reduces the baffle effect if the tubes are
of the same diameter as the loudspeaker drivers. HF, high frequency; LF, low frequency.
HF lsp
Direction
∆l
to listener
LF lsp
(a) (b) (c)
FIGURE 22.4 Tilting the loudspeaker box or displacing the drive units (b and c) may
compensate for the difference in effective distance, Δl, in (a) between two loudspeaker drive
units. HF, high frequency; LF, low frequency.
can be attenuated filling the tube with a suitable sound absorber as discussed for
transmission line loudspeakers in Chapter 18.
22.4.2 Delay
The delay characteristics will depend on the way in which the loudspeaker drivers
are placed on the box front. Typically, HF loudspeakers have shallower cones than
LF loudspeakers, so that HF direct sound will, on the average, arrive earlier than
LF direct sound. This can be compensated by simply giving the enclosure front a
recessed shape for the mounting of the HF loudspeaker as shown in Figure 22.4.
A drawback of this method is that the diffraction effects may be stronger when the
drive unit is mounted in this way.
It should be noted that there are many other driver-mounting possibilities
available. LF drivers may be mounted on box sides and back. MF and HF drivers
may be mounted on box tops, on baffles, or be used with acoustic mirrors.
22.4.3 Directivity
Changes in directivity will mostly affect the size of the optimum listening volume
(sometimes called the “sweet spot”) and the frequency response of early reflections
and reverberation.
The directivity of piston loudspeakers was shown in Chapter 12 to be more
pronounced when the diameter of the rigid piston is comparable to or larger than the
wavelength. For a constant piston to wavelength ratio, an infinite number of drivers
would be necessary. The ratio can to some extent be controlled by the design of the
diaphragm, for example, by discrete or continuous decoupling of the HF vibration
toward the diaphragm edges.
Multiway systems use separate drivers and there will be jumps in directivity
between their different frequency ranges. The differences may to some extent be
controlled by mounting the HF driver at the bottom of a shallow recess or horn as
shown in Figure 22.5a. The recess width should be similar to the effective diameter
HF Isp Recess width HF Isp
Hole in
LF Isp LF Isp compressed
glass fiber sheet
(a) (b)
FIGURE 22.5 (a) Using a recess to make HF driver has directivity similar to that of LF
driver. Recess width is similar to effective diameter of LF loudspeaker driver. (b) Using a
holed sheet of compressed glass fiber to control directivity of LF driver. The recess may be
thought of as a short horn.
of the LF loudspeaker. Different recess shapes (conical, exponential, etc.) can be

used to tweak the frequency response and directivity.
Sometimes it is desirable to reduce the directivity of an LF loudspeaker while
retaining the volume velocity capability. An approach reminiscent of that found
in compression driver horns can then be used as shown in Figure 22.5b. The LF
driver is simply covered with a sheet of glass fiber having suitable flow resistivity
and an appropriate hole. The limited flow resistivity will result in good damping
of the resonances of the cavity between the driver cone and the fiber glass sheet.
This technique is particularly useful in simulating the characteristics of the mouth
for “speaking” manikins. For this use, the hole can be shaped as a 1 cm by 3 cm
rectangle. The air volume between the diaphragm and the sheet together with the air
of the hole will function as a lowpass filter as discussed in Chapter 7.
22.4.4 Concentric Drivers

An obvious way to mount the HF and LF drivers is on a common cylindrical axis.
Figure 22.6 shows some alternatives. The use of a dual-cone arrangement (“whizzer”
cone) can be thought of a crude way of using the same voice coil to drive two separate
diaphragms. The large LF cone can be mechanically decoupled at high frequencies
by attaching it to the voice coil former using an elastic joint or by making the part of
the cone closest to the voice coil flexible.
Two other alternatives are shown in (b) and (c) of this figure. In the dual-cone
case (a), it is difficult to arrange for efficient decoupling of the LF driver cone that
can be used in production, and in cases (b) and (c) the HF driver must be very
small limiting its use to very high frequencies. In case (d), the loudspeaker will
become quite large, but the HF horn will have about the same directivity as the
LF driver. For the LF cone forms and extension of the horn for better coupling,
see Chapter 18.
(a) (b) (c) (d)
FIGURE 22.6 Four concentric HF/LF arrangements: (a) Dual-cone (“whizzer” cone), (b)
HF driver mounted on bridge over LF driver, (c) extra HF driver on center magnet piece of LF
driver, and (d) HF driver coupled to a horn that uses the LF cone as an extension of its length.
22.5 THERMAL AND LINEARITY ASPECTS

Each loudspeaker driver must be fed its own signal so that it does not overload
mechanically or thermally and so that the output from the loudspeaker drivers is
minimized in those small frequency ranges where they overlap.
The spectrum of speech and much classical music typically has a peak in the
0.5–1 kHz octave bands. Since many HF loudspeaker drivers are intended to be used
up to 20 kHz, they need to have quite small diaphragms and voice coils to reduce
the moving mass. Such voice coils have little heat capacity and are easily damaged
by the audio energy available in the 1 kHz range and below for many sound types.
Dome-type high and midrange drivers have larger voice coils than cone drivers for
these frequency ranges so they have better power-handling capability.
Small LF drivers need to make large excursions to radiate sufficient sound
power, so nonlinear distortion can be a problem in design and operation. All of
these factors contribute to the frequent use of crossovers around a few hundred
hertz for low/mid-frequency range drivers and around a few kilohertz for low/
high- and mid/high-frequency range drivers to achieve a reasonable compromise
between frequency response linearity, directivity, nonlinear distortion, and power
handling.
22.6 LOUDSPEAKER AND LISTENING ENVIRONMENT

The design of loudspeakers must be also done with an eye toward the properties
of the acoustics of listening rooms, as mentioned earlier. The acoustic properties
of rooms can vary widely, so an important part of crossover design is to tailor the
frequency response for direct, early reflection, or reverberant fields as preferred.
The early reflections will affect the auditory source width and the localization of
phantom sources. The reverberant sound field is used to achieve the desired sound
field spaciousness and envelopment. The reverberant field frequency response

contributions by the loudspeaker drivers will be different from that of the direct sound
since the diffuse reverberant field consists of sound radiated to all possible angles by
the loudspeaker.
Direct field loudspeakers such as near-field monitors are usually designed for a
maximally flat frequency response whereas loudspeakers that are designed to make
use of room surface reflections (early reflections and reverberation) are often designed
for a compromise between flat direct signal frequency response and a “soft” roll-off
of HF response over about 2 kHz.
Since the response of most loudspeakers becomes directional at the upper fre-
quency limit of the speaker, crossing over to a smaller and more omnidirectional
loudspeaker driver will result in dips in the frequency response to the side affecting
the relative level and the frequency response of the early reflections. Since many
walls have relatively flat frequency response reflection characteristics over the mid-
frequency range, the initial reflections of loudspeaker sound radiated toward these
surfaces will also show frequency response dips. The effect on early reflections can
be minimized by using diffusers or sound absorbers placed at the mirror image loca-
tions on the room walls. Later reflections that make up the reverberation of the room
will of course be reflected so many times by room surfaces and objects that minor
response irregularities will be averaged in the reverberation summation of many
sound field contributions.
For near-field monitor crossover filters, the target is to find a set of filters that
will provide flat sound pressure frequency response in the listening area. Phase and
amplitude characteristics are then chosen by simple mathematical rules. Phase errors
that lead to group delay differences may be compensated for electronically.
For the reverberant field of a room, the reflection characteristics of the room
surfaces must be taken into account. Once the direct sound level is 10 dB below
the early reflections and reverberation in the room, the precedence effect is no lon-
ger as important and hearing is no longer dominated by the direct sound from the
loudspeakers [1,2]. In this case, the power summation of the loudspeaker signal
contributions is instead important.
22.7 CROSSOVER FILTERS

22.7.1 System Considerations
The need for multiway loudspeakers that require crossover filters was mentioned
several times in this chapter. In the design of crossover filters, the following factors
generally need to be considered:
• Driver frequency and phase responses

• Driver power handling
• Driver nonlinear distortion characteristics
• Driver directivity characteristics
• Loudspeaker enclosure shape and size
• Availability of components in suitable values
• Mounting conditions
• Listening conditions-direct or reverberant field
• Available space
• Cost
These factors will influence the choice of
• Type of multiway system: two, three, or more frequency ranges

• Type of crossover filter function
• Crossover frequencies
• Crossover slopes
• Time alignment
• Passive or active filters
• Placement of drivers on loudspeaker enclosure
If the drivers are assumed to have flat frequency response above their fundamental
resonance, have no baffle effect, are linear, have the same sensitivity, can be
mounted infinitely close, listening is done at a point, and that room acoustics is
of no importance, then the choice of target filter functions that will “suit” the
measured driver frequency response characteristics is simple. It is clear from the
initial example shown in Figure 22.1 that drivers have complicated characteristics
that are difficult to reconcile even when considering only the frequency response
in an anechoic chamber.
22.7.2 High- and Low-Impedance Active and Passive Filters

Figure 22.7 shows two different approaches to crossover networks. On the left
is shown the traditional approach where the network is inserted between power
amplifier and loudspeakers. This is often called a passive crossover although
the term low-impedance high-current crossover would be more adequate.
H(ω) H(ω)
H(ω)
Power
amplifier
ω
ω HF driver HF driver
H(ω) H(ω)
Power Power
amplifier amplifier
ω MF driver
ω MF driver
H(ω) H(ω)
Power
amplifier
ω ω
LF driver LF driver
Crossover network with Crossover network with
low impedance level filters high impedance level filters
FIGURE 22.7 Two types of crossover networks, large-signal low-impedance and small-
signal high-impedance circuits. Low-impedance filters are usually passive and high-
impedance filters are usually active.
Passive filters with low impedance need to handle large electric currents due to
the power needed to drive the loudspeakers. Such crossovers will be called large-
signal crossovers in this book.
The figure on the right shows how the network is inserted ahead of the ampli-
fiers, with each driver being fed power from its own amplifier. Since separate
amplifiers are used for each frequency range, these systems are improperly called
active filters, but the crossover that precedes the amplifier can be either passive
or active. A passive filter does not use any electronic amplification to achieve its
filter function.
The active crossover filters may be analog or digital designs, the latter will not be
considered in this text. Analog high-impedance input filters may be passive or active,
but the limited availability of suitable inductance components input leads to active
filters generally being used. An advantage of active high-impedance input filters is
the freedom they give the loudspeaker designer to look away from the impracticalities
caused by low-impedance level and to achieve a more ideal filter. Active filters are
typically implemented as a high-impedance input small-signal filters, with output
source impedances low relative the power amplifier. These crossover filters will be
called small-signal crossovers in this book.
While once there would have been a clear economic reason for choosing high
impedance small-signal passive over the corresponding active filters, advances in
electronics, cost, and production have made active filters an attractive choice. In
many cases the active filter and the associated power amplifier circuitry are built into
the loudspeaker enclosure so there will be a need for an electric power connection
to the loudspeaker.
22.7.3 Large-Signal Filters
The low-impedance, large-signal passive filter approach means that the current
that feeds the drivers goes through passive electrical filters built from inductors,
capacitors, and resistors. Figure B.26 shows some typical filter designs.
Many loudspeakers are designed with a certain nominal impedance largely
determined by the driver voice coil windings. For manufacturing and compatibility
reasons, these resistances are usually in the range of 3–15 Ω since they are intended
to be used with commercially available power amplifiers that have output impedances
below 0.1 Ω for best driver damping and power transfer efficiency. Large signal
passive crossover filters are usually placed in the loudspeaker enclosure.
Resistances in such filters will be in the range from almost zero to a few Ohm,
inductances about 1–100 mH, and capacitances about 1–100 μF, depending on
impedance level, crossover frequency, and frequency response correction needed.
Inductors in the range mentioned will require low-gauge wire and many turns
unless wound on a magnetic core such as a metallic or ferrite core. Small-value
inductors may be wound using no core; sometimes theses are called air-cored
inductors. A disadvantage of air-cored inductors is the stray magnetic fields that
surround the inductor and cause coupling between inductors, as in a transformer
so these must be placed orthogonally to one another. Metal or ferrite cores may
be toroidal or have C shape to minimize the magnetic leakage and the coupling.
A distinct drawback of any metal or ferrite core is the nonlinearity of the magnetic
materials and the nonlinear distortion the magnetic nonlinearity generates in the
signal. Magnetic leakage from the drivers may also contribute to the nonlinearities
that characterize the metal or ferrite core inductors in the filters. The core material’s
nonlinearity and the cost of the copper coil are important reasons to move from
large-signal passive filters to the use of small-signal active filters.
Capacitors used for high-level crossover networks may be bipolar electrolytic
capacitors or polypropylene and other nonelectrolytic capacitors. The latter give
negligible signal distortion but are bulky and may feature unacceptable inductance.
Bipolar or other electrolytic capacitors are plagued by nonlinearity and high losses
but are small and inexpensive compared to nonelectrolytic capacitors. They form
another reason to move away from large-signal filters in crossover filters.
Resistors that pass high current are lossy components that reduce the loudspeaker’s
efficiency. Power resistors tend to be wire wound to withstand the large electric
currents. Unless wound using a bifilar technique, such wire wound resistors may
introduce undesired inductance into the circuit. Wire wound resistors may overheat
and ignite material.
22.7.4 Driver Electric Impedance and Zobel Networks

Figure 15.4 shows an example of the electric impedance magnitude of a typical loud-
speaker system. We note that for the LF loudspeaker shown here, there is a large
impedance peak that corresponds to the resonance frequency of the diaphragm of
the voice coil driver mounted in its enclosure. Mid- and high-frequency loudspeakers
will have similar behavior in their respective frequency ranges.
Simple conventional crossover filter design assumes the loudspeaker has an
impedance that is primarily resistive over its working frequency range. The reason
for this is so the driver looks like a resistance load, with minimum reactance, to the
power amplifier. To make the impedance primarily resistive, it is necessary to use
compensation circuits such as the example shown in Figure 22.8.
22.7.5 Small-Signal Filters
Figure B.27 shows some basic arrangements of active filters. Electric noise can be
added by analog active filters but is usually negligible compared to room noise from
other internal and external sound sources, however, the possibility to filter the audio
signal without the nonlinearities inherent in high-level filters make analog active
filters attractive.
The power amplifiers must be able to supply the required voltages and currents and
tolerate the impedances of the driver loads. Zobel and other impedance-correcting
networks may be used to make the loudspeaker impedance more agreeable to the
amplifiers if necessary. Many amplifiers have a DC voltage output that may be
harmful to the drivers, particularly HF and some midrange units, so in practice
a capacitor is needed in series with each driver. This creates a passive crossover
network with the problems already discussed in the previous section.
Power amplifier
LZobel
Z
Rcable Lcable
RZobel
Ccable
CZobel
Zobel network Compensation for

compensation for resonance peak and
high frequency high frequency driver
impedance rise impedance rise
FIGURE 22.8 Example of electrical circuit components to compensate for the loudspeaker
load impedance Zel resonance peak and to protect the amplifier from the capacitive load of the
cable and inductive load of the driver (Zobel network).
Active filters and other aspects of active loudspeakers are further discussed in
Chapter 23.
22.8 SUMMARY
There are many computer programs available designed to model the response of
multiple-driver direct radiator loudspeakers; however, successful crossover design is
somewhat of an art and still requires much critical listening. An engineering approach
to crossover filters is presented in Appendix B and can serve as a starting point for
design. Crossover networks (and most other things concerning loudspeakers) are
discussed in detail in Refs. [3,4] and in many papers in the Journal of the Audio
Engineering Society.
REVIEW QUESTIONS
22.1 Why is there a need to use dedicated drivers for each frequency range of a
loudspeaker that should function over the full audio spectrum?
22.2 Give an example of why loudspeaker group delay characteristics are important.
22.3 Which factors influence the choice of mounting positions for drivers on a
loudspeaker enclosure?
22.4 Discuss the design for a loudspeaker that is to be used for sound reproduction
as a near-field monitor in a control room.
22.5 Discuss the design for a loudspeaker that is to be used for sound reproduction
in a reverberant listening environment.
22.6 Discuss the reasons for the choice of small- or large-signal crossovers for a
loudspeaker.
PROBLEMS
22.1 The figure below shows a simple electronic circuit that has a voltage source
and two resistors connected via a passive large-signal crossover filter that
splits the LF and HF signals to the respective resistor. For information on
filters, see Appendix B.
1V LF 7Ω HF 7Ω
Tasks:
a. Design a first-order Butterworth crossover filter that has a crossover
frequency of 2000 Hz.
b. Design a third-order Butterworth crossover filter that has a crossover
frequency of 400 Hz.
c. Calculate the sum of the voltages over the resistors as a function of
frequency for each of the cases (a) and (b) and draw the frequency
responses of the sums.
d. Calculate the sum of the power dissipated in the resistors as a function
of frequency for each of the cases (a) and (b) and express the sum as a
level in dB.
22.2 The figure below shows the circuit of a passive large-signal crossover filter
that splits the LF and HF signals from a voltage source to the respective driver.
Each driver is assumed to have an impedance that is purely resistive with a
value of 7 Ω.
LF HF
2.6 mH 5.4 µF
1V
10 µF 7 Ω 0.58 mH 7Ω
Task:
a. Calculate the sum of the voltages over the drivers as a function of
frequency, express the sum as a level in dB, and draw the frequency
response curve for the voltage sum.
b. Calculate the sum of the power dissipated in the drivers as a function
of frequency and draw the frequency response curve for the total power
dissipated.
22.3 The figure below shows the frequency responses of two drivers to be used for
a full-range loudspeaker system. Each driver is assumed to have an impedance
that is purely resistive with a value of 7 Ω.

–5
–10
–15
–20
0.2 k 0.5 k 1k 2k 5k 10 k
Frequency [Hz]
Task:
Design a crossover filter (of your own liking) that has a suitable crossover
frequency for the drivers considering only their respective frequency
responses. Explain your reasons for the design choices.
22.4 The figure below shows the impedance magnitude as a function of frequency
for a driver.
30
Impedance magnitude [Ω]
25
20
15
10
0
20 50 100 200 500 1000 2000
Frequency [Hz]
Task:
Design an electrical circuit that, in parallel with the loudspeaker’s electrical
terminals, will adjust the driver’s electric impedance so that its magnitude
will be 6 Ω ± 1 Ω within the frequency range shown.
REFERENCES
and Rooms, Focal Press, London, U.K. (2008) ISBN-13: 978–0240520094.l
2. Blauert, J., Spatial Hearing—Revised Edition: The Psychophysics of Human Sound
Localization, The MIT Press, Cambridge, MA (1996) ISBN-13: 978–0262024136.
3. Borwick, J. (Ed.), Loudspeaker and Headphone Handbook, 3rd edn., Focal Press,
Oxford, U.K. (2001) ISBN-13: 978–0240515786.
4. Colloms, M., High Performance Loudspeakers, 6th edn., Wiley-Blackwell, Chichester,
West Sussex, U.K. (2005) ISBN-13: 978–0470094303.
23 Active Loudspeakers
23.1 INTRODUCTION
Along with built-in power amplifiers, active loudspeakers use analog or digital signal
processing to modify the transfer function properties, shape optimized crossover
filters, prevent overheating, etc. The signal processing system may be located at some
distance or be installed in the loudspeaker. Analog processing is not only limited to
improvement of transfer function behavior and reduction of nonlinear transduction, it
can also be used for active impedance modification of driver parameters, a possibility
unique to analog techniques. Digital signal processing on the other hand can be used
to equalize the loudspeaker and room transfer functions and to limit overheating
based on models of the thermal properties of the drive mechanism, etc.
23.2 LOUDSPEAKER SOUND FIELD CHARACTERIZATION

Many active loudspeaker systems for audio will be used in rooms so the measurement
of the loudspeaker and room transfer functions forms an important initial step
in the design and use of active loudspeaker systems. Since hearing can separate
early arrivals of a sound from later “copies,” the precedence effect, psychoacoustic
considerations lead to the following three approaches:
• Measurement and possibly global equalization of sound pressure

distribution and decay characteristics at frequencies below a few hundred
hertz. Alternatively, the equalization may be directed toward equalization
of the power radiated by the system.
• Measurement and equalization of direct sound for medium and high
frequencies above about 0.5 kHz.
• Measurement and equalization of early reflected and/or reverberant sound
for medium and high frequencies above about 0.5 kHz.
In either case, a filter function is needed to form the filter that will equalize the
system. A measurement is needed to form this inverting filter function. It is worth
noting that it is more important to remove response peaks than dips [1].
23.2.1 Transfer Function Measurement

Normally, the transfer function will be measured using digital techniques that can be
well specified, but both the loudspeaker (that is to be equalized) and the microphone
(that forms an added filter) are in the measurement chain. No linearization can
be better than the measurement that is used for normalization. The properties of
the microphone are often assumed to be ideal, but to measure low-frequency (LF)
439
response of a room, the microphone should have a much lower cutoff frequency, at
least one-tenth of the cutoff frequency of the system to be measured. For an audio
system ranging down to 20 Hz, this means a cutoff frequency below 1–2 Hz. The
upper frequency is similarly limited by microphone size. Diaphragm resonance and
microphone directivity properties typically cause a more than 12 dB/octave cutoff at
high frequencies. In this case, it would be prudent not to use the microphone signal
above one-tenth of this cutoff frequency. When this is not practical the errors due to
measurement must be taken into account in some way.
The measurement of the transfer function can be done using either maximum
length sequence (MLS) and swept frequency (chirp) techniques, for example, among
other techniques. The measurement can also yield many other room acoustic metrics
such as reverberation time, etc., if measurement is done with a sufficiently long time
window. An intensity measurement may be of interest to understand the power flow
from the loudspeakers. Some active loudspeakers use the measurement of the input
impedance of the room so that they can output the same acoustic power.
23.2.2 Low Frequencies
The modal properties of the room cause the sound pressure to change as a
measurement, and source locations are changed. Reciprocity is a good method
for transfer function measurement at low frequencies. It is not feasible to measure
all combinations of source and microphone, so when optimizing the loudspeaker
placement for example, a compromise must be sought that is based on the assumed
position of the listener.
For measurement of the transfer function, it is sufficient to measure in a room
grid that has a resolution smaller than λ/2 according to the sampling theorem. If
one considers the upper frequency limit of the modal range from the viewpoint of
hearing as about 200 Hz (λ ≈ 1.7 m), measurements should be done at no more than
0.5 m distance for sufficient accuracy at 200 Hz.
23.2.3 Direct Sound at Medium and High Frequencies

For characterization of direct sound, it is necessary to measure the impulse response,
either directly or by measurement of the transfer function, and then converting the result
to the time domain using the inverse Fourier transform. Direct measurement can be
done using impulsive sounds, but this requires many repetitions of the impulse signal to
obtain sufficient signal to noise ratio and so a dual-channel broad band noise or swept-
sine technique is more common. When using transient signals with repeated pulse,
MLS or chirp signals, one must make sure that the measurement window is sufficiently
large so that the tail of the preceding measurement does not contaminate the beginning
of the next measurement. This also applies to transfer function measurements; the
measurement window must be sufficiently long for the time-domain response to be
unambiguous. Good measurements of the first 10 ms of impulse response require time
windows of the order of the reverberation time of the room.
Transfer function measurements of transducer response can be done in anechoic
chambers; in this case time gating of measured impulse response will be useful.
Active Loudspeakers 441
Because of diffraction and directivity, it is usually necessary to average the impulse

response measurements obtained also from measurements in the anechoic chamber.
At high frequencies, the directivity of the loudspeaker can be considerable. At a
frequency of 3 kHz, the far-field distance limit of a 0.2 m diameter piston is at about
0.4 m from the piston, and the corresponding beam width of the piston will be about
1.8 m wide at a distance of 3 m.
To minimize directivity and diffraction effects on measurement of the direct
sound field, it is prudent to measure at about 0.3 m distance at the corners of an
imaginary cube or sphere surrounding the sweet spot. Thus, eight measurements will
need to be averaged at each frequency. For high frequencies, it may be necessary to
average measured transfer functions on a power basis.
23.2.4 Early Reflected and Reverberant Sound at

Medium and High Frequencies
At medium and high frequencies, both transfer function and impulse response
are of interest, so the same type of measurements should be done as for the direct
sound. However, we must note that there is a limit to the spatial resolution that
can be had.
For measurements in the reverberant field, it is important that the sound pressure
at the measurement points is uncorrelated for appropriate averaging. The graph
in Figure 23.1 shows the correlation between points in a three-dimensional and a
two-dimensional sound field as a function of wavelength λ and distance between
points d [2]. In both cases, we see that the measurement points should be many
wavelength apart for the sound pressure at the points to be uncorrelated. The negative
interference between the many sound waves that make up the reverberant sound
leads to frequency response peaks and dips so it would be an error to try to invert the
1.00
3D diffuse
0.75 2D diffuse
Correlation factor
0.50
0.25
–0.25
–0.50
0 0.25 0.50 0.75 1 1.25 1.50
x/λ
FIGURE 23.1 Theoretical correlation factor as a function of distance between measurement

points as a function of wavelength for two types of diffuse sound fields (2D sound field curve
is shown for in-plane correlation). (After Kuttruff, H., Room Acoustics, 5th edn., Spon Press,
London, U.K., 2009.)
curve to equalize. The interference leads to a non-predictable frequency response.

The peaks and dip result in a 20 dB wide 95% confidence interval for the frequency
response of the reverberant sound [2].
23.3 ANALOG SIGNAL PROCESSING

23.3.1 Frequency Response Compensation
The simplest form of analog signal processing is frequency response compensation
using one or more ±6 dB/octave low-pass and high-pass filters. These can be used
also to build shelving filters as discussed in Appendix B. The appendix shows
examples of such circuits. Some loudspeaker enclosure adjustments require the use
of +6 dB/octave or even +12 dB/octave filters at LF as noted in Chapter 20. Some
loudspeakers that use drivers intentionally placed for sound reflections by walls will
need various types of equalization as well for a subjectively satisfying sonic character.
Equalization of resonance or interference effects is often included in passive or active
crossover filters, but the inevitable mechanical production spread in driver and box
production means that these resonant filters cannot be very sharply tuned.
An alternative in frequency response equalization by dedicated filters is to use a
general-purpose parametric or other equalizer. Such equalizers however use reso-
nant circuits, and their application is a Band-Aid approach, hiding the problem rather
than solving it.
23.3.2 Time Delay Compensation

This type of compensation is also discussed in Appendix B and circuits given there.
Unless audio needs to be synchronized with video or similar, the time delay in audio
systems is seldom of any large interest, so it is possible to introduce compensating
delays. Such time delay equalization usually uses digital rather than analog
techniques.
23.3.3 Current Amplifiers

Normally high-current capability voltage amplifiers (power amplifiers) are used to
drive electrodynamic loudspeakers. A common problem in the use of such amplifiers
is that as the voice coil heats, its resistance increases, the current drops, and the force
available to drive the mechanical system diminishes. The result is a compression
effect that reduces the dynamic range of the sound reproduction. Most loudspeakers
also suffer from the variation of voice coil inductance as the voice coil moves in the
air gap of the magnet system, which creates nonlinear distortion [3].
Both of these problems may be successfully addressed using the current drive
principle [4]. By using a current instead of voltage source, the drive current into the
loudspeaker becomes constant, and—if the Bl product does not change—the force
acting on the voice coil will remain constant. A drawback is the need for a linear,
wide-band amplifier to supply the current. Such an amplifier can be designed along
the lines shown in Figure 23.2.
+
Gain = A Zout
ediff
Aediff
ZEC
ein
ZEC >>REF
REF
FIGURE 23.2 Basic current-generating amplifier using a high voltage gain (A) voltage
amplifier with current-sensing feedback. With the feedback shown, the circuit will try to keep
current through R EF and ZEC constant. The loudspeaker’s electric impedance is ZEC.
The current drive principle has a considerable drawback in the need for separate
amplifiers for the different drive units in a multiway loudspeaker. An additional
drawback is that the electric damping of the mechanical system is lost. This
damping would otherwise have been provided by the low output impedance of a
voltage amplifier. With current drive, the mechanical and acoustic components must
by themselves provide the required damping. This issue could be addressed by an
electrical filter preceding the amplifier, special driver design, or diaphragm motion
feedback (described later in this chapter).
23.3.4 Electroacoustic Component Synthesis

Because of the tight coupling between the electric and the mechanical side in
most piezoelectric, piezoceramic, and electrodynamic transducers, the mechanical
properties can be sensed on the electrical side of the transducer as impedance
variations that are a direct result of the mechanical system. This property is used
for the function of crystal and other oscillators. The right side of Figure 22.3 shows
Power amplifier Loudspeaker
MMD rMS YMRb

i REC F CMS YMRf
Bl :1
iG e uD
FIGURE 23.3 The electroacoustic analogy of an electrodynamic driver and a current drive
amplifier combination.
the electrical analogy of an electrodynamic LF loudspeaker. Clearly, we can use the

analogy to calculate the mass, compliance, and damping of the mechanical system.
By including a second impedance network on the primary side of the electrodynamic
transformer, we can adjust the apparent mechanical properties of the loudspeaker
without adding mechanical mass or compliance.
One way in which this can be achieved electronically by giving the power amplifier
that drives the loudspeaker a suitable output impedance. For the amplifier output
impedance to be really “visible” to the driver, the amplifier also needs to have negative
output resistance –REC to compensate for the voice coil resistance REC so that the
parallel reactances can be joined. Figure 23.4 shows one possible output impedance
configuration. The addition of a virtual compliance (CMsynt) will lead to an increased
driver resonance frequency while adding virtual mass (MMsynt) will lower the resonance
frequency. An advantage of the scheme is that the circuit is fairly insensitive to
variations in driver and amplifier component values. The electronically synthesized
components act as real mechanical components in the vibrating system [5].
Figure 23.5 shows the circuit now with the mechanical side converted to the
electrical side analogy.
MMsynt rMSynt MMD rMS YMRb

i CMSynt –REC REC F CMS YMRf
Bl : 1
iG e uD
FIGURE 23.4 The electroacoustic analogy for the electrodynamic loudspeaker implemented
as a mechanical mobility analogy with the power amplifier output impedance added on the
electrical side.
MMsynt rMSynt MMD rMS

i CMSynt –REC REC (Bl)2 (Bl)2CMS (Bl)2
iG
FIGURE 23.5 The electroacoustic analogy for the driver and power amplifier output
impedances but with the radiation components neglected.
R8
RG CCp RRp R7
ZEVC >>R4
CLp
RRs
RA R6 ZEVC
ein
R5
R4
FIGURE 23.6 An example of a circuit showing the feedback principle [5]. All amplifiers are
operational amplifiers that have high gain, high input impedance, and low output impedance.
The loudspeaker’s electric impedance is ZEVC. Note circuit similarity of the output stage to
that of the current amplifier shown in Figure 23.2.
An amplifier configuration that has an output impedance that allows the

implementation of this analogy is shown in Figure 23.6, more fully described in
Ref. [5]. The reference also gives design formulas for the synthesized mechanical
components as functions of the circuit’s electric components.
23.3.5 Diaphragm Motion Feedback

All electrodynamic and other drivers have nonlinearities. Many methods have been
invented to remove or compensate for the nonlinearities. Magnetic and compliance
nonlinearities are difficult to linearize without sensing the diaphragm motion. The
aim is ultimately to control and linearize the acceleration of the diaphragm since the
far-field sound pressure is proportional to diaphragm acceleration. One such system
principle is “motional feedback” (MFB) [6,7], and other forms of diaphragm motion
feedback have been investigated for many years. The basic principle of diaphragm
motion feedback is shown in Figure 23.7.
The diaphragm or cone of electrodynamic drivers is not rigid so the motion is
usually not pistonic. If it were, control would be simple. At low frequencies, the
diaphragm may bend because of excessive air load. At midrange and high frequencies,
a practical driver diaphragm or cone has modal motion so it would be inappropriate to
try to control globally. Capacitive sensing of the displacement the entire diaphragm,
for example, would not correctly sense modes that have displacement a number of
whole wavelengths. Sensing of the sound pressure inside the loudspeaker box would
need linearization due to the nonlinear relationship between volume and pressure
(and would work only for sealed-box loudspeakers at that).
The sensing of diaphragm motion is instead often done at the voice coil, since at
low frequencies the diaphragm is tightly coupled to the voice coil. The motion sensing
+
Gain = A Zout
ediff
A ediff
–
ein Motion sensor
on driver
Signal from ZEVC
H(ω)
motion sensor
FIGURE 23.7 Basic diaphragm motion feedback that uses a current amplifier and voice coil
motion sensing. The loudspeaker’s electric impedance is ZEVC. With the feedback shown, the
circuit will try to keep current through ZEVC constant but modified according to the feedback.
can be done using many methods such as displacement, velocity, or acceleration

sensing, but ultimately the feedback system needs to control the current through
the voice coil for preserved linearity as discussed in a previous section. Since any
electronic system will generate inherent noise and distortion, it is imperative to select
a low noise, linear sensing method.
One method is the acceleration sensing by inertia that uses a piezoelectric,
internally amplified accelerometer. It is a method that is inherently quite linear but
not optimal from a signal-to-noise viewpoint since acceleration can be small at low
frequencies.
Voice coil velocity can be sensed using a second voice coil, much in the same way
as in the dual voice coil loudspeaker shown in Chapter 16. Another possibility is to
sense voice coil velocity via the electromagnetically induced voltage in the voice coil
by including the voice coil in an electronic bridge circuit.
The displacement of the voice coil can be sensed by sensing the sound pressure
inside the cavity formed by the diaphragm center cap and the center pole piece of the
magnet system. Some other methods are the use of capacitive or optical techniques.
A differential transformer arrangement will also work.
Since the voice coil acceleration is proportional to the force acting on the voice
coil and thus to the drive current, the voice coil sensing must provide a voltage that
is suitable for the measurement amplifier. This can be achieved by using some kind
of signal processing; for example, voltages proportional to either displacement or
acceleration can be converted to acceleration by electronic filters and amplifiers.
Both analog and real-time digital techniques can be used since the frequencies are in
the audio range. The reader is directed to control theory and relevant literature [6,7].
23.4 DIGITAL SIGNAL PROCESSING

