Rfcircuitsdesigntheoryandapplicationsludwig 140506103907 Phpapp02

RF Circuit Design
Theory and Applications
Reinhold Ludwig
Worcester Polytechnic Institute
Pavel Bretchko
Worcester Polytechnic Institute
Prentice Hall
Upper Saddle River, NJ 07458
Manufacturing coordinator:
Editorial assistant: Jennie uuLn.ma
All
means,
book may
reproduced. in any fonn
writing from the publisher.
10 9 8 7 6 5 4 3 2 1
ISBN D-13-095323-7
PRENTICE-HALL
PRENTICE-HALL OF JAPAN, INc., Tokyo

PRENTICE-HALL (SINGAPORE) PTE.
PRENTICE-HALL DO BRASIL,
DEDICATION
To our families and the memory of my father F. Ludwig
design is receivsignificant industrial attention

(RF) and
(MW) applications.
semiconductor devices have made
a proliferation
and analog
as observed in
communication,
global
RADAR,
electrical and computer engineering disci. . . ."........,. This interest
translated into a strong
for engineers
comprehensive knowledge high-frequency
design
the student,
engineer, and
faculty member teaching
material
a general
The majority
existing textbooks
appear to
two separate
advanced
with a broad theoretical
and
technologists with
interest in
mathematical and
a result, RF circuit design has been presented
two
different formats. For
advanced students the entry into this field is
through an electromagnetic field approach, while for the
basic circuit
embedded
laws
the
treatment.
approaches make difficult
address
theoretical and practical
surrounding
design principles. The
circuit approach lacks, or only
superficially covers,
wave nature of currents and voltages whose reflection
uu.:~.u.u.:.., ......., ..... properties constitute indispensable
of the RF
behavior.
approach
covers the wave
and transmission
falls far short reaching the important aspects of designing
line aspect,
amplifier, oscillator, and mixer circuits.
The objective of
textbook to
the
circuit design aspects in such a
way that the need for transmission line principles
clear without adopting an
electromagnetic
approach. Therefore, no EM background necessary beyond a
first year
physics course fields and waves as provided by most -~~--~~
and universities. Students equipped with the
basic
and/or an
.u.u.""'........'"'e<~trcmtc~s can use
and cover the
spectrum from
basic principles of transmission
microstrip lines to the various high-frequency
to the
design procedures. Lengthy mathematical derivations are
appendices or placed in
the main
and thus
on the main concepts.
sian of some the dry
Accepting the
of providing a high degree of design
we have
included many examples that discuss in considerable
in many cases .........
over
pages,
philosophy and the intricacies
approaches.
ji:..U...................""-
'11.., .... "'......
VI
Preface
This has caused some problems as well, specifically with respect to the circuit simulations. Obviously, we cannot expect the reader to have ready access to modern computer
simulation tools such as MMICAD or ADS to name but two of the popular choices.
Professional high-frequency simulation packages are generally expensive and require
familiarity to use them effectively. For this reason we have created a considerable number of MATLAB M-files that the interested student can download from our website listed
in Appendix G. Since MATLAB is a widely used relatively inexpensive mathematical
tool, many examples discussed in this book can be executed and the results graphically
displayed in a matter of seconds. Specifically the various Smith-Chart computations of
the impedance transfonnations should appeal to the reader. Nonetheless, all design
examples, specifically the ones presented in Chapters 8 to 10, have been independently
simulated and verified in MMICAD for the linear circuit models, and ADS for the nonlinear oscillator and mixer models.
In terms of material coverage, this textbook purposely omitted the high-speed digital circuits as well as coding and modulation aspects. Although important, these topics
would simply have required too many additional pages and would have moved the book
too far away from its original intent of providing a fundamental, one- or two-semester,
introduction to RF circuit design. At WPI this does not tum out to be a disadvantage,
since most of the material can readily be acquired in available communication systems
'
. courses.
engmeermg
The organization of this text as follows: Chapter 1 presents a general explanation of why basic circuit theory breaks down as the operating frequency is increased to
a level where the wavelength becomes comparable with the discrete circuit components. In Chapter 2 the transmission line theory is developed as a way to replace the
low-frequency circuit models. Because of the voltage and current wave nature, Chapter 3 introduces the Smith Chart as a generic tool to deal with the impedance behavior
on the basis of the reflection coefficient. Chapter 4 discusses two-port networks with
their flow-chart representations and how they can be described on the basis of the socalled scattering parameters. These network models and their scattering parameter
descriptions are utilized in Chapter 5 to develop passive RF filter configurations.
Before covering active devices, Chapter 6 provides a review of some of the key semiconductor fundamentals, followed by their circuit models representation in Chapter 7.
The impedance matching and biasing of bipolar and field effect transistors is taken up
in Chapter 8 in an effort to eliminate potentially dangerous reflections and to provide
optimal power flow. Chapter 9 focuses on a number of key high-frequency amplifier
configurations and their design intricacies ranging from low noise to high power applications. Finally, Chapter 10 introduces the reader to nonlinear systems and their
designs in oscillator and mixer circuits.
VII
Preface
This book is used in

Electrical and Computer
Department at
as required text
(5
hours per week)
in RF circuit
311 Introduction
Circuit Design). The course has primarily
and 4th year undergraduate students with a
in
microelectronics.
course does
include a laboratory,
six videotapes
Philips Semiconductors and in-class RF
practical
performances
cuit measurements
a network analyzer are included. In addition, MMICAD and
.... u ....... u u......... are incorporated as part of
regular lectures.
chapter is fairly
~"'""'""...... with the goal
providing wide
course
materiaL At WPI
of
one three
pressed
a ?week period
a total of
lectures).
are shown the table below.
J
13, Introduction to
Design
Introduction
4. Single- and Multi-Port Networks
Matching
Chapter 9. RF Transistor Amplifier Designs

remaining
is targeted
a
(7 week) term
more
advanced
as microwave filters, equivalent circuit models, oscillators and
organizational plan provided below.
u ...........,&
of RF
Advanced
Overview
Design
1-5.5
Chapter 6, Active RF Components

Chapter
Transistor Amplifier Designs
apter 10, Oscillators and Mixers
I Sec1tions 9.5-9.8
Sections 10.1-10.4
viii
the entire course u:rgamzat:lun

depending on total ""'~.... . ,........
with
courses.
change
requirements
Hfl'lli.IP,,,.,..,.
to the .............
website
BJ .....
http://www.prenhaU.com/ludwig for
files in
format.
ACKNOWLEDGEMENTS
The authors are grateful to a number
Orr, head of the ECE
he provided the funding

Vennema, Jarek Lucek,
"'""'""u................ expertise, sponsoring
students, and pntcti.CU:lLR

was instrumental in ............r.n..""'
simulation packages.
Philips Semiconductors
student projects, and making available
Profs. John Sullivan,
William Michalson, and
Makarov
.,,.,.,. ..... JLJL.., ... A
help. Linda Gu,
Lai, Joe Plunkett, Dr. Funan
and Josh Resnik

and
graduate stu... a ....... .,.1 ambience and
the EM
lab at WPI.
dents
R. L.
Prof. J. Thomas
of Minnesota's
who
to the
principles
RF coils for
in the .........,~.........
would like to
magnetic
..............,. thanks
discusShirokov of ...,._..,..."""~4
specifisions on
aspects of RF/MW
and
and
Frank, Tom
cally
making this book a
package by Optotek
a
of the MIMICAD
Corporation
gratefully
. . . . . . . . . . . . . . . . &.&
U.A . . . ., _ . . . .
' V U ....... ..., . . .
ontents
v
1
Preface
Introduction
l.l
1
1
1
Importance of Radio frequency Design

and Units
Frequency
Behavior of Passive Components
1 l High-Frequency Resistors
l
1
and Circuit
Considerations
Chip Resistors
Chip
Surface-Mounted
Summary
2
6
8
10
14
21
24
25
26
28
Chapter
Theory?
1 Why Transmission
Transmission
2.2
41
41
42
Microstrip Lines
Equivalent Circuit Representation
Foundation
45
47
47
2.6
Parameters
a ParaHeJ Plate Transmission
Summary Different Line Configurations
General Transmission Line .W"4'"auva
2. 7.1
and Current Law ~".... L'""""'"""'
Traveling Voltage and
Waves
2.7.3 General Impedance Definition
2.7.4 "-'"'"'..,.,.,,... "' Transmission
53
57
58
63
64
2.9
2.10
Transmission Line
Voltage Reflection Coefficient
Propagation Constant and
Standing
Special
Conditions
1 Input Impedance Terminated Lossless Line
ix
69
69
71
75
Contents
2.10.2 Short Circuit Transmission Line

2.10.3 Open-Circuit Transmission Line
2.10.4 Quarter-Wave Transmission Line
2.11 Sourced and Loaded Transmission Line
2.11.1 Pbasor Representation of Source
2.11.2 Power Considerations for a Transmission Line
2.11.3 Input Impedance Matching
2.11.4 Return Loss and Insertion Loss
2.12 Summary
Chapter 3. The Smith Chart
3.1 From Reflection Coefficient to Load Impedance
3.1.1 Reflection Coefficient in Phasor Form
3.1.2 Normalized Impedance Equation
3~ 1.3
Parametric Reflection Coefficient Equation
3.1.4 Graphical Representation
3.2 Impedance Transformation
3.2.1 Impedance Transformation for General Load
3.2.2 Standing Wave Ratio
3.2.3 Special Transformation Conditions
3.2.4 Computer Simulations
3.3 Admittance Transformation
3.3.1 Parametric Admittance Equation
3.3.2 Additional Graphical Displays
3.4 Parallel and Series Connections
3.4.1 Parallel Connection of R and L Elements
3.4.2 Parallel Connection of R and C Elements
3.4.3 Series Connection of R and L Elements
3.4.4 Series Connection of R and C Elements
3.4.5 Example of aT-Network
3.5 Summary
Chapter 4. Single- and Multiport Networks
4.1 Basic Definitions
4.2 Interconnecting Networks
4.2.1 Series Connection of Networks
4.2.2 Parallel Connection of Networks
4.2.3 Cascading Networks
4.2.4 Summary of ABCD Network Representations
4.3 Network Properties and Applications
4.3.1 Interrelations between Parameter Sets
4.3.2 Analysis of Microwave Amplifier
4.4 Scattering Parameters
4.4.1 Definition of Scattering Parameters
76
79
81
84
85
87
90
91
93
101
102
102
104
106
108
110
110
113
115
119
122
122
125
126
127
128
128
129
130
133
143
144
153
153
154
155
156
161
161
164
168
168
Contents
4.4.2 Meaning of S-Parameters

4.4.3 Chain Scattering Matrix
4.4.4 Conversion between Z- and S-Parameters
4.4.5 Signal Flow Chart Modeling
4.4.6 Generalization of S-Parameters
4.4.7 Practical Measurements of S-Parameters
4.5 Summary
Chapter 5. An Overview of RF Filter Design

5.1
Basic Resonator and Filter Configurations

5. 1.1 Filter T)rpes and Parameters
5. 1.2 Low-Pass Filter
5.1.3 High-Pass Filter
5.1.4 Bandpass and Bandstop Filters
5 .1.5 Insertion Loss
5.2 Special Filter Realizations
5.2.1 Butterworth-Type Filters
5.2.2 Chebyshev-Type Filters
5.2.3 Denormalization of Standard Low-Pass Design
5.3 Filter Implementation
5.3.1 Unit Elements
5.3.2 Kuroda's Identities
5.3.3 Examples of Microstrip Filter Design
5.4 Coupled Filter
5.4.1 Odd and Even Mode Excitation
5.4.2 Bandpass Filter Section
5.4.3 Cascading bandpass filter elements
5 .4.4 Design Example
5.5 Summary
Chapter 6. Active RF Components

6.1
Semiconductor Basics
6.1.1 Physical Properties of Semiconductors
6. 1.2 PN-Junction
6.1.3 Schottky Contact
6.2 RF Diodes
6.2.1 Schottky Diode
6.2.2 PIN Diode
6.2.3 Varactor Diode
6.2.4 IMPATI Diode
6.2.5 Tunnel Diode
6.2.6 TRAPATI, BARRITT, and Gunn Diodes
6.3 Bipolar-Junction Transistor
6.3.1 Construction
xl
171
175
177
178
184
188
194
201
202
202
206
209
210
217
220
221
224
231
241
243
243
245
253
254
257
258
260
263
271
272
272
279
289
293
293
296
302
305
307
311
312
312
XII
Contents
6.3.2
6.3.3
Functionality
Frequency Response
6.3.4 Temperature Behavior
6.3.5 Limiting Values
6.4 RF Field Effect Transistors
6.4.1 Construction
6.4.2 Functionality
6.4.3 Frequency Response
6.4.4 Limiting Values
6.5 High Electron Mobility Transistors
6.5.1 Construction
6.5.2 Functionality
6.5.3 Frequency Response
6.6 Summary
Chapter 7. Active RF Component Modeling

7 .I
Diode Models
7.I. l Nonlinear Diode Model
7 .1.2 Linear Diode Model
7.2 Transistor Models
7 .2.1 Large-Signal BJT Models
7 .2.2 Sma11-Signa1 BJT Models
7 .2.3 Large-Signal FET Models
7.2.4 Small-Signal PET Models
7.3 Measurement of Active Devices
7. 3. 1 DC Characterization of Bipolar Transistor
7.3.2 Measurements of AC Parameters of Bipolar Transistors
7.3.3 Measurements of Field Effect Transistor Parameters
7.4 Scattering Parameter Device Characterization
7.5 Summary
Chapter 8. Matching and Biasing Networks

8.1
Impedance Matching Using Discrete Components

8.1 .1 Two-Component Matching Networks
8.1.2 Forbidden Regions, Frequency Response, and Quality Factor
8.1.3 T and Pi Matching Networks
8.2 Microstrip Line Matching Networks
8.2.1 From Discrete Components to Microstrip Lines
8.2.2 Single-Stub Matching Networks
8.2.3 Double-Stub Matching Networks
8.3 Amplifier Classes of Operation and Biasing Networks
8.3.1 Classes of Operation and Efficiency of Amplifiers
8.3.2 Bipolar Transistor Biasing Networks
8.3.3 Field Effect Transistor Biasing Networks
314
321
323
327
328
329
331
337
337
338
339
339
343
343
351
352
352
354
357
357
366
378
382
385
385
387
392
393
397
405
406
406
415
426
431
431
435
440
444
444
449
455
Contents
8.4
XIII
Summary
Chapter 9. RF Transistor Amplifier Designs
9.1
9.2
Characteristics of Amplifiers
Amplifier Power Relations
9.2.1 RF Source
9.2.2 Transducer Power Gain
9.2.3 Additional Power Relations
9.3 Stability Considerations
9.3.1 Stability Circles
9.3.2 Unconditional Stability
9.3.3 Stabilization Methods
9.4 Constant Gain
9.4.1 Unilateral Design
9.4.2 Unilateral Figure of Merit
9.4.3 Bilateral Design
9.4.4 Operating and Available Power Gain Circles
9.5 Noise Figure Circles
9.6 Constant VSWR Circles
9.7 Broadband, High-Power, and Multistage Amplifiers
9.7.1 Broadband Amplifiers
9.7.2 High-Power Amplifiers
9.7.3 Multistage Amplifiers
9.8 Summary
Chapter 10. Oscillators and Mixers
10.1 Basic Oscillator Model

10.1.1 Negative Resistance Oscillator
10.1.2 Feedback Oscillator Design
10.1.3 Design Steps
10.1.4 Quartz Oscillators
10.2 High-Frequency Oscillator Configuration
10.2.1 Fixed-Frequency Oscillators
10.2.2 Dielectric Resonator Oscillators
10.2.3 YIG-Tuned Oscillator
10.2.4 Voltage-Controlled Oscillator
10.2.5 Gunn Element Oscillator
10.3 Basic Characteristics of Mixers
10.3.1 Basic Concepts
10.3.2 Frequency Domain Considerations
10.3.3 Single-Ended Mixer Design
10.3.4 Single-Balanced Mixer
10.3.5 Double-Balanced Mixer
10.4 Summary
456
463
464
465
465
466
468
470
470
473
480
483
483
490
492
495
502
506
511
511
522
526
529
539
540
541
543
546
550
552
556
563
569
570
573
574
575
578
580
588
590
590
xlv
Contents
Appendix A. Useful Physical Quantities and Units
597
Appendix B. Skin Equation for a Cylindrical Conductor
601
Appendix C. Complex Numbers

C. I Basic Definition
C.2 Magnitude Computations
C.3 Circle Equation
Appendix D. Matrix Conversions
603
603
603
604
605
Appendix E. Physical Parameters of Semiconductors
608
Appendix F. Long and Short Diode Models

F.l Long Diode
F.2 Short Diode
609
610
Appendix G. Couplers
G.1 Wilkinson Divider
G.2 Branch Line Coupler
G.3 Lange Coupler
612
612
616
619
Appendix H. Noise Analysis

H.l Basic Definitions
H.2 Noisy Two-Port Networks
H.3 Noise Figure for Two-Port Network
H.4 Noise Figure for Cascaded Multipart Network
620
Appendix I. Introduction to MATLAB

I.l Background
1.2 Brief Example of Stability Evaluation
1.3 Simulation Software on Compact Disk
!.3.1 Overview
1.3.2 Software Installation
1.3.3 File Organization
Index
610
620
623
625
629
631
631
633
635
635
635
636
637
CHAPTER
Introduction
I t is common knowledge that both analog and digital design engineers are continually developing and refining circuits for increasingly
higher operational frequencies. Analog circuits for wireless communication in the gigahertz (GHz) range and the ever-increasing clock speeds of computer circuits in highperformance mainframes, workstations, and, of course, personal computers exemplify
this trend. Global positioning systems require carrier frequencies in the range of
1227.60 and 1575.42 MHz. The low-noise amplifier in a personal communication system may operate at 1.9 GHz and fit on a circuit board smaller in size than a dime. Satellite broadcasting in the C band involves 4 GHz uplink and 6 GHz downlink systems. In
general, due to the rapid expansion of wireless communication, more compact amplifier, filter, oscillator, and mixer circuits are being designed and placed in service at frequencies generally above 1 GHz. There is little doubt that this trend will continue
unabated, resulting not only in engineering systems with unique capabilities, but also
special design challenges not encountered in conventional low-frequency systems.
This chapter reviews the evolution from low- to high-frequency circuit operations. It
motivates and provides the physical rationales that have prompted the need for new engineering approaches to design and optimize these circuits. The example of a cellular phone
circuit, components of which will be analyzed in more detail in later chapters, serves as a
vehicle to outline the goals and objectives of this textbook and its organization.
The chapter begins with a brief historical discussion explaining the tr~nsition from
direct current (DC) to high-frequency modes of operation. As the frequen"cy increases
and the associated wavelengths of the electromagnetic waves becomes comparable to
the dimensions of the discrete circuit components such as resistors, capacitors, and
inductors, these components start to deviate in their electric responses from the ideal
frequency behavior. It is the purpose of this chapter to provide the reader with an appre-
Chapter 1 Introduction
ciation and understanding of high-frequency passive component characteristics. In particular, due to the availability of sophisticated measurement equipment, the design
engineer must know exactly why and how the high-frequency behavior of his or her circuit differs from the low-frequency realization. Without this knowledge it will be impossible to develop and understand the special requirements of high-performance systems.
1.1 Importance of Radiofrequency Design

The beginning of electrical circuit design is most likely traced back to the late
eighteenth and early nineteenth centuries when the first reliable batteries became available. Named after their inventor A. Volta (1745-1827), the Voltaic cells permitted the
supply of reliable DC energy to power the first crude circuits. However, it soon became
apparent that low-frequency alternating current (AC) power sources can transport electricity more efficiently and with less electric losses when transmitted over some distance and that rerouting the electric energy could be facilitated through transformers
that operate in accordance with Faraday's induction law. Due to pioneering work by
such eminent engineers as Charles Steinmetz, Thomas Edison, Werner Siemens, and
Nikolas Tesla, the power generation and distribution industry quickly gained entry into
our everyday life. It was James Maxwell ( 1831-1879) who, in a paper first read in 1864
to the Royal Society in London, postulated the coupling of the electric and magnetic
fields whose linkage through space gives rise to wave propagation. In 1887 Heinrich
Hertz experimentally proved the radiation and reception of electromagnetic energy
through air. This discovery heralded the rapidly expanding field of wireless communication, from radio and TV transmissions in the 1920s and 1930s to cellular phones and
Global Positioning Systems (GPS) in the 1980s and 1990s. Unfortunately, the design
and development of suitable high-frequency circuits for today's wireless communication applications is not so straightforward. As will be discussed in detail, conventional
Kirchhoff-type voltage and current law analysis tools, as presented to first- and secondyear undergraduate electrical engineering students, apply strictly only to DC and lowfrequency lumped parameter systems consisting of networks of resistors, capacitors,
and inductors. They fail when applied to circuits governed by electromagnetic wave
propagation.
The main purpose of this textbook is to provide the reader with theoretical and
practical aspects of analog circuit design when the frequency of operation extends into
the radio frequency (RF) and microwave (MW) domains. Here conventional circuit
analysis principles fail. The following questions arise:
At what upper frequency does conventional circuit analysis become inappropriate?
Importance of Radlofrequency Design
What characteristics make the high-frequency behavior of electric components so

different from their low-frequency behavior?
What "new" circuit theory has to be employed?
How is this theory applied to the practical design of high-frequency analog circuits?
This book intends to provide comprehensive answers to these questions by developing
not only the theoretical framework but also delivering the practical applications through
a host of examples and design projects.
To identify more clearly the issues that we will address, let us examine the generic
RF system shown in Figure 1-1.
Antenna
r- - ---- -,
r -
- -:- - - -
..
- - - - - - - -
,.......__..___, :
t ........ ------ - -- - -- --- - ------ ............
- l-t - - - - - - - - - - - - - - - - - - - - - - - - - - - - -.
.I:.
: :
Digital-to-Analog :
Converter
.:
OSC .
Transmitter
Power Amplifier
Switch
: Local Oscillator
LPF
L------" :
lAnalog-to-Digital!
Low-Pass
Filter
Receiver Power
Amplifier
.~ Converter
.~ .i
--------------------------------- ' .......... ...... ----------- ........................................... ---............ -........ --- -. -----------------
\....
--.......r
Mixed Signal
Circuits
Figure 11
./ - - - - - - - - - - - - - - -
Analog Signal Circuits
Block diagram of a generic RF system.
Typical applications of this configuration are cellular phones and wireless local
area networks (WLANs). The entire block diagram in Figure 1-1 can be called a
transceiver, since it incorporates both transmitter and receiver circuits and uses a single
antenna for communication. In thi"sconfiguration the input signal (either a voice or a
digital signal from a computer) is first digitally processed. If the input signal is a voice
signal, as is the case in cellular phones, it is first converted into digital form; then compressed to reduce the time of transmission; and finally appropriately coded to suppress
noise and communication errors.
After the input signal has been digitally preprocessed, it is converted back to analog form via a digital-to-analog converter (DAC). This low-frequency signal is mixed
with a high-frequency carrier signal provided by (}local oscillator. The combined signal
is subsequently amplified through a power amplifier (PA) and then routed to the
antenna, whose task is to radiate the encoded information as electromagnetic waves
into free space.
In the block diagram of Figure 1-1 let us focus on t he transmitter PA. This could be
a 2 GHz PA for cellular phones that may be implemented as a dual-stage amplifier.
Details of the circuit diagram for the first stage PA are shown in Figure l-2(a).
-=-Vn
8.2 pF-=RFC ;----
-=-Vc
8.2pF
RFC 1________
Stability
To the Second
Stage
Resistor
RFin o----tlr-+
DC Blocking -----)
Capacitor
\
.
.....................................................
.
......
Input Matching
Network
Figure 1-2(a)
Interstage
Blocking Capacitor
Interstage Matching
Network
Simplified circuit diagram of the first stage of a 2 GHz power

amplifier for a cellular phone.
We notice that the input signal is fed through a DC blocking capacitor into an
input matching network, needed to match the input impedance of the transistor (type
BFG425W of Philips Semiconductors), operated in common emitter configuration, to
the output impedance of the mixer that precedes the PA. The matching is needed to
ensure optimal power transfer as well as to eliminate performance degrading reflections. The interstage matching network must then match the output impedance of the
transistor to the input impedance of the second stage of the PA. Key components in the
matching networks are microstrip lines shown by the shaded rectangles in Figure
1-2(a). At high frequency these distributed elements exhibit unique electric properties
that differ significantly from low-frequency lumped circuit elements. We also notice
Importance of Radlofrequency Design
additional networks to bias the input and output ports of the transistor. The separation
of high-frequency signals from the DC bias conditions is achieved through two RF
blocking networks that feature so-called radio frequency coils (RFCs ).
The actual dual-stage circuit board implementation is given in Figure 1-2(b),
. which shows the microstrip lines as copper traces of specific lengths and widths.
Attached to the microstrip lines are chip capacitors, resistors, and inductors.
0.5 inch
___________ ..,
Interstage Matching
... .....;; Network
;
Second Stage
Transistor
Input Matching
Network, ...,_
....
-...
-...-.. ~
Output Matching
Network
DC Bias Network
Figure 12(b)
Printed circuit board layout of the power amplifier.
To understand, analyze, and ultimately build such a PA circuit requires knowledge

of a number of crucial RF topics discussed in this textbook:
Microstrip line impedance behavior is discussed in "Transmission Line Analysis"
(Chapter 2) and its quantitative evaluation is considered in Chapter 3, "The Smith
Chart."
The ability to reduce a complicated circuit into simpler constituents whose inputoutput is described through two-port network description. This is discussed in
Chapter 4, "Single- and Multiport Networks."
Strategies of generically developing particular impedance versus frequency
responses as encountered in filter design. Chapter 5, "A Brief Overview of RF Filter Design," outlines the basic discrete and distributed filter theories, and Chapter
8, "Matching Netwqfks," delves into a detailed circuit implementation as related
to Figure 1-2(b).
High-frequency bipolar junction and field effect transistors as ~ell as RF'diodes

are investigated in "Active RF Components" (Chapter 6) in terms of their physical
basis followed by "Active Circuit Device Models" (Chapter 7), where large signal
and small signal circuit models are analyzed.
The overall amplification requirements, as related to gain, linearity, noise, and stability, are basis of "RF Transistor Amplifier Design" (Chapter 9).
In addition to amplifiers, Chapter 10, "Oscillators and Mixers," focuses on additional important RF circuit design concepts, as shown in Figure 1-1.
A successful RF design engineer knows about and applies all these concepts in the
design, construction, and testing of a particular RF circuit project. As the preceding
example implies, our concern in this textbook is mostly geared toward analog RF circuit theory and applications. We purposely neglect mixed and digital RF signals since
their treatment would exceed the size and scope of this textbook.
1.2 Dimensions and Units

To understand the upper frequency limit, beyond which conventional circuit theory can no longer be applied to analyze an electric system, we should recall the representation of an electromagnetic wave. In free space, plane electromagnetic (EM) wave
propagation in the positive z-direction is typically written in sinusoidal form:
Ex = E 0 xcos(rot- ~z)
(1.1a)
HY = H 0Ycos(rot- ~z)
(1.1b)
where Ex and HY are the x-directed electric and the y-directed magnetic field vector
components, as shown qualitatively in Figure 1-3. Here E0 x and H 0 Y represent constant amplitude factors in units of VIm and Aim.
These waves possess an angular frequency ro , and a propagation constant ~ that
defines the spatial extent in terms of the wavelength A, such that ~
2nl A. Classical
field theory based on Maxwell's equations reveals that the ratio between electric and
magnetic field components is defined in terms of the so-called intrinsic impedance Z 0
!X = Zo = JIUE = J<JloJl,)/(eoe,) = 377 n.JJl/E,
(1.2)
based on the material dependent permeability j.l = !J.ol-lr and permittivity e = e0 er,
with !J.o and e0 being absolute permeability and permittivity of free space and 1-lr and er
denoting relative values. We also point out that the field components are orthogonal to
each other and both are orthogonal to the direction of propagation. This is known as
transverse electromagnetic mode (TEM) and, since we deal exclusively with RF, it is
Dimensions and Units
Figure 1-3 Electromagnetic wave propagation in free space. The electric and
magnetic fields are recorded at a fixed instance in time as a function of space
(Q, 9 are unit vectors in x- andy-direction).
the only mode that is considered in this text. TEM wave propagation is in stark contrast
to the various transverse electric (TE) and transverse magnetic (TM) wave modes,
which are the underlying principles of MW and optical communication. In these cases
the field vectors are no longer perpendicular to the direction of propagation.
The phase velocity vP of the TEM wave can be found via
v
p
= -(1) = - 1
~
(1.3)
Relevant quantities, units and symbols used throughout the book are summarized in
Tables A-1 and A-2 in Appendix A. Although we are dealing here with rather abstract
concepts of electromagnetic wave quantities, we can immediately relate (1.1) to circuit
parameters by observing that the electric field, as the unit of V/m already implies, can
intuitively be understood as a normalized voltage wave. Similarly, the magnetic field,
given in units of Aim, is a normalized current wave.
----------------------------RF&JA~
Example 1-1: Intrinsic wave impedance, phase velocity, and

wavelengths
Compute the intrinsic wave impedance, phase velocity, and wavelengths of an electromagnetic wave in free space for the frequencies
f = 30 MHz, 300 MHz, 30 GHz.
Solution:
Relative permeability and permittivity of free space
are equal to unity. Therefore, from (1.2) we determine that intrinsic
impedance in this case is equal to
41tX10-7
12
- 377
8.85xl0The phase velocity according to (1.3) is equal to

1
v = P
~
Je.oJlo
= 2.999x108 rnls
which happens to be the speed of light v P = c . The wavelength is

evaluated by the following expression:
A = 21t = 21tV p = v p
~
(J)
f
(1.4)
where f is the operating frequency. Using equation (1.4), we find

that the wavelength for an electromagnetic wave propagating in free
space at a frequency of 30 MHz is equal to A, = 10 m; at 300 MHz it
is already reduced to A = 1 m; and at 30 GHz the wavelength is a
minute A, = 1 em.
This example conveys an appreciation of how the wavelength

changes as a function of frequency. As the frequency increases, the
wavelength reduces to dimensions comparable to the size of circuit
boards or even individual discrete components. The implication of
this fact will be analyzed in Chapter 2.
1.3 Frequency Spectrum

Because of the vast scope of applications, engineers have to deal with a broad
range of frequencies of circuit operation. Over the years several attempts have been
made to classify the frequency spectrum. The first designations for industrial and government organizations were introduced in the United States by the Department of
Defense during and shortly after World War II. However, the most common frequency
spectrum classification in use today was created by the Institute of Electrical and Electronic Engineers (IEEE) and is listed in Table 1-1.
Frequency Spectrum
Table 1-1
Frequency Band
IEEE Frequency Spectrum

Frequency
Wavelength
ELF (Extreme Low Frequency)
30-300 Hz
10,000-1000 km
VF (Voice Frequency)
300-3000 Hz
1000-100 km
VLF (Very Low Frequency)
3-30kHz
100-10km
LF (Low Frequency)
30-300 kHz
10-1 km
MF (Medium Frequency)
300-3000 kHz
1--0.1 km
HF (High Frequency)
3-30MHz
100-10 m
VHF (Very High Frequency)
30-300MHz
10-1m
UHF (Ultrahigh Frequency)
300-3000 MHz
100-10 em
SHF (Superhigh Frequency)
3-30 GHz
10-1 em
EHF (Extreme High Frequency)
30-300 GHz
1--0.1 em
Decimillimeter
300-3000 GHz
1--0.1 mm
PBand
0.23-1 GHz
130-30 em
LBand
1-2GHz
30-15 em
SBand
2-4 GHz
15-7.5 em
CBand
4-8 GHz
7.5-3.75 em
X Band
8-12.5 GHz
3.75-2.4 em
KuBand
12.5-18 GHz
2.4-1.67 em
KBand
18-26.5 GHz
1.67-1.13 em
KaBand
26.5-40GHz
1.13--0.75 em
Millimeter wave
40-300GHz
7.5-1 mm
Submillimeter wave
300-3000 GHz
1--0.1 mm
Based on Table 1-1 and calculations carried out in Example 1-1 we note that the
VHF/UHF band, as typically encountered in television sets, constitutes the point at
which the wavelength first reaches dimensions equivalent to the physical extent of the
electronic system. It is this region where we need to begin to take into account the wave
nature of current and voltage signals in the respective electronic circuits. The situation
becomes even more critical when for instance 30 GHz frequency in the EHF band is
considered. Without being able to assign exact limits, the RF frequency range is customarily associated from VHF to the S band. The MW frequency range has been traditionally associated with radar systems operating in the C band and above.
10
1.4 RF Behavior of Passive Components

From conventional AC circuit analysis we know that a resistance R is frequency
independent and that a capacitor C and an inductor L can simply be specified by their
reactances Xc and XL as follows:
(1.5a)
Xc = roC
(1.5b)
The implications of ( 1.5), for example, are such that a capacitor of C = 1 pF and an
inductor of L = 1 nH at low frequencies of 60 Hz represent, respectively, either an open
or short circuit condition because
1
Xc(60 Hz) =
21t. 60. 10
_12 .:: 2.65xl0 .Q:;:: oo
(1.6a)
=3.77x10-7 Q::::: 0
(1.6b)
-9
XL(60 Hz) = 2n 60 10
It is important to point out that resistances, inductances, and capacitances are not only
created by wires, coils, and plates as typically encountered in conventional low-frequency electronics. Even a single straight wire or a copper segment of a printed circuit
board (PCB) layout possesses frequency dependent resistance and inductance. For
instance, a cylindrical copper conductor of radius a, length l, and conductivity a cond
has a DC resistance of
l
2
(1 .7)
1ta 0 cond
For a DC signal the entire conductor cross-sectional area is utilized for the current flow.
At AC the situation is complicated by the fact that the alternating charge carrier flow
establishes a magnetic field that" induces an electric field (according to Faraday's law)
whose associated current density opposes the initial current flow. The effect is strongest
at the center r = 0, therefore significantly increasing the resistance in the center of the
conductor. The result is a current flow that tends to reside at the outer perimeter with
increasing frequency. As derived in Appendix B, the z-directed current density Jz can be
represented by
1
2
= _p_I _1o_(p_r_)
z
2naJ 1 (pa)
(1.8)
where p = -jrof.HJcond, and J 0 (pr) , 1 1 (pa) are Bessel functions of zeroth and first
order, and I is the total current flow in the conductor. Further calculations reveal that the
11
RF Behavior of Passive Components
normalized resistance and inductance under high-frequency conditions(/~ 500 MHz)

can be put in the form
Rl Rvc
=a/(20)
(1.9)
and
( roL) / Rvc
=a / (28)
(1.10)
In these expressions () is the so-called skin depth
() = (1t/Jlcrcond)-ll 2
(1.11)
which describes the spatial drop-off in resistance and reactance as a function of frequency f, permeability Jl, and conductivity crcond . For the equations ( 1.9) and ( 1.1 0) to
be valid it is assumed that () a . In most cases, the relative permeability of the conductor is equal to unity (i.e., Jl, = 1 ). Because of the inverse square root frequency behavior, the skin depth is large for low frequencies and rapidly decreases for increasing
frequencies. Figure 1-4 exemplifies the skin depth behavior as a function of frequency
for material conductivities of copper, aluminum, and gold.
1
0.9
0.8
0.7
0.6
0.5
c-0
0.4
0.3
0.2
0.1
0 4
10
105
106
107
108
109
~Hz
Figure 1-4 Skin depth be~aviorofcopper Ocu

crAI = 40.0x10 S/m, and gold a Au
= 64.516x10~ S/m, aluminum

= 48.544x10 S/m.
If we consider the conductivity of copper, we can plot the AC current density ( 1.8)
normalized with respect to the DC current density J zo = I I ( 1ta 2) as schematically
shown for the axisymmetric wire depicted in Figure 1-5(a).

For a fixed wire radius of, let us say, a = 1 mm we can now plot J z/ J zO as a
function of radius r for various frequencies as given in Figure 1-5(b).
12
Low current
density
High current
density
' ---
-a
Current Flow
Figure 1-S(a)
Schematic cross-sectional AC current density representation

normalized to DC current density.
2
1.8
1.6
1.4
......'?,
.....,.,
1.2
I kHz
0.8
lOkHz
0.6
0.4
100kHz
0.2
00
Figure 1-S(b)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

r,mm
Frequency behavior of normalized AC current density for a

copper wire of radius a = 1 mm.
We notice the significant increase in current flow at the outer perimeter of the wire
even for moderate frequencies of less than 1 MHz. At frequencies around 1 GHz, the
current flow is almost completely confined to the surface of the wire with negligible
radial penetration. An often used high-frequency approximation for the z-directed current density is
lz=
lp
j2rta,fr
a-r
-(1 +J)-
(1.12)
13
As seen in (1.12), the skin depth has a simple physical meaning. It denotes the
reduction in the current density to the e- 1 factor (approximately 37o/o) of its original
DC value. If we rewrite (1.9) slightly, we find
(1.13)
This equation shows that the resistance increases inverse proportionally with the
cross-sectional skin area, see Figure 1-6.
Figure 1-6 Increase in resistance over the cross sectional surface area. The
current flow is confined to a small area defined by the skin depth o.
To standardize the sizes of wires, the American Wire Gauge (AWG) system is
commonly used in the United States. F~r-instance, the diameter of the wire can be
determined by its AWG value. A complete listing of all AWG values and their corresponding diameters is given in Table A-4 in Appendix A. The general rule is that in the
AWG system, the diameter of the wire roughly doubles every six wire gauges starting
with 1 mil for a AWG 50 wire (see Table A-4).
--------------------------~RF&JM~
Example 1-2: Conversion between wire diameter and AWG
size
Determine the radius of the AWG 26 wire if the diameter of the
5
AWG 50 wire is 1.0 mil (or 2.54x10- m).
14
The increase in diameter is computed as follows:

AWG 50 d= 1 mil
AWG44 d= 2 mils
AWG38 d = 4 mils
AWG32 d= 8 mils
AWG26 d =16 mils
Thus we detennined that the diameter of a AWG 26 wire is equal to
16 mils. Therefore, the radius is
Solution:
-5
8 mil = 8 x (2.54x10 m) = 0.2032 mm
Even in today's increaSingly metric world, AWG has retained

its importance, and knowledge of how to convert mil-based AWG
size wires into millimeters often proves indispensable.
I
1.4.1
High-Frequency Resistors
Perhaps the most common circuit element in low-frequency electronics is a resistor whose purpose is to produce a voltage drop by converting some of the electric
energy into heat. We can differentiate among several types of resistors:
Carbon-composite resistors of high-density dielectric granules
Wire-wound resistors of nickel or other winding material
Metal-film resistors of temperature stable materials
Thin-film chip resistors of aluminum or beryllium based materials
Of these types mainly the thin-film chip resistors find application nowadays in RF and
MW circuits as surface mounted devices (SMDs). This is due to the fact that they can
be produced in extremely small sizes, as Figure 1-7 shows.
As the previous section has shown, even a straight wire possesses an associated
inductance. Consequently, the electric equivalent circuit representation of a highfrequency resistor of nominal value R is more complicated and has to be modified such
that the finite lead dimensions as well as parasitic capacitances are taken into account.
This situation is depicted in Figure 1-8.
The two inductances L model the leads, while the capacitances are needed to
account for the actual wire arrangement, which always represents a certain charge sepa-
15
Figure 1-7 One- and quarter-watt thin-film chip resistors in comparison with a
conventional quarter-watt resistor.
Figure 1-8 Electric equivalent circuit representation of the resistor.
ration effect modeled by capacitance C a , and interlead capacitance C b . The lead resistance is generally neglected when compared with the nominal resistance R. For a wirewound resistor the model is more complex, as Figure 1-9 shows.
CI/
Figure 1-9
Electric equivalent circuit representation for a high-frequency wirewound resistor.
16
Here, in addition to the lead inductances L2 and the contact capacitance, we need
to include the inductance L 1 of the wire-wound resistor, which acts as a coil, and the
stray capacitance C 1 established between the windings. The interlead capacitance C2
(or C b in Figure 1-8) is usually much smaller than the internal or stray capacitance and
in many cases can be neglected.
----------------------------~~~
Example 1-3: RF impedance response of metal film resistors
~
Find the high frequency impedance behavior of a 500 .Q metal film

resistor (see Figure 1-8) with 2.5 em copper wire connections of
AWG 26 and a stray capacitance Ca of 5 pF.
Solution:
In Example 1-2 we have determined that the radius of

an AWG 26 wire is a = 2.032x10-4 m. According to (1.10) the
inductance of the straight wire at high frequency is approximately
equal to L = Roca/(2roo). Substituting (1.11) for the skin depth,
we get the following expression of a lead inductance (we set the
6 -1
-1
conductivity of copper to be <Jcu = 64.516x10 .Q m ):
L = Roc -
2 ro
J1t/Jlo<Jcu = -21-
a
f
24
<J1ta 1t
J1tfflo<Jcu = - 21 ~of
4 1ta 1t<JCu
= 1.54
!1 JlH
A/
where the length of the leads is doubled to account for two connections. The preceding formula for the computation of the lead inductance is applicable only for frequencies where the skin depth is
smaller than the radius of the wire [i.e., o = ('TtfJ..lcr)- 112 a] or in
terms of frequency f 1/(1tJ..lcrcua 2 ) = 95kHz.
Knowing the inductance of the leads, we can now compute the
impedance of the entire circuit as
z = jroL + jroC ~ 1/R

The result of the computation is presented in Figure 1-10, where the
absolute value of the impedance of the resistor is plotted versus
frequency.
As seen, at low frequencies the impedance of the resistor is
equal to R. However, as the frequency increases and exceeds
17
10
10
10
inductive effect
capacitive effect
10
10- 1
10-2
10-3
10
Frequency, Hz
Figure 1-10 Absolute impedance value of a 500-Q thin-film resistor as a

function of frequency.
10 MHz, the effect of the stray capacitance becomes dominant,

which causes the impedance of the resistor to decrease. Beyond the
resonance at approximately 20 GHz, the total impedance increases
due to the lead inductance, which represents an open circuit or infinite impedance at very high frequencies.
\
This example underscores the care that is required when dealing with the ubiquitous, seemingly frequency-independent resistors.
While not all resistors exhibit exactly the same response as shown in
Figure 1-10, it is the single, often multiple, resonance point that
occurs when the frequency reaches into the GHz range.
1.4.2
High-Frequency Capacitors
In most RF circuits chip capacitors find widespread application for the tuning of
filters and matching networks as well as for biasing active components such as transistors. It is therefore important to understand their high-frequency behavior. Elementary
circuit analysis defines capacitance for a parallel plate capacitor whose plate dimensions are large compared to its separation as follows:
18
A
C =cA::;:
d
ord
(1.14)
where A is the plate surface area and d denotes the plate separation. Ideally there is no
current flow between the plates. However, at high frequencies the dielectric materials
become lossy (i.e., there is a conduction current flow). The impedance of a capacitor
must thus be written as a parallel combination of conductance Ge and susceptance roC:
(1.15)
Ge+JcoC
In this expression the current flow at DC is due to the conductance Ge - a dielA/ d,

with a diel being the conductivity of the dielectric. It is now customary to introduce the
series loss tangent tan~ s = co/ adiei and insert it into the expression for Ge to yield
=
cocA
roC
= tan~s
(1.16)
Some practical values for the loss tangent are summarized in Table A-3. The corresponding electric equivalent circuit with parasitic lead inductance L, series resistance Rs
describing losses in the lead conductors, and dielectric loss resistance Re = 1I G e, is
shown in Figure 1-11.
Figure 1-11
Electric equivalent circuit for a high-frequency capacitor.
------------------------------~~~
Example 1-4: RF impedance response of capacitor

Compute the high frequency impedance of a 47 pF capacitor whose
dielectric mediu~-~oiisists of an aluminum oxide (AL 20 3) possessing a series loss tangent of 10-4 (assumed to be frequency independent) and whose leads are 1.25 em AWG 26 copper wires
6 -1
-1
( acu = 64.516x10 Q m ).
19
Solution:
Similar to Example 1-3, the inductance associated
with the leads is given by
a
JJ5o
2l
771
L = Rnc2- 1tf~ocrcu = 4f = ,yf1:.f nH

co
1ta 1t0cu
The series resistance of the leads is computed from ( 1.13) to be
a = 2 2l
I llof
r;
Rs = Roc2~
1t/~ocrcu = - - - = 4.8,yf ~Q
u
1tacrcu
a 1tOcu
Finally, in accordance with (1.16), the parallel leakage resistance is
equal to
R-_!_-
1
e - Ge - 21t/Ctan~s
33.9x10 MQ
The frequency response of the magnitude of the impedance based on

equation (1.15) for the capacitor is shown in Figure 1-12.
10'
Real \acitor
Ideal capacitor
2
10- L__~~~~__.__,_J.-~~~~.........._---~.._._.__._j
9
11
8
10
10
10
10
10
f, Hz
Figure 1-12 Absolute value of the capacitor impedance as a function of

frequency.
In computing the parallel leakage resistance Re we have

assumed the loss tangent tan~s to be frequency independent. In
reality, however, this factor may significantly depend upon the operating frequency. Unfortunately, data sheets often do not, or only
very incompletely, report this behavior.
20
Since the loss tangent can also be defined as the ratio of an

equivalent series resistance (ESR) to the capacitor's reactance, many
data sheets list ESR instead of tan~ s . The ESR value is thus given as
ESR
tan~s
= roc
This indicates that ESR ~ 0 as
tan~s ~
0.
As already known from the RF resistor impedance response in

Example 1-3, the capacitor reveals a similar resonance behavior
due to the presence of dielectric losses and finite lead wires.
The construction of a surface-mounted ceramic capacitor is shown in Figure 1-13.

The capacitor is a rectangular block of a ceramic dielectric into which a number of
interleaved metal electrodes are sandwiched. The purpose of this type of packaging is
to provide a high capacitance per unit volume by maximizing the electrode surface
area. Capacitance values range from 0.47 pF to 100 nF with operating voltage ranging
from 16 V to 63 V. The loss tangent is usually listed by the manufacturer as
3
tan~s ~ 10- at a 1 MHz test frequency. Again, this loss tangent can significantly
increase as the frequency reaches into the GHz range.
Terminations
Ceramic material
Figure 1-13
Actual construction of a surface-mounted ceramic multilayer

capacitor.
21
Besides capacitance and loss tangent, manufactures list a nominal voltage that
cannot be exceeded at a particular operating temperature (for instance, T ~ 85 oc ). Furthermore, the capacitance is temperature dependent, as further discussed in the problem
section of this chapter.
1.4.3
High-Frequency Inductors
Although not employed as often as resistors and capacitors, inductors generally

are used in transistor biasing networks, for instance as RF coils (RFCs) to short circuit
the device to DC voltage conditions. Since a coil is generally formed by winding a
straight wire on a cylindrical former, we know from our previous discussion that the
windings represent an inductance in addition to the frequency-dependent wire resistance. Moreover, adjacently positioned wires constitute separated moving charges, thus
giving rise to a parasitic capacitance effect as shown in Figure 1-14.
Figure 1-14 Distributed capacitance and series resistance in the inductor coil.
The equivalent circuit model of the inductor is shown in Figure 1-15. The parasitic shunt capacitance Cs and series resistance Rs represent composite effects of distributed capacitance Cd and resistance Rd, respectively.
I ~----c,'I/IA. . . .I-0
..
L
Figure 1-15
Rs
II
Equivalent circuit of the high-frequency inductor.
22
----------------------------RF&JA~
Example 1-5: RF impedance response of an RFC

Estimate the frequency response of an RFC formed by N = 3.5 turns
of AWG 36 copper wire on a 0.1 inch air core. Assume that the
length of the coil is 0.05 inch.
Figure 1-16 Inductor dimensions of an air-core coil.
Solution:
The dimensions of the coil are shown in Figure 1-16.
From Table A-4 in Appendix A, we find that the radius of the
AWG 36 wire is a= 2.5 mils= 63.5 Jlm. The radius of the coil core
is r = 50 mils = 1.27 mm. The length of the coil is l = 50 mils
= 1.27 mm. The distance between two adjacent turns is
d = liN=== 3.6x10-4 m.
To estimate the inductance of the coil we will use a wellknown formula for the inductance of an air core solenoid:
2
1tr JloN
L =
(1.17)
Strictly speaking, this formula is valid only for the case when r l
and the number of turns N is large. In our case, the length of the coil
is comparable with its radius and the number of turns is relatively
small. Therefore, (1.17) will not give an exact value for the inductance, but a rather good approximation. Substituting the given values into (1.17), we obtain L = 61.4 nH.
To approximate the effect of the capacitance Cs , we will use
the formula for an ideal parallel-plate capacitor (1.14). In our case
23
the separation d between the plates is assumed to be equal to the

distance between the turns d = l/ N::::: 3.6x10-4 m, and the area A
can be estimated as A = 2alwire, where /wire = 2rcrN is the length
of the wire. We conclude that
Eo
Cs =
2
2rcrN 2a
raN
- 4rcE0 l
= 0.087 pF
liN
Since the radius of the wire is only 63.5 Jlm, we can neglect the skin
effect and compute the series resistance Rs as a DC resistance of the
wtre.
Rs
lwrre
0' Cu rca
2rcrN
0'Cu rca
= 0.034 .Q
The frequency response of the RFC impedance just analyzed is

Ideal inductor
Real inductor
~
J, Hz
Figure 1-17
Frequency response of the impedance of an RFC.
RFCs find widespread use for biasing RF circuits. However, as

Figure 1-17 shows, the frequency dependency can form complicated
resonance conditions with additional elements in an RF system.
Indeed, certain matching circuits rely on the RFCs as tuning elements.
24
As can be seen from Figure 1-17, the behavior of the RFC deviates from the
expected behavior of an ideal inductance at high frequencies. First, the impedance of
the RFC increases more rapidly as the frequency approaches resonance. Second, as the
frequency continues to increase, the influence of the parasitic capacitance Cs becomes
dominant and the impedance of the coil decreases.
If the RFC had zero series resistance, then the overall impedance behavior at resonance would reach infinity, but due to the nonzero value of Rs the maximum value of
the impedance is of finite value. To characterize the impact of the coil resistance, the
quality factor Q is commonly used:
Q=x
Rs
(1.18)
where X is the reactance and Rs is the series resistance of the coil. The quality factor
characterizes the resistive loss in this passive circuit, and for tuning purposes it is desirable that this factor is as high as possible.
1.5 Chip Components and Circuit Board Considerations
The practical realization of passive components on printed RF circuit boards is
primarily accomplished in chip form and placed on specially fabricated board materials. In the following section we examine the three most common passive chip elements
in terms of their sizes and electric characteristics.
1.5.1
Chip Resistors
The size of chip resistors can be as small as 40 by 20 mils (where 1 mil= 0.001
inch= 0.0254 mm) for 0.5 W power ratings and up to 1 by 1 inch for 1000 W ratings in
RF power amplifiers. The chip resistor sizes that are most commonly used in circuits
operating up to several hundred watts are summarized in Table 1-2.
A general rule of thumb in detennining the size of the chip components from the
known size code is as follows: the first two digits in the code denote the length Lin
terms of tens of mils, and the last two digits denote the width W of the component. The
thickness of the chip resistors is not standardized and depends on the particular component type.
The resistance value range from 1/10 n up to several MQ. Higher values are difficult to manufacture and result in high tolerances. Typical resistor tolerance values range
from +5% to +0.01% . Another difficulty that arises with high-value resistors is that they
are prone to produce parasitic fields, adversely affecting the linearity of the resistance versus frequency behavior. A conventional chip resistor realization is shown in Figure 1-18.
25
Chip Components and Circuit Board Considerations
Table 1-2
Geometry
Standard sizes of chip resistors
Size Code
Length L, mils
Width W, mils
0402
40
20
0603
60
30
0805
80
50
1206
120
60
1218
120
180
Marking
\End contact
End contact
Inner electrodes
Ceramic substrate
"-----
Figure 1-18
Cross-sectional view of a typical chip resistor.
A metal film (usually nichrome) layer is deposited on a ceramic body (usually

aluminum oxide). This resistive layer is trimmed to the desired nominal value by reducing its length and inserting inner electrodes. Contacts are made on both ends of the
resistor that allow the component to be soldered to the board. The resistive film is
coated with a protective layer to prevent environmental interferences.
1.5.2
Chip Capacitors
The chip capacitors are implemented either as a conventional single-plate configuration, as shown in Figure 1-19, or a multiple-layer design (see Figure 1-13).
Frequently, single-plate capacitors are combined in clusters of two or four elements sharing a single dielectric material and a common electrode, as shown in Figure
1-20.
26
Chip capacitor
Ribbon lead or wire
Circuit traces
Figure 1-19 Cross section of a typical single-plate capacitor connected to the

board.
Dual capacitor
Quadrupole capacitor
11
1111
I
Figure 1-20 Clusters of single-plated capacitors sharing a common dielectric

material.
The standard sizes of the capacitors range from a minimum of 15 mils square in a
single layer configuration to 400 by 425 mils at higher values. Typical values for commercial capacitors range from 0.1 pF to several J..LF . The tolerances vary from 2% to
50%. For small capacitances tolerances are usually expressed in terms of pF instead
of percent; for example, we often encounter capacitors with the nominal values given as
(0.5 0.25) pF.
1.5.3
Surface-Mounted Inductors
The most common implementation of surface-mounted inductors is still the wirewound coil. A typical example of such an inductor with air core is shown in Figure 1-21.
Modem manufacturing technology allows us to make these inductors extremely small.
Their dimensions are comparable to those of chip resistors and capacitors. Typical sizes
Chip Components and Circuit Board Considerations
27
Figure 121 Typical size of an RF wire-wound air-core inductor in comparison

with a cellular phone antenna (courtesy of Coilcraft, Inc.).
of the surface-mounted wire-wound inductors range from 60 by 30 mils to 180 by 120

mils. The inductance values cover the range from 1 nH to 1000 J..LH .
When thickness constraints of the circuit play a major role, flat inductors are often
employed that can be integrated with microstrip transmission lines. A generic configuration of a flat coil is shown in Figure 1-22. Although such thin-wire coils have relatively low inductances on the order of 1 to 500 nH, it is the frequency in the GHz range
that helps push the reactance beyond 1 kn . The physical construction can be as small
as 2 mm by 2 nun.
Air bridge
Figure 122
Flat coil configuration. An air bridge is made by using either a wire

or a conductive ribbon.
Flat coils are used in both integrated and hybrid circuits. Hybrid circuits are very
similar to an ordinary circuit, but discrete semiconductor elements are placed on the
dielectric substrate in die form (without case) and are connected to the conductors on
28
the board using bond wires. After the entire circuit is assembled, it is then placed into a
single case to protect it from environmental interferences. Resistors and capacitors for
hybrid circuits can directly be implemented on the board by metal-film deposition. This
approach permits significant reduction in the size of the circuit.
1.6 Summary
In this chapter the evolution from low- to high-frequency systems is discussed and
placed in a historical context. A key concept when dealing with high-frequency applications is the fact that the electromagnetic wave nature begins to dominate over Kirchhoff's current and voltage laws. Issues such as propagation constant and phase velocity,
1
= 27t/A and v =\ ~ = - P
/p
gain importance.
A consequence of the electromagnetic wave nature is the skin effect, which forces
the current to flow close to the surface of the conducting structures. The depth of penetration from the surface can be determined via the skin depth equation:
3=
Jrcjp.,cr
With the skin depth we can approximately characterize the frequency dependent resistance and reactance of components at RF frequency. As an example, the simple cylindrical lead wires exhibit resistances and reactances that become a function of frequency
R ::= Roc
and X = roL ::= R0 c
20
20
These wires, in conjunction with the respective R, C, and L elements, fonn electric equivalent circuits whose performance markedly deviate from the ideal element behavior. We
find that the constant resistance at low frequency is no longer constant, but displays a
second-order system response with a resonant dip. The dielectric material in a capacitor
becomes lossy at high frequencies (i.e., allows the flow of a small conduction current).
The degree of loss is quantified by the loss tangent, which is tabulated for a range of
engineering materials. Therefore, a capacitor exhibits an impedance behavior that follows an inverse frequency response only at low frequencies. Finally, inductors represent
an impedance response that follows a linear increase at low frequencies before deviating
from the ideal behavior by reaching a resonance peak and then turning capacitive.
A passive RF component vendor will always attempt to keep the physical dimensions of resistors, capacitors, and inductors as small as possible. This is desired since
the wavelength of high-frequency voltage and current waves becomes ever smaller,
Further Reading
29
eventually reaching the characteristic sizes of the circuit components. As discussed in

subsequent chapters, when the wavelength is comparable in size with the discrete electronic components, basic circuit analysis no longer applies.
Further Reading
I. Bohl and P. Bhartia, Microwave Solid State Design, John Wiley, New York, 1988.
C. Bowick, RF Circuit Design, Newmes, Newton, MA, 1982.

D. K Chen, Fundamentals of Engineering Electromagnetics, Addison-Wesley, Reading,
MA, 1993.
R. A. Chipman, Transmission Lines, Schaum Outline Series, McGraw-Hill, New York,
1968.
L. N. Dworsky, Modern Transmission Line Theory and Applications, Robert E. Krieger,
Malabar, FL, 1988.
M. F. Iskander, Electromagnetic Fields and Waves, Prentice Hall, Upper Saddle River,
NJ, 1992.
T. S. Laverghetta, Practical Mircowaves, Prentice Hall, Upper Saddle River, NJ, 1996.
H. P. Neff, Basic Electromagnetic Fields, 2nd ed. Harper & Row, New York, 1987.
K. F. Sander, Microwave Components and Systems, Addison-Wesley, 1987.
K. F. Sander and G. A. L. Read, Transmission and Propagation of Electromagnetic
Waves, 2nd ed. Cambridge University Press, Cambridge, UK, 1986.
W. Sinnema, Electronic Transmission Line Technology, 2nd ed., Prentice Hall, Upper
Saddle River, NJ, 1988.
F. T. Ulaby, Fundamentals ofApplied Electrqmagnetics, Prentice Hall, Upper Saddle

;
River, NJ, 1997.
Problems
1.1 Compute the phase velocity and wavelength in an FR4 printed circuit board
whose relative dielectric constant is 4.6 and where the operational frequency
is 1.92 GHz.
30
1.2
The current flowing in a microstrip line (assumed to be infinite and lossless)

9
is specified to be i(t) = 0.6cos(9 x 10 t - 500z) A. Find the (a) phase
velocity, (b) frequency, (c) wavelength, and (d) phasor expression of the current.
1.3
A coaxial cable that is assumed lossless has a wavelength of the electric and
magnetic fields of A = 20 em at 960 MHz. Find the relative dielectric constant of the insulation. "'\
1.4
The electric wave field of a positive z-traveling wave in a medium with relative dielectric constant of r =4 and with frequency of 5 GHz is given by
Ex = E0xcos(rot- kz) V/m
6
(a) Find the magnetic field if Eox = 10 V/m.

(b) Determine phase velocity and wavelength.
(c) Compute the spatial advance of the traveling wave between time intervals
t 1 = 3J..LS and t 2 = 7J..LS.
1.5
Find the frequency response of the impedance magnitude of the following

series and parallel LC circuits:
L= lOnH
L = lOnH
~~
C= lOpF
C= lOpF
Compare your results to the situation when the ideal inductance is replaced
by the same inductance and a 5 .Q resistance connected in series. Assume
that these circuits operate in the VHF/UHF frequency band of 30-3000
MHz.
1.6
For the circuit shown, derive the resonance frequency and plot the resonance
frequency behavior as a function of the resistance R.
L= lOnH
Problems
1.7
31
Repeat Problem 1.6 for the following circuit.

R
C= lpF
1.8
For the following circuit we chose R ( JLI C)/2.

R
Find IV0 /Vil as a function of frequency and identify the dominant circuit
portions for the low-, mid-, and high-frequency domains.
1.9
One of the objectives of Chapter 1 is to sensitize the reader to high-frequency phenomena that are usually neglected in a low-frequency circuit analysis. One such phenomenon is the skin effect. To show its importance in RF
calculations, (a) compute the frequency behavior of an inductor formed by
10 turns of AWG 26 copper wire on a 5 mm air core. The length of the coil is
5 mm. (b) repeat the computations by first neglecting the skin effect and then
including it.
1.10 The leads of a resistor in an RF circuit are treated as straight aluminum wires
7
( crAI = 4.0 x 10 S/m) of AWG size 14 and of total length of 5 em. (a)
Compute the DC resistance. (b) Find the AC resistance and inductance at
100 MHz, 1 GHz, and 10 GHz operating frequencies.
1.11
Compute the skin depths for copper ( <Jcu = 64.516 x 10 S/m), aluminum
6
6
( cr AI = 40 x 10 S/m ), and gold ( crAu = 48.544 x 10 S/m) at 1 GHz and
10 GHz, and find the resistance of a 10 em wire with diameter of 1 mm.
1.12 A typical PCB substrate consists of Al 20 3 with a relative dielectric constant

of 10 and a loss tangent of 0.0004 at 10 GHz. Find the conductivity of the
substrate.
32
1.13 For the series RLC circuit with R = 1 .Q, L = 1 nH, and C = 1 pF,
compute the resonance frequency and quality factor at 10% of the resonance frequency. Does the presence of the resistor affect the resonance frequency?
1.14 A 4.7 pF capacitor with relative dielectric constant of 4.6 and series loss tangent of 0.003 is used in a circuit operated at 10 GHz. For a combined copper
lead lengtllof 6 em and diameter of 0.5 mm, determine (a) the lead resistance and lead reactance, and (b) the conductance and the total imRedance.
6
1
The conductivity of copper is given as O'cu = 64.516x10 .n- m- .
1.15 A manufacturer data sheet records the series loss tangent of a capacitor to be
2
2
10-4 at 5 GHz. For a total plate dimension of 10- cm and plate separation
of 0.01 mm and a relative dielectric constant of 10, find the conductance.
1.16 A two-element impedance of the generic form
Z
= R+jX
has to be converted into an equivalent admittance form Y = liZ such that

Y
= G+jB
Find the conductance G and susceptance B in terms of resistance R and reactance X.

1.17 A more elaborate model of a capacitor is sometimes represented by the following circuit:
Here the loss tangent is specified as consisting of two parts involving the
admittance YP = 1/RP+jroC with a parallel-circuit loss tangent
tan~P = !Re{YP}/Im{YP}I and series impedance Zs = Rs + 1/(jroC)
with a series-circuit loss tangent tan~s = !Re{Zs}/Im{Zs}l (it is noted
that Rs is different from Example 1-4). Show that for low-loss capacitances
we
approximately
obtain
tan~~ tan~s + tan~P,
where
tan~ = IRe{Z}/Im{Z}I and Z is the total impedance.
1.18 When recording the capacitance with a measurement equipment, the user
has often the choice to select a suitable circuit representation. For the series
representation, the instrument attempts to predict Rs and C s, while for the
Problems
33
parallel representation the prediction involves R P and C P. Which mode

should be chosen if large capacitors of more than 100 ~F are to be measured? Is this mode also suitable for small values of less than 10 ~F ?
Explain your answers.
1.19 The ability to store electric charge, expressed through the capacitance,
depends on the operating temperature. This behavior can be quantified
through the relation C = C0 [1 + a(T- 20C)], where C 0 is the nominal
capacitance and a is a temperature coefficient that can be positive or negative. If the capacitance C at T = 20C is recorded to be 4.6 pF, which
increases to 4.8 pF at T = 40C, what is the temperature coefficient a?
Determine the capacitance at ooc and 80C.
1.20 When measuring impedance at low frequency we connect the measurement
equipment to a device using a pair of wires and assume that the reading
reflects the impedance of the device under test (DUT). As we have seen in
this chapter, at high frequencies we have to take into account the influence of
the parasitic elements. The typical circuit representation of the measurement
arrangement is as follows.
ZDUT
Measurement
Equipment
DUT
Cables and Fixture

Measurement
Plane
Device
Plane
Here the fixture and cables are replaced by an equivalent circuit of the lead
impedance (Rs+}roLs) and stray admittance (Gp+}roCp). Ideally, we
would like to perform the measurement at the device plane. However, due to
the influence of the fixture, the measurement plane is shifted away from the
DUT.
To measure accurately the impedance of the DUT, the test fixture with
connecting cables has to be taken into account. The methodology adopted by
most manufacturers is to compensate for these undesired, fixture-related
influences through an open- and short-circuit calibration. The first step is to
34
replace the DUT by a short circuit and record the resulting impedance. Due
to the influence of the fixture, the measured impedance will not be equal to
zero. Next, the short circuit is replaced by an open circuit and the impedance
is recorded again. These two measurements allow us to quantify the parasitic
influence of the fixture.
After calibration, we can connect the DUT and measure the input
impedance. The equivalent circuit in this case is as follows.
Zs =Rs+jroLs
Knowing the values of the parasitic elements ( Z s and Yp ), we can now

compute the true impedance of the OUT.
Explain the procedure with all necessary equations, and specify under
what conditions such a calibration is possible. Next, develop the formula that
allows us to find the desired DUT impedance in the absence of the fixture.
1.21
The results of a frequency sweep impedance measurement of an unknown

passive device are shown in the following figure.
10 5 ~~~~~~~~~~~-r~~~~~~~
10
a
~
N 10
0
0
"'0
0
-S'
10
w
10 4
10
10
10
10
10
10
Frequency, Hz
1010
10
11
1012
Problems
35
Based on the shape of the impedance response, develop a circuit that can be
used as an equivalent circuit to replicate this device under test. What device
can it be: resistor, inductor, or capacitor?
1.22 To measure the impedance of a passive component at RF frequencies is quite
a challenge. Conventional techniques such as bridge circuits and resonance
techniques fail beyond a few MHz. A technique pursued by several instrument manufactures is the current voltage recording based on the following
simplified schematic.
Here the voltages are measured with vector voltmeters that allow the recording of magnitude and phase. Explain how the impedance of the component
under test is determined and discuss the purpose of the transformer and
operational amplifier.
1.23 An RFC is constructed by winding four turns of AWG 38 copper wire on a
2 mm ceramic core diameter (Jlr = 1) of 0.1 mm length. Based on Example
1-5, estimate the inductance, stray capacitance, resistance, and resonance
frequency.
1.24 Using data and the equivalent circuit diagram developed in the previous
problem, find values of the equivalent circuit parameters for the magnitude
of the impedance if the device is 100 Q under DC conditions and 1257 Q
at 100 GHz. Assume the resonance frequency point to be at 1.125 GHz.
1.25 A quadrupole capacitor as shown in Figure 1-18 consists of four equal-size
electrodes of 25 mils square separated 5 mils from a common ground plane
through a dielectric medium of a relative dielectric constant of 11 . Find the
individual and total capacitance that can be achieved.
36
1.26 Consider the following diode circuit.

D~IASo---------,
RFC
-___,1---..-------o RFour
RFIN o - - - -........
5nH
As will be shown in Chapter 6, a reverse biased diode can be represented as

a series combination of a resistor Rs and junction capacitor C, where the
capacitance is bias dependent. Its value is approximately given by the
expression
co(l - v
bias)-1/2
v cliff
Assuming that RFC and blocking capacitor C B have infinite values, find the
biasing voltage such that the circuit exhibits a resonance at the frequency of
1 GHz. The diode is characterized as follows: C 0 = 10 pF, Rs = 3 Q,
and barrier voltage V diff = 0.75 V.
CHAPTER
Transmission Line Analysis
A s we already know, higher frequencies imply

decreasing wavelengths. The consequence for an RF circuit is that voltages and currents
no longer remain spatially uniform when compared to the geometric size of the discrete
circuit elements: They have to be treated as propagating waves. Since Kirchhoff's voltage and current laws do not account for these spatial variations, we must significantly
modify the conventional lumped circuit analysis.
The purpose of this chapter is to outline the physical reason for transitioning from
lumped to distributed circuit representation and, in the process, develop one of the most
useful equations: the spatially dependent impedance representation of a generic RF
transmission line configuration. The application of this equation to the analysis and
design of high-frequency circuits is going to assume central importance in subsequent
chapters. Developing the background of transmission line theory in this chapter, we have
purposely attempted to minimize (albeit not eliminate) the reliance on electromagnetics.
The motivated reader who would like to delve deeper into the concepts of electromagnetic wave theory is referred to a host of excellent books listed at the end of this chapter.
2.1 Why Transmission Line Theory?

Let us once again return to the wave field representation (l.la):
Ex = E 0xcos(rot- ~z). Here we have an x-directed electric field propagating in the
positive z-direction. For propagation in free space the orthogonality between electric
field and direction of propagation is always assured. If, on the other hand, we assume
that the wave is confined to a conducting medium that is aligned with the z-axis, we will
find that the electric field has a longitudinal component Ez that, when integrated in zdirection, gives us a voltage drop (i.e., V = -JEzdlz, where dlz is the line element in the
37
Chapter 2 Transmission Line Analysis
38
z-direction). Let us now consider more closely the argument of the cosine term in
(1.1a). It couples space and time in such a manner that the sinusoidal space behavior is
characterized by the wavelength A along the z-axis. Moreover, the sinusoidal temporal
behavior can be quantified by the time period T = 1If along the time-axis. In mathematical terms this leads to the method of characteristics, where the differential change
in space over time denotes the speed of evolution, in our case the constant phase velocity in the form vP:
v
- co - Aj -
p-~-
(2.1)
-~-M
For a frequency of, let us say, f = 1 MHz and medium parameters of E, = 10 and
7
flr = 1 ( v P = 9.49x10 rnls ), a wavelength of A= 94.86 m is obtained. This situation is
spatially and temporally depicted in Figure 2-1 for the voltage wave
V = -Jcos(rot- ~z)dz = sin( rot- ~z)/~.
20.-~--~~--~--~~--~--~~---
z=O
<:'
10
-10
-20~~--~~--~--~~--~--~~--~
0.2
0.4 0.6
0.8
1.0 1.2
t, )J.S
1.4 1.6
1.8 2.0
2o--~--~~--~--~~--~--~~--
t=
10
-10
-2o~~--~~--~--~~--~--~~~
Figure 2-1
20
40 60
80 100 120 140 160 180 200

z,m
Voltage distribution as a function of time (z = 0) and as a function of

space (t = 0).
We next direct our attention to a simple electric circuit consisting of load resistor
RL and sinusoidal voltage source V G with internal resistance Ra connected to the load
by means of 1.5 em long copper wires. We further assume that those wires are aligned
along the z-axis and their resistance is negligible. If the generator is set to a frequency
of 1 MHz, then, as computed before, the wavelength will be 94.86 m. A 1.5 em long
wire connecting source with load will experience spatial voltage variations on such a
minute scale that they are insignificant.
39
Why Transmission Line Theory?
When the frequency is increased to 10 GHz, the situation becomes dramatically

10
different. In this case the wavelength reduces to 'A = vP/10 m = 0.949 em and thus
is approximately two-thirds the length of the wire. Consequently, if voltage measurements are now conducted along the 1.5 em wire, location becomes very important in
determining the phase reference of the signal. This fact would readily be observed if an
oscilloscope were to measure the voltage at the beginning (location A), at the end (location B), or somewhere along the wire, where distance A-B is 1.5 em measured along the
z-axis in Figure 2-2.
v
VA
~
<:]
.,
z =I
-..1
R::
z
~
-~
Figure 2-2 Amplitude measurements of 10 GHz voltage signal at the beginning

(location A) and somewhere in between a wire connecting load to source.
We are now faced with a dilemma. A simple circuit, seen in Figure 2-2, with a
voltage source VG and source resistance RG connected to a load resistor RL through a
two-wire line of length 1, whose resistance is assumed negligible, can only be analyzed
with Kirchhoff's voltage law
N
L= vi= o
(2.2)
when the line connecting source with load does not possess a spatial voltage variation,
as is the case in low-frequency circuits. In (2.2) Vi (i = 1, ... , N) represents the voltage
drops over N discrete components. When the frequency attains such high values that the
40
spatial behavior of the voltage, and also the current, has to be taken into account, Kirchhoff's circuit laws cannot be directly applied. The situation can be remedied, however,
if the line is subdivided into elements of small, (mathematically speaking) infinitesimal
length, over which voltage and current can be assumed to remain constant, as depicted
in Figure 2-3.
z
J(z)!
t'
V(z)
R1
z+Az
! J(z+Llz)
tV(z+dz)
!
!
z
z+Llz
Figure 2-3 Partitioning an electric line into small elements & over which
Kirchhoff's laws of constant voltage and current can be applied.
For each section of length &, we can devise an equivalent electric circuit representation. With reference to our discussions in Chapter 1 it is immediately concluded
that there will be some series resistance and inductance associated with each wire. In
addition, due to the relative proximity of the two wires, a capacitive effect will also be
observed. Since in reality no perfect insulator does exist, a small current flow through
the dielectric occurs. A more accurate analysis of all these effects will be given in
Section 2.2. At this point we need to stress that equivalent elements, briefly described
here, represent only a small segment of the line. To build the complete model of the
entire line we would have to replicate & a large number of times. Therefore, the transmission line in general cannot be represented in terms of lumped parameters, but must
be viewed as distributed parameters R, L, C, and G, where all circuit parameters are
given in terms of unit length.
The question of when a wire, or a discrete component, has to be treated as a transmission line cannot precisely be answered with a single number. The transition from
lumped circuit analysis obeying Kirchhoff's laws to distributed circuit theory involving
Examples of Transmission Lines
41
voltage and current waves depends on the wavelength in comparison with the average
component size. The transition takes place gradually as the wavelength becomes
increasingly comparable with the circuit elements. As a rule of thumb, when the average size lA of the discrete circuit component is more than a tenth of the wavelength,
transmission line theory should be applied (LA~ IJlO). For the example of the 1.5 em
wire we would detennine the following frequency estimation:
:J:_ = 9.49 x 10 m/s
= 633 MHz
10l
0.15 m
Can the RF design engineer deal with the simple circuit in Figure 2-2 as a lumped element representation at 700 MHz? Perhaps. Can Kirchhoff's circuit theory be applied to
the circuit at 1 GHz? Not without having to take into account a significant loss in precision. Additional reasons why the use of transmission line theory is needed will become
apparent in later chapters.
2.2 Examples of Transmission Lines

2.2.1
Two-Wire Lines
The two-wire transmission line discussed in Section 2.1 is one example of a system capable of transporting high-frequency electric energy from one location to
another. Unfortunately, it is perhaps the most unsuitable way of transmitting high-frequency voltage and current waves. As shown schematically in Figure 2-4, the two conductors separated over a fixed distance suffer from the drawback that the electric and
magnetic field lines emanating from the conductors extend to infinity and thus influence
electronic equipment in the vicinity of the line.
Electric Field
(solid lines)
Magnetic Field
(dashed lines)
Figure 2-4 Geometry and field distribution in two-wire parallel conductor

transmission line.
42
Further, due to the fact that the wire pair acts as a large antenna, radiation loss
tends to be very high. Therefore, the two-wire line finds only limited applications in the
RF domain (for instance, when connecting private TV sets to receiving antennas). However, it is commonly used in 50-60 Hz power lines and local telephone connections.
Even though the frequency is low, the distance can easily extend over several kilometers, thus making the wire size comparable to the wavelength (as an example,
8
A = c/f = 3 x 10 /60 = 5000 km). Here again, distributed circuit behavior may
have to be taken into account.
2.2.2
Coaxial Line
A more common example of a transmission line is the coaxial cable. It is used for
almost all cases of externally connected RF systems or measurement equipment at frequencies of up to 10 GHz. As shown in Figure 2-5, a typical coaxial line consists of an
inner cylindrical conductor of radius a, an outer conductor of radius b, and a dielectric
medium layered in between. Usually the outer conductor is grounded, thus minimizing
radiation loss and field interference. Several of the most commonly used dielectric
materials include polystyrene (, = 2.5, tan.1s = 0.0003 at 10 GHz), polyethylene
(, = 2.3, tan.1s = 0.0004 at 10 GHz), or teflon(,= 2.1, tan~s = 0.0004 at 10 GHz).
Figure 2-5 Coaxial cable transmission line.

2.2.3
Microstrip Lines
It is a common practice to use planar printed circuit boards (PCBs) as the basic
medium to implement most electronic systems. When dealing with actual RF circuits,
we need to consider the high-frequency behavior of the conducting strips etched on the
PCBs, as depicted qualitatively in Figure 2-6.
The ground plane below the current carrying conductor traces helps prevent excessive field leakage and thus reduces radiation loss. The use of PCBs simplifies the access
I!DmpiM of Transmission Lines
43
(a) Printed circuit board section

(b) Microstrip line
Figure 2-6 Microstrip transmission line representation.
to the active and passive devices on the board and reduces the cost of the manufacturing
process. In addition, PCBs allow the tuning of circuits by simply changing the position
'>fthe components and manually adjusting variable tuning capacitors and inductors.
One of the disadvantages of single layered PCBs is that they have rather high radi.ttion loss and are prone to "crosstalk" (interference) between neighboring conductor
traces. As noted in Figure 2-7, the severity of field leakage depends on the relative
dielectric constants, as shown qualitatively in the electric field line displays for teflon
epoxy (7 =2.55) and alumina ( 7 = 10.0) dielectrics.
(a) Teflon epoxy ( e7 = 2.55 )
(b) Alumina ( 7 = 10.0)
Figure 2-7 Electric field leakage as a function of dielectric constants.

Direct comparison of the field lines in Figure 2-7 suggests that to achieve high
board density of the component layout, we should use substrates with high dielectric
constants since they minimize field leakage and cross coupling.
Another way to reduce radiation losses and interference is to use multilayer techniques to achieve balanced circuit board designs where the microstrip line is "sandwiched" between two ground planes, resulting in the triple-layer configuration seen in
Figure 2-8.
A microstrip configuration that is primarily used for low impedance, high-power
applications is the parallel-plate line. Here the current and voltage flow is confined to two
plates separated by a dielectric medium. This configuration and the corresponding field
distribution are shown in Figure 2-9 for the dielectric medium of teflon epoxy ( 7 = 2.55).
Chapter 2 Trantmlulon Line Analyalt
44
(a) Sandwich structure (er
= 2.55)
(b) Cross-sectional field distribution
Figure 2-8 Triple-layer transmission line configuration.
(a) Geometric representation
(b) Field distribution (Er = 2.55)
Figure 2-9 Parallel-plate transmission line.
There are many more transmission line configurations used for a number of special-purpose applications. However, a detailed coverage of the pros and cons of all possible combinations would go beyond the objectives of this book.
The preceding transmission line examples all have the commonality that the electric and magnetic field components between the current-carrying conductors are transversely orientated (or polarized); that is, they form a transverse electromagnetic (TEM)
field pattern similar to the one shown in Figure 1-3. As mentioned in Chapter 1, the
TEM behavior has to be seen in contrast to guided modes, where the electromagnetic
wave propagation is accomplished through wave reflections and refractions between
conducting plates or indexed dielectric media in optical fibers. The analysis is broken
down into so-called transverse magnetic (TM) and transverse electric (TE) modes.
Such modes of operation are of major interest in the microwave range for satellite communication, radar, and remote sensing applications. Due to their extremely high frequency of operation, well above the RF range, waveguides and optical fiber cables
require special electromagnetic treatment and are not considered further. Instead, we
refer the reader to a number of references listed at the end of this chapter.
Equivalent Circuit Representation
45
2.3 Equivalent Circuit Representation

As mentioned previously, voltages and currents are no longer spatially constant on
the geometric scale of interest to RF circuit design engineers. As a consequence, Kirchhoff's voltage and circuit laws cannot be applied over the macroscopic line dimension.
However, this problem can be circumvented when the transmission line is broken down
into smaller (in the limit infinitesimally small) segments. Those segments are still large
enough to contain all relevant electric characteristics such as loss, as well as inductive
and capacitive line effects. The main advantage of this reduction to a microscopic representation is the fact that a distributed parameter description can now be introduced
whose analysis follows Kirchhoff's laws on a microscopic scale. Besides providing an
intuitive picture, the approach also lends itself to a two-port network analysis, as discussed in Chapter 4.
To develop an electric model, let us consider once again a two-wire transmission
line. As Figure 2-10 indicates, the transmission line is aligned along the z-axis and segmented into elements of length ~ .
Z
~ at .
...
'
'
...... ...
... ... ..
Z + Llz
:'
1
2
- ........
.......... .......
.~4
......
r;R"; ~ 1
1(~)Ji~
'
. ,,
,.,.......
i ~!~
t
i
G ~-:~:. c ..l.. v(:Z )
... : ..... . .'"
:~:~
.~~
'
~
~---,t- ,, _>l,.U
.~? ..
..~:...:..~~:
- l ( ':'
! ...
':.~
.,_i
~ ----.i.-- - - ---
Y1
i:
R2
..l-----------: _.,,,. __
z+
~z
Figure 2-10 Segmentation of two-wire transmission line into L\2'-long sections

suitable for lumped parameter analysis.
If we focus our attention on a single section residing between z and z + ~' we
notice that each conductor (1 and 2) is described as a series connection of resistor and
inductor (R 1 , L 1 , and R 2 , L 2 ). In addition, the charge separation created by conductors 1 and 2 gives rise to a capacitive effect denoted by C. Recognizing that all dielectrics suffer losses (see our discussion in Section 1.4.2), we need to include a
46
conductance G. Again attention is drawn to the fact that all circuit parameters R, L, C,
and G are given in values per unit length.
Similar to the two-wire transmission line, the coaxial cable in Figure 2-11 can
also be recognized as a two-conductor configuration with the same lumped parameter
representation.
:\a
ft.";i~---
.......--_~::....:
-----
J(z}J?.~....-
..
Jh .- ~~.-~;\/\. . . . . . . . .
.. . . . ....
."J(z-~&) R
1
-.
--r
~, . . . . .. . .....,
. . .. ... . . . . . ;_
L
1
-- :
1
\ :. '
...., ..... ,.....
.
.....+--:vv\:
---rrt
_....
~~......--
G ,_l_:.. Cl... Vi( t)
+-uu,....-
.. .... ____
-...........
........ " _.... .:-.. ......:-i'((~----~--- ~ ----'-i'+
., . ,.
,/+dz
V(z + Az)
_,,.,.__
_1:.
.~2
1
.:V\:--)..~?..
r ) :.......
..... ..........
l li
:_~: C:=t
1
~ - --- - - --' -- --- r- -
z
Figure 2-11
Segmentation of a coaxial cable into M length elements suitable for

lumped parameter analysis.
A generic form of an electric equivalent circuit is developed as shown in Figure 212, where the resistances and inductances of the two conductors are usually combined
into single elements. This representation is not suitable for all transmission line applications. For instance, when dealing with transient wave propagation and signal integrity
issues of inductive and capacitive crosstalks, it generally makes more sense to retain the
parameter representation shown in Figure 2-11. However, for our treatment of transmission lines we will exclusively use the model shown in Figure 2-12.
l(z
+ L\z)
_....
V(z)
t
z
Figure 2-12
V(z + L\z)
t
z + L\z
Generic electric equivalent circuit representation.
Theoretical Foundation
47
It should be recalled from the discussion in Chapter 1 that the aforementioned R,

L, C, and G elements are frequency-dependent parameters that change significantly
depending on the operational frequency and the employed transmission line type. Further, L not only incorporates the inductance of the wire (self-inductance; see Section
1.4.3) but also takes into account the mutual inductance between the wires. In general,
the self-inductance is so small compared with the mutual inductance that it can be
safely neglected. To summarize the advantages of the electric circuit representation, we
observe that it
provides a clear intuitive physical picture
lends itself to a standardized two-port network representation
permits the analysis with Kirchhoff's voltage and current laws
provides building blocks that allow the expansion from microscopic to macroscopic fonns
There are also two significant disadvantages worth noting:
It is basically a one-dimensional analysis that does not take into account field
fringing in the plane orthogonal to the direction of propagation and therefore cannot predict interference with other components of the circuit.
Material-related nonlinearities due to hysteresis effects are neglected.
Despite these disadvantages, the equivalent circuit representation is a powerful mathematical model for describing the characteristic transmission line behavior. With this
model in place, we can now embark on developing generalized transmission line
equations.
2.4 Theoretical Foundation

2.4.1
Basic Laws
The next question that we should ask ourselves is how to determine the distributed
circuit parameters if we know the physical dimensions and electric properties of the
transmission line. The answer is provided through the use of two central laws of electromagnetics: Faraday's law and Ampere's law.
Rooted in experimental observations, Faraday's and Ampere's laws establish two
fundamental relations linking electric and magnetic field quantities. As such, both laws
provide cornerstones of Maxwell's theory by stating so-called source-field relations. In
other words, the time-varying electric field as a source gives rise to a rotational magnetic field. Alternatively, the time-varying magnetic field as a source results in a timevarying electric field that is proportional to the rate of change of the magnetic field. The
48
mutual linkage between electric and magnetic fields is ultimately responsible for wave
propagation and traveling voltage and current waves in RF circuits.
By stating Faraday's and Ampere's laws in integral and differential forms, we possess the necessary tools to calculate, at least in principle, the line parameters R, L, C,
and G for the electric circuit elements. They are needed to characterize various transmission line systems. By going through the subsequent calculations, we will observe
how abstract theoretical laws can be used as a starting point to derive practical circuit
parameters for a particular type of transmission line.
Ampere's Law
This fundamental law states that moving charges, which are characterized by the
current density J, give rise to a rotational magnetic field H surrounding the charge flow
as expressed by the integral relation
fH
dl
= JJJ dS
(2.3)
where the line integral is taken along the path characterized by the differential element
dl that defines the edge of the surface element S in such a manner that the surface S
always stays on the left side. In equation (2.3) the total current density can be written as
J = J 0 + crE + d(EE)Idt. It is comprised of (a) the impressed source current density
J0 , (b) the conduction current density crE, which is induced by an electric field E in the
conductor and is responsible for conduction losses; and (c) the displacement current
density d(EE)Idt, which is responsible for radiation losses. Here and in the following
equations we use again bold letters to denote vector quantities such that
E(r, t) = ExCx, y, z, t)x + Ey(x, y, z, t)y + E/x, y, z, t)z
where Ex, EY, Ez are the vector components and x, y, are unit vectors in x, y, z directions in a Cartesian coordinate system. Figure 2-13 illustrates the meaning of equation
(2.3).
Perhaps less intuitive than the integral relation, nonetheless perfectly identical to
(2.3), is Ampere's law in differential or point form:
(VxH) n = lim .l!H dl
~ ~ o.1Sj
= ~lim
.lJJJ dS
~ o.1S
= J n
(2.4)
where Vx is the curl operator and n is a unit vector perpendicular to the surface element dS. When using vector components in a rectangular coordinate system, this differential operator can be represented in the matrix form
49
Theoretical Foundation
........
I =
Figure 2-13
fH
dl
Ampere's law linking the current flow to the magnetic field.
Vx =
a
az
a
ay
a a
az ay
a
0
ax
a 0
ax
(2.5)
Therefore, by applying the curl to the vector field H, we obtain
VxH
= a
az
a
ay
a a
az ay
a
0
ax
a 0
ax
Hx
HY
Hz
Jx
Jy
(2.6)
Jz
where Hx, HY' Hz and Jx, JY' Jz are x, y, and z components of the magnetic field vector H,
and the current density J.
so
-----------------------------~&uM~
Example 2-1: Magnetic field generated by a constant current
flow in a conductor
Plot the graph of the radial magnetic field H ( r) inside and outside
an infinitely long wire of radius a = 5 mm aligned along the z-axis
and carrying a DC current of 5 A. The surrounding medium is
assumed to be air.
Solution:
This is a typical example for Ampere's law in integral
form as given by (2.3). Inside the conductor the current density J is
uniform and is equal to J = I /(1ta 2 )z . Therefore, the application
of (2.3) yields the following result:
2
= ~Ttr
2
=>
H = ....!..!:_2
1ta
21ta
where 0 $ r $ a . Outside of the conductor the current density is
equal to zero and the surface integral in (2.3) gives the total current I
flowing through the conductor. Thus, the magnetic field H outside
the wire is obtained as
I
H21tr =I
H==>
H21tr
21tr
where r ~a. The total magnetic field inside and outside of the infinitely long wire is thus
lr
H(r) =
-2na 2
<
,r_a
I
,r
2nr
~a
31 .83r kNm, r $5 mm
0.196/r Aim, r ~ 5 mm
The graph of this radial magnetic field distribution is plotted in

Figure 2-14.
We make the important observation that inside the wire the
magnetic field linearly increases from the center to the outer con-
1blorttlcal Foundation
51
140
Ir
H=-
120
21ta2
100
~~
80
60
40
20
0
Figure 2-14
10
15
20
25
r,mm
30
35
40
45
50
Magnetic field distribution inside and outside of an infinitely long wire

of radius a = 5 mm carrying a current of 5 A.
ductor periphery since more current contributes to the magnetic

field.
Faraday's law
This law implies that the time rate of change of the magnetic flux density B = J..LH
.<J.L = J.loJ.lr) as a source gives rise to a rotating electric field
(2.7)
The line integral is again taken along the edge of the surface S as previously described
p..r Ampere's .law. Th~ integration of the electric field aJon~ a wire loop as shown in
i : 11
-~ figure 2-15 yields an mduced voltage V =
-fE
dl = d -f B dS .
1
Similar to Ampere's law, we can convert (2.7) into a ch erential, or point form:
VxE
dB
= -dt
(2.8)
~-Equation (2.8) makes it clear that we need a time-dependent magnetic flux density to
f;.obtain an electric field,

L
,!
which in tum creates a magnetic field according to Ampere's law.
52
Figure 2-15 The time rate of change of the magnetic flux density induces a
voltage.
--------------------------~RF~~
Example 2-2: Induced voltage in a stationary wire loop
Find the induced voltage of a thin wire loop of radius a = 5 mm in
air subjected to a time-varying magnetic field H = H 0 cos (rot) ,
where H 0 = 5 Aim , and the operating frequency 1s
f = 100 MHz.
Solution:
The voltage induced in the loop is equal to the line
integral of the electrical field E along the loop. Employing Faraday's
law (2.7) results in the following:
V =
-fE dl = : JfBdS
1
Since the surrounding medium is air, the relative permeability f..Lr

equals
unity
and
the
magnetic
flux
density
is
B = f..LoH = JloH 0 cos ( rot)z. Substituting B into the preceding
integral leads to an expression for the induced voltage V in the loop:
:tJJB
= : 1fl 0 H 0 cos(rot)1ta 2 = -1ta 2 rof.10H 0 sin(rot)

8
This can be further simplified to V = -0.31 sin(6.28x10 t) V.
V
dS
Ctrcult Parameters for a Parallel Plate Transmission line
53
The result of this example is also known as the transfonner

fonn of Faraday's law whereby a time-varying field produced by a
primary coil induces a voltage response in a secondary loop.
2.5 Circuit Parameters for a Parallel Plate Transmission Line

Our goal is to compute the line parameters R, L, C, and G for a section of a transmission line seen in Figure 2-16. To avoid any confusion we explicitly use a cond and
odie! to denote, respectively, conductivity in the conductor and conductivity in the
dielectric medium.
Figure 2-16
Parallel-plate transmission line geometry. The plate width wis large

compared with the separation d .
.,
for the analysis we must assume that the plate width w is large compared with the plate
separation d for a one-dimensional analysis to apply. Further, we assume that the skin
depth is small compared to the thickness d P of the plates to simplify the derivation of
the parameters. Under these conditions we are able to cast the electric and magnetic
fields in the conducting plates in th~ form
(2.9a)
H
1001
The term e
= yHix, z)eJror
(2.9b)
represents the time dependence of the sinusoidal electric and magnetic

fields, and phasors Ez(x,z) and Hy(x,z) encode spatial variations. We do not have any
field dependency upon y, because the plates are assumed very wide, and thus the electromagnetic fields do not change appreciably along they-axis. Application of the differential forms of Faraday's and Ampere's laws
54
VxE :::: -!l-
aH
(2.10)
VxH :::: O'condE
(2.11)
at
results in two differential equations:
a a
az ay
a 0 a
ax
az
0
a a
ay ax
[~ l=_dE, =-~~~; l
ax
dt
Ez
::::
= -/Hy
dt
-jWJlHy
(2.12)
and
a a
:Z ~
0
ay
a a
ay ax
:x [ol
:y=-d~y = [ol
0
cond ; ,
condz
(2.13)
By differentiating (2.13) with respect to x and substituting (2.12), we find

d
Hy _
--2 -
.
_
jWO'condf.l.Hy -
p 2H Y
(2.14)
dx
2
where p
jroocondJ..l The general solution for this second-order ordinary differential
equation (2.14) is H Y ( x) :::: A e - px + B epx . The coefficients A and B are integration
constants. We can now perform the following manipulations:
= JjrocrconctJl = J)JroocondJl = (1 + J)j(roocondJl) / 2 = (1 + J) I O (2.15)

where o = J2! (roocondll) is recognized as the skin depth. Since p has a positive real
P
component, constant A should be equal to zero to satisfy the condition that the magnetic
field in the lower plate must decay in amplitude for negative x. A similar argument can
be made for the upper plate by setting B = 0. Thus, for the magnetic field in the lower
conducting plate we have a simple exponential solution
HY
_ H
oe
px _
oe
( l + j)xl o
(2.16)
55
ClrcuH Parameters for a Parallel Plate Transmission Line
where B = H 0 is a yet to be determined constant factor. Since the current density can
be written as
aH y
J z = aEz = - =
ax
( 1 + j)Ho
(1 + J)xlo
(2.17)
we are now able to relate the current density Jz to the total current flow I in the lower
plate
I =
JJ
s
Jzdx = wH0 e
Jzdxdy = w
-d
(1 + J)xlol
_
dp
= wH0 ( 1- e
-(1 + J)d
P
/o
(2.18)
where Sis the cross sectional area of the lower plate and dP is the thickness of that plate.
Since we assume that d P
the exponential term in (2.18) drops out and I = w H 0 .
From this we conclude that H 0 = I I w . The electric field at the surface of the conductor (x 0) can be specified as
o,
E (0) =
z
lz(O)
(J cond
(1+j)H 0
(J cond 0
l+j I
(J cond 0 W
(2.19)
Equation (2.19) allows us to compute the surface impedance per unit length, Zs, by
eliminating the current I as follows:
(2.20)
The surface resistance and surface inductance per unit length are then identified as
(2.21)
L =
s
(2.22)
wcrcondroo
o.
Both are dependent on the skin depth

It is important to point out that (2.21) and
(2.22) apply for a single conductor. Since we have two conductors in our system (upper
and lower plates) the total series resistance and inductance per unit length will be twice
the value of Rs and Ls, respectively.
To obtain the inductive and capacitive behavior of the mutual line coupling, we
p1ust employ the definitions of capacitance and inductance:
C=Q
fn. dS
v
eJJExdS _
- JExd/x -
EW
(2.23)
Chapter 2 Transmission Line Analysla
56
and
JJB. dS = ff~HydS = ~Hyd = ~Hyd = ~d

I
Hyw
(2.24)
where we have used the result of (2.18) to compute the current I = wHY. Both in (2.23)
and (2.24) the capacitance and inductance are given per unit length.
Finally, we can express the conductance Gin a similar way as derived in (2.23):
G
= ~ = adierffExdS = adieiExw = adielw

V
fExdl
Exd
(2.25)
Thus we have succeeded in deriving all relevant parameters for the parallel-plate transmission line. From a practical point of view, at RF frequencies the magnitude of Ls is
typically much smaller than L and therefore is neglected.
; ~~~-o
~~-....g..
\.
'
" ..........L.>/
---------------------------~~~
Example 2-3: Line parameters of a parallel-plate transmission
line
For a parallel copper-plate transmission line operated at 1 GHz, the
following parameters are given: w =6 mm, d =1 mm, Er =2.25, and
odiel = 0.125 mS/m. Find the line parameters R, L, G, and C per unit
length.
Solution:
The skin depth for copper with conductivity
6
ocond = 64.516x10 .Q- 1m- 1 at operating frequency of 1 GHz is
<> = 1I Jna condllof = 1.98 Jlm , which is assumed to be much
smaller than the thickness of the conductor. Therefore, the resistance
of each plate is determined by (2.21). Since we have two plates, the
total resistance is R = 2R5 = 2/(wO'condo) = 2.6 .Q/m. The
series
inductance
due
to
the
skin
effect
is
Ls = 2/(wO'condro()) = 0.42 nH/m, where the factor 2 takes into
account both plates. The mutual inductance between plates is determined by (2.24) and for our problem is equal to L = 209.4 nH/m.
As seen, the series inductance is much smaller than the mutual
inductance and therefore can safely be neglected. According to
Summary of Different Line Configurations
57
(2.23), the capacitance of the line

IS
gtven by
C = (0 ,w)/d = 119.5 pF/m. Finally, the conductance G is
determined from (2.25) and equals G = 0.75 mS/m.
The RF surface resistance due to the skin depth phenomenon

does, in general, contribute much more significantly to the line
losses than does the DC resistance.
2.6 Summary of Different Line Configurations

The previous computations were carried out for the relatively simple case of a
parallel-plate transmission line. Similar analyses apply when dealing with more complicated line geometries, such as coaxial cables and twisted wire pairs. Table 2-1 summarizes the three common transmission line types.
Table 2-1
Transmission line parameters for three line types

Coaxial Line
Parallel-Plate Line
Unit
0/m
1tacrcond~
1 (1 1)
21tcrcond~ a+ b
wcrcond~
~acosh( 2a
D)
1t
~In(~)
21t
a
JLw
1t0diel
acosh(D/(2a))
21tcrdiel
ln(b/ a)
adield
1tE
acosh (DI (2a))
21tE
Parameter
Two-Wire Line
ln(b/a)
w
d
E-
Him
S/m
F/m
The geometric dimensions for the two-wire (D, a), coaxial (a, b), and parallelplate (w, d) lines are depicted in Figures 2-4, 2-5, and 2-16. The term acosh in Table 2-1
denotes the inverse hyperbolic cosine function. For more complex transmission line
configurations, significant mathematical effort must be exerted, and resorting to numerical analysis procedures is often the only available solution. This is seen when dealing
with microstrip transmission lines (Section 2.8).
58
2.7 General Transmission Line Equation

2.7.1
Kirchhoff Voltage and Current Law Representations
Having developed the background of Faraday's and Ampere's laws in

Section 2.4.1, we are well positioned to exploit both equations from a circuit point of
view. This is identical with applying Kirchhoff's voltage and current laws (KVL and
KCL, respectively) to the loop and node a shown in Figure 2-17.
I(z+&)
V(z+&)
t
z
Figure 2-17
Z +&
Segment of a transmission line with voltage loop and current node.
Adopting phasor notation, we can use Kirchhoff's voltage law to conclude

(R + jroL)I(z)~z + V(z + ~z) = V(z)
(2.26)
which is re-expressed as a differential equation by combining the voltage drop on either

side of the differential transmission line segment into a differential quotient:
lim ( V(z + 8z)- V(z))
Llz~o
8z
= _dV(z)
= (R + jroL)l(z)
dz
(2.27)
or
- dV(z)
dz
= (R + jroL)I(z)
(2.28)
where R and L are the combined resistance and inductance of the two lines. Applying
Kirchhoff's current law to the designated node a in Figure 2-17 yields
I(z)- V(z + ~)(G + jroC)8z
= l(z + .1z)
(2.29)
which can be converted into a differential equation similar to (2.27). The result is
_ (G . C)V( )
. /(z + 8z)- l(z) -_ dl(z)
1Im
-- - + ]00
z
Liz~ o
.1z
dz
(2.30)
Equations (2.28) and (2.30) are coupled first-order differential equations. They can also
be derived from a more fundamental point of view, revealing the definitions of R, G, C,
fk......
Transmission Line Equation
'
59
'
~Las
discussed in Example 2-4 for the previously analyzed parallel-plate transmisklion line example.
i~
~1
---------------------------~8UM~
Example 2-4: Derivation of the parallel-plate transmission line
equations
Establish the transmission line equations for the parallel-plate conductors.
Solution:
The purpose of this example is to show how the transmission line equations (2.28) and (2.30) can be derived from the
fundamental physical concepts of Faraday's and Ampere's laws.
Let us first consider Faraday's law (2.7). The surface element
over which the line and surface integrations are performed is shown
as a shadowed area in Figure 2-18.
2
plate 1
Figure 2-18
Integration surface element for Faraday's law application.
The line integral in (2. 7) is taken along the edge of the shaded
region with the integration direction denoted by arrows in Figure 218. Evaluation of this line integration yields the following contributions:
fE dl = E 1 z1lz + E(z + Llz) id- E 2

i
( -z)Llz
+ E(z) (-x)d
'~
-------------------------------- :~
60
Chapter 2 Transmission Line Analylle '
where E} = E 1 and E'f = E 2 are the electric fields in the

lower (denoted by index 1) and upper (index 2) plates, respectively;
and Ex(z) = E(z) x and Ex(z + ~z) = E(z + ~z) x are the
electric fields in the dielectric medium between locations z and
z + ~ . It is important to note that the direction of the electric field
in the upper conductor is opposite to that of the electric field in the
lower conductor, whereas the direction of the field in the dielectric is
the same regardless of position. The minus sign in front of the unit
vectors indicates that the integration is performed counterclockwise.
Combining terms, we obtain
fE dl
E}~ + E'f~ + Ex(z + ~)d- Ex(z)d
Since the magnetic field in the dielectric is assumed uniform, the

integration over the surface in (2. 7) gives
ffJ.LH dS = J!Hy~d
Substitution of these two integrals into (2.7) results in
d
E}~ + E'f~ + Ex(z + ~)d- Ex(z)d = - dtJ!Hy~d
Similar to discussions in Section 2.5, the magnetic field in the

dielectric can be expressed as H Y = I I w . The electric field in the
conductor at high frequency is dependent on the skin effect and is
E}
E'f = I /(wcrcondB) + ji /(wcrcondB)
Ez. At low frequency, the skin effect does not affect the electric field behavior. The
field is solely determined by the DC resistivity of the plates and current I: Ez = I /(wcrconddp). Since we are primarily concerned with
the high-frequency performance, we must assume that the skin
depth B is much smaller than the thickness of the plates. Thus, d P
has to be replaced with B. Combining expressions for H Y and E z,
and taking into account the relation for the potential between the
plates, V = E xd, we obtain
2(wcrcondB
I
+
ji )~ + V(z + &) - V(z) = -ll dt,.zdi = -jroJl dilzl
wcrcondB
w dt
w .
or
V(z)
-- V(z + dz)~
61
..,T,.,.mlulon Une EquaUon
where R s = 1 I ( wa cond 0) is the surface resistance of the plates,

Ls = 11(wacondroo) is the high-frequency self-inductance of the
plates, and L = lld I w is the mutual inductance between both plate
conductors.
For the application of Ampere's law (2.3) we use the surface
element shown in Figure 2-19.
plate 2
X
'plate I
Figure 2-19 Surface element used to apply Ampere's law.
The surface integral of the current density J in the dielectric

medium results in the following expression
ff
aEx
J. dS = Jx~w = crdieiExw~ +at w~
where the adieiExw~ term represents the conduction current in the

dielectric, and (aExlat)w~ is the contribution of the displacement current. The line integration of the magnetic field yields
fH dl = - Hy(z + ~)w + Hy(z)w = - l(z + ~) + l(z)

Taking into account the relation between the electric field and the
potential drop Vbetween z and z +~,that is, Ex = VI d, we combine both integrals:
adielw V
wdV _
+ddt-
I(z + ~)- l(z)
or, after introducing the differential quotient,

_ aJ = crdietw V + wdV = 0 dietw V w. V
(G + ,roc)V
d
d dt
d
+ d ]ffi =
az
62
Chapter 2 Tranamlaalon Line Analyala
Thus, we succeeded in deriving the equations for the parallel-plate

transmission line. To obtain the voltage and current distribution of
such a line, the following system of coupled first-order differential
equations must be solved:
- ~~ = [2Rs + jro(L + 2Ls) ]I
-ai
az = (G + jroC)V
Usually, the self-inductance due to the skin effect Ls is much
smaller than the mutual inductance L and is often neglected.
This example underscores the effort and assumptions required

to develop closed-form expressions for the parallel-plate transmission line. However, if w is comparable in size to d, the preceding
treatment breaks down and one has to resort to numerical
simulations.
2. 7.2
Traveling Voltage and Current Waves
The solution of equations (2.28) and (2.30) is greatly facilitated if these first-order
differential equations are decoupled. This can be accomplished by spatially differentiating both sides of (2.28) and substituting (2.30) for the space derivative of the current.
The result is a standard second-order differential equation
2
V~z)
k2V(z) =
(2.31)
dz
describing the voltage behavior in phasor form. Here the factor k is known as a complex
propagation constant
k = k, + jk; = J(R + jroL)(G + jroC)
(2.32)
that depends on the type of transmission line. For simple line configurations, Table 2-1
provides explicit parameters. Reversing the order of decoupling by differentiating
(2.30) and substituting (2.28) results in an identical differential equation describing the
current:
63
Trantmlsslon Line Equation
d I(z) k2 I(z)
dz
_
...vu~
=0
(2.33)
to these decoupled equations are two exponential functions for the voltage
+ -kz
V( z ) = V e
+ v- e+kz
I(z) = I+ e -kz +I-e +kz
(2.34)
(2.35)
:important to observe that (2.34) and (2.35) are general solutions for transmission
aligned along the z-axis. The convention is such that the first term represents
v.,I'Pn",T~ propagating in the +z-direction, whereas the second term denotes wave
OJ)81gat1on in the -z-direction. This makes physical sense since the negative sign in
,..,u..."'"vu with kr ~ 0 ensures diminishing amplitudes for the positive (+z) traveling
Conversely, negative traveling waves are attenuated due to the diminishing expo-
2.7.3
General Impedance Definition
Equation (2.35) is related to (2.34). This can be seen if (2.34) is substituted into
. Differentiating and rearranging provides us with a current expression in the folform:
(2.36)
voltage and current are generally related via an impedance, we can introduce the
.....-u~ characteristic line impedance Z 0 by defining
Zo
= (R + jroL) =
k
lllgo ...,u.n.u.&J:O,
(R + jroL)
(G + jroC)
(2.37)
the current expression (2.35) into the left-hand side of (2.36), we also find
v+
v-
Zo = - - - I+
I
(2.38)
impedance allows us to express the current (2.36) in the concise form

(2.39)
64
The importance of Z 0 will become apparent in the following sections. Here it is noteworthy to point out that Z 0 is not an impedance in the conventional circuit sense. Its
definition is based on the positive and negative traveling voltage and current waves. As
such this definition has nothing in common with the total voltage and current expressions used to define a conventional circuit impedance.
2.7.4
Lossless Transmission Line Model
The characteristic line impedance defined in (2.37) is, in general, a complex quantity and therefore takes into account losses that are always present when dealing with
realistic lines. However, for short line segments, as mostly encountered in RF and MW
circuits, it does not create an appreciable error to deal with lossless line conditions. This
implies R = G = 0 and the characteristic impedance (2.37) simplifies to
(2.40)
Since Z 0 is independent of frequency, current and voltage waves are only scaled by a
constant factor. It is instructive to substitute values for a particular transmission line
type. If we use the parallel-plate transmission line with L and C given in Table 2-1, we
find the explicit form
Zo =
~d
~ew
(2.41)
where the square root term is known as the wave impedance, which yields ( ll = llo ,
e = e0 ), a value of approximately 377 .Q in free space. This value is typical when dealing with radiation systems whereby an antenna emits electromagnetic energy into free
space. However, unlike electromagnetic field radiation into open space, the transmission line introduces geometric constraints as expressed through w and d for the parallelplate line configuration.
2.8 Microstrip Transmission Lines

As we have seen in Figures 2-6 and 2-7, a simple treatment of the strip line as a
parallel-plate capacitor that formed the basis of computing C in Table 2-1 does not
apply in the general case. If the substrate thickness h increases or if the conductor width
w decreases, fringing fields become more prominent and cannot be ignored in the mathematical model. Over the years a number of researchers have developed approximate
expressions for the calculation of the characteristic line impedance, taking into account
conductor width and thickness. As often encountered in engineering, we have to strike a
balance between complexity and the accuracy of our computations. The most precise
65
111>1res~;tm1s
describing microstrip lines are derived by using conformal mapping, but

expressions are also the most complex, requiring substantial computational
For the purposes of obtaining fast and generally reliable estimations of the line
simpler empirical formulas are more beneficial.
As a first approximation, we assume that the thickness t of the conductor forming
line is negligible compared to the substrate height h (tlh < 0.005 ). In this case, we
use empirical formulas that depend only on the line dimensions (wand h) and the
aetc~1l't'c constant r. They require two separate regions of applicability depending on
_ ...........,.. the ratio wlh is larger or less than unity. For narrow strip lines, w I h < 1 , we
"""........... the line impedance
Z0
w)
4
ZJ~In ( 8-+
h
h
21t tV'eff
(2.42)
Z f = J~ 0 1 0 = 376.8 .Q is the wave impedance in free space, and

...-a~ dielectric constant given by
ft
,; 1 ,; 1[(1 12:r
112
Eeff
+ o.o4(
1- ~YJ
teff
is the
(2.43)
La w~de line, w/h > 1, we need to resort to adifferent characteristic line impedance
Lpress10n:
Z0 --
Eeff
zf
h(1.393 + ~ + j1n(~ + 1.444))

=
+1
r2
r2
1(
h)-1/2
1 + 12W
(2.44)
(2.45)
. is important to note that the characteristic impedances given by (2.42) and (2.44) are
approximations and do not produce continuous functions over the entire range of
. h . In particular, we notice that at w I h = 1 the characteristic impedance computed
to (2.42) and (2.44) displays a small discontinuity. Since the error introduced
this discontinuity is less than 0.5%, we still can use the preceding expressions for the
~u~"~""'.... of both the characteristic line impedance and the effective dielectric conas shown in Figures 2-20 and 2-21. In these figures the quantities Z 0 and eff are
!dotted as functions of w I h ratios and r values. The parameter range of w I h and r
" " ".............JO.
is chosen such that it spans the domain of typically encountered practically relevant circuit values.
Er =
12
Line width to dielectric thickness ratio, w /h
Figure 2-20
w~
Characteristic line impedance as a function of wlh.
10
6L----4
=4
3
=2
r =
211------------------------~r~~~
r = 1
Figure 2-21
Effective dielectric constant as a function of wlh for different

dielectric constants.
~I; .
------------------------------.._..,.p
'.. :...,
Transmission Lines
67
In the preceding formulas the effective dielectric constant is viewed as the dielec,tric constant of a homogeneous material that fills the entire space around the line,
. replacing dielectric substrate and surrounding air. With the knowledge of the effective
dielectric constant we can compute the phase velocity of the strip line as
This leads to an expression for the wavelength of
vP = c/
Ji:r.r.
vP
'A=-=
Ao
=-
th Ji:rr'
where, as before, c is the speed of light and f is the operating frequency.

For design purposes we would like to have a relation that allows us to compute
wlh ratios based on a given characteristic impedance Z 0 and dielectric constant er of
the substrate. Assuming an infinitely thin line conductor, we can write (see Sobol's article in Further Reading at the end of the chapter) for w I h ~ 2 :
8eA
(2.46a)
where the factor A is given by

A= 21t Zo
2J
J"r 2+ 1 +
E,- 1 ( 0.23+0.11)
r
+1
For wI h ~ 2 we obtain:
~{B-1-ln(2B-l)+ er-l[ln(B-1)+0.39- 0 61 ]}
W =
h
1t
2er
er
(2.46b)
where the factor B is given by
-----------------------------F~~
Example 2..5: Design of a microstrip line
A particular RF circuit requires that a line impedance of 50 Q is to
be maintained. The selected PCB board material is FR-4 with a relative dielectric constant of 4.6 and a thickness of 40 mil. What are the
width of the trace, phase velocity, and wavelength at 2 GHz?
68
Solution:
At first we can use Figure 2-20 to determine an
approximate ratio of w I h . Choosing a curve corresponding to
E, = 4.6' we find that for Zo = 50 n' wlh is approximately 1.9.
Therefore, in (2.46) we have to chose the case where w I h < 2 . This
leads to
A
= 21t Zo JEr + 1 + Er- 1 (o.23 + 0.11) = 1.5583
z1
e, + 1
e,
Substituting this result into (2.46a), we find

w
SeA
::;::; 1.8477
A
h
e 2 -2
Then, by using (2.45), we obtain the effective dielectric constant
to be
Eeff
= Er 2+ 1 + Er 2-
1(
h)-112
1 + 12 W
= 3.4575
We can compute the characteristic impedance of the line (2.44) to

verify our result:
Z0
zf
JE:~1.393 +~+~In(~+ 1.444))
= 50.2243 Q
which is very close to the target impedance of 50 Q and therefore

indicates that our result is correct.
Using the obtained ratio for w I h , we find the trace width to be
w
73.9 mil. Finally, the effective dielectric constant just computed allows us to evaluate the phase velocity of the microstrip line
vP
= cl h
= 1.61x108 rnls
and the effective wave length at 2 GHz
A= vplf
= 80.67 mm
Strictly speaking, this example focuses on a single trace of infinite length only. In reality, proximity to neighboring traces and
bends is an issue of practical importance that is most easily
accounted for in RF/MW computer aided design (CAD) programs.
Terminated Loaeleu Transmission Line
69
For many applications the assumption of zero thickness of the strip line may not be
valid and corrections to the preceding equations are needed. The effect of nonzero copper
strip thickness is approximated as an increase in effective width weff of the conductor
since more fringing fields will occur. In other words, a finite thickness is modeled by simply replacing the width of the strip in (2.42)-(2.45) with an effective width computed as
t(
2x)
weff = w + ~ 1 + In t
(2.47)
where t is the thickness of the conductor, and either x = h if w > h/(21t) > 2t, or
x = 21tw if h/(21t) > w > 2t .
The influence of nonzero thickness on the characteristic line impedance for a standard FR-4 substrate with h = 25mil is illustrated in Figure 2-22.
150
c:
FR-4
h = 25 mil
r = 4.6
G)
~ 100
G)
c..
.e
.9
.....0
50
f!
0
o.1
0.3
3
1
10
Figure 2-22
Effect of conductor thickness on the characteristic impedance of a

microstrip line placed on a 25 mil thick FR-4 printed circuit board.
As seen in the figure, the effect is most noticeable for narrow strips, while it become
almost negligible for cases when the width is greater than the thickness of the dielectric.
2.9 Terminated Lossless Transmission Line

2.9.1
Voltage Reflection Coefficient
High-frequency electric circuits can be viewed as a collection of finite transmission line sections connected to various discrete active and passive devices. Therefore,
70
Chapter 2 Transmls1lon Line Analyala
let us take at first a closer look at the simple configuration of a load impedance connected to a finite transmission line segment of length l depicted in Figure 2-23. Such a
system forces us to investigate how an incident voltage wave propagating along the
positive z-axis interacts with a load impedance representing a generic line termination.
l
J
z ..
z =- 1
Figure 2-23
Terminated transmission line at location
z = 0.
Without a loss of generality, the load is assumed to be located at z = 0 and the

voltage wave is coupled into the line at z = -l . As we know, the voltage anywhere
along the line is generically given by (2.34). The second term in (2.34) has the meaning
of a reflection from the terminating load impedance for values z < 0. We introduce the
voltage reflection coefficient r 0 as the ratio of reflected to incident voltage wave
vro =v+
-
(2.48)
at the load location z 0. As a consequence of this definition, the voltage and current
waves can be re-expressed in terms of the reflection coefficient as
(2.49)
and
V
l( z) = -(e
Zo
-kz
- r oe
+kz
(2.50)
If (2.49) is divided by (2.50), we find the impedance as a function of space Z(z) anywhere along the z-axis -l ~ z ~ 0 . For instance at z = -l the total input impedance Zin
is recorded, and for location z = 0 the impedance becomes the load impedance
(2.51)
Terminated Loaelels Transmission Line
71
.Equation (2.51) can be solved for the reflection coefficient
r0
ZL-Zo
ro = ZL
--+Zo
with the result
(2.52)
This is a more useful representation than (2.48) since it involves known circuit quantities independent of particular voltage wave amplitude ratios.
We conclude that for an open line ( Z L -7 oo) the reflection coefficient becomes 1,
which means the reflected wave returns with the same polarity as the incident voltage.
In contrast, for a short circuit ( Z L = 0 ) the reflected voltage returns with inverted
amplitude, resulting in r 0 = -1 . For the case where the load impedance matches the
line impedance, Z 0 = Z L , no reflection occurs and r 0 = 0 . If there is no reflection we
have the case where the incident voltage wave is completely absorbed by the load. Tills
can be regarded as if a second transmission line with the same characteristic impedance, but infinite length, is attached at z = 0.
2.9.2
Propagation Constant and Phase Velocity
The definition of the complex propagation constant (2.32) assumes a very simple
fonn for the lossless line (R = G = 0 ). For this case we obtain
k
= kr + jki = jroJLC
(2.53)
This is identified in generally accepted engineering notation as

a=kr = 0
(2.54)
and
(2.55)
where a represents the attenuation coefficient and P is the wave number or propagation constant for lossless lines. The propagation constant is now purely imaginary,
resulting in
(2.56)
and
+
I( z) = -V ( e -JPz Zo
r 0 e+JPz)
(2.57)
Here, the characteristic impedance is again given by (2.40). Furthermore, from (2.1) it is
known that the wavelength A. can be related to the frequency f via the phase velocity vP :
72
A.= vPI f
(2.58)
and the phase velocity vP is given in terms of the line parameters L, C as

v
1
= --
(2.59)
JLC
Because of (2.55), we can relate the wave number to the phase velocity:
(2.60)
0)
vP
Substituting the appropriate line parameters from Table 2-1, it is noticed that for all
three transmission line types the phase velocity is independent of frequency. The implication of this fact is as follows: If we assume a pulsed voltage signal propagating down
a line, we can decompose the pulse into its frequency components, and each frequency
component propagates with the same fixed phase velocity. Thus, the original pulse will
appear at a different location without having changed in shape. This phenomenon is
known as dispersion-free transmission. Unfortunately, in reality we always have to
take into account a certain degree of frequency dependence or dispersion of the phase
velocity that causes signal distortion.
2.9.3
Standing Waves
It is instructive to insert the reflection coefficient for a short-circuit line

= -1) into the voltage expression (2.56) and change to a new coordinated representation such that z = 0 in the old system coincides with the origin of the new coordinate system but extends in opposite, -z direction, as shown in Figure 2-24.
cro
~
'
Figure 2-24
I
I
d=l
Short-circuit transmission line and new coordinate system d.
Equation (2.56) now reads

V(d) = V\e+J~d - e-J~d)
(2.61)
f::ftrmlnated Louless Tranamlsslon Line
73
'
We notice that the bracket can be replaced by 2jsin(~d), and upon converting the pha801' expression back into the time domain, we obtain
v(d, t)
= Re{ V e
= Re{ 2jV+ sin(~d)e
1 001
}
1001
(2.62)
= 2 v+ sin(~d) cos ( rot+ 1t/ 2)
The sin-term ensures that the voltage maintains the short circuit condition for d = 0 at
all time instances t, see Figure 2-25. Because time and space are now decoupled, no
wave propagation, as discussed in Chapter 1, occurs. This phenomenon can physically
be explained by the fact that the incident wave is 180 out of phase with the reflected
wave, giving rise to fixed zero crossings of the wave at spatial locations 0, IJ2, A., 3/J2,
and so on.
0.8
rot
I/21t + 21tn
0.6
-r
~
0.4
0.2
-0.2
-0.4
-0.6
-0.8
-!~--~----~--~----~--~~--~--~
0.51t
1t
1.51t
21t
2.51t
31t
3.51t
~d
Figure 2-25
Standing wave pattern for various instances of time.
Introducing the new coordinate d into (2.56), this equation becomes
V(d) = V+e+J~d(l +f e- 1 2 ~d) = A(d)[l +f(d)]
(2.63)
where we set A (d) = V+ e+J~d and define a reflection coefficient

[(d) =
r oe-j2~d
(2.64)
74
valid anywhere along the length of the line d. The far-reaching implications of equation
(2.64) as part of the Smith Chart will be subject of Chapter 3. Similarly, the current in
the new spatial reference frame can be defined as
+
l(d) = .!::_e+j~d(l- r e-j 2 ~d) = A(d)[l- r(d)]
Zo
(2.65)
Zo
Under matched condition (r0 = 0) the reflection coefficient r(d) is zero, thus maintaining only a right-propagating wave. To quantify the degree of mismatch, it is customary to introduce the standing wave ratio (SWR) as the ratio of the maximum
voltage (or current) over the minimum voltage (or current) as follows:
SWR
= IVmaxi = IImax[
IVmini
(2.66)
jlmini
We note that the extreme values of (2.64) can only be + 1 and -1. Knowing that the
exponential function has a magnitude of 1, we find for (2.66) the form
lrol
1-lrol
1+
SWR =
(2.67)
which has a range of 1 ~ SWR < oo, as seen in Figure 2-26.

20~~--~--~~--~--~~--~--~~
18
16
14
~ 12
r/.l
10
8
6
4
2L-----0o
0.1 0.2 o.3 0.4 o.s o.6 o.7 o.8 o.9
1ro1
Figure 2-26 SWR as a function of load reflection coefficient
Ir ol .
75
. In many cases engineers use the term voltage standing wave ratio (VSWR)
110~w of SWR by defining it as the ratio of the maximum absolute voltage value to its
~...."" It is concluded from the definition (2.66) and from Figure 2-26 that the ideal
of matched termination yields an SWR of 1, whereas the worst case of either open
short-circuit termination results in SWR ~ oo. Strictly speaking, SWR can only be
to lossless lines, since it is impossible to define a SWR for lossy transmission
.....""~ This is because the magnitude of the voltage or current waves diminishes as a
IUilcuon of distance due to attenuation and thus invalidates (2.67), which, as a single
Cril1'tor is independent of where along the transmission line the measurement is
_ ...... Because most RF systems possess very low losses, (2.67) can be safely applied.
inspection of the exponent in (2.64) we see that the distance between the maxi1t or d = A./ 4 and the disand minimum of the reflection coefficient is 2 ~d
81JIIIIICC between two maxima is d = /.../ 2 .
.,...,.....loiU
1,
10 Special Termination Conditions

2.1 0.1 Input Impedance of Terminated Lossless Line
At a distance d away from the load the input impedance is given by the expression
v (d)
z in(d)
y+ ej~d (1 + roe-2j~d)
= l(d) = Z o y +ej~d(l-roe-2j~d)
(2.68)
~where
(2.63) and (2.65) are used for the voltage and current expres.sions. Equation
. (2.68) can be converted into the form
~ -
1 + r(d)
zin(d) = Zo1- r{d)
(2.69)
:and, upon using (2.52) to replace r 0 , we obtain
. .
'
t-..
~.
~
~ :'
..
'
l. .
(2.70)
Chapter 2 TranemJeelon Line Analyala
76
Division by the cosine term gives us the final form of the input impedance for the terminated transmission line:
ZL + jZ0 tan(J3d)
Zin(d)= Zozo + jZrtan(~d)
(2.71)
This important result allows us to predict how the load impedance Z L is transformed along a transmission line of characteristic impedance Z 0 and length d. It takes
into account the frequency of operation through the wave number (3. Depending on the
application, ~ can be expressed either in terms of frequency and phase velocity,
J3 = (21tf)lv P, or wavelength, J3 = 21tiA.
2.10.2 Short Circuit Transmission Line
If ZL = 0 (which means the load is represented by a short circuit) expression

(2. 71) simplifies to
(2.72)
Equation (2.72) can also directly be derived by the division of voltage through current
wave for the short circuit condition (r0 = -1 ):
V(d) = V+[e+j~d-e-j~d] = 2jV+sin(J3d)
(2.73)
and
+
V
-~d
-~d
2V
/(d) = -[e+J + e-1 ] = -cos(J3d)
Zo
Zo
(2.74)
so that Zin(d) = VII = jZ0 tan(J3d). A plot of voltage, current, and impedance as a
function of line length is shown in Figure 2-27.
It is interesting to note the periodic transitions of the impedance as the distance
from the load increases. If d = 0, the impedance is equal to the load impedance, which
is zero. For increasing d the impedance of the line is purely imaginary and increases in
magnitude. The positive sign of the impedance at this location shows that the line
exhibits inductive behavior. When d reaches a quarter-wave length, the impedance is
equal to infinity, which represents an open-circuit condition. Further increase in distance leads to negative imaginary impedance, which is equivalent to a capacitive behavior. At distanced = A-12 the impedance becomes zero and the entire periodic process
is repeated for d > A-I 2 .
From a practical point of view, it is difficult to conduct electric measurements at
various locations along the line, or alternatively by considering a multitude of lines of
different lengths. Much easier (for instance, through the use of a network analyzer) is
Special Termination Conditions
1.5
1
0.5
-0.5
-1
-1.5
-2 0
f~
Short
circuit
Figure 227
1 ~A
-r-1-vr--y-~ r-1-v-r- t
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
open
circuit
Short
circuit
Open
circuit
Short
circuit
Voltage, current, and impedance as a function of line length for a

short circuit termination.
the recording of the impedance as a function of frequency. In this case d is fixed, and
the frequency is swept over a specified range, as discussed in the following example.
----------------------------~~~
Example 2-6: Input impedance of a short-circuit transmission
line as a function of frequency
For a short-circuit transmission line of l = 10 em compute the magnitude of the input impedance when the frequency is swept from
f = 1 GHz to 4 GHz. Assume the line parameters are the same as the
ones given in Example 2.3 (i.e., L = 209.4 nH/m and C = 119.5
pF/m).
Solution:
Based on the line parameters Land C, the characteristic impedance is found to be Z 0 = JLI C = 41.86 .Q. Further, the
phase velocity is given by vP = I I JLC and is equal to
78
1.99x10 m/s. The input impedance of the transmission line

Z in(d = l) as a function of frequency can then be expressed in the
furm
(2.75)
The magnitude of the impedance is shown in Figure 2-28 for the frequency range of 1 GHz to 4 GHz. Again we notice the periodic
short- or open-circuit behavior of this line segment. In other words,
depending on the frequency, the line exhibits an open-circuit behavior (for instance at 1.5 GHz) or a short-circuit behavior (for instance
at 2 GHz).
500~--~~--~----~~--~-----=-----,
450
400
350
a~
3oo
.a
N 250
200
150
100
50
1.5
Figure 2-28
2.5
f,GHz
3.5
Magnitude of the input impedance for a 10 em long, short-circuit

transmission line as a function of frequency.
Practical measurements with a network analyzer permit the

recording of graphs as the one seen in Figure 2-28. Had we fixed the
frequency and varied the line length, we would have gotten an identical response.
lptelal Termination Conditions
79
2.1 0.3 Open-Circuit Transmission Line
If ZL -too the input impedance (2.71) simplifies to the expression

Zin(d) = -
JZotan;~d)
(2.76)
which can be directly derived when we divide the voltage (2.63) by the current wave
(2.65) for the open circuit condition ( r o== + 1 ):
2 V+cos (~d)
(2.77)
2v+
V+
/(d) = -[e
+j ~d- e -J ~d] = - 1- sin(J3d)
(2.78)
V(d) = V +[e +jf3d + e- Jf3d] =
and
Zo
Zo
so that Zin (d) = V I I = - jZ 0 cot ( ~d) . Plotting voltage, current, and impedance as a
function of line length is shown in Figure 2-29.
2
~~--~~~~~-.~~--~~
. ~--~~
z (d) j
vr)Zo~
1
1.5
Ul
'
Open
circuit
Short
circuit
....... -..... .. ...... . ......... . ;_ ............. "----- . .............. :' . .................. ...... .... .
~~
'
0.5
0
-0.5
- 1 ......... - ----.... ---------------.. -------- 1.5
Open
circuit
Figure 2-29
Short
circuit
Open
circuit
Voltage, current, and impedance as a function of line length for an

open-circuit termination.
It is again of interest to keep the length d fixed, and sweep the frequency over a specified range, as the next example illustrates.
80
------------------------Rf&MW4
Example 2-7: Input impedance of an open-circuit transmission
line as a function of frequency
For an open-circuit transmission line of l = 10 em, repeat the calculations of Example 2-6.
Solution:
All calculations remain the same, except that the input
impedance is changed to
2
Zi0 (d = 1) = -jZ0 cot(~l) = - jZ0 cot( ::'
(2.79)
z)
The magnitude of the impedance is displayed in Figure 2-30 for the

frequency range of 1 GHz to 4 GHz. The points where the cotangent
approaches infinity correspond to values where the argument
reaches 90, 180, 270, and so on. In reality, small losses due to
the presence of R and G tend to limit the amplitude to finite peaks.
The physical reason for these peaks is due to a phase shift between
500~--~----~----~----~----~--~
450
400
350
a~
-:a
3oo
N 250
200
150
100
50
01
Figure 2-30
1.5
2.5
J,GHz
3.5
Magnitude of impedance for a 10 em long, open-circuit transmission

line as a function of frequency.
lplclll Termlnlltlon Condlt1ona
81
voltage and current wave. Specifically, when the current wave

approaches zero and the voltage is finite, the line impedance
assumes a maximum. This is equivalent to the mechanical effect
where, for instance, a sound wave at particular discrete frequencies
(so-called eigen frequencies) forms standing waves between the
walls of a confining structure.
Figures 2-28 and 2-30 teach us that impedance matching to a
particular impedance value is only possible at a fixed frequency.
Any deviations can result in significantly different impedances.
2.1 0.4 Quarter-Wave Transmission Line
As evident from (2. 70), if the line is matched, ZL = Z 0 , we see that

Zin(d) = Z 0 regardless of the line length. We can also ask ourselves the question: Is it
possible to make the input impedance of the line equal to the load impedance
(Zin(d) = ZL )? The answer is found by setting d = A./ 2 (or more generally
d = A,/ 2 + m(A/2), m = 1, 2, . .. ), i.e.,
z in(d= A./2) =
Zo
ZL + jZ0
tan(T~)
(2 A)
ZL
(2.80)
Z 0 + jZLtan : 2.
In other words, if the line is exactly a half wavelength long, the input impedance is
equal to the load impedance, independent of the characteristic line impedance Z 0 .
As a next step, let us reduce the length to d = A.! 4 (or d = 'AI 4 + m(A./ 2),
m = 1, 2, . .. ). This yields
ZL + jZ0
Zi0 (d= A./ 4) = Z 0
tan( ~)
(
A)
2
Z 0 + jZLtan : 4.
(2.81)
The implication of (2.81) leads to the lambda-quarter transformer, which allows the
matching of a real load impedance to a desired real input impedance by choosing a
transmission line segment whose characteristic impedance can be computed as the geometric mean of load and input impedances:
82
Chapter 2 Transmission Line Analyal
(2.82)
This is shown in Figure 2-31, where Zin and Z L are known impedances and Z 0 is
determined based on (2.82).
z,.Desired
:J_
L
Figure 2-31
Z 0 =JZLZin
A./4
4:
:r
ZL
..
Input impedance matched to a load impedance through a 'A/4 line

segment Z 0 .
The idea of impedance matching has important practical design implications and
is investigated extensively in Chapter 8. In terms of a simple example we place the preceding formula in context with the reflection coefficient.
-----------------------------RF&JA~
Example 2-8: Impedance matching via a 'A/4 transformer
A transistor has an input impedance of Z L = 25 Q which is to be

matched to a 50 .Q microstrip line at an operating frequency of
500 MHz (see Figure 2-32). Find the length, width, and characteristic impedance of the quarter-wave parallel-plate line transformer for
which matching is achieved. The thickness of the dielectric is
d = 1 mm and the relative dielectric constant of the material is
r = 4 . Assume that the surface resistance R and shunt conductance G (see Table 2-1) can be neglected.
Solution:
We can directly apply (2.81) by using the given
impedances from the problem statement. For the line impedance we
find
Special Termination Conditions
Figure 232
83
Input impedance of quarter-length transformer.
On the other hand, the characteristic impedance of the parallel-plate

line is
= J LI C
Z line
= (dpl w ) JJl/e
Thus, the width of the line is
w-
dp~O = 5.329 mm
z line
EoEr
From Table 2-1 we find the values for capacitance and inductance of
the line:
= J..Ldpl w == 235.8 nH/m

C = ewl d P = 188.6pF/m
The line length l follows from the condition
'A
l == -=
4 4/JLC
== 74.967 mm
The input impedance of the combined transmission line and the load
is shown in Figure 2-32.
ZL + jZline tan(~d)
z in == zlinez line
+ jZLtan ( ~d)
1 + r(d)
= zline 1- r(d)
where d == l = ')J4 and the reflection coefficient is given by
r(d)
= r 0 e-2j~d
ZL- Zline exp ( -J'22rtfd)

Z L + zline
vp
The reflection coefficient is next inserted into the expression for

Z in. Plotting the impedance magnitude is shown in Figure 2-33.
We note that Z in is matched to the line impedance of 50 Q not
only at 500 MHz, but also at 1.5 GHz. Since the quarter-wave trans-
84
Chapter 2 Transmission Line Analysts
so~~--~~~--~--~~--~~~--~~
45
40
35
30
q 25
c:l
t--f 20
15
10
5
O0
Figure 2-33
0.2
0.4
0.6
0.8 I
1.2
J,GHz
1.4
1.6
1.8
Magnitude of Zin for frequency range of 0 to 2 GHz and fixed

length d.
former is designed to achieve matching only at 500 MHz for a particular line length l, we cannot expect matching to occur for
frequencies away from the 500 MHz point. In fact, for circuits
required to operate over a wide frequency band, this approach may
not be a suitable strategy.
The A.! 4 transformer plays an important role in many applications as an easy-to-build, narrowband matching circuit.
2.11 Sourced and Loaded Transmission Line