Real loudspeakers, both electrodynamic and others, are plagued by resonances.
These are due both to unavoidable fundamental resonance of the driver’s mass and
Response (dB)
Response (dB)
0.00 0.00
(ms)
(ms)
0 0.68 0 0.68
–5 –5
delay
delay
1.59 1.59
–10 –10
2.49 2.49
e
–15 –15
Fram
Fram
3.40 3.40
–20 –20
63 125 250 500 1 k 2 k 4 k 8 k 16 k 63 125 250 500 1 k 2 k 4 k 8 k 16 k

(a) Frequency (Hz) (b) Frequency (Hz)
FIGURE 23.8 Cumulative spectral decay of the impulse response of a loudspeaker (a)
unequalized and (b) equalized. (From Dalbjörn, K. and Åsnes, K., Digital filtering for the
improvement of loudspeaker performance in a listening environment, Master’s thesis E91-02,
Division of Applied Acoustics, Chalmers University of Technology, Gothenburg, Sweden, 1991.)
compliance system and to the high-frequency (HF) resonant behavior, which is a

result of wave propagation in the diaphragm and compliance components.
The fundamental resonance is typically in the range of 20–100 Hz for bass voice
coil drivers, and in this frequency range, the reverberation time of rooms may be
long and the temporal resolution of hearing poor. Midrange and HF loudspeakers
have similar technical problems, but the properties of hearing are different at these
frequencies. From the midrange frequencies and higher, the temporal resolution
allows hearing of flutter echo and other comb-filter effects. All of these factors lead
to the need for frequency response equalization.
An efficient way of studying the resonances of electroacoustic systems is to use
“waterfall” plots such as the one shown in Figure 23.8a, which clearly shows the
multitude of resonances over about 2 kHz.
While it is feasible to equalize a few low-Q peaks and dips in the response, the
multitude of resonances shown to exist at medium and high frequencies of this
loudspeaker are virtually impossible to remove without the use of digital signal
processing.
Digital signal processing can be used as a tool to compensate for both linear and
nonlinear distortion in drivers. Before using digital linearization, it is prudent to
apply the “real-time” analog improvements to the loudspeaker such as current drive
technology, bandwidth extension by electroacoustic component synthesis, as well as
linearization of the mechanical and magnet systems. Component aging due to use,
temperature, atmospheric substances, etc., may change the loudspeaker’s properties.
Loudspeaker’s properties may also be a function of placement, transportation, etc.
Because of the many effects any practical signal processing linearization system
should be adaptive. This considerably complicates the process since the loudspeaker
must now either use a self-contained measurement system or be supplied with
computer software for external measurement and linearization. Any linearization
method must also take into account that the loudspeaker is listened to in a room, so
loudspeaker directivity and the damping and diffusion effected by room surfaces
must also be addressed. Averaging over many listening positions may be necessary,
as pointed out previously.
The possibilities for linearization fall into three broad categories: frequency
response or transfer function improvement (direct or reverberant sound), control
of radiated power, and elimination of nonlinear distortion. Informal listening has
shown that one of the prime benefits of linearization is that the loudspeakers in a
stereo or multichannel system can become virtually identical transducers.
23.4.1 Transfer Functions
The problems of direct or reverberant sound equalization were discussed at the
beginning of this chapter. Here we will discuss only the transfer function of the
direct sound. Any measurement of the transfer function of a loudspeaker will include
a minimum-phase part and an allpass part. The allpass part may be thought of as a
pure delay that is seldom considered and that can be eliminated from the equaliza-
tion process unless synchronization to other stimuli is needed.
The transfer function Hpe(ω) of a loudspeaker may be defined as complex ratio
of sound pressure at some point to the input voltage to the loudspeaker system or its
dedicated amplifier. The transfer function is characterized by magnitude and phase:
H pe ( ω ) = H pe ( ω ) e ∠
j H pe (ω )
(23.1)

Linearization then means that the amplitude and phase of the transfer function
are adjusted so that (within the desired frequency range) there will not be any mag-
nitude variation and that the phase variation will increase linearly with frequency,
i.e., have the form of “pure” time delay. A simple inverse filter Hinv(ω) could then be
1
Hinv ( ω ) = (23.2)
H pe ( ω )

The transfer function of the driver or loudspeaker may be measured in several

ways, for example, (1) outdoors with the loudspeaker recessed in a plane, large, hard
surface; (2) in an anechoic chamber that has a lower cutoff frequency than that of the
loudspeaker; and (3) in a sufficiently large room using time-gating techniques. In all
the methods, there will be an influence of the environment on the measured transfer
function. In the first alternative, the diffraction by the cabinet will not be included
correctly, and in the second and third alternatives, the loudspeaker’s radiation will
depend on the modal structure of the chamber and room.
In many cases, the ambition will be to linearize the direct sound’s transfer
function in the range 20 Hz–2 kHz. Few anechoic chambers have a low cutoff
frequency of 20 Hz, so measurements might need to be done outdoors. The LF end of
transfer function measurement using anechoic chambers can be extended by signal
processing techniques but are likely to introduce measurement errors. However,
most listening is done in small rooms, so anechoic measurements are not a good way
of determining the practical transfer function in such rooms.
The transfer function or impulse response measurement can be done using MLS,
chirp, or other techniques to maximize signal-to-noise ratio. The MLS technique
provides direct measurement of the impulse response from which the transfer
function can be calculated but is sensitive to loudspeaker and other system non-
linearities. By extending the length of the measured impulse response, a longer
measurement and possibly a larger measurement chamber can be simulated. Chirp
techniques are based on the use of Fourier transforms to obtain the transfer function
first and from that the impulse response. Chirp techniques have an advantage in that
the nonlinear distortion of the loudspeaker can be measured at the same time as the
frequency response.
The start of the linearization can be said to begin here with the decision of which
frequency range that is to be linearized. There is little use in trying to “squeeze out”
more of a loudspeaker system than it can reasonably provide.
Let us use the loudspeaker frequency response shown in Figure 22.1 as an
example. For this loudspeaker, it is reasonable to define a desired passband that
has its lower cutoff frequency at about 100 Hz since the measured response of this
closed box, near-field monitor falls off by −12/dB octave below 100 Hz. The upper
equalization frequency limit is more difficult to specify but from psychoacoustic
considerations—particularly the natural reproduction of speech and musical
transients—one might want to extend the linear frequency range to 3–5 kHz. We
note that the frequency response of the LF driver starts to fall off considerably
already at 2 kHz so a good alternative would be to use a two-way linearization, one
for each driver and then connecting these using a digital crossover filter, for example,
one that works along constant pressure lines or one that uses Linkwitz–Riley filters
(for Linkwitz–Riley filter properties, see Appendix B).
The LF part of the linearization could then be a bandpass filter function HBPLF(ω)
that covers 0.1–2 kHz for the LF driver with a filter function of, for example,
a second-order highpass Butterworth or Chebyshev characteristic at 0.1 kHz
and a fourth-order Linkwitz–Riley characteristic at 2 kHz. The HF part of the
linearization could be a bandpass transfer function HBPHF(ω) that covers 2–15 kHz
for the HF driver with fourth-order Linkwitz–Riley characteristics at 2 kHz and a
third-order Butterworth characteristic at 15 kHz. The filters would also act as cross-
over filters. A symmetric or asymmetric constant-voltage filter could also be possible
at the crossover frequency of 2 kHz.
The filtering can then be done by two target filters TLF(ω) and THF(ω) given by
H BPLF ( ω )
T LF ( ω ) = (23.3)
H pe ( ω )

H BPHF ( ω )
T HF ( ω ) = (23.4)
H pe ( ω )

One aim in any linearization process is to avoid compensating for deep dips due
to interference. Such dips have little audible relevance when they are present in the
unequalized loudspeaker’s frequency response as mentioned before [1]. Improper
linearization will turn them into annoying audible peaks in the response. Spatial aver-
aging can be used to remove interference dips from transfer function measurements.
Since it is difficult to say at exactly where the listener will be relative to the
loudspeaker, it is advisable to make measurements at many points in the “target” vol-
ume and average these together. Some details are bound to be lost by this procedure,
but response dips that are due to interference will be eliminated. In this way, the
loudspeaker’s frequency response will be equalized, but the location-dependent dif-
fraction and other interference effect may remain. This is a good compromise.
REVIEW QUESTIONS
23.1 Why is it necessary to use different equalization approaches for LF and other
sounds in rooms?
23.2 How can one change loudspeaker frequency response by analog signal
processing?
23.3 What are the advantages and disadvantages of current drive of electrodynamic
loudspeakers?
23.4 Why would it be advantageous to use motional feedback in the case of
electrodynamic loudspeakers?
23.5 How can motional feedback be implemented in the case of electrodynamic
loudspeakers?
23.6 How can one change loudspeaker impulse response by digital signal processing?
REFERENCES
1. Toole, F. T. and Olive, S., The modification of timbre by resonances: Perception and
measurement, J. Audio Eng. Soc., 36(3), 122–142 (1988).
2. Kuttruff, H., Room Acoustics, 5th edn., Spon Press, London, U.K. (2009) ISBN-13:
978–0415480215.
3. Klippel, W., Loudspeaker nonlinearities—Causes, parameters, symptoms, in
Proceedings of the 119th Audio Society Convention, New York (2005), Paper 6584.
4. Mills, P. G. L. and Hawksford, M. O. J., Distortion reduction in moving-coil loudspeaker
systems using current-drive technology, J. Audio Eng. Soc., 37(3), 129–148 (1989).
5. Stahl, K. E., Synthesis of loudspeaker mechanical parameters by electrical means:
A new method for controlling low-frequency loudspeaker behaviour, J. Audio Eng. Soc.,
29(9), 587–596 (1981). Also U.S. Patent #4118600 (1978).
6. De Boer, E., Theory of motional feedback, IRE Trans. Audio, 9(1), 15–21 (1961).
7. Klassen, J. A. and de Koning, S. H., Bewegingstegenkoppeling bij luidspeakers
(Feedback circuits for loudspeakers, in Dutch), Philips Technisch Tijdschrift, Philips
Gloeilampenfabriek, Eindhoven, Vol. 29, pp. 152–161 (1968) ISBN-13: 978-0470094303.
8. Dalbjörn, K. and Åsnes, K., Digital filtering for the improvement of loudspeaker
performance in a listening environment, Master’s thesis E91-02, Division of Applied
Acoustics, Chalmers University of Technology, Gothenburg, Sweden (1991).
24 Headphones and
Earphones
24.1 INTRODUCTION
Personal listening devices such as headphones and earphones offer sound reproduction
that is not subject to the acoustics of the environment. Since the listening room is
eliminated from the sound reproduction chain, the audio rendering can in a sense be
considered more exact. Because headphone and earphone transducers need generate
only relatively small volume velocities, they tend to have less nonlinear distortion
than loudspeakers achieving therefore a more exact sound reproduction.
A disadvantage of personal listening devices is that the sound field is static
around the head; as we move and listen, the sound field stays constant relative to the
listener’s head whereas with loudspeaker sound reproduction the sound field is part
of the environment. Headphones and earphones can also provide sound isolation so
that the sounds of the environment are suppressed.
This chapter will not review these effects but will concentrate on the technical
properties of headphone sound reproduction systems. A wealth of additional material
on headphone technology may be found in Ref. [1] and in the pages of the Journal of
the Audio Engineering Society and the Journal of the Acoustical Society of America.
24.2 CATEGORIZATION
Since there are many types of personal listening devices, it is convenient classify
them for the discussion that follows. Figure 24.1 shows such a common classification
system.
Headphones are worn over the ear and may be subdivided into two groups:
circumaural and supraaural. A circumaural headphone has one or more transducers
mounted in cups that have a reasonably airtight seal against the head. Circumaural
headphones may be designed to be open or closed. A closed headphone will insulate
against outside noise although the resonance between the cup masses and the
headband compliance will reduce insulation at low frequencies. Open-cup designs
have no sound isolation but allow for an environmental acoustic “awareness” and
tend to sound more natural for this reason. A supraaural headphone on the other
hand usually rests on a soft sound-transparent cushion on top of the ear and may be
more comfortable. Open-cup circumaural headphones provide virtually no sound
isolation to the environment.
Earphones, often called “earbuds,” are small transducers that are used close to the
ear canal. Insert earphones, often called “in-ear phones,” are worn in the ear canal
opening but, in contrast to earphones, provide some environmental noise reduction.
451
Circum–aural (closed) Circum–aural (semi-open) Circum–aural (open)
Insert earphone Supra–concha (open) Supra–aural (open)
FIGURE 24.1 Some types of personal listening devices. The transducers in supra-concha and
supraaural headphones rest on sound-transmissive open pore cushions. The insert earphone
transducer is mounted on a soft plug so that it can be inserted comfortably into the ear canal.
All loudspeaker transducer types can be used for headphones and earphones,
although some types are impractical, for example, since they may require high
voltages, have large mass, or be sensitive to humidity and dirt. Low mass is an
important aspect of devices that are to be used while walking or running. The
two most common types for personal listening devices are electrodynamic and
electromagnetic transducers. These can also, in practice, be classified as low-
impedance and high-impedance drivers respectively.
In addition to their “natural” sound isolation, circumaural headphones and insert
earphones can provide further sound isolation by active noise control techniques. If
the device is provided with a microphone, electronic and signal processing circuitry,
it may offer selective “intelligent” listening to the environment, for example, allowing
speech and warning signals to be heard while stopping noise. For special purposes,
such as when the ear canal needs to be left open, a bone conduction receiver may be
a better personal listening device than a headphone [2].
24.3 DESIGN CONSIDERATIONS

Headphones and earphones should have a frequency response tailored to the
properties of the audio signal to be heard. When we listen to sound from loudspeakers
in rooms, the aural impression is formed by the total response of the direct sound,
early reflected sound, and reverberant sound. Since most living rooms have quite
short reverberation times, the direct and early reflected sound will be dominant.
Headphones and Earphones 453
The sound at the listener’s ears depends on the loudspeakers used, their direct sound
frequency response, the response in other directions, the reflection characteristics of
the room surfaces, scattered sound from objects, etc. The sound will also depend on
the reflection by the listener’s body and head.
The sound in the listener’s ears will additionally depend on the angle of the
incoming sound waves toward the head [3,4]. The various resonances of the concha
and ear canal are excited differently depending on the angle of incidence of the
sound at the ear, similar to the directivity of an electronic antenna. This excitation
cannot be fully simulated using headphone excitation.
The head-related transfer functions (HRTFs) are used to characterize the
influence of the listener’s head and torso on sound waves that arrive at the listener.
The HRTFs are strictly defined as the complex ratios of the sound pressure at each
of the listener’s ears to that of a plane wave at the listener’s place in the listener’s
absence. The term is often used synonymously with the head-related impulse
responses to which they are related through the Fourier transform. Each person
is used to his or her own HRTFs, and listening to those of others—using, for
example, auralization—upsets the externalization and localization of sounds so
binaural sound reproduction relies on small personal sound field cues to make
the impression of the surrounding sound field and to appear natural and outside
the head.
Measurements of HRTFs are complicated by a lack of agreement of which
point at or in the ear that is to be used for the measurement. One reasonable point
of measurement is at the entrance of the ear canal. This point could be used as a
reference point for in-ear recording and measurement [5].
In practice, HRTFs are measured using a small sound source, with little diffraction,
that provides a smooth spherical wave field at 1.5–2 m from the listener’s head. It can
be shown that the wave field curvature at this and greater distance has little influence
on measured HRTFs. The HRTFs depend on the angle of arrival of the sound waves,
and each angle has its unique set.
Conventional listening to sound by natural sound sources and sound reproduced
by loudspeakers also relies on the properties of the individual’s HRTFs. As an
example, consider listening to stereo sound reproduction situation with the listener
and the loudspeakers forming an equilateral triangle. A sound in the front, recorded
by flat frequency response microphones but reproduced by loudspeakers at a different
lateral angle, will sound different when listening over loudspeakers because of the
frequency response differences induced by the body.
Figure 24.2 shows an example of the frequency response of the HRTF for frontal
sound incidence as well as for diffuse-field incidence for one ear of a manikin head.
Clearly, the frequency responses are far from linear. Linear frequency response for
headphones (or earbuds or in-ear phones) is therefore not a realistic design target.
Many commercially successful headphones use an equalization curve similar to the
diffuse field shown in the figure.
Headphones and earphones should be able to handle music peak sounds with little
distortion, although including some type of “soft” limiting might be useful to reduce
the risks. “Hard” limiting the signal will introduce nonlinear distortion that should
be avoided. There are also other factors to be considered.
20
Frontal incidence
15 Diffuse field incidence
Sound pressure level increase
at entrance to ear canal (dB)
(1/3 octave band data)
10
–5
31 63 125 250 500 1k 2k 4k 8k 16 k

Octave band center frequencies (Hz)
FIGURE 24.2 The frequency response of the HRTF for an ear on a manikin head measured
in an anechoic chamber for frontal sound field incidence (narrowband data) and in a
reverberation chamber for diffuse-field incidence (third-octave band data). No torso was used
with the head for this measurement; if used, it would have increased levels in the 250 Hz
range. (From Kleiner, M., Acustica, 41, 183, 1978.)
The smaller nonlinear distortion of headphones and earphones sometimes

leads listeners to use sound levels that are higher than those in loudspeaker sound
reproduction. The lack of room reverberation contributes to this practice since the short
peak sounds of music are not perceived as strongly as they would had reverberation
been present. Popular music by amplified instruments such as electric guitars can
produce significant sound level peaks with durations as short as a few milliseconds.
Such peaks are not perceived as loud as would a continuous sound with the same level.
For a 5 ms long tone burst at 1 kHz, the perceived-level “deficiency” is about 20 dB
so the peak/rms ratios are likely to be about 10–20 dB higher than that in loudspeaker
listening.
If listeners adjust the average reproduced sound level to about the same
subjectively satisfying loudness for both types of listening, it is likely that peaks are
likely to be much stronger than while listening to loudspeaker-reproduced sound.
Short-duration sounds are therefore more likely to contribute to the risk of hearing
impairment or even damage. This is also a serious problem in mobile phone use
where the phone is also used for music reproduction over headphone/earphones. The
risk can be minimized by including some form of reverberation added by signal
processing in the signal chain or by the headphone/earphone electronics, but this
may not be practical.
Listening with only one ear, as when using a cell telephone or telephone receiver,
requires about 10 dB higher sound pressure level to give the same subjective
impression of loudness as when using two ears. This clearly further increases the
risk to hearing on the ear used.
24.4 ACOUSTIC ENVIRONMENT

Two types of acoustic load conditions are of particular interest to the designer of
personal listening devices. The transducers of closed circumaural headphones and
insert earphones drive what is essentially a small closed cavity with some leakage.
Because of head shape characteristics, hair, and the bows of glasses, leaks are a larger
problem in headphone than in earphone use. The compliance of human skin around
the ears is particularly difficult to simulate [6,7]. Supraaural headphones and earbuds
are similar to dipoles that operate in an “open” environment, but that are listened to
at minimal distance.
Headphones are usually much smaller than the wavelengths of sound over most
of the audio frequency range. This means that the transducer (depending on con-
struction) can act as an almost ideal piston changing the volume of the cavity. The
impedance of the cavity as seen by the transducer can be described as a closed but
somewhat lossy volume by the expressions given in Chapter 3 up to frequencies of
about 0.8 kHz. Above this frequency, the sound pressure at the concha will start to
be a function of the shape and impedance properties of the cavity formed by the ear
and the headphone. It is important to note that the contact pressure provided by the
head band will influence the compression, and thus the flow resistance, of any pads
used to seal the leak between head and headphone. In addition, the compliance of the
head band and the mass of the headphones will cause a resonance, usually around a
few hundred hertz.
The cup interior may be empty or it may contain some open cell foam or other
sound-absorbing material. If the cup is empty, it is difficult to avoid strong modal
behavior in the front chamber of the headphone. One of the curves in Figure 24.4 shows
an example of the frequency response of an electrostatic circumaural headphone that
has a large diaphragm. It is seen that in spite of the large diaphragm that drives most
of the cavity in phase from one side, the modal behavior is still strong at frequencies
over 3 kHz.
The use of a correctly designed artificial ear or coupler is essential for an accurate
measurement of the response of headphones and earphones [8]. Many couplers are
designed for telephony and will not adequately model the acoustic properties of
the ear to be useful for the full frequency range of audio. Couplers for telephony
Rear chamber
Transducer
Front chamber
Cup
Ear cushion
Concha
Ear canal
Ear drum
FIGURE 24.3 Cup, transducer, and ear combination.

30
Insert earphone
Electrostatic circumaural
20
10
–10
–20
20 50 100 200 500 1k 2k 5k 10 k 20 k
Frequency (Hz)
FIGURE 24.4 The frequency response of some high-quality head- and earphones measured
using a simple small-volume coupler to avoid influence of the ear canal impedance.
FIGURE 24.5 The anthropometric KEMAR manikin is a head and torso simulator for
binaural sound recording and measurement based on worldwide averaged human male and
female head and torso dimensions. (Photo courtesy of G.R.A.S., Holte, Denmark)
equipment usually have a volume of about 1–2 cm3. Sometimes such couplers are
augmented with various resonant circuits. For audio frequency range measurements,
it is better to use an anthropometric head manikin, an example of which is shown
in Figure 24.5. The manikin shown has an artificial ear with a soft rubber pinna,
as well as a concha, ear canal, and eardrum simulator but does not simulate the
compliance of human flesh [9]. Also important to remember is that human pinnae
Air volume to
simulate eardrum
compliance
Air volume to Acoustic resistance
simulate eardrum
Ear canal tube compliance Ear canal tube
Concha Concha
Pressure
sensing
microphone
Helmholtz resonator Pressure

side branches × 4 sensing
(only 2 shown) microphone
FIGURE 24.6 Two types of couplers for headphone/earphone measurement and calibra-
tion as well as binaural sound recording. To the left is the Zwislocki coupler design. (After
Zwislocki, J.J., An ear-like coupler for earphone calibration, Report LSC-S-9, Laboratory of
Sensory Communication, Syracuse University, Syracuse, NY, 1971.) On the right is a simple
coupler. Both allow the concha and ear canal modes to be relatively similar to those in a real
head. (After Kleiner, M., Acustica, 41, 183, 1978.) The latter type is easier to equalize for use
in binaural sound recording.
vary in size so several sets of artificial pinnae are necessary. Another problem
in using anthropometric simulators is that it may be difficult to ensure the same
placement of the headphone at every measurement so repeated measurements need
to be made and averaged.
The ear canal and eardrum Zwislocki-type ear simulator offer very good ear
canal entrance impedance simulation and are designed as shown to the left in
Figure 24.6 [10]. On the right is shown a simple coupler that still allows the con-
cha and ear canal modes to be relatively similar to those of a headphone on a real
head [11].
Any headphone design will need to be optimized to the operating
distance between head and headphone. This distance is determined not only by
head shape and dimensions but also by head band clamp force and the design
of the headphone cup and cushions. This may be a problem for supraaural head
phones. Figure 24.7 shows the large influence of leakage on the resulting frequency
response of a headphone. Clearly minimum leak sensitivity is an important aspect
of headphone design.
Since headphones and earphones are small devices and operate into small volumes
compared to audio range wavelengths, it is practical to simulate the entire system
including the exact transducer mechanism using finite element method modeling.
Finite element or boundary element modeling can be used to great advantage for
open or supraaural headphones and earbuds [12].
30
Electrodynamic, supra-aural, normal pressure
Electrodynamic, supra-aural, extra pressure
20
10
–10
–20
20 50 100 200 500 1k 2k 5 k 10 k 20 k
Frequency [Hz]
FIGURE 24.7 An example of the influence of leakage due to different contact pressures on
the frequency response of a supraaural headphone using a foam plastic seal.
24.5 ELECTRODYNAMIC HEADPHONES

Electrodynamic low-impedance headphones may be designed for various load
impedances such as circumaural open or closed cups and cushions or for supraaural
cups and cushions.
The closed cup corresponds to the closed-box loudspeaker but with a different
load impedance since there is little sound radiation. The acoustic impedance analogy
will be as shown in Figure 24.8.
Here R AB and CAB are the impedance components of the back cavity, Z AE front
cavity and ear impedance, and R AL and MAL the leakage between head and ear
cushion. The ear impedance Z AE is capacitive at frequencies below about 0.8 kHz
since the volume is much smaller than the wavelength. The sound pressure p̲E in the
cavity is then the “voltage” over this component. To handle higher frequencies, one
e Bl
G
(REG + REC) SD MAD CAS RAS
U
D
RAB (Bl)2 MAL

(REG + REC)SD2
p Z
E AE
RAL
CAB
FIGURE 24.8 The acoustic impedance analogy for a closed-back electrodynamic

headphone showing the impedance components due to front cavity and ear impedance Z AE
and leakage between head and ear cushion R AL and MAL .
e Bl
G
(REG + REC)SD MAD CAS RAS
U
D
MAC
(Bl)2 MAL
AR
CAC (REG + REC)SD2 p Z
E AE
MAR RAL
FIGURE 24.9 The acoustic impedance analogy for an open-back electrodynamic

headphone.
e Bl
G
(REG + REC)SD MAD CAS RAS RAF
U
D
(Bl)2
ARB ARF
(REG + REC)SD2
ZAEC
MARB MARF
FIGURE 24.10 The acoustic impedance analogy for a supraaural electrodynamic

headphone.
can build a network of inductances and capacitances to handle the unavoidable wave
propagation in the cavity, concha, and ear canal or one can include the ear canal as
a transmission line the circuit.
The open-cup version corresponds to a conventional loudspeaker with the
radiation impedance components ℜAR and MAR and is shown in Figure 24.9. Here
the “front” cavity and the ear form the interior impedance and components. MAC
corresponds to the air in the grille and CAC to the interior volume between the grille
and the transducer.
When the headphone rests on the ear—as does a supraaural one—its impedance
analogy essentially corresponds to that of a conventional loudspeaker with the
radiation impedance components ℜARB and MARB and is shown in Figure 24.10. Here
the “front” cavity and the ear form the interior impedance and components ℜARF and
MARF. Usually, the part of the earphone that rests against the ear is made from open
cell foam so it adds a resistive component R AF to the impedance circuit. Finally, Z AEC
corresponds to the air in the concha and the ear canal. Note that because of the head,
the impedances ℜARB, MARB, ℜARF, and MARF will be different from those sensed by
a free disk in air.
Leak across ear bud to the outside

e Bl MAL RAL
G
(REG + REC)SD U MAD CAS RAS MACF RACF
D
MACB RACB (Bl)2

AR
CACB (REG + REC)SD2 CACF ZAED
MAR
Back cavity Front cavity Ear
Electrodynamic driver Ear canal
and grille and grille drum
Radiation impedance
FIGURE 24.11 The acoustic impedance analogy of an electrodynamic earbud placed close
to the ear canal opening.
Bass-reflex port
Ear cushion
and auxiliary
diaphragm
Tuning port Diaphragm

Electrodynamic
transducer
Cavity
FIGURE 24.12 An example of a modern electrodynamic headphone using a bass-reflex

system. (After Poldy, C.A., Headphones, in Borwick, J. (Ed.), Loudspeaker and Headphone
Handbook, 3rd edn., Focal Press, Oxford, U.K., 2001.)
This impedance analogy also applies to the earbud. In the case of the earbud
analogy, however, it is often necessary to consider the impedances that are formed
by the grilles and cavities inside the bud as shown in Figure 24.11.
It is also possible to design an electrodynamic headphone similar to a ported
loudspeaker enclosure as shown in Figure 24.12. The parasitic resonators enhance
the low-frequency performance in this design similar to the operation of the port in
bass-reflex loudspeaker enclosures.
A special type of supraaural electrodynamic headphone is the isodynamic headphone
“dipole” design shown in Figure 24.13. In this type of headphone, the transducer’s magnet
and voice coil have been replaced by an electrically conductive meandering foil path on
a thin plastic film placed in a distributed magnetic field. The isodynamic headphone is
similar to electrostatic headphones in that the diaphragm is driven with approximately
the same force per unit area. As a consequence of the light weight necessary for this
type of headphone, the magnetic field is weak and the headphone needs higher current
than other types. The primary advantage of this design is that there are no “rocking
mode” resonances, only the resonances of the tensioned diaphragm. Since the film has
low mass, these resonances will be well damped by the radiation impedance.
Cushion
SN NS
Magnets NS SN
Plastic foil SN NS
NS SN
Conductive foil strips
SN NS
Acoustic resistance Acoustic resistance
NS SN
SN NS
NS SN
FIGURE 24.13 The isodynamic headphone is a distributed force and balanced magnetic
field design.
24.6 ELECTROMAGNETIC HEADPHONES

Electromagnetic earphones (receivers) were used extensively in old-style
telephone equipment because of their sturdiness and high sensitivity. They then
had closed backs and were used as supraaural headphones, resting on the ear but
without a cushion. Because of the high stiffness and mass of the diaphragm, the
transducer’s a resonant. An advantage of electromagnetic earphones is that they
can be made with a high electric impedance since the coil winding does not need
to move. This high sensitivity is an important factor in the design of hearing
aid equipment.
Insert earphones often use magnetically balanced reed armature transducers as
shown in Figure 24.14. Although the design shown is based on a reed, other “fully
mechanically balanced” designs also exist. Up to its first resonance frequency,
Suspension Diaphragm (piston) Suspension Pivot point Diaphragm (piston)

and drive rod and drive rod
Center
N pole piece N
S
N
To ear canal To ear canal
N S
S
Rocking
Magnets armature
Drive coil Air Flexing balanced Drive coils Magnet Air cavity
cavity armature
(a) (b)
FIGURE 24.14 Two types of electromagnetic insert earphones using a balanced magnetic
field: (a) with a reed armature driving the piston, (b) using a balanced “rocking” armature for
a lower resonance frequency. In (b), the restoring force is provided by torsion arms.
SD2LEC
(KEM)2
–SD2CME M
AD CAS RAS
UD
(KEM)2 Armature Tube to Ear canal
eGKEM CAF ear canal ZAED
(REG+REC)SD2
(REG+REC+jωLEC)SD Volume
CAB in front Ear
of piston drum
Back cavity
FIGURE 24.15 The acoustic impedance analogy of a balanced armature earphone with ear
canal and ear drum impedances.
20
10
–10
–20
–30
20 50 100 200 500 1k 2k 5k 10 k 20 k
Frequency [Hz]
FIGURE 24.16 An example of a frequency response curve for an electromagnetic

headphone. (After data from Knowles BK [balanced armature] series, Unknown coupler,
Knowles Acoustics, Itasca, IL.)
the reed will be a compliance component. The response will show multiple
resonances due to the higher-order modes of the reed.
The balanced armature transducer’s drive mechanism was discussed in Chapter 9.
The acoustic impedance analogy of an insert earphone based on the drive mechanism
shown in Figure 24.14a is shown in Figure 24.15 [13]. Here the armature is drawn as
a simple resonant system so the analogy is only valid to frequencies slightly above
the first armature resonance.
Figure 24.16 shows a typical frequency response curve for a balanced armature
reed transducer for hearing aid use. By tuning the armature so that its first resonance
occurs at a high frequency, for example, around 5 kHz, most of the audio range can
be covered quite well.
The rocking armature transducer’s drive mechanism is mechanically similar to
that of electrodynamic devices in that the resonance frequency can be made much
lower than that in the reed transducer.
24.7 PIEZOELECTRIC HEADPHONES

Piezoelectric headphones may be made using either piezoelectric films or piezoelectric
ceramics. Small piezoelectric transducers are usually designed as earbuds or insert
earphones, although direct drive designs using a metal disk having piezoelectric
material on one or both sides is also possible. Such disks are very resonant, and they
are usually avoided in high-quality audio equipment.
Piezoelectric film loudspeakers are discussed in Chapter 25. Piezoelectric films
are made so that the piezoelectric effect is between an electric field applied over the
film and a length extension in the film plane. This requires the film to be slightly
curved to vibrate in air and function as a headphone [14]. An example of a piezoelec-
tric film headphone of the circumaural type is shown in Figure 24.17.
24.8 ELECTROSTATIC HEADPHONES

Two types of electrostatic headphones are common: those based on charge
provided by bias and those based on permanent charge, as discussed in Chapter 9.
From a mechanical viewpoint, both types are identical in that they use a moving
film, tensioned and mounted at its edges to a rigid frame. The frame holds one
acoustically transparent electrode on each side of the film as shown in Figure 24.18.
Unless the transducer uses a push-pull balanced drive, the nonlinear distortion
becomes high and the device is not used for high-quality audio. In commercial
push-pull balanced drive designs, one often finds that the major part of the nonlinear
distortion comes from the transformer that is used to convert low-level audio to the
high voltages necessary to provide the dynamic electric field between the electrodes
of the transducer.
Besides the open-back circumaural designs discussed, the electret principle can
also be used for closed-back insert ear phones. Closed-back electrostatic headphones
typically have many acoustic resonances because of the back cavity air volume.
These are not damped by radiation as in the case of the open-back headphones, but
need to be damped using an acoustic resistance material in the cavity.
Cushion
Electrodes
Piezoelectric
plastic foil
Protective grilles
FIGURE 24.17 A circumaural piezoelectric headphone.