Thus far our discussion has only relied on the transmission line and its termination through a load impedance. In completing our investigation, we need to attach a
source to the line. This results in the added complication of not only having to deal with
an impedance mismatch between transmission line and load, but having to take into
considerations possible line-to-source mismatches as well.
SOurced and Loaded Transmission Line
85
2.11.1 Phasor Representation of Source
The generic transmission line circuit is shown in Figure 2-34 and involves a voltage source consisting of a generator voltage V 0 and source impedance Z 0 .
r tn.
Figure 2-34
Generic transmission line circuit involving source and load

terminations.
The input voltage recorded at the beginning of the transmission line can be written
in general form
yin
+
= yin+
yin
= yin( 1 + rin)
= VG ( z
)
z
in+ G
Z in
(2.83)
where the last expression follows from the voltage divider rule. The input reflection
coefficient rin is obtained by looking from the source into the transmission line of
length d = l:
r in = r(d= l) =
z.m - Z o
2 R.l
= roe-
zin +Zo
}p
(2.84)
In (2.84), r 0 is the load reflection coefficient as defined in (2.52). In addition, it is often

useful to introduce transmission coefficients, which take the form
(2.85)
at the beginning of the line, and
To = 1 + ro =
2ZL
zL+ z0
(2.86)
at the load end. The formal derivation of the transmission coefficient for a terminated
transmission line is presented in the following example.
Chapter 2 TranemlaaJon Line Analyele
----------------------------~&)A~
Example 2-9: Determination of transmission coefficient
Consider a transmission line aligned along the z-axis whose characteristic line impedance is Z 0 and has a load Z L at d = 0 . Derive
the transmission coefficient T 0
To the left of the load impedance (d > 0) we can write
for the voltage wave
Solution:
V(d) = V+ ( e+ j!}d + r e-j!}d)

0
and for the transmitted voltage at the load impedance (d = 0) we set

generically
+
V(d=O)=VT0
Since the voltage has to maintain continuity at d = 0, we obtain
l+r0 = T 0
from which we can find the transmission coefficient
T0 = 1 +
ZL -Z0
ZL+Zo
2ZL
ZL+Zo
The argument of matching incident voltage with transmitted voltage

wave can be applied to any discontinuity between two lines involving different characteristic impedances.
Reflection and transmission coefficients are easier to measure
at high frequency than impedances. They are therefore more commonly used to characterize an inteiface between two dissimilar
transmission line segments.
In addition to the preceding reflection and transmission coefficients, the connected

source introduces an additional difficulty. Since the voltage reflected from the load is
traveling toward the source, we need to consider a mismatch between the transmission
87
SOUrced and Loaded Transmlnlon Line
line and the source impedance. Accordingly, when looking from the line into the source
we can define the source reflection coefficient:
Zo-Zo
(2.87)
rs = - - -
Za+Zo
The output reflection coefficient shown in Figure 2-34 is then computed similar to
2 1
(2.84), but moving in opposite direction: rout = r se -) 1}
2.11.2 Power Considerations for a Transmission Line
From the definition of time-averaged power
= 21Re{ VI* }
P av
(2.88)
we can compute the total power at the beginning of the transmission line. To accomplish this task, the complex input voltage yin = v:(l + rin)
/in = ( v:IZ0 )( 1- rin) have to be inserted in (2.88). The result is
+
Pin = P;n +Pin=
1!v:n1
and current
2-z( 1 -Jr;nJ
0
(2.89)
We notice here again that, just like voltage and current, power is also treated as being
comprised of a positive and negative traveling wave.
Since V~ in (2.89) is not directly accessible, it is more useful tore-express (2.89)
in terms of the generator voltage V 0 as follows:
+
vin
vin
(2.90)
= 1 + r. =
m
where (2.83) is used. As already known from (2.69), the input impedance is rewritten
1 + rin
zin = Zol- r.
(2.91)
The generator impedance follows from (2.87) as
1 +rs
Zo
= Zol- rs
Inserting (2.91) and (2.92) into (2.90) yields, after some algebra,
(2.92)
88
Chapter 2 Tranamlulon Line Analyala
v in
V G (1- fs)
= T <1 in)
r sr
(2.93)
Using (2.93) in (2.89), the final expression for the input power is therefore
p. =
m
lv 12 II - r s12 (1 -lr 12)

!____Q_
8 Z 0 I1- rsr. 12
m
m
(2.94)
Upon using (2.84), we obtain the following expression for the input power for a lossless
line:
p. =
10
11- fsl
1-jr e-2j~zj2
z0 I1- rsroe-2j~[l2 < o
)
!IVal
8
(2.95)
Since the line is lossless, the power delivered to the load will be equal to the input
power. If source and load impedances both are matched to the transmission line impedance (implying r s = 0 and r 0 = 0 ), then (2.95) simplifies to
p. =
10
! IVol
8
z0
! IVcl
8
zG
(2.96)
which represents the power produced by the source under perfectly matched conditions
and which constitutes the maximum available power provided by the source. When the
load ZL is matched to the transmission line, but the source impedance Z 0 is mismatched, then part of the power will be reflected and only portion of the maximum
available power will be transmitted into the line at location d = l:
p. =
lD
! IVal211- rsl2
8
z0
(2.97)
For the case where both source and load impedances are mismatched, reflections will
occur on both sides of the transmission line and the power that will be delivered to the
load is defined by (2.95). Besides watts (W), the unit that is widely used to quantify
power in RF circuit design is dBm, which is defined as follows:
P[dBm] = lOlog P[W]
lmW
In other words, power is measured relative to 1 milliwatt.
(2.98)
...,_and Loaded Transmission Line
89
--------------------------~~~~
Example 2-10: Power considerations of transmission line
For the circuit shown in Figure 2-34, assume a lossless line with
Z0
75 .Q, ZG = 50 Q, and ZL
40 Q. Compute the input
power and power delivered to the load. Give your answer both in W
and dBm. Assume the length of the line to be A./2 with a source
voltage of V G = 5 V .
Solution:
Since the line is lossless, the power delivered to the

load is exactly the same as the input power. To find the input power,
we use expression (2.95). Because the length of the line is /..,/ 2, all
ex~onential terms in (2.95) are equal to unity; that Is,
e- j~l = e-Z j(Z1ti A)(A.IZ) = 1 and (2.95) can be rewritten as
~~.
'
Pin =
where
! lvGlz II - r s12 ( 1 - lrol2)

8 Zo II- rsrol2
the
reflection coefficient at the source end is

r s = (Za- Z0 )/(ZG + Z 0 ) = -0.2 and the reflection coefficient
at the load is r 0 = (ZL- Z 0 )/(ZL + Z 0 ) = -0.304. Substitution
of the obtained values into the preceding equation yields
PL
= P in = 61.7 mW
or
PL =Pin = 17.9 dBm .
Most RF data sheets and application notes specify the output

power in dBm. It is therefore important to gain a "feel" of the relative magnitudes of m Wand dBm.
The previous analysis is easy to extend to a lossy transmission line. Here we find
that the input power is no longer equal to the load power due to signal attenuation.
90
Chapter 2 Transmission Line Analyal
However, with reference to Figure 2-34 the power absorbed by the load can be
expressed similarly to (2.89) as
(2.99)
where the voltage 1v~l for a lossy transmission is 1v~l = 1v~l e -al, with a again being
the attenuation coefficient. Inserting (2.93) into (2.99) gives as the final expression
lv 12 /1 - r s12
= !_G_
P
L
e-2al(l-/fL/2)
(2.100)
8 Zo /1-fsfin/2
where all parameters are defined in terms of the source voltage and the reflection
coefficients.
2.11.3 Input Impedance Matching
Employing an electric equivalent circuit representation for the transmission line

configuration shown in Figure 2-34 allows us to examine optimal conditions for the
matching of the generator to the line.
z.
v
Figure 2-35
10
Equivalent lumped input network for a transmission line

configuration.
In a lumped parameter expression, and consistent with Figure 2-35, we can

express (2.95) as
2
2
v.* )} = ! jvG/
P = ! Re v. ~
z
in
in
2 { m ( z~
2Re{z;} ZG + zin
(2.101)
If we assume the generator impedance to be of fixed complex value Z 0 = R 0 + jX0 ,

we can find the conditions that have to be imposed on Zin to obtain maximum power
transfer into the transmission line. Treating Pin as a function of two independent variables Rin and X in, we find the maximum power value by taking the first derivatives of
Pin with respect to Rin and Xin and setting the values to zero:
91
Loaded Transmission Line
aPin _ aPin _
----0
(;)Rin
(2.102)
axin
two conditions that result are
Rb- R~ + (Xb + 2X0 Xin +X~)
= 0
Xin(X0 + Xin) = 0
(2.103a)
(2.103b)
(2.103b) gives Xin = -X0 and, upon substituting this result into (2.103a),
Rin = R0 . This derivation shows that optimal power transfer requires conjugate
1Dt111ex matching of the transmission line to the generator impedance:
(2.104)
IUJ.VW.~a~.
this is done for the case of generator to input impedance matching, an identican be carried out to match the output impedance to the load impedance.
we will find that the impedances require conjugate complex matching for maxipower transfer:
= Zl
represents the impedance looking into the transmission line from the load
Zout
Zout
2.11.4 Return Loss and Insertion Loss
. Practical circuit realizations always suffer a certain degree of mismatch between

'llllwiUoV.lv source power and power delivered to the transmission line; that is, r in in
is not zero. This mismatch is customarily defined as return loss (RL), which is
Jatio of reflected power, Pr = P~ , to incident power, Pi = P~ , or
RL =
-lOlog(~~) = -10logJrinl 2 = -201ogJr1nl
(2.105a)
(2.105b)
RL = -lnlrinl
equation (2.105a) specifies the return loss in decibel (dB) based on the logarithm
base 10, whereas (2.105b) specifies RL in Nepers (Np) based on the naturallogaA conversion between Np and dB is accomplished by noting that
RL = -20loglrinl = -20(lnlrinl>/(ln10) = -(201oge)Inlrinl
(2.106)
Chapter 2 Tranemleelon Line An..ytll
Therefore, 1 Np = 20loge = 8.686 dB. As seen from (2.106), if the line is matched
rin ~ 0 , then RL ~ oo
---------------------------~&)A~
Example 211: Return loss of transmission line section
For the circuit in Figure 2-35 a return loss of 20 dB is measured.
Assuming real impedance values only, what is the source resistance
R0 if the transmission line has a characteristic line impedance of
Rin = 50 n? Is the answer unique?
Solution:
The reflection coefficient is found from (2.1 05a) as
Jrinl = 10- RL/20 = 10-20/20 = 0.1

The source resistance is now computed by using (2.91):
RG =
1 +fin
(11-0.1
+0.1)
Rin1- fin = 50
.Q
= 61.1 Q
In the preceding calculations, we assumed that the reflection coefficient rin is positive and therefore is equal to its absolute value.
However, it can also be negative, and in that case the source resistance would be
1 + rin
(1- 0.1)
RG = Rin 1 _fin = 50 1 + 0. 1 Q = 40.9 Q
The return loss, which can be recorded with a network analyzer, provides immediate access to the reflection coefficient and
thus the degree of impedance mismatch between the transmission
line and generator.
In addition to the return loss, which involves the reflected power, it is useful to introduce the insertion loss (IL) defined as a ratio of transmitted power P 1 to incident
power Pi . In practice insertion loss is measured in dB according to the following
formula:
lummary
93
p
P.-P
2
IL = -lOlog p~ = -10log 'p. ' = -10log(l-lrinl )
I
(2.107)
The meaning of (2.107) in circuit design is straightforward. As the name implies, if an

unmatched circuit is connected to an RF source, reflections occur that result in a loss of
power delivered to the circuit. For instance, if the circuit represents an open- or shortcircuit condition, the insertion loss reaches a maximum ( IL ~ oo ). Altemativelyt if the
circuit is matched to the source, all power is transferred to the circuit, and the insertion
loss becomes a minimum (IL = 0 ).
2.12 Summary
In this chapter a detailed description is given of the fundamental concepts of distributed circuit theory. The topic is motivated by the fact that when the wavelengths of
the voltage and current waves shrink to roughly l 0 times the size of the circuit compollentst a transition must be made from lumped element analysis, based on Kirchhoff's
current and voltage laws, to distributed theory according to wave principles. This transi_\ion from low- to high-frequency circuit analysis may not be as clear-cut as the definiless than or equal to I 0 'A implies; in fact, a considerable "gray area" does exist.
JaD~eies:s, starting at a particular frequency a transition is needed to obtain meaningresults.
The underlying concepts of distributed theory can best be understood by developequivalent circuit representation (Section 2.3) of a microscopic section of the
"""'.,.,~."'u line. The required circuit parameters per unit length R, L, G, C are
directly from Table 2-1 for three common transmission line types (Section
~:...-A......,..... going into much theoretical detail. However, for the readers who are inter. in how the parameters can be found, Section 2.4 introduces the necessary tools of
and Ampere's laws, followed by Section 2.5, which derives all four circuit
melters for the parallel-plate transmission line .
. In either case, the knowledge of the circuit parameters ultimately leads to the
~~................ line impedance of a generic transmission line system:
(R
+ jroL)
(G + jroC)
this representation the input impedance of a terminated transmission line is devel. The result is perhaps one of the single most important RF equations:
ZL + jZ 0 tan(~d)
zin(d)= Zozo + jZLtan(~d)
Chapter 2 Tranamlulon Line Anelylll
The application of this equation for the special cases of open, short, and matched load
impedances are investigated in terms of their spatial and frequency domain behaviors.
Furthermore, the lambda-quarter or quarter-wave transformer is introduced as a way of
matching a load impedance to a desired input impedance.
As an alternative to the input impedance equation, it is often very useful to represent the line impedance in terms of the reflection coefficients at load and source end:
It is found that the reflection coefficient is spatially dependent, as shown by
r(d)
= roe -j
~d
The reflection coefficient concept allows concise expressions for power flow considerations. Similar to the input impedance we found the input power
p.
10
= !IV al
8
II- rsl (l -jr e-2JPII 2 )

z 0 I1- rsroe -2JPLI2
This equation permits the investigation of various matching or mismatching conditions

at the load/source side. Chapter 2 concludes with a brief discussion of insertion loss and
return loss.
Further Reading
R. Collin, Foundations ofMicrowave Engineering, McGraw-Hill, New York, 1966.

G. Gonzales, Microwave Transistor Amplifiers, Analysis an.d Design, 2nd edition, Prentice Hall, Upper Saddle River, New Jersey, 1997.
H. A. Haus and J. R. Melcher, Electromagnetic Fields and Energy, Prentice Hall,
Englewood Cliffs, NJ, 1989.
M. F. lskander, Electromagnetic Fields and Waves, Prentice Hall, Upper Saddle River,
New Jersey, 1992.
C. T. A. Johnk, Engineering Electromagnetic Fields and Waves, 2nd ed., John Wiley &
Sons, New York, 1989.
J. A. Kong, Electromagnetic Wave Theory, 2nd ed., John Wiley & Sons, New York,
1996.
95
Probltrne
S. Y. Liao, Engineering Applications of Electromagnetic Theory, West Publishing Company, St. Paut MN, 1988.
D. M. Pozar, Microwave Engineering, 2nd ed., John Wiley & Sons, New York, 1998.
P. A. Rizzi, Microwave Engineering, Passive Circuits, Prentice Hall, Englewood Cliffs,
New Jersey, 1988.
H. Sobol, "Applications of Integrated Circuit Technology to Microwave Frequencies,"
Proceedings of the IEEE, August 1971.
D. H. Staelin, A. W. Morgenthaler, and J. A. Kong, Electromagnetic Waves, Prentice

Hall, Upper Saddle River, New Jersey, 1994.
Problems
2.1
To estimate the effective relative pennittivity e, of a dielectric material used

in a transmission line, you decide to measure the voltage distribution along
the line using a similar setup as depicted in Figure 2-2. Your measurements
at 1 GHz excitation frequency have shown that the wavelength of the signal
in the cable is equal to 10m. Using this infonnation, compute the effective
relative permittivity of the materiaL Discuss how this experimental setup
could be used to measure the attenuation factor a .
2.2
As discussed in this chapter, a single signal trace on a printed circuit board

(PCB) can be treated as a transmission line and can be modeled using an
equivalent circuit, shown in Figure 2-12. Nevertheless, when the size of the
PCB gets smaller, the distance between the traces decreases and they can no
longer be treated as separate transmission lines. Therefore, the interaction
between them has to be accounted for. Using the configuration shown in
Figure 2-7, suggest a new equivalent circuit that takes into account interaction between two signal traces.
2.3
In Example 2-1 we showed how to compute the magnetic field distribution

produced by the wire carrying current /. Repeat your computations for a
system consisting of two parallel wires each of radius 5 mm and carrying a
current of 5 A in the same direction. Plot the field distribution of the magnetic field H(r) as a function of distance r starting at the center-line position between the two wires.
2.4
Consider a system consisting of a circular loop of radius r = 1 em of thin

wire (assume the radius of the wire to be equal to zero) and carrying a con-
Chapter 2 Transmission line Analysts
stant current I = 5 A . Compute the magnetic field along the center line of
the loop as a function of distance h from the center of the loop.
2.5
Find kr and ki in tenus of L, C, G, R, and ro in equation (2.32).
2.6
In the text we have derived the transmission line parameters (R, L, G, and C)
for a parallel-plate line. Derive these parameters for a two-wire configuration, see Figure 2-4. Assume that D >> a.
2.7
Repeat Problem 2.6 for a coaxial cable, see Figure 2-5.
2.8
An RG6AIU cable has a characteristic impedance of 75 .Q. The capacitance

of a 0.5 m long cable is measured and the value is found to be 33.6 pF. What
is the cable inductance per unit length, if the cable is lossless?
2.9
Assuming that dielectric and conductor losses in a transmission line are

small (i.e. G roC and R roL ), show that propagation constant k can be
written as
= a+ jj3
~(: + Gz0)
+ jroJLc
where Z 0 = )LI C is the characteristic impedance of the line m the

absence of loss.
2.10
Using the results from the previous problem and the transmission line
parameters given in Table 2-1,

(a) show that the attenuation constant in a coaxial cable with small losses is
{ 1 ( 1 1) crdie! ~
2crcond~~~ln(b/ a) a+ b + 2 ~e
=1
where cr die! and cr cond are the conductivities of the dielectric material
and the conductors, respectively.
(b) show that the attenuation in this case is minimized for conductor radii
such that x lnx = 1 + x , where x = b I a .
(c) show that for a coaxial cable with dielectric constant er = 1 the condition of minimum losses results in the characteristic impedance of
Z 0 = 76.7Q.
2.11
Compute the transmission-line parameters for a coaxial cable, which characteristics are listed as follows:
r.
.<.------------------------------1:1robltml
97
~.
:..
Inner Conductor: Copper

6
a = 0.5mm, O'cu = 64.516x10 S/m
Dielectric: Polyethylene
-14
b = 1.5mm, O'Poly = 10 S/m
Outer Conductor: Copper
6
t = 0.5mm, O'cu = 64.516x10 S/m
2.12 An RG58AIU cable has a characteristic line impedance of 50 0 . The measurements performed on a section of this cable produce the following results
capacitance of 1 meter of cable: 101 pF

"
phase velocity: 66% of speed of light
attenuation at 1 GHz: 0. 705 dB/m
outer diameter of the insulation layer: 2.95 mm
center conductor is made out of AWG 20 copper wue,
6
O'cu = 64.516x10 S/m
14
dielectric layer is made out of polyethylene, O'Poly = 10- S/m
From this list of information, find the following quantities:
(a) inductance L per unit length of the cable assuming that cable is lossless
(b) relative permeability r of the dielectric material
!):
(c) resistance R per unit length of the cable at operating frequency of
t.
1 GHz (Hint: use the formula for the attenuation constant derived in
Problem 2.10)
(d) conductance G of the dielectric per unit length
2.13 Using the coaxial cable from the previous problem, compute its characteristic impedance. Plot the frequency behavior of the real and imaginary components of the characteristic impedance. Is the result what you expected to see?
Explain any discrepancies.
:
'
.~.~:
< .
)'",
= =
2.14 A distortionless transmission line results if R

G
0, which results in
k = jwJLC = a+ j~, or a = 0 and ~ = wlvP with the phase velocity
independent of frequency [i.e., vP = 1I ( Ji) ]. A signal propagating along
this transmission line will not suffer any pulse distortion or attenuation. If we
allow R :; G 0 , find the condition for which a = JifG and ~ = oo JLC.
In other words, the line is attenuative but remains distortionless.
215
. It is desired to construct a 50 n microstrip line. The relative dielectric constant is 2.23 and the board height is h = 0.787 mm. Find the width, wavelength, and effective dielectric constant when the thickness of the copper
trace is negligible. Assume an operating frequency of 1 GHz.
98
Chapter 2 TranmiMion Line Analyll
2.16
Starting with basic definition for the standing wave ratio (SWR)
SWR =
IVmaxi
IVmini
IImaxi
IImini
show that it can be re-expressed as
lfol
1-lrol
SWR = 1 +
2.17
The characteristic impedance of a coax cable is 50 .Q and assumed lossless.

If the load is a short circuit, find the input impedance if the cable is 2 wavelength, 0.75 wavelength, and 0.5 wavelength in length.
2.18
An experiment similar to the one shown in Figure 2-2 is performed with the
following results: The distance between successive voltage minima is
2.1 em; the distance of the first voltage minimum from the load impedance
is 0.9 em; the SWR of the load is 2.5. If Z 0 = 50 .Q, find the load impedance.
2.19
In this chapter we have derived the equation for the input impedance of the
loaded lossless line, (2.65). Using the same approach, show that for a loaded
lossy transmission line (i.e., R "# 0, G ' 0) the input impedance is
ZL + Z 0 tanh(yd)
zin(d)= Zozo + ZLtanh(yd)
where 'Y is the complex propagation constant and tanh denotes the hyperbolic tangent
X
-X
e -e
tanh(x) = X
-X
e +e
2.20
Using the result from the previous problem, compute the input impedance of
a 10 em long lossy coaxial cable connected to a ZL = (45 + j5) 0 load
impedance. The system is operated at I GHz frequency, and the coaxial
cable has the following parameters: R = 123(!-LO/m), L = 123(nH/m),
G = 123{J.LS/m), and C = 123(pF/m).
2.21
Show that the input impedance of a lossless transmission line repeats itself
every half wavelength [i.e., zin(ld) = zin{ld+m(A./2)} ], where ld is an
arbitrary length and m is an integer 0, 1, 2, ...
:t.Nblems
99
2.22 A radio transmitter is capable of producing 3 W output power. The transmitter is connected to an antenna having a characteristic impedance of 75 0.
The connection is made using a lossless coaxial cable with 50 Q characteristic impedance. Calculate the power delivered to the antenna if the source
impedance is 45 n and the cable length is 11 A..
2.23 For an RF circuit project an open-circuit impedance has to be created with a
75
rnicrostrip line placed on a circuit board with relative dielectric constant of 10 and operated at 1.96 GHz. The line is terminated with a short circuit on one side. To what length does the line have to be cut to measure an
infinite impedance on the other side?
2.24 A short-circuited microstrip line of Z 0 = 85 n and (3/ 4 )A. in length is

used as a lumped circuit element. What is the input impedance if the line is
assumed lossless?
2.25 For the following system, compute the input power, power delivered to the
load, and insertion loss. Assume that all transmission lines are lossless.
Zr
400
2.26 Repeat Problem 2.25 for a 50 !l load impedance.

2.27 The complex load impedance Zr = (75- j50)0 is attached to a lossless
transmission line of 100
characteristic impedance. The frequency is
selected such that the wavelength is 30 em for a 50 em long line. Find (a) the
input impedance, (b) the impedance looking toward the load 10 em away
from the load, and (c) the voltage reflection coefficient at the load and 10 em
away from the load.
2.28 A 100 Q microstrip line is connected to a 75 Q line. Determine r, SWR,

percentage power reflected, return loss, percentage power transmitted, and
insertion loss.
2.29 A 50 Q transmission line is matched to a source and feeds a load of
ZL = 75 n. If the line is 3.4A. long and has an attenuation constant
100
Chapter 2 Tranmlaalon Une Analy
a = 0.5 dB/A., find the power that is (a) delivered by the source, (b) lost in
the line, and (c) delivered to the load. The amplitude of the signal produced
by the source is 10 V.
2.30 A measurement technique is proposed to determine the characteristic line

impedance of a coaxial cable via the determination of open, Z~ , and short
circuit, zrg input impedances with a network analyzer. It is assumed that the
line impedance is real. How does one have to process these impedances to
obtain Z 0 ?
2.31
A signal generator is used to feed two loads, as shown in the following

figure.
Z0 = 50 .Q
I.
0.35A ..j
z,,2
50
.Q
Find the both the power produced by the source and the power delivered to
each load.
2.32 A lossless 50 .Q microstrip line is terminated into a load with an admittance
of 0.05 mS. (a) what additional impedance has to placed in parallel with the
load to assure an input impedance of 50 n? (b) If the input voltage is 10 V,
find the voltage, current, and power absorbed by the combined load.
2.33
Show that return loss and insertion loss can be expressed in terms of the
voltage standing wave ratio SWR as
RL
= 201ogSWR
SWR + 1l and IL = 201og SWR + l
-
2JSWR
CHAPTER
!rfhe Smith Chart
transrmsswn line changes its impedance

~nding on material properties and geometric dimensions. Typical practical realiza;Jions include microstrip line, coaxial cable, and parallel-plate line. In addition, both the
length and operating frequency of the transmission line significantly influence the input
;impedance. In the previous chapter we derived the fundamental equation describing the
'input impedance of a terminated transmission line. We found that this equation incorpotates the characteristic line impedance, load impedance, and, through the argument of
1he tangent function, line length and operating frequency. As we saw in Section 2.9, the
_input impedance can equivalently be evaluated by using the spatially dependent reflection coefficient. To facilitate the evaluation of the reflection coefficient, P. H. Smith
developed a graphical procedure based on conformal mapping principles. This
approach permits an easy and intuitive display of the reflection coefficient as well as the
tine impedance in one single graph. Although this graphical procedure, nowadays
known as the Smith Chart, was developed in the 1930s prior to the computer age, it has
retained its popularity and today can be found in every data book describing passive
and active RF/MW components and systems. Almost all computer-aided design programs utilize the Smith Chart for the analysis of circuit impedances, design of matching
networks, and computations of noise figures, gain, and stability circles. Even instruJnents such as the ubiquitous network analyzer have the option to represent certain
measurements in a Smith Chart format.
This chapter reviews the steps necessary to convert the input impedance in its
standard complex plane into a suitable complex reflection coefficient representation via
a specific conformal transformation originally proposed by Smith. The graphical dis-
101
Chapter 3 The Smith Chart
102
play of the reflection coefficient in this new complex plane can then be utilized directly
to find the input impedance of the transmission line. Moreover, the Smith Chart facilitates evaluation of more complicated circuit configurations, which will be employed in
subsequent chapters to build filters and matching networks for active devices.
The following sections present a step-by-step derivation of the Smith Chart followed by several examples of how to use this graphical design tool in computing the
impedance of passive circuits.
3.1 From Reflection Coefficient to Load Impedance

In Section 2.9 the reflection coefficient is defined as the ratio of reflected voltage
wave to incident voltage wave at a certain fixed spatial location along the transmission
line. Of particular interest is the reflection coefficient at the load location d == 0 . From a
physical point of view this coefficient r 0 describes the mismatch in impedance between
the characteristic line impedance Z 0 and the load impedance Z L as expressed by (2.52).
In moving away from the load in the positive d-direction toward the beginning of the
transmission line, we have to multiply r 0 by the exponential factor exp(- j2~d), as
seen in (2.64), to obtain f (d) . It is this transfonnation from r 0 to r(d) that constitutes
one of the key ingredients in the Smith Chart as a graphical design tool.
3.1 .1
Reflection Coefficient in Phasor Form
The representation of the reflection coefficient

complex notation.
ZL- Zo
ro = z
+z =
0
r0
can be cast in the following
JaL
ror + ]fo; = lrol e
(3 .1)
where eL = tan- (f0 ;1r 0 r). We recall that pure short- and open-circuit conditions in
(3 .1) correspond to 0 values of - 1 and +I , located on the real axis in the complex
plane.
r-
--------------------------~~&uM~
Example 3-1: Reflection coefficient representations
A transmission line with a characteristic line impedance of
Z 0 = 50 Q is terminated into the following load impedances:
(a) Z L = 0 (short circuit)
103
From Reflection Coefficient to Load Impedance
(b) ZL ~ oo (open circuit)

(c) ZL = 50 Q
(d) ZL = (16.67- j16.67) Q
(e) ZL = (50+ jl50) Q
Find the individual reflection coefficients
the complex r -plane.
r0
and display them in
Solution:
Based on (3.1) we compute the following numbers for

the reflection coefficients:
(a) r 0 = -l (short circuit)
(b) r 0 = 1 (open circuit)
(c) ro = 0 (matchedcircuit)
(d)f0 = 0.54L221
(e) r 0 = 0.83L34 o
The values are displayed in polar form in Figure 3-1.
90
... 0.8 ..
..
..
...
.-..-
. '.
'
.
. . . . . .: . . . . .
. ...
';,
...
r0 = 0.54 L221
:
240
.. ..
270
Figure 3-1
Complex
r -plane and various locations of r 0
The reflection coefficient is represented in phasor form as done

when dealing with the conventional voltages and currents in basic
circuit theory.
104
3.1.2
Normalized Impedance Equation
Let us return to our general input impedance expression (2.69), into which we
substitute the reflection coefficient
(3.2)
This results in
(3.3)
In order to generalize the subsequent derivations, we normalize (3 .3) with respect to the
characteristic line impedance as follows
1+
r, + jfj
1 - r r - 1r.t
(3.4)
The preceding equation represents a mapping from one complex plane, the zin -plane,
to a second complex plane, the r -plane. Multiplying numerator and denominator of
(3.4) by the complex conjugate of the denominator allows us to isolate real and imaginary parts of zin in terms of the reflection coefficient. This means
Z.
10
= r + }.X =
1- r
r~ + 2jr.
I
c1- r,)
(3 .5)
+ r;
can be separated into

2
1- r r
- r.2
1
o-r,) + r,
(3.6)
and
X=
2f.l
2
(I-f,) + ri
(3.7)
Equations (3 .6) and (3. 7) are explicit transformation rules of finding zin if the reflection
coefficient is specified in terms of r r and r i Therefore, the mapping from the complex r -plane into the zin -plane is straightforward, as the following example underscores.
105
----------------------------~~~
Example 3-2: Input impedance of a terminated transmission
line
A load impedance Z L = ( 30 + j60) Q is connected to a 50 Q
transmission line of 2 em length and operated at 2 GHz. Use the
reflection coefficient concept and find the input impedance Zin
under the assumption that the phase velocity is 50% of the speed of
light.
Solution:
We first determine the load reflection coefficient
r 0 = ZzLL-+ Zo
Z
= 30 + j60- 50
30 + j60 + 50
=0 2
06
+1
= ~
j71.56
215
"'/,/ .J e
(3.8)
Next we compute f(d = 2cm) based on the fact that
~ = 2 1t = 21tf
A
vP
This results in
cient
2~d
= 21tf = 83.77 m -I
0.5c
= 191.99 and yields for the refection coeffi-
r = r 0 e-j 2 ~d = r r + jri = - o.32- jo.ss = Me -j 120.43
Having thus determined the reflection coefficient, we can now

directly find the corresponding input impedance:
Zin = Z 0
!: ~ =
R + jX
= 14.7- j26.7 Q
We note that the reflection coefficient phasor foml at the load,

r 0' is multiplied with a rotator that incorporates twice the electric
line length ~d. This mathematical statement thus conveys the idea
that voltage/current waves have to travel to the load and return back
to the source to define the input impedance.
Example 3.2 could have been solved just as efficiently by using the impedance
equation (2.65) developed in Section 2.9.
106
3.1.3
Parametric Reflection Coefficient Equation
The goal of our investigation is to pursue a different approach toward computing

the input impedance. This new approach involves the inversion of (3.6) and (3.7). In
other words, we ask ourselves how a point in the zin -domain, expressed through its normalized real, r, and imaginary, x, components, is mapped into the complex r -plane,
where it then can be expressed in terms of thereat r r' and imaginary, ri' components
of the reflection coefficient. Since r appears in the numerator and denominator, we
have to suspect that straight lines in the impedance plane Zin may not be mapped into
straight lines in the r -plane. All we can say at this point is that the matching of the load
impedance to the transmission line impedance Zin = Z 0 , or Z;n = 1, results in a zero
reflection coefficient (i.e., r r = ri = 0) located in the center of the r -plane.
The inversion of (3.6) is accomplished by going through the following basic algebraic operations:
(3.9a)
(3.9b)
r 2r -
~r + r~ = 1 -
(3.9c)
rr
can be written as a complete bino-
r+l r
r
r+l
At this point the trick consists in recognizing that

mial expression (see also Appendix C)
r2
+ r~ (r - -'-)2r+l
(r+I)2
1
1-r
r+I
(3.9d)
This finally can be cast in the form
+ r~
(r - -'-)2
r+I
r
(-~-)2
r+I
(3.10)
In an identical way as done previously, we proceed to invert (3.7). The result for
the normalized reactance is
(r,-d+(r;-D2 = GY
(3.11)
Both (3 .10) and (3 .11) are parametric equations of circles in the complex r -plane that
2
2
2
can be written in the generic form ( r r - a) + ( r; - b) = c . Here a, b denote shifts
along the real and imaginary r axes, and cis the radius of the circle.
107
Figure 3-2 depicts the parametric circle equations of (3.10) for various resistances. For example, if the normalized resistance r is zero, the circle is centered at the
origin and possesses a radius of 1, since (3 .1 0) reduces to
+
= 1 . For r = 1 we
2
find (f,-1/2) +
= (1 1 2/, which represents a circle of radius 1/2 shifted in the
positive r, direction by 1/ 2 units. We conclude that as r increases, the radii of the circles are continually reduced and shifted further to the right toward the point 1 on the
real axis. In the limit for r ~ oo we see that the shift converges to the point
2
r l (r + 1) ~ 1 and the circle radius approaches 1/(r + 1 ) ~ 0 .
It is important to realize that this mapping transforms fixed values of r only and
does not involve x. Thus, for a fixed ran infinite range of reactance values x, as indicated by the straight lines in the z-plane, maps onto the same resistance circle. The
mapping involving r alone is therefore not a unique point-to-point correspondence.
r; r;
r;
z-plane
r-plane
Constant resistance lines (r
Figure 32
= const)
Parametric representation of the normalized resistance r in the

complex r -plane.
A different graphical display results for the circle equation (3.11), which involves
the normalized reactance. Here the centers of the circles reside all along a line perpen2
dicular to the r r = I point. For instance, for X = 00 we note that (fr - I ) + r: = 0 '
which is a circle of zero radius, or a point located at r r = 1 and r j = 0 . For X = 1
2
2
we see that the circle equation becomes (r r - 1) + (ri- I ) = 1 . As x --7 0 the radii
and shifts along the positive imaginary axis approach infinity. Interestingly, the shifts
108
can also be along the negative imaginary axis. Here for x = -1 we notice that the cir2
2
cle equation becomes ( r r - 1) + ( r i + 1) = 1 with the center located at r ,= 1 and
r, = -1 . We observe that negative x-values refer to capacitive impedances residing in
the lower half of the r -plane. Figure 3-3 shows the parametric form of the normalized
imaginary impedance. For better readability the circles are displayed inside the unit circle only. In contrast to Figure 3-2 we notice that fixed x-values are mapped into circles
in the r -plane for arbitrary resistance values 0 ~ r < oo, as indicated by the straight
lines in the impedance plane.
The transformations (3.10) and (3.11 ) taken individually do not constitute unique
mappings from the normalized impedance into the reflection coefficient plane. In other
words, impedance points mapped into the r -plane by either (3 .10) or (3 .11) cannot
uniquely be inverted back into the original impedance points. However, since the transformations complement each other, a unique mapping can be constructed by combining
both transformations, as discussed in the next section.
X
3 - - - ------ - ----- -- -
- l/3
- 3 - - -- ----- - - - - - - ---1
z-plane (r > 0)
Constant reactance lines (x
Figure 3-3
3.1.4
1-plane
= const)
Parametric representation of the normalized reactance x in the

complex r -plane.
Graphical Representation
Combining the parametric representations for normalized resistance and reactance

circles (i.e., Figures 3-2 and 3-3) for lr! ~ 1 results in the Smith Chart as illustrated in
,._Won Coefficient to Load Impedance
109
'Figure 3-4. An important observation of the Smith Chart is that there is a one-to-one
mapping between the normalized impedance plane and the reflection coefficient plane.
file notice also that the normalized resistance circles r have a range 0 ~ r < oo and the
normalized reactance circles x can represent either negative (i.e., capacitive) or positive
(i.e., inductive) values in the range -oo < x < +oo
It should be pointed out that the reflection coefficient does not have to satisfy
111 S I. Negative resistances, encountered for instance as part of the oscillation condition for resonators, lead to the case In > 1 and consequently map to points residing
eutside the unit circle. Graphical displays where the reflection coefficient is greater than
.1 are known as compressed Smith Charts. These charts, however, play a rather limited
role in RFIMW engineering designs and are therefore not further pursued in this text.
1be interested reader may consult specialized literature (see the Hewlett-Packard application note listed at the end of this chapter).
If
---~x_=_+l
r=
r +jx- 1
r+ jx + 1
r,
z-plane
=-1
r-p]ane
Figure 34 Smith Chart representation by combining rand
x circles for lrl ~
1.
In Figure 3-4 we must note that the angle of rotation 2~d introduced b~ the length
1
of the transmission line is measured from the phasor location of r 0 = 1rol e L in clockwise (mathematically negative) direction due to the negative exponent ( -2jJ3d) in the
reflection coefficient expression (3.2). For the computation of the input impedance of a
terminated transmission line, the motion is thus always away from the load impedance
toward the generator. This rotation is indicated by an arrow on the periphery of the
chart. We further observe that a complete revolution around the unit circle requires
110
Chapter 3 Tbe Smith Chart
21t
2f3d = 2Td = 21t
where d = "A./2 or 180. The quantity f3d is sometimes referred
length of the line.
Lo
as the electrical
3.2 Impedance Transformation

3.2.1
Impedance Transformation for General Load
The determination of the impedance response of a high-frequency circuit is often

a critical issue for the RF design engineer. Without detailed knowledge of the impedance behavior, RFIMW system performance cannot adequately be predicted. In this
section we will elaborate on how the impedance can be determined easily and efficiently with the aid of the previously introduced Smith Chart.
A typical Smith Chart computation involving a load impedance Z L connected to a
transmission line of characteristic line impedance Z 0 and length d proceeds according
to the following six steps:
1. Normalize the load impedance ZL with respect to the line impedance Z 0 to deterIlllne zL.
2. Locate zL in the Smith Chart.
3. Identify the corresponding load reflection coefficient r 0 in the Smith Chart both
in tenns of its magnitude and phase.
4. Rotate r 0 by twice its electrical length t}d to obtain rin (d) .
5. Record the normalized input impedance zin at this spatial location d.
6. Convert Zin into the actual impedance Z in .
Example 3-3 goes through these steps, which are the standard procedure to arrive at the
graphical impedance solution.
----------------------------RF~~
Example 3-3: Transmission Hne input impedance determination with the Smith Chart
Solve Example 3-2 by following the six-step Smith Chart computations given in the preceding list.
llnpedlnce Transfonnatlon
111
with the load impedance

Solution:
We commence
ZL = (30 + j60) Q and proceed according to the previously outlined steps:
1. The normalized load impedance is
ZL
= (30 + j60)/50 = 0.6 + jl.2
2. This point can be identified in the Smith Chart as the intersection of the circle of constant resistance r = 0.6 with the circle of
constant reactance x = 1.2, as seen in Figure 3-5.
3. The straight line connecting the origin to point zL determines

the load reflection coefficient r 0 . The associated angle is recorded
with respect to the positive real axis.
4. Keeping in mind that the outside circle on the Smith Chart corresponds to the unity reflection coefficient ( ol = I), we can find
its magnitude as the length of the vector connecting the origin to zL.
Rotating this vector by twice the electrical length of the line (i.e.,
2 x ~d = 2 x 96 = 192 ) yields the input reflection coefficient
1r
rin.
5. Tills point uniquely identifies the associated normalized input

impedance zin = 0.3 - j0.53 .
6. The preceding normalized impedance can be converted back
into actual input impedance values by multiplying it by
Z0 = 50 Q, resulting in the final solution: Zin = (15- j26.5 )Q.
We recall that the exact value of the input impedance obtained
in Example 3-2 is ( 14.7- j26.7) Q. The small discrepancy is
understandable because of the approximate processing of the graphical data in the Smith Chart. The entire sequence of steps leading to
the detennination of the input impedance of the line connected to
the load is shown in Figure 3-5.
112
Figure 3-5 Usage of the Smith Chart to determine the input impedance for
Example 33.
These steps appear at first cumbersome and prone to error if

carried out by hand. However, using mathematical spreadsheets and
relying on modem computer-based instrumentation, the calculations are routinely done in seconds and with a high degree of
accuracy.
tlpldlnce Transformation
3.2.2
113
Standing Wave Ratio
From the basic definition of the SWR in Section 2.8.3 it follows that for an arbitrary distance d along the transmission line, the standing wave ratio is written
SWR(d) = 1 + lf(d)l
(3.12)
1 -lf(d)l
where f(d)
= r 0 exp(-j2~d). Equation (3.12) can be inverted to give

lr(d)l =
SWR - l
SWR+ I
(3.13)
This fonn of the reflection coefficient permits the representation of the SWR as circles
in the Smith Chart with the matched condition f(d) = 0 (or SWR = 1) being the
ongm.
It is interesting to note that equation (3 .12) is very similar in appearance to the
expression for detennining the impedance from a given reflection coefficient:
Z(d) =
z0 1 + f(d)
1-f(d)
(3.14)
This similarity, together with the fact that for lf(d)l ::;; 1 the SWR is greater or equal to
unity, suggests that the actual numerical value for the SWR can be found from the
: Smith Chart by finding the intersection of the circle of radius lf(d)l with the right: hand side of the real axis.
-----------------------------~~~
Example 3-4: Reflection coefficient, voltage standing wave
ratio, and return loss
Four different load impedances:
(a) ZL = 50 Q , (b) ZL :::: 48.5 Q, (c) ZL = (75 + j25) Q, and
(d) Z L = (10- j5) Q, are sequentially connected to a 50 Q transmission line. Find the reflection coefficients and the SWR circles,
and determine the return loss in dB.
Solution:
The normalized load impedances and corresponding

reflection coefficients, return loss, and SWR values are computed as
follows:
114
(a) ZL
= 1 ' r = (zL- 1 )l(zL + 1) = 0, RLdB =
(b) ZL
= 0.97' r
SWR
= (zL- 1)/(zL + 1)
= 1.03
SWR
= -0.015' RLdB
r = (zL -1)/(zL + 1)
SWR = 1.77
(c) ZL = 1.5 + j0.5,
RLdB = 11.1,
00 '
=1
= 36.3'
= 0.23 + j0.15,
(d) ZL = 0.2-j0.1, r = (zL -1) / (zL + 1) = -0.66-j0.14,

RLdB
3.5, SWR = 5.05
To determine the approximate values of the SWR requires us to

exploit the similarity with the input impedance, as discussed previously. To this end, we first plot the normalized impedance values in
the Smith Chart (see Figure 3-6). Then we draw circles with centers
at the origin and radii whose lengths reach the respective impedance
points defined in the previous step. From these circles we see that
the load refection coefficient for zero load reactance ( x L = 0) is
ZL- 1
rL- 1
ro = ZL + 1 = rL + 1
rr
The SWR can be defined in tenn of the real load reflection coefficient along the real r -axis:
1 + rr
1-rr
This requires Ir ol = r r > 0 . In other words, for r r > 0 we have to
enforce rL ~ 1, meaning that only the intersects of the right-handside circles with the real axis define the SWR.
As a graphical design tool, the Smith Chart allows immediate

observation of the degree of mismatch between line and load impedances by plotting the radius of the SWR circle.
~ce Tranformatlon
Figure 3-6
3.2.3
115
SWR circles for various reflection coefficients.
Special Transformation Conditions
The amount of rotation by which the point of the normalized transmission line
impedance circles around the Smith Chart is controlled by the length of the line, or
alternatively the operating frequency. Consequently, both inductive (upper plane) and
capacitive (lower plane) impedances can be generated based on the line length and the
termination conditions at a given frequency. These lumped circuit parameter representations, realized through distributed circuit analysis techniques, are of significant practical importance.
The cases of open- and short-circuit line termination are of particular interest in
generating inductive and capacitive behavior and are examined in more detail next.
116
Open Circuit Transformations

To obtain a pure inductive or capacitive impedance behavior, we need to operate
along the r = 0 circle. The starting point is the right-hand location ( r 0 = 1 ) with
rotation toward the generator in a clockwise sense.
A capacitive impedance - j X c is obtained through the condition
_l_J_ = z.
jroCZ
= -Jcot(Ad
)
P 1
(3.15)
as direct comparison with (2. 70) shows. The line length d 1 is found to be
d, =
Hcoc'(ro~zJ +n1tJ
(3.16)
where n1t (n = I) 2) ... ) is required due to the periodicity of the cotangent function.
Alternatively, an inductive impedance j XL can be realized via the condition
jroLd = Zin
0
= -jcot(~d2 )
(3.17)
The line length d 2 is now found to be

d2 =
~[1t-coc'(~~)+n1t]
(3.18)
Both conditions are schematically depicted in Figure 3-7. How to choose a particular
open-circuit line length to exhibit capacitive or inductive behavior is discussed in the
following example.
----------------------------~~~
Example 3-5: Representation of passive circuit elements
through transmission line section
For an open-ended 50 0 transmission line operated at 3 GHz and
with a phase velocity of 77% of the speed of light, find the line
lengths to create a 2 pF capacitor and a 5.3 nH inductor. Perform
your computations both by relying on (3.16) and (3.18) and by using
the Smith Chart.
Solution:
constant is
For a given value of phase velocity, the propagation
- a Tranafonnatton
117
= 27tflvP = 27t//(0.77c) = 81.6m- 1
Substituting this value into (3.16) and (3.18). we conclude that for
the representation of a 2 pF capacitor we need an open-circuit line or
stub with line length d 1 = 13.27 + n38.5 nun. For the realization of
a 5.3 nH inductor, a d 2 = 32.81 + n38.5 mm stub is required.
The alternative method for computing the lengths of the
required stubs is through the use of the Smith Chart (see Figure 3-7).
At a 3-GHz frequency, the reactance of a 2 pF capacitor is
Xc
1/(roC) = 26.50. The corresponding normalized imped-
L = 5.3 nH
C=2pF
--1r
Figure 3..7
Creating capacitive and inductive impedances via an open-circuit

transmission line.
Chapter 3 Tbe Smith Chart
118
ance in this case is zc = - j X c = - }0.53 . From the Smith Chart

we can deduce that the required transmission line length has to be
approximately 0.172 of one wavelength. We note that for the given
phase velocity, the wavelength is A = v PI f = 77 mm . This
results in a line length of d 1 = 13.24 mm which is very close to the
previously computed value of 13.27 mm. Similarly, for the inductance we obtain zL = }2. The line length in this case is 0.426 of
one wavelength, which is equal to 32.8 mm.
Circuits are often designed with lumped elements before converting them into transmission line segments, similar to the procedure described in this example.
Short-Circuit Transformations
Here the transformation rules follow similar procedures as outlined previously,
except that the starting point in the Smith Chart is now the r 0 = -1 point on the real
axis, as indicated in Figure 3-8.
A capacitive impedance - j X c follows from the condition
. lczl
Jffi
zin
jtan(~dl)
(3.19)
where use is made of (2.66). The line length d 1 is found to be

1
d1 =
~[1t- tan- (ro~zJ + n1tJ
(3.20)
Alternatively, an inductive impedance j XL can be realized via the condition
jtan(~d2 )
(3.21)
Htan-~(~~)+nltJ
(3.22)
jroLi
zin
The line length d 2 is now found to be

d2 =
At high frequencies, it is very difficult to maintain perfect open-circuit conditions

because of changing temperatures, humidity, and other parameters of the medium surrounding the open transmission line. For this reason short-circuit conditions are more
preferable in practical applications. However, even a short-circuit termination becomes
119
C=2
---itFigure 3-8
Creating capacitive and inductive impedances via a short-circuit

transmission line.
problematic at very high frequencies or when through-hole connections in printed circuit boards are involved, since they result in additional parasitic inductances. Moreover~
a design engineer may not have a choice if the circuit layout area is to be minimized by
tequiring the selection of the shortest line segments. For instance, the realization of a
capacitor always yields the shortest length for an open-circuit line.
3.2.4
Computer Simulations
There are many computer aided design (CAD) programs available to facilitate the
RF/MW circuit design and simulation processes. These programs can perfonn a multitude of tasks, varying from simple impedance calculations to complex circuit optimizations and circuit board layouts. One commercial software package that is used throughout
120
this textbook is called Monolithic and Microwave Integrated Circuit Analysis and Design
(MMICAD) (Optotek Ltd., Kanata, Ontario, Canada), which is a linear simulator program with optimization tools. Another well-known program with advanced features is
EESof's Libra package (Hewlett-Packard Corporation, Westlake Village, CA, USA),
which is capable of performing linear as well as nonlinear analyses and optimizations.
It is not the purpose of this textbook to review and discuss the various CAD programs presently in industrial and academic use. However, to reproduce the subsequent
simulation results, Appendix I provides a brief introduction to the basic features of
MATLAB, which was chosen as a tool to carry out most simulations presented in this book.
The main reason for using MATLAB is its wide-spread use as a mathematical
spreadsheet which permits easy programming and direct graphical display. This eliminates the need to rely on complex and expensive programs accessible to only a few
readers. The benefit of a MATLAB routine will inunediately become apparent when the
Smith Chart computations have to be perfonned repetitively for a range of operating
frequencies or line lengths as the following discussion underscores.
In this section we revisit Example 3-2, which computed the input reflection coefficient and input impedance of a generic transmission line connected to a load. We now
extend this example beyond a single operating frequency and a fixed line length. Our
goal is to examine the effect of a frequency sweep in the range from 0.1 GHz to 3 GHz
and a change in line length varying from 0.1 em to 3 em. The example MATLAB routine,
which performs the analysis of the transmission line length changing from 0.1 em to
3 em at a fixed operating frequency 2 GHz, is as follows:
smith_chart;
Set_ZO(SO);
d=0.0:0.00l:Q . 03;
% plot smith chart

% set characteristic impedance to 50 Ohm
% set load impedance to 30+j60 Ohm
% compute phase velocity
% set frequency to 2 GBz
% set the line length to a range from 0 to
betta~2*pi*f/vp;
% 3 em in 1 mm increments
% compute propagation constant
s_Load(30+j*60);
vp=0.5*3e8;
=2e9;
Gamma=(ZL-ZO) / (ZL+ZO); % compute load reflection coefficient

rd=abs(Gamma);
% roaqnitude of the reflection coefficient
alpha=angle(Gamma)-2*betta*d;
% phase of the reflection
% coefficient
plot(rd*cos(alpha),rd*sin(alpha)); %plot the graph
In the first line of the MATLAB code (see file fig3_9.m on the accompanying CD)
we generate the Smith Chart with the necessary resistance and reactance circles. The
next lines define the characteristic line impedance Z 0 :::: 50 Q, load impedance
~tlipedence Transformation
121
ieL = (30 + j60) .Q,
operation frequency f = 2 x 10 Hz, and phase velocity

8
:'vp 0.5 x 3 x 10 rnls. The command line d=O. o:o. 001: o. 03 creates an array d rep;,resenting the transmission line length, which is varied from 0 mm to 3 em in 1-mm
;increments. After all parameters have been identified, the magnitude and phase of the
Jnput reflection coefficients have to be computed. This is accomplished by determining
i:the propagation
constant
~ = 2rcf lv P'
load
reflection
coefficient
T 0 = (ZL-Z 0 )/(ZL +Z0 ) and its magnitude
and the total angle of rotation
.:' a = L(f0 ) - 2 ~d. Finally, the display of the impedance as part of the Smith Chart is
done through the plot command, which requires both real and imaginary phasor argucos(
and
sin( a). The final result is shown in Figure 3-9.
lfol
ments lfol
a)
Ifol
I
. i
zL= (30 +160) n
Figure 3-9 Input impedance of a loaded line of 2 em length for a sweep in

operating frequency from 0.0 to 3 GHz. If the operating frequency is fixed at 2 GHz
and the line length is varied from 0.0 to 3 em, the same impedance curve is obtained.
For the case where the length of the line is fixed to be 2 em and the frequency is
. swept from values ranging from 0.0 to 3 GHz, the only necessary modification to the
above input file is to set d=O. 02, followed by specifying the frequency range in incre:: ments of 100 MHz (i.e., f=O. o: le7: 3e9). We should note that in both cases the electri. cal length (~d) of the line changes from 0 to 144 . Therefore, the impedance graphs
produced for both cases are identical.
At the end of the rotation, either by fixing the frequency and varying the length or
Vice versa, the input impedance is found to be Z in = ( 12.4 + j 15.5) .Q . It is reassuring
that for a fixed frequency f = 2 GHz and a line length ranged= 0 ... 2 em, we ulti-
122
Chapter 3 The Smith Chill
mately arrive at the same input impedance of Z10 = ( 14.7- j26.7) Q as obtained in
Example 3-2.
3.3 Admittance Transformation

3.3.1
Parametric Admittance Equation
From the representation of the normalized input impedance (3.4), it is possible to

obtain a normalized admittance equation by simple inversion:
Yin
Yin
y 0 = Zin
I- r(d)
1 + r( d)
(3.23)
where f 0 = 1/ Z 0 To represent (3.23) graphically in the Smith Chart, we have several

options. A very intuitive way of displaying admittances in the conventional Smith Chart
or Z-Smith Chart is to recognize that (3.23) can be found from the standard representation (3.4) via
1- r(d) - 1 + e-j1tr(d)
(3.24)
1 + r(d) - 1- e-}1tr(d)
In other words, we take the normalized input impedance representation and multiply
the reflection coefficient by -1 =
complex r -plane.
which is equivalent to a 180 rotation in the
e -jrc,
--------------------------~~~~
Example 3-6: Use of the Smith Chart for converting impedance to admittance
Convert the normalized input impedance zin = 1 + jl = Jiej(rc/
into normalized admittance and display it in the Smith Chart.
Solution:
The admittance can be found by direct inversion, that is

Y =
m
-j(rt/ 4 )
-e
J2
1 .1
= --J2
2
In the Smith Chart we simply rotate the reflection coefficient corresponding to z:in by 180 to obtain the impedance. Its numerical
value is equal to Yin as shown in Figure 3-10. To denormalize Yin
Admittance Transformation
Figure 310
123
Conversion from impedance to admittance by 180 rotation.
we multiply by the inverse of the impedance normalization factor.

Thus,
1
- zYin - YoYin
0
Rotations by 180 degrees to convert from the impedance to the

admittance representation require only a reflection about the origin
in the T-plane.
In addition to the preceding operation, there is a widely used additional possibility.

Instead of rotating the reflection coefficient by 180 in the Z-Smith Chart, we can
124
rotate the Smith Chart itself. The chart obtained by this transfonnation is called the
admittance Smith Chart or the Y-Smith Chart. The correspondences are such that
nonnalized resistances become normalized conductances and nonnalized reactances
become normalized susceptances. That is,
R
Zo
G
= Z0G
Yo
r = - => g = and
x = - => b = - = Z 0 B
Zo
Yo
This reinterpretation is depicted in Figure 3-11 for a particular normalized impedance
point z 0.6 + jl.2.
-1
(a) Z-Smith Chart

(b) Y..Smith Chart
Figure 3-11 Reinterpretation of the Z-Smith Chart as a Y-Smith Chart.
As seen in Figure 3-11, the transfonnation preserves (a) the direction in which the
angle of the reflection coefficient is measured and (b) the direction of rotation (either
toward or away from the generator). Attention has to be paid to the proper identification
of the extreme points: A short-circuit condition zL = 0 in the Z-Smith Chart is
yL
oo in the Y-Smith Chart, and conversely an open-circuit zL
oo in the Z-Smith
Chart is y L = 0 in the Y..Smith Chart. Furthermore, negative values of susceptance are
plotted now in the upper half of the chart, corresponding to inductive behavior, and positive values in the bottom half, corresponding to capacitive behavior. The real component of the admittance increases from right to left.
Mnlltlnce Transformation
125
To complete our discussion of the Y-Smith Chart, we should mention an addi. tiona!, often employed definition of the admittance chart. Here the admittance is repre: sented in exactly the same manner as the impedance chart without a 180 rotation. In
this case the reflection coefficient phase angle is measured from the opposite end of the
chart (see the book by Gonzalez listed in Further Reading at the end of this chapter).
3.3.2
Additional Graphical Displays
In many practical design applications it is necessary to switch frequently from

impedance to admittance representations and vice versa. To deal with those situations a
combined, or so-called ZYSmith Chart, can be obtained by overlaying the Z- and Y. Smith Charts, as shown in Figure 3-12.
=::
Figure 3-12 The ZY-Smith Chart superimposes the z- and Y-Smith Charts in one
graphical display.
126
This combined ZY-Smith Chart allows direct conversion between impedances and
admittances. In other words, a point in this combined chart has two interpretations
depending on whether the Z-Chart or Y-Chart display is chosen.
------------------------~&MW4
Example 3-7: Use of the combined ZY-Smith Chart

Identify (a) the normalized impedance value z = 0.5 + j0.5 and (b)
the normalized admittance value y = 1 + j2 in the combined ZYSmith Chart and find the corresponding values of normalized admittance and impedance.
Solution:
Let us first consider the nonnalized impedance value
z = 0.5 + j0.5 . In the combined ZY-Smith Chart we locate the
impedance by using circles of constant resistance r = 0.5 and constant reactance x = 0.5, as shown in Figure 3-12. The intersection of
these two circles detennines the specified impedance value
z = 0.5 + j0.5. To find the corresponding admittance value we
simply move along the circles of constant conductance g and susceptance b. The intersection gives us g = 1 and jb = -jl (i.e., the
admittance for part (a) of this example is y = 1 - jl ). The solution
for the normalized admittance y = 1 + j2 is obtained in identical
fashion and is also illustrated in Figure 3-12.
The ZY-Smith Chart requires a fair amount of practice due to

its busy'' appearance and the fact that inductors and capacitors
are counted either in positive or negative units depending on
whether an impedance or admittance representation is needed.
3.4 Parallel and Series Connections

In the following sections several basic circuit element configurations are analyzed
and their impedance responses are displayed in the Smith Chart as a function of frequency. The aim is to develop insight into how the impedance/admittance behaves over
a range of frequencies for different combinations of lumped circuit parameters. A prac-
Parallel and Series Connections
127
tical understanding of these circuit responses is needed later in the design of matching
networks (see Chapter 8) and in the development of equivalent circuit models.
3.4.1
Parallel Connection of Rand L Elements
Recognizing that g = Z 0 / R and bL = +Z0 /(roL), we can locate the normal

ized admittance value in the upper Y-Smith Chart plane for a particular, fixed normalized conductance g at a certain angular frequency roL :
Y..m(roL)
= g- jroZo
L
(3.25)
As the angular frequency is increased to the upper limit ro u , we trace out a curve along
the constant conductance circle g. Figure 3-13 schematically shows the frequencydependent admittance behavior for various constant conductance values g = 0.3, 0.5,
0.7, and I and for frequencies ranging from 500 MHz to 4 GHz. For a fixed inductance
value of L = 10 nH and a characteristic line impedance Z 0 = 50 n, the susceptance
always starts at -1.59 (500 MHz) and ends at --D.20 (4 GHz).
In Figure 3-13 and the following three additional cases, the transmission line
characteristic impedance is represented as a lumped impedance of Z 0 = 50 Q . This is
permissible since our interest is focused on the impedance and admittance behavior of
different load configurations. For these cases the characteristic line impedance serves
only as a normalization factor.
Figure 3-13
Admittance response of parallel RL circuit for roL :S ro :S o.>u at

constant conductances g = 0.3, 0.5, 0.7, and 1.
128
3.4.2
Parallel Connection of Rand C Elements
Here we operate in the lower Y..Chart plane because susceptance be = Z0 roC

remains positive. To locate the normalized admittance value for a particular, fixed normalized conductance g and angular frequency ro L we have
Yin(O>L)
= g + jZ0 roLC
(3.26)
Figure 3-14 depicts the frequency-dependent admittance behavior as a function of various constant conductance values g =0.3, 0.5, 0.7, and l. The normalized susceptance
for C = 1 pF and characteristic line impedance Z 0 = 50 Q always starts at 0.16
(500 MHz) and ends at 1.26 (4 GHz).
Zo lin
~
II
\..)
~
0
N
II
jb = j1.26
Figure 314 Admittance response of parallel RC circuit for roL < ro s; rou at
constant conductances g =0.3, 0.5, 0.7, and 1.
3.4.3
Series Connection of Rand L Elements
When dealing with series connections, we can conveniently choose the Z-Smith
Chart for the impedance display. Identifying the normalized reactive component as
x L = roLl Z 0 , it is straightforward to locate the normalized impedance value for a particular, fixed normalized resistance r at a given angular frequency ro L :
Zin( roL)
= r + jroLLIZ0
(3.27)
In Figure 3-15 the frequency-dependent impedance behavior is shown as a function of

various constant resistance values r =0.3, 0.5, 0.7, and I. For the same inductance of
10 nH and characteristic line impedance of 50 Q as used in Figure 3-13, we now pick
129
Plnllleland Sertes Connection
reactance circles associated with 0.63 (500 MHz) and with 5.03 (4 GHz). Because the
reactance is positive and since we use the Z-Smith Chart, all impedances have to reside
in the upper half plane.
Zo Zin
10 nH
R =rZ0
Figure 3-15
3.4.4
Impedance response of series RL circuit for roL $ ro $ rou and

constant resistances r = 0.3, 0.5, 0. 7, and 1.
Series Connection of R and C Elements
We again choose the Z-Srnith Chart for the impedance display. The normalized
reactive component is xc = +1 / (roCZ0 ), indicating that all curves will reside in the
lower half of the Smith Chart. The normalized impedance value for a particular, fixed
normalized resistance rat an angular frequency roL reads
zin(mL)
= r - j COL ~Z0
(3.28)
Figure 3-16 displays the frequency-dependent impedance behavior as a function of various constant resistance values r =0.3, 0.5, 0.7, and 1. The capacitance of 1 pF in series
with the variable resistance connected to a characteristic line impedance of 50 .Q now
yields circles associated with the reactances of --6.03 (500 MHz) and -0.8 (4 GHz),
which intersect with the four resistance circles, uniquely detennining upper and lower
impedance values.
130
Zo .?in
1 pF
=r~
f=500MHz
jx =-)6.37
Figure 3-16
3.4.5
Impedance response of series RC circuit for roL:::;; ro:::;; rou at

constant resistances r = 0.3, 0.5, 0.7, and 1.
Example of aT-Network
In the previous examples only pure series or shunt configurations have been analyzed. In reality, however, one often encounters combinations of both. To show how
easily the ZY Chart allows transitions between series and shunt connections, let us
investigate by way of an example the behavior of a Ttype network connected to the
input of a bipolar transistor. The input port of the transistor is modeled as a parallel RC
network as depicted in Figure 3-17. As we will see in Chapter 6, R L approximates the
base-emitter resistance and C L is the base-emitter junction capacitance. The numerical
parameter values are listed in Figure 3-17.
LI
''--~---
_ _ _ _ _j\.
T-type network
Figure 3-17
ZL
--
y-
Transistor
input
T network connected to the base-emitter input impedance of a

bipolar transistor.
a
'
filnllllll'ld Series Connections
131
To use the Smith Chart for the computation of the input impedance of this more
complicated network, we first analyze this circuit at 2 GHz and then show the entire
RSp<>nse of the circuit for a frequency range from 500 MHz to 4 GHz by employing the
commercial MMICAD software simulation package.

~-
To obtain the load impedance, or the input impedance of the transistor, we use the
Y..Smith Chart to identify the conductance point corresponding to the load resistor
RL 31.25 0. Assuming a 50 Q characteristic line impedance, we determine the
DOrmalized admittance for this case to be g A = 1.6, which corresponds to point A in
Figure 3-18.
Figure 3-18 Computation of the normalized input impedance of the T network

shown in Figure 3-17 for a center frequency f = 2 GHz .
Chapter 3 The Smith Chllt ~
132
J.
The next step is to connect the capacitance CL = 9 ~ pF in shunt with the resis- ;
tor Rr. At the angular frequency of roL = 21t2 x 10 s , the susceptance of this ~
capacitor becomes Be = roLCL = 24 mS, which corresponds to a rotation of the ;
L
original point A into the new location B. The amount of rotation is detennined by the
normalized susceptance of the capacitor be = Be Z 0 = 1.2 and is carried out along :
L
L
the circle of constant conductance in theY-Smith Chart (see Figure 3-18).
Re-evaluating point B in the Z-Smith Chart, we obtain the normalized impedance '.
of the parallel combination of resistor R L and capacitor C L to be z8 = 0.4- j0.3. The
series connection of the inductance L 1 results in the new location C. This point is
obtained through a rotation from xB = -0.3 by an amount xL = roLL 11Z0 = 1.1 to
I
xc = 0.8 along the circle of constant resistance r = 0.4 in the Z-Smith Chart as discussed in Section 3.4.3.
Converting point C into a Y-Smith Chart value results in Yc = 0.5- jl.O. The
shunt connected capacitance requires the addition of a normalized susceptance
be = roCZ0 = 1.5, which results in the admittance value of Yv = 0.5 + j0.5 or
point D in theY-Smith Chart. Finally, converting point D into the impedance value
zD = 1 - j 1 in the Z-Smith Chart allows us to add the normalized reactance
xL = roLL 2 /Z 0 = 1 along the constant r = 1 circle. Therefore, we reach zin = 1
2
or point E in Figure 3-18. This value happens to match the 50 Q characteristic transmission line impedance at the given frequency 2 GHz. In other words,
Zin = Z 0 = 50 Q .
When the frequency changes we need to go through the same steps but will arrive
at a different input impedance point zin. It would be extremely tedious to go through
the preceding computations for a range of frequencies. This is most efficiently done by
the computer.
Relying on the previously mentioned CAD program MMICAD we are able to
produce a graphical display of the input impedance in the Z-Smith Chart over the entire
frequency range in preselected increments of 10 MHz, as shown in Figure 3-19. This
figure can also be generated as part of the MATLAB software (see file fig3_18.m on the
accompanying CD).
We notice that the impedance trace ranging from 0.5 to 4 GHz is in agreement
with our previous calculations at 2 GHz. Also, as the frequency approaches 4 GHz, the
capacitor of C = 2.39 pF behaves increasingly like a short circuit in series with a single
inductor L 2 . For this reason, the normalized resistance r approaches zero and the reactance grows to large positive values.
Summary
133
Figure 3-19 CAD simulation of the normalized input impedance Zin for the
network depicted in Figure 3-17 over the entire frequency range
500 MHz ~ f~ 4 GHz.
3.5 Summary
This chapter has derived the Smith Chart as the most widely used RF graphical
design tool to display the impedance behavior of a transmission line as a function of
either line length or frequency. Our approach originated from the representation of the
normalized input impedance of a tenninated transmission line in the form
Zin
=r + ]X =
l+f(d ) l+rr+Jfi
1 - f (d) = l - r r - j f i
which can be inverted in terms of the reflection coefficient to yield two circle equations
(3.10) and (3.11), which take on the following expressions for the normalized
resistance r:
+ r~
(r r - r_r_)2
+1
'
and for the normalized reactance x
(fr -l)2 + ( r; - D2
= (~f
Superimposing the circles described by both equations over the complex polar form of
the nonnalized impedance z-plane on the unit circle yields the Smith Chart. The key
feature to remember is that one full rotation is equal to half a wavelength because of the
134
Chapter 3 The Smfth Chart
exponent 2~d in the reflection coefficient expression (3.2). In addition to observing the
impedance behavior, we can also quantify in the Smith Chart the degree of mismatch
expressed by the standing wave ratio (SWR) equation (3.12), or
SWR(d)
= 1 + lr(d)l
1 -lr(d)l
which can be directly obtained from the chart.
To facilitate computer-based evaluation of the Smith Chart, a wide range of commercial programs can be utilized. Due to its ease of implementation on a PC and its
user-friendly interface, throughout this book we have used the package MMICAD
developed by Optotek. However, for the relatively incomplicated circuits analyzed in
this Chapter, one can also create a custom-tailored Smith Chart and perform simple
computations by relying on mathematical spreadsheets such as Mathematica, MATLAB,
or MathCad. To demonstrate the procedure, a number of MATLAB modules have been
developed, and the use of these so-called m.files as part of a basic Smith Chart computation is demonstrated in Section 3.2.4.
A transition to the admittance, or Y-Smith Chart, can be made via (3.23):
Yin
Yin
= yo = Zin =
1- r(d)
1 + r( d)
and it is found that the only difference to (3.4) is a sign reversal in front of the reflection
coefficient. Consequently, rotating the reflection coefficient in the Z-Smith Chart by
180 results in theY-Smith Chart. In practice, this rotation can be avoided by turning
the chart itself. Superimposing the rotated chart over the original Z-Smith Chart provides a combined Zf-Smith Chart display. The benefit of such a display is the easy transition from parallel to series connection in circuit designs. This ease is demonstrated by
a T-network configuration connected to the input port of a bipolar transistor consisting
of a parallel RC network. To investigate the impedance behavior as a function of frequency sweep, however, is most easily accomplished through the use of CAD programs.
Further Reading
G. Gonzalez, Microwave Transistor Amplifiers: Analysis and Design, 2nd ed., Prentice
Hall, Upper Saddle River, NJ, 1997.
K. C. Gupta, R. Garg, and I. J. Bohl, Microstrip Lines and Slotlines, Artech House,
Dedham, MA, 1979.
J. Helszajn, Passive and Active Microwave Circuits, John Wiley, New York, 1978.
Probtlma
138
Hewlett-Packard Application Note 154, "S-Parameter Design," 1972.

H. Howe, Stripline Circuit Design, Artech House~ Dedham, MA, 1974.
S. Y Liao, Microwave Devices and Circuits, Prentice Hall, Englewood CJiffs, NJ, 1980.
MMICAD for Windows, Reference Manual, Optotek, Ltd., 1997.
D. M. Pozar, Microwave Engineering, 2nd edition, John Wiley, New York, 1998.
P. A. Rizzi, Microwave Engineen'ng, Passive Circuits, Prentice Hall, Englewood Cliffs,

NJ, 1988.
P. H. Smith, ''Transmission-Line Calculator," Electronics, Vol. 12, pp. 29-31, 1939.
P. H. Smith, '~n Improved Transmission-Line Calculator," Electronics, Vol. 17, p. 130,
1944.
Problems
3.1 Consider a load Z L = ( 80 + j40) n connected to a lossy transmission line

with characteristic line impedance of
O.l + j200
0.05 - }0.003
Detennine the reflection coefficient and the standing wave ratio (SWR) at
the load.
3.2
A coaxial cable of characteristic line impedance Z 0 = 75 0 is terminated

by a load impedance of z L = ( 40 + j35) n . Find the input impedance of
the line for each of the following pairs of frequency f and cable length d
assuming that the propagation velocity is 77% of the speed of light:
(a) f = 1 GHz and d = 50 em
(b) f = 5 GHz and d = 25 em
(c) f = 9 GHz and d = 5 em
3.3
The attenuation coefficient of a transmission line can be detennined by

shortening the load side and recording the VSWR at the beginning of the
line. We recall that the reflection coefficient for a lossy line takes on the form
r(d) = roexp(- kl) = roexp(-a/)exp(-j~/). If the line is 100 m in
length and the VSWR is 3. find the attenuation coefficient a in Np/m, and
dB/m.
136
Chapter 3 The Smith Chlft
3.4
A load impedance of ZL = (150- j50) 0 is connected to a 5 em long

transmission line with characteristic line impedance of Z 0 = 75 0. For a
wavelength of 6 em, compute
(a) the input impedance
(b) the operating frequency, if the phase velocity is 77% the speed of light
(c) the SWR
3.5
Identify the following normalized impedances and admittances in the Smith

Chart:
(a) z = 0.1 + j0.7
(b) y = 0.3 + j0.5
(c) z = 0.2+j0.1
(d) y = 0.1 + j0.2
Also find the corresponding reflection coefficients and SWRs.
3.6
An unknown load impedance is connected to a 0.3A. long, 50 0 lossless
transmission line. The SWR and phase of the reflection coefficient measured
at the input of the line are 2.0 and -20, respectively. Using the Smith Chart.
detennine the input and load impedances.
3.7
In Section 3.1.3 the circle equation (3.10) for the normalized resistance r is
derived from (3.6). Start with (3.7); that is,
2f.I
and show that the circle equation
1)2 = (1)2
~
(f,-1) 2 + ( ri-~
can be derived.
3.8
Starting with the equation for normalized admittance

1- r
.
y = g+jb = - l+f
prove that the circle equations for the Y-Smith Chart are given by the follow
ing two formulas:
(a) For the constant conductance circle as
)2 2 ( )2
g - + f = -1( fr+l+g
'
1+g
(b) For the constant susceptance circle as
137
3.9
A lossless transmission line (Z 0 :.::: 50 Q ) is 10 em long (/

800 MHz,
vP = 0.77 c). If the input impedance is Zin = } 60 Q
(a) Find Z L (using the Smith Chart)
(b) What length of a short-circuit transmission line would be needed to
replace ZL?
3.10 A transmission line of characteristic impedance Z 0 = 50 Q and length

d = 0.15A. is tenninated into a load impedance of ZL = (25- }30) Q.
Find 0 , Zin(d) , and the SWR by using the Z-Smith Chart.
3.11
A short-circuited 50 .Q transmission line section is operated at 1 GHz and

possesses a phase velocity of 75% of the speed of light. Use both the analytical and the Smith Chart approach to determine the shortest lengths required
to obtain (a) a 5.6 pF capacitor, and (b) a 4.7 nH inductor.
3.12 Determine the shortest length of a 75 Q open-circuit transmission line that

equivalently represents a capacitor of 4.7 pF at 3 GHz. Assume the phase
velocity is 66% of the speed of light.
3.13 A circuit is operated at 1.9 GHz and a lossless section of a 50 n transmission line is short circuited to construct a reactance of 25 Q. (a) If the phase
velocity is 3/4 of the speed of light, what is the shortest possible length of
the line to realize this impedance? (b) If an equivalent capacitive load of 25
Q is desired, determine the shortest possible length based on the same phase
velocity.
3.14 A microstrip line with 50 Q characteristic line impedance is tenninated into
a load impedance consisting of a 200 Q resistor in shunt with a 5 pF capacitor. The line is 10 em in length and the phase velocity is 50% the speed of
light. (a) Find the input impedance in the Smith Chart at 500 MHz, 1 GHz,
and 2 GHz, and (b) use the MATLAB routine (see Section 3.2.4) and plot the
frequency response from l 00 MHz to 3 GHz in the Smith Chart.
3.15 For an FM broadcasting station operated at 100 MHz, the amplifier output
impedance of 250 Q has to be matched to a 75 n dipole antenna.
(a) Detennine the length and characteristic impedance of a quarter-wave
transformer with vP = 0. 7 c .
(b) Find the spacing D for a two-wire loss less transmission line with AWG
-------------------------------;
Chapter 3 The Smith Chlrt .~

J
138
26 wire size and a polysterene dielectric ( E, = 2.55 ).

i
3.16 Consider the case of matching a 73 .Q load to a 50 n line by means of a 1

'AI 4 transformer. Assume the matching is achieved for a center frequency of 1
f c = 2 GHz. Plot the SWR for the frequency range 1/3::; f If c::; 3.
3.17
A line of characteristic impedance of 7 5 .Q is terminated by a load consist- !

ing of a series connection of R = 30 .Q , L = 10 nH, and C = 2.5 pF. :
Find the values of the SWR and minimum line lengths at which a match of
the input impedance to the characteristic line is achieved. Consider the fol- :
lowing range of frequencies: (a) 100 MHz, (b) 500 MHz, and (c) 2 GHz.
3.18
A 50 .Q lossless coaxial cable ( , = 2.8) is connected to a 75

antenna .
operated at 2 GHz. If the cable length is 25 em, find the input impedance by
using the analytical equation (2.71) and the Z-Smith Chart.
3.19 A balanced to unbalanced (balun) transfonnation is often needed to connect

a dipole antenna (balanced) to a coaxial cable (unbalanced). The following
figure depicts the basic concept.
Unbalanced
coaxial cable
Balanced
: antenna
'
139
'nbteme
As an alternative of using a transformer, one often uses the following

antenna connection.
(a) Explain why one leg of the dipole antenna is connected a distance A./ 4
away from the end of the coax cable.
(b) For an FM broadcast band antenna in the frequency range from 88 to
I 08 MHz, find the average length where the connection has to be made.
3.20 Using the ZY-Smith Chart, find the input impedance of the following network at 2 GHz.
What is the input impedance of this network at 1 GHz?
3.21 A Z0
= 500
transmission line is O.SA. in length and terminated into a load

of ZL = (50- j30)Q. At 0.35A. away from the load, a resistor of
R = 250 is connected in shunt configuration (see figure below). Find the
input impedance with the help of the ZY Smith Chart.
140
Ct.pter 3 The Smith Clllrt
3.22
A 50- n transmission line of 3/4 wavelength in length is connected to two

transmission line sections each of 75 0 in impedance and length of 0.86 and
0.5 wavelength, respectively, as illustrated in the following figure.
The termination for line 1 is zl = (30 + j40) n and z2 = (75-j80) Q

for line 2. Employ the Smith Chart and find the input impedance.
3.23
Repeat the previous problem if all characteristic line impedances are

Z 0 = 50 Q and all transmission line sections are 'A/ 4 in length.
3.24
A dipole antenna of impedance ZL = (75 + }20) 0 is connected to a 50 0

lossless transmission whose length is "A./3. The voltage source V G = 25V
is attached to the transmission line via an unknown resistance RG. It is
detennined that an average power of 3W is delivered to the load under loadside matching (Zratch = 50 0). Find the generator resistance RG, and
detennine the power delivered to the antenna if the generator impedance is
matched to the line via a quarter-wave transformer.
141
Problem
3.25 Determine the values of the inductance L and the capacitance C such that
they result in a 50 Q input impedance at 3 GHz operating frequency for the
following network.
L
ZL
(150 + j50)0
3.26 An open-circuit transmission line (50 Q) is operated at 500 MHz (vp = 0.7c).
Use the ZY Smith Chart and find the impedance Zin if the line is 65 em in
length. Find the shortest distance for which the admittance is lin= -j0.05S.
3.27 Find the minimum line length 11 and the minimum length of the short-circuited stub 12 in terms of wavelength A., such that the input impedance of
the circuit is equal to 50 n .
ZL
(50+ j50)0
Z0 =50fl
3.28 Find the input impedance in terms of magnitude and phase of the following
network at an operating frequency of 950 MHz.
Z:n
t-1
2 pF
R
750
Z0 =50 n
3.29 Repeat your computation and solve Problem 3.28 for a 1.5 GHz operating
frequency. Comment on the differences in your results.
142
3.30
A specific transmission line configuration is as follows:
The characteristic line impedance for all three elements is Z 0 = 50 n . The

load impedance has a value of ZL = (20 + }40) 0, and the electrical
lengths of the corresponding line segments are e 1 = 164.3 ' e2 = 57.7 '
and 9 3 = 25.5.
(a) Find the input impedance.
(b) Find the input impedance if transmission line segment 9 2 is open
circuit.
(This problem and Problem 3.27 become very important in Chapter 8, when
we discuss the problem of matching a particular load impedance to a desired
input impedance.)
CHAPTER
'Single- and Multiport

Networks
J,;
:.,..
Ever since single- and multiple port networks were

~t introduced into the electrical engineering profession through Guillemin and Feld~ller. they have quickly become indispensable tools in restructuring and simplifying
~licated circuits as well as in providing fundamental insight into the performance
~active and passive electronic devices. Moreover, the importance of network modelIng has extended far beyond electrical engineering and has influenced such diverse
~Ids as vibrational analysis in structural and mechanical engineering as well as bioPerucine. For example, today's piezoelectric medical transducer elements and their
i*ctrical-mechanical conversion mechanisms are most easily modeled as a three-port
letwork.
The ability to reduce most passive and active circuit devices, irrespective of their
j:omplicated and often nonlinear behavior, to simple input-output relations has many
~antages. Chief among them is the experimental determination of input and output
parameters without the need to know the internal structure of the system. The
f'tack box" methodology has tremendous appeal to engineers whose concern is mostly
l?cused on the overall circuit performance rather than the analysis of individual compoThis approach is especially important in RF and MW circuits, where complete
...
~d theoretical solutions to Maxwell's equations are either too difficult to derive or the
a.Iutions provide more information than is normally needed to develop functional,
-. tical designs involving systems such as filters, resonators, and amplifiers.
,, In the following sections our objective is to establish the basic network input-out..~ parameter relations such as impedance, admittance, hybrid, and ABCD-parameters.
: then develop conversions between these sets. Rules of connecting networks are pre. ted to show how more complicated circuits can be constructed by series and parallel
f.ascading of individual network blocks. Finally, the scattering parameters are presented
itnts.
~
ji:
143
------------------------------------------------------------------~
144
Chapter 4 Single- and Murtlport Netwoltll j
as an important practical way of characterizing RF/MW circuits and devices through
the use of power wave relations.
4.1 Basic Definitions

Before embarking on a discussion of electrical networks we have to identify some .
general definitions pertaining to directions and polarity of voltages and currents. For .
our purposes we use the convention shown in Figure 4- 1. Regardless of whether we ."
deal with a single-port or an N-port network, the port-indexed current is assumed to
flow into the respective port and the associated voltage is recorded as indicated.
One-port
Network
..-=--
+
v!
Two-port
N etwork
Port 2
Port 1
Port 1
_..
+--=-
v2 + Port 2
vl
i4
.._,...
Port 3
+ vJ
...
lN--.
Port N - I + VN- l
Figure 4-1
+
v2
Multiport
v4....
Network
...
lN
......VN
Port4
+ Port N
Basic voltage and current definitions for single- and

multipart network.
In establishing the various parameter conventions we begin with the voltage-cur-

rent relations through double-indexed impedance coefficients Z,m, where indices nand .
m range between 1 and N. The voltage at each port n = 1 . .. N is given by
(4.la) :
for port 1,
(4.lb) '
for port 2, and
vN
zNI i1
+ z N2 i2 + ... + z NN iN
(4.tcr
for port N . We see that each port n is affected by its own impedance Znn as well as by a.
linear superposition of all other ports. In a more concise notation, (4.1 ) can be con-.
verted into an impedance or Z -matrix form:
Blalc Deflnftlona
145
vl
v2
VN
zll z12
ztN
ll
Zzt Z22
.. . z2N
l2
ZN I Z N2 ... ZNN
lN
(4.2)
or, in matrix notation,
{V} = [Z]{I}
(4.3)
where {V} and {I} are vectors of voltages v 1, v 2, ... , vN and currents i 1, i 2, ... , iN,
respectively, and [Z] is the impedance matrix.
Each impedance element in (4.2) can be detennined via the following protocol:
vn
znm
(4.4)
= -:l
m ik =0 (fork ;em)
which means that the voltage v n is recorded at port n while port m is driven by current
im and the rest of the ports are maintained under open tenninal conditions (i.e. ik = 0
where k# m).
Instead of voltages as the dependent variable, we can specify currents such that
ll
lz
lN
Yu
y12
YIN
VI
y2l
y22
y2N
v2
y Nl y N2 . .. YNN
VN
(4.5)
or
= [Y]{V}
{I}
(4.6)
where, similar to (4.4), we define the individual elements of the admittance or

Ymatrix as
ln
y nm =V-
(4.7)
m vk
= 0 (fork :;em )
Comparing (4.2) and (4.5), it is apparent that impedance and admittance matrices are
the inverse of each other:
[Z]
= [Yr 1
(4.8)
146
Chaptw 4 Single- and Multlport . . . _ .
---------------------------RF&MW4
Example 4-1: Matrix representation of Pi-network
For the pi-network (the name of the network comes from the resemblance with the greek letter TI) shown in Figure 4-2 with generic
impedances Z A , Z 8 , and Z c find the impedance and admittance
matrices.
--------------- --,
i!
i2
I
I ____._.
+--=-!
+I
port 1
YJ
I+
Zs
I
1
I
I
I
ZA
Zc
; v2
I
I
I
port2
I _ - ------- ---- -- - --I
Figure 4-2
Pi-network as a two-port network.
Solution:
The impedance elements are found by using (4.4) and

the appropriate open- and short-circuit termination conditions.
To find Z 11 we must compute the ratio of the voltage drop v1
across port 1 to the current i 1 flowing into this port when the current
into port 2 equals zero. The requirement i 2 = 0 is equivalent to an
open-circuit condition. Thus. the impedance Z 11 is equal to the parallel combination of impedances Z A and Z JJ + Z c .
vl
::;; ZAII(ZB+Zc) =
Zu - .
'1 .
ZA(ZB + Zc)
A+
zB + z C
=0
The value for Z 12 can be found as the ratio of voltage drop v 1 measured across port I to the current i2 . In this case we must ensure that
the current i 1 remains zero (i.e., we must treat port 1 as open). Voltage v 1 is equal to the voltage drop across impedance Z A and can be
obtained using the voltage divider rule:
12
= Z
ZA
Z VAB
B+ C
where v AB is a voltage drop across impedances Z A and Z 8 con~
nected in series and computed as v AB = i 2 [Zcii(ZA + Z 8 )]. Thus,
VI
Bale Definitions
147
Similarly, we can obtain the rematrung two coefficients of the

impedance matrix:
'z ,,. = 0
Thus, the impedance matrix for the generic pi-network is written in
the form
[Z]
[ZA(ZB + Zc)
ZAZc
ZA + Zs + Zc
ZAZc
Zc(ZA + ZB)
The coefficients for the admittance matrix can be derived using

(4.7). To find the value for Y 11 we must find the ratio of current flow
into port 1 to the voltage drop across this port when the second port
is shortened (i.e., v2 = 0 ).
The value for coefficient Y 12 of the admittance matrix can be

obtained by shortening port 1 (i.e., forcing vi = 0) and measuring
the ratio of the current i I to the voltage drop across port 2. We note
that, when a positive voltage is applied to port 2, the current i 1 will
flow away from port 1, resulting in a negative current:
li
YI2=v2
v, =0
1
--ZB
The rest of the admittance matrix can be derived in the same way,
leading to the following final form
148
Chapter 4 Slngte- and Multlport Networkl
ZB
[Y]
ZB
-1
1
ZB
-+-1
Zc
-1
where YA = Z A , Y8 = Z B , and Y c = Z c .
Direct evaluation shows that the obtained impedance and
admittance matrices are indeed inversely related, which supports the
validity of (4.8).
The practical determination of the matrix coefficients can be

accomplished easily by enforcing open- and short-circuit conditions. However, as the frequency reaches RF limits, parasitic terminal effects can no longer be ignored and a different measurement
approach becomes necessary.
Example 4-1 indicates that both impedance and admittance matrices are symmetric. This is generally true for linear, passive networks. Passive in this context implies
not containing any current or voltage sources. We can state the symmetry as
(4.9)
which also applies for admittances because of (4.8). In fact, it can be proved that any
reciprocal (that is, nonactive, linear) and lossless N-port network is symmetric.
Besides impedance and admittance network descriptions, there are two more useful parameter sets depending on how the voltage and currents are arranged. Restricting
our discussion to two-port networks and with reference to Figure 4-1 , we define the
chain or ABCD-matrix as
(4.10)
and the hybrid or h-matrix as
(4.11)
149
Basic Definitions
The determination of the individual matrix coefficients is identical to the method introduced for the impedance and admittance matrices. For instance. to find h 12 in (4.11),
we set i 1 to zero and compute the ratio of v 1 over v 2 ; that is,
hl2
v.
=-
v2 .
'1 :::
It is interesting to note that in the hybrid representation parameters h 21 and h 12 define

the forward current and reverse voltage gain, respectively. The remaining two parameters determine the input impedance ( h 11 ) and output admittance ( h 22 ) of the network.
These properties of the hybrid representation explain why it is most often used for lowfrequency transistor models. The following example shows the derivation of the hybrid
matrix representation for a bipolar-junction transistor (BJT) for low-frequency
operation.
----------------------------RF~~
Example 4-2: Lowfrequency hybrid network description of a
BJT
Describe the common-emitter BJT transistor in terms of its hybrid
network parameters for the low-frequency, small-signal transistor
model shown in Figure 4-3.
la 1
B~
-~._
I~
_.__.-;:.,~c
z8 1
~.
==-;>
lva ii;~ r,,
o--..-~.,_
_ _ _ _..__ _._ _....,-.oE
._ _ _ _ _ _ _ _ 4
~-----------------------
Figure 43
Common-emitter low-frequency, small-signal transistor model.
Solution:
In the transistor model shown in Figure 4-3 'BE' 'Be'
and r CE represent base-emitter, base-collector, and collector-emitter internal resistances of the transistor. The current through the current-controlled current source is dependent on the current iB'
flowing through the base-emitter resistance.
150
Chapter 4
Slngl~ and
Multlport Networkl
To evaluate the h II parameter of the hybrid matrix according

to (4.10) we must short-circuit the coUector and emitter terminals,
thus setting v2
v CE
0, and compute the ratio of the base-emitter voltage to the base current. Using the notation established in
Figure 4-3, we notice that h 11 is equal to the parallel combination of
rBE
and
rBc:
(input impedance)
Following a similar procedure, the relations for the remaining three
parameters of the hybrid representation can be established as follows:
h12
VBE
rBE
rBE
VCEio=O
h21
tc
LB
rBE
tc
h22 =
PrBc- rBE
= Q
VCE
(voltage feedback ratio)
+ rB c
+ rB c
(small-signal current gain)
= _1_+
1+~
TeE
rBE+rBC
(output admittance)
=0
In the majority of all practical transistor designs, the current amplifiVeE .
'B
cation coefficient p is usually much greater than unity and the

collector-base resistance is much larger than the base-emitter resistance. Keeping these relations in mind, we can simplify the expressions derived for the h-matrix representation of the transistor:
hll
VBE
= rBE
ZB
VC E
=0
VB
hl2
=v
CE.
'o
h21
v ('E
=0
(voltage feedback ratio)
=p
(small-signal current gain)
=O
tc
lB
(input impedance)
=0
..
, . . Dlflnltlona
151
1
- - - + JL (output admittance)
rcE rsc
The hybrid network description is a very popular way to characterize the Bn; and its h-parameter coefficients are widely
reported in 11lflny data sheets.
,t
Due to the presence of the current source in Example 4-2, the h-matrix is no
ger symmetric (h 12 * h21 ) and the transistor model is nonreciprocal. In low-frenc~ electronic circuit design the coefficients of the hybrid matrix representation are
. n hsted as hie for h 11 , hre for h 12 , hfe for h 21 , and hoe for h22 .
Up to this point we considered the problem of deriving the matrix representation
on a known topology and element values of the circuit. However, in practical
tasks it is often required to solve an inverse problem and obtain the equivalent
for an unknown or incompletely defined device based on a few measurements.
becomes extremely important when the characterization of the device is perunder a particular set of operating conditions, but it becomes necessary to evalits performance under completely different circuit conditions. In this case the use
equivalent circuit representation enables an engineer to predict with reasonable
,.,.,,..,!:lt'U the response of the device or circuit under changing operating conditions. In
following example we will derive the values of the internal resistances of the BJT
known h-matrix parameters.
-:
----------------------------~&)A~
Example 4-3: Determination of internal resistances and cur..
rent gain of a BJT based on h-parameter mea
surements
Use the equivalent circuit representation of the BIT shown in
Figure 4-3 and employ the following measured hybrid parameters:
-4
hie = 5 kQ, hre = 2x10 , hfe = 250, hoe= 20 J..LS (these
parameters correspond to the 2n3904 transistor manufactured by
152
Motorola). Find the internal resistances r BE , rBe, and r eE, and the
current gain ~ .
Solution:
As derived in Example 4-2, the values of the h-matrix
for the equivalent circuit shown in Figure 4-3 are given by the following four equations:
'BE
hrt = - - 'BE+ 'Be
(input impedance)
(4.12)
(voltage feedback: ratio)
(4.13)
~'Be- 'BE (small-signal current gain)

'BE+ 'ae
1
1 + 13
hoe = - +
(output admittance)
h fe =
(4.14)
(4.15)
'eE 'BE+ 'Be
If we divide (4.12) by (4.13), we detennine that the base-collector
resistance is equal to the ratio of hie over hre. Accordingly, for
values given in the problem formulation, we obtain:
'Be
h;el hre
71 MQ. Substituting this value into either equation (4.12) or (4.13), we find 'BE = h;el( 1- hre) = 5 kQ. Knowing r 8 c and r 8 E, (4.14) allows us to find the current gain
coefficient ~ = (hrt- h 1e)/(h,e- 1) = 300.02 . Finally, the collector-emitter resistance can be evaluated from (4.15) as
'cE
hie
2
= 63.35 kQ
hoehie- h,eh ft + 2hre- hre
We note from the obtained values that rBE is indeed much smaller
than 'Be.
This example provides a first idea of how the measured hparameters can be used as a basis to characterize the BJT circuit
model. The concept of "inverting" the measurements to determine
circuit model parameters will be further analyzed in Chapter 7.
Interconnecting Networks
153
4.2 Interconnecting Networks

4.2.1
Series Connection of Networks
A series connection consisting of two two-port networks is shown in Figure 4-4.

The individual networks are shown in impedance matrix representation.
r - - - - - - - - - - - - - - - - - - - I
I
__,___. I
vl
I
I
I
I
I
I
I
I
' I
J2 ,.....__
,.....__
lt
--=.....-
v'[
v'2
[z']
.,
.,
ll
__,___.
+
v2
l2
,.....__
""
I
I
I
v"1
v"2
[Z"]
-I - - - - - - - - - - - - - - - - - - - Figure 4-4 Series connection of two two-port networks.
In this case the individual voltages are additive while the currents remain the
same. This results in
(4.16)
where the new composite network [Z] takes the form
[Z] = [Z']+[Z"]= [Zll'+Zll"
ZI2'+Zl2'~
(4.17)
Z 21 ' + Z 21 "Z 22 ' + Z"

22
Caution has to be exercised in not indiscriminately connecting individual networks, as short circuits may be created. This situation is exemplified in Figure 4-5 (a).
The problem can be avoided by including a transformer, as seen in Figure 4-5 (b). The
transfonner in this case decouples input and output ports of the second network. However, this approach will only work for AC signals since the transformer acts as a highpass filter and rejects all DC contributions.
When two networks are connected with the output interchanged, as shown in
Figure 4-6, the most suitable representation is the hybrid form.
In the network connection that is shown in Figure 4-6, the voltages on the input
ports and currents on the output ports are additive (i.e., v 1 = v 1' + v1" and
i 2 = i2' + i 2" ), while the voltages on the output ports and currents on input ports are
154
Chapter 4 Single- and Multlport Networks
~~~
,._.__ I
I+
I
I
IV'
I J
I
z~
I
I
I
I+
I
I
I
.,
ll
"
VI
I
I
I I
v21
_I
I
I
I
I
v2
II
+
I
Z'c
v"l
21
v"l
21
.,
v2
r.+:
~I
1--- ---- - ------ - ---I
z;
VI
+I
I
t I
V2 1
I
1 -
..._,.=...
I
I
I
I
(a}
(b)
(a) Short circuit in series connection. (b) Transformer to avoid short
circuit.
Figure 45
r~------- - ------~--
lt
1 lt
12
I
i2
I ~
~~~
~
1
I
I
I
Vt
I
I
I
+,
v'I
+,
[h']
~2
>---
.,
..
lt
___,__...
I
I +,
:VI
[h'j
I
I
.,
I
I
l2
v2
-
Figure 4-6
v2
I
I
I
I
I- - -
- - - - - - - - - - - - - - - _I
Connection of two-port networks suitable for hybrid representation.
the same (i.e., v2 = v2 ' = v/' and i 1 = i 1' = it"). From this observation we can
conclude that the resulting h-matrix for the overall system is equal to the sum of the hmatrices of the individual networks:
{ ~l }
= {
l2
~< + ~1~'} =
l2 + 12
[hll: ++ hll: hl2: ++ hl2:lj{ il }

h21
h21
h22
h22
(4.18)
V2
An example of this type of connection 1s the Darlington transistor prur Q1 and

Q2 shown in Figure 4-7.
4.2.2
Parallel Connection of Networks
A parallel connection of two dual-port networks is shown in Figure 4-8 for the
admittance matrices Y' andY", where, unlike (4.16), the currents are now additive
155
........................
.......................'
.
\ -
- :._:.:_:_..:.. :::_:_ -
- I
Figure 4-7 Series connection of two hybrid networks.
(4.19)
and the new admittance matrix is defined as the sum of the individual admittances
y 11 + y 11 y 12 + y 12
[Y] = [Y' ] + [Y"]=
[Y21 , + Y 21" Y 22 ' + Y 22 "
I
i{
+
port 1
----'-+
v;
Figure 4-8
4.2.3
[y']
.,
(4.20)
v2
port 2
ii'
II
----'-+
v;'
fl
[Y"]
v2
Parallel connection of two two-port networks.
Cascading Networks
The ABCD-parameter description is most suitable when cascading networks. as

depicted in Figure 4-9 for the example of a two-transistor configuration. In this case the
cment on the output of the first network is equal in value, but opposite in sign, to the
input current of the second network (i.e., i/ = -i 1" ). The voltage drop v 2 ' across the
output port of the first network is equal to the voltage drop v 1" across the input port of
the second network. Thus, we can write the following relations:
156
v"1
+
port 1
i'
._.1...
v;
[A'c' D'BJ
i"
!-4
v"l
v2
[A"
B"]
C"D"
+
v2' port 2
Figure 4-9
Cascading two networks.
(4.21)
The overall system ABCD-matrix is equal to the product of the ABCD-matrices of the
individual networks.
4.2.4
Summary of ABCD Network Representations
As we will see in subsequent chapters, microwave circuits can usually be represented as the result of cascading simpler networks. It is therefore important to develop
ABCD-matrix representations for simple two-port networks that can be used as building blocks of more complex configurations. In this section several examples are considered for which we will derive ABCD-parameters such as transmission line, series
impedance, and passive T-network. Other very useful circuits, such as parallel impedance, passive pi-network, and transformer, are left as exercises at the end of this chapter
(see Problems 4.10-4.12). The results of all the computations are summarized in Table
4-1 at the end of this section.
157
----------------------------~&JA~
Example 4-4: ABCD network representation of an impedance
element
Compute the ABCD-matrix representation for the following network:
i,
l2
VI
v2
Solution:
Guided by the definition (4.1 0), to detennine parameter A we have to compute the ratio of the voltage drop across port 1
to the voltage drop across port 2 when the current into this port is
equal to zero (i.e., port 2 is disconnected). In this case, it is apparent
that for the circuit under consideration, the voltages on both ports
are equal to their ratio, which is equal to unity
vl
=1
A=v2 .
'2
=0
To obtain the value forB, we need to find the ratio of the voltage
drop across port 1 to the current flowing from port 2 when the tenninals of port 2 are shortened. From the circuit topology, this ratio is
equal to the impedance Z:
B=
vl
-l2
=Z
v2 = 0
The remaining two parameters are found according to (4.1 0) of the

ABeD-representation and can be shown to be
C =
lt
v2 .
12
=0
0 and D =
11
-z2
v2 = 0
The ABCD-matrix coefficients are determined in a similar

manner as the previously discussed Z-, Y-, and h-matrix coefficients.
158
Chapter 4 Single- and Multlport NetwOfU
The accurate prediction of the coefficients again depends on the

ability to enforce open- and short-circuit terminal conditions.
In the following example the ABCD-parameters of the passive T-network are

determined. In the derivation of the parameters we will rely on the knowledge of ABCD
parameters for series and parallel connections of the impedance.
----------------------------~&JA~
Example 4-5: ABCD matrix computation of a Tnetwork
Compute the ABCD-matrix representation for the following T-network:
Solution:
This problem can be solved using two different
approaches. The first approach involves directly applying the definition of the ABCD-matrix coefficients and compute them as done in
the previous example. Another approach is to utilize the knowledge
of the ABCD-parameters for parallel and series connections of a single impedance. If we choose this method, we first have to break the
initial circuit into subcircuits as follows:
ZB
ZA
''
!V2
v, ~
\.
Network A
1~ \...
Network C
NetworkB
.)
159
As discussed previously, the ABCD-matrix representation of the

entire circuit is equal to the product of the ABCD-matrices of the
individual subcircuits. Using the results from Example 4-4 and
Problem 4.8, we can write
1+~
Zc
[ABCD]
l
Zc
1+~
Zc
Here we see the advantage of using the ABCD-matrix representation in that a more complex network can be constructed by
multiplication of simpler building blocks.
As a last example, let us consider the computation of the ABCD parameters for a
transmission line.
----------------------------~~~
Example 4-6: ABCD-matrix coefficient computation of a
transmission line section
Compute the ABCD-matrix representation of the following transmission line with characteristic impedance Z 0 , propagation constant ~ , and length /.
Solution:
Similar to Example 4-4, we have to apply open- and
short-circuit conditions at port 2. For a transmission line these con-
Chapter 4 Single- and Multfport Netwol1cl.
160
ditions are equivalent to the analysis of open- and short-circuit stub

lines. Such lines are simply the open/short-circuit transmission line
representations discussed in Sections 2.9.3 and 2.9.2. In these sections we found that for the open-circuit stub the voltage and current
are given by the following expressions [see (2.71) and (2.72)]:
V(d)
= 2V+ cos(Pd) and l(d) =
2'V+
sin(~d)
where distance d is measured from the open port (i.e., in our case
from port 2).
For a short-circuit stub of length l voltages and currents are
determined by (2.67) and (2.68):
V(d)
= 2jV+ sin(~d)
2
and l(d) = ;
cos(~d)
where distance d is again measured from port 2 to port 1. In addition

to these relations, it is important to recall that the current is defined
as flowing toward the load. Therefore, the current is equal to i 1 at
port 1 and equal to -i 2 at port 2.
Having detennined the relations for voltages and currents, it is
now possible to establish equations for the ABCD-parameters of the
transmission line. Parameter A is defined as the ratio of the voltages
at ports 1 and 2 when port 2 is open (i.e., we have to use the formulas for the open-circuit stub):
A
= 2V+ cos(~/) = cos(~l)
= v1
v2 .
12
=0
2V+
where we employ the fact that d = 0 at port 2 and d = l at port 1.

Parameter B is defined as the ratio of the voltage drop across
port 1 to the current flowing from port 2 (i.e., toward the load) when
port 2 is shorted. For this case we have to use the formulas for voltage and current defined for a short-circuit stub. This yields
B
vt
= -.
-l2
v2
-_ 2jV+ sin(~/) -_ Jz0 sm. (Rf)

,..,
2V+ IZo
The remaining two coefficients are obtained in a similar manner:
Network Properties and Appltcatlons
161
~v sin(~!)
0
-----
2V+
2V+
z-cos(~l)
0
2V+
- cos (~l)
Zo
Thus, a transmission line with characteristic impedance Z 0 , propagation constant ~ , and length l has the following matrix representation:
A
[C
Bl
nj
= [ cos(~l) jZ 0 sin(~l}l
jY0 sin(~l) cos(~/)
The ABCD transmission line representation has the expected

periodic parameter behavior similar to the line input impedance
formula derived in Chapter 2 .
.?
In Table 4-1 six of the most common circuit configurations are summarized in
Htenns of their ABCD two-port network representations. From these six basic models,
f!more complicated circuits are readily constructed by suitably combining these elemenfltary networks.
..
!l
; !4.3 Network Properties and Applications
t (
;
,!
~
ft
~
4.3.1
Interrelations between Parameter Sets
Depending on the particular circuit configuration, we may be forced to convert

f. between different parameter sets to arrive at a particular input/output description. For
t.' instance, the low-frequency transistor parameters are often recorded in h -matrix form.
However, when cascading the transistor with additional networks, a more useful
!
f ABCD-matrix form may be appropriate. Thus, converting the h-matrix into an ABCDf ~trix form and vice versa can greatly simplify the analysis .
i
.~
r...
L
Chapter 4 Single- and Multipart Networkt
162
Table 4-1
ABCD..Parameters of Some Useful Two-Port Circuits.
Circuit
ASCD-Parameters
A= 1
c~
A= 1
C= y
B~
D= 1
B~
D= 1
A= 1 + ____:!
Zc
c~ _!_
Zc
D= 1 + _!!.
Zc
1
B=-
Yc
y
D= 1+~
Yc
;,
,...
;2
A= cos~~
V1
c~ jsin~!
'4 ~
V~
Zo
B= }Z0 sin~l
D= cos~/
A= N
C= 0
B= 0
1
D= N
To show how the conversion between the individual parameter sets can be accorn~
plished, let us find an ABCD-matrix representation of a given h-matrix. From the defi
nition (4.11) we can express parameter A as follows
(4.22)
Network Properties and Applications
163
In this expression we are able to re-express the current i 1 in ( 4. 11) in terms of the voltage v2 because i 2 = 0. The result is
(4.23)
= h 11 h 22 -
h 12 h 21 denotes the determinant of the h-matrix. Similarly, for

the remaining coefficients we compute
where Ah
h 11 i 1
= ---
VI
= --:-
'z
'z
=0
v2
hll(::J
lz
hl]
= --
(4.24)
hzt
-h22
--vz
C=
ll
vz 12:::0
.
hzt
vz
h22
(4.25)
h21
'2
't
D = -:-
lz
hzl
lz
=0
v2
(4.26)
h21
This concludes the conversion from h-parameters to ABCD form. A similar procedure
could have been performed from ABCD-parameters to h-matrix form.
As an additional case, let us investigate the conversion from ABCD-parameters to
the Z-representation. Starting with (4.2) and using (4.11), we can develop the following
relations:
z11 = v-l
ll
i2 =:: 0
Av2
Cv2
- c-A
(4.27)
BC
Zn
= vl
Av2 - Bi 2
c
-v2
l2 il ::: 0
Z21
vz
ll
D
=
i2:::::
Av2- D Vz
v 1/
Cv 2
c
-v2
AD-BC
AABCD
(4.28)
Av2 /A
Cv 2
= c
(4.29)
164
(4.30)
where MBCD = AD- BC is the detenninant of the ABCD..matrix.

By relying on the respective defining voltage and current relations, it is relatively
straightforward to work out all parameter conversions. For convenience, Table 4-2 summarizes the formulas for the previously defined four network parameter sets (see also
Appendix H for a complete list of all conversion formulas).
Table 4-2
Conversion between Different Network Representations
Z22
zll z 12
(Z]
[h)
[Z]
Z21 Z22
t;Z
z12
-t;Z
Zu
Z21
-t;Z
- -
tlZ
y22
yll
y21
[h]
h12
h22
h22
1
h21
-h22
h22
AMBCD
[ABCD]
4.3.2
z12
-
Z22
Z22
Z22 Z22
yll
yl2
-y 11
y2l
-yll -ar
yll
y 21 y22
LlY
tJz
-
t;Z
Z21
yll y12
- - -~y
-- -
YI2
-i\Y LlY
[ABCD]
hu
--B1
hll hl2
tlh
A
B
1 Z22
Z21 Zzt
y22
y21
y21
----
ar Yu
---y2l
y21
-!lh
--h21 h22
hll
D MBCD
- B
B
Z21 Z21
hu
hl2
1
-hll hll
h21
zll KZ
-
h21
h2l
1
-h22
--h21
h21
B MBCD
D
1
-D
AB
CD
Analysis of Microwave Amplifier
In this section we consider, by way of an example, the usage of the conversion

between different network representations to analyze a relatively complicated circuit.
Basis of the analysis is the circuit diagram of a particular microwave amplifier shown in
Figure 4-10.
Ntlwork Properties and Applications
1&5
Figure 410
Microwave amplifier circuit diagram.
The first step is to break down the circuit into smaller, simpler subnetworks. This
can be accomplished in several ways, one of which is shown in Figure 4-11.
Feedback loop
r-----R-------1
Input matching
network
-----------
I
I
~------------J
I
I
I
I
Output matching
network
~-- ---- ---- -- --- ----- ----- :
L
L
~~
r-------,_--_-__.~---H~~
~
:c::::Jt:~~~~:~~
i ~l. Zo l
I
l
I
I
------------
'-----------'
'--------------------------Subnetwork representation of the microwave amplifier.
Figure 4-11
i.
As shown in this figure, the amplifier is divided into a set of four subcircuits. The
input matching network consists of a transmission line (for convenience only the upper
trace is shown) and is cascaded with a parallel combination of the transistor and a feedback loop. This circuit is then cascaded with an output matching network.
For the transistor we will use a high-frequency hybrid pi-network model (see also
Chapter 7), which is shown in Figure 4-12.
[
~.
f.
18:
-- - - -- - - -- - ---- - ---- -- - - - - --
B ___. :
CBC
: 1c
.~
E~------~~--~------------~~----~--------4
L-- -- - - - -- --- -- - -- - -- - - - -- -- 1
Figure 4-12
High-frequency hybrid transistor model.
Chapter 4 Slngl and MuiUport
1H
The derivation of the h-parameters is left as a problem (Problem 4.13 at the end~
this chapter). Here we only list the resulting h-matrix for the transistor:
jroCBcrBE
h22
h12
= hre = - - - - - - - 1 + jro(CBE + Coc)rBE
h21
= h fe
rBE(8m- }roCBe)
= _............;;;;.,__,;,:,:___ ____;;__
1 + jro(CBE+ CBc)rBE
gmrBE+
jroCBErBE)
= - 1 + jroCBc(l
_ _...;;...__+_
___
_....;;,...;..__:._.::.._
=h
oe
rcE
l+jro(CBE+Coc)rBE
To compute the matrix for the parallel combination of the transistor and the feed .
back loop resistor we have to convert the h-matrix into a Y-matrix called [Y]tr in ord
to apply the summation rule (4.20). To accomplish this, we can use formulas fro
Table 4-2 and add the result to theY-matrix of the feedback resistor. The admittan
matrix for the feedback resistor can be derived either directly using the definition oftli~
Y-matrix or by converting the ABCD-parameters derived in Example 4-4 into the Y4
form. The result of these computations is
(4.32)
After the summation we obtain the admittance matrix for the parallel combination ol
the transistor and the feedback resistor [ Y] tr + R
-~
The same result could have been obtained if we had noticed that the feedback
resistor is connected in parallel with the capacitor C BC of the transistor. Thus, to obtaitl
the admittance matrix of the parallel combination of the feedback resistor and the tran
sistor, we simply need to replace CBe in the h-matrix of the transistor with
CBC + 1 I (jroR) and then convert the resulting matrix into -representation.
The final step in the analysis is to multiply the ABCD-matrices for the input
matching network (index: IMN), the transistor with feedback resistor (index: tr + R),
and the output matching network (index: OMN)
(4.33)
Network Properties and Applications
167
where the ABCD-matrices for the matching networks are found using the results from
Table 4-1:
[~~LN
[~ ~tMN
cos~l
jsin~l
Zo
2
1- ro LC
jroC
jZ 0 sin~~
(4.34)
cos~l
2jroL- j u/ L
2
1- ro LC
(4.35)
Due to rather lengthy expressions we are not presenting the final result for the
ABCD-parameters of the entire amplifier. Instead we urge the interested readers to perfonn these computations by relying on a mathematical spreadsheet program of their
choice (MathCad, MATLAB, Mathematica, etc.). One of the results of these computations is shown in Figure 4-13, where the small-signal current gain for the amplifier with
short-circuited output (inverse of the D-coefficient) is plotted versus frequency for different values of the feedback resistor.
4o ======~===-7.-==~====~--~~~--~~
R = lkn~=]OkQ
I
f
3s t--_
.s
~
E
30 -
r--
.R=soo n ~
R = 3oo
R = 200 Q
~
~
-----
g 25
()
l0
I0
Frequency, Hz
Figure 4-13
Small-signal current gain of the amplifier versus frequency for

different values of the feedback resistor.
Chapter 4 Single. and Muttlport Networks
168
The computations are based on the circuit in Figure 4-11 with L = l nH,
C = 10 pF, transmission line length of l = 5 em, and phase velocity equal to 65% of
the speed of light. The transistor is described by the following set of values:
rBE = 520 n, rcE = 80k n, CBE = 10 pF, CBC = 1 pF,and gm = 0.192 s.
4.4 Scattering Parameters

In almost all databooks and technical literature regarding RF systems, the scattering or S-parameter representation plays a central role. This importance is derived from
the fact that practical system characterizations can no longer be accomplished through
simple open- or short-circuit measurements, as it is customarily done in low-frequency
applications and as discussed at the beginning of this chapter. We should recall what
happens when we attempt to create a short circuit with a wire: The wire itself possesses
an inductance that can be of substantial magnitude at high frequency. Also, the open
circuit leads to capacitive loading at the terminal. In either case, the open/short-circuit
conditions needed to detennine Z-, Y-, h-, and ABCD-parameters can no longer be guaranteed. Moreover, when dealing with wavepropagation phenomena, it is not desirable
to introduce a reflection coefficient whose magnitude is unity. For instance, the terminal
discontinuity will cause undesirable voltage and/or current wave reflections, leading to
oscillations that can result in the destruction of the device. With the S-parameters, the
RF engineer has a tool to characterize the two-port network description of practically
all RF devices without requiring unachievable tenninal conditions or causing harm to
the device under test (OUT).
4.4.1
Definition of Scattering Parameters
Simply put, S-parameters are power wave descriptors that pennit us to define the
input-output relations of a network in terms of incident and reflected power waves.
With reference to Figure 4-14 we define an incident nonrtlllized power wave an and a
reflected normalized power wave b n as follows:
an
bn
(4.36a)
1
17(Vn -Zoln)
(4.36b)
2 -JL-0
where the index n refers either to port number 1 or 2. The impedance Z 0 is the charac~
teristic impedance of the connecting lines on the input and output side of the network.
Under more general conditions the line impedance on the input side can differ from the
169
Scattering Parameters
line impedance on the output side. However, for our initial discussion, we will keep
things simple and assume that both impedances are the same.
[S)
Figure 4-14
Convention used to define 8-parameters for a two-port network.
Inverting (4.36) leads to the following voltage and current expressions:
vn =
JZo(an + bn)
(4.37a)
1
(4.37b)
!7(an-bn)
A./Zo
The physical meaning of (4.36) becomes clear when we recall the equations for power:
In=
Pn =
~Re{Vnl:}
~(lanj 2 -lbnl 2)
(4.38)
Isolating forward and backward traveling wave components in (4.37), we immediately
see
(4.39a)
v-
bn =
fzo = -JZol~
(4.39b)
which is consistent with the definitions (4.37) since

(4.40)
Based on the directional convention shown in Figure 4-14 we are now in a position to
define the S-parameters:
(4.41)
where the terms are
Chapter 4 Single- and Multlport Networb
170
= bt
a1
= b2
a1
s
22
S
12
= b2
a2
= b1
a2
a2 = 0
a2
=0
a 1= 0
=reflected power wave at port 1

- incident power wave at port 1
(4.42a)
=transmitted power wave at port 2

incident power wave at port 1
-
(4.42b)
=reflected power wave at port 2

- incident power wave at port 2
(4.42c)
=
a1 = 0
transmitted power wave at port 1