Cushion
Electrodes
Plastic foil
Protective grilles
FIGURE 24.18 A circumaural electrostatic headphone.
Outside leak across headphone
MAL RAL
–CAE MAD CAS MACF RACF
UD
MACB RACB
AR pD
CACB CACF ZAE
MAR
Back cavity Electrostatic driver, diaphragm Front headphone
and grille mass and compliance cavity and grille
Ear load
impedance
FIGURE 24.19 The acoustic impedance analogy of a push-pull circumaural open-back

electrostatic headphone.
The acoustic impedance analogy of the open-back headphone has similarities to

those of electromagnetic and isodynamic headphones and is shown in Figure 24.19.
REVIEW QUESTIONS
24.1 Define the HRTF and the head-related impulse response.
24.2 What would be the difference between free-field and diffuse-field HRTFs?
24.3 How will coupler impedance characteristics influence the measurement of the
characteristics of (a) supraaural and (b) circumaural headphones?
24.4 Why does the piezoelectric film need to be curved in the headphone shown in
Figure 24.17?
24.5 Where should the transfer function or frequency response be measured for
different types of headphones to be the most representative for listening to
music over loudspeakers in the home?
PROBLEMS
24.1 An electrodynamic circumaural headphone is mounted on a Zwislocki-type
coupler.
Task:
Draw the acoustical impedance analogy for the system for frequencies where
the coupler can be considered an acoustic compliance.
24.2 Assume that you have a small electrodynamic loudspeaker that has a flat
(pressure) frequency response in the relevant audio range and want to use it in
an intercom as both a loudspeaker and a headphone.
Task:
What will be its microphone response characteristics?
24.3 Most personal phones have the electrodynamic transducer mounted inside the
cover so that they face (a) to the front a volume inside the phone that couples
to the ear through small holes in the cover, and (b) to the back the volume of
the phone that in turn is open to the surrounding air through some small vent
holes.
Task:
Draw an analogy that includes the electrical, mechanical, and acoustical sides
of the system including the compliance of the front headphone chamber and
the concha of the user’s ear (see Figure 24.3).
24.4 In contrast to the situation when dealing with many loudspeaker transducers,
the front and back diaphragm areas must be separated in the analysis of
electrodynamic transducers for headphone since much of the volume velocity
drives the volume in the magnet system.
Task:
Draw an analogy that includes the electrical, mechanical, and acoustical sides
of such a system that includes separate impedances for the back cover and
magnet air volumes.
24.5 An earphone coupler has the schematic acoustical representation shown in the
figure shown below. The component values are as follows: CA1 = 1.3 · 10 −11 m5/N,
CA2 = 5.3 · 10 −11 m5/N, CA3 = 1.8 · 10 −11 m5/N, R A1 = 6.5 · 106 Ns/m5, R A2 = 2.0 · 107
Ns/m5, R A3 = 5.0 · 108 Ns/m5, MA1 = 1.0 · 104 kg/m4, MA2 = 5.0 · 102 kg/m4.
Task:
Draw a sketch of how such an impedance could be implemented in practice.
MA2 MA1
RA3 CA3 RA2 RA1
CA2 CA1
REFERENCES
1. Poldy, C. A., Headphones, in Borwick, J. (Ed.) Loudspeaker and Headphone Handbook,
3rd edn., Focal Press, Oxford, U.K., pp. 585–692 (2001) ISBN-13: 978-0240515786.
2. Håkansson, B., The balanced electromagnetic separation transducer: A new bone
conduction transducer, J. Acoust. Soc. Am., 113(2), 818–825 (February 2003).
3. Wightman, F. L. and Kistler, D. J., Headphone simulation of free-field listening I:
Stimulus synthesis, J. Acoust. Soc. Am., 85, 858–867 (1989).
4. Shaw, E. A. G. and Teranishi, R., Sound pressure generated in an external-ear replica
and real human ears by a nearby point source, J. Acoust. Soc. Am., 44(1), 240–249
(1968).
5. Hammershøi, D. and Møller, H., Binaural technique—Basic methods for recording,
synthesis, and reproduction, in Blauert, J. (Ed.) Communication Acoustics, Springer,
Berlin, Germany, pp. 223–254 (2005) ISBN-13: 978-3540221623.
6. Håkansson, B. et al., The mechanical point impedance of the human head, with and
without skin penetration, J. Acoust. Soc. Am., 80(4), 1065–1075 (1986).
7. Bauer, B. B., Rosenbeck, A. J., and Abbagnaro, L. A., External-ear replica for acoustical
testing, J. Acoust. Soc. Am., 42, 204 (1967).
8. Delaney, M. E., The acoustical impedance of human ears, J. Sound Vib., 1, 455–467
(1964).
9. Burkhardt, M., Manikin measurements, Manikin Measurements Methods Conference
Proceedings, Zürich and Washington (1976), G.R.A.S. Sound & Vibration, Holte,
Denmark. Available from www.gras.dk (sampled 2011).
10. Zwislocki, J. J., An ear-like coupler for earphone calibration, Report LSC-S-9,
Laboratory of Sensory Communication, Syracuse University, Syracuse, NY (1971).
11. Kleiner, M., Problems in the design and use of dummy-heads, Acustica, 41, 183–193
(1978).
12. Biba, D. and Opitz, M., Development of a finite element headphone model, Proceedings
of the 122nd Audio Engineering Society Convention, Vienna, Austria, paper 7103 (2007).
13. Merhaut, J., Theory of Electroacoustics, McGraw-Hill College, New York (1979) ISBN-
13: 978-0070414785.
14. Tamura, M. et al., Electroacoustic transducers with piezoelectric high polymer films,
J. Audio Eng. Soc., 23(1), 21–26 (1975).
25 High-Frequency
Transducers
25.1 BANDWIDTH AND POWER

25.1.1 Bandwidth
High-frequency transducers, such as high-frequency-range loudspeakers and
transducers for ultrasound, are usually designed for much smaller bandwidths than
those common in audio. For audio applications, one usually needs a bandwidth of
about three decades, that is, 20 Hz–20 kHz. Many high-frequency audio transducers
are similar to ultrasound transducers and only cover the top two octaves of the
audio range. Ultrasonic scale modeling for room acoustics, also needs a large
bandwidth, although typically only two decades of bandwidths will be achieved.
Some common uses are for airborne ultrasound in sonar applications for sensing
and localization.
A major characteristic of sound propagation at ultrasonic frequencies in air
are the large propagation losses, already discussed in Section 3.7, that limit practical
applications to frequencies below 200 kHz. A small bandwidth, such as 10% of
the center frequency, is, for example, often sufficient in static sensing applications.
Various forms of sonar applications can need a wider bandwidth. Some of the
bandwidth-induced limitations in sensitivity and spatial resolution can be overcome
by selecting a suitable signal and using digital signal processing.
25.1.2 Transducer Choices
High-intensity technical applications of ultrasound are common for dispersion of
fluids, degassing of fluids, fluid cleaning of the surfaces of solid objects, welding and
riveting of plastics and metals, as well as agglomeration of particles in gases [1]. An
application of high-intensity airborne ultrasound for audio is found in parametric
loudspeakers.
Conventional audio transducers have fairly limited power-handling capacity
since acoustic applications typically need very little acoustic power and the power
amplification in the audio range is very easy and inexpensive to achieve. Ultrasonic
cleaning, on the other hand, uses hundreds of watts of acoustic power to be effective
in large cleaning vats. Sonar ranging needs to reach far from mobile platforms and
needs to conserve power, that is, it needs to be highly efficient.
To generate large acoustic power, it is usually necessary to use specialized res-
onant transducers. Many high-power transducers for ultrasound use piezoelectric
467
and magnetic materials. Because of their impedance characteristics, it is usually

necessary to use various mechanical or electrical matching and transformation net-
works. Bending wave transducers can also generate large sound power when suitably
coupled to the air [2]. When very high acoustic power at ultrasonic frequencies is
required in air or other gases, whistles, such as Hartmann whistles, may be conve-
nient [3].
For low-power applications, many types of transducers are available. Because
transducer membranes must have very low mass to operate efficiently at high
frequencies, many ultrasound transducers use very low mass, thin plastic films.
Examples of such transducers are capacitive Sell transducers [4] and piezoelectric
film transducers [5–7]. More conventional are whizzer cone and horn loaded
piezoelectric ceramic transducers [8]. In the ionophone, on the other hand, the
membrane has been eliminated entirely and replaced by a modulated volume of
ionized air [9,22–25].
25.2 SEMI-RESONANT CAPACITIVE TRANSDUCERS

Electrostatic transducers are usually used where a wide frequency range high-
quality sound reproduction is desired. Electrostatic loudspeakers intended for the
audio range must be very large to achieve reasonable sound levels. Electrostatic
headphones tend to be circumaural designs so that they can generate sufficient sound
levels for listening to music. For ultrasound, the unbalanced Sell transducer is very
common since it can be made to have high sensitivity and has a built-in feature in the
form of the frequency doubling sound generation, which is a result of its nonlinearity
when unbiased.
A nonlinear capacitive Sell transducer has been introduced in Chapter 9. The
transducer type is very useful for generating sound in gases at ultrasonic frequencies
because of its semi-resonant properties. Recently, Sell transducers for ultrasound
have been called capacitive micromachined ultrasonic transducers (CMMUT). Sell
transducers were used historically in tabletop radio sets with vacuum tube amplifiers
as inexpensive high-frequency audio transducers.
Figure 9.16 shows the basic idea of a Sell transducer. A schematic view of cross sections
through two designs is shown in Figure 25.1. The film needs to have one isolating side
and one conductive side. The second electrode can be an etched or micromachined solid
Air pockets Metallized isolating flexible film diaphragm Air pockets
Solid electrode
(Random depth back electrode) (Micro-machined back electrode)
FIGURE 25.1 Two types of electrodes used for Sell transducers.

High-Frequency Transducers 469
electrode, but because of dust and micro-irregularities in the surface of the electrode,
almost any electrode surface will work to some extent. An etched film can also be used.
One property of the quasi-random airspace pits is that the resonance frequency of each
“cell” will vary over the surface of the transducer [10]. This will help in broadbanding the
transducer but at the same time also reduces the sensitivity.
The transducer is ideally suited for use as a two-dimensional shaded array. By
using micromachined electrodes, the resonance frequency and the sensitivity of the
transducer can be accurately designed and changed over the area of the transducer for
array shading, thus controlling directivity. The electrode can be split into concentric
rings to obtain a focused transducer similar to a Fresnel plate.
Sizes for Sell tranducers typically range from a few to tens of square centimeters
but always should be many wavelengths large at least in one dimension. Center fre-
quencies from a few kHz to several tens of MHz are used. Bandwidths range from
about 20% to 100% of the center frequency.
25.2.1 Acoustical Properties
In contrast to conventional electrostatic transducers, the Sell transducer uses a limp,
lightly stretched film membrane acting as a movable electrode. The film is made of an
electrically isolating material that has been made electrically conductive on one side.
The film is mounted with its isolating side toward the back electrode which has a series
of wells or other irregularities. The film patches are free to move against the air in the
wells and form a multitude of first-order resonant mass-spring systems. The resonance
frequency f0 of the system (or cells) is approximately given by Equation 25.1.
1 c ρ
f0 ≈ ≈ (25.1)
2π M AC A 2 π m′′dair

Here, m″ is the mass per unit area of the membrane, and d is the effective distance
between the membrane and the back electrode. The practical resonance frequency
will differ from that given in the equation since the effective volume depends not
only on the air tapped between the electrodes but also on the membrane tension. The
tension is to some extent a result of the attraction force of the electric field between
the electrodes. In practice, the difference is usually negligible since the electric field
cannot be very strong.
25.2.2 Damping
The damping of the resonance is generally achieved by the radiation impedance.
Because of the damping, the frequency response is usually fairly flat over a wide
frequency range. If all losses are due to sound radiation (the transducer being large
compared to the wavelength), the Q factor of the resonance is as given in Equation 25.2.
2πf0 m′′
Q= (25.2)
ρc
It is common to find Q factors in the range 1–3.

25.2.3 Electrical Properties

Electrically the transducer appears as a capacitance with its value being determined
by the well shape, the dielectric coefficients of air and film, possible bias voltage,
and cables. The sensitivity is low because of the weak electromechanical coupling.
Conventional linear amplifiers may have difficulty driving a capacitive load and
some form of electric impedance matching network may be needed to optimally
drive the transducer.
The electroacoustical analogy for the Sell-type transducer is the same as that for
the condenser microphone studied in Chapter 14. As with the conventional elec-
trostatic transducer described in Chapter 9, there is a risk of membrane collapse
which usually can be avoided by pre-tensioning the membrane. Since the size of the
air-filled pits can be made very small, the risk is usually small since there is a limit
to the curvature of the film due to its bending stiffness. The capacitance CEO of the
transducer is [11]
ε film ε 0 S
CEO = (25.3)
ε 0 d film + ε film dair

Here, dfilm is the membrane thickness, d the effective air space thickness, and S the
transducer area. The dielectric constant of the insulating film membrane is ɛfilm. The
Sell transducer is nonlinear if used without a bias voltage. When the bias voltage E 0
is large compared to the signal voltage and the membrane excursions small com-
pared to d, the transformation factor KEM will be [11]
E0 ε film ε 02 S
K EM = 2 (25.4)

(ε d
0 film + ε film dair )
As a receiver the Sell transducer needs to be biased the same way as conventional
condenser microphones for more linear response. The film exterior is usually
attached to ground and the bias voltage is applied to the back electrode.
When the transducer is used without bias as a transmitter, the force generation
mechanism will behave as a full-wave rectifier (giving an infinite number of
overtones of the drive frequency) unless the transducer is provided with a bias
voltage as discussed in Chapter 9. This means that the Sell transducer can be used
as an acoustomechanical filter to select the overtone desired. The Sell transducer’s
mechanical resonance reduces the need to remove the overtones generated by the
rectification process and can be used to select the overtone desired, thus simplifying
the design of the drive electronics. The bias voltage can either be applied separately,
as shown in Chapter 9, or be provided when the ultrasonic signal is demodulated. The
signal fed to the CMMUT can, for example, be pulse width or pulse rate modulated
and demodulated by the integrating action of the Sell transducer’s low-pass or band-
pass filter properties. The transformer is optional and only used when the voltages
needed are higher than those that can otherwise be provided from the drive electronics.
The film is usually made conductive on one side by sputtering metal onto the
film or by painting or spraying the film with an electrically conductive paint. The
drawback is if the film dielectric breaks down creating a short circuit between
the two electrodes, no part of the film will function because of the high electrical
conductivity obtained by either of these two treatments. This can be avoided by
making the film (or backplate) only weakly conductive so that a local short circuit
will only affect a minor part of the film. Of course, the sensitivity and directivity
will be affected in this case but the transducer will still function. The polyamide
and polytetrafluoroethylene films used in CMMUT manufacture typically have a
bulk dielectric strength of about 20 kV/mm, but the dielectric strength depends on
material thickness and on whether the voltage is AC or DC, so manufacturer’s data
should be consulted; a starting point for 10 μm polyamide film is 2 kV [12], and for
polytetrafluoroethylene film about 1 kV.
It can be shown that both the static and dynamic performance of the Sell
transducer can be affected by the membrane metallization, and this should be taken
into account in CMMUT design.
25.3 PIEZOELECTRIC TRANSDUCERS

25.3.1 Introduction
The piezoelectric properties of some materials were discussed in Chapter 9. Most of
the discussion here follows that of Ref. [13].
Quartz is a single crystal piezoelectric material that is primarily used for frequency
stable oscillators. Lead zirconium titanate, Pb(Ti,Zr)O3 (also called “PZT”),
ceramic is mostly used in sensors and transducers since it has excellent piezoelectric
properties. Polycrystalline piezoelectric materials such as barium titanate, BaTiO3,
are also widely used. All of these materials have mechanical properties that make
them difficult to use because of large impedance mismatch to air. Some form of
mechanical transformation is needed.
Two types of piezoelectric transducers dominate for use in transmitting and
receiving airborne ultrasound, those that are based on piezoceramic bars, disks, or
sheets using PZT and couple to a diaphragm using a point force, and those that use a
distributed force in a piezoelectric film diaphragm.
25.3.2 Piezoceramic Bars

Let us first study a piezoelectric bar as that shown in Figure 9.22. One end is fixed,
the other end being free. We can set up longitudinal vibration in the bar by applying
an alternating voltage between its sides. What will be the movement of the bar as a
function of the voltage and the properties of the bar?
The movement at the end at y = ly can either be allowed to radiate by itself or
be coupled to a diaphragm to improve the radiation. By joining two bars along the
y-axis—operating with opposite polarity—one can make a bimorph bar that will
vibrate strongly in the x-direction at y = ly.
The resonance conditions for compressional waves in the bar can be obtained
just as for the acoustical standing waves along the length of a narrow air-filled tube.
Assume the longitudinal standing wave to have a vibration displacement η along the
y-axis described by
η( y, k ) = Ae − jkq y + Be jkq y (25.5)

where kq is the wave number for longitudinal waves in the y-direction in the bar.
Assume that the end at y = 0 is kept fixed and only the end at y = ly is moving.
The sides of the bar at x = 0 and x = lx are metallized. The voltage ex is applied at the
metallized sides. If the moving end is much larger than the wavelength of the sound,
it will see a real-valued radiation impedance
Z MR ≈ ρc lxlz (25.6)

If the moving end is small compared to wavelength, the relevant impedance value
can be inserted. In most cases, at audio frequencies, the bar end will be coupled to
a piston or diaphragm and the impedance will be the transformed radiation imped-
ance. The boundary condition for y = 0 gives
A = − B (25.7)
The movement can now be described by
η( y, k ) = −2 jA sin ( kq y ) (25.8)

After some further calculation, one obtains the displacement amplitude at the first
resonance as
K ME eˆx
2 jA = (25.9)
ω1ρclxlz
The piezoelectric transformation factor K ME is
d l
K ME = s12 z (25.10)
22

At the lowest resonance frequency, the bar’s length ly is equal to one quarter of the
wavelength in the crystal
ly
λ q1 = (25.11)
4
so its natural resonance frequency can be calculated from
cq 1 1
f1 = = (25.12)
4ly 4ly ρq s22

Since the amplitude may become very large if the resonance is weakly damped,
it is often necessary to pre-stress piezoelectric transducers so that they do not crack
when driven at resonance.
25.3.3 Power Radiation

The intensity of the wave transmitted from the bar end (assuming it is large com-
pared to wavelength) can be calculated from the value of the velocity u of the
vibrating end
K ME e x
u= (25.13)
Z MR
The intensity I at y = ly is a function of the rms value of the applied voltage e x̴
K ME
2
ex2
I= (25.14)
lxlz Re ( Z MR )

To radiate airborne sound at high intensity the diaphragm must make large move-
ments and/or be large. Often, however, the situation in radiating airborne sound is
that the bar end is small compared to wavelength and a diaphragm of some sort is
needed. The piezoceramic bar discussed previously must also have some mechanism
for converting the small movement into a large diaphragm movement. Several pos-
sibilities are available, such as a bimorph construction, where two bars are attached
but operate with different polarity to create a bending motion of the bar system. If
there is no need to radiate the ultrasound in any specific direction, a radially resonant
bending disk can be attached, as shown in Figure 25.2 [2]. Attaching a horn or other
matching device to the diaphragm is also common.
If the disk is reasonably damped (which is likely going to be the case because of
the radiation and internal losses), the point mechanical input impedance of the disk
will be nearly resistive, similar to that of an infinite disk, about
Z M = 8 m′′B′ (25.15)
Piezoelectric disc Solid horn

Solid
¼ λ -blocker
Solid cylinder Bending disk

for support and
impedance matching
FIGURE 25.2 High-efficiency ultrasound source or ultrasonic vibration device.

Here, B′ is the bending stiffness per unit length and m″ the mass per unit area
of the disk. Impedance matching between the piezoceramic disk and the horn can,
if needed, be achieved by adjusting the length of the impedance matching cylinder
shown in Figure 25.2. This cylinder can also be used for fastening the source.
The Q factor of the system can be calculated by inserting Equation 25.15 into
Equation 25.6. It is then also possible to optimize the efficiency of the system relative
to the electrical side.
25.3.4 Electromechanical Impedance Analogy

At frequencies much below the first resonance frequency, the electrical properties of
the crystal will be those of a capacitor with a capacitance CE0.
ε 0 ε′xlylz (25.16)
CE 0 =
lx
It is shown in Ref. [13] that for frequencies close to the first resonance the electric
impedance of the bar can be described by
Z MR ρcll
Z EM = − j q q2 x z cos ( kqly ) (25.17)
K ME
2
K ME
in parallel with the CE0 capacitance. A series expansion of the cosine for frequencies
close to the resonance frequency shows that the bar can be represented by an
electrical network similar to that shown in Figure 25.3.
The resistance ℜER [Ω], capacitance CEM [F], and inductance LEM [H] are given by
ρclxlz ℜ MR
ℜ ER = = 2 (25.18)
K ME
2
K ME
8K ME
2
s22ly
CEM = = K ME
2
C M (25.19)
π 2l x l z
L C R
CE0
FIGURE 25.3 Electrical circuit that will have the impedance properties of the bar for
frequencies close to one of its resonance frequencies.
d12lz LM CM
1: ——— MR
S22 u
e F
CE0
FIGURE 25.4 Mechanical impedance analogy of the bar close to one of its resonance
frequencies.
ρqlxlylz L
LEM = = 2M (25.20)
2 K ME
2
K ME
This means that the circuit can be redrawn to include a transformer as shown in
Figure 25.4.
In many applications, it is necessary to also include the losses in the piezoelectric
material. Another resistor must then be added to the circuit on the mechanical side.
Sometimes it is also necessary to take the electrical losses into account, in which
case these will be represented by a resistor in parallel with the capacitance. This
gives us the electric impedance analogy circuit shown in Figure 25.5.
A measure of the electromechanical activity of the piezoelectric material is given
by the ratio (which is constant for a particular type of cut and crystal type)
CEM 8k 2
= 2 E 2 (25.21)

CEO π 1 − kE ( )
This ratio should be high so that the voltage required for a certain acoustical output
of the bar will be small.
When operating the bar at the first resonance frequency the effective mass of the
bar will be half that of the “normally weighed” mass. The compliance will also dif-
fer between static and dynamic conditions.
LEM CEM REM
RE0 CE0 ER
FIGURE 25.5 Electric impedance analogy of the bar close to one of its resonance
frequencies.
25.3.5 Q Factors
The Q factor of the bar, oscillating longitudinally and radiating from one end only, is
ω1LEM π ρq cq
Q≈ = (25.22)
ℜ ER 4 ρc
Since the radiation impedance into air is much smaller than the characteristic
impedance of longitudinal waves in the bar, the Q factor can become very high even
if the bar is thick.
In some applications, such as oscillators, quartz crystals are common and it is
important that the crystal circuit has a high Q factor. Quartz crystals used for oscilla-
tors and other timing circuits such as watches will usually be mounted in a vacuum
container to avoid damping due to sound radiation. The internal losses of the crystal
then become important.
Any oscillator will exhibit noise around the frequency of oscillation. This noise
corresponds to timing imperfections. The smaller the bandwidth of this noise, the
more precise the frequency will be. Since quartz oscillators for timing are used in
low-power circuits there is normally no need to worry about the crystal breaking due
to mechanical stress. The relation between the dynamic displacement at resonance
and the static displacement applying the same voltage (peak) at zero frequency is
ηdyn 2 Z 0q
= (25.23)
ηstat π Z0

Here, Z 0q is the characteristic impedance of the crystal. The resonances of piezo-
electric crystals and ceramics that radiate into oil or water, for example, have much
smaller Q factors than those radiating into air. The stresses will in these cases be
relatively small, even at resonance.
For high-power applications involving light loads it will be usually necessary to
pre-stress the piezoceramic disk so that it does not break under conditions of high
Q resonance.
25.3.6 Piezoceramic Disk Vibrators

To radiate airborne sound at high intensity, the source diaphragm must make
large movements. This can be easily achieved by using bending disks rather than
longitudinally vibrating bars. However, even with disks the volume acceleration
that can be achieved is low. Some form of impedance matching to the air must be
arranged. In the case of disks, this is achieved easily by using a Helmholtz resonator.
A piezoceramic material is attached to a metal disk so that the piezoceramic layer is
out of the neutral plane, as indicated in Figure 25.6. When the piezoceramic expands
sideways, the disk will bend.
In many cases, the ultrasound to be generated is a buzzing noise or beep. Such
a sound generator can have its oscillating frequency controlled by the electronic
system. This can be arranged either by sensing the electric impedance of the disk or
by using part of the disk as a sensor, as shown in Figure 25.7.
Helmholtz
resonator
Resonant compliance
brass disk cavity
Piezoelectric Helmholtz
ceramic resonator
mass hole
Cavity
FIGURE 25.6 Resonant band-pass box enclosure for a piezoelectric beeper.
Resonant brass disk
Main bending
piezoceramic Sensing
layer piezoceramic
layer
FIGURE 25.7 Resonant brass disk shown in Figure 25.6 may have part of its piezoceramic
layer used for a sensor strip for feedback into an electronic oscillator circuit.
25.4 SERIES AND PARALLEL RESONANCE

The circuit shown in Figure 25.3 will have both a series and parallel resonance. The
input impedance of the circuit is
1  1 
R + j ωL +
jωC0  jωC  (25.24)
Z EL =
1 1
+ R + jω L +
jω C 0 jωC

The magnitude of this impedance for a simulated crystal is shown in Figure 25.8.
The dip in the impedance curve is the result of the series resonance and the peak
a result of the parallel resonance. We find the following expressions for the series and
parallel resonance frequencies:
1
ω transmit,series = (25.25)
LEM CEM
106
|ZEL| [Ω] 104
102
1
1k 2k 5k 10 k 20 k 50 k 100 k
Frequency (Hz)
FIGURE 25.8 Magnitude of the electric impedance for a piezoelectric resonator as

shown in Figure 25.5 for L EM = 1 × 10 −1 H, CEM = 1 × 10 −9 F, ℜ EM + ℜER = 1 Ω, ℜ E0 = 1 × 106 Ω,
and CE0 = 5 × 10 −10 F.
1
ω transmit , parallel = (25.26)
LEM CEP
Here CEP is
CE 0CEM
CEP = (25.27)
CE 0 + CEM
The series transmit and the series parallel frequencies will differ slightly. For a
transmitter application, where more acoustic power is desired, it is advantageous to
drive the piezoelectric bar at its series resonance frequency. At the series resonance,
the current will be large through the “moving parts” of the bar, that is, through the
branch containing LEM and CEM. This is equivalent to a high velocity at the free end
of the transducer crystal bar, as will be explained earlier.
When the piezoelectric crystal or ceramic transducer is receiving an acoustic
signal, the circuit is driven by a voltage generator inserted in the series component
branch, as shown in Figure 25.9. Assume that the amplifier sensing the transducer’s
output has a high input impedance. The resistance RE0 of the input stage in the
amplifier can often be neglected so that only the CE0 capacitance will remain. This
will simply slightly reduce the available voltage. The resonance frequency in the
receive mode will then be
1
ω receive,series = (25.28)
LEM CEP
Equation 25.28 shows that the series receive resonance frequency will be the same
as the parallel transmit frequency. Whether the slight frequency difference will be
a problem or not depends on the application. In a monostatic sonar application, in
which we use the same transducer for both the transmit and receive operation, we
ER REM CEM LEM
d12lz
pl x l z —–— CE0 RE0 e
S22
FIGURE 25.9 Electrical impedance analogy of the piezoelectric bar close to a resonance
frequency, including the voltage generator due to the received sound pressure.
usually want the transmit and receive frequencies to be quite similar so that we can
use the same transducer for both operations.
If the Q factor of the transducer is low, then there may be sufficient overlap
between the two frequency ranges available for the transducer to be used for both
tasks, given the frequency band requirement. For a high Q factor application, it will
probably be necessary to use two separate transducers.
All sonar applications will need a certain signal bandwidth to allow detection and
separation of objects in time or space. Sometimes the difference between the transmit
and receive frequencies can be useful; for example, it will help in using the Doppler
shift to separate the backward and forward movements of a reflector.
25.5 BANDWIDTH AND RANGING

Airborne sonar pulse-echo systems are used in robotics, object recognition systems,
and other remote measurements. The sonar is used to measure the location of an
object by measuring the time lapse from transmit to receive as well, with multiple
transducers, for triangulation to determine directions.
Sonar systems often provide angular separation by an array that sends out and
receives beams of sound in desired directions. Radial separation uses timing differ-
ences between the transmitted and received reflected signals.
The requirement for signal bandwidth is related to radial separation in the
following way. Assume a pulsed, transmitted signal to have a frequency f and a band-
width Δf, and also assume the medium to carry the sonar signals at a velocity c. The
relationship between time and frequency resolution is typically
∆f ⋅ ∆t ≈ 1 (25.29)

Here, Δt is the duration of the pulse envelope of the signal. The spatial extension of
the pulse envelope will be cΔt, which will define the smallest object separation pos-
sible in the radial direction.
The detection will be more robust if the signal is coded in time or swept in
frequency. This will require more bandwidth without giving better spatial resolution,
and the spatial detection capability will then depend on the duration or sweep rate
of the coded signal.
The reason to strive for a large value Q factor is that this raises the signal level and
improves the signal-to-noise ratio of the transmit/receive system. The signal-to-noise
ratio can however also be increased by averaging the received echoes if time is
available. Reducing the Q factor by 10 will require 10 times the averaging for the
same effective bandwidth. In practice, the Q factor of the transducers will depend
on their mass and radiation conditions and the signal frequency will need to be
chosen for the signal/noise ratio desired and frequency channels available. The
“reverberation time” of the transducer’s resonances is an important design parameter
since a high Q factor resonance leads to long decay time which makes the target echo
less distinct.
A low value of Q means that the same transducer can probably be used for
both transmit and receive operations. This will lower the cost and complexity of
the system. On the other hand, this approach may require a more advanced signal
detection scheme, such as the use of swept or pulsed signals and time averaging,
which in turn will result in a stronger possibility of the transmitted signal, and the
transmitter’s location, being detected.
The situation is slightly more complex when the target is moving relative to the
transmitter and/or receiver. In these cases, there will be a Doppler effect that leads to a
frequency shift. The shift is largest when the transmitter is also working as the receiver
and the distance between the two varies as they move along the line joining them.
c − vO
f′ = f (25.30)
c − vS
Here, f is the frequency emitted by the sound source, f ′ the frequency sensed by the
observer, vO and vS are the speeds of the object and the source respectively relative
one another in the reference frame, and c is the speed of sound in the medium.
If the sonar receiver is sharply tuned so that the resonance has a high Q factor, or
if a narrowband filter is used to enhance sensitivity and reduce noise, the reflected
sound may well have a frequency offset that makes it fall out of the sensitive range
of the resonant system when the source or object is moving. Consequently, a sonar
ranging system for airborne sound is best built using a wideband transmitter and
receiver system.
25.6 PIEZOCERAMIC LOUDSPEAKERS

Piezoceramic transducers by themselves, as disks or bars, are brittle, and large
stresses will lead to mechanical failure. Overvoltage may cause arcing and general
electric insulation difficulties in humid and dirty environments.
Most practical piezoceramic devices for the audio and low-frequency ultrasonic
ranges are based on the use of a piezoceramic layer being added to a thin metal disk
as described earlier. The metal disk is usually used as one of the electrodes. The
layer is treated so that it expands sideways and thus tries to bend the metal disk. To
increase the force, several piezoelectric layers can be used either on one side or on
both sides of the metal disk. In a loudspeaker application, the sides of the disk must
be acoustically separated to avoid an aerodynamic short circuit around the disk. This
can be done using a loudspeaker box approach, for example, isolating one side of a
disk by a closed chamber, as shown in Figures 25.6 and 25.10. Since the metal disk is
Resonant conical
diaphragm
Circular
Resonant cavity
piezoceramic
bender Elastic support ring
(a) (b)
FIGURE 25.10 (a) Common piezoelectric loudspeaker design for high-frequency audio; (b)
common transducer design for narrowband ultrasound transmission and reception.
stiff, and has a high resonance frequency, the air chamber can be made small without
seriously affecting the frequency response. The frequency response of the disk can
be further tailored using a resonator or horn for the front-side radiation to extend or
filter the frequency response.
Horns or multiple matching layers can be used for broadband matching, as
shown in Figure 25.10a. One must be careful when using resonators and horns for
piezoceramic benders. Benders do not generate a plane wave so there will be higher-
order propagation modes in the horn. These will distort the directivity pattern unless
the horn is split up into sectors to prevent such modes.
Figure 25.10b shows a common transducer design for narrowband ultrasound
transmission and reception. In this type of transducer is a resonant, bending piezo-
ceramic disk that rests on a flexible ring located at a disk node. The top of the disk
holds a “loudspeaker” cone that is set in vibration by the disk much like an electro-
dynamic loudspeaker’s diaphragm is driven by its voice coil. The design offers little
disk damping so the disk’s primary bending resonances are still very noticeable.
Resonant acoustic circuits, such as a quarter wave or Helmholtz resonators, and
matching layers can be used for narrowband blocking and matching. However, the
decay rate of the resonances must be controlled. For ultrasonic transducers the rela-
tive bandwidth is often 10% or less so additional analog electrical elements such as
inductance, capacitance, and resistance can be used to tailor the desired band-pass
frequency response to suit the required filter characteristic.
25.7 PIEZOELECTRIC FILM LOUDSPEAKERS

25.7.1 Introduction
Many of the problems involved in the use of piezoceramic transducers for high-fre-
quency audio and low-frequency ultrasound applications can be addressed by using
a piezoelectric film instead of a piezoelectric disk or bender, since the flexible film
can be thin, have a large radiating surface, and low mass. It is difficult however to
move large amounts of air with piezoelectric films, since the force available is small,
so their use as loudspeakers is limited to the frequency range above 1 kHz for con-
ventional audio material. Such limits do not apply for their use for microphones and
closed cavity headphones. Two advantages of piezoelectric film microphones are
their potential for low cost and immunity of the film to moisture.
Piezoelectric film can also be used for force or accelerometer designs as an
alternative to traditional piezoelectric ceramic materials. The film is then used
in its thickness d33 mode, for example, as a transducer under the bridge in string
instruments.
25.7.2 Function
The piezoelectric film will expand and contract due to the attraction or repulsion of
internal dipoles to the applied field. Piezo film is typically operated in the d31 mode
that leads to expansion and contraction in the film plane.
Clearly a flat sheet of film cannot move much air. The film is instead arranged to
be statically slightly curved or formed to make circular transducers and fastened at
its edges to a rigid holder, as shown in Figures 25.11 and 25.12. When the AC volt-
age is applied, there will be considerable transverse contraction and expansion of the
film. Since the film can be driven over its entire surface, the movement will be in
phase over the surface except when the film exhibits higher order modes.
It is important to remember that the film will have many bending wave reso-
nances. Their excitation depends on the force distribution over the film but also on
the acoustic load on the film by the surrounding air. If the film has low mass per unit
area, its impedance may be low compared to the air load impedance and the reso-
nances will be well damped. The resonances may also be damped by using a sound-
absorbing material on the back side of the film, close to but not touching the film, as
shown in Figure 25.11b, to arrange the film for a toroidal-type directivity. High static,
but low dynamic impedance, at the edges that hold the film is useful.
Porous sound-absorber
Plastic foam on
support grille
Piezoelectric film
AC voltage supply Piezoelectric film

(a) (b)
FIGURE 25.11 (a) Schematic rendering of how transverse film motion in the d31 mode can
be converted to a pulsating, approximately radial, vibration. (b) A toroidally shaped high-
frequency loudspeaker driver using a piezoelectric film.
Plastic foam on Piezoelectric film

porous support Porous sound-absorber
FIGURE 25.12 Strip of piezoelectric film bent to form an array using construction similar
to that in Figure 25.11.
The maximum displacement of the film arc shown in Figure 25.11a will be at its
center. The displacement η center is [5]
3
ηcenter ≈ rcurve S (25.31)
2
Here, rcurve is the radius of the film sector with no voltage applied and S the
circumferential strain in the film. The strain S = Δl/l is given by
e (25.32)
S = d31
h
Here, e is the electric potential difference over the film, h its thickness, and d31 the
piezoelectric constant for transverse motion of the film.
The film can also be formed into arrays useful for low-frequency ultrasound
applications such as the one shown in Figure 25.12. The directivity pattern is
determined by the number of elements and their diameter.
25.7.3 Electrical Properties

A common model for the piezoelectric film as a receiver—that applies for
audio frequencies—is that of a strain-dependent voltage source in series with a
capacitance. Leakage over the film, connectors, and in the amplifier results in a
simple RC high-pass filter characteristic.
For transmitter use, the film requires high voltages and it is necessary to consider
electrical breakdown of the film, its connectors, and associated arcing. The break-
down voltage of the film material is about 80 kV/mm but the film is generally thin
when used for loudspeakers (about 20–100 μm thick). At high frequencies, the losses
in the film may result in considerable heating. With sputtered films the resistivity of
the sputtered layer can become important, thus contributing to localized heating, arc-
ing, and film breakdown. Screen-printed silver ink will withstand high voltage and
high currents better than sputtered surfaces. Screen-printed silver ink electrodes are
also less likely to develop cracks in the conductive layer.
In some applications such, as electric guitar mechanical vibration pickups, it is
necessary to introduce electrical shielding to avoid hum and other noise. This can
be done using a fold-over design, as shown in Figure 25.13. One of the electrodes is
grounded and acts as the shield. The interior electrode is slightly smaller than the out-
side shielding electrode. The drawback of any shielding approach is a higher moving
mass, so the technique is primarily used for vibration sensing.
25.7.4 Physical Configurations

For high-frequency audio loudspeaker applications, the film usually is used in the
d31 mode as discussed previously, clamped at two sides so that it forms an arc or
cylinder, as shown in Figure 25.11. This configuration is also suitable for air
ultrasound applications up to frequencies of about 50 kHz. The displacement caused
Electrical connection strips
Piezoelectric film
Ground (shield) side metallization
Signal side metallization folded in
FIGURE 25.13 Fold-over design in which one of the electrodes is grounded and acts as
the shield.
by in-plane strains is converted to approximately radial motion along the radius of

curvature and perpendicular to the film surface. As in conventional loudspeakers, it
is necessary to isolate the back radiation from the film unless a dipole loudspeaker
is needed. This can be done by some kind of enclosure or by letting the film back
radiate into sound-absorbing material as shown earlier.
For a rough estimation, the low-frequency resonance cutoff of a curved film
transducer can be shown to be proportional to the square root of the ratio of Young’s
modulus E of the piezoelectric film to its density ρ and inversely proportional to the
radius of curvature rcurve [5]
1 E
f0 = (25.33)
2πrcurve ρ

Note that Equation 25.33 does not consider the mass loading effect of the
electrodes. The equation should be modified when the film is especially thin and
relatively heavy electrode materials, such as silver ink, are used.
Clearly if the film radius is reduced, the relative displacement at the film
center becomes larger. At the same time the resonance frequency f0 is increased.
As the resonance frequency is reduced, the sensitivity will be lower, as shown in
Figure 25.14 [7].
25.8 POWER REQUIREMENTS OF PIEZOELECTRIC

LOUDSPEAKERS AND TRANSMITTERS
In contrast to most electrodynamic and electromagnetic devices, piezoelectric
transducers are often characterized by high electroacoustic efficiency. Three types
of loudspeaker/transmitter uses are common. For loudspeaker applications these are:
(1) wide frequency range audio below the first resonance frequency of the piezo disk
and (2) narrow frequency range audio over the first resonance of the piezo disk.
For alarms and sensing these are: (3) narrow band audio or ultrasonic range. In
applications 1 and 2, the disk electric impedance is primarily a capacitance, and in
application 3 it is primarily resistive.
The capacitance of piezoceramic disks is usually between 5 × 10 −8 and 1 × 10 −6 F,
which can be a difficult electrical load for many electronic power circuits. To make
110
100
rcurve
5 cm
90 10 cm
SPL (dB)
20 cm
80 30 cm
70
60
0.1 k 0.2 k 0.5 k 1k 2k 5 k 10 k 20 k 50 k 100 k
Frequency (Hz)
FIGURE 25.14 Measured frequency response curves for a piezoelectric film having length
and width of 5 cm for various radii of curvature (at 1 m distance). (After Measurement
Specialties, Inc. (MSI), Piezo film sensors technical manual, http://www.msiusa.com/
sampled, November 2011.)
the load more manageable for the drive electronics it is common to insert a resistor
in series with the disk. This leads to a low-pass filter characteristic for the circuit.
This is seldom a problem since the disk fundamental resonance frequency is in the
range 0.5–1 kHz, and the audio is typically low-pass filtered, for example, in cellular
phones, to the range below 6 kHz.
The frequency range where maximum power is needed is usually around 0.5 kHz
for speech and the required level drops by about 6 dB per octave above this fre-
quency. The duty cycle for speech consonants is also very low, about 10%–20%.
Since the listening distance in the case of cellular phones is only a few centimeters,
the power requirements are low. For computer applications the listening distance
in the case of loudspeakers is often about 0.5 m, but then more electrical power is
likely to be available. In the latter application, a lower disk resonance frequency is
also desirable.
For music sound reproduction, the capacitive impedance leads to large drive cur-
rents at high frequencies. This causes even greater problems in the design of the
drive electronics. The capacitive reactance of the piezoelectric transducers can be
compensated for in the feed circuit using Zobel networks, for example, Π, L, or T
links (see Chapter 23). Insertion of a simple series resistance is adequate in many
instances. Additionally, series inductance may be added on the electrical side to
increase the bandwidth around the piezo series resonance [14]. Transformers are
usually used to help provide the voltages required and to achieve galvanic separation.
25.9 PARAMETRIC LOUDSPEAKERS FOR AUDIO

The nonlinear effects of sound propagation were not included in the analysis in
Chapter 1. For a review of these effects, the reader is directed to Refs. [15–20].
When two collimated sound beams travel along the same path, at high levels, the
nonlinearity of air will generate new signal components that are the sum and the
difference of the frequencies of the signals carried in the sound beams. If the signals
have the same carrier frequency but one signal is amplitude-modulated or single-
sideband-modulated, the nonlinear properties of air can be made useful. If the sound
intensities of the two beams are sufficiently high, audible sound is generated in the
common volume of the beams. This is sometimes called parametric sound genera-
tion. The devices used are sometimes called parametric loudspeakers.
The sound pressure pdem of the low-frequency signal demodulated by the nonlin-
earity can be shown to be proportional to the second time-derivative of the envelope
of the double sideband signal [17]
pdem ∝ pˆ u2lz Sω 2c e mod (25.34)

Here, p̂u is the amplitude of the modulated ultrasonic signal, emod the modulating
envelope signal, ωc the carrier’s angular frequency, S the cross-sectional area of the
transducer beam, and lz the axial distance from the source along the beam to the
observation point.
There will be mainly two types of distortion products: intra- and inter-sideband
distortion. The intra-sideband distortion is caused by the ultrasound frequency com-
ponents within a specific sideband. Inter-sideband distortion is the result of the cross-
modulation between the frequency components in the two different sidebands of the
double sideband signal. The two sidebands interact and more distortion is added
to the demodulated signal. This inter-sideband distortion can be reduced by audio
signal preprocessing before modulation [21], which is often done in some practical
application of parametric loudspeakers. If a single sideband approach (using either
sideband) is used, the inter-sideband distortion can be removed. This approach also
allows for more efficient use of the bandwidth afforded by the ultrasound transmitters.
Since the nonlinearity is small the audio components from the cross-modulation
are weak, typically about 70–80 dB below the level of the ultrasound beams. For
useful audio sound levels, the ultrasound pressure levels of the beams need to be
about 140 dB or more. While such sound pressure levels are not harmful, they can
be difficult to generate. The low level also requires the audio to be preprocessed to
reduce the peak to rms level.
For the ultrasound beams to have high intensity, it is necessary to use the volume
within the near-field zone of the transducers. The limiting distance z of the near-field
zone is approximately
d 2 (25.35)
z≈
λ
The larger the transducer diameter and the shorter the wavelength, the
more efficient interaction between the two beams. Note that the strength of the
demodulated signal increases with distance from the transducer until the drop in
intensity in the Fraunhofer zone starts to take effect. With typical transducer arrays
having a diameter of 0.4 m operating at 40 kHz, the working distance is about 4–5 m.
10
5
0
Relative SPL (dB)
–5
–10
–15
–20
–25
–30
–35
30
25
30
20 25
y 15 15
20
10 x
5 10
5
0 0
FIGURE 25.15 Measured relative demodulated audio SPL as a function of x and y at z = 1 m
for SSB modulation over a 0.3 by 0.3 m quadratic grid with 2 cm resolution. (From Barbagallo,
M., et al., Modulation and demodulation of steerable ultrasound beams for audio transmission
and rendering, Proc. of the 11th Int. Conference on Digital Audio Effects (DAFx-08), Espoo,
Finland, 2008.)
Figure 25.15 shows the measured level of the demodulated audio for a plane array
having a diameter of about 0.2 m operating at 40 kHz at 1 m distance.
In practice, it is necessary to keep the ultrasound frequencies of the beams as low
as practically possible, typically in the 30–50 kHz range, to reduce damping by air. If
desired, the beams may be focused using the geometrical design of the transducer (zone
plate or curved transmitter) or with signal processing using an array of small transmitters.
25.10 IONOPHONES
Occasionally there is a need for a transducer that has very low internal impedance
and can operate at very high frequencies. Ionophone loudspeakers fill these needs
since the sound source is a corona discharge, an ionized globule of air. Such ionized
air is usually called a plasma and is characterized by electrons being able to move
around freely since they are not attached to any nuclei. There are two types of iono-
phone working principles, cold and hot plasma.
“Cold plasma” is the type that exists when the air is ionized by a strong DC electric
field. One way of obtaining a cold plasma is by using a sharp pin—functioning as an
electrode—that is kept near a second electrode that is preferably sound transparent,
for example, a metal wire mesh or grill. At a sufficiently high voltage, the air will
become ionized and electrically conductive. A large-value resistance is inserted in
the series electrical circuit to limit the current. The air is permanently ionized as
long as the electric current is sufficiently large.
Grid holding Plasma Anode

cathode needles globules grid
and resistors at needle tips
Modulated
current sink
+
Signal High voltage
source bias supply
R Needles
FIGURE 25.16 “Cold plasma” ionophone for operation with a DC current with an overlaid
modulation connected to the grid holding the cathode needles.
To generate sound, the globule of ionized air must expand and contract, which
is achieved by injecting an AC component into the electric circuit, as shown in
Figure 25.16. Since the sound generated by a single plasma globule is very weak, a
functioning cold plasma ionophone usually consists of an array—flat or focused—of
thousands of needles. The audio signal can be supplied to all needles in parallel.
Array shading and time delay techniques can be used to direct the sound. The “cold
plasma” ionophone also suits itself to be used as a “digital loudspeaker” since the
needles can be supplied with different currents in a binary fashion [22–24].
A “hot plasma” is generated by subjecting the air to an intense AC electric
field. Typically the AC frequency is chosen to be at radio frequencies, for example,
30 MHz or higher, but not conflict with broadcast and other radio transmissions.
The high AC frequency also serves to keep the air constantly ionized, which would
not be the case for carrier frequencies below approximately 3 MHz. Two major
types of hot plasma loudspeakers exist: those that operate with a horn, shown in
Figure 25.17, and those that use a point-type electrode similar to the cold plasma
type. In both cases, however, the plasma-generating AC field is modulated with the
Quartz tube Electrode ring

at ground potential
Center electrode holder
Platinum electrode
Plasma globule
Horn
FIGURE 25.17 “Hot plasma” ionophone for operation with a high-frequency amplitude
modulated electric carrier signal connected to the platinum electrode and the electrode ring.
The small modulated plasma globule appears in the air at the tip of the center electrode inside
the quartz tube. The modulation frequency can be as high as permitted by the carrier and
modulation scheme.
audio signal so that the size of the plasma globule varies with the audio amplitude
and frequency [9,25].
The horn attached to the ionophone acts more as a concentrator than as an
impedance-matching device. A major problem in any ionophone is the generation
of ozone, O3. Since ozone is poisonous in high concentrations, it is necessary
to convert the ozone molecules back to regular oxygen molecules. This can be
achieved by containing the ozone-generating plasma inside a fine mesh metal
sphere. As the ozone molecules hit the mesh, they are converted to regular oxygen.
REVIEW QUESTIONS
25.1 Why is the bandwidth required in ultrasonics applications often much more
limited than that in audio applications?
25.2 What will be the effect on the Sell transducer’s response of a randomized back
electrode to one that is made with a homogeneous ridge pattern?
25.3 What is the purpose of the blocker shown in Figure 25.2?
25.4 How can the device be fastened using the matching cylinder shown in
Figure 25.2?
25.5 Most piezo film transducers have film working in the d31 mode. Why must
such transducers have the film curved for transverse film movement?
25.6 What are the main limitations of ultrasound parametric sound generation?
25.7 What are the main limitations of sound generation by ionophones?
REFERENCES
1. Cracknell, A. P., Ultrasonics, Wykeham Publications Ltd., London, U.K. (1980)
ISBN-13: 978-0844813301.
2. Sasaki, Y. et al., High efficiency ultrasonic sound source in air, J. Acoust. Soc. Jpn. (E),
1(3), 209–210 (1980).
3. Gooberman, G. L., Ultrasonics, The English Universities Press Ltd., London, U.K.
(1968) ISBN-13: 978-0340051467.
4. Kuttruff, H., Physik und Technik des Ultraschalls, Hirzel Verlag, Stuttgart, Germany
(1988) ISBN-13: 978-3777604275.
5. Tamura, M. et al., Electroacoustic transducers with piezoelectric high polymer films,
J. Audio Eng. Soc., 23(1), 21–26 (1975).
6. Naono, H. et al., Design of an electro-acoustic transducer using piezoelectric polymer
film, Proc. Audio Eng. Soc., 77th Conv., New York, Preprint #1271 (1988).
7. Measurement Specialties, Inc. (MSI), Piezo film sensors technical manual, http://www.
msiusa.com (sampled November 2011).
8. Murata Super Tweeter Driver Brochure, http://www.murata.com (sampled November
2011).
9. Klein, S., Ionophone, Comptes Rendus de Academie des Sciences, 222, 1282 (May
1946).
10. Wykes, C., Capacitive acoustic transducers for air-borne ultrasonics, Proceedings of
Sonics: A Sound Basis for Testing, University of Southampton Institute of Transducer
Technology, Southampton, U.K. (1994).
11. Ladabaum, I. et al., Surface micromachined capacitive ultrasonic transducers, IEEE

Trans. Ultrason. Ferroelectr. Freq. Control, 45(3), 678–690 (1998).
12. DuPont Teijin Mylar Polyester Films, DuPont Teijin Films, 1 Discovery Drive (P.O. Box
411) Hopewell, VA 23860. http://www.dupontteijinfilms.com (sampled November 2012).
13. Kinsler, L. E. and Frey, A. R., Fundamentals of Acoustics, 2nd edn., J. Wiley & Sons,
New York (1962) Library of Congress Card Number 62-16151.
14. Waanders, J. W., Piezoelectric Ceramics—Properties and Applications, Philips
Components, N.V. Philips Gloeilampenfabrieken, Eindhoven, the Netherlands (1991).
15. Meyer, E. and Neumann, E. G., Physical and Applied Acoustics, Academic Press,
London, U.K. (1972) ISBN-13: 978-0124931503.
16. Beyer, R. T., Nonlinear Acoustics, Naval Sea Command, Department of the
Navy (1974). ASIN: B007WAN53K, available at http://www.dtic.mil/cgi-bin/
GetTRDoc?AD=ADA098556 (Accessed on November 2012).
17. Berktay, H., Possible exploitation of nonlinear acoustic in underwater transmitting
arrays, J. Sound Vib., 2(4), 435–461 (1965).
18. Bennet, M. B. and Blackstock, D. T., Parametric array in air, J. Acoust. Soc. Am., 57(3),
562–568 (1975).
19. Yoneyama, M. and Fujimoto, J., The audio spotlight: An application of nonlinear
interaction of sound waves to a new type of loudspeaker design, J. Acoust. Soc. Am,
73(5), 1532–1536 (1983).
20. Mellert, V., The origin of the audio signal in a beam of modulated ultrasound in air,
Proceedings of CFA/DAGA’04, Strasbourg, France (2004).
21. Kite, T. D. et al., Parametric array in air: Distortion reduction by preprocessing, ICA/
ASA Proceedings, Seattle, WA (1998).
22. Deraedt, A., Electroacoustic transducer using corona effect, Proceedings of the 90th
Audio Engineering Society Convention, Paris, France, Paper 3037 (1991).
23. Tombs, D. M., Corona wind loudspeaker, Nature, 176(4489), 923 (1955).
24. Fransson, F. J. and Janson, E. V., STL–Ionophone: Transducer properties and
construction, J. Acoust. Soc. Am., 58(4), 910–915 (1975).
25. Chew, J. R., The Klein-Plessey ionophone loudspeaker, BBC Engineering Division
Report M.018, #1954/13 (1954).
Appendix A: Introduction
to Electric Components and
Classic Circuit Theory
A.1 VOLTAGE, CURRENT, POWER, AND ENERGY
Electric networks can carry energy from a supply to a load by a combination of
electric potential difference and electric charge. The movement of electric charge is
called electric current, which is confined to electric conductors, usually called wiring
or cables. The unit for current i is the ampere (A). The circuit needs to be complete
for charge to flow and for power to be transported. The potential difference across a
circuit element or elements is called the voltage or tension. The unit for voltage e is
the volt (V). The power P is expressed in watts (W) and the energy W is the sum of
power over time and is expressed in watt-seconds (Ws) or joule (J).
A.2 POWER, DIRECT, AND ALTERNATING CURRENT

Current and voltage can be static or time-varying. Circuits carrying only non-
time-varying voltage and current are called direct current (DC) circuits. Those
carrying oscillating voltages and currents are called alternating current (AC) circuits.
Circuits can carry a combination of both types, superimposed on one another. These
can be treated separately if the system is linear; audio and acoustic systems are usually
linear. In acoustics, as applied to audio engineering and sonics, i.e., electroacoustics,
we are primarily concerned with dynamic systems having frequencies in the range
from a few hertz (Hz) to some hundred thousand Hz. Linear AC circuits are usually
analyzed one frequency at a time, i.e., using a Fourier approach where the AC current
is frequency analyzed and assumed to be the superposition of various sinusoidal
components having different frequencies.
In DC circuits, the power dissipation is calculated as the product of voltage and
current. In AC circuits, the power dissipation also depends on the phase angle between
the voltage across the circuit element and the current through the element. The phase
angle is usually frequency dependent. For AC circuits, the power dissipated in an
element is the product of voltage across the element, the current through the element,
and the cosine of the phase angle between the voltage and the current.
Upper case letters are often used to write voltage and current in DC circuits, and
lower case letters are typically used for AC circuits. Because of the many symbols
needed in electroacoustics, it will not be possible to use this type of notation in
this book.
491
492 Appendix A: Electric Components and Classic Circuit Theory
A.3 “SYMBOLIC” OR “jω-METHOD”

We can analyze an AC electric circuit one frequency at a time, using the jω-method,
sometimes also called the symbolic method. Using this method, we assume a time
dependence sinusoidal, described by a factor
e jωt = cos(ωt ) + j sin(ωt ) (A.1)

where t is the time and ω is the angular frequency, expressed in radians per second
(r/s), related to the frequency f given in hertz by
ω = 2πf (A.2)

We use this complex representation of the time dependence and can then find the real
part of the time dependence by using the factor
cos(ωt ) (A.3)
and the imaginary part of time dependence by using the factor
sin(ωt ) (A.4)
The phase difference ϕ (in radians) between two time dependencies is described by
adding a phase angle as follows:
cos(ωt + φ) (A.5)

In using the symbolic method, we assume the time dependence is sinusoidal and
therefore not expressly written in the variables. By using the underline under the letters
used to denote the physical quantity, we understand that the voltages e and currents
i are complex, have real and imaginary parts, and a sinusoidal time dependence.
We use the convention of ê = |e|√2.
Important to note is that impedance Z, the ratio of voltage to current, has both real
and imaginary parts but is not time dependent.
A.4 CIRCUIT ELEMENTS AND NETWORKS

Electric networks consist of circuit elements that—ideally—are dependent only on
the electric properties of the components that they symbolize. In reality, the circuit
elements will depend on the physical implementation of the components, for exam-
ple, a capacitor that is sensitive to vibration will exhibit a circuit element that displays
that property (i.e., microphonics as a condenser microphone).
In this book, we idealize the circuit elements and assume them linear, i.e.,
they will remain the same whatever voltage or current that is applied to them.
A passive circuit element receives power from the network that is dissipated
Appendix A: Electric Components and Classic Circuit Theory 493
(or in the acoustic case radiated). An active circuit element delivers power to the
network. Active circuit elements may be dependent on some property in a specific
place in the network, for example, as the voltage between two nodes; such an
element is called a voltage amplifier.
An electric network will consist of both active and passive circuit components.
Active components are characterized by their ability to deliver energy to other parts
of the circuit where energy may be stored, dissipated, or radiated. Passive components
can exhibit all three characteristics, but from the viewpoint of circuit analysis, it is
often practical to analyze components as if they are characterized by just one of
the aforementioned properties. Using electric circuits that include amplifiers, it is
possible to simulate ideal components so that they behave in a more ideal fashion
than do the actual conventional components.
Electric components include resistors, inductors, capacitors, transmission lines,
voltage sources, current sources, and switches. These are idealized, assumed to be
characterized by a “pure” property such as resistance, inductance, and capacitance.
Non-ideal characteristics are modeled from these same components.
One usually looks at the properties of electric components as lumped elements,
in the same way that mechanical engineering looks at the mass of a body being a
pure mass without, for example, any elasticity. Lumped elements are components
such as resistors, capacitors, and inductors, which are usually assumed to be
linear. Electric networks that contain only voltage or current sources, resistors,
capacitors, inductors, and transmission lines can be analyzed algebraically
to determine their response so that one can calculate the voltage and current
distribution in the network.
A.5 ELECTRIC CIRCUIT FUNDAMENTALS

The behavior of electric voltages and currents, both DC and AC, follow
• Kirchhoff’s current law: The sum of all electric currents entering a junction
is equal to the sum of all currents leaving the junction
• Kirchhoff’s voltage law: The directed sum of the electric potential
differences around a closed circuit must be zero
• Ohm’s law: The voltage across a resistance is the product of its resistance
and the current flowing through it
A.5.1 Kirchhoff’s Current Law

Conceptually, electric current flows in a conductor like water in pipes. The symbol
I (or i) is generally used for electric current. The current is assumed to move from a
junction at a higher electric potential to a junction at a lower potential. The direction
of a current is generally indicated by an arrow and assumed positive. Sometimes
current between two junctions a and b, or nodes, is written Iab indicating the junctions
and the reference direction for positive current. For example, Iab = −5 A means that a
current of 5 A is flowing from b to a.
Junction i2
i1
i3
i5
i4
FIGURE A.1 A junction where five conductors meet.
Figure A.1 shows a junction where five conductors meet. The sum of all currents
entering a junction is equal to the sum of all currents leaving the junction. This is
called Kirchhoff’s current law.
A.5.2 Kirchhoff’s Voltage Law

The voltage can be thought of as a height difference in a gravitational field. The
symbol E (or e) is often used for voltage. Assume a voltage difference between two
junctions a and b. The voltage between the junctions is written Eab defining both
the junctions and the voltage differential positive reference. For example, Eab = −5 V
means that junction b is 5 V greater than the voltage at a. This means that one can
write Eab = −Eba. In Figure A.2, the junctions are marked by black dots.
One electric junction is connected to another electric junction by an electric cir-
cuit component, for example, a current or voltage source, forming an electric circuit.
A closed circuit will require at least two components. Figure A.2 shows examples
of closed circuits. The meaning of the graphical symbols of the components in the
circuit will be explained in the next section.
The sum of all voltages around a closed circuit must be zero. In the figure shown,
the sum of voltages around the circuits a-b-c-f-a and f-c-d-e-f is both zero. This is
called Kirchhoff’s voltage law.
a b
Junctions f c
e d
FIGURE A.2 The sum of all voltages around a closed circuit is zero. This circuit features
three closed circuits: (1) a-b-c-d-e-f-a, (2) f-c-d-e-f, (3) a-b-c-f-a.
a b c d e
FIGURE A.3 Five ways of drawing conductors that cross. Mode a should be used to mark
conductors that cross without making a connection. Avoid using modes b (which makes a
connection) and c (which could easily be misinterpreted to be a connection if the printers do
a bad job). To mark conductors that meet at junctions, use modes d and e.
A.5.3 Avoidance of Ambiguities
In simple circuits containing only circuit components in series, the dots are often
neglected. However, because of the many ways in which circuit diagrams may be
drawn, there is always a risk for misprints.
Consider two conductors that cross but do not connect, as shown in Figure A.3a.
There would normally be a dot if there was a junction, as shown in Figure 3b,
Sometimes people draw nonconnecting conductors as shown in Figure A.3c.
To avoid ambiguities, it is best always to let conductors meet at “T-junctions.”
Whether there is a dot or not does, not matter; one would not have drawn the
conductors this way if they were not intended to connect and make a junction.
Figure A.3d and e show two ways of drawing which help one avoid mistakes.
A.6 ELECTRIC COMPONENTS

Electric components have properties that enable them to change the amplitude and
phase relationships between voltages and currents in a circuit. Common components
are resistors, inductors, capacitors, transformers, and gyrators. Resistors, inductors,
and capacitors are two-port components, i.e., they have two leads or terminals.
Transformers and gyrators are four-port components, having four terminals.
A.6.1 Voltage Sources, Batteries, and Ground

The DC voltage source and ground symbols are shown in Figure A.4. The battery
symbol is usually taken to represent a DC voltage source. The DC voltage is
symbolized by the letter E. The positive terminal is the one at the long line side of the
I
E R
(a) (b) (c)
FIGURE A.4 Battery (a) and ground (b) symbols, and a simple circuit (c).
symbol. Note that the symbol is somewhat similar to the symbol for capacitance. The
AC voltage source symbol is shown in Figure A.6. The AC voltage is symbolized by
the letter e.
The ground of a circuit refers to a common reference point that should not disturb
the working of the circuit; it is not necessarily connected to earth. Typically, the
ground connects to the negative terminal of the battery. Ground is often taken to
be fixed at zero potential in the circuit. Note that in amplifiers and other electronic
equipments, there may be separate analog and digital “grounds.” Transformers are
often needed to separate grounds so that noise-sensitive ground loops are avoided.
The design of the ground layout in circuits is important to minimize loops and stray
paths that can affect circuit function.
A.6.2 Resistance and Resistors

A common component in electric circuits is the resistor. The resistance acts to
limit the current flowing between two junctions at a potential (voltage) difference
connected by the resistor, as described by Ohm’s law. The unit of resistance is the
ohm [Ω]. The letter symbol R is generally used for resistors. Figure A.5 shows
some graphical symbols for resistance. From the viewpoint of electroacoustics, it is
essential to know that wire wound resistors (unless bifilar wound) have considerable
inductance, particularly unwanted for low-value resistors.
The inverse of resistance is called conductance, which is given in units of siemens
(S) and usually denoted by the letter G. The same graphical symbols are used for
conductance as well as for resistance. In old literature, conductance is often given
in units of (mho) and denoted by the lower case letter r. A similar system is used in
acoustics.
A.6.3 Inductance and Inductors

Physically an inductor is an electrically conductive wire. The wire can be straight
or coiled up. Inductance is the primary property of inductors and is due to the
buildup of a magnetic field around the conductor as a result of the current flowing
through the inductor. A characteristic of an inductor is that it takes some time for
a current to start flowing through the inductor when a voltage is applied across
the inductor terminals. The current through the inductance is said to be lagging
the voltage over the inductance, i.e., the phase of the voltage will be ahead of the
phase of the current.
(a) (b) (c)
FIGURE A.5 Three different graphical symbols for resistance/resistors. Symbol a is

commonly used as is symbol b, which is easier and faster to draw, but symbol b is also commonly
used as the graphical symbol for impedance. Symbol c is an older symbol, analogous to a.
i
i
e L
e C
(a) (b)
FIGURE A.6 (a) Circuit with an inductance. (b) Circuit with a capacitance.
Inductance is measured in units of henry (H). The letter symbol used for inductance
is L. The graphical symbol for inductance is shown in Figure A.6a. Mathematically,
the inductance depends on voltage e over and current i through the inductance L in
the following way:
di
e=L e = jωL i (A.6)
dt
A.6.4 Capacitance and Capacitors

Capacitance is the primary property of capacitors. The term capacitor is a more
modern term than condenser, which can also be used. A characteristic of a capacitor
is that it takes some time for voltage to build up when charge is fed into, via
the current, the capacitor. The voltage across the capacitance is said to be lagging
the current through the capacitance, i.e., the phase of the current will be ahead of the
phase of the voltage.
The letter symbol for capacitance is C, and electric capacitance is measured in
units called the farad (F). The graphic symbol for capacitance is shown in Figure A.6b.
Mathematically, the relationship between voltage e over and current i through a
capacitor having a capacitance C is given by the following:
1 1

e=
C ∫
i dt e=
j ωC
i (A.7)
A.6.5 Impedance and Admittance

Most real components in electric circuits will be characterized by a combination
of capacitance, inductance, and resistance. Impedance is expressed in ohms (Ω),
the letter symbol is Z, and the typical graphical symbols are those shown in
Figure A.7.
Z R X Y G B
FIGURE A.7 Typical graphical symbols for impedance Z and admittance Y.