incident power wave at port 2
(4 .42d)
We observe that the conditions a 2 = 0 and a 1 = 0 imply that no power waves are
returned to the network at either port 2 or port I. However, these condition can only be
ensured when the connecting transmission lines are tenninated into their characteristic
impedances.
Since the S-parameters are closely related to power relations, we can express the
nonnalized input and output waves in terms of time averaged power. With reference to
Section 2.10.2 we note that the average power at port 1 is given by
P1 =
~~~;1\I-Ir;il = ~~~;I\I-1Sill 2 l
0
(4.43)
where the reflection coefficient at the input side is expressed in terms of S 11 under
matched output according to the following argument:
bi
r.m = -V!
=
y+
at
1
= s 11 .
(4.44)
a2 = 0
This also allows us to redefine the VSWR at port 1 in terms of S 11 as
ISlll
1-ISlll
1+
VSWR =
(4.45)
Furthennore, based on (4.39a) we can identify the incident power in (4.43) and express
it in terms of a 1 :
(4.46)
Scat!erlng Parameters
171
which is the maximal available power from the generator. Using (4.46) and (4.44) in
(4.43) finally gives us the total power at port 1 (under matched output condition)
expressed as a combination of incident and reflected powers:
2
P1
latl -1rinl 2 )
= Pine+ Pref = 2Cia11 - Ib1l 2 ) = -2-(1
1
(4.47)
If the reflection coefficient, or S11 , is zero, all available power from the source is delivered to port 1 of the network. An identical analysis at port 2 yields
P2
4.4.2
la2l
2
= 2<!a2! 2-1h2l 2 ) = 2(1-lro
utl )
1
(4.48)
Meaning of $-Parameters
As already mentioned in the previous section, the S-parameters can only be determined under conditions of perfect matching on the input or output side. For instance, in
order to record S11 and S 21 we have to ensure that on the output side the line impedance Z 0 is matched for a2 = 0 to be enforced, as shown in Figure 4-15.
[S]
a,
-:0
:.----_z_o
z,
~
Figure 4-15
Measurement of 5 11 and
~ 1 by
matching the line impedance Zo at
port 2 through a corresponding load impedance ZL = Zo
This configuration allows us to compute S 11 by finding the input reflection

coefficient:
z. - Z 0
s11 = r in = Z_in_+_Z_o
m
(4.49)
In addition, taking the logarithm of the magnitude of S11 gives us the return loss in dB
RL = -20log!S 11 1
Moreover, with port 2 properly terminated, we find
(4.50)
172
v;; JZo
(4.51)
Since a 2 = 0 , we can set to zero the positive traveling voltage and current waves at
port 2. Replacing V 1 by the generator voltage V G! minus the voltage drop over the
source impedance Z 0 , V Gl- Z 0 1 1 gtves
2V2
2V 2
V Gl
V Gl
S21 = - =
(4.52)
Here we observe that the voltage recorded at port 2 is directly related to the generator
voltage and thus specifies the forward voltage gain of the network. To find the forward power gain, we square (4.52) to obtain
Go = IS2tl2 =
v2
(4.53)
VGl/2
If we reverse the measurement procedure and attach a generator voltage V G2 to

port 2 and properly terminate port l , as shown in Figure 4-16, we can detennine the
remaining twoS-parameters, S22 and S 12 .
[S]
~I
Figure 416 Measurement of ~2 and 5 12 by matching the line impedance~ at
port 1 through a corresponding input impedance ZG =Zo
To compute S22 we need to find the output reflection coefficient rout in a similar
way as already discussed for S11:
s22
and for
s12
r out -
(4.54)
kltW!ng Parameters
173
V~/ flo
(4.55)
The term S 12 can further be manipulated through the substitution of V 2 by

VG2 - Z0 / 2 , leading to the form
s1z =
2V~
V G2
2V 1
(4.56)
V G2
2
known as the reverse voltage gain and whose square jS12 j is identified as reverse
power gain. While detennining S 11 and S22 can be directly computed as part of the
impedance definitions, S 12 an S21 require the replacement of the defining voltages by
the appropriate network parameters. In the following example, the S-parameters are
computed for a simple, three element network.
-----------------------------~&)A~
Example 4-7: Determination of a T-network elements
Find the S-parameters and the resistive elements for the 3 dB attenuator network shown in Figure 4-l7(a) assuming that the network is
placed into a transmission line section with a characteristic line
impedance of Z 0 = 50 Q .
Solution:
An attenuator should be matched to the line impedance and must therefore meet the requirement S11 = 5 22 = 0. As a
result, based on Figure 4-17(b) and consistent with (4.49), we set
z in
Rt + (
R3 (R 2 +50 Q)
R 50 Q = 50 Q
R3 + 2 +
)
Because of symmetry, it is immediately clear that R 1 = R 2 . We
now investigate the voltage V 2 =
at port 2 in terms of
V1 =
According to the circuit configuration shown in Figure
4-17(c), the following expression is obtained
v; .
v;
Chapter 4 Single- and Multlpon Networks
174
R3
Port I
Port 2
(a)
R1
R2
= fR~ ~}son
Port 1
Port 2
Port 2
Port 1
(c)
(b)
Figure 4-17 S.parameter computation for a T-network. (a) circuit diagram;

(b) circuit for S11 and S21 measurements; (c) circuit for S12 and S22 measurements.
R3(R 1 +50 Q)
R3 + R 1 +50 Q
R 3 (R 1 +50 0)
-----+Rt
R 3 + R 1 +50 Q
For a 3 dB attenuation, we require
2V2
v2 1
S21 = = - = - = o.707 = s 12
VGl
Vl
J2
Setting the ratio of V 2 /V 1 to 0.707 in the preceding equation
allows us, in combination with the input impedance expression, to
detennine R 1 and R 3 After simplification it is seen that
R1
= R 2 = J2-l
M
Z 0 = 8.58 Q
AJ2 + 1
and R 3 = 2AJ2 Z 0
= 141.4 Q
The choice of the resistor network ensures that at the input and
output ports an impedance of 50 Q is maintained. This implies that
this network can be inserted into a 50 Q transmission line section
without causing undesired reflections. resulting in an insertion loss.
scattering Parameters
175
The definitions for the S-parameters require appropriate termination. For instance,
if S11 is desired, the transmission line connected to port 2 has to be terminated into its
characteristic line imped ance. This does not necessarily mean that the output impedance
Zout of the network has to be matched to the line impedance Z 0 . Rather, the line impedance must be matched to ensure that no wave is reflected from the load, as implied by
a2 = 0 . If this is not the case, we will see in Section 4.4.5 how S11 is modified.
4.4.3
Chain Scattering Matrix
To extend the concept of the S-parameter re presentation to cascaded networks, it

is more efficient to rewrite the power wave expressions arranged in terms of input and
output ports. This results in the chain scattering matrix notation. That is,
(4.57)
It is immediately seen that the cascading of two dual-port networks becomes a simple
multiplication. This is apparent in Figure 4- 18, where network A (given by matrix [T]A)
is connected to network B (given by matrix [T] 8 ).
bA
bB
as
:+1
aA
a4
--tt+
port 1:
[TJ4
~~
.-ftbA1
b.4
as
Figure 4-18
(T]B
>ort2
-1++
bB
2
Cascading of two networks A and B.
If network A is described by the relation
(4.58a)
and network B by
{:n
[~!~ ~!21{ :~}

21
22J
we notice, based on the parameter convention shown in Figure 4-18, that
(4.58b)
176
Chapter 4 SlngJe. and Murtlport Networks
(4.59)
Thus, for the combined system, we conclude
:~ } = [~1 T:2l [~1 ~2l{ :~ }

1
(4.60)
T 21 T 2J T 21 T 22J
which is the desired matrix multiplication. Therefore, the chain scattering matrix plays
a similar role as the ABCD-matrix discussed earlier.
The conversion from the S-matrix to the chain matrix notation follows identical
steps as outlined in Section 4.3.1. In particular, to compute T 11 for instance, we see that
(4.61)
Similarly,
(4.62)
(4.63)
(4.64)
Conversely, when the chain scattering parameters are given and we need to convert to
S-parameters, we find the following relations:
SII
sl2
bl
-al
Tztbz
a2 = 0
Tzl
Tll
tJ.T
= T11T22-T21T12
Tu
Tn
1
Tll
Tl2
S22
= T11b2
= -Tu
(4.65)
(4.66)
(4.67)
(4.68)
k:ltterlng Parameter
177
.\ltematively, a matrix manipulation as discussed in the next section could have been
;arried out with the same result.
4.4.4
Conversion between Z and 5-Parameters
We have already seen how certain S-parameters can be defined in terms of input
md output impedances of a network [i.e., equations (4.49) and (4.54)]. In this section,
we go through a formal conversion between the Z- and S-parameter sets. Once this
interrelation is established, we are able to formulate conversion links between all six
network parameter sets (S, Z, Y, ABCD, h, T).
To find the conversion between the previously defined S-parameters and the Zparameters, let us begin with the defining S-parameter relation in matrix notation [i.e .
(4.41)]
{b}
Multiplying by
(4.69)
JZo gives
jZo{b}
Adding { y+}
= [S]{a}
= {V- } = jZo[S]{a} = [S]{V+}
(4.70)
= JZo {a} to both sides results in

{V} = [S]{V+} + {V+} = ([S] + [E]){V+}
(4.71)
where [E] is the identity matrix. To compare this form with the impedance expression
{V} = [Z 1{I} , we have to express { V+ } in terms of {I} . This is accomplished by
first subtracting [S]{V+} from both sides of {V+} = /Zo{a}; that is,
{V+}- [S] {V+}
= JZo< {a}- {b})
= Z 0 {1}
(4.72)
Now, by isolating { v+} , it is seen that

1
{V+} = Z 0 ([E]- [S])- {I}
(4.73)
Substituting (4.73) into (4.71) yields the desired result of

1
{V} = ([S] + [E]){V+} = Z 0 ([S] + [E])([E]- [S])- {I}
(4.74)
or
[Z] = Z 0 ([S] + [E])([E]- [S])-
Explicit evaluation yields
(4.75)
178
(4.76)
=
Identifying individual terms is now easily carried out. A complete summary of all net~
work coefficient sets is given in Appendix C.
4.4.5
Signal Flow Chart Modeling
The analysis of RF networks and their overall interconnection is greatly facilitated

through signal flow charts as commonly used in system and control theory. As originally introduced to seismology and remote sensing, wave propagation can be associated
with directed paths and associated nodes connecting these paths. Even complicated networks are easily reduced to input-output relations in which the reflection and transmission coefficients play integral parts. In this section we will briefly summarize key
principles needed for a signal flow network analysis.
The main concepts required to construct flow charts are as follows:
1. Nodes that are deployed to identify network parameters such as a 1, b 1, a 2, b2

when dealing with S-parameters
2. Branches that are needed when connecting the network parameters
3. Addition and subtraction of branch values in accordance with the directions of the
branches
We will now discuss these three items in detail. To this end let us consider a section
of a transmission line that is terminated in a load impedance Z L , as seen in Figure 4-19.
----o
----
Zo
-------
ZL
b+-H(a)
(b)
Figure 4-19 Terminated transmission line segment with incident and reflected S
parameter description. (a) Conventional form, and (b) Signal flow form.
Scltttt1ng Parameters
179
Even though we could use voltage values as node identifier, it is the S-parameter
representation that finds widespread use. In Figure 4-19(b) the nodes a and b are connected through the load reflection coefficient r L . This makes sense since the reflection
coefficient is the ratio b/a, so that it simply states that node b is found as a result of multiplying node a by r L. This is depicted in generic form in Figure 4-20.
b
a
0
(a) Source node a, which launches wave.
(b) Sink node b, which receives wave.
(c) Branch connecting source and sink.
Figure 4-20
Generic source node (a), receiver node (b), and the associated (c)
branch connection.
In terms of notation, we can encode the situation shown in Figure 4-20 as

b
= ra
(4.77)
A more complicated situation arises when we need to make the transmission line
circuit shown in Figure 4-19 more realistic by including a source term, as seen in Figure 4-21.
Vo~4~
:]z,
a'
bs
(b)
(c)
Terminated transmission line with source. {a) conventional form, (b)
signal flow form, and (c) simplified signal flow form.
{a)
Figure 4-21
Unlike Figure 4-19, the nodes a and bare preceded by two additional nodes that
we shall denote a' and b'. The ratio b'Ia' defines the source reflection coefficient r s as
already discussed in Section 2.11. Here we also see that b' is given by multiplying a'
with the source reflection coefficient. By relying on the concept of summation, we
define b' as the sum of b 5 and a'r s. Thus, the source b 5 is
b5 = b'- a'r s
(4.78)
180
Chapter 4 Single- and Murtlport Networks
An explicit expression for b s is obtained by noting that

(4.79)
Vs = V G + IGZG
based on an outflowing current convention (see Figure 4-21). This can be converted into
the form
+
(v; v:s:J
(4.80)
Vs + Vs = VG + ZG Zo- Zo
JZo gives
______:JZo
.__ v = -v:s: - r -v;
ZG + Zo
JZo s JZo
Rearranging terms and division by
(4.81)
When comparing ( 4.81) with (4. 78), we immediately see that
bs
JZo
= ZG + Z0 V G
(4.82)
An important conclusion can be drawn when expressing a' in (4.78) by

obtain
b'
r Lb'
= b5 + rLfsb' = 1- bs
rLrs
so that we
(4.83)
This is a known as a self- or feedback loop (see Figure 4-22), which allows us to represent the nodes b 5 and b' by a single branch whose value is given by (4.83).
1/(1 -
Figure 4-22
b'
bs
r;_ r:~)
0
b'
A self-loop that collapses to a single branch.
All signal flow chart principles can therefore be reduced to six building blocks, as
summarized in Table 4-3.
By way of an example, let us analyze a more complicated RF circuit consisting of

a sourced and terminated dual-port network.
181
Table 43
Signal flow chart building blocks
Description
Nodal Assignment
Graphical Representation
------Zo
----Zo
-------
-----
====:>
--
.-oa
- _,.
b
0 ..
Branch
Series Connection
.---
:::~z, = ~~=:Jr,
sba
..
scb
sbo scb
sl
Parallel Connection
-Q.
s2
Splitting of Branches
:: ;,
Selfloop
=:>
cs
b
S2
t ~3 c ..
0
S1+S2
0
a b
.-o
=:>
a: s,:
..
s3
:: +s3
11(1-r)
0
a
c
0
..
---------------------------------~&M~
Example 4-8: Flow chart analysis of a dual-port network
For the network shown in Figure 4-23 find the ratios of b 1 I a 1 and
a 1I bs . Assume unity for the multiplication factor of the transmission line segments.
182
Chapter 4 Single- and Murtlport Networb
(a) Circuit representation
~s ~s I ~ j' ~
Su
21
I< I
r,
bl s,2 a2 1
(b) Signal flow chart
1
Figure 4-23
Sourced and terminated two-port network.
Solution:
The process of setting up the individual ratios is
explained best by going through a step-by-step simplification for the
ratio a 11bs employing the rules summarized in Table 4-3. Figure 424 depicts the five steps.
Step 1: Splitting of the rightmost loop between b 2 and a 2 , leading
to the self-loop S22rL
Step 2: Decomposition of the self-loop between branches a 1 and b 2 ,
resulting in the multiplication factor s21 I ( 1 - s22 r L) ' which can be
combined with L and S 12
Step 3: Series and parallel connections between a 1 and b 1 , leading
to the input reflection coefficient
r in
bl
sl2s21
= - = S 11 + 1 S r
at
22 L
rL
Step 4: Splitting the loop into a self-loop, resulting in the multiplication factor
S12S21
)
( u + I - 22 r L r L r s
Step 5: Decomposition of the self-loop at a 1 , leading to the

expressiOn
183
Bcllt8rtng Parameters
~s ! I
rst
; ai, S~,~
<ji; H
b I S 12 :: ...a 2 .......... .......:.
Step 1
Step 2
Step 3
Step 4
Figure 424
Step 5
Step~by-step simplification to determine the ratio a 1I b s .
184
Chpter 4 Single- and Multi port Networkl
Rearranging and simplification leads to the final form:
The preceding derivation follows a pattern similar to finding

the transfer function of a control system or a signal processor. Even
complicated circuits can be reduced efficiently and quickly to establish the nodal dependcies.
The preceding example points out what will happen if the matching condition for
recording the S-parameters is not satisfied. As we know, if we compute S11 we need to
ensure that a 2 = 0 . However, if a 2 '# 0 , as is the case in the preceding example, we see
that S11 is modified by the additional factor S 12S21 L/ ( 1 - S22rL).
4.4.6
Generalization of S..Parameters
In our discussion thus far it was assumed that the characteristic line impedance at
both ports has the same value Z 0 . However, this does not have to be the case. Indeed, if
we assume that port 1 is connected to line impedance Z 01 and port 2 to impedance
Z 02 , we have to represent the voltage and current waves at the respective port
(n = I, 2) as
(4.84)
and
v+ vI = - n __
n_ -an- - -bnn
Zon Zon
hh
(4.85)
where we immediately observe
v+
a. -
iz;.
v-
bn =
1z;.
(4.86)
185
These equations allow the definition of the S-parameters as follows:
s ..
I}
V~!)Z;;
b l.
a.
1 a 11
=O(n 1:- j)
v:;
141 v+
1
(4.87)
= o( n 1:-1)
When compared to the previous S-parameter definitions, we notice that scaling by the
appropriate line impedances has to be taken into account. It should also be apparent that
although the focus of our derivations was a two-port network, the preceding fonnulas
can be extended to an N-port network where n = 1, ... , N .
A second consideration is related to the fact that practical measurements involve
the detennination of the network S-parameters through transmission lines of finite
length. In this case we need to investigate a system where the measurement planes are
shifted away from the actual network, as depicted in Figure 4-25.
bs
a2
aJ
Zs -tt+
v.S:: : I~:~~~1 :
ZOI
b~ ~
~b2
z2 0
ozl
-!,
Figure 4-25
Zo2
ZL
= Zo2
- /2
Two-port network with finite-length transmission line segments.
An incident voltage wave launched from the power supply will have to travel a
distance 11 in order to reach port 1. Consistent with the notation introduced in
Section 2.9, we note that at port 1 the incident voltage is given as
(4.88)
and, at the generator side, as
+
+ -j ~) ( -11 )
vin( zl== -L.) = VIe
(4.89)
The reflected voltage wave at port 1 can be cast in the fonn

(4.90)
and
(4.91 )
Chapter 4 Single- and Multlport Networb
186
where, as usual, ~ 1 stands for the lossless propagation constant of line 1. In an identical
fashion, the voltage behavior at port 2 can be formulated by simply replacing V in in
terms of V out and V 1 in terms of V 2 as well as P1 in terms of jl2 The preceding equations can be combined in matrix form
(4.92)
which links the impinging waves at the network ports to the corresponding voltages
shifted by the electric lengths of the attached transmission line segments. For the
reflected voltage waves we get the matrix form
V~( -1
1)
[e-j~ 111
v:utC -12)
J{ V~ }
0
e
-j~2~2
(4.93)
v;
As the discussion in Section 4.4.1 taught us, the S-parameters are linked to the coefficients an and bn, which in turn can be expressed through voltages (if we assume
Zol = Zoz>
(4.94)
It is apparent that if transmission line segments are added, we have to replace the above
voltages by the previously derived expressions, leading to the form
vin(-ll)
-
v~ut(-12)
e- 1 ... 1
'R
e-1'"' 11
s 21 s22
e-)~212
J[
e-)~212
~[
sll s l 2
r.t /
J{
vin(-l i )
+
v~ut(-12)
(4.95)
This final reveals that the S-parameters for the shifted network are comprised of three
matrices. In terms of the coefficients, we see that
-)2 ~ 1 1 1
[S]SHIFT =
S 11 e
s21 e
-)( (}1 1 1 + ~2 12 )
S12 e
-JC~I / 1 + P 212)
-)2~2 /2
(4.96)
s22e
The physical meaning of this form is easy to understand. The first matrix coefficient
reveals that we have to take into account 2J3 1l 1 or twice the travel time for the incident
voltage to reach port 1 and, upon reflection, return. Similarly, for port 2 we see that the
187
Scanerlng Parameters
phase shift is 2~ 2 12 . Moreover, the cross terms, which are closely related to the forward and reverse gains, require the additive phase shifts associated with transmission
line 1 ( ~ 111 ) and transmission line 2 ( f3 2 ! 2 ), since the overall input/output configuration now consists of both line segments.
----------------------------~~~
Example 4-9: Input impedance computation of a transmission
line based on the use of the signal flow chart
A lossless transmission line system with characteristic line impedance Z 0 and length l is terminated into a load impedance Z L and
attached to a source voltage V G and source impedance Z a, as
shown in Figure 4-26. (a) Draw the signal flow chart and (b) derive
the input impedance formula at port 1 from the signal flow chart representation.
Vo~ ~~:
bl
1..
Zo
I
]z,
b2
..1
Figure 426 Transmission line attached to a voltage source and terminated by a

load impedance.
Solution:
(a) Consistent with our previously established signal

flow chart notation, we can readily convert Figure 4-26 into the form
seen in Figure 4-27.
Figure 4-27
Signal flow chart diagram for transmission line system in

Figure 4-26.
(b) The input reflection coefficient at port 1 is given by
CMpter 4 Single- ..,d Multlport Networkl
188
bl = f
Le - j2JH al
which is exactly in the form given in Section 3.1' with

l = d . Thus
r L = r 0 and
This example shows how the input impedance of a transmission line can be found quickly and elegantly by using signal flow
chart concepts.
4.4.7
Practical Measurements of S.Parameters
Measurement of the S-parameters of a two-port network requires reflection and

transmission evaluations of traveling waves at both ports. One of the most popular
methods is to use a vector network analyzer. The vector network analyzer is an instrument that can measure voltages in tenns of magnitude and phase. Usually network analyzers have one output port, which provides the RF signal either from an internal source
or an external signal generator, and three measurement channels, which are denoted as
R, A, and B (see Figure 4-28).
The RF source is typically set to sweep over a specified frequency range. The
measurement channel R is employed for measuring the incident wave. Channel R also
serves as a reference port. Channels A and B usually measure the reflected and transmitted waves. In general, the measurement channels A and B can be configured to record
any two parameters with a single measurement setup. An example of the test arrangement that allows us to measure S11 and S 21 is shown in Figure 4-28.
In this case the value of S 11 can be obtained by evaluating the ratio A I R, and
S 21 through computing Bl R. To measure S 12 and S 22 we have to reverse the DUT. In
Figure 4-28 the dual-directional coupler allows the separation of the incident and
reflected waves at the input port of the OUT. The bias tees are employed to provide necessary biasing conditions, such as a quiescent point for the DUT. Since the most com-
189
8c:ltlertng Parameters
0
0
0
0
0
0
RF
000
000
000
c::::::J 0
R A B
Dual-Directional
Dual-Directional
son
Coupler
Coupler
DC :Eb\.\er Supply ...___,
Figure 4-28
Measurement system for S11 and S21 parameters using a network

analyzer.
mon use of network analyzers is the characterization of two-port devices, bias tees,
directional couplers, and necessary electronic switches as well as the RF sweep signal
generator are all integral parts of most modern analyzers.
As we can see, a practical test arrangement is more complicated when compared
with the simple ideal system described in Sections 4.4.4 and 4.4.6, where we assume
that the DUT is connected to perfectly matched transmission lines of equal (Section
4.4.4) or unequal (Section 4.4.6) characteristic impedance. In a realistic measurement
:. system we cannot guarantee either matching conditions or ideality of the components.
In fact, we have to consider all effects of the external components connected to the
input and output ports of the OUT. Furthermore, the primary reference plane for measurements of complex voltages, which are then converted into S-parameters, is usually
somewhere inside of the networks analyzer. As a result, it is necessary to take into
account not only attenuation and phase shifts due to the external components, but also
portions of the internal structure of the network analyzer itself.
In general, the measurement test arrangement can be reduced to the cascade of
three networks depicted in Figure 4-29.
180
Chapter 4 Slngte. and Muttfport Networkt
Error
Error
boxB I
box A
Measurement
Reference Plane
~2
c;+i
'
Measurement
Reference Plane
Desired Reference Plane

/
(a)
Ex
RRm
FEr
12
sl2
(b)
Figure 4-29 (a) Block diagram of the setup for measurement of S-parameters of
a two-port network; (b) signal flow chart of the measurement test setup.
In Figure 4-29 the signals R, A, B correspond to the reference port and channels A
and B of the network analyzer. RFin is the output line from the signal source. The
branch denoted Ex represents possible leakage between the output of the signal source
and the channel B.
The network analyzer treats everything between the measurement reference
planes as a single device. Therefore, our task is reduced to finding a way to calibrate the
network analyzer in such a way that it becomes possible to eliminate the effect of all
undesired influences or parasitics. The main goal of a calibration procedure is to characterize the error boxes prior to measuring the DUT. This information can then be used
by an internal computer to evaluate the error-freeS-parameters of the actual DUT.
Assuming that the error box A network is reciprocal, we can state E 12 21 .
Therefore, we have to find six parameters (E 11 , 12,22, Ex, ER, and Er) to characterize the error boxes.
The simplest calibration method involves three or more known loads (open, short,
and matched). The problem with this approach is that such standards are usually imperfect and are likely to introduce additional errors into the measurement procedures.
These errors become especially significant at higher frequencies. To avoid the dependency on the accuracy of calibration standards, several methods have been developed
(see Eul and Schiek and Engen and Hoer, listed in the Further Reading section at the
lclttertng Parameters
191
end of this chapter). In this section we will only consider the so-called ThroughReflect-Line (TRL) technique (see Engen and Hoer).
The TRL calibration scheme does not rely on known standard loads. Instead, it is
based on the use of three types of connections, which are shown in Figure 4-30.
RFin
Ell
21
FEr
ER
22
.A
1
12
{a) Through
REin
(b) Reflect
RFin
Eu
..A
Figure 4-30
~I
E22
e-i ~t
F Er
ER
(c) Line
Signal flow graphs of TRL method: (a) Through, (b) Reflect, (c) Line
configurations.
The Through connection is made by directly connecting ports 1 and 2 of the DUT.
Next, the Reflect connection uses a load with high reflectivity. The reflection coefficient
4oes not have to be known because it will be determined during the calibration process.
The only requirement is that the load possesses the same reflection coefficient for both
input and output ports. The Line connection is made by connecting ports 1 and 2 via a
transmission line matched to the impedance of the error boxes. Usually, this impedance
Chapter 4 Singl and Muttlport Networkl
192
is close to 50 Q . Before we continue with the actual analysis of each particular connection type, let us first consider the system as a general two-port network.
From Figure 4-29(b) it is seen that the signal at node B is a linear combination of
the input RF signal and the signal at node F:
B = Ex+ ErF
(4.97)
Applying the self-loop rule, we can write that signal at node F as
F=
s21
1- ERS22
(4.98)
To compute the signal at port R, the same method as discussed in Example 4-8 can
be used. In this example we first replaced the loop with the signal F through a self-loop
and then performed the same transfonnation for the signal R. The result of these computations is
R
(4.99)
s12S21ER)
1-22( Sn + 1- ERS22
Substituting (4.99) into (4.98) followed by the substitution of (4.98) into (4.97), we
obtain an expression for signal B:
S
B = Ex+ Er
21
1- ERs22
(4.100)
21
s J2s21ER)
I - E 22 S 11 + 1
E S
R 22
Finally, the value for the signal at node A is obtained by using the summation rule:
E 1221
A= Ell+
1-
22(sn
S21
SSE
Sn+Sl 2ER1 - E S
+ 1 ~ ;~s2:)
R
22
(4.101)
If the measurement system does not introduce any errors, then E 12 == 21 = Er == 1

and 11 = 22 = ER = Ex = 0. Substituting these values into (4.99), (4.100), and
(4.101), we find that R = 1, A = S ll, and B = S 12 , which shows the validity of the
formulas.
Now we are ready to investigate the TRL connections in more detail. To avoid
confusion, let us denote the measured signals R, A, and B for Through by subscript T,
for Reflect by R, and for Line by L.
Sc:ltttrlng Parameters
193
For the Through connection we know that Sn

Setting E 12 = 21 it follows that
= S22 = 0
2
12
AT = 11 + 1 E E ER
-
Br
= S21
= 1.
(4.102b)
22
= Ex+ Erl -
and S12
12
E E
22
For the Reflect connection we have S 11 = S22 =

in the equations
(4.102c)
R
and S 12 = S21 = 0. This results
E12r
AR = Eu + _ Ezzr
1
BR
(4. 103b)
= Ex
(4.103c)
Finally, for the Line connection we see that S 11 = S22 = 0 and S 12 = S21 = e-yl,
where l is the transmission line length and y is a complex propagation constant
(y = ex+ jp) that takes into account attenuation effects. The result is
12
RL =
1- E 22 ERe
(4. 104a)
- 2yl
(4.104b)
BL = Ex+ Ere
- yl
E 12
1- E 22 ERe
_2'Y1
(4.104c)
Equations (4.1 02a)-(4.1 04b) allow us to solve for the unknown coefficients of the error
boxes E 11 , 12 , 22 , Ex, ER, Er, the reflection coefficient r, and the transmission
line parameter e-yl . Knowing the error coefficients we are then in a position to process
the measured data in order to obtain an error-freeS-parameter set of the DUT.
Ct.pter 4 Single- nd Multlport Nltworb
194
4.5 Summary
Networks play an integral part in analyzing basic low-frequency circuits as well as
RFIMW circuits. For instance, the admittance or Y..matrix for anN-port network can be
written in generic form as
't
'2
lN
Yu
yl2
YJN
vl
y21
y22
y2N
v2
y NN
VN
YNI y N2
..
where currents and voltages become the defining external port conditions. The evaluation of the matrix coefficients is accomplished through appropriate terminal conditions:
rn
y nm =Vm vk
=0 (fork;em)
The concepts of Z-, Y-, h-, and ABCD-matrix representations of networks can be
directly extended to high-frequency circuits. Unfortunately, we encounter practical difficulties in applying the required open- and short-circuit network conditions needed
when defining the respective parameter sets. It is for this reason that the scattering
parameters as normalized forward and backward propagating power waves are
introduced:
v-
b. =
io
-JZor.
For a two-port network this results in the matrix form
{:J
[~~: ~~~{::}
Unlike open- or short-circuit network conditions, impedance line matching at the

respective port is now required to establish the S-rnatrix set. The S-parameters can be
directly related to the reflection coefficients at the input and output of the two-port
network (S 11 , S 22 ). Furthermore, forward and reverse power gains are readily identified
2
( IS2tl
'
IS 121 2 ) .
Fufthtr Reeding
195
The S-parameters are also very useful descriptors when dealing with signal flow
diagrams. A signal flow diagram is a circuit representation involving nodes and paths
for the sourced and terminated transmission line as follows:
-j fll
bs 1 a 1
..rol
I~
e-jflt
b2
bl
a2
With signal flow diagrams even complicated systems can be examined in tenns of specific input output relations in a similar manner as done in control system theory.
Chapter 4 finishes with a brief discussion of the practical recording of the Sparameters for a two-port network (DUT) through the use of a vector network analyzer.
To compensate for various error sources associated with the measurement arrangement,
the so-called TRL method is presented. Here the Through, Reflect, and Line calibrations
are shown to account for the various errors and therefore permit the recording of the
actual S-parameters needed to characterize the DUT.
Further Reading
C. Bowick, RF Circuit Design, Howard Sams & Co., Indianapolis, IN, 1982.
R. S. Elliot, An Introduction to Guided Waves and Microwave Circuits, Prentice Hall,

Upper Saddle River, NJ, 1997.
G. F. Engen and C. A. Hoer, "Thru-Refect-Line: An Improved Technique for Calibrating the Dual Six-Port Automatic Network Analyzer," IEEE Trans. Microwave Theory
and Techniques, VoL MTT-27, pp. 987-998, 1979.
H.J. Eul and B. Schiek, "Thru-Match-Reftect: One Result of a Rigorous Theory for Deembedding and Network Analyzer Calibration," Proceedings of the 18th European
Microwave Conference, Stockholm, Sweden, 1988.
G. Gonzales, Microwave Transistor Amplifiers, 2nd ed. Prentice Hall, Upper Saddle
River, NJ, 1997.
S-Parameter Design, Hewlett-Packard Application Note 154, 1972.
D. V. Morgan and M. J. Howes, eds., Microwave Solid State Devices and Applications,
P. Peregrinus Ltd., New York, 1980.
P. A. Rizzi, Microwave Engineering-Passive Circuits, Prentice Hall, Upper Saddle

River, NJ, 1988.
D. Roddy, Microwave Technology, Prentice Hall, Upper Saddle River, NJ.
196
Chapter 4 Single- and Multtport Networb
Problems
4.1
From the defining equations (4.3) and (4.6) for the impedance and admit1
tance matrices, show that [ z] = [ y
.
4.2
For the following generic T-network, find the impedance and admittance
matrices.
4.3
Show that for a bipolar-junction transistor in a common-base configuration

under small-signal low-frequency conditions (whose equivalent circuit is
shown below) a hybrid parameter matrix can be established as follows:
rbe+(l+fl)rce
[h] -
rbe+(l+fl)rce
rbe + flrce
7bc
1
+ -----=--rbe + (1
+ ~)rct
where the individual transistor parameters are denoted in the figure.

r'--- -- - - -,- - - - - - - J
~] 8
,
h:
LX j :B
....._. ,- - - - -
B:
E~~~~
:,
=:>
:I.t
-""'"-
'B rc,
B.---~-----._~-B
4.4
Using the results from Problem 4.3, compute the equivalent circuit parameters for a BJT in common-base configuration if the h-matrix is given as
[h] = [
16.6
- 0.99668
4.5
0.262xl~-3l
9
66.5 x l0 J
Employ the conversion table for the different parameter representations of

the two-port network and find the h-matrix representation for a Darlington
pair shown in Figure 4-7 under the assumption that the transistors are specified by the same h-matrices derived in Example 4-2.
Problems
197
4.6
Using the definition of the ABCD network representation, find the -parameter description.
4.7
From the results of Problem 4.3 and Example 4.2, establish the conversion
equations between the h-matrix parameters for the common-base and common-emitter transistor configurations.
4.8
Unlike the series connection discussed in Example 4-4, derive the ABCDparameters for a two-port network where the impedance Z is connected in
parallel.
4.9
Find the ABCD-parameters for a genenc three-element pi-network, as

depicted in Figure 4-2.
4.10 Compute the ABCD-parameters for an RF transformer with tum ratio

N = N 1IN 2 , where N 1 is the number of turns a the primary winding and
N 2 is the number of turns of the secondary winding.
4.11
Prove that the h-matrix parameters for a high-frequency hybrid transistor

model shown in Figure 4-12 are given by (4.31).
4.12 In this chapter we have mentioned several h-matrix representations of the

bipolar-junction transistor for different frequency conditions. In all cases we
have neglected the influence of the parasitic components associated with the
casing of the transistor. The modification to the equivalent circuit of the transistor that takes into account these parasitics is shown below:
Cec
B'
Intrinsic
Transistor
Model
E'
198
Chapter 4 Single- and Multlport Networkl
Assuming that the intrinsic transistor model is given by a generic h-matrix,

derive the modified model that accounts for the casing.
4.13
Compute the return loss for a 25

transmission line.
resistor connected to a 75 .Q lossless
4.14 Find the forward gain of the circuit discussed in Example 4-8.
4.15
Given that the input of an amplifier has a VSWR of 2 and the output is given
by VSWR = 3, find the magnitudes of the input and output reflection coefficients. What does your result mean in terms of S 11 and S22 ?
4.16
Using the same approach as described in Section 4.4.4, show that the Sparameters of the network can be computed from the known f-parameters
USing
= ([Y] + Y0 [E])- 1(Y 0 [E]- [Y])
[SJ
and the corresponding inverse relation
= Y0 ([E]- [S])([S ] + [E])-1
[Y]
where Y0
4.17
= 1I Z0
is the characteristic line admittance.
The ideal transformer of Problem 4.10 can also be represented in S-parameter form. Show that the S-matrix is given by
[S]
4.18
=(
1 )[(N -1) (2N)

2
1+N
(2N) ( 1 - N 2 )
For the following two circuits, prove that the S-parameters are given as
[S]
= [ rt
1-r1
respectively, where f
1-
r ll
1
and [S]
f 2 I + r 2l
1+f2 r 2
=[
= (1 + 2Z0 /Z 1)- 1 and f 2

and
o------oo
= -( 1 + 2Y0 /Y 1)- 1
Problema
199
4.19 For the following T-network inserted into a transmission line with characteristic impedance of Z 0 = 50 Q, the three resistances are
R 1 = R 2 = 8.56 Q , and R3 = 141.8 Q. Find the S-parameters of this
configuration and plot the insertion loss as a function of inductance L for
the frequency of f = 2 GHz and L changing from 0 to 100 nH.
4.20 In practice, the resistors in the T-network of the previous problem are not
frequency independent. At RF frequencies parasitic effects have to be taken
into account. Compute the S-parameters at 2 GHz when all resistors have a
0.5 nH parasitic series inductance. Assume L is fixed at 10 nH.
4.21
A BJT is operated in a 50 Q circuit at 1.5 GHz. For the bias conditions of 4

rnA collector current and collector-emitter voltage of 10 V, the manufacturer
provides the S-parameters in magnitude and angle as follows:
S 11
=0.6 L -127; S 21 =3.88 L 87; S 12 = 0.039 L 28; S22 =0.76 L -35.
Find (a) the Z-parameter and (b) the h-parameter representation.
C\-\"P1'ER
An Overview of RF Filter
Design
Arter the discussion in Chapter 4, we are ready to

extend and apply our knowledge of one- and two-port networks to develop RF filters. It
is of particular interest in any analog circuit design to manipulate high-frequency signals in such a way as to enhance or attenuate certain frequency ranges or bands. This
chapter examines the filtering of analog signals. As we know from elementary circuit
courses, there are generally four types of filters: low-pass, high-pass, bandpass, and
bandstop. The low-pass filter allows low-frequency signals to be transmitted from the
input to the output port with little attenuation. However, as the frequency exceeds a certain cut-off point, the attenuation increases significantly with the result of delivering an
amplitude-reduced signal to the output port. The opposite behavior is true for a highpass filter, where the low-frequency signal components are highly attenuated or reduced
in amplitude, while beyond a cut-off frequency point the signal passes the filter with little attenuation. Bandpass and bandstop filters restrict the passband between specific
lower and upper frequency points where the attenuation is either low (bandpass) or high
(bandstop) compared to the remaining frequency band.
In this chapter we first review several fundamental concepts and definitions pertaining to filters and resonators. Specifically, the key concept of loaded and unloaded
quality factors will be examined in some detail. Then, we introduce the basic, multisection low-pass filter configuration for which tabulated coefficients have been developed
both for the so-called maximally fiat binomial, or Butterworth filter, and the equi-ripple
or Chebyshev filter. The intent of Chapter 5 is not to introduce the reader to the entire
filter theory, particularly how to derive these coefficients, but rather how to utilize the
information to design specific filter types. We will see that the normalized low-pass filter serves as the basic building block from which all four filter types can be derived.
201
202
Chapter 5 An Overview of RF Filter Dealgn
Once we know the procedures of converting a standard low-pass filter design in

Butterworth or Chebyshev configuration into a particular filter type that meets our
requirements, we then need to investigate ways of implementing the filter through
distributed elements. This step is critical, since at frequencies above 500 MHz lumped
elements such as inductors and capacitors are unsuitable. Relying on Richards transformation, which converts lumped into distributed elements, and Kuroda's identities, we
are given powerful tools to develop a wide range of practically realizable filter
configurations.
5.1 Basic Resonator and Filter Configurations

5.1.1
Filter Types and Parameters
It is convenient to begin our discussion by introducing the ideal behavior of the

four basic filter types: low-pass, high-pass, bandpass, and bandstop. Figure 5-l summarizes their attenuation a versus normalized angular frequency behavior.
a, dB
a., dB
CX-+oo
(l~oo
I
I
I
0
0
Low-pass filter
I
High-pass filter
a., dB
a, dB
(l~oo
a~oo
Figure 5-1
Four basic filter types.
We have chosen the parameter Q = ro/ roc as a normalized frequency with

respect to the angular frequency roc, which denotes cut-off frequency for low-pass and
high-pass filters and center frequency for bandpass and bandstop filters. As we will
203
Bulc Resonator and Filter Conflg ... ratlons
see, this normalization will greatly simplify our task of developing standard filter
approaches. Actual attenuation profiles are shown in Figure 5-2 for the so-called binomial (Butterworth), Chebyshev, and eUiptic (Cauer) low-pass filters.
a, dB
a, dB
1
Chebyshev filter
Binominal filter
a, dB
1
Elliptic filter
Figure 52
Actual attenuation profile for three types of low-pass filters.
The binomial filter exhibits a monotonic attenuation profile that is generally easy
to implement. Unfortunately, to achieve a steep attenuation transition from pass- to stop
band, a large number of components is needed. A better, steeper slope can be implemented if one permits a certain degree of variations, or ripples, in the passband attenuation profile. If these ripples maintain equal amplitude, either in the stopband or
passband, we speak of a Chebyshev filter since the design relies on the so-called Chebyshev polynomials. For both the binomial and the Chebyshev filter we observe that the
attenuation approaches infinity as Q ~ oo This is in contrast to the elliptic filters,
which allow the steepest transitions from passband to stopband at the expense of ripples
in both bands. Because of their mathematical complexity in designing elliptic filters, we
will not investigate them any further (for more information see Rizzi, listed in Further
Reading at the end of this chapter).
In analyzing the various trade-offs when dealing with filters, the following parameters play key roles:
Chapter 5 An Overview of RF Filter Design
204
Insertion loss. Ideally, a perfect filter inserted into the RF circuit path would
introduce no power loss in the passband. In other words, it would have zero insertion loss. In reality, however, we have to expect a certain amount of power loss
associated with the filter. The insertion loss quantifies how much below the 0 dB
line the power amplitude response drops. In mathematical terms it states
pin
I 2
IL = lOlogp
= -IOlog(l- rinl )
(5.1)
where PL is the power delivered to the load, Pin is the input power from the
source, and jrinl is the reflection coefficient looking into the filter.
Ripple. The flatness of the signal in the passband can be quantified by specifying
the ripple or difference between maximum and minimum amplitude response in
either dB or Nepers. As already mentioned, and as will be discussed further, the
Chebyshev filter design allows us to precisely control the magnitude of the ripple.
Bandwidth. For a bandpass filter, bandwidth defines the difference between upper .
and lower frequencies typically recorded at the 3 dB attenuation points above the :
passband:
(5.2)
'
Shape factor. This factor describes the sharpness of the filter response by taking ;
the ratio between the 60 dB and the 3 dB bandwidths:
:
f~odB - tf!dB
B w60dB
SF=
BW3dB
f!dB _ fldB
(5.3)
}
!
Rejection. For an ideal filter we would obtain infinite attenuation level for the t
undesirable signal freque~cies. However, in reality we expect an. upper ~ound due f
to the deployment of a fimte number of filter components. Practical designs often i
specify 60 dB as the rejection rate since it can readily be combined with the shape l
factor (5.3).
The preceding filter parameters are best illustrated by way of a generic bandpass :
attenuation profile, as summarized in Figure 5-3. The magnitude of the filter's attenua~
tion behavior is plotted with respect to the normalized frequency Q. As a result, the:
center frequency f c is normalized to Q = 1 . The 3 dB lower and upper cut-off fr~:
quencies are symmetric with respect to this center frequency. Beyond these 3 dB points,
we observe the attenuation response rapidly increasing and reaching the 60 dB rejection.
points at which the stopband begins.
Basic Resonator and FIHar Configurations
205
ex., dB
BW60dB
Rejection
60dB
Figure 5-3
Generic attenuation profile for a bandpass filter.
There is one additional parameter describing the selectivity of the filter. This
parameter is known as the quality factor Q, which generally defines the ratio of the
average stored energy to the energy loss per cycle at the resonant frequency:
Q = ro average stored energy

energy loss per cycle
average stored energy

power loss
00----:---------~
- ro wstored
p
loss
ro =
00c
(5.4)
where the power loss P 1085 is equal to the energy loss per unit time. In applying this definition, care must be taken to distinguish between an unloaded and loaded filter. What is
meant here is best seen by viewing the filter as a two-port network connected to a
source at the input side and a load at the output, as shown in Figure 5-4.
VoO
Figure 5-4
Filter
[]z,
Filter as a two-port network connected to an RF source and load.
It is customary to consider the power loss as consisting of the power loss associated with the external load and the filter itself. The resulting quality factor is named
loaded Q, or QLD . Interestingly, if we take the inverse of the loaded Q, we see that
Chapter S An Overview of RF Fitter O..lgn
208
_ 1 ( power loss in filter )

QLD - co average stored energy
1
ro = co,
1 ( power loss in load )

+ ro average stored energy
(5.5)
ro = ro,
since the total power loss is comprised of the power losses due to the presence of the filter and the load. This can be written in the concise form
1
QLD
QF
QE
-=-+-
(5.6)
where QF and QE are the filter Q and the external Q. The precise meaning of (5.6)
will be analyzed in Section 5.1.4. As we will also see in this section, (5.6) can be cast in
the form
fc
fc
(5.7)
JtdB _/~dB = BW3dB
Qt.D =
where f c is the center or resonance frequency of the filter. In the following sections a
summary is given of the salient features of the three most common filters. Emphasis is
placed on the network description as previously developed in Chapter 4.
5.1.2
Low-Pass Filter
As one of the simplest examples we start our investigation by analyzing a firstorder low-pass filter connected to a load resistance, as depicted in Figure 5-5.
Zo
r- -- --- ---- --- -.

: Jl
:
--
~- --- --
c!lv,
ZL
... -~-- -
(a) low-pass filter with load resistance.
Va~
Val
Filter
Dv,
(b) Network with Input/output voltages
Figure 5-5 Low-pass fitter connected between source and load resistance.
The focal point in any filter design is to find the output V 2 due to the input voltage V 1 , or even better, the generator voltage V G . For our simple circuit this can best be
accomplished by cascading four ABCD-networks (labeled 1 through 4) as suggested in
Figure 5-6.
tulc Resonator and Filter Configurations
------
l
Figure 5-6
207
- --- .... ...... .. ___ ..... _,

2
3
._____4 _...
Cascading four ABCD-networks.
The overall ABCD-network is therefore
Bl _ [1 Ral ll Rll 1 ol [ 1 ol
0 1Jlo 1j l;roc1j 1/ RLIJ
A
[CDj
1 + (R + Ra)(jroC +
=
. c + R1
Jro
;J
Ra + RL
(5.8)
where we use the fact that both source and load impedances are resistive, i.e., Z 0 = RG
md ZL = Rv Since A is already the ratio V 0 /V 2 , we only have to invert this single
~fficient:
=A
1
------------------1 + (R + Ra)(jroC +
;J
(5.9)
Equation (5.9) can be examined for the limiting cases where the frequency is either zero
()1' approaches infinity. For ro ~ 0 we obtain
(5.10a)
and for ro ~ oo .
(5.10b)
In the first case we notice that the voltage divider rule applies for the DC condition,
while for the second case the filter exhibits the expected low-pass behavior of zero output voltage at high frequencies. Further, if the load resistance goes to infinity
(RL ~ oo ), the filter becomes unloaded and in the limit a pure first-order system
results:
v2 =
V0
H(ro) - - - - - -1 + jro(Rc + R)C
(5.11)
Chapter 5 An Overview of RF Filter Dellgn
208
where H ( ro) is known from system theory as transfer function. Besides specifying
the transfer function, it is more common to compute the attenuation factor in Neper
(Np) such that
1
2
a(ro) = -lniH(ro)l = - IniH(ro)l
(5.12a)
or in dB as
a(ro) = -20logiH(ro)l = -10log iH(ro)l
The corresponding phase is
(5.12b)
q>( ro) = tan- I (lm{H( ro) })
(5.12c)
Re{H(ro)}
Directly related to phase is the so-called group delay t g , which is defined as the frequency derivative of the phase .
= dq>( (J))
t
g
(5.12d)
dO>
It is often desirable to design a filter with nearly linear phase (i.e., 4> oc roA , with A
being an arbitrary constant factor). The group delay is then simply a constant t g oc A .
A typical filter response for C = 10 pF, R = 10 Q, R0 = 50 n and various load
resistances is shown in Figure 5-7.
30
25
R =SQ
~
20
'j
~
15
~--~R~
=~2o~n=------------
lO
R
5
=500
RL=lOOQ
RL =2500
0~======~======~--~~~~=-~~
9
to6
to'
to8
Frequency, Hz
to
t010
(a) Attenuat;on profile of the low-pass filter for various load resistances
Figure 57 First-order low-pass filter response as a function of various load
resistances.
209
Bulc Resonator and Filter Configurations
-10
-20
-30
if
'"0
-40
(IJ~
"'
~
a:;
-50
-60
-70
-80
-90~--~~~--~~~~~~~~--~--~
106
10
10 9
10 8
10
10
Frequency, Hz
(b) Phase response of the low-pass filter for various load resistances
Figure 5-7 First-order low-pass filter response as a function of various load
resistances. (Continued)
5.1.3
High-Pass Filter
Replacing the capacitor with an inductor in Figure 5-5 pennits the construction of
a first-order high-pass filter, as depicted in Figure 5-8. The analysis follows the same
steps as outlined in (5.9), except that the capacitive reactance is replaced by an inductive reactance. The result is
+(R+Rc)c~L +
. c +If1
Jro
;J
Rc+R
(5.13)
I
Zo
-------- -- ... -- ---,

; R
:
t--cl"""i'-'
-- -- ---
'
'I
'
(a) High-pass filter with load resistance

(b) Network and input/ou put voltages
Figure 5-8 First-order high-pass filter.
Chapter 5 An Overview of RF Finer Design
210
This gives us directly the result

1
-- =
A
I
+(R+Ra>C~L +
(5.14)
;J
As ro ---7 0 , it is seen that
(5.15a)
and for ro ---7 oo we conclude
-VG
1 + (R + Rc)l RL
(5.15b)
Rc + R + RL
which reveals that the inductive influence can be neglected. The filter response for
L = 100 nH, R = 10 n, RG =50 n, and various load resistances is shown in Figure 5-9.
5.1.4
Bandpass and Bandstop Filters
A bandpass filter can be constructed through an RLC series circuit or through a

parallel connection of an RLC shunt circuit. The generic series circuit diagram, including generator and load impedances, is displayed in Figure 5-10.
The network representation in ABCD notation takes on the form
(5.16)
1
where impedence Z is specified from conventional circuit analysis as
z = R + i(roL -
c)
1
00
(5.17)
The transfer function H ( ro) = 1I A is found to be
~ = H(ro)
Vc
(ZL + Zc)
zL
+ R + j[roL- 1/ (roC)]
(5.18)
Explicit plots of the transfer function and the attenuation profile are discussed in the
following example.
211
40
35
30
~
"0
;::
25
R =5 Q
.sa
~ 20
c4)
t::
<
15
10
5
0
10
107
10
Frequency, Hz
(a) Attenuation profile of the filter for various load resistances

90~~;:~~~~~~~~~
80
70
60
~
Cl.l
'"0
50
"'ce
40
ll-.
30
20
10
0 6
10
10
10 8
Frequency, Hz
10
1010
(b) Phase response of the filter for various load resistances

Figure 5-9 Low-pass filter response as a function of various load resistances.
Figure 5-10
Bandpass filter implemented in series configuration.
212
----------------------------~~~
Example 5-1: Bandpass filter response
For a bandpass filter with zL = ZG = 50 n the following components are selected: R = 20 n , L = 5 nH , and C = 2 pF . Find
the resonance frequency, and plot the frequency response of the
phase of the transfer function and the associated attenuation profile
in dB.
Solution:
To solve this problem we use the definition of the transfer function for the bandpass filter presented in (5.18). The attenuation profile of the filter expressed in dB is computed as
a = 20log[H(rof 1] = -20log[H(ro)]. Both the attenuation and
phase profiles of the filter are shown in Figure 5-11. From the graph
we can estimate the resonance frequency f 0 of the filter to be approxIS
imately
1.5 GHz.
The
exact
numerical
value
/ 0 = l / (21tJi) = 1.59 GHz .
100
45
... ...
40
80
' ....
''
35
~ 30
t:
<
20 ~
'
:::s
s::
~
40
( 'a
'
s:::f
..... 25
60
20
0 ..01
Q.
01
-20 <!CI
\
\
-40
15
....
10
5
10 7
10 8
10 9
Frequency, Hz
Figure 5-11
10
'
... ... __
10
Bandpass filter response.
--60
-80
-100
11
10
213
Bulc Resonator and Filter Configurations
As expected, our bandpass filter assumes a minimum attenuation at the resonance point, but the transitions from stopband to
passband are very gradual.
If the series circuit is replaced by a shunt circuit, as shown in Figure 5-12, we only
have to replace Z by 1/Y in (5.17), which leads to
V2
V
0
ZL
ZL+Z 0 +I I Y
(5.19)
j(roc - ~L)
(5.20)
where the admittance is
Y = G+
and upon insertion into (5.19) yields
(5.21)
A typical transfer function response of magnitude and phase for the values listed in
Example 5-1 is seen in Figure 5-12.
Working with energy storage systems or LC-based networks, we can use the quality factor as introduced in Section 5 .1. 1 to specify the bandwidth of the 3 dB passband
or stopband of a filter:
BW = fo
Q
(5.22)
where f 0 is the resonance frequency. This quality factor is the inverse of the dissipation factor d, which depends on whether we deal with a series (RLC) or a parallel connected (GLC) circuit. Table 5-l summarizes all relevant definitions for the series and
parallel resonance circuits.
The quality factor provides important insight into the losses generated by a particular resonator circuit configuration. The circuits shown in Table 5-1 depict unloaded filters (i.e., filters in the absence of any external load connections).
214
Chapter 5 An Overview of RF Filter Detlgn
l6r-~~~~~~~~~~~~~--~~
15
14
~13
g 12
.::::
!
l(;=ZL =50 0
R=2000
C=2pF
11
!10
L=SnH
9
8
7
6~------~--~~~~~~~------~
107
10 8
10q
1010
10 11
Frequency, Hz
(a) Magnitude of transfer function
20
oil
~
~
10
Or-----
-10
-20
-30~--~--~~~~~--~~~----~~
10
lOs
10
10
10
1011
Frequency, Hz
(b) Phase of transfer function
Figure 5-12 Bandstop filter response.
When dealing with the loaded situation we are confronted with the additional
complication of generator and load impedances attached to the resonator. With reference to Figure 5-10, let us take a more detailed look at how the three different quality
factors arise. To this end, our aim is to analyze the series resonance, or bandpass filter,
connected to the generator resistance R 0 and load RL. Without loss of generality, we
can combine both resistances into the configuration shown in Figure 5-13.
215
B11lc Resonator and Filter Configurations
Table 5-1
Series and parallel resonators
c
~rR
Parameter
~
L
Impedance or Admittance
1
z = R + j roL + ]00
. C
roo =
Reaonance Frequency
Dissipation Factor
-1
JLC
= -roRL = Rro0 C
ro 0 L
BW
Bandwidth
G
d= - - - Gro0 L
ro C
0
= Rro0 C
Q=
1R
= fo = -Q
21tL
BW
Q= -
= -JLC
roo
Quality Factor
Y = G + jooC + _l_
jroL
ro C
1
Goo0L
= fo
= _1 G
21tC
0
--G
'--- -- -- -- ------ --- --
Figure 5-13
where RE
Circuit used for the definitions of loaded and unloaded quality factors.
= Rc + RL
and V G is understood as a Thevenin-equivalent source. The

losses can now be partitioned as originating from an external resistance RE, an internal
resistance R , or both. Therefore, we need to differentiate three cases:
External quality factor (RE :F. 0, R = 0)
ro0 L
RE
REro0 C
QE=- =
Internal or filter quality factor (RE = 0, R "#. 0)
_ ro0 L _
QF - -R -
1
Rro0 C
218
Loaded quality factor (RE ' 0, R ' 0)
ro0L
= R +RE =
QLD
Identical expressions are derived for a shunt resonator circuit if we replace R and Re
by G and GE. It is customary to introduce the normalized frequency deviation from
the resonance point
ro roo
= --roo ro
(5.23)
and expand it as follows:
f 0 + f- I 0fo
I0
fo+f-fo
(t + ~!)(t +~~)-I
~ 2~1
fo
lo
lo
(5.24)
with tt.j = f 0 - f. Equation (5.24) leads to the expression of the differential change in
quality factor
= QLD ~ 2~~ QLD
f!QLD
If (5.25) is solved for
QLD,
and using X
QLD
= roL, we obtain
fo
tt.QLD
dX
= 2(RE + R) df
(5.25)
(5.26a)
I -=lo
for the series circuit configuration.

Alternatively, for a parallel circuit with B = 1I ( roL) we have
ilQLD
QLD =
dB
2( G + G) df
fo
(5.26b)
I
=fo
The equations (5.26a) and (5.26b) show that generically the loaded quality factor for
complex impedances (or admittances) can be computed as
Q
_ ft.QLD _
LD -
f o dlm{Z}
2Re{Z}
df
fo
dim{ Y}
df
I=
fo
or
QLD
ilQLD
= 2Re{Y}
I= lo
217
where Re{Z}, lm{Z}, Re{Y}, and lm{Y} are real and imaginary parts of the total
impedance or admittance of the resonance circuit.
5.1.5
Insertion Loss
The previously developed quality factor expressions are very useful in RF circuit
design, since the Q of a filter can more easily be measured (for instance, with a network
analyzer) than the actual impedance or admittance. It is therefore helpful tore-express
the impedance or admittance values of bandpass or bandstop filters in terms of the various Q-factors. For example, the impedance of the series resonance circuit can be rewritten as:
(5.27)
which leads to
(5.28)
Following the same steps as described for a series resonator, a very similar expression
can be derived for the admittance Y of a parallel resonator:
Y =
(GE+Gl[~:+jQwE]
(5.29)
We now tum our attention to the following situation: a transmission line system
with characteristic line impedance Z 0 is matched at the load and generator sides
(ZL = Zc = Z 0 ) as seen in Figure 5-14(a).
(a) Matched transmission line system
Zo
f__i ___ ___L__ .. _c1
~'"'V"\..--fl-f-o-c:===t-O---,
-- ---- --------------'
(b) Inserted bandpass filter
Figure 5-14 Insertion loss considerations.
Chapter 5 An Overview of RF Filter O..lgn
218
In Figure 5-14(a) the power delivered to the load PL is the total available power from
the source Pin :
(5.30)
If the filter is inserted as shown in Figure 5-14(b), the power delivered to the load
becomes
PL
=!
V G 2Zo
2 2Zo + Z
IVGl
/(8Zo)
LD + jQLD
.
- 1 2 2Z0 + (220 + R) -Q
4Z0
F
[Q
(5.31)
which, after some algebra and the use of (5.6), yields

PL
= pin
1
2
(5.32)
(1 + QLD)QE/ QLD
The insertion loss in dB due to the presence of the filter is then computed as
IL
2 2
2
= lOlog ( 1 +2 f?QZDJ
=
lOlog(l
+
QLv)-lOlog(l- QLD/QF)
2
(5.33)
QLD/QE
At resonance, = 0 , the first term drops out and the second term quantifies the associated resonator losses. However, if the filter is off resonance, then the first term quantifies the sensitivity. If we consider the frequency at which the power delivered to the
load is half, or 3 dB, of the power at resonance frequency, we can immediately write
2
that 1 + QZv = 2, or, taking into account relation (5.24), we obtain
BW3dB
= 2~f = /o = fo i QLD
Recalling Section 2.11, we notice that (5.33) can be related to the input reflection coefficient:
2
1-
zin
-ZG 2 = QLD / QE _
zin
+ ZG
1 + 2QZD
(5.34)
LF
where LF is known as the loss factor. This loss factor plays a central role when devel~
oping the desired filter attenuation profiles.
219
Bulc Resonator and Filter Conflguratlona
---------------------------~~~
Example 5..2: Calculation of various quality factors for a filter
For the filter configuration shown in Figure 5-14(b), the following
Zo, R
10 n,
parameters are given: Zo = 50 n, ZG = ZL
L = 50 nH, C = 0.47 pF, and the generator voltage is
V 0 = 5 V. Find the loaded, unloaded (filter), and external quality
factors; power generated by the source; power absorbed by the load
at resonance; and plot the insertion loss in the range of 20% of the
resonance frequency.
Solution:
The first step in the solution of this problem is to find

the resonance frequency of the filter:
f0 =
rr;:.
= 1.038 GHz
2rt"-/LJC
Knowing this value we are now capable of computing the various

quality factors of the filter:
External quality factor:
Internal or filter quality factor:
ro0L
Loaded quality factor:
QLD = R + 2Z = 2.97
0
To determine the input power, or maximum available power from

the source, we use (5.30):
Pin
IV ol 2 /(820 )
= 62.5 mW
Due to nonzero internal resistance of the filter (R = 10 .Q), the signal will suffer some attenuation even at the resonance frequency and
the power delivered to the load will be less than the available power:
pL =
1
pin
(1 + QLD)QE/QLD
f=fo
1
pin
= 5 1.7 mW
QE/QLD
Finally, substituting the loaded and external quality factors into

(5.33), we proceed to find the insertion loss of the filter in the range
220
of +20% of the resonance frequency by plotting the frequency

response of IL, as shown in Figure 5-15. As we see from the graph,
the 3 dB bandwidth of this filter is approximately equal to 350 MHz,
which agrees with the result obtained using our formula derived earlier in this section (i.e., BW 3dB = f 0 /QLD = 350.07MHz).
5.5
5
4.5
~
fJ)
--=
BW=350MHz
4
3.5
(II
t::
2.5
Q)
2
1.5
0.5
0.8
0.85
0.9
0.95
1.05
l.l
1.15
1.2
1.25
Frequency, GHz
Figure 515 Insertion loss versus frequency.

Although not as distinctively observed in practice, the example
shows that the loaded quality is lower than both the external and
internal filter quality factors.
5.2 Special Filter Realizations

The analytical synthesis of special filter characteristics such as low-pass, high-pass,
and bandpass/bandstop filters is generally very complicated. In our brief introductory
treatment we are going to concentrate on two filter types: the maximally flat Butterworth
and the equi-ripple Chebyshev filter realizations. Both filter types are analyzed first in a
normalized low-pass configuration, before the low-pass behavior is frequency scaled to
implement the remaining filter types through a frequency transformation.
221
Special Filter Realizations
5.2.1
Butterworth-Type Filters
This filter type is also known as maximally flat filter since no ripple is permitted in
its attenuation profile. For the low-pass filter, the insertion loss is determined through
the loss factor,
IL = -10log(l-jfinj 2 )
= lOlog{LF}
= lOlog{ 1 +a
n 2N }
(5.35)
where Q is again the normalized frequency as introduced in Section 5 .1.1 and where N
denotes the order of the filter. It is customary to select the constant a = 1 so that at
0 = ro/ roc = 1 the insertion loss becomes IL = 10log{2}, which is the 3 dB point
at the cut-off frequency. In Figure 5-16 the insertion loss for several values N is plotted.
30
25
!gv) 20
en
0
15
'0
..s
10
5
oL-~~~~~~L-~~--~~~
0.2 0.4 0.6 0.8
1.2
1.4
1.6
1.8
Normalized frequency, 11
Figure 5-16 Butterworth low-pass filter design.
Two possible realizations of the generic normalized low-pass filter are shown in
Figure 5-17, where we set R 0 = 1 .
The element values in the circuits in Figure 5-17 are numbered from g 0 at the
generator side to g N + 1 at the load location. The elements in the circuit alternate
between series inductance and shunt capacitance. The corresponding elements g are
defined as follows:
g = {
0
internal generator resistance for circuit in Figure 5-17 (a)

internal generator conductance for circuit in Figure 5-17(b)
222
(a)
G =g
g1
ct=E~~~gN'I
(b)
Figure 517 Two equivalent realizations of the generic multisection low-pass

filter with normalized elements.
gm
gN+ I
inducance for series inductor

capacitance for shunt capacitor
(m= 1, ... ,N)
load resistance if the last element is a shunt capacitor

= {
load conductance if the last element is a series inductor
The values for the g's are tabulated and can be found in the literature (see Pozar and Rizzi,
listed in Further Reading). For N up to 10, Table 5-2 summarizes the respective g-values
for the maximally flat low-pass filter based on g 0 = 1 and cut-off frequency roc = 1.
The corresponding attenuation versus frequency behavior for various filter orders
N is seen in Figure 5-18. We note that Q = 1 is the 3 dB cut-off frequency point. The
attenuation curves in Figure 5-18 are very useful in determining the required order of
the filter. For instance, if a maximally flat low-pass filter is to be designed with attenuation of at least 60 dB at Q
2, we see that an order of N
10 is required.
Figure 5-18 exhibits a steep increase in attenuation after cut-off. We notice that
2
for Q 1 or ro roc the loss factor increases as Q N, which is a rate of 20N dB per
decade. However, nothing is said about the phase response of such a filter. In many
wireless communication applications, a linear phase behavior may be a more critical
issue than a rapid attenuation or amplitude transition. Unfortunately, linear phase and
rapid amplitude change are opposing requirements. If linear phase is desired, we
demand a functional behavior similar to (5.35)
(5.36)
223
Coefficients for maximally flat low-pass filter (N =1 to 1O)
Table 5-2
g,
g2
g4
83
8s
g6
87
g9
88
8to
g II
1 2.0000 1.0000
2 1.4142 1.4142 1.0000
3 1.0000 2.0000 1.0000 1.0000

4 0.7654
1.8478
1.8478
0.7654
1.0000
5 0.6180 1.6180 2.0000 1.6180 0.6180
1.0000
6 0.5176 1.4142 1.93 18 1.9318 1.4142 0.5176 1.0000

7 0.4450 1.2470
1.8019 2.0000
I .8019
1.2470
0.4450
1.0000
8 0.3902
J.llll
1.6629
1.9615
1.9615
1.6629
1.111 J
0.3902
9 0.3473
1.0000
1.5321
1.8794 2.0000
1.8794
1.5321
1.0000 0.3473
1.0000
1.0000
10 0.3129 0.9080 1.4142 1.7820 1.9754 1.9754 1.7820 1.4142 0.9080 0.3129 1.0000
50
li
0
.....
40
---- ......................... -. ......_..
------
...
.''
..'.__ _________.'...
____ ____ ,..
s 30
Q)
20
10
1.2
Figure 5-18
1.3
3
4
Normalized frequency, Q
1.5 1.7
11
Attenuation behavior of maximally flat low-pass filter versus

normalized frequency.
224
with A 1 and A 2 being arbitrary constants. The associated group delay t g is

t
= dcj)(Q) = A [ 1 +A (2N + 1 )Q 2 N]
dO.
(5.37)
In Table 5-3, the first 10 coefficients for a linear phase response with group delay
t g = 1 are listed.
Since steep filter transition and linear phase are generally competing requirements, it has to be expected that the shape factor is reduced. The question of how a linear phase design based on Table 5-3 compares with a standard design of Table 5-2 is
discussed in Example 5-3 for the case N = 3.
Table 5-3
Coefficients for linear phase lowpass filter (N = 1 to 10).
g3
g4
g5
g6
gg
gl
g2
g7
2.0000
1.0000
1.5774 0.4226 1.0000
1.2550 0.5528 0.1922
1.0000
1.0598 0.5116 0.3181
0.1104
0.9303 0.4577 0.3312 0.2090 0.0718
0.8377 0.4116 0.3158 0.2364 0.1480 0.0505
0.7677 0.3744 0.2944 0.2378 0.1778 0.1104 0.0375
0.7125 0.3446 0.2735 0.2297 0.1867 0.1387 0.0855 0.0289
0.6678 0.3203 0.2547 0.2184 0.1859 0.1506 0.1111
89
8to
811
1.0000
1.0000
1.0000
1.0000
1.0000
0.0682 0.0230
1.0000
10 0.6305 0.3002 0.2384 0.2066 0.1808 0.1539 0.1240 0.0911 0.0551 0.0187 1.0000
5.2.2
Chebyshev-Type Filters
The design of an equi-ripple filter type is based on an insertion loss whose functional behavior is described by the Chebyshev polynomials TN ( n) in the following
form:
IL
= IOlog{LF} = lOlog{ I +a2T 2N(Q)}
where
1
T N(O.) = cos{N[cos- (0)]}, for
101 ~ 1
1
T N(Q) = cosh{N[cosh- (0)]}, for
101 >I
(5.38)
;peclel Filter Realizationa
225
To appreciate the behavior of the Chebyshev polynomials in the normalized frequency

range -1 < Q < 1 , we list the first five terms:
2
T0 = l,T 1 = Q,T2 = -1+20 ,T3
=-30+403,T4 = 1-802 +804
The functional behavior of the first two terms is a constant and a linear function, and
the subsequent three terms are quadratic, cubic, and fourth-order functions, as seen in
Figure 5-19.
'
o.8 ,
0.6
0.4
',
1I
,~
'
'I
7,
13
I,
o I '\
o.2
-0.2
\
\
-0.4
-0.6
\
-0.8
'
-lL-~~--~~'~~;~~-u~~~
-I -0.8 -0.6-0.4 -0.2 0
0.2 0.4 0.6 0.8
Normalized frequency, n
Figure 5-19
Chebyshev polynomials T 1( n) through T 4 ( Q) in the normalized

frequency range -I ~ 0 ~ I .
It can be observed that all polynomials oscillate within a 1 interval, a fact that is
exploited in the equi-ripple design. The magnitude of the transfer function IH(jil)l is
obtained from the Chebyshev polynomial as follows:
IH(il)l = JH(Q)H(Q) * =
1
J1 + a
T~(Q)
(5.39)
where TN(Q) is the Chebyshev polynomial of order Nand a is a constant factor that
allows us to control the height of the passband ripples. For instance, if we choose
a =1 ' then at n = 1 we have
IH(O)I
= j21 = 0.707
226
which is the 3 dB level that applies uniformly throughout the passband (equiripple). We do
not go any further into the general theory of Chebyshev filter design, but rather refer the
reader to a classical textbook that covers this topic comprehensively (see Matthaei et al.).
In Figure 5-20 the loss factor and insertion loss are plotted for a Chebyshev filter
with coefficient a = 1 , which again results in a 3 dB attenuation response at resonance
frequency ( n = 1 ) .
6~~--~--~--~~~~~~~--~--~
N=4
5.5
4.5
v
~
V)
V)
3.5
3
2.5
1.5
~~~~~~~~~~--~--~~~_.--~
0.2 0.4 0.6
0.8
1.2
1.4
1.6
1.8
1.6
1.8
Normalized frequency, n
25
~ 20
fll
VJ
....0c
rl'.l
.s
15
10
5
0.2 0.4 0.6 0.8
1.2
1.4
Figure 5-20
Frequency dependence of the loss factor and insertion loss of the

Chebyshev low-pass filter.
227
As mentioned, the magnitude of the ripple can be controlled by suitably choosing

factor a. Since Chebyshev polynomials oscillate in the range from -1 to +1 for
-1 ~ n s; 1 , the squared value of these polynomials will change from 0 to + 1 in the
same frequency range. Therefore, in the frequency range of - 1 ~ n s; 1 the minimum
attenuation that is introduced by the filter is 0 dB and the maximum attenuation, or
2
equivalently the magnitude of the ripples, is /L = 10log( 1 + a ). Thus, if the desired
magnitude of the ripples is denoted as RPLda , then a should be chosen as
a =
J10
RPL0 BI 10
- 1
For instance, to obtain a ripple level of 0.5 dB we have to select
51 10
112
a = (10
- 1)
= 0.3493. The associated attenuation profiles for the first 10
orders are shown in Figure 5-21 for a 3 dB ripple, and in Figure 5-22 for a 0.5 dB ripple.
o~~~~~~--~~~~~--~--~~~
1.01 1.02 1.04
1.1
1.2
1.4 1. 7 2
8 11
Figure 5-21
Attenuation response for 3 dB Chebyshev design.
Upon comparing Figure 5-21 with 5-22, it is apparent that the disadvantage of a
higher ripple in the passband has as an advantage a steeper transition to the stopband. For
instance, a fifth-order, 3 dB ripple Chebyshev filter design at Q = 1.2 has an attenuation of20 dB, whereas the same order 0.5 dB ripple filter reaches only 12 dB at the same
frequency point. The trend remains the same for higher frequencies and different orders.
As a case in point, at n = 5 the fourth-order, 0.5 dB filter has an attenuation of 65 dB
compared with the 3 dB design, which has an attenuation of approximately 73 dB.
With reference to the prototype filter circuit, Figure 5-17, the corresponding coefficients are listed in Table 5-4.
228
0~~~~~~~=~~~~~
1.01 1.02 1.04

Figure 522
Table 5-4 (a)
1.1 1.2 1.4 l. 7 2

Normalized frequency, 0
8 11
Attenuation response for 0.5 dB Chebyshev design.
Chebyshev filter coefficients; 3 dB filter design (N
=1 to 10)
8!
82
1.9953
1.0000
3.1013 0.5339 5.8095
3.3487 0.7117 3.3487 1.0000
3.4389 0.7483 4.3471
3.4817 0.7618 4.5381 0.7618 3.4817
1.0000
3.5045 0.7685 4.6061
0.6033 5.8095
3.5182 0.7723 4.6386 0.8039 4.6386 0.7723 3.5182 1.0000
3.5277 0.7745 4.6575 0.8089 4.6990 0.8018 4.4990 0.6073 5.8095
3.5340 0.7760 4.6692 0.8118 4.7272 0.8118 4.6692 0.7760 3.5340 1.0000
10 3.5384 0.7771
83
84
85
86
87
8s
89
810
8u
0.5920 5.8095
0.7929 4.4641
4.6768 0.8136 4.7425 0.8164 4.7260 0.8051
4.5142 0.6091 5.8095
Table 5-4 (b)

N
g,
gz
229
Chebyshev filter coefficients; 0.5 dB filter design (N = 1 to 10)
g3
g4
g5
g6
g7
0.6986 1.0000
1.4029 0.7071 1.9841
1.5963 1.0967 1.5963 1.0000
1.6703 1.1926 2.366 1 0.8419 1.9841
1.7058 1.2296 2.5408 1.2296 1.7058
1.7254 1.2479 2.6064 1.3137 2.4758 0.8696 1.9841
1.7372 1.2583 2.6381
1.7451
1.7504 1.2690 2.6678
1.3444 2.6381
g8
g9
giO
g 11
1.0000
1.2583 1.7372 1.0000
1.2647 2.6564 1.3590 2.6964 1.3389 2.5093 0.8796 1.9841

1.3673 2.7939
1.3673 2.6678
10 1.7543 1.2721 2.6754 1.3725 2.7392 1.3806 2.7231
1.2690 1.7504 1.0000

1.3485 2.5239 0.8842 1.9841
Unlike the previously discussed Butterworth filter, the Chebyshev filter approach
provides us with a steeper passband/stopband transition. For higher normalized frequencies Q 1, the Chebyshev polynomials Tty(Q) can be approximated as
(l/2)(2Q)N. This means that the filter has an improvement in attentuation of roughly
2
(2 N)/4 over the Butterworth design.
------------------------RF&MW4
Example 5-3: Comparison between Butterworth, linear phase
Butterworth, and Cbebyshev filters
Compare the attenuation versus frequency behavior of the thirdorder low-pass filter for (a) standard 3 dB Butterworth, (b) linear
phase Butterworth, and (c) 3 dB Chebyshev design.
Solution:
If we choose the first element of the filter to be an
inductor connected in series with the source, then the circuit topology of the third order filter is given by
230
where the inductances and the capacitor are obtained from Tables
5-2, 5-3, and 5-4. Specifically,
Standard Butterworth: L 1 = L 2 = 1 H, C 1 = 2 F
Linear phase Butterworth: L 1 = 1.255 H, C 1 = 0.5528 F,
L 2 = 0.1922 H
3 dB Chebyshev filter: L 1 = L 2 = 3.3487 H, C 1 = 0.7117 F
generator and load: Ra = RL = 1 n
As we can see from the preceding circuit diagram, under DC
condition the inductances become short circuits and the capacitor
acts like an open circuit. The voltage across the load is equal to onehalf of the voltage at the source due to the voltage divider formed by
the load and source impedances (i.e. V 2 = 0.5 V G ). When the frequency is not equal to zero the voltage across the load can be
obtained by applying the voltage divider rule twice; first, to obtain
the voltage at node A:
+ RL)
Zcll(ZL
VA=
Zcii(ZL2 + RL) + ZL 1 + R 0
Va
and, second, to obtain the voltage across the load with reference to
VA:
where Zc = R0 +jro1 If we find the ratio of the circuit gain at AC to

the gain under DC conditions, it is possible to compute the attenuation that is introduced by the filter:
a = 2
RL
Zcii(ZL2 + RL)
RL + ZL2 Zcli(ZL2 + RL)
+ ZL 1 + RG
The graph of the attenuation coefficient expressed in dB for the

three filter realizations is shown in Figure 5-23. As expected, the
Chebyshev filter has the steepest slope of the attenuation profile,
while the linear phase filter exhibits the lowest roll-off with fre-
Special Fitter Realizations
231
12~--------~~--------~----~----~
10
0.5
Normalized frequency, .a
Figure 5-23 Comparison of the frequency response of the Butterworth, linear

phase, and 3-dB Chebyshev therd-order filters.
quency. Therefore, if a sharp transition from passband to stopband is

required, and ripples can be tolerated, the most appropriate choice
would be a Chebyshev filter implementation. We also note that the
attenuation of the Chebyshev filter at cut-off frequency is equal to
the ripple size in the passband.
Even though the linear phase Butterworth filter suffers from a
shallow transition, it is the linear phase that makes it particularly
attractive for modulation and mixer circuits.
5.2.3
Denormallzatlon of Standard Low-Pass Design
To arrive at realizable filters, we have to denormalize the aforementioned coefficients to meet realistic frequency and impedance requirements. In addition, the standard low-pass filter prototype should be convertible into high-pass or
bandpass/bandstop filter types depending on the application. Those objectives can be
achieved by considering two distinct steps:
Chapter 5 An Overview of RF Filter De1lgn
232
Frequency transformation to convert from nomalized frequency
to actual
frequency ro . This step implies the scaling of the standard inductances and capacitances.
Impedance transformation to convert standard generator and load resistances
g 0 and g (N + 1) to actual resistances R L and R G
We begin by examining the frequency transformation and its implications in terms of
the various filter types. To eliminate confusing notation, we drop the index denoting
individual components (i.e., Ln(n = 1, ... , N) ~ L and Cn(n
1, ... , N) ~C).
This makes sense since the transformation rules to be developed will be applicable to
all components equally.
Frequency Transformation
A standard fourth-order low-pass Chebyshev filter with 3 dB ripples in the passband response is shown in Figure 5-24, where we have included negative frequencies to
display more clearly the symmetry of the attenuation profile in the frequency domain.
Furthermore, by appropriately scaling and shifting, we notice that all four filter types,
Figures 5-25, 5-26, 5-28, and 5-29, can be generated. This is now examined in detail.
30~--~--~---.---.----.---~---.---.
25
fg
20
15
~
<
10
Normalized frequency, .Q
Figure 5-24 Fourth-order low-pass Chebyshev filter with 3 dB ripples in the

passband.
233
For the low-pass filter we see that a simple multiplication by the angular cut-off
frequency roc accomplishes the desired scaling (see Figure 5-25):
ro
= .Qroc
(5.40)
30
25
20
t:f
.....
....0
~
15
10
0.2
Figure 5-25
0.4
0.6
0.8
1.2 1.4
Frequency, GHz
1.6
1.8
Conversion of standard low-pass filter prototype into low-pass

realization. Cut-off frequency is fc 1 GHz.
For the scaling we picked an arbitrary cut-off frequency of 1 GHz. In the corresponding
insertion loss and loss factor expressions, n is simply replaced by Qroc. For the inductive and capactive elements, we have to compare normalized with actual reactances:
(5.41a)
x
1
1
1 c = jQC - j(ro/ roc)C -
-ro-C
1
(5.41b)
This reveals that the actual inductance and capacitance L and C are computed from the
nonnalized L and C as
(5.42a)
(5.42b)
For the high-pass filter the parabolically shaped frequency response has to be
mapped into a hyperbolic frequency domain behavior. This can be accomplished
through the transformation
234
-roc
ro = -
(5.43)
The correctness of this transformation is immediately apparent when the normalized

cut-off frequency n = +1 is substituted in (5.43). This assigns the actual cut-off frequency ro = =t=roc to the high-pass filter, consistent with Figure 5-26.
30
25
20
"
=
.=
15
<
10
~
c
~
0
0
0.2 0.4 0.6 0.8
1.2 1.4
Frequency, GHz
1.6
1.8
Figure 5-26 Conversion of standard low-pass filter prototype into highwpass

realization. Cut-off frequency is fc =1 GHz.
Care has to be exercised in de-normalizing the circuit parameters. We note
xL = ju.
AL = -}.rocL = -1
(5.44a)
(5.44b)
0)
J c
= jO.C =
ro
jrocC
jroC
. -
= JroL
Thus, it follows that

1
C=-
rocL
(5.45a)
(5.45b)
This makes intuitive sense since it is known from fundamental circuit theory that a firstorder high-pass filter can be obtained from a low-pass filter by replacing the inductors
with capacitors or vice versa. Equations (5.45) are the logical extension to higher-order
filters.
The bandpass filter requires a more sophisticated transformation. In addition to
scaling, we also have to shift the standard low-pass filter response. The mapping from
the normalized frequency Q to the actual frequency ro is best explained by considering
Figure 5-27.
1.5
c:
-r----- :-----------i----------:~-------:
'I
'
'I
'
1
I
"t:S
I
4
" ~~:
Q)
t=
0 --- ~ ~ ~--- --; - ~ - ~-- - - --.. :. . -------.. --- ~ --- ------- ~- ----------
0~
";
OOu
I
-rou
.0.5
-1 ,..,..,. .... ,.,.,..,.,.,.I .. .,,.,..,..,.., ..........: ...................... ~
....................
.~
1I
II
II
-1.5
:
'
'
"'
ro
/.
,.,.~ ...................... ..
:
II
'I
'
0
0
I
I
I
-2LL~--~~--~--~~--~--~~~
-5
-4
-3
-2
1
0
-I
Frequency, ro
Figure 5-27 Mapping from standard frequency !l into actual frequency ro.
Lower cut~off frequency is roL = 1 and upper cut-off frequency is rou = 3.
The functional relation that achieves scaling and shifting is
1
ro
1 )
.Q = ffiuf (J)c- roLf <Oc O>c- ro/ roc
roc ( ro roc)
= ffiu- roL roc- CO
(5.46)
where the upper and lower frequencies O>u, roL define the bandwidth expressed in
rad/s (BW = O>u- roL) of the passband located at roc = ro0 . In other words, the cutoff frequency roc now defines the center frequency ro0 as mentioned earlier. Using ro0
and (5.23), it is possible to rewrite (5.46) as
ro_o__ E
Q = __
C.Ou- roL
The upper and lower frequencies are the inverse of each other:
(5.47)
236
rou
roo
= roo
(5.48)
00L
a fact that can be employed to specify the center frequency as the geometric mean of
the upper and lower frequencies, ro0 = Jro uc.oL. The mapping of this transformation is
verified if we first consider Q = 1 . Equation (5.46) is unity for ro = O>u and
0) :: (1) L . For n :: 0 we obtain (1) = roo . The frequency transformations are therefore as follows:
- 1 ~ .Q ~ 0 ~
O>L
< 0> < - O>o
The result of this transformation applied to the low-pass filter prototype is shown in
Figure 5-28.
25
~
a0
20
15
c::
10
5
o~~--~~~~--~~~~~--~~
0.2 0.4
0.6 0.8 1 1.2 1.4 1.6

Frequency, GHz
1.8
Figure 528 Conversion of standard lowpass filter prototype into bandpass

realization with lower cut-off frequency fL = 0. 7 GHz, upper cut-off frequency
fu 1.3 GHz, and center frequency of 10 1 GHz.
The circuit parameters are next transfonned according to the assignment
nL :: 1(
1'XL :: 1u
<.Oo
) L
Wu- O>L
. L- + -1= ]W
-
}roC
(5.49)
which yields for the series inductor Lin (5.49) the denormalized series inductor L
237
L=
L
rou- roL
(5.50a)
and the denonnalized series capacitance C
C=
rou- roL
(5.50b)
ro0L
The shunt capacitor is transformed based on the equation
Ac = 1(
J.Bc = Ju.
roo
rou- roL
. c- + -1e)c = JOl
jroL
(5.51)
to the following two shunt elements:
L = rou- roL
ro~C
C=--rou- roL
(5.52a)
(5.52b)
Referring to Figure 5-17, we see that a normalized inductor is transformed into a series
inductor and capacitor with values given by (5.50). On the other hand, the normalized
capacitor is transformed into shunt inductor and capacitor, whose values are stated by
(5.52).
The bandstop filter transformation rules are not explicitly derivedt since they can
be developed through an inverse transform of (5-47) or by using the previously derived
high-pass filter and applying (5.49). In either case, we find for the series inductor the
series combination
(5.53a)
C = l/[(rou-roL)L]
(5.53b)
and for the shunt capacitor the shunt combination
L = 1/[(rou-roL)C]
(5.54a)
(5.54b)
The resulting frequency response for the band-stop filter is shown in Figure 5-29.
Table 5-5 summarizes the conversion from the standard low-pass filter to the four
filter realizations.
238
25
fg
20
c:f
.:::
m
~
<
15
10
5
o~~--~~~~--~~~~~--~~
0.2
0.4
0.6 0.8 1 1.2 1.4 1.6

Frequency, GHz
1.8
Figure 5-29 Conversion of standard low-pass filter prototype into band-stop

realization with center frequency of t0 = 1 GHz. Lower cut-off frequency is
fL = 0.7 GHz and upper cut-off frequency is fu =1.3 GHz.
Table 5-5
Transformation between normalized low-pass filter and actual bandpass

and bandstop filter ( B W = ro u - roL )
Low-pass
prototype
1c
T
= gk
= gk
Low-pass High-pass
-roc
1 -rocL
T
Bandstop
Bandpass
..
L
-BW
-BW
ro02 L
1 -rocc ) -rocc _f_+

T
~BW
;
~
2
f rooC
BW
tl
(BW)L
(BW)LT_
J
1
...
..
..
1
(BW)C
(BW)C
roo
roo
Dpedance transformation
In the original filter prototype shown in Figure 5-17 we have unit source and load
~istances
except for the even-numbered Chebyshev filter coefficients listed in

able 5-4. If, however, either the generator resistance go or the load resistance RL is
:quired to be unequal to unity, we need to scale the entire impedance expression. This
1accomplished by scaling all filter coefficients by the actual resistance RG . That is,
RG = 1RG
(5.55a)
= LRG
c- =c-
(5.55b)
RG
- = RLRG
RL
(5.55c)
(5.55d)
'
pnere the tilde expressions are again the resulting actual parameters and L, C, and R L
~ the values of the original prototype.
t In Example 5-4 we demonstrate the design of a Chebyshev bandpass filter based
the low-pass prototype.
----------------------------~&uM~
Example S-4: Chebysehev bandpass filter design
An N =3 Chebyshev bandpass filter is to be designed with a 3 dB
passband ripple for a communication link. The center frequency is at
2.4 GHz and the filter has to meet a bandwidth requirement of 20%.
The filter has to be inserted into a 50 0 characteristic line impedance. Find the inductive and capactive elements and plot the attenuation response in the frequency range 1 to 4 GHz.
Solution:
'
'
.
'
.
:.
From Table 5-4(a) we find that the coefficients for a

standard low-pass N = 3 Chebyshev filter with 3 dB ripples in the
g 0 = g4 = 1 ,
g 1 = g 3 = 3.3487,
and
pass-band
are
g2 = 0.7117. In this filter prototype we assumed that both generator and load impedances are equal to unity. In our problem, however,
we have to match the filter to 50 n line impedances. Thus we must
Chapter 5 An Overview of RF Filter De1lgn
240
apply scaling as described by (5.55). The resulting circuit is shown

in the following figure:
C2 = 14.234 mF
This is still a low-pass filter with cut-off frequency of roc

1 or
fc = l/(27t) = 0.159 Hz. We can next apply the frequency transformation to change the low-pass filter into a bandpass filter:
9
rou = 1.1(27t2.4xl0 ) = 16.59x10

roL = 0.9(27t2.4x10
= 13.57x10
and
ro0 = JroLrou
= 15xi09
The actual inductive and capacitive values are defined in (5.50) and
(5.52):
= Wu 2-
roL
= 0.94 nH
ffioC2
C2 =
c2
C.Ou- ffiL
= 4.7 pF
The final circuit is shown in Figure 5-30 together with the resulting
graph for the attenuation response.
Filter Implementation
241
60
50
... 40
.a
30
20
10
o~----~----~~~~--~----~----~
1.5
2.5
3.5
Frequency, GHz
Figure 530
Attenuation response of a third-order 3-dB ripple bandpass
Chebyshev filter centered at 2.4 GHz. The lower cut-off frequency is

GHz and the upper cut-off frequency is f u = 2.64 GHz.
f L = 2.16
The filter design becomes almost a cook-book approach if we

start from the standard low-pass filter and subsequently apply the
appropriate frequency transformation and component scaling.
5.3 Filter Implementation

Filter designs beyond 500 MHz are difficult to realize with discrete components
because the wavelength becomes comparable with the physical filter element dimensions, resulting in various losses severely degrading the circuit performance. Thus, to
arrive at practical filters, the lumped component filters discussed in Section 5.2 must be
Chapter 5 An 0Y8f'VIew of RF Filter Design
242
converted into distributed element realizations. In this section, some of the necessary
tools are introduced-namely, Richards transformation, the concept of the unit element, and Kuroda's identities.
To accomplish the conversion between lumped and distributed circuit designs,
Richards proposed a special transformation that allows open- and short-circuit transmission line segments to emulate the inductive and capactive behavior of the discrete
components. We recall that the input impedance Zin of a short-circuit transmission line
(ZL = 0) of characteristic line impedance Z 0 is purely reactive:
(5.56)
Here, the electric length e can be rewritten in such a way as to make the frequency
behavior explicit. If we pick the line length to be A.0 / 8 at a particular reference frequency f 0 = v PI 'A 0, the electric length becomes
e = J3 A.o = 2rt/ vP = ~ L = ~ n
8
vP
Sf 0
4f0
(5.57)
By substituting (5.57) into (5.56), a direct link between the frequency-dependent inductive behavior of the transmission line and the lumped element representation can be
established:
(5.58)
where S = j tan (1t!l/ 4) is the actual Richards transform. The capacitive lumped element effect can be replicated through the open-circuit transmission line section
(5.59)
Thus, Richards transformation allows us to replace lumped inductors with short-circuit
stubs of characteristic impedance Z 0 = L and capacitors with open-circuit stubs of
characteristic impedance Z0 = 1 I C .
It is interesting to note that the choice of /...0 / 8 as line length is somewhat arbitrary. Indeed, several authors use ')...0 / 4 as the basic length. However, /...0 / 8 is more
convenient since it results in smaller physical circuits and the cut-off frequency point in
the standard low-pass filter response is preserved (i.e., S = j 1 for f = f 0 = f c). In
FUtlt' Implementation
243
Section 5.5.3 we will encounter a bandstop filter that requires a A.0 / 4 line length to
meet the expected attenuation profile.
Richards transformation maps the lumped element frequency response in the
range of 0 S f < oo into the range 0 ~ I S 4f 0 due to the periodic behavior of the tangent function and the fact that all lines are A.0 /8 in length, a property that is known as
commensurate line length. To obtain the inductive responses, we need to restrict the
domain to 0 $; f $; 2/0 . Because of this periodic property, the frequency response of
such a filter cannot be regarded as broadband.
5.3.1
Unit Elements
When converting lumped elements into transmission line sections, there is a need
to separate the transmission line elements spatjaJiy to achieve practically realizable
configurations. This is accomplished by inserting so-called unit elements (UEs). The
unit element has an electric length of 9 = ~(I I I 0 ) and a characteristic impedance
ZuE. The two-port network expression in cliain parameter representation is immediately apparent from our discussion in Chapter 4. We recall that the transmission line
representation is
[UE] =
[AvE BvEJ CuE DuE
cos9
}sine
cose
(5.60)
1
ZuE
where the definition of Sis given by (5.58). The use of the unit elements is discussed
best by way of a few examples, as presented in Section 5.3.4.
5.3.2
Kuroda's Identities
In addition to the unit element, it is important to be able to convert a practically

difficult-to-implement design to a more suitable filter realization. For instance, a series
inductance implemented by a short-circuit transmission line segment is more complicated to realize than a shunt stub line. To facilitate the conversion between the various
transmission line realizations, Kuroda has developed four identities which are summarized in Table 5-6.
We should note that in Table 5-6 all inductances and capacitances are represented
by their equivalent Richards transformations. As an example we will prove one of the
identities and defer proof of the remaining identities to the problems at the end of this
.chapter.
244
Table 5-6
Kuroda's Identities
Initial Circuit
Yc
Kuroda's Identity
=S/~
==
ZL =SZ/N
Unit
element
!--"
'I'
Yc:
'I'
Unit
element
z,
'I'
Z2 /N
ZL =Z,S
S/(NZ2 )
'I''I'Y~
Unit
element
Unit
element
z2
NZ 1
I:
Yc = S/(NZ2)
Yc=SIZ2
0
Jl
Unit
element
Unit
element
z,
II
NZ,
i[
N: 1
ZL =SZ/N
)
Z 1-
= ZS~
I
(
'l
Unit
element
Unit
element
Z 21N
z2
"
~
~
...
~[
1 :N
N = 1 + Z 2 /Z 1
--------------------------~RF8UMW4
Example 5-5: Prove the fourth of Kuroda's identities from

Table 5-6
Solution:
It is convenient to employ chain parameter representation of the shunt connected inductor (see Table 4-1 for the corresponding ABCD-matrix) and the unit element as given in (5.60) to
write the left-hand side as follows:
245
" ' Implementation
[~ ~L
:::
1 JI-S2 s
z2
sz 1
z2s
:::
2 -+ J1-S sz
Z
1
z2s
z
1 + _3
zl
Similarly, we can write the ABCD-matrix for Kuroda's fourth identity, or the right-hand side:
0
[
1/N
0 ]
trans
ind
UE
where subscripts UE, ind, and trans indicate chain parameter matrices for unit element, inductor, and transformer, respectively. After
carrying out the multiplication between the matrices, we obtain the
following ABCD-matrix describing Kuroda's identity:
A
[
M ~~)
1
cnJR J1-s2 ~+-~
z2 sz 1
which is identical with the left-hand side, if we set

N = 1 + Z 2/ Z 1 . The remaining three Kuroda identities can be
proved in a similar fashion.
We see again the importance of the ABCD network representation, which allows us to directly multiply the individual element
networks.
Examples of Microstrlp Filter Design

In the following two examples we will concentrate on the design of a low-pass
bandstop filter. The bandstop design will be conducted based on the aforemenRichards transformation followed by employing Kudora's identities. Specifithe bandstop design requires some attention in converting from lumped to
, ..u .......~ elements.
The practical filter realization proceeds in four steps:
Chapter 5 An Overvtew of RF Filter Dealgn
246
1. Select the normalized filter parameters to meet the design criteria.