The impedance of an inductance L is described by Z = jωL, that of a capacitance C by

Z = 1/jωC. A real capacitor will not exhibit pure capacitance but will be better described
by a combination of real and imaginary parts. The real part is the resistance and the
imaginary part is called the reactance, symbolized by R and X respectively. A typical
model is a series circuit written Z = R + jX. Both R and X are expressed in units of Ohm.
Sometimes we are interested in the reciprocal of impedance, which is called
admittance and usually written Y. The unit of admittance is siemens (S). The
admittance will in general be composed of a real part and a frequency-dependent
imaginary part, symbolized by conductance G and susceptance B respectively. We
write Y = G + jB. For mechanical circuits, admittance is the ratio between velocity
and force, often called mobility and written M.
A.6.6 Ohm’s Law
The voltage across a resistor is the product of its resistance and the current flowing
through it; this is called Ohm’s law. For DC conditions (zero frequency), having a
voltage E over and a current I through the resistance R, Ohm’s law is symbolically
written as
E = RI (A.8)
Ohm’s law can be applied to AC conditions, having a current i through and voltage e
across an impedance Z, and is then (one frequency at a time)
e = Z i (A.9)

The phase ϕ between the current i through and voltage e across an impedance
Z = R + jX is given by
 X
φ = tan −1   (A.10)
 R

The current through a capacitance leads the voltage by π/2 (i.e., 90°) whereas the
current through an inductance lags the voltage by π/2 (i.e., −90°).
A.6.7 Power
The power P dissipated in a resistance R by an AC current i is
P = Ri2 (A.11)
A.6.8 Voltage Sources
Until now, we have discussed only voltages and currents without discussing how
they are generated. The constant voltage source is characterized by the property that,
irrespective of whatever component or circuit that is placed in parallel with it, the
Zint
e e
FIGURE A.8 The symbol for a voltage source with and without internal impedance.
voltage across the source will always remain the same. A common graphic symbol
for a voltage source is that shown in Figure A.8a.
Any practical source will be characterized by an internal impedance that will limit
its ability to keep the voltage constant over the generator’s accessible terminals, as
indicated in Figure A.8b. An ideal voltage source has zero internal impedance. Many
audio amplifiers, both preamplifiers and power amplifiers, are designed to operate at
voltage sources, having negligible output impedance.
A.6.9 Constant Current Sources

The constant current source is characterized by the property that, irrespective of
whatever component or circuit that is placed in series with it, the current flowing into
that node will always remain the same. The symbol for a current source is shown in
Figure A.9a. An ideal current source has infinite internal impedance.
A practical current source will be characterized by having internal impedance,
reactive or resistive, which will limit its ability to keep current constant at its
terminals. Such a source is symbolically shown in Figure A.9b.
A.6.10 Transformers
As noted before, electric components can be of two types, two-ports and four-ports.
A common fourport circuit element is the transformer. A transformer contains at least
two windings, made of coils of wires which can be of various topologies and that,
by themselves, are two-ports. The 2 two-ports are converted into a four-port by the
magnetic coupling between the windings. A graphic symbol for the transformer is
shown in Figure A.10.
i i Zint
FIGURE A.9 The symbol for a current source with and without internal impedance.
i1 N1:N2 i2
a c
eab ecd
b d
FIGURE A.10 The graphical symbol for a transformer.
Terminals a and b attach to the primary winding of the transformer and terminals
c and d to the secondary winding. The primary winding is typically taken to be that
on the left side of the graphic symbol.
The transformer will convert the voltage at terminals a and b to another voltage
at terminals c and d. The ratio between these two voltages is determined by the
transformer’s turns ratio, which is N1:N2. For an electric transformer, with a magnetic
flux conductive core, N1 and N2 are the number of turns of wire of each winding
around the core.
As indicated by the dots appearing next to the windings in Figure A.10, the
windings can be related to one another in two ways. This determines the relative
phase of the primary two-port to the secondary. If the voltage e ab is positive, the
voltage e cd will also be positive if both dots are at the top of the graphic symbol.
These dots are usually not shown unless the transformer windings are connected so
that they should appear “in reverse.”
The ideal transformer is defined by the following equations:
N1
e ab = e cd (A.12)
N2
N2
i1 = i 2 (A.13)
N1
Note the convention for currents going in and out of the transformer. The positive
direction for current is in principle into the transformer at the top connections a and
c when the transformer is regarded as a “black box” four-port. In reality, however, the
current at c is directed in the other direction for a current going into the transformer
at a. This results in the necessity of a minus sign in the current relationship.
The transformer will act as an impedance converter, as implied in Figure A.11.
A “load” impedance Zload attached to the secondary winding of the transformer, i.e.,
terminals c and d, will be converted to the impedance Zin “as seen into” the terminals
a and b:
2
N 
Z in =  1  Z load (A.14)
 N2 

N1:N2
Zin Zload
FIGURE A.11 The transformer acts as an impedance converter.
A.6.11 Gyrators
A gyrator is a specialized component that can convert the impedance of an inductance
into that of a capacitance and vice versa. Gyrators in electric circuits are not real
physical components but are composed of components connected in networks.
Gyrators can be both two-port and four-port components, but in electric circuits they
need to be simulated by an active circuit containing amplifiers. Gyrators are often
used in analog audio frequency filters since they enable the realization of inductance
with high values and low losses using a network comprised of amplifiers, resistance,
and capacitance. The circuit symbol for the gyrator is shown in Figure A.12.
In electroacoustics, we find that some components, such as various electroacoustic
transducers, sometimes behave as if they are gyrators, converting a mass-type
impedance into a compliance-type impedance and vice versa.
A.6.12 Parallel and Series Impedances

Series and parallel connections of impedances are common in networks. For series-
coupled impedances as shown in Figure A.13, the resulting impedance Zseries is given by
Z series = ∑ Z (A.15)
i =1
i
For parallel-coupled impedances as shown in Figure A.14, the total impedance

Zparallel is given by
N
1 1
Z parallel
= ∑Z
i =1 i
(A.16)

G:1
Zin Zload
FIGURE A.12 The gyrator acts as a special type of impedance converter.

Z1 Zi ZN
Zseries
FIGURE A.13 Series connected impedances.
Zparallel Z1 Zi ZN
FIGURE A.14 Impedances connected in parallel.
In most networks, there will be both series- and parallel-connected impedances.

In this case, it is generally best to start “from the bottom” and work one’s way up
using the earlier relationships as appropriate, as one combines the impedances to
form series and parallel subnetworks that are solved one by one.
A.6.13 Conversion between Y and Δ Impedance Networks

In the analysis of the function of circuits, it is sometimes advantageous for further
simplification of the circuit to be able to reconfigure the impedances of the circuit.
Conversion of Y-type, “wye,” networks into Δ-type, “delta,” networks is one such
conversion, called a transformation. Seen from the outside, a four-port circuit element
can have any internal arrangement as long as the input and transfer impedances are
the same. The networks shown in Figure A.15 can be transformed into one another
by using the following simple equations:
Z1 = Z a Z c ( Z a + Z b + Z c )
Z 2 = Z b Z c ( Z a + Z b + Z c ) (A.17)
Z3 = Z a Zb ( Z a + Zb + Zc )

Z a = ( Z1Z 2 + Z1Z 3 + Z 2 Z 3 ) /Z 2
Z b = ( Z1Z 2 + Z1Z 3 + Z 2 Z 3 ) /Z1 (A.18)
Z c = ( Z1Z 2 + Z1Z 3 + Z 2 Z 3 ) /Z 3

Z1 Z2 Zc
Z3 Za Zb
(a) (b)
FIGURE A.15 Two networks: (a) A Y-network, (b) a Δ-network.
A.7 SOME IMPORTANT NETWORK THEOREMS

A.7.1 Superposition Theorem
In linear circuits, one may consider two or more sources acting jointly by studying
each one separately and then superimpose the results.
A.7.2 Inoperative Sources
Inoperative voltage sources are considered to be short circuits, i.e., have zero internal
impedance, and inoperative current sources are considered to be open circuits,
i.e., have infinite internal impedance.
A.7.3 Reciprocity Theorem
If a voltage applied to one branch produces a certain current in a second branch, then
applying the same voltage in the second branch will result in the same current in the
first branch.
A.7.4 Substitution Theorem
If a branch in a network carries a current Iab and the voltage across the branch Uab, the
branch components may be exchanged provided the voltage and the current remain
the same.
A.7.5 Substitution of Power Sources

A constant current source may be substituted for a constant voltage source provided
the internal impedances remain the same (except for the case of zero internal
impedance).
A.7.6 Millman’s Theorem

Multiple coherent sources in a network may be combined into a single source.
A.7.7 Thévenin’s Theorem
A “black box” containing a source and an arbitrary linear network will, seen from
the outside, behave as a constant voltage source with a certain internal impedance.
A.7.8 Norton’s Theorem
A “black box” containing a source and an arbitrary linear network will, seen from
the outside, behave as a constant current source with a certain internal impedance.
A.7.9 Maximum Power Transfer Theorem

Maximum power output from a source is obtained when the internal impedance of
the source Zi = Ri + jXi is the conjugate of the load impedance Zl = Rl + jXl, i.e., Zi = Zl*
that is Ri = Rl and Xi = −Xl.
A.8 AMPLIFIERS
A.8.1 Ideal Amplifiers
One usually differs between various types of ideal amplifiers as follows:
• Voltage amplifiers: Infinite input impedance, zero output impedance,

constant gain Vout is a constant times Vin irrespective of output load.
• Current amplifiers: Zero input impedance, infinite output impedance,
constant gain Iout is a constant times Iin irrespective of output load.
The circuit representation of the ideal voltage amplifier is shown in Figure A.16 along
with its common circuit symbol. The output voltage source has a value determined
by the input voltage and the amplifier-unloaded gain A. The amplifier gain A is
typically larger than unity, although unity gain amplifiers are common being used
as isolation amplifiers, as well as in filter and gyrator circuits. The amplifier gain
will generally be complex, but in many situations, it is reasonable to assume it is real
valued, as we do in this text. Of course, amplifiers of this type may be cascaded to
obtain any gain desired. A voltage amplifier can have multiple inputs, for example,
each with different gain. An inverting voltage amplifier has negative gain.
iout
+ + +
ein Aein eout Zload

ein A eout
– – –
(a) (b)
FIGURE A.16 The circuit diagram for the ideal voltage amplifier (a) and its circuit symbol (b).
In (b) a common ground is understood.
This type of amplifier is also a power amplifier in the sense that the amplifier
input does not consume any electric power and the amplifier output is able to feed
any power to the load. Assume the load impedance can be written in the following
form:
Z L = RL + jX L (A.19)

The power output from the amplifier will then be
P = iout
2
RL (A.20)
The output current iout is calculated from the circuit as

eout
i out = (A.21)
ZL
The circuit representation of the ideal current amplifier is shown in Figure A.17
along with its common circuit symbol.
It may be necessary to connect several amplifiers in series to obtain the desired
signal amplification. For ideal amplifiers (since they do not electrically load one
another), the total amplification Atot is easy to calculate as the product of the
amplification of the amplifiers Ai as
Atot = ∏ A (A.22)
i
i
A.8.2 Practical Amplifiers

Amplifiers are common in most types of electronics. Some common amplifiers for
audio are as follows:
• Preamplifiers
• Isolation amplifiers
• Charge amplifiers
• Power amplifiers
iin iout iin iout
i = Aiin Zin i = Aiin Zout
(a) (b)
iin A iout
(c)
FIGURE A.17 The circuit diagram for the ideal current amplifier (a), more realistic model
(b), and a possible circuit symbol (c).
An amplifier design can be to increase signals to overcome noise, power loss in long
lengths of cable, and to drive capacitive or inductive loads. At the same time, it is
important to realize that any amplifier can generate noise of its own as well as signal
distortion that may be both linear and nonlinear.
A.8.2.1 Noise
Since the components that are used to build amplifiers exhibit resistance, there will
be “thermal” noise induced at many places in the amplifier circuit. It is customary
to refer all of this noise to the amplifier input by simply taking the measured output
noise voltage and divide it by the appropriate gain of the amplifier. It is important to
note that amplifier noise will depend on the type of source connected to the input.
Generally, any amplifier will exhibit minimum noise for a signal source having zero
internal impedance, i.e., with the amplifier input short-circuited.
For radio-frequency amplifiers, it is quite common to indicate the input noise
characteristic as a noise resistance, i.e., the resistance that gives the same noise as
the amplifier. This is quite convenient since it immediately indicates the lowest
suitable value for the signal source resistance. This is also useful for various forms of
preamplifiers, which need to work with signals in the microvolt to millivolt ranges.
The noise generated by a resistor is the thermal agitation noise; this noise has the
same power density at all frequencies (“white noise”). The RMS value of this noise
voltage for a resistor of RE Ohm is
e = 4kTRE B (A.23)

where
k is the Boltzmann’s constant
T is the temperature in K
B the bandwidth studied in Hz
Besides this noise, there may also be other noise sources in resistors and other
components.
While not noise in the statistical sense, power mains frequency hum and its har-
monics must be counted as noise since they are undesirable additions to the signal.
(The voltage available on the power mains is in general very noisy as well.) Another
source of noise is demodulation of radio-frequency signals by nonlinearities in the
amplifier, bad solder joints, or connectors. Such radio-frequency signals can enter
the amplifier due to lack of insufficient shielding, improper ground routing, and
through the inputs and outputs.
A.8.2.2 Linear Distortion

The audible frequency range is nominally 20 Hz–20 kHz. The frequency range
of interest for audio is sometimes taken to be larger, for example, 2 Hz–200 kHz,
particularly for amplifiers.
Linear distortion is the term used to describe changes to the signal that do
not contribute new frequency components (at least within the frequency range of
interest). Such changes include the filter action of the amplifier. The amplifier will
(in its linear range) act as a signal filter that amplifies or attenuates various frequency
ranges. The amplitude filter action is also observable as a change in the phase of the
signal. The phase change will extend beyond the frequency range in which amplitude
change is noticeable. As a rule of thumb, one can say that (for a “benign” amplifier)
phase change can extend one decade below the low-frequency cutoff of the amplifier
and similarly one decade above the high-frequency cutoff.
It is customary to give the cutoff frequencies where the amplitude response has
dropped by 3 dB referred to the center frequency of the amplifier (typically 1 kHz).
Most audio amplifiers will act as high-pass filters, cutting off at low frequencies in the
range of 2–20 Hz, since low frequencies are difficult to render through loudspeakers
or headphones, and the signal sources have been subject to high-pass filtering in
the recording studio. At the high-frequency end, the amplifier will act as a low-pass
filter, cutting off in the frequency range between 20 and 200 kHz.
The response at the cutoff frequencies is often associated with considerable
signal delay. The group delay of a signal τ(ω) is associated with the phase response
(in radians) of the amplifier or filter given by
dϕ ( ω )
τ (ω ) = − (A.24)
dω
The time delay depends on the steepness of the filter roll-off at the low- and high-
frequency limits of the amplifier. Time delays of 20 ms at the low-frequency end and of
a few ms at the high-frequency end are likely to be audible. Remember that an acoustic
delay of 3 ms approximately corresponds to a distance of 1 m. The audibility of a time
delay difference will also depend on the signal and its rise time.
Unless a filter has constant group delay for all frequencies in the signal, the
waveform of the signal will be distorted. The distortion will be most visible on signals
that are transient and have a wide spectrum. Most conventional classical filters such
as Butterworth and Chebyshev filters are associated with large phase change near the
passband edges. Bessel filters are phase linear, i.e., have constant group delay, but do
not show as rapid roll-off at the passband edges as do the previous ones. Noise-type
signals are usually analyzed for power content and so the waveform distortion may
be of little interest.
A.8.2.3 Harmonic Distortion

Any signal processing device such as an amplifier or filter that has some form
of nonlinear transfer characteristics (i.e., the output signal is not constantly
proportional to the input signal) will generate harmonics to the signal if the
signal is in the nonlinear range. This type of distortion is sometimes also called
amplitude distortion, although the term harmonic distortion (HD) (because of its
effect) is the common term used. A simple form of nonlinearity is clipping when
the output signal is made to become higher than the maximum output voltage
capability of the device. Other forms of nonlinearity are zero-point nonlinearity
(crossover distortion), slew-rate limiting, and hysteresis. A special form of acoustic
nonlinearity is Doppler distortion, which sometimes is noticeable in electrodynamic
loudspeakers.
When discussing HD, the input signal is assumed to be sinusoidal and free from
noise and harmonics. The harmonics generated by the amplitude nonlinearity can
be both regular harmonics, i.e., the frequencies generated are integer multiples of
the signal frequencies, and subharmonics, i.e., the frequencies generated are integer
fractions of the signal frequencies.
Harmonics that are outside the audible frequency range might seem nonprob-
lematic, but since subsequent signal processing stage may be very nonlinear in
those frequency ranges, for example, due to slew-rate nonlinearity, they can cause
subharmonics and various forms of cross-modulation products as discussed in the
next section.
It is also important to note that the distortion characteristics of amplifiers
also depend on the nature of the load impedance. A resistive load impedance
is less likely to cause distortion than a predominantly reactive one. Impedance-
correcting networks, discussed later, are often used to overcome unsuitable load
impedances.
The HD is given in percentage by dividing the power of the harmonics by the
total power. For small amounts of HD (typically less than 10%), the percentage may
be calculated from
HD = 100
∑ i≥2
si2
≈ 100
∑ i≥2
si2
(A.25)
s2

∑ i ≥1
si2 1
Here s1 is the RMS value of the signal at the fundamental frequency, the input
signal frequency, and si is the RMS value of the harmonic component at a frequency
of if1.
A.8.2.4 Intermodulation Distortion

Intermodulation distortion is the name given to those distortion products that are a
result of the effects of nonlinearities on signals carrying two or more frequencies.
Any circuit that generates HD will also generate intermodulation distortion. The
audibility of intermodulation distortion is often higher than that of HD.
The intermodulation distortion (IMD) is found by dividing the power of the har-
monics by the total power and presented in percent. For small amounts of intermodu-
lation distortion (typically less than 10%), the percentage may be calculated from
IMD ≈ 100
∑ si2
i ≠1 or 2
(A.26)
s + s
2
1
2
2

Here s~1 and s~2 are the respective RMS values of the two input signals, and s~i is
the RMS value of the various intermodulation distortion components. Typically, the
two tonal components are chosen so that they are relevant to the test situation. One
common set used for audio amplifier testing are two tones at 60 Hz and 6 kHz,
another set are two tones at 10 and 11 kHz. In the first case, the intermodulation
components of interest will be the tonal components at 6 kHz ± N·60 Hz, and in the
second case they will be N kHz.
A.8.2.5 Multitone Distortion Measurement

To measure distortion similar to real-life situations, one can assemble a test signal of
pure tones over a desired range of frequencies, while ensuring that all tones have a
full period within the time window used. One can then do a spectrum analysis of the
distorted signal and (if one has used a rectangular time window) remove the original
tonal components and simply sum up the distortion power that will now be spread
almost like noise over a wide frequency range.
A.8.3 Input and Output Impedance

Practical amplifiers typically have an input impedance that is primarily resistive,
although one must keep in mind that the input impedance is, in practice, combined
with the capacitive load of the cable connecting the signal source to the amplifier. The
cable capacitance is about 100 pF per meter for coaxial cables used to interconnect
low-level devices. The input impedance of preamplifiers and power amplifiers is in
the range of 47 kOhm to 1 MOhm.
The output impedance will often vary over the frequency range. The reason for this
is the varying degree of negative feedback used in the amplifier. Most preamplifiers
have an output impedance of less than 1 kOhm. Many power amplifiers have an
impedance-correcting network at their output to stabilize the amplifier performance
so that the amplifier will be stable, i.e., not oscillate, for typical loudspeaker loads.
This filter will typically have the configuration shown in Figure A.18. Since a power
amplifier will drive a load consisting of the loudspeaker and cable combination, it
can be expected that the amplifier output will see load impedances of a few Ohms.
The amplifier output impedance is likely to be much lower than the resistance of the
cables, which is determined by their length and cross-section area.
Figure A.19 shows a variety of measured data for commercial loudspeaker cable
capacitance and inductance. It is best to use low capacitance cables for line level
connections since this will minimize the reactive load on the amplifier output.
L
C
To amplifier To loudspeaker
R
FIGURE A.18 A typical output filter network to stabilize a power amplifier, sometimes
called a Zobel network. The network also protects the amplifier against radio-frequency
signals entering the amplifier.
2
Inductance [µH/m]
1
0.5
0.2
0.1
0.05
10 20 50 100 200 500 1000
Capacitance [pF/m]
FIGURE A.19 Capacitance and inductance per meter of some commercial loudspeaker
cables. The dashed line in the figure represents the theoretical relationship between capaci-
tance and inductance for a cable at the propagation velocity of electromagnetic waves in
vacuum. (After Lesurf, J., Investigation of speaker cables, HiFi News, November 2008.)
The cable capacitance, in practice, does not have a series resistance (the cable
resistance being low) and is thus more difficult for the amplifier to drive. The extra
inductance introduced by the cable will be in series with the loudspeaker that
will (in the simplest case of a single electrodynamic loudspeaker) exhibit quite
high inductance anyway. The solid line in the figure represents the theoretical
relationship between capacitance and inductance for a cable at a propagation
velocity of electromagnetic waves in vacuum. The slower propagation speed is due
to the dielectric of the cable. The distributed capacitance and inductance may be
represented approximately in a circuit diagram by, for example, T-links as shown in
Figure A.20.
For low-level coaxial signal cables, the capacitance per meter and the relative
signal propagation speed are usually specified by the manufacturer. Values in the
range of 50–100 pF/m for the capacitance and a signal propagation speed of about
70% of the speed of light are typical.
L/2 L L L/2
To loudspeaker
To amplifier
C C C
FIGURE A.20 Typical discretized circuit representation of capacitance and inductance

of a cable.
BIBLIOGRAPHY
Ballou, G. M. (ed.), Handbook for Sound Engineers, Howard Sams, Carmel, IA (1991).
Hood, J. L., Audio Electronics, Newnes, Oxford, U.K. (1999).
Hood, J. L., The Art of Linear Electronics, Butterworth-Heinemann, Oxford, U.K. (1994).
Lesurf, J., Investigation of speaker cables, HiFi News (November 2008).
Skilling, S., Electric Engineering Fundamentals, Wiley, New York (1965).
Appendix B: Filters and
Filter Functions
B.1 INTRODUCTION
Acoustical, mechanical, and electrical systems are frequency dependent. The
properties of some variables as a function of frequency—at some point in a system—
are related to those at some other point in the system by a transfer function. The
transfer function describes the relative change in amplitude and phase between the
action at the two points.
Electrical filters are used to modify the transfer functions, but filters may also be
designed to have overall gain. Voltage and/or power gain is accomplished by using
amplifiers to increase the voltage, current, and driving power capability of the filter.
Electrical filters may use amplifiers for their operation, but such amplifiers are not
discussed in detail in this book.
The properties of acoustical, mechanical, or electrical systems may be converted
into one another by means of the theory presented in Chapters 5 through 8. There
is well-established theory for the filter effects of electrical circuits. This is a major
reason to convert acoustical and mechanical systems into equivalent electrical
circuits so that they may be studied by this theory.
B.2 TRANSFER FUNCTION AND FREQUENCY RESPONSE

A signal s(t) can be described by its complex spectrum S(ω) meaning that the
magnitude and phase components of the spectrum at each frequency ω = 2πf can be
described by
S (ω) = S (ω) ∠ S (ω) (B.1)

The complex spectrum can be also described at each frequency by its real and
imaginary parts
S (ω) = Re [ S (ω)] + Im [ S (ω)] (B.2)

Consider the signal processing done by the filter shown in Figure B.1. The action of
the filter on the signal Sin(ω) is described by the filter’s transfer function H(ω). The
transfer function will depend on the filter characteristic and will generally affect
both magnitude and phase of the signal.
513
514 Appendix B: Filters and Filter Functions
Filter action
Sin(ω) H(ω) Sout(ω)
FIGURE B.1 To find the amplitude and phase of the output signal each frequency is treated
separately to calculate the influence of the filter function on the amplitude and phase of the
input signal.
The filter action is mathematically described by
Sout (ω) = HИ(ω)Sin (ω) (B.3)

The complex multiplication is done for each frequency, one at a time. In audio,
the filter action is typically described by the filter’s response in (dB) relative to a
reference frequency as
 H (ω ) 2 
Frequency response = 10 log10  2
(B.4)
 H (ω Ref ) 

The frequency response is a colloquial term used to describe a filter’s transfer

function. Expressed in decibel (dB) values, it shows the response of the filter action
referred to some reference frequency, typically 1 kHz. For wideband bandpass filters
the geometrical center frequency is used, while for narrowband filters, the arithmetic
mean frequency is often used.
The cutoff frequency of a lowpass (LP), highpass (HP), or bandpass filter (BP)
is defined as the frequency at which the frequency response has dropped by 3 dB
relative to the maximum response or relative to the response at some nominal
frequency. Bandpass filters can be made to have different cutoff rates in their HP
and LP zones.
Modern active filters will typically be preceded and followed by amplifiers to
provide (at least) a high-impedance input (to reduce load on previous systems) and
a low-impedance output (to provide possibility for the electrical load of long cables
and other system components).
B.3 IDEAL FILTERS

Filters may be loosely classified as shown in Figure B.2. Real physical filters cannot
have the sharp response characteristics at the various transition frequencies since
such characteristics require the filter to be noncausal, i.e., to have response before the
audio signal has entered the filter.
Appendix B: Filters and Filter Functions 515
Lowpass Bandpass
1 1
H(ω)
H(ω)
Band edge
Band
Pass band Pass band edges
0 ω 0 ω
ωcut−off ωl ωu
Highpass Band stop
1 1
H(ω)
H(ω)
Band edge
Band edges
Pass band Stop band
0 ω 0 ω
ωcut−off ωl ωu
Shelving 1 Shelving 2
1 1
H(ω)
H(ω)
Transition Transition
band band
0 ω 0 ω
ωl ωu ωl ωu
FIGURE B.2 Some common filter functions.
B.4 PHYSICAL FILTERS

The order of a physical filter denotes the falloff rate beyond the cutoff frequency. For
a filter of order N the rate is N·6 dB per octave or N·20 dB/decade. The frequency
response of a practical LP filter is likely to be reminiscent of that shown in Figure B.3.
10
0
–10
–20
–30
–40
10 20 50 100 200 500 1000 2000 5000
Frequency [Hz]
FIGURE B.3 Characteristics of a practical LP Chebyshev filter designed for a cutoff

frequency of 1000 Hz, a 3 dB ripple, and a roll-off rate of −18 dB per octave.
The associated filter transfer functions H(ω) will generally be described by ratios
such as
Aω 2 + Bω + C
H (ω ) = (B.5)
Dω 2 + Eω + F
where coefficients A, B, C, D, E, and F are real or complex numbers chosen to

achieve the desired filter action. Clearly, filters that have a higher order function in the
numerator than in the denominator will be LP filters. Filters that have the same order
in numerator and denominator may be shelving filters, or bandpass/band stop filters.
B.5 FREQUENCY TRANSFORMATION
AND LOWPASS TO HIGHPASS
In practical filter design, it is common to “start” with an LP filter designed for a
nominal cutoff frequency of ω0 = 1. This filter is then converted to the desired cutoff
frequency of fc by replacing ω0 in the filter transfer function by ω0/2π fc. This LP design
is then converted to an HP filter by replacing all occurrences of ω0/2π fc by 2π fc/ω0.
B.5.1 Bandpass Filter Example

Inserting the coefficients values A = 0, B = 1·102, C = 0, D = 1, E = 1·102j, and F = −1·106 into
Equation B.5, one obtains the filter’s frequency response as shown in Figure B.4. This
filter can be considered a simple bandpass filter although the passband is more shaped
as a peak and does not have the “flat” section one normally expects of a bandpass filter.
By using transfer functions with ratios between higher-order polynomials, one can
achieve better approximations to the nominal transfer functions shown in Figure B.2.
The frequency response of a typical octave band wide bandpass filter centered on
1 kHz is shown in Figure B.5.
10
0
–10
–20
–30
–40
500 630 800 1000 1250 1600 2000
Frequency [Hz]
FIGURE B.4 The frequency response properties of a simple bandpass filter centered
on 1 kHz.
10

0
–10
–20
–30
–40
500 630 800 1000 1250 1600 2000
Frequency [Hz]
FIGURE B.5 Frequency response of an analog commercial octave band wide bandpass filter
centered on 1 kHz.
The filter shown has a frequency response characteristic of many bandpass filters.
There is passband ripple, dips in the response outside the passband and the roll-off
rate in the transition bands is limited.
B.6 TIME SIGNAL AND TRANSFER FUNCTION

The time and the frequency responses of a system are related by Fourier transforms.
The impulse response is the idealized time response of a system or filter to a Dirac
function δ(t) (also called a “delta” function). The Dirac function is defined by
+∞ t=0
δ (t ) =  (B.6)
0 t≠0
and must also fulfill the requirement

∞
∫ δ(t )dt = 1 (B.7)

−∞
The impulse response h(t) of a system or filter is related to its transfer function H(ω)
by the Fourier transform integrals
∞
H (ω) = FT [ h(t )] =
∫ h(t ) e − jωt
dt (B.8)
−∞
∞
1
h(t ) = FT −1 [ H (ω )] =
2π ∫ H (ω ) e jωt
dω (B.9)
−∞
where FT denotes the Fourier and FT−1 the inverse Fourier transform.
Impulse response h(t)
–1
0 2 4 6 8 10 12 14 16 18 20
Time [ms]
FIGURE B.6 Impulse response of filter function shown in Figure B.5.
Looking carefully at the filter response in Figure B.5, we note the filter has ripple
in its frequency response and sharp band edge cutoffs and so the filter’s impulse
response will be characterized by ringing as shown in Figure B.6.
B.6.1 Group Delay
All real filters will delay the signal. The delay is a function of the phase change as
a function of frequency as described earlier. Theoretically, one can design all-pass
filters that provide smooth and almost constant delay over a large frequency range.
The signal delay is given by the group delay τg, which depends on the phase response
of the filter as
d
τg = − ∠ H (ω) (B.10)
dω
The delay will usually be the largest at the filter band edges since the resonances
shaping the filter at those frequencies will tend to be the least damped.
B.7 SOME COMMON FILTER CONTOURS

B.7.1 Three Classical Filters
Some “standard” classical filter characteristics are often used in filter design. Since
filters are extensively used in electroacoustics, particularly in loudspeaker design,
the frequency response and transient response of these filters are of great interest.
Figure B.7 shows three third-order classical LP filter characteristics: Chebyshev,
Butterworth, and Bessel, all with a nominal cutoff frequency ω0 = 1. Chebyshev
(many spellings exist) typically has equal “ripple” (frequency response irregularity
in dB) in the passband and ringing in the impulse response. The ripple is needed
as one of the specifications of a Chebyshev filter. In the graphs shown here all
10
0
–10
–20
Butterworth (B3)
–30 Bessel (BE3)
Chebyshev (C3)
–40
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.7 The normalized frequency response of the three third-order LP filters designed
for the same cutoff frequency ω0.
0
–2
–4
Butterworth (B3)
–6 Bessel (BE3)
Chebyshev (C3)
–8
0.1 0.2 0.5 1 2
ω/ω0
FIGURE B.8 Detail of frequency responses close to cutoff frequency of the third-order
filters shown in Figure B.7.
Chebyshev C3 filters have 3 dB ripple. Butterworth filters do not have ripple, have a
“maximally flat” frequency response in the passband, and a small amount of ringing
in the impulse response. Bessel filters finally have no ringing and a fairly “non-flat”
frequency response in the passband. Figure B.8 shows the frequency response
characteristics of these filters close to the cutoff frequency.
These LP filters have the following transfer functions, Butterworth HB3LP(ω),
Chebyshev HC3LP(ω), and Bessel HBE3LP(ω) shown as follows:
1
H B3 LP (ω) = (B.11)
1 + 2 jω + 2( jω )2 + ( jω )3
1
HC 3 LP (ω ) = (B.12)
1 + 3.70 jω + 2.38( jω)2 + 3.98( jω)3
15
H BE 3 LP (ω) = (B.13)
15 + 26.33jω + 18.49( jω)2 + 5.411( jω )3
The reader is directed to Ref. [4] for the theory behind these transfer functions and to
obtain the coefficients needed for other filter orders and types. Butterworth filters of
other orders will be discussed further later in this appendix since they are commonly
used for crossover filters to provide signals within a suitable frequency range to
loudspeaker drivers.
The phase, ringing, and ripple properties are increasingly pronounced the larger
the roll-off rate of the filter. The phase curves for the three filters in Figure B.7 are
shown in Figure B.9.
We note that the phase dependency on frequency is quite similar for the three
filter types. While phase is of great importance in sound reproduction at frequen-
cies below about 0.5 kHz, the group delay is much more important over most of
the audio range. Figure B.10 shows the group delay for the three filters discussed.
Note that the sharper the cutoff, the more pronounced the group delay will be.
A smooth and flat group delay curve results in poor frequency response in the
form of slow cutoff. For these reasons, Chebyshev and Bessel functions are seldom
used in crossover filter design. Butterworth and Chebyshev HP filters are often
used as target designs for the frequency response of low-frequency loudspeakers
as described in Chapter 17.
Conversion of LP filters to HP filters is done by the replacement of ω in the LP
transfer function by 1/ω to obtain the HP filter transfer function. The corresponding
frequency response curves are shown in Figure B.11. Figures B.12 and B.13 show the
associated phase and delay curves.
π
2
Phase [radians]
–π Butterworth (B3)
2
Bessel (BE3)
Chebyshev (C3)
–π
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.9 The normalized phase response of the three third-order LP filters designed for
the same cutoff frequency ω0 shown in Figure B.7.
8
Butterworth (B3)
Bessel (BE3)
Chebyshev (C3)
6
Delay/ω0 [s]
0
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.10 The normalized group delay response of the three third-order LP filters
designed for the same cutoff frequency ω0 shown in Figure B.7.
10
0
–10
–20
Butterworth (B3)
–30 Bessel (BE3)
Chebyshev (C3)
–40
0.1 0.2 0.5 1 2 5 10 20 50 100
ω/ω0
FIGURE B.11 The normalized frequency response of three third-order HP filters designed
for the same cutoff frequency ω0.
For most crossover filters, the sum of the signals of HP and LP filters is different
from the original signal sent to the pair of filters. This applies to both the impulse
and the frequency responses. The sum of the transfer functions is not unity. When
the filtered signals have been radiated by the loudspeakers, the distance between
the drivers and the presence of early reflections, reverberation, and diffraction will
further distort the summation.
The LP filters discussed earlier can easily be converted to HP filters as described
previously. This technique results in HP filters that are very useful and well
documented in the literature. However, this is not an optimal way to design crossover
filters. Even if one designs crossover filters using the combination of LP and HP
filters of one “family” and with the same cutoff frequency, the sum of the LP and
HP filter transfer functions often shows unsuitable amplitude and phase properties in
π
Phase [radians] 2
π
– Butterworth (B3)
2
Bessel (BE3)
Chebyshev (C3)
–π
0.1 0.2 0.5 1 2 5 10 20 50 100
ω/ω0
FIGURE B.12 The normalized phase response of three third-order HP filters designed for
8
Butterworth (B3)
Bessel (BE3)
Chebyshev (C3)
6
Delay/ω0 [s]
0
0.1 0.2 0.5 1 2 5 10 20 50 100
ω/ω0
FIGURE B.13 The normalized delay response of three third-order HP filters designed for
the frequency region around the cutoff frequency. Crossover filters are studied in the
next section and more optimal designs such as Linkwitz–Riley (LR) and asymmetric
constant-voltage designs are discussed and shown.
B.7.2 Peak and Dip Equalization