2. Replace the inductances and capacitances by equivalent A.0 / 8 transmission lines.
3. Convert series stub lines to shunt stubs through Kudora's identities.
4. De-normalize and select equivalent microstrip lines (length, width, and dielectric
constant).
Specifically, step 4 requires knowledge of the appropriate geometric dimensions of the
respective microstrip lines, a subject that is discussed in detail in Chapter 2. According
to these four steps, let us now discuss the two examples.
The first design task involves a low-pass filter which is formulated as follows:
Project I
Design a low-pass filter whose input and output are matched to a 50 Q impedance
and that meets the following specifications: cut-off frequency of 3 GHz; equi-ripple of
0.5 dB; and rejection of at least 40 dB at approximately twice the cut-off frequency.
Assume a dielectric material that results in a phase velocity of 60% of the speed of
light.
In solving this problem, we proceed according to the previously outlined four
steps.
Step 1 From Figure 5-22, it is seen that the filter has to be of order N = 5, with
coefficients
g1
= 1.7058 = g 5, g 2 = 1.2296 = g 4 , g 3 = 2.5408, g6
= 1.0
The normalized low-pass filter is given in Figure 5-31.
cl = c5 = 1.7058
Figure 5-31
c3= 2.5408
L2= L4= 1.2296
Normalized low-pass filter of order N = 5.
Step 2 The inductances and capacitances in Figure 5-31 are replaced by open
and short circuit series and shunt stubs as shown in Figure 5-32. This is a direct consequence of applying Richards transformation (5.58) and (5.59). The characteristic line
impedances and admittances are
Y1
= Ys = gl , Y3 = g3, Z 2 = Z4
= g4
Filter Implementation
247
S.C.
r0
Figure 5-32
S.C.
=1
Replacing inductors and capacitors by series and shunt stubs

(o.c. =open circuit line, s.c. =short circuit Una).
Step 3 To match source and load sides, and to make the filter realizable, unit
elements are introduced with the intent to apply the first and second of Kudora's identities (see Table 5-6) to convert all series stubs into shunt stubs. Since we have a fifthorder filter we must deploy a total of four unit elements to convert all series connected
short-circuited stubs into shunt connected open-circuit stubs. To clarify this process we
divide this step into several substeps.
First, we introduce two unit elements on the input and output ends of the filter, as
S.C.
rG = 1
Figure 5-33
ZuEt = 1
S.C.
Zu2 = 1
Deployment of the first set of unit elements (U.E. = unit element).
The introduction of unit elements does not affect the filter performance since they
~.
~
are matched to source and load impedances. The result of applying Kuroda's identities
to the first and last shunt stubs is shown in Figure 5-34.
248
z, = 0.6304
Z2
S.C.
= 1.2296
S.C.
z4 = Z2
S.C.
Z 5 =Z1
S.C.
U.E.
Z3 = 0.3936
Figure 5-34
Converting shunt stubs to series stubs.
This version of the circuit is still nonrealizable because we have four series stubs.
To convert them to shunt connections, we have to deploy two more unit elements, as
2 )= 0.6304
Z2= 1.2296
S.C.
S.C.
z4 = ~
S.C.
U.E.
z3:; ;: 0.3936
Figure 535
Deployment of the second set of unit elements to the fifth-order filter.
Again, the introduction of unit elements does not affect the performance of the filter since they are matched to the source and load impedances. Applying Kuroda's iden
tities to the circuit shown in Figure 5-35, we finally arrive at the realizable filter design,
Step 4
De-normalization involves scaling the unit elements to the 50 Q input

and output impedances and computing the length of the lines based on (5.57). Using
Flltlr Implementation
249
Zu3 = 1.6304 ZvEI = 1.5992 Zun = 1.5992 Zu4 = 1.6304
r0 = 1
Z1 = 2.5862
Figure 5-36
Z2
0.4807
z3 = 0.3936 z4 = 0.4807
Zs = 2.5862
Realizable filter circuit obtained by converting series and shunt stubs

using Kuroda's identities.
=0.6c = 1.8 X 108 rn/s,
the length is found to be

l = (}.,o/8) =
vp/(8/0 ) = 7.5 mm. The final design implemented in microstrip lines is shown in
Figure 5-37(a). Figure 5-37(b) plots the attenuation profile in the frequency range 0 to
3.5 GHz. We notice that the passband ripple does not exceed 0.5 dB up to the cut-off
frequency of 3 GHz.
The second design project involves a more complicated bandstop filter, which
requires the transformation of the standard low-pass prototype with a unity cut-off frequency into a design with specified center frequency and lower and upper 3 dB frequency points.
Vp
Project II
Design a maximally flat third-order bandstop filter whose input and output are
matched to a 50 Q impedance that meets the following design specifications: center
frequency of 4 GHz and bandwidth of 50%. Again, we assume a dielectric material that
results in a phase velocity of 60% of the speed of light.
This design requires a careful analysis when converting from lumped to distributed
elements. Specifically, when dealing with bandstop designs, we require either maximal
or minimal impedance at the center frequency f 0 depending on whether series or shunt
connections are involved. With our previous definition of Richards transformation based
on A.0/8 line segments, we encounter the difficulty that at f = f 0 , (5.58) yields a tangent value of 1, and not a maximum. However, if a line length of A.0 / 4 is used, then the
tangent will go to infinity as required for a bandstop design. Another aspect that we have
to take into account is the fact that we want the Q = 1 cut-off frequency of the lowpass prototype filter to be transformed into lower and upper cut -off frequencies of the
bandstop filter. This is done by introducing a so-called bandwidth factor bf
bf == cot
(1t2ro000L)
= cot
[1t2(1 - s2bw)J
(5.61)
----------------------------------------------------------~
Chapter 5 An Overview of RF Ftlter Dnlglt ,
250
son
81.5
80.0 0
80.00
19.7 0
24.00
129.3 0
81.5
son
129.3 n
24.00
(a) Microstrip line low-pass filter implementation

5
4.5
4
3.5
-~
2.5
d'
=
11)
1.5
1
0.5
-....
_.... ~
0
-0.5 0
0.5
~J
"'- /
2.5
2
Frequency, GHz
1.5
3.5
{b) Attenuation versus frequency response

Figure 537 Final microstrip line low-pass filter.
ro0
ro0
the
where sbw = (O>u- O>r)l

is the stopband width and
= (rou + O>r)/ 2 is
center frequency. Multiplying the Richards transformation for A-0 / 4 line lengths by bf
at the lower or upper frequency points reveals that the magnitude of the product is equal
to unity. For instance, for the lower frequency point 0>L, it follows that
(bf)S j _
ro - ro L
cot(~roL)tan(~O>L):: 1
2 ro0
2 ro0
This corresponds to a n = 1 cut-off frequency in the normalized low-pass filter :

response. Similarly, for the upper cut-off frequency O>u we have
;j
(bf)SI
ro
"'u
= cot(~roL)tan(~rou) = cot(~roL)tan[~(2 ro0 - roL)] =-1

2 ro0
2 ro0
2 ro0
ro0
lI
Fllttr Implementation
251
which corresponds to a .Q = -1 cut-off frequency in the normalized low-pass filter.

With these preliminary remarks, we are now ready to proceed according to the four steps.
Step 1 From Table 5-2, the coefficients for a maximally flat normalized lowpass filter prototype of third order are
gl = 1.0
= g3, g2
= 2.0, g4 = 1.0
Thus, the normalized low-pass filter has the form shown in Figure 5-38.
Figure 538
Normalized third-order low-pass filter.
Step 2 The inductances and capacitances in Figure 5-38 are replaced by open
and short circuit series and shunt stubs, as depicted in Figure 5-39. The line impedances
and the admittance are multiplied by the bandwidth factor (5.61).
zt
= z3 = bf gt,
S.C.
Figure 5-39
Y2 = bf g2
S.C.
Replacing inductors and capacitors by series and shunt stubs.
Step 3 Unit elements of 'A0 / 4 line length are inserted and Kudora's identity is
used to convert all series stubs into shunt stubs as seen in Figure 5-40.
252
Chapter 5 An overview of RF Filter Dtslgn
S. C.
S.C.
U.E.
Z uEJ
r0
(a) Unit elements at source and load sides

1
Zur; 1 = 1.4142 Zu2 = 1.4142
z, = 3.4142
Figure 5-40
rI. = 1
=I
z3= 3.4142
(b) Conversion from series to shunt stubs

Introducing unit elements and converting series stubs to shunt
stubs.
De-normalization the unit elements and explicit computation of the individual line lengths can now be conducted. Using the phase velocity
8
vP = 0.6c = 1.8x 10 m/s,
the length is computed to be l = (A.0 / 4)=
v PI ( 4f 0 ) = 15 mm. Thus, the resulting design in micros trip line implementation is
as shown in Figure 5-41.
Finally, for this bandstop filter we can also utilize a commercial simulation package such as MMICAD to simulate the filter response of the microstrip line configura
tion shown in Figure 5-41. The attenuation profile is given in Figure 5-42, and shows
that the filter specifications are met.
Step 4
Coupled Fitter
253
son
70.7 n
170.7 n
Figure 5-41
son
70.711
60.4 n
170.7 n
Characteristic impedances of final microstrip line implementation of

bandstop filter design.
'
4.5
3.5
2.5
<
1.5
0.5
0
-0 5
'2
\
\
_.,.
2.4
I
I
_\
\
J
2.8
\
3.2
3.6
4.4
4.8
5.2
""
...._
5.6
Frequency, GHz
Figure 542 Attenuation versus freguency response for third-order bandstop
filter.
5.4 Coupled Filter
The literature is extensive when dealing with coupled filter designs and analyses.
For our cursory treatment we will introduce only the most salient points and refer the
( reader to the references listed at the end of this chapter.
Our discussion briefly covers the odd and even wave coupling of transmission
: lines through a common ground plane, which results in odd and even characteristic line
254
impedances. This sets the stage to an understanding of the coupling between two strip
lines and their input/output impedances as part of a two-port chain matrix representation. Cascading these elements gives rise to bandpass filter structures that are most easily designed with the aid of RF circuit simulation packages.
5.4.1
Odd and Even Mode Excitation
A simple modeling approach of coupled microstrip line interaction is established

when considering the geometry depicted in Figure 5-43. The configuration consists of
two lines separated over a distance S and attached to a dielectric medium of thickness d
and dielectric constant ,. The strip lines are W wide, and the thickness is negligible
when compared with d. The capacitive and inductive coupling phenomena between the
lines and ground is schematically given in Figure 5-44. Here equal indices denote selfcapacitances and inductances, whereas index 12 stands for coupling between line 1 and
line 2 (which is equal to coupling between line 2 and line 1).
Figure 5-43 Coupled microstrip lines.
We can now define an even mode voltage Ve and current le and an odd mode voltage Vod and current I od in terms of the total voltages and currents at terminals 1 and 2
such that
(5.62a)
and
(5.62b)
This is consistent with the voltage and current convention shown in Figure 5-44. For
even mode of operation (Ve, /e), voltages are additive and currents flow in the same
direction. However, for odd mode of operation (V0 d, / 0 d) the terminal voltages are sub- ,
tractive and currents flow in opposite directions.
255
Coupled Filter
Figure 5-44
Equivalent circuit diagram and appropriate voltaQe and current

definitions for a system of two loss less coupled transmission lines.
The benefit of introducing odd and even modes of operation is seen when establishing the fundamental equations. It can be shown that for two lines we get a set of
first-order, coupled ordinary differential equations similar in form to the transmission
line equations in Chapter 2 :
dVe
- dz = JOl(Lll + L 12)/e
(5.63a)
(5.63b)
and
(5.64a)
(5.64b)
What is important to notice is the fact that even and odd modes allow us to decouple the
governing equations. The characteristic line impedances Zoe and Z 00 for the even and
odd modes can be defined in terms of even and odd mode capacitances Ce, Cod and the
respective phase velocities as follows:
256
zoe =
1
C ' Zoo =
V pe
1
C
Vpo
(5.65)
od
If both conductors are equal in size and location, we can conclude for the even mode
ce = c11 = c22
(5.66a)
and for the odd mode
(5.66b)
The capacitances are, in general, difficult to find since fringing fields and different
media have to be taken into account. For instance, even the strip line conductor over a
dielectric substrate cannot be computed based on the simple capacitance per unit length
formula C 11 = e0 er( wId) because the width-to-thickness ratio is not sufficiently
large for this formula to apply. Moreover, the cross-coupling capacitance C 12 requires
a very intricate treatment. For this reason, it is common practice to resort to a numerically computed impedance grid, such as the one shown in Figure 5-45.
180
160
N~
Q)
140
120
--
100
Q)
aQ)
"'s
=
0
80
;>
~
60
40
20 ..... j.....d
20
~--~~~~~~MM~
40
60
80
100
120
Odd mode impedance Z00, !l
Figure 545
Even and odd characteristic impedance for microstrip lines.
~pled Finer
5.4.2
257
Bandpass Filter Section
We turn our attention to two microstrip lines as the main building block of a band:pass filter shown in Figure 5-46. Both the geometric arrangement with input and output
:ports and open-circuit conditions and the corresponding transmission line representa;~on are depicted.
'
'
:!.
I
..
(a) Arrangement of
~~o microstrip lines

Figure 5-46
(b) Transmission line representation

Bandpass filter element.
Without delving into details of the rather complicated treatment (see Gupta in
er Reading), this configuration has the impedance matrix coefficients for open
~smission line segments in the form
= -j~(Zoe + Z00 )cot(~l)
= Z 22
(5.67a)
zl2 = -j~(Zo,- Zo) sin~~/) = Zzl
(5.67b)
Z 11
When cascading these building blocks into multiblock filter configurations, our
sire is to match both ports of this segment to the adjacent elements. This is also
, own as finding the image impedance. For the input impedance at port 1 we can write
,;
zin
v 1 AZL +B
T : : cz + D
1
(5.68a)
,d the output impedance at port 2

j .
)
-V2
ZL = 12
DZin +B
= CZin +A
(5.68b)
.ince we require Zin = Z L, we find from (5.68) that A = D and

Zm =
(5.69)
258
Chapter 5 An Overview of RF Fffter Des(fl
If (5.67) is converted into a chain matrix form the coefficients A, B, C, D can be determined. Inserting Band C into (5.69), one finds for the input, or image impedance
zin
= 2sin(l}l) (Zoe- Zoo) -(Zoe + Zoo) cos (~l)
(5.70)
The bandpass filter behavior of (5.70) becomes apparent when plotting the real part of
the input impedance response as a function of the electric length in the range
0 < f3l ~ 21t, as depicted in Figure 5-47.
50~------~--------~----------~------~
a..
-Ns 40
,_,
.
-a
j
]Zoe- Zoolf2
30 .......................
20
10
0.57t
L51t
1t
Electrical length, ~~
Figure 5-47
Input impedance behavior of equation (5.70). Zoe and Z 00 are

arbitrarily set to 120 n and 60
respectively.
.a ,
According to Figure 5-47, the characteristic bandpass filter performance is obtained-~

when the length is selected to be "A/4 or f3/ = n/2 . For this case the upper and lowe(
cut-off frequencies are found as
.:
- 1[Zoe -
(pt),,2 = el,2 = cos
Oe
Zoo]
z
(5.71)-'
Oo
Also noticeable is the periodic impedance response in Figure 5-47, which indicates that~
the upper operating frequency has to be band limited to avoid multiple bandpass filter :.
responses at higher frequencies.
5.4.3
Cascading bandpass filter elements
A single bandpass element as discussed in the previous section does not result in a :1
good filter performance with steep passband to stopband transitions. However, it is the }
/
:OUpled Filter
259
tbility to cascade these building blocks that ultimately results in high-performance

ilters. Figure 5-48 depicts a generic multielement design.
0, 1
fiUUUjlfff~Uiiillilliil Z(A>, Z0v
l, 2
lu jUUUfUUUUHU!I
z z<
(I.,
ijJUUUiUQnJifHIJI$11
z z
Oo
fu
l$jiiii11J111iftlllllllll
N - I,N RIIIW~
N,N + I
Figure 5-48
Multielement configuration of a fifth-order coupled-line

bandpass filter (N = 5).
To design such a structure that meets a particular bandpass filter specification, a

number of computations have to be performed. The following sequence of steps is
needed to translate a set of design requirements into a practical filter realization (see
Matthaei et al. in Further Reading).
Selection of standard low-pass .filter coefficients. Depending on whether a Butterworth or Chebyshev design with desired rejection and ripple is needed, the
designer can directly select the appropriate standard low-pass filter coefficients
go, g 1, ... , g N' g N + 1 listed in Tables 5-2 to 5-6.
Identification of normalized bandwidth, upper, and lower frequencies. From the
desired filter specifications for lower and upper frequencies roL, rou and the center frequency ro 0
= (rou + roL) / 2, we define the normalized bandwidth of the
filter as
BW
rou- roL
roo
(5.72)
This factor a1low us to compute the following parameters:
(5.73a)
260
1tBW
1;, i + 1
JN,N+ I
J
o2 KiKi+ 1
(5.73b)
1tBW
= Zo
-
(5.73c)
2gNgN+ 1
which in tum pennit us to detennine the odd and even characteristic line impedances:
Zool l,..I+ ]
= Zo[l-Zolii+l+(Zolii+l)]
'
(5.74a)
'
and
(5.74b)
where the indices i. i + 1 refer to the overlapping elements seen in Figure 5-48.
Here Z 0 is the characteristic line impedance at the beginning and the end of the
filter structure.
Selection of actual strip line dimensions. Based on Figure 5-45, the individual odd
and even line impedances can be converted into strip line dimensions. For
instance, if the dielectric material and the thickness of the PCB board are given,
we can detennine separationS, and width W of the copper strips. Normally, the
width will conform with the width of the other microstrip lines. Therefore, the
separation is the most common parameter that can be varied to achieve the impedances required by (5.74). The length of each coupled line segment has to be equal
to 'A,/ 4 at the center frequency, as described in Section 5.4.2.
The preceding steps result in a first and often crude design, which can be made
more precise by introducing length and width corrections to account for fringing field
effects. In addition, the use of simulation packages often allows further adjustments and
fine-tuning to ensure a design that actually performs according to the specifications.
5.4.4
Design Example
In the following example we go through the steps outlined in the previous section
by designing a particular bandpass filter.
COupled F11ter
261
----------------------------~&)A~
Example S-6: Bandpass filter design with coupled line transmission line segments
A coupled-line bandpass filter with 3 dB ripples in the passband is
to be designed for a center frequency of 5 GHz and lower and upper
cut-off frequencies of 4.8 and 5.2 GHz, respectively. The attenuation
should be at least 30 dB at 5.3 GHz. Select the number of elements
and find odd and even mode characteristic impedances of the coupled transmission lines.
Solution:
According to Section 5.4.3, the first step in the design

of this filter is to choose an appropriate low-pass filter prototype.
The order of the filter can be selected from the requirement of 30 dB
attenuation at 5.3 GHz. Using frequency conversion for the bandpass filter (5.46), we find that for 5.3 GHz the normalized frequency
of the low-pass filter prototype is
Q
roc ( ro - roc) :::: 1.4764

(l)c
ro
rou- roL
From Figure 5-21 we determine that the order of the filter should be
at least N == 5 to achieve 30 dB attenuation at Q = 1.4764. The
coefficients for an N = 5 Chebyshev filter with 3 dB ripples are
gl
= g5 = 3.4817, g2 = 84
= 0.7618, g3
= 4.5381, g6 = 1.
The next step in the design is to find the even and odd excitation mode characteristic impedances of the coupled transmission
lines as described by (5.74). The results of theses computations are
listed in the following table.
l
ZolI, 1'+I
Zoo(Q)
Zoe(Q)
0.1900
42.3056
61.3037
1
2
0.0772
46.4397
54.1557
0.0676
46.8491
53.6077
0.0676
46.8491
53.6077
0.0772
46.4397
54.1557
0.1900
42.3056
61.3037
282
To confirm the validity of our theoretical design, we can use

MMICAD to analyze the performance of the bandpass filter just
designed. The result of the simulation is shown in Figure 5-49.
30------~-------r------.-------~----------~~
25~-4---+------~------~------~-----4---4--~
~ 20
.;::: 15 1--------\~-------+--------+--------~------+--+----~
<
10~----~------~------+-------~-----4~----~
o~----~~--~L----~~------~~--~------~
4.7
4.8
4.9
5.0
5.1
5.2
5.3
Frequency, GHz
Figure 549 MMICAD simulations of the fifth-order coupled-line Chebyshev

bandpass filter with 3 dB ripple in the passband. The lower cut-off frequency is
4.8 GHz and the upper cut-off frequency is 5.2 GHz.
The filter response in Figure 5-49 confirms that the specifications

are met for fL and fu and that the attenuation at 5.3 GHz even
exceeds the 30 dB requirement.
Often the theoretical filter design leads to coefficients whose
validity must be double-checked against an RF circuit simulator to
test the actual performance.
Another reason for resorting to a simulation package is the need to verify the
design methodology independently and to test the filter performance over a range of
parameter variations in terms of geometry and dielectric properties. Most of these parametric studies can be accomplished with little effort on a computer. After the initial theoretical design idea, the computer simulations typically precede the actual board
construction and testing.
Summary
263
5.5 Summary
Our emphasis in this chapter has been an exposure to filter design concepts that
are ubiquitous in many RF/MW circuit designs. Rather than going into detailed derivations, the intent of this chapter is to present a generic discussion of some of the key
issues facing a design engineer in the construction of practical filter types.
Beginning with a general classification of high-pass, low-pass, bandpass, and
bandstop filters, we introduce a common terminology that is needed to understand the
common descriptors when developing filter specifications. Terms that are often used
such as cut~off, lower, upper, and center frequencies, shape factor, bandwidth, insertion
loss, and rejection, are defined and placed in context with simple first-order high- and
low~pass filters as well as series and parallel resonant circuits. Since the resonator circuits permit the realization of bandpass and bandstop designs, the sharpness of the
impedance or admittance behavior is quantified through the so-called quality factor:
Q = ro averaged stored energy

energy loss per cycle
a measure that can be further broken down into the filter QF and external QE quality
factors. Specifically, the notation of insertion loss
IL
pin
= IOlogp
= -lOlog(l - Irin 12 )
L
which defines the amount of power lost by inserting the filter between the source and
load ports, is of central important in the design of high-frequency filters. Depending on
the attenuation profile necessary to realize the various filter types, the loss factor
LF =
l
1 -jrinl2
is employed to realize a particular response.

To enable a more comprehensive approach, the low-pass filter design based on a
nonnalized frequency scale is chosen as the standard type. Through frequency scaling
and shifting, all filter types can then be readily realized. The benefit of this approach is
that only a few sets of standard low-pass filter coefficients have to be derived depending
on whether a Butterworth filter with a maximally fiat profile or Chebyshev filter with an
equiripple attenuation profile is desired.
The practical implementation is achieved through Richards transformation:
s = jtan(~n)
264
This transformation is central in establishing a link between lumped capacitive and

inductive elements and distributed transmission line theory. The various series and .
shunt transmission line segments can be spatially separated through unit elements
before Kudora's identities permit the conversion of some of the transmission sections
into easy-to-implement segmental elements. In particular, series inductive configurations are often easier to implement than stub elements. With the aid of Kudora's identi
ties this can be accomplished elegantly.
The fact that the proximity of strip lines causes electromagnetic coupling is
exploited to design bandpass and bandstop filters. Without delving into the theoretical
explanations too deeply, two line segments are used as the basic building block of a
two-port network representation. Through odd and even mode impedance analysis we
can find the image impedance
zin
1
J(Zoe- Zoo) 2 -(Zoe+ Zoo)2cos2 (~!)
= 2sin(~l)
as the characteristic bandpass response. This single element can be cascaded into multiple section filters to fulfill various design requirements. By using an RF/MW simulation
package, the same example is revisited and the coupled filter response is computed as a
function of various element numbers and geometric dimensions of the microstrip lines.
Although the topic of filter design could only cursively be covered, Chapter 5
should convey the basic engineering steps needed to arrive at a functional high
frequency filter realization. We attempted to make the process of picking the appropriate filter coefficients, scaling the results to actual frequencies, and implementing the
process in microstrip lines as much of a cookbook approach as possible. However,
Chapter 5 should also make clear the usefulness of commercial simulation packages in
carrying out a detailed numerical analysis. Indeed, for most modern filter design examples, an RF/MW simulation package is an indispensable tool to predict the filter performance. Moreover, from the circuit schematic it is relatively straightforward to use
special layout programs to generate the actual PCB layout file that becomes the basis
for the physical board construction.
Further Reading
S. Butterworth, "On the Theory of Filter Amplifiers," Wireless Eng., Vol. 7, pp. 536541, 1930.
K. C. Gupta, R. Garg, and I. J. Bahl, Microstrip Lines and Slot Lines, Artech House~
Dedham, MA, 1979.
Problems
265
G. L. Matthaei, et al., Microwave Filters, Impedance-Matching Networks, and Coupling Structures, McGraw-Hill, New York, 1964.
C. G. Montgomery, R. H Dicke, and E. M. Purcell, Principles of Microwave Circuits,

MIT Radiation Laboratory Series, Vol. 8, McGraw-Hill, New York, 1948.
D. M. Pozar, Microwave Engineering, 2nd ed., John Wiley, New York, 1998.
P. A. Rizzi, Microwave Engineering: Passive Circuits, Prentice Hall, Englewood Cliffs,

NJ, 1988.
L. Weinberg, Network Analysis and Synthesis, McGraw-Hill, New York, 1962.
Problems
5.1 For the simple integrator circuit shown,

50Q
2pF
50Q
determine the following quantities:

Transfer function H ( ro) = V 2 / V 0
Attenuation versus frequency behavior a ( ro)
Phase versus frequency behavior <p ( ro)
Group delay t g
Plot these factors for the frequency range from DC to 1 GHz.
5.2
Derive expressions for internat external, and loaded quality factors for the
standard series and parallel resonance circuits discussed in Section 5.1.4.
5.3
In Section 5.1.5 the admittance of the parallel resonance circuit is expressed

in terms of a quality factor expression. Prove the resulting equation (5 .29).
5.4
For the filter circuit shown,

500
lOnH
lOpF
500
Find the loaded, unloaded, and external quality factors. In addition determine the power generated by the source as well as the power absorbed by the
load at resonance. Furthermore, plot the insertion loss as a function of frequency in the range of +50% of the resonance frequency.
5.5
Repeat Problem 5.4 for the following filter circuit:

lOnH
500
Z0 =50Q
lOpF
Z0 =50Q
son
5.6
You are required to build a low-pass Butterworth filter that provides an

attenuation value of at least 50 dB at f = 1.5f 3dB Which filter order is
required? How many components (inductors and capacitors) do you need to
realize this filter?
5. 7
Design a prototype low-pass Butterworth filter that will provide at least

20 dB attenuation at the frequency off = 2/JdB.
5.8
Plot the insertion loss of a low-pass Chebyshev filter that has 6 dB ripple in
the passband and at least 50 dB attenuation at f = 2f cut-off.
5.9
Using the low-pass prototype developed in Problem 5.7, design a high-pass

filter with cut-off frequency of l GHz. Plot the attenuation profile.
5.10 To suppress noise in a digital communication system a bandpass RF filter is

required with a passband from 1.9 GHz to 2 GHz. The minimum attenuation
of the filter at 2.1 GHz and 1.8 GHz should be 30 dB. Assuming that a
0.5 dB ripple in the passband can be tolerated, design a filter that will use a
minimum number of components.
5.11
In the design of an amplifier for cellular phone applications it is discovered

that the circuit exhibits excessive noise at 3 GHz. Develop a bandstop filter
with a center frequency of fc = 3GHz and bandwidth of 10% at fc with
30 dB minimum attenuation in the stopband.
Problema
287
5.12 In previous chapters we examined the input impedance behavior of an opencircuit stub assuming that open-circuit conditions at the end are ide al. In
practical realizations, due to fringing fields, leakage occurs. This can be
modeled as an additional parasitic capacitance, as shown below:
Using your favorite mathematical program, find the input impedance of the
50 n open-circuit stub of length l = 1 em for frequencies ranging from 10
MHz to 100 GHz. In your computations assume that an equivalent load
capacitance is C oc = 0.1 pF and the phase velocity of the line is
8
vP = 1.5x10 m/s. Compare your results to the input impedance behavior
of the ideal open-circuit and short-circuit stubs.
5.13
Assuming all physical parameters of the open-circuit stub to be the same as

in Problem 5.1 2, find the effective fringing capacitance Coc if the lowest
frequency (at which the input impedance of the stub is equal zero) is
3.3 GHz.
5.14 After reconsidering the design in Problem 5.12, it is decided to use an opencircuit stub of half of the length (i.e., l = 5 mm). Since the board is already
manufactured with a I em stub, you cut a slit in the middle so that the length
of the resulting stub is 5 mm, as shown below.
Due to proximity effects, the equivalent circuit in this case is as follows:
~-J 5 mm.1
Zo
lr---Cs_,.........t:::::;:5
nun::::::l
-rc~ -rc~
C.'i'=50 fF
Zo
CDC = 100 fF
268
Using a mathematical spreadsheet, compute the input impedance of this configuration for frequencies ranging from 10 MHz to 20 GHz, assuming that
the characteristic line impedance is 50 n and the phase velocity is
8
v P = 1.5x10 m/s. Compare the results with the input impedance behavior
of the 5 mm open-circuit stub taking into account a fringing capacitance
coc = 100 tF.
5.15
In Chapter 2 we introduced a quarter-wave strip line transformer that is able

to transform any real load impedance into any other real value. In our analysis we always assumed that there are no parasitic elements involved. In reality, the connection of two transmission lines with different impedances leads
to discontinuity in the line width as follows:
Due to this discontinuity, additional parasitic elements have to be taken into

account. The equivalent circuit for the above configuration is:
A, /4
For a load impedance of L = 25

and a Zo = 100
line impedance
find a characteristic impedance ztr of the quarter-wave transformer and
compute the input impedance Zin of the entire system for a frequency range
from 10 MHz to 20 GHz, assuming that the transmission line is quarterwave long and parasitic elements have the following values: L = 10 pH, C =
100 fF.
5.16
Prove the first three Kuroda's identities given in Table 5-6 by computing the
appropriate ABCD matrices.
5.17
Develop a low-pass filter with cut-off frequency of 200 MHz and attenuation
of 50 dB at 250 MHz. The flatness of the filter response is not a design consideration. Choose the filter implementation that requires the least number of
components.
Problema
269
5.18 Design a three-section bandpass filter with 3 dB ripples in the passband. The
center frequency is 900 MHz and the bandwidth is 30 MHz. Use a mathematical spreadsheet and plot the insertion loss of the filter.
5.19 In Project I of Section 5.3.3 we designed a microstrip realization of the
Chebyshev-type low-pass filter with 3 GHz cut-off frequency. Repeat this
design using an FR-4 substrate with dielectric constant of Er = 4.6 and
thickness of h = 25 mil. In addition, obtain the physical width and length of
each microstrip line.
5.20 Design a five-section bandstop filter having a maximally flat response. The
bandwidth of the filter should be 15% with a center frequency of 2.4 GHz.
The filter has to be matched to a 75 .Q impedance at both sides.
5.21
Design a fifth-order low-pass filter with linear phase response. The cut-off
frequency of the filter is 5 GHz. Provide two designs: the first one using
lumped elements and the second design using microstrip lines. In both cases
assume that a FR-4 substrate is used ( Er = 4.6, h =20 mil).
5.22 As a part of a satellite communication link, a bandpass filter for image rejection in the downconversion stage has to be designed. The bandwidth of a signal is 300 MHz and the center frequency is 10 GHz. It is essential to provide
maximally fiat response in the passband and obtain at least 40 dB attenuation at 10.4 GHz.
5.23 Prove equations (5.68a) and (5.68b) and show how (5.70) results.
CHAPTER
Active RF Components
ur focus in the first five chapters has been primarily geared toward passive RF devices and their electric circuit behavior. In this
chapter we extend and broaden our scope to include an investigation of various active
, circuit elements. Of specific interest for the design of amplifiers, mixers, and oscillators
are solid-state devices such as diodes and transistors. What complicates a unified treatment is the wealth of special purpose components developed and marketed by a range
of companies for a wide host of industrial applications. We cannot adequately address
the multitude of technological advances currently shaping the RF/MW commercial
markets. This is not the intent of this text~ rather we emphasize a number of key concepts driving the technological RF/MW evolution. These concepts are utilized later for
the design of amplifiers, mixers, oscillators, and other circuits developed in subsequent
chapters. Our approach intends to enable the reader to formulate and develop his or her
own network descriptions as part of an integrated strategy to construct suitable models
of analog RF circuits.
Before developing appropriate network models for active devices, a short discussion of solid-state physics involving pn and metal-semiconductor junctions is
presented. The aim is to provide a so1id-state perspective of the electric circuit representations derived from the physical device level. This is needed because
at high-frequency modes of operation, additional capacitive and inductive effects
enter the solid-state devices and affect their performance
the high-frequency behavior of many active devices markedly departs from that of
low-frequency components and therefore requires special treatment
to utilize simulation tools such as SPICE, or more dedicated RF CAD programs, a
working knowledge of the physical parameters must be obtained that directly or
indirectly influence the circuit behavior
271
Chapter 6 Active RF Components
272
Chapter 6 provides a concise summary of the most important semiconductor fundamentals that are encountered at high frequencies.
By analyzing the pn-junction and the Schottky contact, we gain a more complete
picture of electronic circuit functions that form the foundation of rectifier, amplifier,
tuning, and switching systems. In particular, the metal-semiconductor interface is
shown to be especially useful for high-frequency operations. It is the RF domain that
has seen many specialized diode developments. Chief among them are the Schottky,
PIN, and tunnel diode, to name but a few.
Next, our attention is turned toward the bipolar and field effect transistors, which
are more complex implementations of the previously investigated pn-junction and
Schottky contact. We learn about the construction, functionality, temperature, and noise
performance of the bipolar and the metal-semiconductor field effect transistors.
6.1 Semiconductor Basics

6.1.1
Physical Properties of Semiconductors
The operation of semiconductor devices is naturally dependent on the physical

behavior of the semiconductors themselves. This section presents a brief introduction to
the basic building blocks of semiconductor device modeling, particularly the operation
of the pn-junction. In our discussion we will concentrate on the three most commonly
used semiconductors: germanium (Ge), silicon (Si), and gallium arsenide (GaAs). Figure 6-1 (a) schematically shows the bonding structure of pure silicon: Each silicon atom
shares its four valence electrons with the four neighboring atoms, forming four covalent
bonds.
In the absence of thermal energy (i.e., when the temperature is equal to zero
degree Kelvin (rK :::: 0 or roc : : -273.15, where roK = 273.15 +roc]) all electrons are bonded to the corresponding atoms and the semiconductor is not conductive.
However, when the temperature increases, some of the electrons obtain sufficient
energy to break up the covalent bond and cross the energy gap W 8 = W c - W v, as
shown in Figure 6-l(b) (at room temperature r = 300K the bandgap energy is
equal to 1.12 eV for Si, 0.62 eV forGe, and 1.42 eV for GaAs). These free electrons
form negative charge carriers that allow electric current conduction. The concentration
of the conduction electrons in the semiconductor is denoted as n. When an electron
breaks the covalent bond it leaves behind a positively charged vacancy~ which can be
occupied by another free electron. These types of vacancies are called holes and their
concentration is denoted by p.
Electrons and holes undergo random motion through the semiconductor lattice as
a result of the presence of thermal energy ( T > 0K ). If an electron happens to meet a
Semiconductor Buies
273
Forbidden Band
or Bandgap
(b) Energy band levels

(a) Planar representation of covalent bonds
Figure 6-1 Lattice structure and energy levels of silicon.

(a) schematic planar crystal arrangement with thermal breakup of one valent bond
resulting in a hole and a moving electron for T > 0K.
(b) equivalent energy band level representation whereby a hole is created in the
valence band Wv and an electron is produced in the conduction band W0 The energy
gap between both bands is indicated by Wg.
hole, they recombine and both charge carriers disappear. In thermal equilibrium we
( have equal number of recombinations and generations of holes and electrons. The concentrations obey the Fermi statistics according to
n
= Ncexp [
We- WFJ
kT
(6.la)
(6.1b)
rf
r where
I'
~
2 3/ 2
N c, v = 2(2mn, PrtkT I h )
(6.2)
t are the effective carrier concentration in the conduction ( N c) and valence (N v )
~~!
bands, respectively. The terms W c and W v denote the energy levels associated with
the conduction and valence bands and W F is the Fermi energy level, which indicates
the energy level that has a 50% probability of being occupied by an electron. For
intrinsic (i.e., pure) semiconductors at room temperature the Fermi level is very close
'
274
m;
to the middle of the bandgap. In (6.2), m~ and

refer to the effective mass of elec
trons and holes in the semiconductor that are different from the free electron rest mass
due to interaction with the crystal lattice; k is Boltzmann's constant; h is Planck's constant; and T is the absolute temperature measured in Kelvin.
In an intrinsic semiconductor the number of free electrons produced by thennal
excitation is equal to the number of holes (i.e. n = p = n; ). Therefore, electron and
hole concentrations are described by the concentration law
np
= n;2
(6.3)
where n; is the intrinsic concentration. Equation (6.3) is true not only for intrinsic but
also for doped semiconductors, which are discussed later in this section.
Substitution of (6.1) into (6.3) results in the expression for the intrinsic carrier
concentration:
(6.4)
The effective electron and hole masses as well as the concentrations N c, N v, and
ni for T = 300K are summarized in Table 6-1 and are also listed in Table E-1 in
Appendix E.
Table 6-1
Effective concentrations and effective mass values at T =300K
Semiconductor
-3
-3
0.56
2.8xl0 19
1.04x1019
1.45xl0 10
0.55
0.37
L04x10 19
6.0x10 18
2.4x10 13
0.067
0.48
4.7x10 17
7.0xl0 18
1.79x106
m~lm 0
m;/m 0
Silicon (Si)
1.08
Germanium (Ge)
Gallium Arsenide (GaAs)
Nc(cm ) Nv (em
-3
n; (em
Classical electromagnetic theory specifies the electrical conductivity in a material

to be o = J IE, where J is the current density and E is the applied electric field. The
conductivity in the classical model (Drude model) can be found through the carrier concentration N, the associated elementary charge q, the drift velocity vd, and the applied
electric field E:
(6.5)
In semiconductors, we have both electrons and holes contributing to the conductivity of
the material. At low electric fields the drift velocity vd of the carriers is proportional to
275
&ernlconductor Basics
the applied field strength through a proportionality constant known as mobility J.l.
Thus, for semiconductors we can rewrite (6.5) as
a =
(6.6)
qnJln + qpjlp
l where !ln, llp are the mobilities of electrons and holes, respectively. For intrinsic semiconductors we can simplify (6.6) further by recalling that n = p = ni, that is,
i
!i.
~
(6.7)
I
:~
--------------------------~~&)A~
Example 6-1: Computation of the temperature dependence of
the intrinsic semiconductor conductivity
It is desired to find the conductivities for the intrinsic materials of
Si, Ge, and GaAs as a function of temperature. To make the computations not too difficult, we assume that the bandgap energy and the
mobilities for ho1es and electrons are temperature independent over
the range of interest -50C $ T ~ 200C.
,.1
(
Solution:
As a first step it is convenient to combine into one

parameter o 0 (T) all factors without the exponential term in (6.7);
that is,
a 0 (T) = q J N eN v<lln + Jlp)

where electron and hole mobilities are found from Table E- 1:
Jln
= 1350(Si), 3900(Ge), 8500(GaAs)
Jlp = 480(Si), 1900(Ge), 400(GaAs)

2
All values are given in units of em I (V s). N c, N v are computed according to (6.2) as
N c, y(T) : N c,
y(300oK)(3~r/2
This leads to the form
cr : cr0 (T) exp h:t) : q(~. + ~P) jN eN v(3~r\xph:t)
276
where the bandgap energy W 8 :: W c- W v is, respectively, 1.12 e V

(Si), 0.62 eV (Ge), and 1.42 eV (GaAs). The three conductivities are
plotted in Figure 6-2.
102
100
IB
10-2
10-4
0
0
10-{)
10-8
>
.:
"'d
- 14....__ _..___ _....___
10
- 50
50
_.____ __,____ ____.__
100
150
200
___.
250
Temperature, o C
Figure 6-2
Conductivity of Si, Ge, GaAs in the range from
-sooc to 250C.
The electric properties of semiconductors are strongly influenced by the ambient temperature. In this example we have
neglected the temperature dependence of the bandgap energy, which
is discussed in Chapter 7. Knowledge of the temperature behavior of
active devices is an important design consideration where internal
heating, due to power dissipation, can easily result in temperature
values exceeding 100-150C.
A major change in the electrical properties of a semiconductor can be initiated by

introducing impurity atoms. This process is called doping. To achieve n-type doping
(which supplies additional electrons to the conduction band) we introduce atoms with a
larger number of valence electrons than the atoms in the intrinsic semiconductor lattice
that they substitute. For instance, the implantation of phosphorous (P) atoms into Si introduces loosely bound electrons into the neutral crystal lattice, as shown in Figure 6-3(b).
2n
Conduction band
Conduction band
Conduction band
/ U . t / / / l / . l / / / / / / . l / 1 / 1 TU
rr ('
---------- w,.
l.tlljll.t.t.t.t/11/IVLLL Til'
rr c
1//1.1111111/111111//1
w,. - - - - - - - - - - .
----------------w,)
m
rr c
w,. ------------------~
---------.
Wv """"''"""'""Wv
"""""'""'""'
~' ""'""""'"""'
Valence band
Valence band
Valence band
(a} Intrinsic
(b) n-type
(c) p-type
Figure 6-3 Lattice structure and energy band model for (a) intrinsic, (b) n-type,
and (c) p-type semiconductors at no thermal energy. W0 and WA are donor and
acceptor energy levels.
It is intuitively apparent that the energy level of this "extra" electron is closer to
the conduction band than the energy of the remaining four valence electrons. When the
temperature is increased above absolute zero, the loosely bound electron separates from
the atom, forming a free negative charge and leaving behind the fixed positive ion of
phosphorous. Thus, while sti11 maintaining charge neutrality, the atom has donated an
electron to the conduction band without creating a hole in the valence band. This results
in an increase in the Fermi level since more electrons are located in the conduction
band. Contrary to the intrinsic semiconductor (ni, P i) we now have ann-type semiconductor in which the electron concentration is related to the hole concentration as
(6.8)
where N D is the donor concentration and Pn represents the minority hole concentration. To find nn and Pn we have to solve (6.8) in conjunction with (6.3). The result is
N + JN~ + 4ni
D
(6.9a)
278
Chapter 6 Active RF Componentl
(6.9b)
If the donor concentration N D is much greater than the intrinsic electron concentration
n;, then
(6.10a)
2
n.I
(6.10b)
Let us now consider adding impurity atoms with fewer valence electrons than the
atoms fonning the intrinsic semiconductor lattice. These types of elements are called
acceptors, and an example of such an element for the Si lattice is boron (B). As seen in
Figure 6-3(c), one of the covalent bonds appears to be empty. This empty bond introduces
additional energy states in the bandgap that are closely situated to the valence band.
Again, when the temperature is increased from absolute zero, some electrons gain extra
energy to occupy empty bonds but do not possess sufficient energy to cross the bandgap.
Thus, impurity atoms will accept additional electrons, forming negative net charges. At
the sites where the electtons are removed, holes will be created. These holes are free to
migrate and will contribute to the conduction current of the semiconductor. By doping the
semiconductor with acceptor atoms we have created a p-type semiconductor with
(6.11)
where N A, nP are the acceptor and minority electron concentrations. Solving (6.11)
together with (6.3), we find hole p P and electron n P concentrations in the p-type semiconductor:
(6.12a)
nP
-NA +
JN! +4n:
(6.12b)
Similar to (6.9), for high doping levels, when N A ni, we observe
(6.13a)
Pp=:.NA
n.2I
(6.13b)
Slmlconductor Basics
279
Minority and majority concentrations play key roles in establishing the current
fl.ow characteristics in the semiconductor materials.
6.1.2
PH-Junction
The physical contact of a p-type with an n-type semiconductor leads to one of the
most important concepts when dealing with active semiconductor devices: the pn-junction. Because of the difference in the carrier concentrations between the two types of
semiconductors a current flow will be initiated across the interface. Tills current is commonly known as a diffusion current and is composed of electrons and holes. To simplify our discussion we consider a one-dimensional model of the pn-junction as seen in
Figure 6-4.
Electric field
___"'
p-type
r---~~.._
Hole
n-type
r---~~---"'
IF
Electron
diffusion
diffusion
current
current
'-
/'---~
Space
charge
Space
charge
x =O
Figure 64 Current flow in the p~rjunction
The diffusion current is composed of In

I<liff
= In .
+ /p dff.
dtff
ditT
and IP
diff
components:
dn
dp)
= qA ( Dnd+
DPdX
X
(6.14)
where A is the semiconductor cross-sectional area orthogonal to the x-axis, and Dn ,

DP are the diffusion constants for electrons and holes in the fonn (Einstein relation)
kT
D n,p = J..ln, p -q = J..ln, p VT
(6.15)
The thermal potential V T = kT I q is approximately 26 mV at room temperature of

300K.
Since the p -type semiconductor was initially neutral, the diffusion current of holes
is going to leave behind a negative space charge. Similarly, the electron current flow
280
from the n-semiconductor will leave behind positive space charges. As the diffusion
current flow takes place, an electric field E is created between the net positive charge in
the n-semiconductor and the net negative charge in the p-semiconductor. This field in
turn induces a current IF = oAE which opposes the diffusion current such that
IF+ I diff = 0. Substituting (6.6) for the conductivity, we find
(6.16)
Since the total current is equal to zero, the electron portion of the current is also equal
to zero; that is,
In.d!ff +In F
dn + qnJ.tnAE = qJ.tnA ( V rd-dn n-d

dV) = 0
= qDnAdX
X
X
(6.17)
where the electric field E has been replaced by the derivative of the potential
E = -dVI dx. Integrating (6.17), we obtain the diffusion barrier voltage or, as it is
often called, the builtin potential:
C"
dV =
vdiff
vr>-
dn =
Vrtn(::)
(6.18)
where again nn is the electron concentration in then-type and nP is the electron concentration in the p-type semiconductor. The same diffusion barrier voltage could have
been found had we considered the hole current flow from the p to the n-semiconductor
and the corresponding balancing field-induced current flow I PF. The resulting equation
describing the barrier voltage is
V diff = V
rtn(~:)
(6.19)
If the concentration of acceptors in the p-semiconductor is N A>> n; and the concentration of donors in the n-semiconductor is N D n;, then n11 = N D, nP = n7 IN A, see
(6.13b), and by using (6.18) we obtain
(6.20)
Exactly the same result is obtained from (6.19) if we substitute p P = N A and
2
Pn = n;IND.
Stmlconductor Bastes
281
------------------------~&M~
Example 6-2: Determining the diffusion barrier or built-in
voltage of a pn-junction
For a particular Si pn-junction the doping concentrations are given
18
15
.h
. . .
to be N A = 10 em-:r and N ~ = 5 x 10 em- J Wit
an mtrms1c
concentration of n; = 1.5 x 10 em -J . Find the barrier voltages for
T == 300K.
Solution:
The barrier voltage is directly determined from (6.20):
V diff = V TIn
NAND) kT (NAN2 D)
( n 2. = -ln
q
n.
l
= 0. 796 (V)
We note that the built-in potential is strongly dependent on the doping concentrations and temperature.
For different semiconductor materials such as GaAs, Si, Ge,

the built-in voltage will be different even if the doping densities are
the same. This is due to significantly different intrinsic carrier concentrations.
If we want to determine the potential distribution along the x-axis, we can employ
Poisson's equation, which for a one-dimensional analysis is written as
2
_ dE
- -p(x)
----
d V(x) _
dx2
ro
dx
(6.21)
where p(x) is the charge density and er is the relative dielectric constant of the semiconductor. Assuming uniform doping and the abrupt junction approximation, as
shown in Figure 6-5(b), the charge density in each material is
p(x)
= -qNA
for -d P ~ x ~ 0
(6.22a)
(6.22b)
where dP and dn are the extents of the space charges in the p- and n-type semiconductors.
282
p-~e
_ _ _,
~--_......,..._.._
i-----
-------------------------
r ___n-~e
_.,..,._ _ _,
++++++
++++++
++++++
++++++
++++++
++++++
++++++:
., - - I '
j
sJ;Ce
sJ;ce
! charge
,:
-- /
charge
X
-dp
X=
d,
(a) pn-junction with space charge extent

n,p
pP =
(majority carrier)
n 11 =No (majority carrier)
-~~~~~~~~============~x
-dp
d,
(b) Acceptor and donor concentrations

p(x)
dp
d,
++++
++++
++++~
---qNA--L- ~===
--Figure 65
qND
(c) Polarity of charge density distribution

The pn-junction with abrupt charge carrier transition in the absence of
an externally applied voltage.
283
(d) Electric field distribution

V(x)
~iff
----------~~~~----+--------+x
-tfp
dn
(e) Barrier voltage distribution
Figure 65 The pn-junction with abrupt charge carrier transition in the absence of
an externally applied voltage. (Continued)
The electric field in the semiconductor is found by integrating (6.21) in the spatial
limits -dP ~ x ~ dn such that
(6.23)
qND
---(dn- x), for 0 S: x

Er 0
dn
The resulting electric field profile is depicted in Figure 6-5(d). In deriving (6.23)
we used the fact that the charge balance law demands that the total space charge in the
semiconductor equals zero, which for highly doped semiconductors is equivalent to the
condition
(6.24)
To obtain the voltage distribution profile we now carry out the integration of
(6.23) as follows:
Chapter e
284
V(x)
= -[- d
E(x)dx
Active RF Component~
(6.25)
Since the total voltage drop must be equal to the diffusion voltage V diff, it is
found that
(6.26)
Substituting dP = dnN D/ N A and solving (6.26) for dn, we obtain the extent of the
positive space charge domain into then-semiconductor:
(6.27)
where E = EoEr. An identical derivation involving dn
charge extent into the p-semiconductor:
= dPN AI N D
gives us the space
(6.28)
The entire length is then the addition of (6.27) and (6.28):
1/ 2
_
ds-
2 EV diff( 1
1)
-+- }
{
q
NA ND
(6.29)
We next tum our attention to the computation of the junction capacitance. This is
an important parameter for RF devices, since low capacitances imply rapid switching
speeds and suitability for high-frequency operations. The junction capacitance can be
found via the well-known one-dimensional capacitor formula
= EA
ds
Substituting (6.29) for the distance d 5 , we express the capacitance as
c-
A{
NAND
2 vdiffNA + N D
}1/2
(6.30)
285
If an external voltage VA is applied across the junction, two situations arise that
explain the rectifier action of the diode, as shown in Figure 6-6.
p
n
++ I I
+ + I I
++ I....I
+ +
++
I I
1+-1
n
+I I
+I I
+ 1+-1
+I
+I
I ,I....I1,_
I
Space charge distribution in the pn-junction

E
_.....
_.....
+-
'
+-
,,
''
Electric field distribution in the pn-junction
Voltage distribution in the pn-junction

(a) Reverse biasing (VA < 0)
(b) Forward biasing (VA > 0)
Figure 6-6 External voltage applied to the pn-junction in reverse and forward
directions.
The reverse polarity [Figure 6-6(a)] increases the space charge domain and prevents the flow of current except for a small leakage current involving the minority carrier concentration (holes in the n-semiconductor, and electrons in the p-semiconductor).
In contrast, the forward polarity reduces the space charge domain by injecting excessive
electrons into the n- and holes into the p-type semiconductor. To describe these situations, the previously given equations (6.27) and (6.28) have to be modified by replacing
the barrier voltage V diff with V diff - VA ; that is,
(6.31)
288
Chapter 6 Active RF Componenta
(6.32)
which leads to a total length of the space charge or depletion domain
1
_ {2(V diff- VA)( -+~

1
1 )}
dsNA
2
(6.33)
ND
Depending on the polarity of VA, we notice from (6.31)-(6.33) that either the space
charge domain is enlarged or diminished.
----------------------------~&JA~
Example 6-3: Computation of the junction capacitance and
the space charge region length of a pn-junction
For an abrupt pn-junction Si semiconductor at room temperature
10
3
(, = 11.9, n; = 1.5 x 10 em- ) with donor and acceptor con15
15
3
3
centrations equal to N D = 5 x 10 cm- and N A = 10 cm- ,
we desire to find the space charge regions d P and d n and the junction capacitance at zero biasing voltage. Show that the depletionlayer capacitance of a pn-junction can be cast into the form
c,
c,o(l- VA )-1/2
vdiff
and detennine CJO Sketch the depletion capacitance as a function

of applied voltage. Assume that the cross-sectional area of the pn.
. . A = 10-4 Cffi2 .
JUDCtlOn lS
Solution:
We return to the capacitance expression (6.30) where
we introduce the applied voltage VA . Thus,
c,
=A [
q
2 v diff( 1 - v A I
NAND
]112
v diff) N A + N D
which is immediately recognized as the preceding formula, if we set
Semiconductor Baslea
287
q NAND
C10 = A
----[ 2VdiffNA
+ND
112
Substituting V diff = V rln(NANDin;) = 0.6159 V, it IS found

that C10 = 10.68 pF.
For the space charge extents we use (6.28) and (6.29):
A(
2E V diffN
dn =
Nv NA+Nv
)}1/2 = 0.1643J.Lm
The dependence of the junction capacitance on the applied voltage

is depicted in Figure 6-7.
70
60
~
0..
c.}
50
....
.....
u
40
~ 30
.....0
0
20
10
0_5
-4
-3
-2
Applied voltage
-1
~.
Figure 6-7 The pn-junction capacitance as a function of applied voltage.
In Figure 6-7 the junction capacitance for applied voltages

near the built-in potential will approach infinity. However, in reality
the value begins to saturate, as further discussed in Chapter 7.
288
For the current flow through the diode we list the Shockley diode equation, which
is derived in Appendix F:
(6.34)
where I 0 is the reverse saturation or leakage current. The current-voltage characteristic, often called the 1-V curve, is generically depicted in Figure 6-8.
1110
7
-+
5
4
--~~--~--~~~~---r--,---r-+~ / ~-2.0 - 1.5 j.;2--0.5

1.0
1.5
2.0
I ; ............................. ... -1
Figure 68 Current-voltage behavior of prrjunction based on Shockley equation.
This curve reveals that for negative voltages a small, voltage independent, current
-I0 will flow, whereas for positive voltages an exponentially increasing current is
observed. The function shown in Figure 6-8 is an idealization since it does not take into
account breakdown phenomena. Nonetheless (6.34) reveals clearly the rectifier property of the pn-junction when an alternating voltage is applied.
The existence of the depletion layer or junction capacitance requires a reversebiased pn-junction diode. This implies, with reference to Example 6-3, the condition
that V A < V diff . However, under forward bias condition we encounter an additional dif
fusion capacitance due to the presence of diffusion charges Qd (minority carriers)
stored in the semiconductor layers which become dominant if VA > V diff. This charge
can be quantified by realizing that the charge Qd can be computed as diode current I
multiplied by the transition time of carriers through the diode 'tr or
Qd
= l'tr = 'trl0 ( e V AIV T -
1)
(6.35)
It is apparent that the diffusion capacitance assumes a nonlinear relation with the
applied voltage and the junction temperature. The diffusion capacitance is computed as
289
(6.36)
and is seen to be strongly dependent on the operating voltage.