A special type of stop filter is the notch filter that is designed to only remove a very
small part of the spectral content in a signal. This type of filter is not designed using
the previously mentioned classical filter methods but rather using simple resonant
circuits, which can be passive or active as discussed later in the section on peak
L L L L L
ein C C C C C ZE = R lsp
N LC-links
FIGURE B.14 A delay line consisting of N LC-links.
and dip equalization (see also Chapter 22 for passive circuits). These are generally
described by a center frequency and a −3 dB bandwidth. Active implementations of
peak and dip filters are discussed later in this appendix.
B.7.3 Delay Equalization
When using multiple drivers in a loudspeaker time delay equalization may be
necessary. The delay characteristics can be modified by using high- and low-level
electrical filters as shown in Figure B.14. The N sections of LC links give a delay
τ = N LC (B.14)
Since the filter has a low pass character, its component values must be chosen so that
its cutoff frequency is much higher than its operating frequency range. Many links
will be needed to provide useful delay so relocation of the drivers should be used
first. The delay line impedance must be chosen so that the signal is not reflected at
the loudspeaker end; matching is needed. The delay line characteristic impedance
is given by
L
Z= (B.15)
C
The amount of delay that is acceptable is a subject of discussion, but a delay

variation of 1 ms can be considered large since it corresponds to a displacement
of 0.34 m of the sources. Experiments have shown that the audibility of group
delay distortion is about 1 ms. Group delay distortion is particularly noticeable on
transient sounds 11.
Some types of loudspeakers, such as horn and transmission line loudspeakers,
will have very frequency-dependent group delay characteristics that may be possible
to equalize only with filters using digital signal processing. The group delay of horns
will vary with the length of the horn and with the expansion function of the horn
cross-section area.
Active circuits for delay equalization are discussed later in this appendix.
B.8 CROSSOVER FILTERS

B.8.1 Nomenclature
In simple filter design, it is common to regard each crossover filter for a pair of
drivers individually and similarly for three- and four-way systems. The design would
include the frequency response characteristic, amplitude, and phase of the filters.
The target functions for the loudspeaker drivers will initially often be chosen so that
the LP and HP filters conform to one of the standard filter shapes.
The crossover network are roughly categorized by their filter characteristics such as
• Overall frequency response characteristic, Butterworth, Linkwitz-Riley,

constant voltage, etc.
• Highpass HP, lowpass LP, or bandpass BP characteristic
• Cutoff frequency ω0
• Cutoff rate beyond the cutoff point ±N·6 dB per octave
• M number of driver sections (M-way)
• Possible separation of HP and LP cutoff frequencies
Proper crossover filter design must be done with the properties of the individual
drivers in mind, as well as their mounting, the intended use of the loudspeaker, and
the acoustic properties of the listening environment.
In addition to the crossover filters many networks also include shelving circuits,
attenuators, peak and dip resonance circuits, and impedance compensation circuits.
These are discussed in Ref. [1] and below.
B.8.2 Target Functions
For crossovers, the target is to find a set of filters that will provide flat sound pressure
frequency response in the listening area. Phase and amplitude characteristics are
then chosen by simple mathematical rules. Phase errors that lead to group delay
differences may be compensated for electronically. The sum pressure response
HΣ,pressure(ω), assuming drivers have transfer functions DLP(ω) and DHP(ω) and the
filters have transfer functions HLP(ω) and HHP(ω), is
H Σ, pressure (ω ) = DLP (ω) H LP (ω ) + DHP (ω) H HP (ω ) (B.16)

If the listening space is in the reverberant field of a room, the reflection

characteristics of the room surfaces must be taken into account. Once the direct
sound level is 10 dB below the early reflections and reverberation in the room,
the precedence effect is no longer as important and the hearing impression is
no longer dominated by the direct sound from the loudspeakers [10,13]. In this
case, the power summation of the loudspeaker contributions becomes important,
particularly if the loudspeakers are mounted more than one half-wavelength away
from one another. In this case, the approximate summed frequency response needs
to be added on a power basis as
2 2
H Σ, power (ω ) = DRLP
2
(ω ) H LP (ω ) + DRHP
2
(ω) H HP (ω ) (B.17)

Here DRLP(ω) and DRHP(ω) are the rms responses of the drivers in the reverberant
field. In the following, all D(ω) functions will be assumed to be unity.
B.8.3 Butterworth
Butterworth filters, sometimes called maximally flat filters, are easy to design and
implement since component values are not as critical compared to, for example,
Chebyshev filters. Butterworth LP filters have transfer functions that fulfill
2 1
H (ω ) = (B.18)
1 + ω2 N
Here N is the filter order. The first three orders of Butterworth LP and HP filter
transfer functions HB1LP(ω), HB1HP(ω), HB2LP(ω), HB2HP(ω), HB3LP(ω), and HB3HP(ω)
are given by the following equations. For higher orders, the reader is advised to
consult, for example, Ref. [4]
1
H B1LP (ω) =
1 + jω
1
H B1HP (ω) = (B.19)
1
1+
jω
1
H B 2 LP (ω) =
1 + 2 jω + ( jω ) 2
1
H B 2 HP (ω) = 2
(B.20)
 1   1 
1+ 2   +  
 jω   jω 

1
H B3 LP (ω) =
1 + 2 jω + 2( jω)2 + ( jω)3
1
H B3 LP (ω) = 2 3
(B.21)
 1   1   1 
1+ 2  + 2  +  
 jω   jω   jω 

First-order Butterworth (B1) filters for crossovers are in practice not very useful for
crossover networks since their slopes of 6 dB per octave are usually not sufficient,
in most cases, except where the drive units complement the B1 filters. An attractive
feature of B1 filters for crossovers is that the sum pressure response is linear. The
sum response for the in-phase condition is HΣB1+(ω) and the out-of-phase condition
is HΣB1−(ω) is given by
1 + jω
H Σ1B + (ω) = (B.22)
1 + jω
1 − jω
H Σ1B − (ω) = (B.23)
1 + jω
2 2 1 + ω2
H Σ1BP (ω) = H LP (ω) + H HP (ω) = = 1 (B.24)
1 + ω2
The frequency-normalized transfer function magnitudes and their associated group

delays are shown in Figure B.15.
The transfer function for the second crossover using Butterworth B2 filters is
1 − ω4 ω + ω3
H 2 BOP (ω) = −j (B.25)
2ω + (1 − ω )
2 2 2
2ω + (1 − ω 2 )2
2
The frequency response curves for the out-of-phase B2 filter are shown in Figure B.16
and the delay in Figure B.17. It is seen that the summed response has a slight hump at
the crossover frequency. This hump can be minimized by using slightly offset values
for the crossover frequency of the LP and HP filters.
Corresponding curves for the third-order crossovers using Butterworth filters (B3)
are shown in Figures B.18 and B.19.
Higher-order Butterworth filters are unusual in passive (high-level) crossover
networks since the delay starts to become excessive and the added complexity
and cost.
10
0
–10
–20
B1LP
–30 B1HP
B1LP + B1HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.15 The frequency response curves for the B1 LP and HP filters and the summed
filters. The crossover frequency is ω0.
10
0
–10
–20
B2LP
–30
B2HP
B2LP – B2HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.16 The frequency response curves for the out-of-phase B2 filter for LP and HP
filters and for the out-of-phase summed filters. The crossover frequency is ω0.
1.5
Delay/ω0 [s]
0.5
0
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.17 The delay introduced by the B2 out-of-phase combined condition. The
crossover frequency is ω0.
B.8.4 Linkwitz–Riley
The Linkwitz–Riley (LR) group of filters are similar to B1 and B3 filters that
give unity summing of their LP and HP signal parts, but the LR filters are usually
used with an even order of sections as second- and fourth-order filters (LR2 and
LR4), see Refs. [6,7]. Since B2 crossover filters are usually unsuitable because
of their interference at the crossover frequency, the LR2 filters give the designer
a possibility to work with 12 dB per octave filters. The equations governing the
LR2 filters are
1
H LR 2 LP (ω) = (B.26)
(1 + jω)2
10
0
–10
–20
B3LP
–30
B3HP
B3LP–B3HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.18 The frequency response curves for B3 filters for LP and HP filters and the
response for the out-of-phase summed filters (dotted). The crossover frequency is ω0.
5
B3LP + B3HP
B3LP – B3HP
4
Delay/ω0 [s]
0
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.19 The delay introduced by the B3 in-phase (dashed) and out-of-phase (solid)
combined condition. The crossover frequency is ω0.
−ω 2
H LR 2 HP (ω) = (B.27)
(1 + jω)2
1 − ω2 1 − jω
H LR 2OP (ω) = = (B.28)
(1 + jω)2 1 + jω
The corresponding equations governing the LR4 filters are
1
H LR 4 LP (ω) = (B.29)
(1 + )
2
2 jω − ω 2

ω4
H LR 4 HP (ω) = (B.30)
(1 + )
2
2 jω − ω 2

1 + ω4
H LR 4 IP (ω) = (B.31)
(1 + )
2
2 jω − ω 2

The associated curves for the frequency response are shown in Figures B.20 and B.21.
The delay characteristics are shown in Figure B.22.
10
0
–10
–20
LR2LP
–30 LR2HP
LR2LP–LR2HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.20 The frequency response curves for LR2 filters for LP and HP filters and the
response for the out-of-phase summed filters. The crossover frequency is ω0.
10
0
LR4LP
–10 LR4HP
LR4LP + LR4HP
–20
–30
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.21 The frequency response curves for LR4 filters for LP and HP filters and the
response for the in-phase summed filters. The crossover frequency is ω0.
4 LR4LP + LR4HP
Delay/ω0 [s] LR2LP – LR2HP

3
0
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.22 The delay introduced by the LR2 out-of-phase and LR4 in-phase summed
filters. The crossover frequency is ω0.
We see that the LR4 filter introduces considerable delay compared to the LR2
filter. An important property of the LR2 and LR4 filters is that they crossover at
−6 dB in contrast to the B1 and B3 filters that crossover at −3 dB. This results in a
3 dB dip in the power response of the LR2 and LR4 filters at the crossover frequency.
B.8.5 Asymmetric and Symmetric Constant Voltage

A logical way of specifying crossover filter is to specify the desired LP filter and
form the associated HP filter by subtraction of the LP filter function from unity gain.
An example of an asymmetric constant-voltage filter, ACV, given in Ref. [14], has the
following LP HACV3LP(ω) and HP HACV3HP(ω) functions.
1
H ACV 3 LP (ω) = (B.32)
1 + 2 jω − 2 ω 2 − jω 3
2 jω − 2 ω 2 − jω 3
H ACV 3 HP (ω) = (B.33)
1 + 2 jω − 2 ω 2 − jω 3
The combined crossover of course has an all-pass unity response with some delay.
Figure B.23 shows the ACV3 filter function magnitudes. Such asymmetric constant
voltage filters always have one of the filter slopes with an attenuation of 6 dB per
octave. This type of filter is very difficult to use in practice since electrodynamic
loudspeakers have a falloff rate of 12 dB per octave at low frequencies. Using this
filter would require amplification below the crossover frequency.
In fact, the “mirrored” asymmetric constant-voltage design, MACV, shown in
Figure B.24 may be superior from a practical viewpoint since their LP- and HP-frequency
responses correspond much better to the needs when using electrodynamic loudspeakers
Ref. [12].
10
0
–10
–20
ACV3LP
–30 ACV3HP
ACV3LP + ACV3HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.23 The frequency response curves for ACV3 LP and HP filters and the response
for the in-phase summed filters. The nominal crossover frequency is ω0.
10
0
–10
–20
MACV3LP
–30
MACV3HP
MACV3LP + MACV3HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.24 The frequency response curves for MACV3 LP and HP filters and the
response for the in-phase summed filters. The nominal crossover frequency is ω0.
The transfer function equations for these mirrored LP HMACV3LP(ω) and HP

HMACV3HP(ω) filters are
1 + 2 jω − 2 ω 2
H MACV 3 LP (ω) = (B.34)
1 + 2 jω − 2 ω 2 − jω 3
− jω 3
H MACV 3 HP (ω) = (B.35)
1 + 2 jω − 2 ω 2 − jω 3
A symmetric constant-voltage third-order filter design SVC3 given in Ref. [14] has
LP HSCV3LP(ω) and HP HSCV3HP(ω) functions as
10
0
–10
–20
SCV3LP
SCV3HP
–30
SCV3LP+SCV3HP
–40
0.1 0.2 0.5 1 2 5 10
ω/ω0
FIGURE B.25 The frequency response curves for SCV3 LP and HP filters and the
flat frequency response of the in-phase summed filter responses. The nominal cross-
over frequency is ω 0.
1 + 3.7 jω
H SCV 3 LP (ω) = (B.36)
1 + 3.7 jω − 3.7ω 2 − jω 3
−3.7ω 2 − jω 3
H SCV 3 HP (ω) = (B.37)
1 + 3.7 jω − 3.7ω 2 − jω 3
The transfer function magnitudes of the SCV3 filter are shown in Figure B.25. The
summed response is unity since one of the filters is formed by subtraction from
unity.
B.9 IMPLEMENTATION OF CROSSOVER FILTERS

B.9.1 Passive Filters
Typical filter configurations for passive crossovers are the first-, second-, and third-
order filters shown in Figure B.26.
The following design rules give constant input resistance Butterworth filters that
are suitable for power amplifiers under the assumption that the loudspeaker electrical
impedances are of the same resistance R [2].
Simplest B1 filter (Figure B.26b):
R
L= (B.38)
ω0
1
C= (B.39)
Rω 0
C
ein R HF lsp
L R HF lsp
e in
C R LF lsp R LF lsp
L
(a) (b)
C2 C2 C3
e in L2 R HF lsp e in L3 R HF lsp
C1 R LF lsp C1 R LF lsp
L1 L1 L2
(c) (d)
FIGURE B.26 Examples of basic first-, second-, and third-order LP and HP filters used
in loudspeaker crossover networks. Drivers are assumed to be impedance compensated to
have resistance only. (a) Simplest second-order combined HP and LP filter; (b) simplest first-
order combined HP and LP filter; (c) second-order combined HP and LP filter; (d) third-order
combined HP and LP filter.
Simplest B2 filter (Figure B.26c):
2R
L1 = L2 = (B.40)
ω0
1
C1 = C2 = (B.41)
2 Rω 0
Simplest B3 filter (Figure B.26d):
3R
L1 = = 3L2 = 2 L3 (B.42)
2ω 0
4 2
C1 = = C3 = 2C2 (B.43)
3Rω 0 3
In the design of passive filters, it is often assumed that the filter building blocks are
ideal and that the drivers appear as pure electrical resistances. This is not the case;
inductance cannot be had without resistance, and capacitance cannot be had without
resistance and leakage. Typically, real capacitors are more ideal capacitances than
real inductors are ideal inductances.
B.9.2 Some Active Filters

Various aspects of active filters are discussed in Chapter 22. Typically, high-input
and low-output impedances are the goal in the active filter. Simple non-inverting
impedance converters are used for this purpose as shown in Figure B.27.
Active first-order filters are trivial and are also shown in Figure B.27, which shows
simple HP and LP filters that have 6 dB per octave slopes and are maximally flat.
The transfer function of the LP filter is
1
H LP (ω) = (B.44)
1 + jωR1C1
The transfer function of the HP filter is
jωR1C1
H HP (ω) = (B.45)
1 + jωR1C1
Amplifier with
high input and
low output
Cin impedance
Cout
+
–
ein eout A = +1
R
e in Unity gain voltage amplifier eout
with high input and
low output impedance
Unity gain impedance converter circuit with DC blockers at input and output.
The need for R, Cin, and Cout depends on the complete circuit requirements but should
initially be chosen so as not to affect response in the audio frequency range.
C1
R1
A = +1 A = +1
ein R1 Unity gain voltage amplifier eout ein C1 Unity gain voltage amplifier e
out
with high input and with high input and
low output impedance low output impedance
Highpass filter circuit Lowpass filter circuit
FIGURE B.27 Impedance converter and simple first-order maximally flat HP and LP filters.
The cutoff frequency ω0 (same for both LP and HP) is given by
1
ω0 = (B.46)
R1C1
The transfer function resulting from in-phase summation of HLP (ω) and HHP (ω) is
unity. If the files are summed out-of-phase, the transfer function magnitude is still
unity, but the summed filters then give a phase shift that can be used to generate
a delay in a circuit. For smooth delay over a wide frequency range, several such
sections with different center frequencies can be series-connected. This is a use-
ful alternative to the delay line shown in Figure B.14. An interesting observation
is that the frequency response resulting from the power summation of HLP (ω) and
HHP (ω) is also unity.
For higher-order filters, combinations in the form of series-connected filters are
convenient. The Sallen–Key filters shown in Figure B.28 are commonly used to
obtain second-order filters and can be cascaded Ref. [3]. The transfer function of the
generic Sallen–Key filter is
e out (ω) Z 3 (ω ) Z 4 (ω )
H (ω ) = = (B.47)
e in (ω) Z1 (ω)Z 2 (ω ) + Z 3 (ω) ( Z1 (ω) + Z 2 (ω)) + Z 3 (ω)Z 4 (ω)

Z3
C1
Z1 Z2 R1 R2
A = +1 A = +1
ein Z4 Unity gain voltage amplifier e ein C2 Unity gain voltage amplifier eout
with high input and out with high input and
low output impedance low output impedance
(a) (b)
Amplifier
R1 with high input and
R1 C2 R5 low output impedance
+
C1 C2 –
A = +1 R3
ein C1 R2
eout
ein R2 Unity gain voltage amplifier
with high input and eout R4
low output impedance
(c) (d)
FIGURE B.28 Various Sallen–Key filters: (a) generic filter circuit; (b) lowpass filter circuit;
(c) highpass filter circuit; (d) bandpass filter circuit.
The LP second-order Sallen–Key filter has a frequency response given by
ω 20
H (ω ) = (B.48)
2 jω 0 ω 2
ω0 + −ω
Q
and a Q factor
R1R2C1C2
Q= (B.49)
( R1 + R2 )C2
Here ω0 is the natural resonant frequency of the filter, which is usually close to the
−3 dB cutoff frequency of the filter
1
ω0 = (B.50)
R1R2C1C2
The HP second-order Sallen–Key filter has a frequency response given by
ω2
H (ω ) = (B.51)
jω ω
ω 2 − 0 − ω 20
Q
and a Q factor
R1R2C1C2
Q= (B.52)
(C1 + C2 ) R1
For the Butterworth characteristics, often preferred in audio circuits, Q = √2.

Several first- and second-order filters can be cascaded to obtain filters that have
higher slope rates. Other Q factors will be needed in these sections, and some of the
sections may become quite resonant, and the group delay large. HP and LP filters
can be used in series to obtain bandpass characteristics. For relatively narrowband
characteristics, the filter shown in Figure B.28d can be used.
With the cascaded networks, one can build various types of filter functions such
as Butterworth, Chebyshev, Bessel, etc. The transfer function is factorized into first-
and second-order filter factors. High Q filters may generate large voltages and cause
clipping at resonance frequencies.
B.9.3 Peak and Dip Equalization

Filters that generate peaks or dips can also be implemented using active filters; a
simple example using a gyrator is shown in Figure B.29.
Amplifier
with high input
and low
C1 R3 output impedance
R4
+
A=+1
–
P D
R1
Zgyr ein eout
Zgyr
R2 R5
C2
Basic gyrator circuit Simple circuit that can give peaks or dips.
Typically R3 = R4 = R5 and peak height or
dip depth depends on R4 slider position.
FIGURE B.29 Simple gyrator circuit. left: Simulation of inductance, right: narrowband
filter.
The circuit topology is similar to that of the Sallen–Key filter. The input imped-
ance of the gyrator is
1
Z gyr = R1 + R2 + + jωR1R2C2 (B.53)
jωC1
The resonance frequency is given by
1
ω0 = (B.54)
R1R2C1C2
For R1 = R2, the Q factor is given by
C2 (B.55)
Q=
4C1

B.9.4 Phase and Group Delay Equalization

A passive circuit to delay signals is shown in Figure B.14. A major drawback of this
circuit is the use of inductance. Figure B.30 shows an active circuit that eliminates
the inductance and will produce delay.
The circuit has the transfer function
1 − jωRC
H (ω ) = (B.56)
1 + jωRC
Amplifier
with high input and
R1 R1 low output impedance
–
+
R
ein C
eout
Delay circuit
FIGURE B.30 Simple circuit for achieving phase adjustment and delay without an inductor.
This transfer function introduces a frequency-dependent delay of the input voltage
2 RC
τ (ω ) = (B.57)
1 + (ωRC )2
For small values of ωRC, the delay will be approximately frequency independent
and equal to τ ≈ 2RC. The operating range of the filter is determined by the choice of
RC. For a wideband delay, the RC values should be chosen so that ωRC < 0.1 at the
highest frequency of interest. A drawback of any active circuit is the electrical noise
and nonlinear distortion that is added by each amplifier stage.
By series combination of this circuit, it is also possible to have a reasonably
constant phase change over a wide range. This circuit is sometimes used to obtain
signals in quadrature, that is, they are 90° offset from one another.
B.9.5 Digital Filters
Using computers it is possible to enhance filter design by starting with a standard
filter and then “tweak” the filter characteristics so that the combined loudspeaker/
crossover filter characteristic is optimized.
Digital filters can be designed to simulate the classical filters described previously.
However, because of the flexibility inherent in digital signal processing, many more
(and more advanced) filter designs can be implemented. One example is the matched
filter used to normalize frequency response functions. Most digital filters are infinite
impulse response filters. A different type of filter is the finite impulse response filter.
Because of signal sampling frequency issues, the limited amplitude resolution
of any digital audio system, digital filters are likely to introduce new, and possibly,
audible signal artifacts.
Since real-world electroacoustic components are likely to change over time, due
to material aging, thermal factors, etc., while the digital signal system stays stable,
the end result of the combined analog and digital systems may be quite disastrous
from an audio viewpoint after some time. Digital audio equalization systems in
particular need to be adaptive so that they can accommodate changes in the systems
they are to equalize.
The reader is directed to Ref. [5] for a comprehensive review of the issues involved
in designing digital filters.
REFERENCES
1. Borwick, J. (ed.), Loudspeaker and Headphone Handbook, 3rd edn., Focal Press,
Oxford, U.K. (2001) ISBN-13: 978–0240515786.
2. Colloms, M., High Performance Loudspeakers, 6th edn., Wiley-Blackwell, Chichester,
U.K. (2005) ISBN-13: 978–0470094303.
3. Sallen, R. P. and Key, E. L., A practical method of designing RC active filters, IRE
Trans. Circuit Theory, 2(1), 74–85 (1955).
4. Zverev, A.I., Handbook of Filter Synthesis, Revised edn., Wiley-Interscience, New York
(2005) ISBN-13: 978–0471749424.
5. Ifeachor, E. C. and Jervis, B. W., Digital Signal Processing: A Practical Approach, 2nd
edn., Prentice Hall, New York (2001) ISBN-13: 978–0201596199.
6. Linkwitz, S. H., Active crossover networks for noncoincident drivers, J. Audio Eng.
Soc., 24(1), 2–8 (1976).
7. Linkwitz, S. H., In-phase crossover network design, J. Audio Eng. Soc., 34(11), 889–
894 (1986).
8. Leach, W. M., Loudspeaker driver phase response: The neglected factor in crossover
network design, J. Audio Eng. Soc., 28(6), 410–421 (1980).
9. Lipshitz, S. P. and Vanderkooy, J., A family of linear-phase crossover networks of high
slope derived by time delay, J. Audio Eng. Soc., 31(1/2), 2–20 (1983).
10. Blauert, J. and Laws, P., Group delay distortions in electroacoustical systems, J. Acoust.
Soc. Am., 63(5), 1478–1483 (1978).
11. Preis, D., Phase distortion and phase equalization in audio signal processing—A tutorial
review, J. Audio Eng. Soc., 30(11), 774–794 (1982).
12. Fleming, L., Frequency-dividing networks, J. Acoust. Soc. Am., 35(9), 1454–1454
(1963).
and Rooms, Focal Press, Oxford, U.K. (2008) ISBN-13: 978–0240520094.
14. Small, R.H., Constant-voltage crossover network design, J. Audio Eng. Soc., 19(1),
12–19 (1971).
Appendix C: Magnetic
Fields and Forces
C.1 INTRODUCTION
Magnetism is one of the fundamental properties of nature and can be due to aligned
atomic dipole magnets in materials (permanent magnets) and electric currents.
Magnetism is characterized by magnetic field strength H [A/m] and magnetic flux
Φ [Wb]. The flux can be imagined as composed of field lines having a vectorial
quantity, the more lines that cross a surface S [m2], the higher the flux density B,
expressed in tesla [T].
Φ=
∫ BdS (C.1)
S

Magnets are said to have poles. For historic reasons, the field lines are defined as
“leaving” the magnet at its “north pole” and entering at its “south pole.” Like magnet
poles repel, and unlike poles attract.
The magnetic field strength in a material depends on its permeability μ [H/m]
B
H= (C.2)
µ
Permeability can be considered analogous to conductivity in electrically resistive

materials. The higher the permeability, the less magnetic field strength is needed to
create a certain flux density in the material.
In contrast to low-frequency electrical circuits where it is usually easy to keep the cur-
rent inside the conductor, magnetic materials, even when they have high permeability,
are surrounded by considerable leaked flux as there are no magnetic insulators.
C.2 RELUCTANCE AND MAGNETOMOTIVE FORCE

The magnetic flux Φ can be modeled as caused by a magnetomotive force MMF [A],
sometimes called the “magnetic strength.” The MMF has strength NI for magnetic
fields induced by a current I in a coil of N turns, and M for fields due to permanent
magnet sources. The MMF in magnetic circuits can be treated as analogous to a battery
in an electric circuit. The MMF in a circuit is related to the magnetic field strength as
MMF =
∫ Hdl (C.3)
l

541
542 Appendix C: Magnetic Fields and Forces
The magnetic reluctance Rm [A/Wb] corresponds to the resistance in an electric

circuit. The magnetic flux in a circuit, such as one that consists of a combination of
magnetically soft and hard materials, can then be calculated knowing the MMF and
reluctance.
MMF = ΦRm (C.4)
The reluctance of a magnetic circuit component depends on its length lm,

permeability μ, and cross-sectional area Sm perpendicular to the flux lines and can
be calculated in a similar way to the resistance in an electrical circuit. (Remember
however that magnetic circuits often involve important nonlinearities.)
lm
Rm = (C.5)
µS m
A permanent magnet also has an internal reluctance. In a circuit, it can be modeled

using either a Thevenin or Norton circuit model. In the Thevenin model, the internal
reluctance appears in series with the MMF. In the Norton model, a magnetomotive
flux or MMΦ source appears in parallel with the internal reluctance, which is more
physically correct.
C.3 PERMEABILITY
The permeability of gases and fluids is nearly the same as that of free space, μ0,
which is 4π·10 −7 H/m. The permeability μ of a magnetic material is usually written as
μ = μrμ0 where μr is the relative permeability of the material. Ferromagnetic materials
are most important magnetic materials and exhibit high magnetic conductivity and
high relative permeability μr. The permeability of ferromagnetic materials depends
on the flux density in the material.
Magnetism in ferromagnetic materials appears as “magnetic domains.” In the
natural state of a ferromagnetic material, the domains are randomized but can be
aligned by an externally applied magnetic field in a soft material or in a material
that is permanently magnetized. If the properties of a material are suitable, the
application of the external field will result in remanence, i.e., the domains will remain
aligned. The material has become permanently magnetic and will be characterized
by a magnetomotive force and an internal reluctance. This is further discussed in
the following section on magnetic hysteresis. Figure C.1 shows magnetic flux lines
around a permanent magnet.
C.4 ELECTROMAGNETS
A conductor that carries an electric current generates a magnetic field that surrounds
the conductor, as shown in Figure C.2. If the conductor is wound to a cylindrical
coil, a solenoid, most of the magnetic flux will remain inside the coil, entering at
Appendix C: Magnetic Fields and Forces 543
N
S
FIGURE C.1 Magnetic flux lines around a permanent magnet. Flux lines are defined to go
from the “north pole” (N) to the “south pole” (S) of the magnet.
Magnetic flux lines
Direction of
electric current
Conductor
FIGURE C.2 Magnetic flux lines around a conductor carrying DC: the “right-hand rule.”
the “south pole” and exiting at the “north pole” as shown in Figure C.3. The flux
lines are directed as shown in the figures from north to south on the outside of the
magnet. The magnetic fields of electromagnets and permanent magnets share the
same properties. A magnetically soft material that is magnetized by a surrounding
coil carrying electrical current will act in the same way as a permanent magnet. The
flux line density will be high in a bar of ferromagnetic material that is inserted inside
the solenoid
MMF = iN (C.6)
The dynamic permeability of a coil that has a ferromagnetic core is different from
the static permeability because of the nonlinearity, including hysteresis described
FIGURE C.3 Magnetic flux lines around a solenoid made of nine turns of wire.
below, of ferromagnetic properties. In ferromagnetic materials, such as iron, steel,

nickel, and cobalt, the magnetic flux density will saturate at some level of the flux
density B, typically in the range 0.3–1.5 T.
C.5 INTERACTION BETWEEN MAGNETIC FIELDS

The interaction force between two objects having magnetic fields depends on the flux
density and direction of the fields. A simple case is that of a wire carrying electric
current immersed in a static homogeneous surrounding magnetic field as shown in
Figure C.4.
The direction of the magnetic field lines around a conductor, the “right-hand
rule,” is shown in Figure C.2. With the surrounding magnetic field directed as shown
in Figure C.4, the force on the conductor will be directed to the left. The force on a
wire of length l carrying the current i in a field having the magnetic flux density B is
given by the equation
F = Bli sin(β) (C.7)

where β is the angle between the direction of the current i and that of the lines
of flux.
C.6 PARALLEL AND SERIES MAGNETIC CIRCUITS

Figure C.5 shows a homogeneous toroidal magnetic core having high permeability
that is driven by a coil with an electrical current. The flux will almost be completely
contained in the core because of its high permeability. From the previous equations,
North
South
FIGURE C.4 An electric conductor carrying current in a magnetic field. The current is
directed perpendicularly into the page. The large arrow indicates the direction of the force
acting on the conductor.
Φ
Rm Sm
IE
MMF = NIE NIE Rm
N turns
Φ
lm
FIGURE C.5 An electric current induces a magnetic field in a ferromagnetic core. The flux
is determined by the reluctance and the MMF.
we can calculate the magnetic field strength H. With data for the length lm , cross
section Sm , and permeability μ of the core, we can calculate the reluctance of the
circuit. It is also important to calculate the flux density B in the core to make sure
it is not saturated.
Figure C.6 shows a magnetic circuit composed of two parallel core sub-circuits.
The magnetomotive force NI driving the two cores is the same, so the magnetic flux
in each core will depend on the core reluctances. The reluctances are calculated
from the core properties and the flux is then determined from the respective
reluctances.
Figure C.7 shows a series-coupled magnetic circuit consisting of a core with
an air gap. The flux is determined by the magnetic strength and the sum of the
Φ1 Φ2
Rm1 Sm1 Rm2 Sm2
N turns
MMF = NIE Rm1 NIE Rm2
Φ1
IE Φ2 lm2
lm1
FIGURE C.6 Two parallel ferromagnetic core circuits being driven by a common MMF.
The flux is split between the cores.
Rm Sm Rm
NIE
IE
dslit Rslit Rslit
MMF=NIE
N turns
Φ
lm
dslit << Sslit
FIGURE C.7 An electric current is driving a ferromagnetic toroidal core that has an air gap.
The system forms a series magnetic circuit. The total reluctance in the circuit is determined
by the sum of the reluctances of the core and the air gap. Here the air gap is assumed to be so
small compared to the width of the air gap that the leakage flux is negligible.
reluctances in the magnetic circuit. The reluctance of the air gap is high because
of the low permeability of free space. In many circuits containing air gaps, the
reluctance of the ferromagnetic object’s part of the circuit is so much lower than
that of the air gap that the flux in the circuit is fully determined by the air gap’s
properties. As long as the air gap opening length is much smaller than the cross
section’s dimensions, most of the flux will be contained within the air gap’s cross-
sectional area.
Most electrodynamic drivers use magnet systems that have axial symmetry. The
magnetic circuit of such a driver using a ferrite magnet has several reluctances, some
of which represent the leakage of flux out of the desired magnetic circuit, as shown
in Figure C.8.
When the reluctance of the driver’s magnet is very high, as in rare earth materials,
leakage paths may well have lower reluctance than the magnet’s internal reluctance
and these can play a considerable role. In practice, a good design approach is to use
finite element modeling of the magnetic motor system.
Rm2
Rm1 Rm3
Rm4
M, Rm8 Rm5
Rm7 Magnet
Rm6 Soft iron

structure
Rm1
MMF Rm8 Rm7 Rm2 Rm3 Rm4
Rm6 Rm5 Φair
FIGURE C.8 The magnetic circuit of an electrodynamic loudspeaker magnet system has
many reluctances. Reluctances: (1) top plate, (2) upper air gap leakage, (3) air gap, (4) lower
air gap leakage, (5) center pole piece, (6) lower plate, (7) leakage around magnet perimeter,
and (8) internal magnet reluctance. The flux in the air gap for the voice coil is Φair.
C.7 HYSTERESIS
The relationship between flux density B and magnetic field strength H of a
ferromagnetic material is described by the material’s magnetization curve.
Figure C.9 shows an example of the BH characteristics for a ferromagnetic material.
The curve shows hysteresis, i.e., the flux density behaves differently when magnetic
field strength diminishes than when it increases, as discussed in the next paragraph.
Hysteresis is caused by the nonlinear losses in the material. Notice that hysteresis is
always present for such a material even when the core is not saturated.
When the material is initially magnetized by a rising electrical current, the BH
relationship follows curve (a), but as the current drops, the magnetic field strength is
reduced and the BH relationship follows curve (b). As magnetization induced by the
alternating current changes sign, the working point flux density returns by curve (c).
When the magnetic field strength is reduced, some magnetic flux remains, which is
called the remanence Br (T). Remanence is present at all levels but is strongest when
the material has been driven into saturation. Remanence can be positive or negative
depending on the current direction.
Figure C.10 shows two typical magnetization curves for magnetically “soft”
and “hard” materials respectively. Coercivity Hc is a measure of the field intensity
required to reduce the magnetization of a material after the magnetization has be
driven to saturation, and is expressed in [A/m]. Magnetically hard materials have
2
Bmax
Br
Magnetic flux density B[T] 1 Hc (a)

Slope is µAC
0
(b) (c)
–1
–2
–200 –100 0 100 200
Magnetic field strength H[kA/m]
FIGURE C.9 A hysteresis curve characteristic for a ferromagnetic material driven close to
saturation.
2
Magnetic flux density B[T]
–1
–2
–200 –100 0 100 200
Magnetic field strength H[kA/m]
FIGURE C.10 Examples of hysteresis curve characteristics for a “hard” ferromagnetic

material (solid line) and a “soft” ferromagnetic material (dashed line). Note that the magnetic
field strength scale is given in (kA/m) here.
a high coercivity, and are considered to be permanent magnets. Magnetically soft

materials have a low coercivity, and return to a relatively low residual magnetism
when the applied field is removed.
It is important to note that the hysteresis curve is not continuous since the
magnetization of the ferromagnetic material is due to the magnetic domains of the
material aligning themselves with the applied magnetic field. Because of the discrete
nature of the domains, the magnetization takes place in small irregular bursts or
steps as shown in Figure C.11. This is called the Barkhausen effect and results in
granulation noise at relatively low magnetization levels in transformers.
Magnetic flux density [T] 1
–1
–2
–200 –100 0 100 200
Magnetic field strength [A/m]
FIGURE C.11 The flux density does not change continuously in a ferromagnetic material.
As the magnetization force changes, the domains switch directions and the flux density
changes in steps, called Barkhausen jumps.
C.8 DYNAMIC MAGNETIZATION

The area inside the hysteresis curve corresponds to the losses in the magnetic
system. When an AC-current induces its magnetic field, this magnetic field
also permeates the core. If the core has high electrical conductivity, there will
be eddy currents induced along the entire core length, which leads to losses
and nonlinearities in the system. The core can be considered a sort of shorted
secondary winding. To reduce the influence of this “winding,” an additional
“shorting ring” of copper or aluminum around the pole piece can be used, i.e., a
“third” winding having higher electrical conductivity than the soft iron core. The
ring acts as a short circuit secondary in a transformer. The copper ring has much
lower electrical resistance than the soft iron steel core and so the eddy current in
the ring will be large so the flux is prevented from entering the material. The eddy
current losses in the core can be further reduced by using thin, electrically isolated
sheets of magnetic material for the core, using iron powder cores, or by using
ferrite cores that intrinsically have no electrical conductivity due to their ceramic
nature. However the low heat conductivity of ferrite cores reduces overall magnet
system cooling and results in more voice coil heating, and thus increased electrical
resistance, leading to audio compression effects.
Flux modulation is also a result of the AC-current-induced magnetic field
in the core. The changing flux results in nonlinear distortion in the drive since
the conductor is in a time-varying magnetic field. To reduce this additional flux
modulation, is another reason to attach a shorting ring to the center pole piece. Since
the ring reduces the flux modulation, distortion will be reduced. For best results in
reducing flux and inductance modulation, the shorting ring should be placed in the
loudspeaker’s air gap. However, this increases the air gap and the reluctance of the
magnetic circuit.
C.9 INDUCTANCE
The inductance of a coil is due to the energy in the magnetic field around the coil
created by the electrical current in the coil. Higher inductance generally results in
higher efficiency and in less current in the voice coil. The voice coil inductance LE
(H) is an important property of electrodynamic loudspeakers. The inductance of the
nonmoving voice coil LE0 is
Φ
LE 0 = N (C.8)
i
The presence of a ferromagnetic core in the coil will increase its inductance. The
dynamic permeability of the core changes depending on where on the hysteresis
curve the magnetization takes place. This means that there will be modulation of the
inductance and the drive current in a loudspeaker’s voice coil depends on its position
in the air gap. The presence of a shorting coil in or close to the air gap reduces
the magnetic field inside the magnet circuit and also the influence of inductance
modulation. For minimal inductance variation, the shorting ring should be placed on
the inside of the air gap, typically on the pole piece.
The effect discussed above can still be observed even if the voice coil is prevented
from moving. This inductance modulation should not be confused with the one
that results from voice coil movement relative to the center pole piece. Again. the
presence of a shorting ring reduces distortion since it reduces the influence of the
core magnetization on the instantaneous inductance, due to the time-dependent
location of the voice coil relative to the core.
C.10 DESIGN OF PERMANENT MAGNET CIRCUITS

Typically, magnetic circuits used in electroacoustics have an air gap where the flux
density should be as large and homogeneous as possible so that the force acting on a
current conductor in the field is constant as it moves in the field.
Figure C.12 shows an example of a thin circular slit in a magnetic circuit
consisting of two soft iron pole pieces and a cylindrical block-shaped magnet
having a length lm and a circular cross-sectional area Sm. For a thin slit dslit, the flux
density will be the same as in the magnetic material if dslit is much smaller than
√Sslit. The purpose of the pole pieces is to keep the magnetic field in the air gap
area Sslit. The pole pieces are assumed not to be saturated and to have negligible
reluctance compared to the slit.
The “power” in both reluctances should be the same to minimize the magnet
volume necessary to achieve the desired flux density in the air gap
dm d
Φ 2m = Φ 2slit slit (C.9)
µ r µ 0 Sm µ 0 Sslit
Φm Φslit
Rmi Sslit
dm MMF dslit Rslit
Sm MMF Rm, i Rm, slit
FIGURE C.12 A cylindrical permanent magnet is driving a ferromagnetic semi-toroidal

core that has an air gap. The air gap is assumed to be so small compared to the width of the
air gap that the leakage flux is negligible. The reluctances of the pole pieces are neglected
since they are assumed small compared to the air gap reluctance.
and since the fluxes Φm and Φslit are
Φ m = Bm Sm
Φ slit = Bslit Sslit (C.10)
Equation C.9 becomes
dm Sm Bm2 dslit Sslit Bslit

2
= (C.11)
µr µ0 µ0
which can be rewritten to give the volume of the magnet
dslit Sslit Bslit

2
dm Sm = (C.12)
µ 0 ( Bm H m )max
The magnetic energy density of a permanent magnet is given by its BH product,

typically given in units of (kJ/m3). Equation C.11 shows that the magnet can be made
small if the BH product of the magnet material is high. Figure C.13 shows the BH
product for the hysteresis curve of a magnetically hard material. The maximum BH
product is (BmHm)max and is obtained for Bm. The magnetic field strength at this point
of maximum energy is Hm.
Since ferrite materials have a very low (BmHm) max typically about 30 kJ/m3,
while for a metal alloy it may be as high as 300 kJ/m3, ferrite cores need to be
much larger than metal alloy cores for the same magnetic energy. This increased
size makes it necessary to place the ferrite on the “outside” part of the loudspeaker
magnet system, and this leads to large stray magnetic fields. This is a problem in
the loudspeaker design shown in Figure C.8. In practice, leakage will play a more
important role in magnet systems that use high internal reluctance magnets than in
1.5 Br
Magnetic flux density B[T]

1.0
Bm
0.5 Hc
0.0
Hm (BmHm)max
–100 –50 0 50
Magnetic field strength H [kA/m] BH [kJ/m3]
FIGURE C.13 The second quadrant of the hysteresis curve is shown here along with a graph
of the BH product at various points on the curve. A simple graphical method can then be used
to determine the operating point on the hysteresis curve that gives the maximum magnetic
energy density.
those that use classical magnet materials, such as Alnico, that have relatively low
internal reluctance.
The low thermal conductivity of the ceramic ferrite results in a lower magnet system
thermal capacity and heat transfer from the voice coil through the magnet system
structure to the surrounding air. Loudspeakers having ferrite magnets have lower
power handling capacity than loudspeakers using interior metal magnets. Another
advantage of internal metal magnets is that the stray flux to the outside is much smaller
than that of loudspeakers using external magnets, which means that the loudspeaker
does not influence magnetically sensitive equipment such as cathode ray tubes.
BIBLIOGRAPHY
Aldoshina, I. et al., Modeling of flux modulation distortion in moving coil loudspeakers by the
finite element method, Proc. 98th Audio Engineering Society Convention, Paris, Paper
3996 (1995). Accessed October 2012
Klippel, W., Loudspeaker nonlinearities (www.klippel.de).
Leach, W. M. Loudspeaker voice-coil inductance losses: circuit models, parameter estimation,
and effect on frequency response, J. Audio Eng. Soc., 50(6), 442–449 (2002).
Mazin, V. Modeling of magnetic hysteresis and its influence on harmonic distortion in
electrodynamic loudspeakers, Proc. 106th Audio Engineering Society Convention,
Munich, Paper 4865 (1999).
Vanderkooy, J., A model of loudspeaker driver impedance incorporating eddy currents in the
pole structure, J. Audio Eng. Soc., 37(3), 119–128 (1989).
Appendix D: Time-Domain
Approach to Directivity
D.1 INTRODUCTION
Point source synthesis is used for the study of sound radiation, particularly with
regard to directivity. Because of reciprocity, the results found for loudspeakers will
also hold for microphones that have the same geometrical and mechanical properties.
Point source synthesis may be used in both the time and frequency domains. The
frequency-domain method is described in Chapter 12. In this appendix, we study how
the same results can be obtained by using time-domain point source synthesis. The
appendix closely follows Kuttruff’s approach in Ref. [3]. This reference is in German
and may not be accessible to English readers. Those knowledgeable in German are
recommended to read Kuttruff’s treatment in the original. More material on this
approach to the study of directivity may also be found in Refs. [1,2,4,5].
Using the frequency-domain approach, we considered sound radiation by a
monopole and wrote the sound pressure at some distance r from a monopole as
A − jkr
p(k, r ) = e (D.1)
r
Here A is the amplitude at unity distance. We note that there is a phase delay with
increasing frequency since
ω
k= (D.2)
c
We know from Appendix B that a phase change in a filter transfer function

corresponds to a group delay τ according to
dϕ
τ=− (D.3)
dω
Differentiating the argument −kr in the exponential in Equation D.1, we find that
r
τ= (D.4)
c
which is the time the sound will take to travel the distance r at speed c.
553
554 Appendix D: Time-Domain Approach to Directivity
D.2 DELTA AND STEP FUNCTIONS

Any physical time function can be regarded as a sum of progressively delayed Dirac
(delta) functions of different amplitudes. The Dirac function has the property
∞
∫ δ(t ) dt = 1 (D.5)
−∞
From the Fourier transform definition, we find that

ω′
1
δ(t ) = lim
ω ′→∞ 2 π ∫e jωt
dω (D.6)
−ω′
The integral of the Dirac function is the step function ε(t)

t
ε (t ) =
∫ δ(t ′) dt ′ (D.7)
−∞
Because the impulse response h(t) is the time derivative of the step function η(t)
of a physical system, the transfer function can be written as (provided that the step
response goes slowly toward zero for large times t, see Ref. [3]).
∞ ∞
dη(t ) − jωt
H (ω ) =
∫ dt
e dt = jω
∫ η(t ) e − jωt
dt = jωF[ η(t )] (D.8)
−∞ −∞
D.3 POINT SOURCE SYNTHESIS IN THE TIME DOMAIN

In Chapter 10, we found that by summing the sound pressure from a distribution of
monopoles, each having a volume velocity udS on a plane and rigid baffle, we could
obtain Rayleigh’s integral
ρω e − jkr
p(ω, r ) = j
2π
u
∫∫
Ssources
r
dS = H (ω, r )u = jωF[ η(r, t )]u (D.9)

We consider H(ω,r) the transfer function of the radiation “system.” The transfer
function of a system is found by applying the Fourier transform F to the
impulse response of the system, but we will see that it is more practical to look
at the step function response. Since the sound pressure is proportional to source
volume acceleration, the sound pressure due to a distribution of identical sources
with velocity u can also be written as
ρu 1  r
p(r, t ) =
2π ∫∫
Ssources
δ t − dS = η(r, t ) ∗ u(t ) (D.10)
r  c 

Appendix D: Time-Domain Approach to Directivity 555
with the step response of the source distribution being
ρ 1  r
η(r , t ) =
2π
Ssources
∫∫ δ t − dS (D.11)
r  c 

We call this the geometrical step function of the source distribution.
D.3.1 Plane Piston in a Baffle

An application of the time-domain theory to a plane rigid piston in a plane rigid
baffle follows. It is practical to introduce a test sphere having a radius ct as shown
in Figure D.1.
The radius ct connects all elements that, at any time, contribute to the sound
pressure at point P. One can then write the preceding equation as
ρc
η(r, t ) = Θ(t ) (D.12)
2π
Here Θ(t) is the length of the circular segment that covers the piston S 0 a a function
of time.
Figure D.2 shows the construction used in the analysis below. The receiving point
P is located as shown at a distance y from the piston’s normal axis and the test sphere
at time t has a circle that is fully within the piston area. As long as y = 0, the step
response will be rectangular with a duration

∆t =
1
c ( )
a 2 + x 2 − x (D.13)
ct
S0
P
Θ(t)
ct
FIGURE D.1 Construction of test sphere used to calculate the step response of a piston with
arbitrary geometry. (After Kuttruff, H., Physik und Technik des Ultraschalls (in German),
Hirzel Verlag, Stuttgart, Germany, 1988.)
ct
s
x P
θ
ct ct
s x P y
y
ct
a a
(a) (b)
FIGURE D.2 Construction to help determine the step response of a circular piston. (a) y < a,
(b) y > a. (After Kuttruff, H., Physik und Technik des Ultraschalls (in German), Hirzel Verlag,
Stuttgart, Germany, 1988.)
If y is in the range 0 < y < a, the step response will start with a constant magnitude
lasting until
s = (ct )2 − x 2 (D.14)

and then monotonically drop as the circle’s radius, where the sphere is cut by the
plane, drops to zero. One finds that the step response can be written
ρc 0 < s < a − y when a > y


 ρc  s2 + y2 − a2 
η( x, y, t ) =  cos −1   a − y < s < a + y (D.15)
π  2sy

0 all other values of s

The geometrical step responses that result from this expression are shown graphically
in Figure D.3 for four different values of y when the observation point P is at a
distance of 2a from the plane.
We will now look at the behavior of the geometrical step function in the far
field where the distances to all points on the piston are about equally long. The cut
between the sphere and the plane will now be approximately a straight line that
moves over the piston as shown in Figure D.4.
Except for y = 0, resulting step response is
ρc

 πr sin 2 (θ) (a 2
sin 2 (θ) − (r − ct )2 ) r − a sin(θ) < ct < r + a sin(θ)
η( x, y, t ) = 

0 all other values of ct

(D.16)
y=0
2 3
ct/a
y = a/2
2 3
ct/a
y=a
2 3
ct/a
y = 3a/2
2 3
ct/a
FIGURE D.3 The step response of a circular piston for various offsets from the axis of
symmetry for a circular piston. (After Kuttruff, H., Physik und Technik des Ultraschalls
(in German), Hirzel Verlag, Stuttgart, Germany, 1988.)
ct
ct r y = r sin(θ)
l
θ
x = r cos(θ) r – a sin(θ) r + a sin(θ)
a
ct
FIGURE D.4 Calculation of the step response of a circular piston as a function of angle
to the normal and the resulting step response. (After Kuttruff, H., Physik und Technik des
Ultraschalls (in German), Hirzel Verlag, Stuttgart, Germany, 1988.)
We note that the shape of the geometrical step response depends primarily on the
distance r and the observation angle θ. On the center normal of the piston, the step
response is that given by Equation D.15 for x < ct < x + a2/2x.
Two cases are of particular interest–the pressure response on and off axis.
Integrating Equation D.9 over the piston surface, one finds the on-axis sound
pressure as

(
p( x, t ) = ρcu e − jkx − e − jk x 2 + a2
) (D.17)
Where the first part corresponds to a plane wave emitted from the piston and the
second part corresponds to the diffracted wave from the edge of the piston.
To calculate the off-axis response, we introduce a new integration variable x′
r − ct
x′ = (D.18)
sin(θ)
From Equations D.8 and D.16 we see the resulting pressure response off axis can be
written as
a
jωρ − jkr
H (ω, r, θ) =
πr
e
∫
−a
a 2 − x ′ 2 e jkx ′ sin(θ )dx ′ (D.19)

The integral is a definition of the first-order Bessel function, and evaluating it, the
transfer function becomes
jωρS − jkr 2 J1 ( ka sin(θ))

H (ω, r, θ) = e (D.20)
2πr ka sin(θ)
If we feed a signal u to a filter that has the transfer function H, we have an output Hu
so the resulting sound pressure p̲ in the far field of the piston for a piston vibration
velocity up will be
 2 J1 ( ka sin(θ))  e − jkr
p(ω, r, θ) = jωρ   ( )
πa 2 u p (D.21)
 ka sin(θ)  2πr
This is the same as Equation 12.16 found by working in the frequency domain. An
important result of the theory presented in this appendix is that only the spherical
“breathing” sphere can have a Dirac-type impulse response.
REFERENCES
1. Stepanishen, P. R. Transient radiation from pistons in an infinite planar baffle, J. Acoust.
Soc. Am., 49(5B), 1629–1638 (1971).
2. Pierce, A.P., Acoustics: An Introduction to Its Physical Principles and Applications,
American Institute of Physics, New York (1989) ISBN-13: 978-0883186121.
3. Kuttruff, H., Physik und Technik des Ultraschalls (in German), Hirzel Verlag, Stuttgart,
Germany (1988) ISBN-13: 978-3777604275.
4. Piwakowski, B. and Delannoy, B., Method for computing spatial pulse response: time-
domain approach, J. Acoust. Soc. Am., 86(6), 2422–2432 (1989).
5. Jiang, C. et al., Impulse response and frequency response of a line loudspeaker array,
Proceedings of the 117th Audio Engineering Society Convention, San Francisco, Paper
6248 (2004).
Appendix E: Sound-
Absorbing Materials
E.1 SOUND ABSORPTION COEFFICIENT
When a sound wave impinges or contacts a material or construction, some of the
incident power Winc will be absorbed, Wabs. The sound absorption of materials is
characterized by the sound absorption coefficient, α (alpha), which is defined as
Wabs
α= (E.1)
Winc
Sound absorbers are grouped into resonant and nonresonant absorbers. Nonresonant
absorbers are usually made of porous materials. This group includes mineral wool,
glass wool, open pore plastics, and similar materials and these are examined in
this appendix.
E.2 POROUS SOUND ABSORBERS

E.2.1 General Properties
The conversion of acoustic to thermal energy in porous absorbers is due to the vis-
cous behavior of the air flowing in the canals, pores, and air pockets of the material.
Porous materials that do not have open pores cannot function as porous absorbers;
an example of such a material is expanded polystyrene. If the flow is blocked by, for
example, paint or other surface coverings, the absorption coefficient will usually
drop considerably, but if the material is elastic, there may still be some absorption.
It is important to differ between the sound absorption and the heat insulation
properties of porous materials. Porous materials may be good thermal insulators and
still have poor sound absorption properties. Expanded polystyrene and many other
porous plastic materials made for heat insulation purposes have poor sound absorp-
tion properties. A survey of modern porous sound-absorbing materials can be found
in Ref. [2].
E.2.2 Models for Sound Absorption

The porosity of an open pore material is important to its sound absorption properties.
The porosity σ is the ratio between the pore volume and the total volume required
to hold the material. Commercial products for sound absorption, such as glass wool,
may have a porosity larger than 0.95.
561
562 Appendix E: Sound-Absorbing Materials
There are many options in the manufacture porous materials using fibers, hairs,
weaves, etc., or foam. In commercial materials, such as glass or mineral wool, the
fibers used are packed so densely that the width of the open canals between fibers is of
the order of a few hundred micrometers. In common fibrous materials, such as glass
and mineral wool, the fibers have diameters in the range 3–6 μm. These materials
have a volume porosity σv (ratio between the air volume to the total volume of the
material) of about 0.92–0.99. The density of the materials is in the range 25–75 kg/m3.
A major characteristic of these fibrous materials is that the body (skeleton) of the
material is relatively rigid. For thick samples, the material does not move at audio
frequencies. Materials such as animal hair (felt, wool) are more pliable than glass or
mineral wool although felts usually have lower volume porosity. The structure factor
χ is a number given to porous materials to describe the influence of the pore shape on
sound propagation in the material. For glass and mineral wool, χ ≈ 1.3 [6].
The most important acoustical feature of a porous material is its specific flow
resistance Ξ, i.e., the flow resistance per thickness unit. It is usually measured with a
static air flow. Manufacturers of porous materials can usually supply flow resistance
data for their products. See also Ref. [4]. The acoustic resistance of a small sample of
the material having thickness d in a tube having cross section area S is
Ξd
RA = (E.2)
S
The Rayleigh model theory for porous materials agrees quite well with measured
data [3]. One finds that the complex wave number kabs in a porous “homogeneous”
absorber material is
ω Ξσ v
k abs (ω) = χ− j (E.3)
c ωρ

Here c is the speed of sound for the free wave. Using the equation of motion, we find
that for plane waves moving in the positive x-direction, the sound field impedance in
the absorber Z abs (and small damping) is
p+ ( x, k ) Ξσ v
Z abs (ω) = = ρc χ − j (E.4)
u+ ( x, k ) ωρ

We can now calculate the attenuation of the wave as it moves through the material.
The sound pressure of the wave can be written
p( x, k abs ) = pˆ e − jk abs x (E.5)

The attenuation of the wave ΔL [dB] per wavelength will be [6]
∆L ≈ 55 dB/λ (E.6)

The damping inside the material is very high per wavelength, so at high frequencies,
the porous sheet can be relatively thin. If the material is placed against a rigid wall,
Appendix E: Sound-Absorbing Materials 563
the wave will be reflected back into the porous material with a reflection coefficient
close to 1 since the wall has much higher impedance than the porous material.
To consider the material as an excellent sound absorber, it needs to attenuate the
wave (round-trip) by at least 20 dB. This corresponds to an absorption coefficient
α greater than 0.99. Of course, some power is reflected at the surface of the porous
material, so obtaining a high sound absorption can require a thick and porous material.
The impedance seen by a wave that is incident on a sound-absorber is of interest
since it determines the reflection and sound-absorption coefficients. For a plane
wave impinging perpendicularly on a rigid wall covered by a sheet of porous sound-
absorptive material of thickness d the impedance seen by the wave is
p+ (d, ω) + p− (d, ω) Ξσ v 1 + e − j 2 kabs d

Z abs (d, ω ) = = ρc χ − j (E.7)
u+ (d, ω) + u− (d, ω) ω ρ 1 − e − j 2 kabs d

We find that there will be maxima in the sound absorption when Z abs is best matched
to the sound field impedance of the free wave. Since k = ω/c, the reflection coefficient
can be written as
Z abs (d, ω) − ρc
r (d, ω ) = (E.8)
Z abs (d, ω) + ρc
By using the relationship between sound absorption coefficient α and reflection

coefficient r, we can calculate α for different thicknesses d and specific flow
resistances Ξ of the porous material
2
α(d, ω) = 1 − r (d, ω) (E.9)

An example of the dependence of α on the specific flow resistance Ξ is shown
in Figure E.1 for a 10 cm thick sheet of porous absorber mounted against a rigid
surface. Figure E.2 shows how α depends on the sheet thickness d for a porous
absorber with Ξ = 16,000 Ns/m4, mounted against a rigid surface. This value of
flow resistivity corresponds to that of a glass wool blanket having a bulk density
of about 30 kg/m3.
We see from the curves that there is a limit to the blanket thickness beyond
which the absorption coefficient does not appreciably increase. Once Ξ becomes
large, it is better to mount the material with an airspace to improve sound absorption
at low frequencies. Marginal improvements in α can be obtained by using layers of
different Ξ, with the largest Ξ value layer closest to the rigid wall.
E.2.3 Influence on Wave Speed

Sound waves in porous materials will travel at a lower speed than sound waves in
free air, with the speed given by the real part of the complex propagation constant.
We see from Equation E.3 that for reasonably high frequencies,
kabs
re
≈ k χ (E.10)

1.0
Ξ = 256
Ξ = 64
0.8 Ξ = 16
Ξ=4
Absorption coefficient
Ξ=1
0.6
0.4
0.2
0.0
20 50 100 200 500 1k 2k 5k 10 k 20 k
Frequency [Hz]
FIGURE E.1 The frequency dependence of α for a 10 cm thick homogeneous porous
absorber assumed mounted against a rigid surface, with specific flow resistance as parameter.
Specific flow resistance Ξ given in units of 103 Ns/m4. Here χ = 1.3 and σ = 0.95.
1.0
d = 0.0125
d = 0.025
0.8 d = 0.05
d = 0.1
Absorption coefficient
d = 0.2
0.6
0.4
0.2
0.0
20 50 100 200 500 1k 2k 5k 10 k 20 k
Frequency [Hz]
FIGURE E.2 The frequency dependence of α for a homogeneous porous absorber

assumed mounted against a rigid surface with absorber thickness d (m) as parameter.
Here Ξ = 16·103 Ns/m4, χ = 1.3, and σ = 0.95.
Since χ is about 1.3 for glass and mineral wool, the speed should decrease by about
15% in these materials. This corresponds to a speed of about 291 m/s rather than
the 343 m/s for the free wave. Measurements on practical glass and mineral wool
materials show even lower speed values. One finds values in the range of 270–280 m/s
for commercial qualities of glass and mineral wool used for heat insulation having a
density of about 70 kg/m3.
RA2 MA1
RA1
CA1
CA2
FIGURE E.3 Element CA1 represents the compliance of the air in the material. Elements
RA1 and CA2 represent the heat exchange with the skeleton. RA2 is the acoustic resistance and
represents the friction between the air and the skeleton and MA1 is due to the mass of the air
in the pores. (After Mechel, F.P. et al., Formulas of Acoustics, 2nd edn., Springer, Berlin,
Germany, 2008.)
If the loudspeaker box is filled by such a material, the effective “acoustical”

volume of the box will increase by about 30% since the acoustic compliance CAB is
proportional to 1/c2. At the same time, there may be considerable damping so that
it will be necessary to introduce an extra component RAB representing the losses in
the box air volume into account.
E.2.4 Electroacoustic Analogy for Porous Materials

The acoustic impedance behavior of the porous material will be associated with
acoustical mass, compliance, and resistance, as shown in the acoustic impedance
analogy of a porous material shown in Figure E.3.
More detailed circuit models for the impedance of porous materials may be found
in Refs. [3,5].
E.2.5 Inhomogeneity of Porous Materials

It is important to remember that glass and mineral wools are manufactured using a
process similar to that one sees at fairs for manufacturing candy floss. The molten
material is thrown out through a rotating nozzle and collected on a conveyor belt,
cooled, treated with resin, and cut to suitable dimensions. Clearly, there may be large
variations in measured Ξ values between samples picked at different positions of the
manufactured material.
E.2.6 Other Considerations
The wave speed in the enclosure will also be reduced when extensive bracing and
honeycomb structures are used to stiffen the box. The bracing can act like the
acoustic lens structures discussed in Chapter 12. Their use can lead to a reduction of
the internal resonances of and in the loudspeaker enclosure.
E.3 MEASUREMENT OF SOUND ABSORPTION

The measurement of the absorption coefficient can be done by several methods. The
two most common ones are the plane wave tube (Kundt’s tube) and the room rever-
beration chamber methods. One must not confuse the sound absorption that can be
obtained inside a loudspeaker box with the sound absorption that is measured using
these methods.
E.3.1 Plane Wave Tube

The plane wave tube method allows direct measurement of the absorption coefficient
α of a material. A wave is incident on a test sample, usually 5–10 cm thick, that is
mounted against a rigid backing. The absorption can be calculated from the ratio of
the maximum and minimum pressures in the standing wave in front of the material.
An alternative measurement method allows convenient measurement of both the
complex reflection coefficient r and α.
This method requires measurement of the complex sound pressures _p1 and _p2 at
two points Δx apart at a distance d to the material as shown in Figure E.4.
Given that _p1 and _p2 are

( )
p1 (d, ∆ x, ω) = pИ 1 + re − jk ( d + ∆x ) (E.11)
2 2

( )
p2 (d, ∆ x, ω) = pИ e − jk∆x + re − jk 2 d (E.12)
The transfer function H21 between the two measurement points is given by the ratio
of the pressures
p2 (d, ∆ x, ω)
H 21 ( ∆x, ω ) = (E.13)
p1 (d, ∆ x, ω)

Using this transfer function, we can now calculate the reflection coefficient r from
Equations E.8, E.11, and E.12
Microphones
Loudspeaker p1 p2 Test sample
∆x d
x
0
FIGURE E.4 The measurement configuration for the two microphone method for measuring
α(ω) and r(ω) using a Kundt’s tube.
H 21 ( ∆x, ω) − e − jk∆x
r (ω) = e jk ( 2 d + 2 ∆x ) (E.14)
e − jk∆x − H 21 ( ∆x, ω)
Once the reflection coefficient r is obtained the sound absorption coefficient α can be
calculated using Equation E.9. By fitting the measured reflection coefficient data for
various thicknesses of the material to the equation model, one can better understand
the material’s acoustic properties.
E.3.2 Room Method
In the reverberation chamber, all surfaces are glossy and rigid so to absorb very little
acoustic energy. The sample to be measured must be large compared to wavelength,
usually 10 m2. The reverberation time is measured for the desired range of frequen-
cies with and without the test sample. Using Sabine’s equation, α can be calculated
from measurement of the reverberation time of the chamber with the absorptive
material present T60,w and absent T60,e
V 1 1 
α ≈ 0.16 − (E.15)
S  T60,w T60,e 


where V is the room volume and S the surface area of the absorptive material (See
Ref. [7]). The use of Sabine’s equation is limited to measurement conditions where
the sound field in the room is diffuse. Because of this the absorption coefficient mea-
sured using the room method is not directly applicable to the case of the material
being used in loudspeaker box. The sound field is only diffuse for frequencies above
the Schroeder frequency given by Equation F.20. This lower frequency limit occurs
at 100–150 Hz in measurement chambers used for this purpose.
The tube and the room methods do not give the same results because of the
different sound field conditions. The sound field inside a loudspeaker box is neither
a plane wave nor a diffuse field, so the best method is to measure the specific flow
resistance of the material (in various directions), average over a large number of
samples, and then use the average in a finite element model of the loudspeaker. If
this possibility is not available, the plane wave tube method is superior to the room
method in the determination of the sound absorption properties that can be expected
when the sound-absorbing material is used inside loudspeaker boxes.
REFERENCES
1. Ingard, U., Noise Reduction Analysis, Jones & Bartlett Publishers, Boston, MA (2009)
ISBN-13: 978-1934015315.
2. Arenas, J. P. and Crocker, M. J., Recent trends in porous sound-absorbing materials,
Sound Vib. Vol. 44 (7), pp. 12–17. (July 2010) (available from www.SandV.com).
3. Mechel, F. P., et al., Formulas of Acoustics, 2nd edn., Springer, Berlin, Germany (2008)
ISBN-13: 978-3540768340.
4. Bies, D. A. and Hansen, C.H., Flow resistance information for acoustical design, Appl.
Acoust., 13, 357–391 (1980).
5. Leach, W. M., Electroacoustic-analogous circuit models for filled enclosures, J. Audio

Eng. Soc., 37(7/8), 586–592 (1989).
6. Cremer, L. et al., Principles and Applications of Room Acoustics, Vol. 2, Applied
Science Publishers, New York (1982) ISBN-13: 978-0853341147.
7. Kleiner, M., Acoustics and Audio Technology, J.Ross, Ft. Lauderdale (2012) ISBN-13:
978-1-60247-052-5.
Appendix F: Resonance
in Boxes and Rooms
F.1 RESONANT MODES IN ROOMS
Many loudspeaker boxes and listening rooms are similar to shoebox-shaped
rectangular volumes. Such an idealized box is shown in Figure F.1 and is assumed to
have rigid walls. Assume further that the box is empty so there is no damping due to
sound absorption by some porous materials, no flow of air (except that due to sound),
no heat flow into the walls, and no viscous losses at the walls (due to particle velocity
components parallel to the walls). We also assume the sound source (if there is one)
to be a harmonic point source.
The instantaneous sound pressure inside the box will be a function of time,
frequency, and the room coordinates (x, y, z). The spatial distribution is described by
a function Ψ(x, y, z) that will depend on frequency.
p( x, y, z ) = pИΨ( x, y, z ) e jωt (F.1)