In general, the total capacitance C of a pn-diode can roughly be divided into three
regiOns:
1. VA < 0 : only the depletion capacitance is significant: C = C 1
2. 0 < VA < Vruff: depletion and diffusion capacitances combine: C = C1 + C d
3. VA > V diff: only the diffusion capacitance is significant: C
Cd
The influence of the diffusion capacitance is appreciated if we consider a diode

that is operated at VA
1 V and that has an assumed transition time of
-10
.
-15
tr
100 ps
I 0 s and a reverse saturatton current of I 0
1 fA
10 A measured at room temperature of 300K (i.e., V T
26 m V ). Substituting these values
into (6.36), we find C
Cd
194 nF which is rather large and for typical resistances
of R 0.1 ... 1 11 results in large RC time constants that restrict the high-frequency
use of conventional pn-junction diodes.
6.1.3
Schottky Contact
W. Schottky analyzed the physical phenomena involved when a metallic electrode

is contacting a semiconductor. For instance, if a p-semiconductor is in contact with a
copper or aluminum electrode, there is a tendency for the electrons to diffuse into the
metal, leaving behind an increased concentration of holes in the semiconductor. The
consequences of this effect are modified valence and conduction band energy levels
near the interface. This can be displayed by a local change in the energy band structure
depicted in Figure 6-9(a).
I
~-----We
~
,,
...
~ - - - - - '""]~- - - - - - - WF
~
Metal
Wv
p-semiconductor
(a) Energy band model

(b) Voltage-current characteristic
Figure 6-9 Metal electrode in contact with p--semiconductor.
Chapter 6 Active RF Component.
Because of the higher concentration of holes, the valence band bends toward the
Fermi level. The conduction band, as the result of a lower electron concentration, bends
away from the Fermi level. For such a configuration we always obtain a low resistance
contact [(see Figure 6-9(b)], irrespective of the polarity of the applied voltage.
The situation becomes more complicated, but technologically much more interesting, when a metallic electrode is brought in contact with an n-semiconductor. Here
the more familiar behavior of a pn-junction emerges: A small positive volume charge
density is created in the semiconductor due to electron migration from the semiconductor to the metal. This mechanism is due to the fact that the Fermi level is higher in the
semiconductor (lower work function) than in the metal (higher work function) when the
two materials are apart. However, as both materials are contacted, the Fermi level again
has to be the same and band distortions are created. Electrons diffuse from the n-semiconductor and leave behind positive space charges. The depletion zone grows until the
electrostatic repulsion of the space charges prevents further electron diffusion. To clarify the issues associated with a metal n-semiconductor contact, Figure 6-10 shows the
two materials before and after bonding.
Free electron energy level
.. \ ..
........
qX
1--- ------ ---JJj..
w,
Metal
l n-semiconductor
Metal
X
(a) Metal and semiconductor do not interact

(b) Metal-semiconductor contact
Figure 6 ..10 Energy-band diagram of Schottky contact, (a) before and (b)
after contact.
The energy W b = q V b is related to the metal work function W M = q V M ( VM is

recorded from the Fermi level to the reference level where the electron becomes a
detached free particle; values of V M for some commonly used metals are summarized
in Table 6-2) and the electron affinity qX, where X is 4.05V for Si, 4.0V for Ge, and
291
4.07V for GaAs and is measured from the conduction band to the same reference level
where the electron becomes a free carrier, according to
(6.37)
Table 6-2
Work function potentials of some metals

Work function potential, V M
Material
Silver (Ag)
4.26V
Aluminum (Al)
4.28V
Gold (Au)
5. 1 v
Chromium (Cr)
4.5V
Molybdenum (Mo)
4.6V
Nickel (Ni)
5.15
Palladium (Pd)
5.12 v
Platinum (Pt)
5.65V
Titanium (Ti)
4.33 v
An expression for a built-in Schottky barrier voltage V d is established just as in

the pn-junction, which involves (6.37) and the additional voltage V c between conduction and Fermi levels:
(6.38)
where V c is dependent on the doping N D and the concentration of states in the conduction band N c according toN c = N 0 exp(V ciV r). Solving for the voltage gives
Vc = V rln(Ncl N 0 ). Although real metal-semiconductor interfaces usually involve
an additional very narrow isolation layer, we will neglect the influence of this layer and
only deal with the length of the space charge in the semiconductor:
1
A)}2
- {2(Vd - V
----q
Nv
ds-
Therefore, it is found that the junction capacitance of the Schottky contact
(6.39)
292
(6.40)
is almost identical to (6.30). A simple computation now can predict a typical value for
V d as illustrated in the following example.
-------------------------RF&MW4
Example 6-4: Computation of the barrier voltage, depletion
capacitance, and space charge region width for a
Schottky diode
A Schottky diode is created as an interface between a gold contact
material and an n- ty~ silicon semiconductor. The semiconductor is
6
doped to N D = 10 cm-3 and the work function V M for gold is
5.1 V. Also, as mentioned above, the affinity for Si is X = 4.05V.
Find the Schottky barrier V d , space charge width d 5 , and capacitance C1 if the dielectric constant of silicon is Er = 11.9. Assume
the cross-sectional diode area to be A = 10-4cm 2 and the temperature equal to 300 o K .
Solution:
Since the concentration of states in the conduction

19
band of silicon is N c = 2.8xl0 cm- 3 , we can compute the conduction band potential as
V c = V rln(N c) = 1.38xl0-23300ln(2.8xlOI9) V = 0.21 V
1.6xl0-19
]Q16
ND
Substituting the obtained value for V c into (6.38), we find the builtin barrier voltage
Vd = (VM-X)-Vc = (5.1 V-4.05 V)-0.21 V = 0.84V
The space charge width is obtained from (6.39)

- 12
2(8.85xl0
)_
11.90.84
_ _ m = 332 Jlm
~---_;.
1.6x10-I9
]Q16
Finally, the junction capacitance according to the formula for the

parallel-plate capacitor, see (6.40), gives us
AF Dfodea
293
This example shows that the metal-semiconductor junction

diode for similar size and doping has a depletion capacitance significantly smaller than that of a pn-junction, which permits higher
operational frequencies of the device.
6.2 RF Diodes
In this section we will review some practical realizations of the diodes that are
most commonly used in RF and MW circuits. As presented in the previous section, a
classical pn-junction diode is not very suitable for high-frequency applications because
of the high junction capacitance. In contrast, diodes formed by a metal-semiconductor
contact possess smaller junction capacitances and consequently reach higher frequency
limits. Today, Schottky diodes find widespread applications in RF detectors, mixers,
attenuators, oscillators, and amplifiers.
After discussing the Schottky diode in Section 6.2.1, we will continue investigating a number of special RF diodes. In Section 6.2.2 the PIN diode is analyzed and
placed in context with its primary use as a variable resistor and high-frequency switch.
Besides relying on the rectifier property of diodes, we can also exploit the dependence of the junction capacitance on the applied voltage to construct voltage-controlled
tuning circuits, where diodes are used as variable capacitors. An example of such a specialized diode is the varactor diode, covered in Section 6.2.3.
At the end of this section we will discuss a few more exotic diode configurations,
such as IM:PATT, Tunnel, TRAPAIT~ BARRITT, and Gunn diodes, which are less frequently used but which are still of interest due to their unique electric properties.
6.2.1
Schottky Diode
Compared with the conventional pn-junction, the Schottky barrier diode has a different reverse-saturation current mechanism, which is detennined by the thermionic
emission of the majority carriers across the potential barrier. This current is orders of
magnitude larger than the diffusion-driven minority carriers constituting the reversesaturation current of the ideal pn-junction diode. For instance, the Schottky diode has a
294
typical reverse-saturation current density on the order of 10- A/cm 2 compared with
11
10- A/cm 2 of a conventional Si-based pn-junction diode. The schematic diagram of a
cross-sectional view of the Schottky diode with the corresponding circuit elements is
given in Figure 6-11.
Metal contact
Depletion
region
~-----+
.
~-------~
n-typ~
epitaxial l-ayer
....................>-.
.
................... -....
, :.~
...... \,','
-.\~etal contact
Figure 6-11
\Metal contact
Cross-sectional view of Si Schottky diode.
The metal electrode (tungsten, aluminum, gold, etc.) is in contact with a weakly
doped n-semiconductor layer epitaxially grown on a highly doped n + substrate. The
dielectric is assumed to be ideal; that is, the conductance is zero. The current-voltage
characteristic is described by the following equation:
I = Is< e
( V -IR)
5
A
I)
(6.41)
where the reverse-saturation current is given by
Is =
A(R'T2 exp[-:~
b])
(6.42)
and R* is the so-called Richardson constant for thermionic emission of the majority
carrier across the potential barrier. A typical value of R* for Si is 100 A/cm 2 K 2 .
The corresponding small-signal equivalent circuit model is illustrated in Figure
6-12. In this circuit we note that the junction resistance R 1 is dependent on the bias current, just as is the diode series resistance, which is comprised of epitaxial and substrate
resistances R 5 = Repi + Rsub. The bond wire inductance is fixed and its value is
approximately on the order of Ls = 0.1 nH. As discussed above, the junction capacitance C 1 is given by (6.40). Because of the resistance R s, the actual junction voltage is
295
RF Diodes
equal to the applied voltage minus the voltage drop over the diode series resistance,
resulting in the modified exponential expression (6.41).
Figure 6-12
Circuit model of typical Schottky diode.
Typical component values for Schottky diodes are Rs = 2 . . . 5 Q ,

C8 = 0.1 ... 0.2 pF, and R1 = 200 ... 2 kQ. Often, the additional IRs tenn in
(6.41) is neglected for small bias currents below 0.1 rnA. However, for certain applications, the series resistance may form a feedback loop, which means the resistance is
multiplied by a gain factor of potentially large magnitude. For this situation, the IRs
tenn has to be taken into account.
In circuit realizations of high-frequency Schottky diodes, the planar configuration
in Figure 6-11 gives rise to relatively large parasitic capacitances for very small metal
contacts of typically 10 J..lm diameter and less. The stray capacitances can be somewhat
minimized through the addition of an isolation ring~ as depicted in Figure 6-13.
p-type
nng
n...-type substrate
\Metal contact
Figure 6-13
Schottky diode with additional isolation ring suitable for very-highfrequency applications.
The small signal junction capacitance and junction resistance can be found by
expanding the electric current expression (6.41) around the quiescent or operating
296
point V Q . That means the total diode voltage is written as a DC bias V Q and a small
AC signal carrier frequency component v d :
V = VQ+vd
(6.43)
The substitution of (6.43) in (6.41) for a negligible IRs tenn yields
= I 5 ( e VI V T -
1) = Is ( e
V Q/ V T v / V T
- 1)
(6.44)
Expanding this equation in a Taylor series about the Q-point and retaining the first two
terms gives
dl
I(V).::IQ+ dV
vd
fsvd
= IQ+ye
VQI Vr
vd
= IQ+(IQ+ls)vT
vd
== IQ+-
RJ
(6.45)
Here the junction resistance R j( V Q) is identified as
R 1 (V Q) = I
VT
I
Q+ s
(6.46)
and the junction capacitance is given by (6.40), with V Q replacing VA.
6.2.2
PIN Diode
PIN diodes find applications as high-frequency switches and variable resistors

(attenuators) in the range from 10 kQ to less than 1 Q for RF signals up to 50 GHz.
They contain an additional layer of an intrinsic (/-layer) or lightly doped semiconductor sandwiched between highly doped p + and n+ layers. Depending upon application
and frequency range, the thickness of the middle layer ranges from 1 to 100 J.lm. In
forward direction, the diode behaves as if it possesses a variable resistance controlled
by the applied current. However, in reverse direction the lightly doped inner layer creates space charges whose extent reaches the highly doped outer layers. This effect takes
place even for small reverse voltages and remains essentially constant up to high voltages, with the consequence that the diode behaves similar to a dual plate capacitor. For
instance, a Si-based PIN diode with an internal /-layer of 20 J1 m and a surface area of
200 by 200 J.l m has a diffusion capacitance on the order of 0.2 pF.
A generic PIN diode and its practical implementation in mesa processing tech
nology is presented in Figure 6-14. The advantage of the mesa configuration over the
conventional planar construction is a significant reduction in fringing capacitance.
The mathematical representation of the/- V characteristic depends on the level and
direction of current flow. To keep things simple, we will rely to a large extent on discussions already outlined for the pn-junction.
RFDiodes
297
r
I= n
I=n
n+-type substrate
(a) Simplified structure of a PIN diode (b) Fabrication in mesa processing technology
PIN diode construction.
Figure 6-14
In the forward direction and for a weakly doped n-type intrinsic layer the current
through the diode is
= A qn~WJ (e VA/ (2Vr ) Nvtp
(6.47)
1)
where W is the width of the intrinsic layer; t P is the excess minority carrier lifetime,
which can be on the order of up to t P = 1 J..LS ; and N D is the doping concentration in
the middle layer of the lightly doped n-semiconductor. The factor 2 in the exponent
takes into account the presence of two junctions. For a pure intrinsic layer N D = n i ,
(6.47) leads to the form
I = A(
W) (
q ni
'tp
VA I ( 2 V T)
- 1
(6.48)
The total charge can be calculated from the relation Q

the diffusion capacitance:
- dQ Cd - dV A
tP
= ItP . This allows us to find
( dl ) - l'tp
dV
- 2V
A
(6.49a)
In the reverse direction, the capacitance is dominated by the bias-dependent space

charge length of the /-layer. For small voltages C1 is approximately
298
c,
I(~)
(6.49b)
where 1 is the dielectric constant of the intrinsic layer.

The dynamic resistance of a PIN diode can be found through a Taylor series
expansion around the Q-point as already discussed for the Schottky diode. The result is
dV
Rj(VQ) = di I
=I Q
==
2Vr
I Q + I PO
(6.50)
where, with reference to (6.47), we have set I PO == A(qn;W)I(ND'tp) .

Based on the PIN diode's resistive behavior under forward bias ("switch on") and
capacitive behavior under reverse bias (Hswitch off" or isolation) we can proceed to construct simple small signal models. For the PIN diode in series connection, the electric circuit model is seen in Figure 6-15 terminated with source and load resistances. The
junction resistance and diffusion capacitance, as derived in (6.49) and (6.50), may in practice model the PIN diode behavior only very approximately. More quantitative infonnation is obtained through measurements or sophisticated computational modeling efforts.
------1
l-----(a) Forward bias
------,
L------
(b) Reverse bias (isolation)

Figure 615
PIN diode in series connection.
The bias point setting required to operate the PIN diode has to be provided
through a DC circuit that must be separated from the RF signal path. The DC isolation
is achieved by a radio frequency coil (RFC)~ representing a short circuit at DC and an
open circuit at high frequency. Conversely, blocking capacitors (CB) represent an
~-
t~--------------------------------------------------------------
299
RFDfodes
open circuit at DC and a short circuit at RF. Figure 6-16 shows a typical attenuator circuit where the PIN diode is used either in series or shunt connection.
Although in the following discussion we will use a DC bias, a low-frequency AC
bias can also be employed. In this case the current through the diode consists of two
components such that I = ( d QI dt) + QI 't P The implication of this is deferred to the
problem section.
DCbias
o-----,
Cs
1----_.,_____. . ,. ~~---------It---o RFout
RFin
PIN Diode
RFC
-(a) Series connection of PIN diode

DCbias
o-------.
PIN Diode
RFC
--
(b) Shunt connection of PIN diode

Figure 616
Attenuator circuit with biased PIN diode in series and shunt

configurations.
For positive DC bias voltage, the series connected PIN diode represents a low
resistance to the RF signal. The shunt connected PIN diode, however, creates a short-circuit condition, permitting only a negligibly small RF signal to appear at the output port.
300
The shunt connection acts like a high attenuation device with high insertion loss. The
situation is reversed for negative bias condition where the series connected PIN diode
behaves like a capacitor with high impedance or high insertion loss, whereas the shunt
connected diode with a high shunt impedance does not affect the RF signal appreciably.
An often used notation is the transducer loss TL conveniently expressed in tenns
of the S parameter 2 1 j so that with (4.52)
IS
TL
= -20log jS2 J!
2V2
(6.51)
= -20log _
VG
The following example computes the transducer loss for a PIN diode in series
configuration.
----------------------------~~~
Example 6-5: Computation of transducer loss of a PIN diode
in series configuration for forward and reverse
bias conditions
Find the transducer loss of a forward and reverse biased PIN diode in
series connection (Zc = ZL =Zo == 50 Q ). Assume the junction resistance R 1 under forward bias ranges between 1 and 20 Q . Furthermore, assume that the reverse bias operating conditions result in the
junction capacitance being C1 0.1, 0.3, 0.6, 1.3, and 2.5 pF, and
the frequency range of interest extends from 10 MHz to 50 GHz.
Solution:
Based on (6.51) and Figure 6-15, the transducer loss is

found with the aid of the voltage divider rule to be
1
TLforward
R )
= - 20log ( 100100
+ RJ) = 20log ( 1 + 100
and
TLreverse
= -20log
~OO
100- J1/(roCp)
= 10log[1 + ( 100 l
ro
C )
2
]
Figure 6-17 plots the transducer loss in dB under forward bias

condition for the given range of junction resistances. In contrast,
Figure 6- 18 graphs the reverse bias condition where the PIN diode
essentially has a purely capacitive response.
RF Dlode1
301
1.4
=g
ii0
-gt
1.2
0.8
e o.6
f-4
0.4
Zt =Zo
I0
12
14
16
18
20
JW1ction resistance R, , .Q
Figure 6-17 Transducer loss of series connected PIN diode under forward bias
condition. The diode behaves as a resistor.
70
60
so
v)
r.l.)
40
....
~
.
~
30
20
10
0
lOMHz
100MHz
l GHz
10 GHz
100 GHz
Frequency
Figure 6-18 Transducer loss of series connected PIN diode under reverse bias
condition. The diode behaves as a capacitor.
Chapter 6 Active RF Com~ ;
302
6.2.3
Varactor Diode
The PIN diode with its capacitive behavior under reverse bias already suggests !
that a variable capacitance versus voltage characteristic can be created by a specific :
middle layer doping profile. A varactor diode exactly accomplishes this task by a suit- !
able choice of the intrinsic layer thickness Win addition to selecting a particular doping :
distribution N v(x).
-------------R.F&.Mw-+ .
Example 6-6: Determination of the required doping profile for
a particular capacitance-voltage behavior
Find the appropriate doping concentration profile N v(x) that
ensures that the varactor diode capacitance changes depending on
the applied reverse biasing voltage as C(VA) = C 0 '1(VA- Vdiff),
12
where C 0 ' = 5 x 10- FV and the cross-sectional diode area
2
is A = 10-4 em .
Solution:
The extent of the space charge length can be predicted
based on (6.39) to be
x =
{2E/(V~f- VA)(~J r/2
which determines the junction capacitance C = e1AIx. In the derivation of the preceding formula we assumed that the doping concentration in the /-layer is much lower than the doping in the adjacent
layers. If the space charge domain is increased by a small increment
the charge is modified to
ax,
iJQ = qNv(x)Aax
This differential increase in length can be expressed by a corresponding decrease in capacitance. By differentiating the capacitor
formula, we obtain
iJx
= -1AoCIC2
ax
Upon substitution of
into the expression for dQ and noting that
iJQ = CiJV A, we have
303
AFDiodes
aQ =cavA
= -qND(x)A 2e/i)C!C2
This gives us the desired expression for the doping profile:

Nv(x)
c3 (avA)
qe!A2
oC
For the desired capacitance, we find
Co'
2xl0 11
-2
ND(x) = - =
em
qAx
x
Naturally we cannot enforce the doping profile to reach infinity
as x approaches the beginning of the /-layer. Nonetheless. by
approximating a hyperbolic function, it is possible to ensure the
desired capacitance-voltage behavior.
Figure 6-19 presents the simplified electric circuit model of the varactor diode
consisting of a substrate resistance and voltage-dependent capacitance of the form
112
. This is the case when the doping profile is constant. Therefore we
(Vdiff- V Af
have for the capacitance in generic representation:
Cv
Q )-112
= Cvo (1 -v-
(6.52)
V diff
.where V Q is the reverse bias.

;
One of the main applications of this diode is the frequency tuning of microwave
{circuits. This is due to the fact that the cut-off frequency fv of the first-order varactor
~model
'
},
::,
t.:
,.
....
(6.53)
l:
~be controlled through the reverse bias V Q .
~:
In addition, the varactor diode can be used to generate short pulses as schematically
l plained in Figure 6-20. An applied voltage VA across a series connection of resistor
ddiode creates a current flow I v . This current is in phase with the voltage over the pos~ 've cycle.
During the negative voltage cycle the stored carriers in the middle layer con. 'bute to the continued current flow until all carriers are removed. At this point the current
304
Chapter 6 Active RF Componenla
0.20
0.18
c...
\.) 0.16
G,)
<.J
0.14
<.J
0.12
.~
0.10
M)
= 0.2 pF,
~iff=
o.sv
0.08 .___.__ _.__....__.____..__ _.__....____.......____.._______.

-2.0 -1.8 -L6 -1.4 -1.2 -LO -0.8 -0.6 -0.4 -0.2
0
Biasing voltage J-Q , V
Figure 619
Simplified electric circuit model and capacitance behavior of

varactor diode.
drops abruptly to zero. A transformer can now couple out a voltage pulse according to
Faraday's law V out = L(dl vi dt). The pulse width can be approximated based on the
length of the middle layer Wand the saturation drift velocity vdmax of the injected carrier
concentration.
Varactor
~------------------------~~~----_.1
1... lv ~/
Figure 6-20
Pulse generation with a varactor diode.
M=DiodH
305
If we assume W = 10 J..Lm and vdmax = 10 crn/s we obtain a transit time that is

equivalent to a pulse width of
tv
6.2.4
= -wvdmax
10 Jlm
4
- 1 ns
(6.54)
10 rnls
IMPATT Diode
IMPATT stands for IMPact Avalanche and Transit Time diode and exploits the
avalanche effect as originally proposed by Read. The principle of this diode construction, which is very similar to the PIN diode, is depicted in Figure 6-21. The key difference is the high electric field strength that is generated at the interface between the n +
and p layer resulting in an avalanche of carriers through impact ionization.
0-+
07----.
Hole
Impact
Electron
+-------~------~
X
(a) Layer structure and electric field profile

(b) Impact ionization
Figure 621 IMPATI diode behavior.
The additional ionization current /ion that is generated when the applied RF voltage VA produces an electric field that exceeds the critical threshold level is seen in Figure 6-22. The current slowly decreases during the negative voltage cycle as the excess
carriers are removed. The phase shift between this ionization current and the applied
voltage can be tailored so as to reach 90 . The total diode current suffers an additional
delay since the excess carriers have to travel through the intrinsic layer to the p + layer.
1be time constant is dependent on the length and drift velocity as given in (6.47).
Choosing the intrinsic layer length appropriately in conjunction with a suitable doping
concentration can create an additional time delay of 90.
306
Chapter 6 Actlw RF ComponenCI
Figure 6-22
Applied voltage, ionization current, and total current of an

IMPATT diode.
The electric circuit diagram of an IMPATT device shown in Figure 6-23 is more :
intricate than the PIN diode and the reactance reveals an inductive behavior below the )
diode's resonance frequency f 0 before turning capacitive above the resonance fre- ;
quency. The total resistance is positive for f < f 0 and becomes negative for f > f 0 . .
Figure 6-23 Electric circuit representation for the IMPATT diode.
The resonance frequency is determined based on the operating current I Q , dielec- :

tric constant, saturation drift velocity v dmax, and the differential change in the ioniza- .
RFDiodes
307
tion coefficient a. with respect to the differential change in electric field strength
a' = iJa.l iJE . The resonance frequency is predicted as
f0
= _1
21t
21 vdmaxa'
(6.55)
The additional circuit parameters are specified as follows

R
= RL + _
_ _ _ ~_'d_rn_a_x_ _ __
(6.56a)
2tt 2 f~ CL w[1- (f I f 0 ) 2 ]
(6.56b)
cion
(6.56c)
= d
(21tj o)
(6.56d)
c ion
where RL is the combined resistance of the semiconductor layers, d is the length of the
avalanche region of the p-layer, and W is the total length, as shown in Figure 6-19. The
negative resistance of this diode above the resonance frequency can be understood in
terms of returning electric energy to the RF or MW resonance circuit; which means the
diode operates as an active device. Thus, the circuit attenuation can be substantially
. reduced to the point where additional power is transferred to the load impedance.
Unfortunately, the 180-degree phase shift comes with a price: The efficiency of converting DC to RF power at operating frequencies of 5 to 10 GHz is very low, with typical values in the range of 10 to 15%.
6.2.5
Tunnel Diode
Tunnel diodes are pn-junction diodes that are made of n and p layers with
.: extremely high doping (concentrations approach 10 19-1020 cm- 3) that create very nar. row space charge zones. This can be seen immediately from equations (6.27) and
: (6.28). The result is that the electrons and holes exceed the effective state concentra tions in the conduction and valence bands. The Penni level is shifted into the conduction band ~V en of the n + layer and into the valence band W Vp of the p +
semiconductor. We notice from Figure 6-24 that the permissible electron states in either
semiconductor layer are only separated through a very narrow potential barrier.
308
Chapter 6 Active RF Componenll
- C)
~-----+~~------~----U'vp
Wr-p
+<>
Figure 624
Tunnel diode and its band energy representation.
Based on quantum mechanical considerations, there is a finite probability that