At very low frequencies where the wavelength of sound is much larger than the box
dimensions, the sound pressure will be constant throughout the box, Ψ(x, y, z) ≈ 1.
At higher frequencies, the box air volume will become resonant and the sound
pressure is determined by the wave equation, which, for a three-dimensional
Cartesian coordinate system, can be written as
∂2 p ∂2 p ∂2 p
+ + + k 2 p = 0 (F.2)
∂x 2 ∂y 2 ∂z 2
The spatial distribution of the sound pressure inside the box is a function of the room
coordinates. Assume the box interior to have the dimensions lx, ly, and lz. Since the
box walls are plane and rigid, the particle velocity perpendicular to the respective
walls must be zero, i.e.,
ux x = 0,l x
= 0; uy y = 0,l = 0; uz z = 0,l = 0 (F.3)
y z

These conditions are fulfilled with a standing wave pattern spatial distribution of
sound pressure Ψ(x, y, z) given by
Ψ( x, y, z) = cos(k x x ) cos(k y y) cos(kz z ) (F.4)

569
570 Appendix F: Resonance in Boxes and Rooms
lz
lx
ly
x
FIGURE F.1 The rectangular box under study.
The coefficients k x, k y, and kz can be regarded as the wave number components of

waves moving around inside the rectangular box, bouncing off the walls.
e − jkx x + e jkx x e − jky y + e jky y e − jkz z + e jkz z

Ψ( x, y, z ) = (F.5)
2 2 2
The waves will interfere and lead to standing wave patterns in the same way as in
the one-dimensional case of the plane wave tube (Kundt’s tube). Applying the wave
equation to Ψ(x, y, z), we find that the second-order derivative of _p with respect to x is
∂2 p
= − k x2 pˆ cos(k x x ) cos(k y y) cos(kz z) e jωt = − k x2 p (F.6)
∂x 2
and we find the second-order derivatives for the y- and z- directions similarly.
Inserting these second-order derivatives into Equation F.2, we obtain the condition
k x2 + k y2 + kz2 = k 2 (F.7)

where k is the wave number determined by the frequency ω of the sound, k = ω/c.
The room will be resonant for the combinations of kx, ky, and kz that fulfill the
Equation F.7. At resonance, the energy in the room will exist indefinitely if there are
no losses. This is called an oscillation mode of the room, an eigenmode, or just a mode.
The function Ψ, which describes the spatial variation of the sound pressure at
resonance frequency of a particular eigenmode, is called an eigenfunction. Figure F.2
shows an example of the normalized magnitude of the eigenfunction for such a mode.
To determine the eigenfunctions, we need to find values for the wave number
components k x, k y, and kz.
The particle velocity perpendicular to the walls, ux, uy, and uz, must be zero. We
start by studying the conditions in the x-direction, i.e., at the walls at x = 0 and x = lx.
The particle velocity is linked to the pressure distribution. Equation 3.2 gives us an
expression for the particle velocity ux(x, y, z) as a function of p̲ (x, y, z). We obtain the
particle velocity ux(x, y, z) by using the equation of motion for the plane wave
Appendix F: Resonance in Boxes and Rooms 571
1
0.5 0.5
ψ(x,y,z) 0
0.4
–0.5
–1 0.3
ly is 0.5 m
0.1 0.2
0.2 0.1
lx is 0.4 m 0.3
0.4
FIGURE F.2 Instantaneous sound pressure distribution for the (2, 4, 0) mode in a rectangular
box. The sound pressure varies sinusoidally with time at the mode’s resonance frequency for
free oscillations. Projected on the bottom of the graph are isobars for the sound pressure.
1 ∂p 1 1 ∂p

ux = −
ρ ∫ ∂x dt = − ρ jω ∂x (F.8)
which gives us
kx 1
ux ( x, y, z) = pˆ sin(k x x )cos(k y y)cos(kz z) (F.9)
ρ jω
Where the conditions for the particle velocity at the wall at x = 0 are fulfilled. In the
same way, we find that the solution F.4 also fulfills the conditions for the particle
velocity at the walls at y = 0 and z = 0.
According to expression F.8, for zero particle velocity at right angles to the wall
at x = lx, the following condition must be met:
sin(k xlx ) = 0 (F.10)
The sine will be zero only when the wave number component k x is
qx π
kx = (F.11)
lx
Here qx is a natural number, 0, 1, 2, etc. In the same way, we find that requirements
for the wave number components k y and kz are
qyπ qz π
ky = and kz = (F.12)
ly lz

Inserting these values for the wave number components into Equation F.7, we find
the condition for the wave equation to be satisfied (with the specified geometry and
boundaries) is
2 2 2
ω 2  q x π   q y π   qz π 
k2 = = +  + (F.13)
c 2  lx   ly   lz 

F.2 RESONANCE FREQUENCIES

There are an infinite number of combinations of qx, qy, and qz. Each of these
combinations will identify a possible mode (also called an eigenmode). The
resonance frequencies of these modes (also called eigenfrequencies) are determined
by the speed of sound and the room dimensions. The triplet (qx: qy: qz) is often called
the mode index. The k-values correspond to the resonance frequencies of the box
according to the equation
2 2 2
c  q x   q y   qz  (F.14)
fq x ,q y ,qz = +  +
2  lx   ly   lz 

For each eigenfrequency, there is at least one eigenmode. For certain room geometries,
many eigenmodes may belong to a certain eigenfrequency. In this discussion, we
have limited ourselves to the case of hard and immovable walls. If the box walls
are not rigid, i.e., have a finite impedance, the resonance frequencies will vary from
those calculated for the ideal case.
Each mode in the rectangular room is characterized by a spatial distribution of
sound pressure
 q πx   q πy   q πz 
pqx ,qy ,qz = pˆqx ,qy ,qz cos  x  cos  y  cos  z  (F.15)
 lx   ly   lz 

At any driving frequency, the sound field in a box with losses will be composed
of a superposition of pressure contributions from an infinite number of modes
since the losses are accompanied by the resonances having finite bandwidths. In
loudspeaker boxes losses at the walls are due to both wall vibration and leakage, for
example at joints. The mechanical properties of the wall sheets mean the walls being
vibratory systems that have their own resonance frequencies and these resonances
may couple to those of the air in the interior of the box. This leads to considerable
changes in both eigenfrequencies and eigenfunctions for the air inside the box. The
results of such coupling can in practice be measured by physical means or modeled
using, for example, the finite element method.
All the eigenfunctions have a large magnitude at the corners of the room for
the simple case studied here. Note also that the ratio between the sound pressure
and the particle velocity is highest at the corners, i.e., the sound field impedance is
highest at the corners. This means that a constant-volume velocity-type source will
excite the room modes best when placed in the corner of a room. Many technical
sound sources such as machines, loudspeakers, etc., are practical constant-volume
velocity-type sources.
Using a loudspeaker and a sine wave generator, one can study the properties of
the modes in a room. The summed sound pressure of several excited room modes
is going to depend on how the sound pressure contributions of each mode add up
with respect to magnitude and phase. This depends on the particular modes that
contribute at a particular location so the sound pressure amplitude varies from one
position to the next. At high frequencies where many modes “leak into one another,”
the sound pressure level will vary considerably from one position to the next due to
the interference between many modes.
When the sound source is switched off, the modes will start to oscillate at their
natural resonance frequencies, given by Equation F.14, if the modes are weakly
damped and uncoupled. Those modes that have resonance frequencies closest to the
excitation frequency are likely to have been excited most strongly and will tend to
dominate the decay process.
F.3 DAMPING
In practice, the room walls will be nonrigid, and there will be sound absorption by
the walls and by the sound-absorptive materials in the room, such as carpets and
furniture. The sound-radiating loudspeaker and other passive, that is non-powered,
loudspeakers in the room will also act as acoustic absorbers. Due to these factors,
the modal frequencies will change and the decay of the sound pressure in the room
can be written as
pq ( x, y, z) = pˆ Ψq ( x, y, z )e jωq −δqt (F.16)

Here q is the mode defining triplet q = [qx, qy, qz], ωq the new resonance frequency
specific for each mode, and δq its decay constant. These depend on the particular
distribution of the wall impedances affecting the sound field in the room. The decay
of each individual mode is exponential.
If one feeds sound into the room at a frequency close to the resonance frequency
of a mode, the sound pressure amplitude in the room, as a function of frequency, will
have a frequency response behavior similar to that of a simple mass–spring system,
as shown in Figure F.3. At sufficiently low frequencies, the modes may not overlap to
any large degree, but as the modal frequencies become higher, modes will increas-
ingly overlap.
How “strongly” any particular mode will be excited depends on how close its
resonance frequency is to the excitation frequency, the damping of the particular
mode, and the location of the sound source.
A metric to describe the “peakiness” of a resonance curve is the Q factor. The
Q factor can be considered the amplitude amplification factor at resonance for a
series resonant system such as the Helmholtz resonator shown in Figure 7.30. The
peak-level amplification will be 20 · log(Q). The Q factor is calculated from the
magnitude of the system response as a function of frequency as shown in Figure F.3.
Relative response [dB] 0
–1
–2
–3
–4
–5
ω0 – Δω ω0 ω0 + Δω
2 2
Normalized frequency [Hz]
FIGURE F.3 The resonance curve for a damped simple mass–spring system.
The Q factor is defined as the resonance frequency ω0 divided by the value of the
frequency width Δω of the resonance curve at −3 dB relative to the peak at the
resonance frequency.
ω0
Q= (F.17)
∆ω
Low and high values of Q correspond to damped and undamped resonances
respectively. The damping of the resonances can also be described by a reverberation
time. In room acoustics, reverberation time, written RT or T60, is used to characterize
the behavior of the decay of sound energy or pressure in a room. The reverberation
time T60 is defined as the time it takes for the energy stored in the resonant system to
drop to 10 −6 of its start value (a drop of 60 dB) and is related to the decay constant,
the loss factor η, and the Q factor as
3 ln(10) 6 ln(10) 6 ln(10) Q

T60 = = = (F.18)
δ ωη ω
Rooms for music and voice that have subjectively pleasant reverberation times are
characterized by modes having reverberation times in the range of 0.4–2 s around
500 Hz. The choice of reverberation time depends on the music to be played.
It is reasonable to assume that any loudspeaker resonance should have a decay
that is considerably shorter than that of the room in which it is being used; that
is, the loudspeaker damping should always be considerably higher than that of
the room.
Designing the room or box not to have parallel surfaces does not eliminate the
modes; it only makes it more difficult to calculate their frequencies fq and associated
spatial functions Ψq.
Most loudspeaker box interiors as well as most rooms may be regarded as

irregularly shaped rooms that may or may not have even distribution of sound
absorption over their interiors. In contrast to rooms, most closed loudspeaker boxes
for high-end audio are filled with an open-pore porous sound-absorptive material
such as a synthetic or natural wool, a glass fiber matt, or an open-pore plastic foam.
For a closed-box loudspeaker, it is better for the sound-absorptive material be evenly
distributed inside the loudspeaker box than only covering the walls, leaving most
of the box interior open. The reason is the material effectively lowers the speed of
sound, and the box therefore has a higher compliance for a given volume than if it
was empty, as was discussed in Appendix E.
F.4 ROOM FREQUENCY RESPONSE

It is useful to study the room response or transfer function for frequencies below and
in the modal region.
At very low frequencies below the modal region, when the wavelength of
the sound λ ≫ lx, ly, and lz, the room acts as an acoustic impedance showing a
compliance characteristic as given by Equation 7.19. The response will depend on
the loudspeaker’s resonance frequency. Below the first mode of the box or room,
the room response is characterized by a drop of −12 dB per octave. This is due to the
room’s acoustic impedance being that of a compliance (−6 dB per octave) and to the
driver’s mechanical system being mass controlled (−6 dB per octave). In real rooms,
the frequency response rise toward low frequencies will be limited by the leakage of
the box or room (that results in a parallel resistance/compliance loading the source
rather than a pure compliance) and will flatten out.
For small sound source, having a volume velocity U, placed at location, r0 in a
room, the expression for the total sound pressure at another point, r, due to all the
room modes, is given by an infinite sum of the pressure contributions of the various
modes according to
Ψqx ,qy ,qz (r )Ψqx ,qy ,qz (r0 )

∞ ∞ ∞
jωρc 2
( )
ptot r r0 =
V
U ∑∑∑ Λ q x , q y , qz ( )
 ω 2qx ,qy ,qz − ω 2 + 2 jω qx ,qy ,qz δ qx ,qy ,qz 

qx qy qz  
(F.19)
The constant Λ for a mode is determined if the standing wave is created by waves
moving in one (Λ = 1/2), two (Λ = 1/4), or three (Λ = 1/8) dimensions. Equation F.19 is
only valid for small values of δ, the decay constant.
In real rooms, the loudspeaker sound source may be considered small compared
to the room dimensions and the wavelengths of sound for low-order modes. We note
that in Equation F.19, the source and observation points r and r0 can be exchanged
without change to the pressure. This property is an example of reciprocity and can be
used to great advantage in finding those loudspeaker positions that give the “flattest”
frequency response between two positions in a room.
Relative frequency response [dB] –10
–20
–30
–40
–50
–60
–70 (a)
(b)
–80
2 5 10 20 50 100
Frequency [Hz]
FIGURE F.4 A frequency response curve in a room for an ideal closed-box loudspeaker that
has its resonance frequency (a) below 1 Hz and (b) at 20 Hz. The lowest room mode frequency
is at 24.5 Hz in this example.
How “strongly” any particular room mode will be excited depends on
• How close any of its resonance frequencies are to the excitation frequency,
• How damped the particular room mode is, and
• Where the loudspeaker box is located in the room.
Figure F.4 shows two examples of theoretical frequency responses for a closed-box
loudspeaker in a room as a function of frequency. The solid line shows the response
for a loudspeaker that is mass controlled to very low frequencies, i.e., has a very low
resonance frequency. The dashed line shows the response for a loudspeaker that has
a resonance frequency somewhat below the first room mode.
F.5 OVERLAPPING MODES AND THE SCHROEDER FREQUENCY

If the box or room is large, the mode frequencies will be close to another. Because
of the damping, many modes will then share excitation at one frequency, i.e., many
modes will be active, and since most modes have different propagation directions
within the room, the sound field will be characterized by the presence of waves
propagating in many directions. If the room does not have an extreme shape, and if
all modes have similar but not extremely high damping, the waves are about equally
strong and the resulting sound field is considered diffuse. The reason is that each
mode is associated with a certain wave vector and consequently the sound propaga-
tion direction in the room.
A lower frequency limit for the diffuse sound field in the statistical sense is the
Schroeder frequency, f S. The Schroeder frequency is approximately given by the
equation
T60
fS ≈ 2000 (F.20)
V
Here T60 is the reverberation time and V the box or room volume. Because of the
sound absorption, the Q factor of the resonance curves at this frequency is less
than 3, which can be shown to correspond to at least three mode resonance frequen-
cies within the half-value-widths of the resonance curve of a single mode.
Figure F.5 shows that the sound pressure level bounds for a 68% confidence
interval (±σ) for various distances from a small omnidirectional sound source in a
reverberant room. Here the reverberation radius rh is the distance at which the direct
sound from the omnidirectional sound source has the same energy density as the
reverberant sound field in the room.
Above the Schroeder frequency the 95% confidence interval of the frequency
response of a room is confined to about a 20 dB wide interval.
20
16
Normalized sound pressure level [db]
12
4
p84%
0
p50%
Reverberant sound
–4
p16%
–8 Direct sound
.125 .25 .5 1 2 4 8
r/rh
FIGURE F.5 The sound pressure level from a small monopole-type sound source in a room
at frequencies above the Schroeder frequency for various distances to the source relative to
the reverberation radius rh. The median and ±σ percentiles for the SPL probability distribution
are shown.
F.6 WAVES IN DUCTS

F.6.1 Rectangular Ducts
Assume a long rectangular duct that has constant cross section and rigid walls and is
much larger in the z-direction than in the crosswise x- and y-directions as shown in
Figure F.6. It is reminiscent of the rectangular box studied previously, the difference
being that we consider the box infinitely long in the z-direction.
Assume the waves being generated by a vibrating piston covering the entire cross
section at z = 0. For zero particle velocity perpendicular to the wall at x = dx and
y = dy, the following conditions must be met:
sin(k x d x ) = 0
sin(k y d y ) = 0 (F.21)

so the wave number components k x and k y will be
qx π
kx =
dx
qyπ
ky = (F.22)
dy

Here qx and qy are the natural numbers 0, 1, 2, etc. The solutions to the wave equation
will then be of the form shown in Equation F.23.
 q πx   q πy 
Ψ( x, y, z, k ) = cos  x  cos  y  e − jkz (F.23)
 dx   dy 

y
dx
dy
FIGURE F.6 The rectangular duct under study.

Substitution of this equation into the wave equation shows that the propagation
constant k x for the x-direction is
2 2 2
 2πf   q x π   q y π 
kz =  − − (F.24)
 c   d x   d y 

For wave propagation in the positive z-direction, the wave number kz must be positive
and real. This can occur only for frequencies f where
2 2
c  qx   qy 
f > + (F.25)
2  d x   d y 

Waves that have qy and qz larger than zero will effectively travel slower in the
z-direction than a free wave since they are waves that bounce between the duct
walls as they propagate. Such waves are called oblique waves and their mode of
propagation is called a higher-order mode.
If, for a certain combination of qy and qz, the frequency f is low and so k is
small, the root in Equation F.24 will become imaginary and waves of that qy and qz
combination cannot propagate.
The lowest frequency oblique wave that can propagate in the duct will be given by
q x = q y = 0 (F.26)

so for oblique modes to exist, the wavelength must be sufficiently small and
λ
≤ max(d x , d y ) (F.27)
2
The frequency at which this occurs is called the oblique mode cutoff frequency.
Below this cutoff frequency, only a plane wave can propagate in the tube. The wave
is often referred to as an axial or zeroth-order mode (or 0,0 mode). Note that unless
the driver has its diaphragm perfectly flat against the z = 0 plane and fills the rect-
angle given by dx and dy, close to the driver, the sound field will carry a near-field
component corresponding to the higher-order modes even at frequencies below the
oblique mode cutoff frequency. These die out exponentially with distance away from
the source. At distances far from the driver, the wave-front will be that of a plane
wave.
The sum of sound pressure of waves propagating in the duct can is described by
∞ ∞
 q πx   q πy 
p( x, y, z, k ) = ∑∑ A
qx = 0 qy = 0
qx ,q y cos  x  cos  y  e − jkz z (F.28)
 dx   dy 

Here kz is given by Equation F.24. Because the oblique waves propagate in both the
positive and negative k x and k y directions, nodal lines will be found in the x- and
y- directions as in a room or box. The wave field inside the rectangular duct will
be similar to that of a rectangular room with the difference being that there are no
modes in the z-direction.
In practical ducts, the wave field will be distorted in amplitude and phase close
to the duct walls because of the sound-absorptive material and leakage of sound
power out through the duct walls because of their vibration or at wall joints. In many
practical applications, these distortions can be neglected unless the duct is many
wavelengths long. In microphone probe tubes the small diameter of the tube leads
to viscous losses that distort the plane wave. It is convenient to take the losses into
account by introducing a complex wave number kz.
In many audio frequency applications, such as in low-frequency transmission line
loudspeakers and in microphone probes, the frequency of sound will usually be well
below the cutoff frequency given by Equation F.25. In high frequency and ultrasonic
applications, however, the higher-order modes may exist and become a nuisance. In
using the tube method for impedance and absorption measurement, modes other than
the plane wave mode must be avoided. This can be done by keeping the measurement
frequency sufficiently low or having a sufficiently small diameter tube and by using spe-
cial microphone probes or arrays that null the contributions from higher-order modes.
Any type of duct irregularity, such as constrictions, expansions, folds, bends, etc.,
will cause reflected waves and excitation of higher-order modes. If the frequency
is below the cutoff frequency, such modes will be attenuate quickly attenuated.
Folds and bends are discussed in a later section. If the tube is cut at an angle, there
will be reflection of sound at the cut end that may cause higher-order modes as
will an asymmetrically mounted driver. The higher-order mode waves are often
characterized by comparatively small sound radiation from the end opening because
of cancellation effects.
We usually want to avoid any higher-order modes. This can be done by mak-
ing the tube sufficiently narrow so that these modes cannot exist. If the tube plane
wave impedance becomes too large, one can use multiple tubes in parallel, similar
to a honeycomb structure. Another alternative is to line the tube walls with sound-
absorptive material. Such materials typically absorb sound better at the higher fre-
quencies, which is advantageous for many applications. If there is no air flow in the
duct, filling the duct with sound-absorptive material is also an alternative.
F.6.2 Cylindrical Ducts

Wave propagation in cylindrical ducts is similar to that in rectangular ducts.
Attenuation and higher-order modes in cylindrical ducts follows the same behavior
as that outlined for rectangular ones.
The spatial function for waves in the positive z-direction can be shown to be of
the form [3]
Ψ(r , z, ϕ, k ) = ( Ap, υ cos( pϕ) + Bp, υ sin( pϕ)) J p ( υr )e − jkz (F.29)

p=2
n=0
–0.5
–1.0
1.0
0.5
p=1
n=0
–0.5
–1.0
1.0
0.5
p=0
n=1
–0.5
–1.0
1.0
0.5
FIGURE F.7 The isobars for a three higher-order modes in a circular tube. Note that
the isobars will show lengthwise variations despite there not being reflected waves in the
z-direction.
Note that the waves in the radial direction have amplitudes determined by Bessel
functions of order p and those in the circular direction by circular functions. The
wave number in the r direction is υ but depends on the number of circular nodes,
i.e., wave propagation radially. The propagation modes are determined by the mode
numbers n and p. The pressure distribution functions for a few such modes are shown
in Figure F.7.
Since the boundary condition is the particle velocity radially against the tube wall
at r = a must be zero, it can be formulated as
∂
J p ( υ p , nr ) =0
∂r r=a
(F.30)
π
ϕp =
p

Values for the nulls of derivatives of Bessel functions can be found in tables such as
Ref. [F.4]. The wave number in the z-direction is
2
 2πf 
kz =  − υ2p,n (F.31)
 c 

Similar to what we obtained for the case of rectangular cross section ducts, one finds,
for a rigid circular tube of radius a, that the cutoff frequency fcutoff below which only
plane waves can propagate in the tube is given by
c
fcutoff = υ p,n (F.32)
2πa
The cutoff occurs when υp,n has values determined by the numbers p and n in a
similar way to qx and qy. The following table lists some of these υp,n values:
n
0 1 2
0 0 3.83 7.01
p 1 1.84 5.33 8.54
2 3.05 6.70 9.96
The lowest cutoff frequency is obtained for radial modes having a υ0,1 value equal to
1.84. For a duct having a diameter d, the cutoff frequency is then
c 1.84 c
fcutoff ≈ ≈ 0.59 (F.33)
π d d
This means that higher-order modes must have wavelength shorter than
λ d
< (F.34)
2 1.172
to propagate. The cutoff frequency of a circular tube is slightly higher than for a
rectangular tube having the same width as the tube diameter.
When we discussed sound propagation in horns in Chapter 19, we were only
looking at propagation of the zeroth-order mode. Because of the unavoidable horn
and driver asymmetries, higher-order modes will always exist in a horn. These
modes can be avoided by splitting the horn, for example as done in multicell
horns. The higher-order modes may cause considerable distortion of the directivity
characteristics of the horn when calculated using only the plane wave mode.
REFERENCES
1. Morse, P. M., Vibration and Sound, 2nd edn., American Institute of Physics, New York
(1991) ISBN-13: 978-0883188767.
2. Skudrzyk, E. J., Foundations of Acoustics, Basic Mathematics & Basic Acoustics,
Springer, New York (1972) ISBN-13: 978-0387809885.
3. Rschevkin, S. N., A Course of Lectures on the Theory of Sound, Pergamon, Oxford,
U.K. (1963) ASIN: B0006AY30W.
4. Abramowitz, M. & Stegun, I. A., Handbook of Mathematical Functions, Dover,
New York (1965) ISBN-13: 978-0486612720.
Appendix G: Level Definitions
G.1 ACOUSTICS
In acoustics, “a constant times the logarithm” (of base 10) for the ratio of two RMS
values of a sound field property is used to describe their difference and called a level
difference. Levels are denoted by L and are expressed in units of “decibel” [dB].
Levels are defined by
 a 
L = 20 log   (G.1)
 aref 

where ãref is the reference value.

It is best to use the internationally standardized reference values, when applicable,
in calculating levels. The sound pressure level reference value has been chosen
so that the number 0 dB corresponds roughly to the human hearing threshold at
1 kHz. The sound pressure level Lp of a sound having the root mean square pres-
sure p̃ expressed in pascal [Pa] is defined as
 p 
L p = 20 log  (G.2)
 2 ⋅ 10 −5 

The sound intensity level L I of a sound field having the intensity I expressed in
watt/m2 [W/m2] is defined as
 I 
LI = 20 log  −12  (G.3)
 10 

The sound power level LW of a sound source radiating the power P expressed in watt
[W] is defined as
 P 
LW = 20 log  −12  (G.4)
 10 

G.2 ELECTRICAL ENGINEERING

Level definitions in electrical engineering follows the same basic pattern as in
acoustical engineering. Some common-level definitions are dBV, where the reference
level is 1 volt, and dBm, where the reference level is 1 mW.
583
584 Appendix G: Level Definitions
The voltage level L [dBV] of a voltage ẽ is defined relative to an RMS voltage

of 1 V as
 e 
L = 20 log   (G.5)
 1

When the power level L [dBm] is used with the reference being 1 mW
 P 
L = 20 log  −3  (G.6)
 10 

However, since the 1 mW reference was into a 600 Ω load it corresponds to a voltage
of 0.775 V. Care is advised when using dBm since sometimes voltage level is referred
to 0.775 V. Best is to avoid using dBm.
G.3 MICROPHONE SENSITIVITY

Since the microphone is an electroacoustic device, the sensitivity of interest is
typically the microphone’s RMS (open load) output voltage ẽ for an RMS sine sound
wave at 1 kHz at a given sound pressure. The standard SPL reference is 94 dB, which
corresponds to an RMS sound pressure of 1 Pa. The sensitivity is usually expressed
in dB relative to 1V/Pa as
 e 
Sensitivity = 20 log   (G.7)
 1

or in dB relative to 1 mV/Pa, that is,
 e 
Sensitivity = 20 log  −3  (G.8)
 10 

G.4 LOUDSPEAKER SENSITIVITY

Loudspeaker sensitivity is the sound pressure level on axis at 1 m distance from the
loudspeaker for a given input voltage or power at some frequency. The reference
voltage typically is 2.83 V, which corresponds to an electric power of 1 W into a
resistive load of 8 Ω. It is advisable to check if the specification is to an applied
electrical power of 1 W or to the voltage of 2.83 V.
Since most power amplifiers give a specified voltage irrespective of their load
within their working range, the voltage level has become the de facto reference, not
the power. Sometimes the loudspeaker may be very large, and it is assumed that the
measurement is done in the far field of the loudspeaker. If the far-field distance is
Appendix G: Level Definitions 585
larger than 1 m, then the measurement is done at the appropriate distance and then
sometimes calculated back to an equivalent distance of 1 m or to a distance of 10 m.
As an example, the sensitivity is expressed as ** dB SPL for 1 W at 1 m distance
where the appropriate value is substituted for **.
Loudspeaker efficiency, i.e., acoustic power for a certain electrical input power,
may be an important property in some applications. One common way of specifying
the electroacoustic efficiency η is to use the ratio between acoustic output power and
electric input power
Wacoustic
η= (G.9)
Welectric
Most direct radiating electrodynamic loudspeakers have efficiencies in the range

of 0.1%–5%. Using horns or other impedance transformers, one can substantially
increase the efficiency.
Electrical Engineering
ElEctroacoustics
Mendel Kleiner
“Finally a modern alternative to the classical books on electroacoustics. The power-
ful technique of electrical analogous circuits is used for the analysis of microphones,
loudspeakers and headphones. Theory as well as practical aspects are covered
in this book by an author with decades of experience in teaching, research and prac-
tical applications. The inclusion of exercises makes this a great textbook for university
teachers. Electroacoustics is a great topic for attracting students to acoustics, and this
modern textbook will serve well in that role.”
—Peter Svensson, Norwegian University of Science and Technology (NTNU)
“... includes comprehensive coverage pertinent to fundamentals of electroacoustics

... Many basic theoretical concepts are explained with the help of a minimal amount
of math, accessible to many students or audio engineers.”
—Ning Xiang, Rensselaer Polytechnic Institute, New York, USA
Electroacoustic devices such as microphones and loudspeakers are used every-

where from cars and mobile phones to homes, places of worship, and sports
arenas. They are a key part of the modern communication society, helping to
transmit information to our ears. A contemporary introduction to the subject,
Electroacoustics explains the scientific and engineering principles behind the
design of these sound transducers. It also examines the compromises that are nec-
essary when designing transducers for use in the real world.
This accessible text uses the concept of electrical circuit analogies to help readers
quickly grasp the fundamentals of acoustical and mechanical systems. It covers
both traditional electrodynamic audio and ultrasonic transducers and includes
up-to-date material on arrays, planar transducers, and loudspeaker enclosure
design. Throughout, the book emphasizes multidisciplinary engineering princi-
ples to prepare students for the broad range of applications they may encounter.
The modern treatment of transducers also makes this a valuable reference for
transducer designers, acoustical consultants, hobbyists, and anyone involved in
electroacoustic design.
K11635
6000 Broken Sound Parkway, NW

Suite 300, Boca Raton, FL 33487
711 Third Avenue
an informa business New York, NY 10017
2 Park Square, Milton Park
www.taylorandfrancisgroup.com Abingdon, Oxon OX14 4RN, UK

Kleiner, Mendel-Electroacoustics-CRC Press (2013)

Uploaded by

Copyright:

Available Formats

Kleiner, Mendel-Electroacoustics-CRC Press (2013)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kleiner, Mendel-Electroacoustics-CRC Press (2013)

Uploaded by

Copyright:

Available Formats

ElEctroacoustics

Boca Raton London New York

CRC Press is an imprint of the

© 2013 by Taylor & Francis Group, LLC

No claim to original U.S. Government works

International Standard Book Number-13: 978-1-4665-9943-7 (eBook - PDF)

and the CRC Press Web site at

Chapter 2 Introduction to Electroacoustic Systems...............................................5

Chapter 3 Sound and Its Properties..................................................................... 11

3.9.3 Normal Reaction.........................................................28

Chapter 4 Waves in Membranes and Plates......................................................... 39

Chapter 5 Circuits and Circuit Components........................................................ 65

Chapter 6 Electromechanical Analogies.............................................................. 69

6.3 Mechanical Impedance and Mobility....................................... 70

Chapter 7 Electroacoustical Analogies................................................................ 83

Chapter 8 Conversion between Analogies......................................................... 121

8.3 “Dot” Method......................................................................... 122

Chapter 9 Transducer Operating Principles....................................................... 129

Chapter 10 Radiation and Impedance.................................................................. 163

Chapter 11 Sound Source and Acoustic Environment......................................... 199

Review Questions.............................................................................. 211

Chapter 12 Directivity......................................................................................... 215

Chapter 13 Microphones and Sound Fields......................................................... 253

13.5 Two Ways to Achieve Directivity...........................................260

Chapter 14 Microphones...................................................................................... 269

Chapter 15 Electrodynamic Loudspeaker Drivers.............................................. 299

15.3 Magnet Air Gap and Voice Coil............................................. 301

Chapter 16 Baffle and Box................................................................................... 319

Chapter 17 Vented Box Loudspeakers................................................................. 337

Chapter 18 Transmission Line Loudspeakers...................................................... 351

Chapter 19 Horns................................................................................................. 367

19.17 Linear and Nonlinear Distortion............................................ 391

Chapter 20 Gradient Loudspeakers..................................................................... 397

Chapter 21 Drivers Using Flexible Diaphragms..................................................407

Chapter 22 Multiway Loudspeakers.................................................................... 423

22.2 Diaphragm Dimensions and Wavelength............................... 424

Chapter 23 Active Loudspeakers......................................................................... 439

Chapter 24 Headphones and Earphones.............................................................. 451

24.4 Acoustic Environment............................................................ 455

Chapter 25 High-Frequency Transducers............................................................ 467

Appendix A: Introduction to Electric Components

Appendix C: Magnetic Fields and Forces........................................................... 541

a radius, acceleration [m/s2]

f function, frequency [Hz]

MARv vent acoustic radiation mass [kg/m4]

R AW box wall acoustic resistance [Ns/m5]

YMR mechanical radiation admittance [m/Ns]

μμ magnet permeability [H/m]

Electroacoustics is formed by the combination of knowledge from three areas, namely,

Ultrasonic signal Sonar

Hearing Noise and

Psychoacoustics Room acoustics

Speech Music Art

1.2 OTHER BOOKS AND REFERENCES

Ripple Flat Rise Roll-off

Marker for level and frequency

0 f1 2f1 3f 1 4f1 and so on Frequency

Fundamental Harmonic distortion

Low-frequency High-frequency Harmonic to f2

Difference tone distortion

Z S (r , k ) ≈ jρckr = jωρr; kr 1 (3.24)