electrons can be exchanged across the narrow gap rather than having to overcome the
potential barrier through an externally supplied voltage. This phenomenon is known as
tunneling. In thermal equilibrium the electron tunneling from the n to p layer is balanced by the opposite tunneling from the p to n layer. No net current flow results.
The peculiar current-voltage response of the tunnel diode is best explained with
reference to the corresponding energy band deformation for four distinct situations, as
shown in Figure 6-25(b}-(e).
I
~pn-junction
:7
...........
diode current
--------~----~----~---------+~
v~ak
J-;i;rr
(a) I-V curve of tunnel diode. At high positive biasing voltages the corresponding current
of the tunnel diode approaches the current of the conventional pn-junction diode.
Figure 6-25 Current-voltage behavior of the tunnel diode and comparison with
energy band structure.
RFDiodes
309
- - - - - Wcp
electrons
(b) Negative current flow for VA< 0
(c) No current flow tor VA= 0
. - - - - - Wep
1
Tunneling
~~~.1
--
W.-,
trtn:~f::r~-:--:~-- wFp
Excess of
free states
Wv, - - - "
(d) Positive tunneling current, 0 < VA<
Figure 6-25
Vdift
(e) Positive current flow for VA> Vditt
Current-voltage behavior of the tunnel diode and comparison with

energy band structure. (Continued)
Unlike the equilibrium condition shown in Figure 6-24 and Figure 6-25(c), for a
.tegative applied voltage VA a higher concentration of electron states is created in the pJayer, which results in a higher probability to tunnel into the n-layer than vice versa.
The consequence is that even for small negative voltages, a steep increase in current can
~observed [Figure 6-25(b)]. For a small positive voltage the reservoir of free electrons
~shifted to the n-semiconductor and an increase in free electron states is created in the
tsemiconductor. The consequence is a positive current flow [Figure 6-25(d)] in
psponse to the tunneling of electrons from the n to the p layer. However, if the applied
;,ttage reaches a critical value VA = V diff no overlapping band structures occur [i.e.,
/;e condition Wen< W Vp responsible for the tunneling effect no longer exists, see Fig~ 6-25(e)]. The current flow through the tunnel diode approaches a minimum. Above
\'
Chapter 6 Aetlve RF Components
310
this critical voltage point V diff the diode behaves again like a conventional pn-junction
diode and current increases exponentially.
The electric circuit of the tunnel diode, Figure 6-26, is very similar to the IMPAIT
diode shown in Figure 6-23. Here Rs and L5 are resistance of the semiconductor layer
and associated lead inductance. The junction capacitance C T is in shunt with a negative
conductance -g = dl I dV, which is utilized in the negative slope of the /- V curve
shown in Figure 6-25(a).
-g
Figure 6-26
Electric circuit representation of a tunnel diode.
A simplified amplifier circuit involving a tunnel diode is depicted in Figure 6-27.

If we consider the power amplification factor Gr as the ratio of the power delivered to
2
the load R L to the maximally available power from the source P s = IVc/ I ( 8R0 ) , we
obtain at resonance
(6.57)
where the influence of R5 is neglected. If g is chosen appropriately (i.e.,
g = 1I RL + 1I RG ), the denominator approaches zero and we have the behavior of an
oscillator.
-g
Figure 6-27
Tunnel diode circuit for amplification/oscillation behavior.
Diode
6.2.6
311
TRAPATT, BARRITT, and Gunn Diodes
For completeness we briefly mention these additional three diode types without
~aoing into any detail of their circuit representation and quantitative electric parameter
;derivations.
j , The TRApped Plasma Avalanche Triggered Transit (TRAPAIT) diode can be
]~nsidered an enhancement of the IMPATT diode in that a higher efficiency (up to
f75%) is realized through the use of bandgap traps. Such traps are energy levels that are
f,situated inside the bandgap and allow the capture of electrons. External circuits ensure
Jabat during the positive cycle a high barrier voltage is generated, resulting in carrier
~multiplication of the electron-hole plasma. The consequence is a breakdown in the rect1ifier properties of the diode during the negative cycle. The operating frequency is
tslightly lower than the IMPATT diode. This is due to the fact that the buildup of the
fclectron-hole plasma is slower than the transit time through the middle layer in an
f IMPAIT diode.
f
For the BARRier Injection Transit Time (BARRIT) we are essentially dealing
with a transit time diode whose p +n p + doping profile acts like a transistor without base
1.:
~:COntact. The space charge domain extends from the cathode through the middle layer
tinto the anode. The small-signal circuit model consists of a resistor and shunt capacitor
~;yfflose values are dependent on the DC current bias. Unlike the IMPAIT diode, this RC
t..
tdrcuit can create a negative phase of up to -90 degrees at a relatively low efficiency of
i
~~% and less. The BARRIT diode finds applications in RADAR mixer and detector
tdrcuits.
t;w The Gunn diode is named after its inventor J. B. Gunn, who found in 1963 that in
~n semiconductors (GaAs, InP) a sufficiently high electric field can cause electrons
~ scatter into regions where the bandgap separation increases. As a result of this
.;
in bandgap energy, the electrons suffer a loss in mobility J.ln . This phenome:<:' n is so dramatic that, for instance ~n GaAs, the drift velocity ( vd = nqJ.ln) can drop
7
~ m 2 x 10 crnls to less than 10 crnls for electric field strengths growing from
.. kV/cm to 7 kV/cm. The negative differential mobility
I
I.
dvd
lln = dE< 0
again used for oscillator circuits as we will see in later chapters. To exploit the Gunn
,.. ect for RF and MW applications, a special doping profile is needed to ensure that
:.\;
~: .
the voltage exceeds the required threshold a stable single-carrier space domain is
ted.
Chapter 6 Acflve RF Componentl
312
6.3 Bipolar..Junction Transistor

The transistor was invented in 1948 by Bardeen and Brattain at the former AT&T
Bell Laboratories and has over the past 50 years received a long lists of improvements
and refinements. Initially developed as a point-contact, single device, the transistor has
proliferated into a wide host of sophisticated types ranging from the still popular bipo
lar junction transistors (BJTs) over the modern GaAs field effect transistors (GaAs
FETs) to the most recent high electron mobility transistors (HEMTs). Although transistors are often arranged in the millions in integrated circuits (ICs) as part of microprocessor, memory, and peripheral chips, in RF and MW applications the single transistor
has retained its importance. Many RF circuits still rely on discrete transistors in lownoise, linear, and high-power configurations. It is for this reason that we need to investigate both the DC and RF behavior of the transistors in some detail.
The constituents of a bipolar transistor are three alternatively doped semiconductors, in npn or pnp configuration. As the word bipolar implies, the internal current flow
is due to both minority and majority carriers. In the following we recapitulate some of
the salient characteristics.
6.3.1
Construction
The BIT is one of the most widely used active RF elements due to its low-cost
construction, relatively high operating frequency, low-noise performance, and highpower handling capacity. The high-power capacity is achieved through a special interdigital emitter-base construction as part of a planar structure. Figure 6-28 shows both
the cross-sectional planar construction and the top view of an interdigital emitter-base
connection.
Because of the interleaved construction shown in Figure 6-28(b) the base-emitter
resistance is kept at a minimum while not compromising the gain performance. As we
will see, a low base resistance directly improves the signal-to-noise ratio by reducing
the current density through the base-emitter junction (shot noise) and by reducing the
random thermal motion in the base (thennal noise), see Chapter 7 for more details.
For frequency applications exceeding 1 GHz it is important to reduce the emitter
width to typically less than I Jliil size while increasing the doping to levels of
1020 1021 cm- 3 to both reduce base resistance and increase current gain. Unfortunately, it becomes extremely difficult to ensure the tight tolerances, and self-aligning
processes are required. Furthermore, the acceptor and donor doping concentrations
reach quickly the solubility limits of the Si or GaAs semiconductor materials, providing
a physical limitation of the achievable current gain. For these reasons, heterojunction
bipolar transistors (HBTs) are becoming increasingly popular. HBTs achieve high
alpolar-Junctlon Transl$tor
313
c
(a) Cross-sectional view of a multifinger bipolar junction transistor
Base bonding pad
p base well
Emitter bonding pad

(b) Top view of a multifinger bipolar junction transistor
Figure 6-28 Interdigitated structure of high-frequency BJT.
current gains without having to dope the emitter excessively. Due to additional semiconductor layers (for instance, GaAlAs-GaAs sandwich structures) an enhanced electron injection into the base is achieved while the reverse hole injection into the emitter
is suppressed. The result is an extremely high emitter efficiency as defined by the ratio
of electron current into the base to the sum of the same electron current and reverse
emitter hole current. Figure 6-29 shows a cross-sectional view of such a structure.
Besides GaAs, heterojunctions have been accomplished with InP emitter and
InGaAs base interfaces; even additional heterojunction interfaces between the GainAs
base and InP collector (double heterojunctions) have been fabricated. The material InP
has the advantage of high breakdown voltage, larger bandgap, and higher thermal conductivity compared to GaAs. Operational frequencies exceeding 100 GHz, and a carrier
314
Figure 629 Cross-sectional view of a GaAs heterojunction bipolar transistor

involving a GaAIAs-GaAs interface.
transition time between base and collector of less than 0.5 ps have been achieved.
Unfortunately, lnP is a difficult material to handle and the manufacturing process has
not yet matured to a level that allows it to compete with the Si and GaAs technologies.
6.3.2
Functionality
In general, there are two types of BITs: npn and pnp transistors. The difference
between these two types lies in the doping of the semiconductor used to produce base,
emitter, and collector. For an npn-transistor, collector and emitter are made of n-type
semiconductor, while the base is of p-type. For a pnp-transistor, the semiconductor
types are reversed (n-type for base, and p-type for emitter and collector). Usually, the
emitter has the highest and the base has the lowest concentration of doping atoms. The
BIT is a current-controlled device that is best explained by referring to Figure 6-30,
which shows the structure, electrical symbol, and diode model with associated voltage
and current convention for the npn-structure. We omit the discussion of the pnptransistor since it requires only a reversal of voltage polarity and diode directions.
The first letter in the voltage designation always denotes the positive and the second letter gives the negative voltage reference points. Under normal mode of operation
(i.e., the forward active mode), the emitter-base diode is operated in forward direction
(with V BE= 0.7 V) and the base-collector diode in reverse. Thus the emitter injects
electrons into the base, and conversely from the base a hole current reaches the emitter.
If we maintain the collector emitter voltage to be larger than the so-called saturation
voltage (typically around 0.1 V), and since the base is a very thin (on the order of
d 8 < 1 J.im ) and lightly doped p-type layer, only a small amount of electrons recombine with the holes supplied through the base current. The vast majority of electrons
reach the base-collector junction and are collected by the applied reverse voltage V BC.
For the reverse active mode, the collector-emitter voltage is negative (typically
V CE < -0.1 V) and the base-collector diode is forward biased, while the base-emitter
Blpolar..JunctJon Translstor
315
c
Electron
recombination
Hole
recombination
~ collector
n-~e
} p-~e
Hole
injection
in-~e
base
Is
Vcr:
-----+
Vc
emitter
E
(a)
(b)
(c)
Figure 6-30 npntransistor: (a) structure with electrical charge flow under forward
active mode of operation, (b) transistor symbol with voltage and current directions,
and (c) diode model.
diode is now operated in reverse direction. Unlike the forward active mode, it is now
the electron flow from the collector that bridges the base and reaches the emitter.
Finally, the saturation mode involves the forward biasing of both the base-emitter and base-collector junctions. This mode typically plays an important role when
dealing with switching circuits.
For a common emitter configuration, Figure 6-31 (a) depicts a generic biasing
arrangement, where the base current is fixed through an appropriate choice of biasing
resistor RB and voltage source V BB, resulting in a suitable Q-point. The base current
versus base-emitter voltage, Figure 6-31 (b), follows a typical diode 1-V behavior, which
C9Dstitutes the input characteristic of the transistor. The base current and base-emitter
voltage at the intersection point between the load line and the transistor input character. istic are identified as I~ and V~E. The collector current versus collector-emitter voltage
~havior as part of the transistor output characteristic follows a more complicated pat~Jem since the collector current must .be ~eated as a parametric curve dependent on the
'r,se current(/BI <I82 .. ) as seen m Ftgure 6-3l(c).
I The quantitative BIT behavior is analyzed by investigating the three modes of
f,peration in tenns of setting appropriate operating points and fonnulating the various
f.urent :flows. For simplicity, we will neglect the spatial extent of the individual space
e domains and assume typical representative voltage and current conditions. To
track of the different minority/majority and doping conditions in the three semi, uctor layers, Table 6-3 summarizes the parameters and corresponding notation.
E
,.
316
+
-==-~c
(a) Biasing circuit for npn BJT in common-emitter configuration
(b) Input characteristic of transistor

Figure 6-31
(c) Output characteristic of transistor
Biasing and input, output characteristics of an npn BJT.
Table 63
Parameter description
Doping level
BJT parameter nomenclature
Emitter (n-type)
NE
D
E
Base (p-type)
Ng
NB
Minority carrier concentration

in thermal equilibrium
p n0 = n 1./ND
Majority carrier concentration

in thermal equilibrium
Spatial extent
Collector (n-type)
nPo = n;INA
Pn0 = ni / ND
nn0
B
Ppo
nn0
dE
dB
de
BlpolaNunctlon Transistor
317
For the following BIT analysis, it is implicitly understood that the concentrations
obey the inequality P!o n:o P~o.
Forward Active Mode (V CE > V CEsat = 0.1 V, I 8 > 0)
To find the minority charge concentrations, we consider the configuration shown
in Figure 6-32. Here the concentration is plotted as a function of distance across the
three semiconductor layers. For predicting the spatial minority carrier concentrations in
the respective layer, we rely on the so-called short diode (see Appendix F) analysis,
which approximates the exponentials as linear charge concentration gradients.
--~--------~--------+---------~---+X
X =
-dE
Figure 6-32
=0
= dB
Minority carrier concentrations in forward active BJT.
The minority charge concentrations in each layer are given as follows:

Emitter:
Pn (-dE) = Pno and Pn (0) = Pnoe
VsEIVr
Base:
The last two concentrations are zero because the base-collector voltage is negative (for
instance, for typical transistor values of V CE = 2.5 V and V BE = 0.7 V we find
VBc = -1.8 V, which yields exp[V8 c1 Vr] = exp[-1.8/ 0.026 ] ~ 0). Based on
the aforementioned carrier concentrations we can now predict the diffusion current density of holes J!diff in the emitter:
E
E
E E
E
EdPn (x)
qDP E
E
qD pPno V sEIV r
d
(e
- 1) (6.58)
J diff = -qD
= --d
[pn(O)- Pn(-dE) ] =
P
P
dx
E
E
Chapter 6 Active RF Componentt
318
For the diffusion current density of electrons in the base layer J!diff we similarly obtain
B
_
Jndiff -
qDn
[dn!(x)]
_ qD!
dx
- dB [np(dB)- np(O)]
B
qD!n!o
dB e
V 8 E1Vr
(6.59)
From the preceding two equations, the collector and base currents can be established as
(6.60)
and
(6.61)
where index F denotes forward current, A is the junction cross-sectional area, and
= (q n
I d 8 is the saturation current. The emitter current is directly found
by adding (6.60) and (6.61). The forward current gain ~F under constant collector
emitter voltage is defined as
Is
D! !oA)
(6.62)
To arrive at (6.62) it is assumed that the exponential function in (6.61) is much larger
than 1, allowing us to neglect the factor -1. Moreover, the ratio between collector and
emitter currents, or a F , is expressed as
(6.63)
------------------------------~&uM~
Example 6-7: Computation of the maximum forward current
gain in a bipolar-junction transistor
Find the maximum forward current gain for a silicon-based BJT
with the following parameters: donor concentration in the emitter,
E
19
-3
N D = 10 em ; acceptor concentration m the base,
B
17
-3
.
N A = 10 em ; space charge extent m the elllitter,
dE = 0.8 J.lm ; and space charge extent in the base, d 8 = 1.2 ~-trn .
Bipolar-Junction Transistor
319
Solution:
To apply (6.62), we need to detennine the diffusion
constants in base and emitter as described by the Einstein relation
(6.1 5). Substituting this relation into (6.62), we obtain the forward
current gain:
B
~F
Jl.nn pOdE
E
J.l pP,odB
Furthermore, using the expressions for the minority carrier concentrations in base and emitter from Table 6-3, we arrive at the final
expression for ~ F :
~F-
- 187.5
As discussed in Section 6.3.3 and in the following chapter; the

current gain is only approximately constant. In general, it depends
on the transistor operating conditions and temperature behavior.
Reverse Active Mode ( V cE < -0.1 V, I 8 > 0)

The minority carrier concentrations are shown in Figure 6-33 with the associated
space charge domains (i.e., the base-emitter diode is reversed biased whereas the basecollector diode is forward biased).
Reverse biased
junction~
Forward biased
~junction
t n: (ds)
p;(dB)
E
F.
/p"=
0
X
--~--------~--------~----------~--~
x=O
Figure 6-33
Reverse active mode of BJT.
320
The minority charge concentrations in each layer are as follows:

E
E VBEIVr
= Pnoe
:::::0
B
B YsEIVr
B
B YsciVr
Base:
nP(O) = nP 0 e
=0 and nP(d8 ) = nP0 e
C
C VsciVT
C
C
Collector: Pn (d 8 ) = Pnoe
and Pn (dB+ de)::::: Pno
Pn (-dE)::::: 0 and Pn (0)
Emitter:
From the diffusion current density, we can find the reverse emitter current
dnpBJ
= -JndiffA = -qDn ( dx A
B
IRE
B B YsciVr
qDnnpo
= dB Ae
= lse
YsciVr
(6.64)
and the reverse base current
IRB
C
-JpdiffA
c)
c c
dpn
qDpPnoA(
YsciVr
)
=-qDP dx A=
de
e
-1
C
(6.65)
In a similar manner as done for the forward current gain, we define the reverse current
gain ~R
(6.66)
and the collector emitter ratio
aR
IRe
(6.67)
Saturation Mode ( V BE' V Be> V p I B > 0)

This mode of operation implies the forward bias of both diodes, so that the diffusion current density in the base is the combination of forward and reverse carrier flows;
that is, with (6.60) and (6.64):
(6.68)
From (6.68) it is possible to find the emitter current by taking into account the forward
base current. This forward base current (6.61) injects holes into the emitter and thus has
to be taken with a negative sign to comply with our positive emitter current direction
convention. Making the exponential expressions in (6.68) compatible with (6.61), we
add and subtract unity and finally obtain
Blpotar-Junetlon Transistor
321
8
7
5 ( V 8 EI V 7
IE = -Is ( e VsEIVr - 1) -1e
- 1) + Is ( eV ciV - 1)
~F
(6.69)
Because the BIT can be treated as a symmetric device, the collector current is expressible in a similar manner as the contribution of three currents: the forward collector and
reverse emitter currents, given by the negative of (6.68), and an additional hole diffusion contribution as the result of the reverse base current I RB. The resulting equation is
I c = Is ( e
V B1V 7
- 1 -
V 8ciV
pIs(
R e
- 1 -/s e
V 8 ci V 7
- 1
(6.70)
Finally, the base current I 8 = - I c - IE is found from the preceding two equations:
I8
_
-
{ 1 ( V 8 c iV7
1 ( VBEI Vr
)
I 5 ~R e
- 1 + ~F e
-
t)}
(6.71)
Here again, it is imponant to recall that the internal emitter current flow is denoted
opposite in sign to the customary external circuit convention.
6.3.3
Frequency Response
The transition frequency f T (also known as the cut-off frequency) of a microwave BJT is an important figure of merit since it determines the operating frequency at
which the common-emitter, short-circuit current gain h fe decreases to unity. The transition frequency f T is related to the transit time 't that is required for carriers to travel
through the emitter-collector structure:
fr
=!'t
(6.72)
This transition time is generally composed of three delays:

't
= 'tE+tB+'tc
(6.73)
where 'tE, -r8 , and 'tc are delays in emitter, base, and collector, respectively. The baseemitter depletion region charging time is given by
'tE
VT
VT
= rEC = T(CE + Cc) =T(CE + Cc)
(6.74a)
where C E, C c are emitter and collector capacitances, and r E is the emitter resistance
obtained by differentiation of the emitter current with respect to base-emitter voltage.
The second delay in (6.73) is the base layer charging time, and its contribution is given as
322
d2
'tB
(6.74b)
11Dn
where the factor 11 is doping profile dependent and ranges from 11 - 2 for uniformly
doped base layers up to 11 = 60 for highly nonuniform layers. Finally, the transition
time 'tc through the base-collector junction space charge zone we can be computed as
'tc
= we
-
(6.74c)
vs
with v s representing the saturation drift velocity. In the preceding formulas we have
neglected the collector charging time 'tee = rcCc, which is typically very small when
compared with 'tE
As seen in (6.74a), the emitter charging time is inversely proportional to the collector current, resulting in higher transition frequencies for increasing collector currents.
However, as the current reaches sufficiently high values, the concentration of charges
injected into the base becomes comparable with the doping level of the base, which
causes an increase of the effective base width and~ in tum, reduces the transition frequency. Usually, BIT data sheets provide information about the dependence of the transition frequency on the collector current. For instance, Figure 6-34 shows the transition
frequency as a function of collector current for the wide band npn-transistor BFG403W
measured at V CE = 2 V, f = 2 GHz, and at an ambient temperature of25C.
20
..
16
/
~
>.
cQ) 12
g.
./~
Q)
.............
....
fll
10
Collector current I c rnA
Figure 6-34 Transition frequency as a function of collector current for the 17 GHz
npn wideband transistor BFG403W (courtesy of Philips Semiconductors).
Bipolar-Junction Transfstor
323
Another aspect of the BJT operated at RF and MW frequencies is that at high frequencies the skin effect physically restricts current flow to the outer perimeter of the
emitter (see also Section 1.4). To keep the charging time as low as possible, the emitter
is constructed in a grid pattern of extremely narrow (less than 1 J.lm) strips. Unfortunately, the trade-off is a high current density over the small surface area, limiting the
power handling capabilities. Additional ways to increase the cut-off frequency are to
reduce the base transition time constant 't 8 by high doping levels and concomitantly
fabricate very short base layers of less than 100 nm. In addition, a small base thickness
has as an advantage a reduction in power loss.
6.3.4
Temperature Behavior
We have seen in this chapter that almost all parameters describing both the static
and dynamic behavior of semiconductor devices are influenced by the junction temperature Tj . As an example of such a dependence, in Figure 6-35 the forward current gain
~F for a given V CE is plotted as a function of collector current I c for various junction
temperatures 1) . As we can see from this graph, the current gain raises from 40 at
Ic =3.5 rnA and Tj = - 50C to more than 80 at Tj = sooc.
140
120
~
~
~
100
.E
C':S
80
..........
I..
I--
"'----
" ----
b()
r. 50(1-
I'--
1--
TJ
E 6o '-----..
-
='wooc
1j 150C
0C
T=
-soc"-'t""
J
1..
1-4
~ r----- -
40
20
0
Figure 6-35
3
4
Collector current l c, rnA
Current gain J3F = aFI( 1 - aF) as a function of collector current

for various junction temperatures at a fixed V CE.
324
Another example that shows the strong temperature influence is the dependence
of the input characteristic of a transistor described by the base current as a function of
base-emitter voltage, as depicted in Figure 6-36.
~--~~~~--~--~--~--~
0.5
1.0
Base-emitter voltage
Figure 6-36
2.0
~,
Typical base current as a function of base-emitter voltage for various

junction temperatures at a fixed VeE
Again, if we compare the behavior of the transistor at T1 = -50C and

T1 = 50C, we notice that at T1 = -50C and a base-emitter voltage of 1.25 V the
transistor is in cut-off state, whereas at T1 = 50C the BJT already conducts 4 rnA
base current. These two examples underscore the importance of temperature considerations in the design of RF circuits. For instance, the design of a cellular phone for
worldwide use must ensure that our circuit preforms according to specifications under
all temperature conditions encountered by the operator. Standard specifications usually
cover the temperature range from -50C to 80C.
The junction temperature also plays an important role when dealing with the maximum power dissipation. In general, the manufacturer provides a power derating
curve that specifies the temperature T s up to which the transistor can be operated at
the maximum available power Ptot . For junction temperatures T j exceeding this value,
the power has to be reduced to values dictated by the thermal resistance between the
junction and the soldering point (or case) Rthjs according to
(6.75)
Blpolar..Junc1lon Transistor
325
where T j max is the maximum junction temperature. Typical BJT values vary between
150 and 200C .
For the RF transistor BFG403W the maximum total power Ptot of 16 mW can be
maintained up to T s = 140 o C . For higher temperatures T s $ T j ~ T j max , the power
must be derated until the maximum junction temperature T jmax of 150C is reached.
The corresponding slope is 820K/W . This value implies that if the power dissipation
of the device decreases by 10 mW, the junction temperature can be increase by 8.2 C
up to the maximum junction temperature. Obviously, transistor cases with such a high
slope (or high thermal resistance) are not acceptable for high-power applications and
manufacturers have to develop effective ways to dissipate the thermal energy generated
by the transistor. Usually, this is done by employing heat sinks and using materials with
high thermal conductivity. Instead of the thermal resistance at the soldering point Rthjs,
the manufacturer may supply additional information involving heat resistances between
junction-to-case (Rlhjc), case-to-sink (Rthcs), and sink-to-air (Rihha) interfaces.
To simplify the thermal analysis it is convenient to resort to a thermal equivalent
circuit with the following correspondences:
Thermal power dissipation = electric current
Temperature
= electric voltage
A typical thermal circuit in equilibrium is shown in Figure 6-37, where the total electric
power supplied to the device is balanced through a thermal circuit involving thermal
resistances. In particular, we recognize the thermal resistance of junction to soldering
point which is assumed to be equal to Rthjc Therefore
Rthjc -
_Ti-Ts _
Rthjs -
Pw
- 'YthAarr
(6.76)
where junction and soldering point temperatures T1 and T s and thermal power P w
determine the thermal resistance in Kelvin per Watt e KJW), and whose value can also
be expressed in terms of the thermal conductivity Yth and the surface area ABJT of the
BJT. The solder point temperature is affected by the transition between casing and heat
sink. This constitutes a thermal resistance Rthcs with values up to 5 KIW. Finally, the
heat sink represents a thermal resistance of
Rthha
0hs
L
A
(6.77)
hs
2
where ()hs is a convection coefficient that can vary widely between 10 W / (Km ) for
2
2
still air, IOOW /(Km ) for forced air, up to lOOOW/ (Km ) for water cooling, and
Abs is the total area of the heat sink.
326
Chapter 6 AcHve RF Componentl
Power BIT
Figure 6-37 Thermal equivalent circuit of BJT.
The following example provides an often encountered design problem.
--------------------------~RF&JA~
Example 6..8: Thermal analysis involving a BJT mounted on a

heat sink
An RF power BIT generates a total power Pw of 15 W at case temperature of 25C. The maximum junction temperature is 150C
and the maximum ambient operating temperature is specified by the
user to be Ta = 60C. What is the maximum dissipated power if
the thermal resistances between case-to-sink and sink-to-air is
2K/W and l0K/W, respectively.
Solution:
With reference to Figure 6-37, we are dealing with

three thermal resistances: Rthjs, Rthcs, and Rthha. The junction-tosoldering resistance can be found based on equation (6.76):
Rth
= Tj- Ts = 15ooc- 25oc = 8 33oK/W
Js
Pw
15 W
Adding up all resistances gives us a total thermal resistance of

Rthtot
= Rthjs + Rthca + Rthhs
= 20.333 oK/W
The dissipated power Pth follows from the temperature drop (junction temperature T J minus ambient temperature T a ) divided by the
total thermal resistance:
Blpolar..Junctlon Transistor
327
150C- 60C
P th =
= 20 33 o K/W = 4 .43 W
thtot
To operate the BJT in thermal equilibrium, we have to reduce the

total electric power Ptot =Pw to the point where it is in balance with
the computed thermal power P 101 = Pth. Thus a reduction from
15 W to 4.43 W is required.
T - Ta
While the design engineer cannot influence the junction-to-soldering point heat resistance, it is the choice of casing and heat sink
that typically allows major improvements in thermal performance.
6.3.5
Limiting Values
The total power dissipation capabilities at a particular temperature restrict the

range of safe operation of the BIT. In our discussion we will exclusively focus on the
active mode in the common-emitter configuration and will neglect the switch-mode
behavior whereby the BIT is operated either in saturation or cut-off mode. For a given
maximum BIT power rating, we can either vary the collector-emitter voltage V CE and
plot the allowable collector current I c = Ptot/V C (here we assume that base current
is negligibly small compared to the collector current due to high ~)or vary I c and plot
the allowable collector-emitter voltage V C = Ptot/ I c. The result is the maximum
power hyperbola. This does not mean that I c and V CE can be increased without
bound. In fact, we need to ensure that I c ~I Cmax and V CE :S V CEmax, as depicted in
Figure 6-38. The safe operating area (SOAR) is defined as a set of biasing points
where the transistor can be operated without risk of unrecoverable damage to the
device. The SOAR domain, shown as a shaded region in Figure 6-38, is more restrictive
than a subset bounded by the maximum power hyperbola, since we have to take into
account two more breakdown mechanisms:
1. Breakdown of first kind. Here the collector current density exhibits a nonunifonn distribution that results in a local temperature increase, which in turn lowers
the resistance of a portion of the collector domain, creating a channel. The consequence is a further increase in current density through this channel until the positive feedback begins to destroy the crystal structure (avalanche breakdown),
ultimately destroying the transistor itself.
328
2. Breakdown of second kind. This breakdown mechanism can take place independently of the first mechanism and affects primarily power BJTs. Internal overheating may cause an abrupt increase in the collector current for constant V CE. This
breakdown mechanism usually occurs at the base-collector junction when the
temperature increases to such high values that the intrinsic concentration is equal
to the collector doping concentration. At this point the resistance of the junction is
abruptly reduced, resulting in a dramatic current increase and melting of the
junction.
Figure 6-38 Operating domain of BJT in active mode with breakdown

mechanisms.
It is interesting to point out that the BJT can exceed the SOAR, indeed even the
maximum power hyperbola, for a short time since the temperature response has a much
larger time constant (on the order of microseconds) in comparison with the electric time
constants.
Additional parameters of importance to a design engineer are the maximum voltage conditions for open emitter, base and collector conditions; that is, V CBO (collectorbase voltage, open emitter), V CEO (collector-emitter, open base), and V EBO (emitterbase voltage, open collector). For instance, values for the BFG403W are as follows:
V CBol
= 10 V, V CEol
= 4.5 V, and V EBOI
= 1.0 V.
max
max
max
6.4 RF Field Effect Transistors

Unlike BJTs, field effect transistors (FETs) are monopolar devices, meaning
that only one carrier type, either holes or electrons, contributes to the current flow
through the channel. If hole contributions are involved we speak of pchannel, other
wise of n-cbannel FETs. Moreover, the FET is a voltage-controlled device. A variable
RF Field Effect Transistors
329
electric field controls the current flow from source to drain by changing the applied
voltage on the gate electrode.
6.4.1
Construction
Traditionally PETs are classified according to how the gate is connected to the
conducting channel. Specifically, the following four types are used:
1. Metal Insulator Semiconductor FET (MISFET). Here the gate is separated

from the channel through an insulation layer. One of the most widely used types,
the Metal Oxide Semiconductor FET (MOSFET), belongs to this class.
2. Junction FET (JFET). This type relies on a reverse biased pn-junction that isolates the gate from the channel.
3. MEtal Semiconductor FET (MESFET). If the reverse biased pn-junction is
replaced by a Schottky contact, the channel can be controlled just as in the JFET
case.
4. Hetero FET. As the name implies (and unlike the previous three cases, whose
constructions rely on a single semiconductor material such as Si, GaAs, SiGe, or
ln.P) the hetero structures utilize abrupt transitions between layers of different
semiconductor materials. Examples are GaAIAs to GaAs or GainAs to GaAlAs
interfaces. The High Electron Mobility Transistor (HEMT) belongs to this
class.
Figure 6-39 provides an overview of the first three types. In all cases the current
flow is directed from the source to drain, with the gate controlling the current flow.
Due to the presence of a large capacitance formed by the gate electrode and the
insulator or reverse biased pn-junction, MISFETs and JFETs have a relatively low cutoff frequency and are usually operated in low and medium frequency ranges of typically up to 1 GHz. GaAs MESFETs find applications up to 60-70 GHz, and HEMT can
operate beyond 100 GHz. Since our interest is geared toward RF applications, the
emphasis will be on the last two types.
In addition to the above physical classification, it is customary to electrically classify FETs according to enhancement and depletion types. This means that the channel
either experiences an increase in carriers (for instance the n-type channel is injected
with electrons) or a depletion in carriers (for instance then-type channel is depleted of
electrons). In Figure 6-39 (a) the FET is nonconducting~ or normaiJy..off, until a sufficiently positive gate voltage sets up a conduction channel. Normally-off FETs can only
be operated in enhancement mode. Alternatively, normally-on PETs can be of both
enhancement and depletion types.
330
Source Gate
Drain
p-type substrate
(a) Metal insulator semiconductor FET (MISFET)

Gate
Drain
n
p substrate
(b) Junction field effect transistor (JFET)
Gate
Drain
S~-insulating layer .
:
~ '
(c) Metal semiconductor FET (MESFET)

Figure 639 Construction of (a) MISFET, (b) JFET, and (c) MESFET. The shaded
areas depict the space charge domains.
6.4.2
331
Functionality
Because of its importance in RF and MW amplifier, mixer, and oscillator circuits,

we focus our analysis on the MESFET, whose physical behavior is in many ways similar to the JFET. The analysis is based on the geometry shown in Figure 6-40 where the
transistor is operated in depletion mode.
Low VDs
High VDs
,_.- 1111~-----+---oD
-----.11 ....+--o D
VGs
+ t-- - o
(a) Operation in the linear region.

(b) Operation in the saturation region.
Figure 6-40 Functionality of MESFET for different drain-source voltages.
The Schottky contact builds up a channel space charge domain that affects the
current flow from source to drain. The space charge extent d s can be controlled via the
gate voltage in accordance to our discussion in Section 6.1.3, where (6.39) is adjusted
such that VA is replaced by the gate source voltage V Gs :
2e(vd- vGs)
q
ND
(6.78)
For instance, the barrier voltage V d is approximately 0.9 V for a GaAs-Au interface.
The resistance R between source and drain is predicted by
R=
L
a(d- d 5 ) W
(6.79)
with the conductivity given by 0' = q~nND and W being the gate width. Substituting
(6.78) into (6.79) yields the drain-current equation:
(6.80)
332
where we have defined the conductance G0 = a W d I L. This equation shows that the
drain current depends linearly on the drain source voltage, a fact that is only true for
small V os
As the drain-source voltage increases, the space charge domain near the drain
contact increases as well, resulting in a nonuniform distribution of the depletion region
along the channel; see Figure 6-40(b). If we assume that the voltage along the channel
changes from 0 at the source location to V DS at the drain end, then we can compute the
drain current for the nonuniform space charge region. This approach is also known as
the gradual-channel approximation. The approximation rests primarily on the
assumption that the cross-sectional area at a particular location y along the channel is
given by A (y) = { d - d 5 ( y)} W and the electric field E is only y-directed. The channel current is thus
dV(y)
10 = - aEA(y) =a dy {d-d 5 (y)}W
(6.81)
where the difference between V d and Vas in the expression for d 8 (y) has to be augmented by the additional drop in voltage V(y) along the channel; that is, (6.78)
becomes
2
ds(Y) = [ qND(Vd-Vas +V(y))
]112
(6.82)
Substituting (6.82) into (6.81) and carrying out the integration on both sides of the
equation yields
J~I 0dy = I 0L = crW( '(d- Jq'7:D Jv + Vr vGs)dv

0
(6.83)
The result is the output characteristic of the MESFET in terms of the drain current as
a function of V ns for a given fixed Vas , or
In
= Go(V DS- j
~( {V
~qNDd2
DS
+ Vd- V Gs}
312
{Vd- V Gs}
3 2
/
]J
(6.84)
This equation reduces for small V DS to (6.80) .

An interesting phenomenon occurs when the space charge extends over the entire
channel depth d. The drain-source voltage for this situation is called drain saturation
voltage V Dsat and is given by
d s(L)
=d
(6.85)
RF Fteld Effect Ttans\stots
333
or, explicitly,
2
V Dsat
qNvd
2E
- ( V d- V GS)
= V P- V d + V GS
(6.86)
= V GS- V TO
2
where we introduced the so-called pinch-otT voltage V P = qNvd /(2E) and

threshold voltage V TO = V d- V p . The associated drain saturation current is found by
inserting (6.86) into (6.84) with the result
In,.,
312
= G0 [~P- (Vd- Vos> + }v;<Vr VGs> ]
(6.87)
The maximum saturation current in (6.87) is obtained when Vas = 0, which we

define as I Dsat( V GS = 0)= I nss. In Figure 6-41 the typical input/output transfer as
well as the output characteristic behavior is shown.
I Dsa.t IIDSS
1
.,..______
I D r--"-->
Linear _____
Saturation
. . ; _ - - - - VGS = 0
.l-----}v~ < o
s
-1
(a) Circuit symbol

Figure 6-41
(b) Transfer characteristic

(c) Output characteristic
Transfer and output characteristics of an n-channel MESFET.
The saturation drain current (6.87) is often approximated by the simple relation
In,, = Ioss(l- ~;:f
(6.88)
How well (6.88) approximates (6.87) is discussed in the following example.
----------------------------~~&)A~
Example 6-9: Drain saturation current in a MESFET
16
A GaAs MESFET has the following parameters: N D = 10 em- ,

d = 0.75 Jlm,
W = 10 )lm,
L = 2 J1m,
E, = 12.0,
2
Vd = 0.8 V, and J..ln = 8500 cm /(Vs). Determine (a) the pinch-
334
off voltage, (b) the threshold voltage, (c) the maximum saturation
current I vss; and plot the drain saturation current based on (6.87)
and (6.88) for V GS ranging from - 4 to 0 V.
Solution:
The pinch-off voltage for the FET is independent of
the gate-source voltage and is computed as
vp
qNDd
= 4.24
Knowing V P and the barrier voltage V d = 0.8 V, we find the

threshold voltage to be V TO = V d- V p = -3.44 V. The maximum
saturation drain current is again independent of the applied drainsource voltage and, based on (6.87), is equal to
IDss
=Go[~P- vd + 3
}y; v/
12
]
=6.89 A
where G0 = aqN0 Wd/ L

q JlnNDWd / L
8.16 S .
Figure 6-42 shows results for the saturation drain current computed by using the exact formula (6.87) and by using the quadratic
law approximation given by (6.88).
7
<
"'"'c~
E
~
~
a
0
-~
Quadratic law
approximation
0
~
.a""'
(/')
~Exact formula
0~--~--=---~--~----~--~--~--~
-4
Figure 6-42
- 3.5
-3
- 2.5
-2
- 1.5 - 1
Gate-source voltage 'Vc,5 , V
-{).5
Drain current versus V65 computed using the exact and the
approximate equations (6.87) and (6.88).
335
Because of the excellent agreementt the quadratic law approximation (6.88) is more widely used in the literature and data sheets
than the exact equation.
If V vs reaches the saturation voltage V Dsat for a given V cs, the space charges
pinch off the channel. This implies that the drain current saturates. Interestingly, pinchoff does not imply a zero I D since there is no charge barrier impeding the flow of carriers. It is the electric field as a result of the applied voltage V DS that "pulls" the electrons across the depletion space charge domain. Any additional increase V DS > V Dsat
will result in a shortening of the channel length from the original length L to the new
length L' = L- ilL. The result is that (6.87) must be modified to
I'D= ID(L:M) = ID(f,)
(6.89)
The change in channel length as a function of VDS is heuristically taken into account
through the so-called channel length modulation parameter A = MI(L'V vs). This
is particularly useful when expressing the drain current in the saturation region:
(6.90)
where measurements show a slight increase in drain current as V DS is increased.
-----------------------------~&uM~
Example 6-10: I-V characteristic of a MESFET
For discrete gate-source voltages V GS = -1, -1.5, -2, and -2.5 V,
plot the drain current I D of a MESFET as a function of drain-source
voltage V DS in the range from 0 to 5 V. Assume that the device
parameters are the same as in the previous example and that the
channel length modulation parameter A is set to be 0.03 v- 1 Compare your results with the case where A = 0 .
Solution:
In the analysis of the MESFET behavior we have to be

careful about choosing the appropriate formulas. At very low drainsource voltages, the drain current can be described by a simple linear relation (6.80). As the voltage increases, this approximation
336
becomes invalid and a more complicated expression for I D has to be

employed; see (6.84). Further increase in V DS ultimately leads to
channel pinch-off, where V DS ~ V Dsat = V cs - V TO . In this case the
drain current is equal to the saturation current given by (6.87). Additional increases in VDs beyond the saturation voltage result only in
minor increases of the drain current due to a shortening of the channel. At this point, I D is linearly dependent on V DS. Substituting
(6.87) into (6.90) for V DS ~ V Dsat, we obtain
cs)3} (l+'AVDs)
2)<V d- v
vP
lv =Go { 3-(Vd-Vcs)+3
,JV;
To provide a smooth transition from normal to saturation region for

nonzero 'A. we multiply (6.84) by ( 1 + 'A.V vs). Thus, the final
expression for the drain current for V DS $ V Dsat is
ID
2)(Vvs+Vd-Vcs )3 -j(Vd - Vcs)

Go { V DS- 3
h!
"'v p
'\
( 1 + 1\. V vs)
The results of applying these formulas to predict I D for zero (dashed

line) as well as nonzero 'A (solid line) are shown in Figure 6-43.
3.5
<.
~ 2.5
- 1.5
~
~
1-<
----------- --- ---- ---- ----- --- -- --- --- - 'Vc;s
0.5
0
0
Figure 6-43
3
Drain-source voltage
-2.5 V
Vvs, V
Drain current as a function of applied drain-source voltage for

different gate-source biasing conditions.
337
The channel length modulation is similar to the Early effect

encountered in a Bfl', where the collector current in saturation
mode increases slightly for increasing collector emitter voltage as
discussed in Chapter 7.
6.4.3
Frequency Response
The high-frequency MESFET performance is determined by the transit time of

charge carriers traveling between source and drain and the RC time constant of the
device. Here we will focus our attention on the transit time only and defer the time constant computation, which requires knowledge of the channel capacitance, to Chapter 7.
Since electrons in silicon and GaAs have much higher mobility than holes, n-channel
MESFETs are used in RF and MW applications almost exclusively. Furthermore, since
the electron mobility of GaAs is roughly five times higher than that of Si, GaAs MESFETs are usually preferred over Si devices.
The transit time t of the electrons traveling through the channel of gate length L
is computed as
t =
(6.91)
vsat
where we have assumed a fixed saturation velocity vsat . As an example, the transition
frequency f r = l/(21tt) for a gate length of LO J.lm and a saturation velocity of
7
approximately 10 cm/s is 15 GHz.
6.4.4
Limiting Values
The MESFET must be operated in a domain limited by maximum drain current

IDmax, maximum gate-source voltage V GSmax, and maximum drain-source voltage
VDSmax . The maximum power P max is dictated by the product of V DS and I D, or
p max = VDS/D
(6.92)
which in tum is related to the channel temperature T c and ambient temperature T a

and the thermal resistance between channel and soldering point Rthjs, according to
T c = Ta + Rthjsp
(6.93)
Figure 6-44 clarifies this point. Also shown in this figure are three possible operating points. Bias point 3 indicates low amplification and possible clipping of the output
current. However, the power consumption is at a minimum. Bias point 2 reveals accept-
Chapter 6 AcUve RF Components
338
Figure 6-44
Typical maximum output characteristics and three operating points

of MESFET.
able amplification at substantially increased power consumption. Finally, bias point 1

shows high amplification at high power consumption and low output current swing.
Choosing appropriate bias points for specific applications will be investigated in-depth
in subsequent chapters.
6.5 High Electron Mobility Transistors

The high electron mobility transistor (HEMT), also known as modulation
doped field effect transistor (MODFET), exploits the differences in band gap energy
between dissimilar semiconductor materials such as GaAIAs and GaAs in an effort to
substantially surpass the upper frequency limit of the MESFET while maintaining low
noise perfonnance and high power rating. At present, transit frequencies of 100 GHz
and above have been achieved. The high frequency behavior is due to a separation of
the carrier electrons from their donor sites at the interface between the doped GaAIAs
and undoped GaAs layer (quantum well), where they are confined to a very narrow
(about 10 run thick) layer in which motion is possible only parallel to the interface.
Here we speak of a two-dimensional electron gas (2DEG) or plasma of very high
2
mobility, up to 9000 cm /(Vs). This is a major improvement over GaAs MESFETs
2
with J.ln::::: 4500 cm /(Vs). Because of the thin layer, the carrier density is often spec~
ified in terms of a surface density, typically on the order of 10 12- 10 13 cm-2
To further reduce carrier scattering by impurities it is customary to insert a spacer
layer ranging between 20 and 100 nm of undoped GaAIAs. The layer is grown through
a molecular beam epitaxial process and has to be sufficiently thin so as to allow the gate
voltage V GS to control the electron plasma through electrostatic force mechanism.
Besides single layer heterostructures (GaAlAs on GaAs), multilayer heterostructures
33t
High Electron Mobility Transistors
involving several2DEG channels have also been proposed. As can be expected, manufacturing an HEMT is significantly more expensive when compared with the relatively
inexpensive GaAs MESFET due to the precisely controlled thin-layer structures, steep
doping gradients, and the use of more difficult to fabricate semiconductor materials.
6.5.1
Construction
The basic heterostructure is shown in Figure 6-45, where a GaAlAs n-doped

semiconductor is followed by an undoped GaAIAs spacer layer of the same material, an
undoped GaAs layer, and a high resistive semi-insulating (s.i.) GaAI substrate.
Source
Gate
Drain
-d
Figure 6-45
Generic heterostructure of a depletion-mode HEMT.
The 2DEG is formed in the undoped GaAs layer for zero gate bias condition
because the Penni level is above the conduction band so that electrons accumulate in
this narrow potential well. As discussed later, the electron concentration can be
depleted by applying an increasingly negative gate voltage.
HEMTs are primarily constructed of heterostructures with matching lattice constants to avoid mechanical tensions between layers. Specific examples are the GaAlAsGaAs and InGaAs-InP interfaces. Research is also ongoing with mis-matched lattices
whereby, for instance, a larger InGaAs Lattice is compressed onto a smaller GaAs Lattice. Such device configurations are known as pseudomorphic HEMTs, or pHEMTs.
6.5.2
Functionality
The key issue that determines the drain current flow in a HEMT is the narrow
interface between the GaAlAs and the GaAs layers. For simplicity, we neglect the spacer
layer and concentrate our attention at the energy band model shown in Figure 6-46.
A mathematical model similar to (6.21) can be developed by writing down the
one-dimensional Poisson equation in the form
340
Schottky
contact
..... ..................... .....

~
.~
/ .
2DEG
;,.
.---------..
..
.. .- -.................. _. ........ ..................... - ::::: ::"~.:::' ..
..
-- -w,:/
_GaAs
__.,.., /
/
: ;)--
-......... GaAIAs : .
'.
-d
.. ..
(a) Energy band diagram

(b) Close-up view of conduction band
Figure 6-46 Energy band diagram of GaAIAs-GaAs interface for an HEMT.
dx
(6.94)
En
where N D and e8 are the donor concentration and dielectric constant in the GaAlAs
heterostructure. The boundary conditions for the potential are imposed such that
V(x =0) =0 and at the metal-semiconductor side V (x = -d) = - V b + V G + t:.. W cl q.
Here V b is the barrier voltage, see (6.38); t:,. W c is the energy difference in the conduction levels between the n-doped GaAIAs and GaAs; and V G is comprised of the gatesource voltage as well as the channel voltage drop V G = - V GS + V (y) . To find the
potential, (6.94) is integrated twice. At the metal-semiconductor we set
qND 2
V(-d) = - x - E (O)d
2eH
(6.95)
which yields
(6.96)
where we defined the HEMT threshold voltage V ro as V ro = V b - d Wcl q - Vp.
2
Here we have used the previously defined pinch-off voltage V P = qNvd /(2E8).
From the known electric field at the interface, we find the electron drain current
(6.97)
As mentioned previously, the current flow is restricted to a very thin layer so that it is :
appropriate to carry out the integration over a surface charge density Q5 at x = 0. The
341
High Electron Mobility Transistors
result is cr
= -JlnQ/(WLd) = -JlnQ 5 /d. For the surface charge density we find with
Gau~~,~ \a"W
Qs -= cHE\\)). \1\.~e.tte\\ \\\ \_().91), "We ()\)\a\:t.\
IDdy =
J.l.w(Ds QsdV
(6.98a)
Upon using (6.96), it is seen that the drain current can be found
Vvsf.H
IDL = flnW
d(V cs- V- Vr 0 )dV
(6.98b)
or
(6.98c)
Pinch-off occurs when the drain-source voltage is equal to or less than the difference of
, gate-source and threshold voltages (i.e., V DS ~ V cs- V TO). If the equality of this condition is substituted in (6.98c), it is seen
WeH
ID = Jln 2Ld(V cs- VTo)
(6.99)
The threshold voltage allows us to determine if the HEMT is operated as an

enhancement or depletion type. For the depletion type we require V TO < 0, or
2
vb- (AW cl q)- v p < 0. Substituting the pinch-off voltage v p = qN Dd /(2) and
solving for d, this implies
c)}
2eH
~W
d> { - ( Vb--qND
112
(6.100)
and if dis less than the preceding expression (i.e., V ro > 0 ), we deal with an enhancementHEMT.
---------------------------~~~
Example 6-11: Computation of HEMT-related electric characteristics
Determine typical numerical values for a HEMT device such as
pinch-off voltage, threshold voltage, and drain current for
V GS = -I, -0.75, -0.5, -0.25, and 0 V as a function of drain-
342
source voltage V D S . Assume the following parameters:

18
-3
N 0 = 10 em , V b = 0.81 V , H = 12.50 , d = 50 nm,
-20
~ W c = 3.5 x 10
Ws, W = 10 Jlm , L = 0.5 IJ-m , and
2
lln = 8500 em /( Vs) .
Solution:
The pinch-off voltage of a HEMT is evaluated as

2
V p = qN vd / (2H)
= 1.81 V
Knowing V P we can find the threshold voltage as

V TO
= Vb -
AW cl q - V p
= -]. 22 V
Using these values the drain current is computed by relying either on

equation (6.98c) for V os :s; V GS- V TO or equation (6.99) for
V os ~ V GS - V TO . The results of these computations are plotted in
Figure 6-47. We notice in this graph that unlike the GaAs MESFET in
Figure 6-43, a channel length modulation is not taken into account. In
practical simulations such a heuristic adjustment can be added.
30
Vas= 0 V
25
~~ 20
Vas = -0.25 V
.....
c
~
5
u
15
-~
0""
Vcs = -0.5 V
10
Vas= -0.75 V
1-Gs = -1 V
00
Drain-source voltage Jt;)s, V

Figure 6-47
Drain current in a GaAs HEMT.
Both GaAs MESFET and HEMT exhibit similar output characteristics and are thus represented by the same electric circuit model.
~mmary
6.5.3
343
Frequency Response
The high-frequency performance of the HEMT is detennined by the transit time

similar to the MESFET. However, the transit time t is expressed best through the electron mobility lin and the electric field E of the drain-source voltage according to
't
= -L
V sat
=-L
JlnE y
L2
lln V DS
(6.101)
We therefore obtain a transit frequency f T = l/(21tt) of a~proximately 190 GHz for

the gate length of 1.0 Jlm and a mobility of Jln = 8000 em /(Vs) at a typical drain
voltage V os of 1.5 V.
6.6 Summary
To understand the functionality and limitations of the most widely employed
active RF solid-state devices, we commenced this chapter with a review of the key elements of semiconductor physics. The concepts of conduction, valence, and Fermi levels
as part of the energy band model are used as the starting point to examine the various
solid-state mechanisms.
We next turned our attention to the pn-junction, where we derived the barrier
voltage
Vdiff
rln(N:~D)
and the depletion and diffusion capacitances C d and C s in the forms

and
Both capacitances are of primary importance when dealing with the frequency response
of a pn-diode whose current is given by the Shockley equation
This equation underscores the nonlinear current-voltage diode characteristics.

Unlike the pn-junction, the Schottky contact involves an n-type semiconductor
and a metal interface. The Schottky barrier potential V d is now modified and requires
the work function of metal, q V M , semiconductor, qX , and the conduction band potential Vc, expressed via
344
Vd = (VM - X)- V c
Unlike the 0.7 V of a pn-junction, we obtain a typical value of 0.84 V for a Si-Au interface. Technologically, this contact is exploited in the Schottky diode, which has become
ubiquitous in many RF applications such as modulators and mixers. The I-V characteristic remains the same as for the pn-junction diode, except that the reverse saturation
current Is is theoretically more intricate.
Additional special-purpose RF diodes are the PIN, varactor, and tunnel diodes.
The PIN diode incorporates an additional intrinsic layer sandwiched between the p and
n layers. This allows the switching between a low-resistance forward bias to a capacitive reverse, or isolation, bias. PIN diodes find applications in switchers and attenuators. The /-V characteristic of a PIN diode is very similar to a pn-junction diode but
differs by the factor 2 in the exponent:
I=
A(q:~W)(//(2V7 )
-I)
The varactor diode incorporates the /-layer based on a special doping profile to achieve
a particular capacitance-voltage behavior. Such a response is beneficial for frequency
tuning and the generation of short pulses. The tunnel diode exhibits a negative slope
during a particular portion of its /- V curve, thus making it suitable for oscillator circuits.
Additional diodes of interest in the RF field are the IMPAIT, TRAPATT, BARRITI,
and Gunn diodes.
The BJT in many ways can be regarded as an extension of our previous diode discussion since the npn-structure constitutes the series connection of two diodes. The
three transistor modes forward active, reverse active, and saturation are reflected in the
emitter, collector, and base current expressions (6.69)-(6.71):
The frequency response of a BJT is determined by the transit or transition frequency

f r = l/(21tt) at which the short-circuit current gain is equal to unity. The time con- :
stant is comprised of three delays 't = t + t 8 +-reassociated with emitter, base, and
collector domains.
Further Readtng
345
Unlike the bipolar BJT, the FET is a monopolar device that displays superior
high-frequency and low-noise performances. In particular, n-channel GaAs MESFETs
are commonly found in many RF amplifiers, mixers, and oscillators. The key equation
that determines the output characteristic of a MESFET is the drain current (6.84):
lv =
Ga(vvs- ~ ~qNvd2
~[ {V os + Vd- VGs} 3/ 2 - {Vd- V Gs} 3/ 1)
2
Additional modifications to the drain current are required when the channel is pinched
off and the FET is operated in the saturation domain with channel length modulation.
Finally, the HEMT device is almost identical in construction with the MESFET,
but exploits the differences in bandgap energies between heterogeneous semiconductors. Here the current flow is restricted to a very narrow, quantum well layer where the
charge mobility can attain twice the value of a MESFET. Because of carrier separation
from the donor sites, extremely high operational frequencies have been reported
(exceeding 100 GHz). The drain-current representation is almost identical with the one
discussed for the MESFET.
Further Reading
R. S. Cobbold, Theory and Applications of Field-Effect Transistors, John Wiley, New

York, 1970.
A.M. Cowley and S.M. Sze, "Surface States and Barrier Height of Semiconductor
Systems," J. Appl. Physics, Vol. 36, pp. 3212-3220, 1965.
M. B. Das, "Millimeter-Wave Performance of Ultra-Submicrometer Gate Field-Effect

Transistors. A Comparison ofMODFET, MESFET, and HBT-Structures," IEEE Trans.
on Electron Devices, Vol. 34, pp. 1429-1440, 1987.
A. S. Grove, Physics and Technology ofSemiconductor Devices, John Wiley, New York,
1967.
G. Massobrio and P. Antognetti, Semiconductor Device Modeling with SPICE,
McGraw-Hill, New York, 1993.
J. L. Moll, Physics of Semiconductors, McGraw-Hill, New York, 1964.

D. V. Morgan and N. Parkman, Physics and Technology of Heterojunction Devices, P.
Peregrinus Ltd., London, UK, 1991.
M. H. Norwood and E. Schatz, "Voltage Variable Capacitor Tuning-A Review," Proceed. IEEE, Vol. 56, pp. 788-798, 1968.
346
R.S. Pengelly, Microwave Field-effect Transistors-Theory, Design and Applications,

Research Studies Press, London, UK, 1982.
C. T. Sah, "Characteristics of the Metal-Oxide Semiconductor Field-Effect Transistor,"
IEEE Trans. on Electron Devices, Vol. 11, pp. 324-345, 1964.
W. Shockley, Electrons and Holes in Semiconductors, Van Nostrand Reinhold, New
York, 1950.
M. Shur, GaAs Devices and Circuits, Plenum Press, New York, 1987.
S. M . Sze, Physics of Semiconductor Devices, John Wiley, New York, 1981.
C. Weisbuch. Physics and Fabrication of Microstructures and Microdevices, SpringerVerlag, New York, 1986.
Problems
6.1
6.2
To appreciate the large number of atoms in semiconductors, let us consider

the following simple calculation: A silicon semiconductor is a bo<.!l-centered
cubic semiconductor with a lattice constant of a = 5.43 x 10 em. The
atom arrangement is such that a comer atom contributes one-eighth plus one
center atom. Find the density of atoms per cubic centimeter.
The conduction and valence band carrier concentration is determined by
integration of the density of states based on the Fermi statistics.
N ==
For effective electron mass

the density function
Jg(E)dE
m;, quantum mechanical considerations lead to
g(E) = 41t(2m; )
312
JE! h
(a) Determine the generic electron concentration of states N for energy values up to 1.5 eV.
31
(b) For an effective electron mass of 1.08mn or 1.08 x 9.11 x 10- kg,
explicitly find the number of states.
6.3
Let us consider a p-type Si semiconductor whose doping concentration at

16
room temperature contains N A = 5 x 10 boron atoms per cubic centime10
3
ter ( ni = 1.5 x 10 em- ). Find the minority and majority carrier concentrations as well as the conductivity of the semiconductor.
.,
.>
":j
l
:1
Problems
347
6.4
The Fermi-Dirac probability for indistinguishable particles is the underlying

statistical theory describing the quantum mechanical distribution of particles
per unit volume and per unit energy N(E) normalized with respect to the
number of quantum states per unit voltage and per unit energy g(E) according to
1
f(E) - g(E} - l + exp([E- EF]I[kT])
_ N(E) _
(a) Plot both the probability of states being occupied, ft.E), as well as the
probability of states being empty, 1 - ft.E), at room temperature and for
EF= 5 eV.
(b) Detennine the temperature at which we have a 5% probability of
encountering an empty state.
6.5
The intrinsic carrier concentration is typically recorded at room temperature.

For GaAs we find at T = 300K the effective densities of state
17
18
N c = 4.7 x 10 em-J, N v = 7.0 x 10 em-J. Assuming that the bandgap energy of 1.42 eV remains constant,
(a) Find the intrinsic carrier concentration at room temperature.
(b) Compute n; at T = 400K.
(c) Compute n; at T = 450K.
6.6
It is interesting to observe that a significant diffusion current density can be

created even for moderate carrier concentration gradients. We can assume
for a p-~pe Si semiconductor a linear hole concentration changing from
17
3
18
3
5 x 10 cm- to 10 cm over a distance of 100 J.!m. Find the current
density if the diffusion coefficient is given at T = 300K to be
2
DP = 12.4 em /s.
6.7
In Section 6.1.2 we derived the expressions for the electric field and potential distributions in the pn-diode with abrupt junction. Repeat these computations for a case of gradual junction where the charge density changes
linearly according to the following relation:
qNA(x/ dp), -dp ~X~ 0
p(x)
6.8
= { qND(xl dn), 0 ~X< dn
The built-in potential barrier of a pn-junction remains relatively constant

even though the doping concentration may change over several orders of
magnitude. We recall that the typical barrier potential in solid state circuits is
348
assumed to be 0.5-0.9 V. In this problem we intend to show how one arrives .

at this voltage. Let us assume a p-type semiconductor with
. . d
.
ductor of concentratiOn
.
N A = 10 18 em-3 JOtne
w1'th an n-type semtcon
15
-3
N D = 5 x 10 em .
10
(a) Find the barrier voltage at room temperature (n ; = 1.45x l0 cm-3 ).
(b) Recompute the barrier voltage if N A
IS
reduced to
16
-3
NA = 5 X 10 em .
6.9
An abrupt pn1unction made of Si has the acceptor and donor concentrations

1
3
15
of N A = 10 em- and N D = 5 x 10 em- respectively. Assuming that
the device operates at room temperature, detennine
(a) the barrier voltage
(b) the space charge width in the p- and n-type semiconductors
(c) the peak electric field across the junction
4
2
(d) the junction capacitance for a cross sectional area of 10- cm and a relative dielectric constant of r = 11.7
6.10
For two pn-diodes with abrupt junction, one of which is made of Si and
17
14
3
another is made of GaAs, with N A = I 0 em-J and N D = 2 x 10 emin both cases:
(a) Find the barrier voltage.
(b) Find the maximum electric field and the space charge region width.
(c) Plot the space charge, potential, and electric field distribution along the
diode axis.
6.11
A silicon pn junction has a conductivity of 10 S/ em and 4 S / em for p and

n layers, respectively. Using the necessary properties of silicon, calculate the .
built-in voltage of the junction at room temperature.
6.12
A Schottky contact between a metal and a semiconductor can be made of

various materials. For both Si and GaAs we would like to investigate the barrier voltage if the metal is either aluminum or gold. Use Table 6-2 and Table
E-1 to find the four barrier voltages and associated depletion layer thicknesses at room temperature.
6.13
Consider a Schottky diode formed by the contact between n-type GaAs and
silver. The diode is operated at forward biasing current of 1 mA. The Richardson constant R* = 4 AI (em 2 K 2 ) , the parasitic series resistance is
15 .Q , and the device cross section A = 1o-2 mm 2 Compute the barrier
voltage V d and plot the magnitude and phase of the impedance of diode ver-
;
,
:
Problems
349
sus frequency ranging from 1 MHz to 100 GHz for two doping densities
N D: 10 15 and 10 17 cm- 3 . Assume that the device is operated at the temperature of 300K.
6.14
It is often of enormous practical interest to investigate the nonlinear current

behavior of a Schottky diode for a given applied voltage. We recall
I=Is ( e
(V A- IR 5 )
-I
11
with the reverse saturation current given to be Is = 2 x 10- A. For a substrate resistance R s = 1.8 Q write a computer program to predict the current if the applied voltage is allowed to vary within 0 ~ VA ~ 10 V .
6.15 A PIN diode is a semiconductor device with an intrinsic layer sandwiched

between two highly doped n-and p-type materials. In the intrinsic layer, the
charged minority and majority carriers possess a finite life time 'tP before
recombination takes place. On the basis of the recombination lifetime a simple PIN model can be constructed involving the diode current I and the
stored charge Q:
I= Q + dQ
'tp
dt
(a) Establish the frequency domain response Q( ro) of this first-order
system.
(b) Plot the normalized charge response 20 log [ Q( ro) I (I 'tP)] versus angular frequency for 'tP of I 0 ps, 1 ns, and 1 J.lS .
Note: For frequencies well below the cut-off frequency f P = 1I 'tP the PIN
diode behaves like a normal pn-junction diode. However, at frequencies
above f P , the PIN diode becomes a pure linear resistor whose value is controlled by the biasing signal.
6.16
The fabrication of two different types of varactor diodes calls for the following two capacitance-voltage behaviors:
(a) c
5 pF Jv AI (V A- Vdiff)
(b) C
5 pF ( VA I ( VA - V diff )) 113
Determine the necessary donor doping profile N v(x) for the intrinsic layer.
2
Assume the cross-sectional area of the varactor diode to be 10-4 cm
=
=
6.17
Consider a Si bipolar junction transistor whose emitter, base, and collector

21
3
are uniformly doped with the followin2 concentrations: N~ = 10 em- ,
8
17
-3
c
19
-:r
N A = 2x 10 em , N D = 10 em . Assume that the base-emitter volt-
350
Chapter 6 Acttve RF Components
age is 0.75 V and the collector-emitter potential is set to 2 V. The cross2

sectional area of both junctions is 10-4 cm and the emitter, base, and collector thicknesses are dE = 0.8 J.l.m , dB = 1.2 f.1m , and d c = 2 J.Lm,
respectively. Assuming that the device is operated at room temperature:
(a) Find the space charge region extents for both junctions.
(b) Draw the energy band diagram.
(c) Compute the base, emitter, and collector currents.
(d) Calculate the forward and reverse current gains ~F and ~R.
6.18
For a GaAs BJT the maximum junction temperature is 420C (which far
exceeds the maximum junction temperature of Si with 200C ). The supplied power is 90 W. The thermal resistance between the BJT and the heat
sink is estimated to be 1.5C/ W
(a) Determine the maximum thennal resistance of the heat sink if the ambient operating temperature does not exceed 50 C .
2
(b) For a heat convection coefficient of 100 W ;oc m find the required
surface area.
6.19
A BIT is encapsulated in a plastic housing and mounted on a heat sink

(Rthha = 3.75C/ W ). Under these conditions the total power dissipation is
supposed to be 20 W at an ambient temperature of 20C. What rating has the
engineer to choose for the BJT casing if the maximum junction temperature
should not exceed 175C?
6.20 Prove that the drain current (6.84) for a MESFET under gradual-channel
approximation reduces to (6.80) for small VDs
6.21
Derive the saturation drain current equation (6.88).
6.22 The junction field-effect transistor with n-type channel has the following
parameters: W/L = 10, J.ln = 1000 m 2/(Vs), d = 2J.lm, Er = 11.7,
and V TO = -3 V . Compute the saturation drain current at V Gs = -1 V
6.23
Compute the output current I D versus V DS characteristics of the transistor

from Problem 6.22 for drain-source voltage ranging from 0 to 5 V. First
assume that channel length modulation effect is negligible (i.e., A. = 0 ), and
then repeat your computation for a case when A. = 0.01 .
CHAPTER
. Active RF Co1nponent
Modeling
i
f
~
t
~
;:
~~.
!most all circuit designs of any complexity have

to be modeled as part of computer-aided design (CAD) programs prior to their practical
realizations to assess quantitatively whether or not these circuits meet design specifications. For the purpose of electric circuit simulation, a large number of software analysis
packages offer a host of equivalent circuit models attempting to replicate the electric
performance of the various discrete elements. Special electric circuit models have been
developed to address such important design requirements as low- or high-frequency
operation, linear or nonlinear system behavior, and normal or reverse mode of operation to name but a few.
It is the purpose of this chapter to examine several active devices in terms of suitable equivalent circuit representations for diodes as well as mono- and bipolar transistors. The physical foundation of these devices is reviewed in Chapter 6. By developing
a close link with the previous chapter, we will be able to observe how a basic understanding of solid-state device physics naturally leads to large signal (nonlinear) circuit
models. Subsequent discussions will focus on modifications that can be made to linearize these models and to refine them for high-frequency operations.
Considering the various BJT models, we restrict our discussion to only the most
popular types such as the Ebers-Moll and Gummel-Poon models. Both types, and a
number of linear derivatives, find widespread applications in such simulation tools as
SPICE, ADS, MMICAD, and others. Often the situation arises where the device manufacturer may not be able to specify all the required electric parameters. since they can
easily exceed 40 independent parameters, and a so-called SPICE model representation
is unattainable. Under those circumstances. the S-parameters are recorded for various
bias conditions and operating frequencies to characterize the high-frequency behavior.
351
352
Chapter 1 Active RF Component Modetlng
In most cases, these S-parameters may provide the design engineer with sufficient
information to complete the simulation task.
7.1 Diode Models

7.1.1
Nonlinear Diode Model
The typical large-scale circuit model treats both the pn and Schottky diode in the
same fashion, as shown in Figure 7-1.
.,
Rs
~ r-[)h
Io
Figure 71
.I
Large-scale diode model.
This model takes into account the nonlinear I-V characteristics of the Shockley
diode equation (6.34) in slightly modified form
(7.1)
where the emission coefficient n is chosen as an additional parameter aimed at bringing

the model in closer agreement with actual measurements. This coefficient for most
applications is close to 1.0. Furthermore, in Section 6.1.2 the diffusion and junction (or
depletion layer) capacitances C d and C 1 are discussed. Both effects are combined in a
single capacitance C, but in a more general form. Specifically, for the junction capacitance, we have to consider the space charge Q 1 , which is differentiated with respect to
the applied voltage, leading to
(7.2)
where m is known as the junction grading coefficient. It assumes a value of 0.5 for the
abrupt junction that is subject of our analysis in Section 6.1.2. For the more realistic
case of a gradual transition m lies in the range 0.2 ~ m ~ 0.5. As mentioned in
Chapter 6, the formula given in (7.2) is applicable only for certain positive applied voltages. If the applied voltage VA approaches the built-in potential V diff' the junction
Diode Models
353
capacitance computed using (7 .2) approaches infinity which is obviously physically

impossible. In practice, the junction capacitance becomes almost linearly dependent on
the applied voltage once it exceeds a threshold potential V m, which is usually equal to
half of the built-in potential, V m ~ 0.5 V diff. Therefore, the approximate formula capable of describing the junction capacitance over the entire range of applied voltages is
given by
(7.3)
We also observe that Cd is dependent on VA . For the diffusion capacitance, we

. can use
(7.4)
with the transition time tr.

In a realistic diode the injection and extraction of charges is accomplished by the
. electric field that constitutes a voltage drop in the charge neutral domains. This voltage
.drop is modeled as a series resistance R 5 . Thus the total voltage in Figure 7-1 is composed of two contributions:
(7.5)
..
Temperature dependencies can also be introduced into this model. Besides the
:,obvious thermal voltage V r = kT I q it is primarily the reverse saturation current Is
. :that is found to be strongly influenced by temperature according to
(7.6)
, ~where T0 is a reference temperature at which the saturation current is recorded. The lit'';erature primarily uses T 0 =300K (or 27C). The reverse saturation current temper
: ature coefficient p1 is either 3 or 2 depending on whether a pn or Schottky diode is
modeled. The model parameter can thus account for the difference in temperature
behavior between the two diode types. Also, the bandgap energy Wg(T) is consid. ered. As the temperature increases, this bandgap decreases, making it easier for charge
. carriers to transition from the valence into the conduction band. The semi-empirical
Chapter 7 Active RF Component Modeling
354
formula assumes a specific bandgap energy W g(O) recorded at T

adjusts this value as follows
= 0K
and then
(7.7)
For instance, the experimentally determined parameters for Si are WgCO) - 1.!6eV,
aT = 7.02 x 10-4 eV /K, and ~T = 1108K. Additional temperature dependencies
affecting the capacitances are usually small and are neglected.
Perhaps the most popular circuit simulation program in industry and academia is
SPICE, which is capable of taking into account the nonlinear diode model depicted in .
Figure 7-1. This simulation program incorporates a range of physical model parameters; some of them are so specialized that they are beyond the scope of our textbook.
The most important ones are summarized in Table 7-1. Also listed are the differences
between the standard pn and Schottky diode.
Table 7-1
Diode model parameters and their corresponding SPICE parameters
Symbol
SPICE
Description
Typical values
Is
IS
saturation current
1 fA-10 J.lA
emission coefficient
'tT
TI
transit time
5 ps-500 IJ.S
Rs
RS
ohmic resistance
0.1-20 Q
vdiff
VJ
barrier voltage
0.6-D.8 V (pn)
0.5--0.6 V (Schottky)
CJO
CJO
zero-bias junction capacitance
5-50 pF (pn)
0.2-5 pF (Schottky)
grading coefficient
0.2--0.5
wg
EG
bandgap energy
l.lleV(Si)
0.69 eV (Si-Schottky)
Pt
XTI
saturation current temperature coefficient
3 (pn)
2 (Schottky)
7.1.2
Linear Diode Model
The nonlinear model is based on the device physics developed in Chapter 6. AB

such, this model can be used for static and dynamic analyses under practically any circuit conditions. However, if the diode is operated at a particular DC voltage bias point
and the signal variations about this point are small, we can develop a linear or small
J)4ode Models
355
signal model. The concept of linearization implies the approximation of the exponential I-V characteristic through a tangent at the bias or Q-point V Q . The tangent at this
Q-point is the differential conductance Gd , which we can find as
1
Gd = Rd
dlv
= dVA
_ IQ+ls = IQ
(7.8)
nVy -nVy
VQ
The tangent approximation is shown in Figure 7-2 along with the simplified, linear circuit model. It is important to emphasis that the differential capacitance is now the diffusion capacitance at bias point V Q , or
(7.9)
Rd( JQ)
(a) tangent approximation at Q-point

(b) linear circuit model.
Figure 7-2 Small-signal diode model.
An apparent benefit of such a linearized circuit model is the ability to decouple the RF
diode operation from the DC bias condition, as the following design example underscores.
--------------------------------~~~
Example 7-1: Derivation of the small-signal pn diode model
A conventional Si-based pn-diode is operated at 300K and has the
following electric parameters at this temperature: 'ty = 500 ps,
- 15
Is
5 x 10
A, R s = 1.5 Q , n = 1.16 . The DC operating conditions are chosen such that I Q = 50 rnA. To characterize the per-
356
formance of a particular RF system in which this diode is used, we

need to obtain
(a) the impedance behavior of the diode in the frequency range
10 MHz <J ~ I GHz at 300K, and
(b) the impedance response of the diode in the same frequency
range, but for temperatures of 250K, 350K, and 400K.
Solution:
1Q
At a temperature of 300K, we first detennine from

50 rnA the corresponding V Q, which is found from (7 .1)
VQ
= nVrln(l +IQils) = 0.898 V
Next we can compute the differential resistance and capacitance as

Rd
n VT
= -1- = 0.6 0
Q
and Cd
I s'tr v Ql(nVr)
n T
= -V e
= 832.9 pF
Knowing these parameters, we can find the impedance of the diode

as a resistor Rs connected in series with the parallel combination of
Rd and Cd:
Rd
= Rs+ I
. C R
+ ]0>
d d
The resulting frequency behavior is shown in Figure 7-3.

As temperature changes and the biasing current I Q is maintained constant, the biasing voltage V Q should change due to the
temperature dependence of the thermal potential V T = kT I q,
bandgap energy W8 given by (7.7), and saturation current Is
described in (7.6). Results of these computations are presented in
Table 7-2, and the corresponding frequency behavior of the diode
impedance is shown in Figure 7-3.
Table 7-2
Diode model parameters for different temperatures
T,oK
250
300
350
400
Wg(n,ev
1.128
1.115
1.101
1.086
5.lxl0- 19
5.0x10- 15
3.3xl0- 12
3.8x10- 10
VQ, V
0.979
0.898
0.821
0.748
RJ,n
0.5
0.6
0.7
0.8
cd,pF
999.5
832.9
713.9
624.7
Is<n. A
Tl"'nalator Models
357
2.4
400 K
2.3
350 K
2.2
2.1
300K
2.0
1.9
1.8
250 K
1.7
1.6
1.5
10
Figure 7-3
100
Frequencyf, MHz
1000
Frequency behavior of the diode impedance for different junction

temperatures.
We observe how the physical parameters developed for the pnjunction in Chapter 6 directly translate into the small-signal circuit
model. The DC bias conditions influence the AC behavior because
they affect the differential capacitance and resistance.
7.2 Transistor Models

Over the years a number of large- and small-signal bipolar and monopolar transistor models have been developed. Perhaps the best-known one is the Ebers-Moll BIT
model, which was initially introduced to characterize static and low-frequency transistor modes. The need to expand into RFIMW frequencies and high power applications
required taking into account many important second-order effects, such as low-current
and high-injection phenomena. This has resulted in the Gummel-Poon model as a more
refined BJT circuit representation.
7.2.1
Large-Signal BJT Models
We begin our discussion with the static Ebers-MoH model, which is one of the
most popular large-signal models. Although this model was first introduced in December of 1954, it still is indispensable to understand the basic model requirements and its
extensions to more sophisticated large-signal models as well as the derivation of most
358
small signal models. Figure 7-4 shows the generic npn transistor with the associated
Ebers-Moll circuit model in the so-called injection version.
(a) Voltage and current convention for npn transistor
Bo----..---t
Eo---__..--------oE
(b) Ebers-Moll circuit model
Figure 7-4 Large-signal Ebers- Moll circuit model.
In Figure 7-4 we encounter two diodes connected in foiWard and reverse polarity, as
already seen in Chapter 6. In addition, two current-controlled current sources pennit the
mutual coupling of the two diodes as part of the base contact. The foiWard and reverse
current gains (in common-base configuration) aF and a.R possess typical values of
aF = 0.95 ... 0.99 and aR = 0.02 ... 0.05 . As a direct extension of the previously
discussed single-diode model, the dual-diode Ebers-Moll equations take on the fonn
IE= a.RIR-IF
(7..10)
Ic = aFIF-IR
(7.11)
with the diode currents

(7.12)
(7.13)
359
Tl'lnslstor Models
where the reverse collector and emitter saturation currents I cs and I ES (whose
numerical values range from 10- 9 A to 10- 18 A) can be related to the transistor saturation current Is as follows:
(7.14)
Despite their simplicity, the Ebers-Moll equations are capable of describing all major
~ physical phenomena developed in Chapter 6. For the important cases of forward and
reverse active modes, the circuit model can be simplified. The following two situations
anse:
Forward Active Mode ( V CE > V CEsat = 0.1 V, V BE== 0.7 V ). With the baseemitter diode IF conducting, and the base-collector diode in reverse direction
(i.e., V Be< 0 V ), we conclude that I R = 0 , and also aRI R = 0. The base-collector diode and the base-emitter current source can thus be neglected.
Reverse Active Mode ( V CE < -0.1 V, V BC =0.7 V ). Here the base-collector
diode IRis conducting, and the base-emitter diode is biased in reverse direction
(i.e., V BE< 0 V ), which results in IF= 0 and aFI F = 0.
Figure 7-5 summarizes these two modes of operation when the emitter is chosen as
common reference point.
B o-----+---{
Eo----<~~~----------oE
Eo------+------oE
(a) Forward active mode

(b) Reverse active mode
Figure 75 Simplified Ebers-Moll equations for forward and reverse active
modes.
This model can be modified to account for dynamic operations by including the
.i, familiar base-emitter and base-collector diffusion ( Cde , C de) and junction ( C Je , C jc)
diode capacitances. Unlike the simple charge analysis presented for the single-diode
. model, a more elaborate treatment is required for the BJT. For instance, the charge
accounting for the emitter diffusion capacitance is comprised of minority charges stored
: in (a) the neutral emitter zone, (b) the emitter-base, (c) the collector-base space charge
regions, and (d) the neutral base zone. An identical analysis applies to the collector diffu. sion capacitance. Figure 7-6 depicts the dynamic Ebers-Moll chip-modeL Further refine=
360
Ebers-Moll
Rcc
model
cdc
RsB'
~.
tIR
B'
cde
~(!
~IF
(a) Dynamic EbersMoll chip model
(b) RF model with parasitic terminal effects

Figure 7-6
Dynamic Ebers-Moll model and parasitic element refinements.
~---------------------------------------------------------------
,~~
ments for RF operations are often made by including the restsuve and inductive
properties of the lead wires as well as parasitic capacitances between the terminal points~
see Figure 7 -6(b).
----------------------------~&uM~
Example 72: Transport versus injection form of the EbersMoll large..signal model
Instead of the injection model, it is the transport model that typically
finds use in SPICE simulations. Let us go through the qualitative
steps to arrive at this important representation.
Solution:
We begin our discussion with the static BIT model,

since the diffusion and junction capacitances can be added later in
the derivation. First, we can show that the injection model Figure
7-4 is equivalent to the transport model in Figure 7-7.
c
~fc
ffEc/O.R
~Iccla.F
tJE
E
Figure 7-7 Transport representation of static Ebers-Moll injection model.
The equivalence of both models can be established if we reexpress collector and emitter currents as follows:
fc
= fcc-IEclaR
IE= -fcclaF+IEc
with the current controlled sources now given as
362
VsciVT
I EC = Is( e
- 1)
A slightly different form can be obtain if both current sources are
combined to a single source I com = Icc - I EC and the diode currents are re-expressed as
1EC
1 - CtR
---7
(lR
CtR
Icc
I - aF
I EC
lEe=~R
Icc
Icc=CtF
~F
This model configuration is shown in Figure 7-8 with base, collector, and emitter resistances. Also shown in Figure 7-8 are the combined diffusion and junction capacitances C be and C be associated
with the base-emitter and base-collector diodes.
---7
O.F
Rcc
cbc
RaB'
tIci~R
~
cbe
/corn
!IcciPF
RF'
Figure 7-8
Dynamic Ebers-Moll transport model with single current source.
The Figure 7-8 configuration is important since it leads

directly to the large-signal BJT model under forward active mode
Translator Models
363
condition. This mode allows us to neglect the base-collector diode

current, but not its capacitative effect. Renaming the electric parameters, we arrive at the circuit depicted in Figure 7-9, where we
replaced the forward-biased diode with an equivalent current source.
B'
c~
cbe
Eo-----------+----oE
Figure 7-9
Large-signal BJT model in forward active mode.
This final form can be found in the SPICE library as a nonlinear representation of the standard BJT.
We notice how the dynamic transport model of the Ebers-Moll

equations naturally lead to the SPICE large-signal model. An inherent difficulty for all circuit models is the unique determination of the
model parameters through appropriate measurement strategies.
The Ebers-Moll model was one of the first BJT circuit representations and has
retained its popularity and wide acceptance. However, shortly after its introduction, it
became apparent that a number of physical phenomena could not be taken into account
by this original model. Specifically, research has shown that (1) j3F and j3R are current
dependent, and (2) the saturation current Is is affected by the base-collector voltage
1
'"' (Early effect). Both effects significantly influence the overall BIT performance. For
, this reason a number of refinements have been introduced to the original Ebers-Moll
model, culminating in the Gurnmel-Poon model shown in Figure 7-10.
In this model we immediately notice the addition of two extra diodes to deal with
the collector-dependent forward and reverse current gains ~F(/ c) and ~R(/ c). Figure
7-11 depicts a typical curve for ~ F .
364
Rcc
REE.
Figure 7-10 Static Gummei-Poon model.

1000r---~--~--~--~--~--~~--~~
BFG403W
fCE = 3 V
~
100
=
~
(.)
10
1~--~--~--~--~--~--~~--~~
w-s
10-4
w-J
1o-2
w-'
10
10 1
I0 2
10 3
Collector current Ic, rnA
Figure 711
Typical dependency of ~ F on the collector current I c for a fixed

collector-emitter voltage V CE.
The two leakage diodes Ll, L2 provide four new design parameters: coefficients
ls 1 , nEL in IL 1
ls 1(exp[V8 EI(nELVT)]-l) forlow-currentnormalmodeoperation, and I s 2 , ncL in I L2 = I s 2( exp [ V Bel (ncL V T)]- 1) for low-current inverse
mode operation. Additionally, the Gummel-Poon model can handle the Early effect,
Transistor Models
365
whereby with increasing collector-emitter voltage the space charge domain begins to
extend far into the base region. The result is an increase in collector current for a fixed
base current. If one draws tangents to each collector current curve (see Figure 7 -12), they
all converge approximately at a single voltage point -VAN known as the forward Early
voltage. An identical analysis can be conducted if the BJT is operated in the reverse
active mode, resulting in a voltage point V BN known as the inverse Early voltage.
~~
.. " . .
..
.... .
..\'~-:-:-:::~-:-~-: _-:-:-::~: ~:::~.

..
-~N
Figure 7-12
Collector current dependence on cEand its approximation through

the Early voltage VAN
Both voltages are incorporated as additional factors in the model. Moreover,

Gummel-Poon also permits the specification of a current-dependent base resistance and
a distributed base-collector junction capacitance C jbc . We will not go into any details
of the various underlying physical reasonings leading to the requirement of these additional model parameters. The interested reader is referred to the sources listed at the
end of this chapter. Converting the static Gummel-Poon model (Figure 7-10) into
dynamic form by including the diode capacitances and C jbc leads to the equivalent circuit shown in Figure 7-13.
This circuit is similar to the large-signal Ebers-Moll form (Figure 7-9) but with
the differences that the base resistance R88 , is current dependent, the collector current
takes into account the Early effect, and a distributed base-collector junction capacitance
Cjbc enters the model.
In SPICE both BJT models can be invoked, with Ebers-Moll requiring the specification of 26 circuit parameters and up to 41 parameters for Gummel-Poon. Generally,
the BJT manufactures supply these parameters in their data sheets. Unfortunately, one
increasingly encounters the situation where instead of the generally applicable SPICE
model parameters, only the measured S-pararneters are given. Since these measure
ments are recorded for particular operating frequencies and under certain bias conditions, it is then left to the circuit design engineer to interpolate the data for a particular
transistor operation not found in the data sheet.
366
B'
cbe
Eo---------_..----oE
Figure 713
7.2.2
Large-signal Gummei-Poon model in normal active mode.
Small-Signal BJT Models
From the large-signal Ebers-Moll equations it is now easy to derive a small-signal

model in the normal active mode. To this end, the large signal model (Figure 7-9) is
converted into the linear hybrid-1t model shown in Figure 7-14.
rll
rn
B
ell
+
rc
vn gmvn
Eo-----------+--------oE
Figure 7-14
Small-signal hybrid-1t Ebers-Moll BJT model.
We see that the base-emitter diode is replaced by a small-signal diode model and
the collector current source is substituted by a voltage-controlled current source. To
make the model more realistic, a resistor r ~ is connected in shunt to the feedback
capacitor C ~. For this model we can directly establish the small-circuit parameters by
expanding the input voltage V BE and output current I c about the biasing or Q-point in
terms of small AC voltage vhe and current ic as follows:
Transistor Models
367
(7.15a)
Ic
= I~+ i c = I sexp[( V~E + vb,)/VT I = /~[ 1 + (~;) + ~(~J + ... ]
(7.15b)
Truncating the series expansion of the exponential expression after the linear term, we
find for the small-signal collector current
ic
(~!}be =
(7.16)
8mVbe
where we identify the transconductance

_
gm -
dl c
= d
(V HF.IV 7 )! ::::: I c
dV
d V I se
Q V
BE Q
BE
T
(7.17)
and the small-signal current gain at the operating point
~F ~ Q
die
= df
BQ
= ~0
(7 .18)
The input resistance is determined through the chain rule:

dVBE
rTt:;;;;
dl
~0
:;;;;
(7.19)
B Q
For the output conductance we have

1
'o
dVcE
=
Q
d
dVcE
Q
=- C
V
( 1 e VRf:fVr[ 1 +CE])
5
VAN
(7.20)
VAN
which includes the Early effect, also known as the base-width modulation because of
the increased depletion layer extent into the base.
It is directly seen that this model in its simplest form at the terminals B'-C'-E'
reduces for the static case and, under negligence of the collector-emitter resistance, to
our familiar low-frequency transistor model. Here the output current can simply be
expressed in term of the input voltage vb e as
(7.21)
368
Often additional small-signal BIT circuit models can be developed on the basis of
the h-parameter network representation. For instance, if we recall the definitions of the hparameters and apply them to a BIT in common-emitter configuration, we obtain
(7.22)
ic
= h21ib + h22Vce
(7.23)
which is encoded in generic form in Figure 7-15.
Figure 7-15
Generic h-parameter BJT representation with two sources.
In this notation the indices denote 11 ~ input , 21 ~ reverse , 21 ~ forward , and

22 ~ output . The individual parameter can be computed via the following relations:
hu
h21
h12
h22
vbe
- -
lb
input impedance
vee= 0
lc
lb
(7.24a}
foward current gain ~ F
(7.24b)
reverse voltage gain
(7.24c)
output admittance
(7.24d)
vee= 0
vbe
vce
tb
=0
lc
vee
ib = 0
It is observed that h 12 represents the influence of the output voltage "fed back" to the
input as part of a voltage-controlled voltage source. Conversely, h 21 models the influence of the input "fed forward" to the output, or gain, as part of a current-controlled
current source. The output to input feedback is modeled by the reverse biased collectorbase junction capacitance C cb, which is generally on the order of 0.1 to 0.5 pF and a
resistor r cb, with values ranging in the low MO. Therefore, for low and intermediary
frequencies up to approximately 50 MHz, this feedback can safely be neglected. However, in the GHz range, it may profoundly affect the BIT operation.
389
Tl'lnslator Models
If the feedback resistor r be is neglected, a high-frequency circuit model results, as

displayed in Figure 7-16. Also shown in this figure is a converted circuit such that the
feedback capacitance C cb appears as the Miller capacitance on the input and output
sides. The Miller effect allows us to decouple the input from the output port by redistributing the feedback capacitance, as the following example shows.
(a) RF circuit model
Figure 7-16
(b) Equivalent circuit model

RF small-signal circuit model and converted circuit model using the
Miller effect.
------------------------------~~~
Example 7-3: The Miller effect
Show that the feedback capacitance Ccb can be expressed as
CMI = ccb(l-vcelvbe)
on the input port and
as
C M 2 = C cb ( 1 - vbel vce ) on the output port. Assume that the input
and output voltages are approximately constant, and keep in mind
that vce is negative under common emitter configuration.
Solution:
We need to convince ourselves that the two generic
circuits shown in Figure 7-17 are equivalent.
The current I P is found by taking the voltage difference
between output and input divided by the feedback impedance
IP = (V 1 - V2 )/Z 12
and for the equivalent input and output impedances
z 11 , z22
370
Chapter 7 Active RF Component Modellnt
lr
I,
12
(a) Circuit with feedback impedance Z 12
(b) Equivalent form
Figure 717
Miffer transformation of feedback impedance.
vl
zll = lp -
Z 12 V 1
(V
1-
2)
= ZI2(1- V2/Vt)
-1
and
Z22
v2
= (-lp) -
Z12v2
-t
(V
V
= 212(1- Vl/V2)
2- 1)
With the assignments Z 12 = ll(jroCcb), Z 11 = l/(jroCM 1),

Z 22 = I I (jooC M 2 ) and V 1 = v be, V 2 = vce, we find the equivalent capacitances
CM1 -
cbe(l- vcelvbe)
(7.25)
and
(7.26)
Decoupling of the input from the output port is accomplished
by computing an equivalent capacitance that depends on a constant
voltage amplification factor vclvbe
Another important factor that is directly related to the BJT frequency behavior is
the short-circuit current gain hfe( ro), which implies the connection of the collector
with the emitter as depicted in Figure 7-18.
Ttanalltor Models
371
(a) Short-circuit hybrid-1t model
lhft (f)l
~ol-----
1 ... -~
........ .
.
'
fo
(b) Amplification versus frequency behavior

Figure 7-18 Short-circuit current gain of BJT model.
Since the output is short circuited and thus vce = 0, the Miller effect does not
enter the analysis. We find h fe ( oo) by computing the ratio of collector to base currents
ic
hfe(ro) = :lb
where
Zin
gmZin( 1- jroC}J/ gm)
= rttl( l + jror1tC1t). Substituting Zin

h fe ( (J))
= 1 +0 J<Orrc(Crc
.
=
+ CJ..L)
~ ( 1- jroCJ..ll gm)
with the maximum frequency
f0
f0
(7.27)
1 + JffiCJ.!Zin
into (7.27) and using (7.19) results in

~ 0 [ 1- j(f If 0 )]
1 + j(fI frl)
and the beta cut-off frequency
1
g m and f 21tCJ..L
rl - 21tr rc< ere+ Cll)
(7 .28)
f IS
(7 29)
.
The transition frequency f T denotes the point where the magnitude of the current gain
is unity (or 0 dB) under short-circuit output condition. Setting the absolute value of
(7.28) equal unity, we find
372
fT
1
= 21t
~~- 1
2
r1t(C1t
(7.30)
+ 2C1tCJ.l)
Since usually (}0 1 and C 1t C J.l we can rewrite (7 .30) as
~0
gm
1T ~ 21tr1tc1t = 2nCn
(7.31)
As already seen in Chapter 6, this frequency is related to the emitter-collector time

delay, which is composed of the delays associated with base, emitter, and collector.
Another name for f T is the gain-bandwidth product, which is specified in data sheets
for a particular collector-emitter voltage and collector current bias condition. Addi2
tional figures of merit can be established when one considers IS21 1 , where the power
gain of the transistor is recorded under zero source and load reflection coefficients. This
condition will be investigated in greater detail in Chapter 9.
Finally, let us discuss a design project involving the BJT. In this project, we will
go through the steps of deciding upon bias conditions, determining the input and output
impedances as a function of frequency, and converting the impedance values to the relevant S-parameters. The transistor parameters used for this example are summarized in
Table 7-3. The MATLAB routine ex7_4.m provides computational details.
-----------------------------~~~
Example 7-4: Setting bias conditions, determining input/out
put impedances, and computing the S-parameters for a BJT
Our task is to design an amplifier for a portable communication system. The system is supposed to operate from a 3.6 V battery source.
Taking into considerations the maximum available current and battery lifetime, we demand that the current for the amplifier should not
exceed approximately 10 rnA. Assuming V CE = 2 V and
I c = l 0 rnA as bias conditions for this transistor, and the BIT
parameters given in Table 7-3, we need to detennine the hybrid-1t
model. In addition, the resulting input/output impedances and the
corresponding S-parameters for the frequency range of
1 MHz </ < 100 GHz have to be found.
Transistor Models
373
Table 7-3
Parameters of the BJT transistor

Description
Symbol
Typical value
~F
forward current gain
145
Is
saruration current
5.5 fA
VAN
forward Early voltage
30V
't;
forward transition time
4 ps
c,co
base-collector )unction capacitance at zero applied junction voltage
16 fF
c lEO
base-emitter junction capacitance at zero applied junction voltage
37 fF
me
collector capacitance grading coefficient
0.2
mE
emitter capacitance grading coefficient
0.35
VdiffBE
base-emitter diffusion potential
0.9V
v ditr
base-collector diffusion potential
0.6V
rB
base body resistance
125 .Q
'c
collector body resistance
15 .Q
rE
emitter body resistance
1.50
Ls
base lead inductance
l.lnH
Lc
collector lead inductance
1.1 nH
LE
emitter lead inductance
0.5 nH
BC
Solution:
We begin this design by developing a standard voltage

divider biasing network, as shown in Figure 7-19.
With the power supply voltage of V cc = 3.6 V, desired collector-emitter voltage of V CE = 2 V, and collector current of
I c = 10 rnA, we can find a value for the collector resistor Rc as
follows:
V cc - V CE
lc
I60 n
374
RF;n
Figure 7-19
Biasing a BJT in common-emitter configuration.
Based on the current gain of ~ = 145 and collector current of

I c = 10 mA ,
we
find
the
base
current
to
be
I 8 = I cl ~ = 69 JlA . The current through the resistor R 81 is
equal to the sum of the current flowing through resistor R82 and I 8 .
In practice, the values of R 8 1 and R 82 are selected such that they
make the magnitude of I 8 equal to 10% of the current through resistor R82 Keeping this in mind and realizing that the base-emitter
voltage drop V8 E is approximately equal to the base-emitter built-in
potential V diffBE , we find
and
RBt
= v cc- vdiffsE = 3560 n

lliB
Now we are ready to compute the hybrid-7t model parameters.

From equations (7.17H7.20) we obtain gm = I ciVT = 386 mS,
r'lt = ~olgm = 375 Q, and ro = VANIIc = 3 kO. To find ell
and
we have to resort to the pn-junction analysis. Since the
base-collector voltage is negative, the base-collector capacitance is
only detennined by the junction capacitance. From (7.3) we find
en
ell =
CJCO
( 1- V BciVdiffsc)mc
13 tF
Tl'lnslstor MOdels
375
Because the base-emitter voltage is positivet Cn is a combination of

both the junction and diffusion capacitances. From (7.3), and by
assuming V m E = 0.5 V diffBE , we have
c
ltJ.U!lCI
= CJ!O (1 + mEVBE-~-~ vdiffB)
0.5
55 fF
and
1tctiff
v BE;v r
= I-Vs'tr
e
T
- 1.085 pF
Thust the total base-emitter capacitance is
Cn
= Crt.
JUIK:I
+ Cn: d1fT
.
= 1.14 pF
After establishing all parameters of the hybrid- 7t model, we

can compute the corresponding h-parameter matrix as described in
(7 .24). The result takes into account only the transistor die hybrid-7t
parameters without incorporating base, col1ector, and emitter resistances and parasitic inductances.
To consider the influence of the lead resistance and inductance,
we can employ a network analysis as described in Chapter 4. Specifically, we can partition the equivalent transistor circuit into four
two-p,ort networks, as shown in Figure 7-20.
Hybrid-1t model
Base
Collector
r------,
.----------------I Ir------,
I
Lo
rs I I
rc
Lc I
B
C
I
"""------~
------- ..
--------~
I
1 Emitter
Eo--------------------+-----_.-----,----------------------------oE
Figure 7-20
Complete transistor model divided into four two-port networks.
376
Relying on this network partitioning we proceed as follows: To

obtain the Z-parameters of the entire transistor we first convert the
h-parameters of the hybrid-1t model into ABCD representation.
Next, this converted hybrid-1t model is multiplied by the ABCDmatrix representations for base and collector leads.
The matrix equation is as follows:
ABl _ [1 r
[C Dj 0
tr
[A Bl
[1
+ jroL8l
r c + jroLcl
1
base C Dj h-model 0
1
collector
Finally, we convert the ABCD representation of the transistor with

the attached base and collector leads into Z-parameter form and add
the resulting matrix to the Z-matrix of the emitter lead.
Z 11
[Zzt
Z 12~
z22
trans
[Z 11
Z 12~
z21 z22
jroLE~
[rE + jroLE rE +
+ rE + jroLE rE + jroLE
tr
enutter
The frequency responses of coefficients Z 11 and Zz2 are shown in

Figure 7-21.
As we see from Figure 7-21(a), the addition of the lead impedance to the basic hybrid-1t model at low frequencies results in a significant increase in the input impedance due to the large base
resistance. At high frequencies the effect of base and emitter inductances become noticeable in terms of a sharp rise in the impedance.
For the output impedance the situation is quite different. Since
the base resistance does not have any effect on Z 22 , the output
impedance remains virtually unaffected by the addition of the leads
and is dominated by the resistance r 0 up to very high frequencies.
At that point the inductive effect of the leads become dominant.
From the known Z-parameter representation of the transistor
we can easily compute the S-parameters using the conversion
described in Chapter 4. The resulting input reflection coefficient,
S 11 , and gain, S 21 , of the transistor are shown in Figure 7-22 as part
of the Smith Chart and a polar plot, respectively.
As we notice in Figure 7-22(b), even though the emitter resistance and inductance seem to be negligible compared to the values of
the other components in the model, their addition results in a significant drop in gain over the entire frequency range. This shows once
again the influence of parasitic elements in RF circuits.
Transistor Models
600~----~------~----~------~----~
500 r---~
Transistor with leads

on all three ports
N"
400
Transistor die with

base and collector
leads only
11)
~
300
Q
."
s
200
100
Transistor die
(hybrid-1t model)
Frequency, Hz
(a) Input impedance of the transistor

3.5~----------------------------------~
a 3.o
~
2.5
2.0
1E 1.5
Q
Transistor with leads

on all three ports
.....
=
~
1.0
Transistor die
wi~
collecto~ ~
base and
leads only
0.5
Frequency, Hz
(b) Output impedance of the transistor

Figure 7-21
Input and output impedances as a function of frequency.
378
....o
Hybrid-1t model
with base and
90
collector leads
-1.0
(a) Input reflection coefficient, S11
Figure 722
170
(b) Gain of the transistor, 8 21
5 11 and 5 21 responses of a BJT for various model configurations.
We have demonstrated an approach of computing the smallsignal parameters of the transistor from known operating conditions
of the underlying SPICE model. Even though a simple topology is
investigated, this method can be directly applied to more complicated internal structures by breaking them down into a set of interconnected two-port networks.
7 .2.3
Large-Signal FET Models
FETs offer a number of advantages but also suffer some disadvantages over BJTs.
In choosing the appropriate active device for a particular circuit, one should take into
consideration the following FET-related benefits:
FETs exhibit a better temperature behavior.
The noise performance of a FET is, in general, superior.
The input impedance of FETs is normally very high, making them ideal for
preamplification stages.
The drain current of a FET shows a quadratic (and thus a more linear) functional
behavior compared with the exponential collector current curve of a BJT.
The upper frequency limit exceeds, often by a substantial margin, that of a BJT.
The power consumption of a FET is smaller.
In terms of the disadvantages one often hears:
FETs generally possess smaller gains.
Tranalator Models
378
Because of the high input impedance, matching networks are more difficult to
construct.
The power handling capabilities tend to be inferior compared with BITs.
The preceding list is debatable, since new device concepts and fabrication improvements continuously affect various transistor perfonnance aspects.
For our FET modeling purposes we will focus on the noninsulated gate FET. To
this group we count the MESFET, often identified as GaAs FET (pronounced "gasfet"),
and the HEMT. Both types are discussed in Chapter 6. In Figure 7-23 the basic n-channel, depletion mode MESFET model (with negative threshold voltage) is shown along
with the transfer and output characteristics.
/Dsat
s
0
(a) FET symbol
(b) Transfer characteristic
'Vas- Vro
~----Vos=O
0
(c) MESFET model
Figure 7-23
(d) Output characteristic

Static n-channe1 MESFET model.
The key equations for the drain current in forward, or normal, mode of operation
follow from the analysis developed in Section 6.4. There we obtained the drain current
for both the linear and saturation regions. These current expressions constitute the starting point of deriving the model for the FET.
Saturation region ( V vs ~ V cs - V TO > 0)
The saturation drain current given by (6.94) is repeated here for convenience
(7.32)
380
If we substitute in (7 .32) the combination of threshold voltage V ro and pinch-off voltage V P (in other words, V d = V TO + V P) an alternate form is obtained:
I Dsat
V P{
GoT
V GS- V TO)
1- 3 1V
+ 2 [ 1 - V GSV V TOJ
(
312
(7.33)
Making a binomial expansion of the square bracketed expression up to the second term
allows us to write (7 .33) as
(7.34)
The constant factors in front of the square term in (7 .34) are combined to the conduc
tion parameter
~n
= !(Go) = J.lnEZ
J3
n
4 Vp
2Ld
(7.35)
where the definitions for the conductance G 0 = aZdl L = JlnNvqZd/ L and the
2
pinch-off voltage V P = (qNvd )1(2e) from Section 6.4 have been used. If the channel modulation effect is included, we arrive at
(7.36)
1
Here the parameter A. == 0.0 I ... 0.1 v- models the slight increase in drain current for
increasing drain-source voltage in the saturation region, see Figure 7-23(d).
Linear region (0 < V vs < V GS- V ro)

Identical steps, as outlined for the saturation region, can be invoked to manipulate
the drain current expression (6.91) to yield
(7.37)
where again the channel modulation is considered to achieve a smooth transition from
the linear into the saturation region. For instance, if V vs = V GS- V TO (that is, the
transition from linear to saturation region) both drain currents are identical.
The FET can also be operated in reverse or inverted mode if Vvs < 0 . For completeness, the two drain current relations are given without further comments.
Reverse saturation region (- V DS ~ V GD - V ro > 0)

(7.38)
Transistor Modela
381
Reverse linear region ( 0 < - V DS < V GD - V TO )

2
ID = ~n[2(VGD- Vro)VDs- VDs](l +AVDs)
(7.39)
Under reverse operation, the gate-drain diode ( GD) is negatively biased.

Making the transition from the static to the dynamic FET model requires only the
addition of gate-drain and gate-source capacitances, as illustrated in Figure 7-24. Also
shown in this model are source and drain resistors associated with source-gate and
drain-gate channel resistances. A gate resistor is typically not included because the gate
current, although substantially higher than for a MOSFET, is still negligible.
Figure 7-24
Dynamic FET model.
A summary of the most relevant SPICE modeling parameters for a MESFET is
presented in Table 7-4.

Table 7..4
Symbol
SPICE modeling parameters for a MESFET

SPICE
Description
Vro
VTO
Threshold voltage
/...
LAMBDA
Channel-length modulation coefficient
BETA
Conduction parameter
Cov
CGD
Zero-bias gate-to-drain capacitance
Cos
cas
Zero-bias gate-to-source capacitance
'v
RD
Drain resistance
's
RS
Source resistance
382
7.2.4
Small-Signal FET Models
A small-signal FET circuit can directly be derived from the large-signal FET
model (Figure 7-24). In this model we simply replace the gate-drain and the gatesource diodes by their small-signal representations derived in Section 7 .1. In addition,
the voltage-controlled current source is modeled via a transconductance g m and a shunt
conductance g 0 = 1I r ds . The model can be tied in with a physical device correspondence, as Figure 7-25 shows.
(a) Idealized MESFET device structure

D
cds
s
(b) Circuit model
Figure 7-25 Small-signal MESFET model.
This model can be described by a two-port Y-parameter network in the form

ig
Y11Vgs
+ Y12Vds
(7.40a)
id
Y21Vgs+Y22Vds
(7.40b)
Transistor Models
383
Under realistic conditions, the input conductance of y 11 and the feedback conductance of y 12 are very small and can thus be neglected. This is consistent with the fact
that the gate current is too small to be of practical consequence. However, for high-frequency operations the capacticances are typically included, resulting in the circuit
model shown in Figure 7-26.
s
Figure 726
High-frequency FET model.
For DC and low-frequency operation, the model in Figure 7-26 simplifies to the
condition where the input is completely decoupled from the output. Transconductance
gm and output conductance g 0 can be readily computed for the forward saturation
region from the drain current equation (7 .36):
Y21
= gm
diD
=dV
= 2~"(VGS- V roHl +A V Ds )
(7.41)
GS Q
Y22
= r ds -
diD
dV
DS Q
= ~"A.(Vcs-Vro)
vgs
,.~
r
...
~
(7.42)
vgs.
with the operating point, or Q-point, denoted by

and
The gate-source and gate-drain capacitances play a crucial role in determining the
frequency perfonnance. For the transition frequency f r we again have to consider the
short-circuit current gain for the situation where the magnitude of the input current I c
is equal to the magnitude of the output current I D , or specifically
'.
IIcl
= COr(Cgs + cgd) !V
csl = IIvi
= gmiV csl
(7.43)
which gives us
(7.44)
! For low-frequency FET applications, it is primarily the charging time defined by these
~
~
r
i.;
capacitances that severely limits the FET frequency response. This is in contrast to the
384
channel transition time, as defined in Section 6.4.3, which for high-frequency applications limits the FET's operation as the following example shows.
-----------------------------~&JM~
Example 7-5: Approximate determination of cut-off frequency of a GaAs MESFET
A GaAs MESFET with a gold gate is fabricated to be 1.0 J.lm in
length and 200 11m in width, and d
0.5 ~J,m in depth. The following
electric characteristics
are
known:
, == 13.1,
16
2
3
N D = 10 cm- , and Jln = 8, 500 cm /Vs. Under suitably chosen approximations, we would like to find the cut-off frequency at
room temperature.
Solution:
To apply (7 .44), it is necessary to find an approximate

expression for the transconductance and capacitances. The transconductance can be found by knowing that the drain saturation current
(7.33) is maximum for V GS = 0, which gives
gm
===
dfDsat
dV
GS v GS
= GoO -...; V d/V p)

=0
where the built-in voltage Vd for the Schottky contact is found from
(6.39) to be
vd
with
X=
= ( vM- X)- v c
V c:::: Vrln(Nc!N 0 ) = O.l V,
VM = 5.1 V,
and
4.07 V. Substituting these values yields V d = 0.93 V . The
pinch-off voltage and the conductance are, respectively,
VP
qNDd
20 ,
= 1.74 V
and G 0 =
q~nNDWd
= 34 mS
Thus, g m ~ 9.1 mS . For the capacitance we can approximately compute the surface area of the channel times the dielectric constant
divided by the channel thickness:
Measurement of Active Devices
385
From these values we can finally estimate
f r = 2 ( C gm C ) =
1t
gs + gd
f r to be
31.5 GHz
In contrast to an approximate channel transit time of 15 GHz

discussed in Section 6.4.3, we now have the situation that the RC
time constant is smaller. In other words, the channel transition time
becomes the limiting factor in the high-speed peiformance of this
MESFET.
An often used approximate formula for (7.44) can be derived if we set gm = G 0 .

The explicit result is
fr=
qJlnNvd
21tr.L
(7.45)
This expression applied to the above example would have yielded 29.3 GHz, a value
very close to the computed frequency of 31.5 GHz.
7.3 Measurement of Active Devices

7.3.1
DC Characterization of Bipolar Transistor
We commence our analysis with the Ebers-Moll equations (7.10) and (7.11), reexpressed as collector and base currents:
(7.46a)
Is VsEIVr
Is VsciV r
18 = ~F(e
-l)+~R(e
-1)
'
!
t
!
iI'
1
t
(7.46b)
The unknown coefficients to be detennined through measurements are Is, ~ R , and ~ F

In addition, forward and reverse Early voltages VAN and V BN become important when
the BIT is operated for large VCE To separate forward and reverse current gain measurements, we resort to two measurement protocols, shown in Figure 7-27.
386
(a) Forward measurements

(b) Reverse measurements
Figure 7-27 Forward and reverse measurements to determine Ebers-Moll BJT
model parameters.
Under the forward measurement condition, the base-collector is short circuited

( V Be == 0 ), simplifying (7 .46) to
(7.47a)
(7.47b)'
Monitoring the base and collector currents as a function of V BE results in the graph
Voc
=0
lnls /
VB
Figure 7-28
lc and 18 versus VaE
Both currents are logarithmically plotted and shown for sufficiently large VBE
values, where the exponential terms dominates over the factor 1. A linear slope of
1I V T for both currents is obtained, since
387
(7.48a)
VB
lnlB = lni5 -ln~F + V
(7.48b)
From these two curves we can first extrapolate the collector current to get lnis and
thus Is . Extrapolating the base current next yields a value for In Is - In~F, from which
we can determine ~F. From Figure 7-28 it is apparent that the current gain is constant
only over a very narrow collector-emitter voltage domain. For low and high current
injections significant deviations occur. The Early effect is expressed as a linear gradient
of the collector current:
Ie
= Is( e vce1vr -
VeE)
VeE)
1) ( 1 + - ::::: I se vc.E1v r ( 1 + VAN
VAN
(7.49)
This allows us to find VAN by projecting the tangent, applied to the collector current in
the saturation region, to the intercept point with the VeE -axis in the second quadrant.
The intercept point is the same for various base currents, as shown in Figure 7-12. The
determination of the reverse mode parameters ~ R , V BN is carried out by interchanging
the collector with the emitter terminal [see Figure 7-27(b)], and then following the
identical procedure as done in the forward direction.
7.3.2
Measurements of AC Parameters of Bipolar Transistors
The determination of the AC parameters is more of a challenge depending on the

model involved and the details required. To extract analytically the large-signal EhersMoll or Gummel-Poon circuit elements is an actively pursued research endeavor. For
our purposes we concentrate on the small-signal, low-frequency circuit model shown in
Figure 7-29.
Figure 7-29 Small-signal, low-frequency h-parameter representation.
This model is related to the hybrid-1t model presented in Figure 7-14, but without
the output feedback ( h 12 = 0) and ohmic contributions r B = r E = r c :::: 0. For a Qpoint in the active forward region, and consistent with (7.15)-(7.20), we can derive the
following parameters:
Chapter 7 Active RF Component Moc~Ming
388
Transconductance
gm
die
IQc
::;;:
dVBE
(7.50a)
VT
VeE = 0
Input capacitance
en
'tbey
IQ
Is
T
V 8 1 V r
'they
(7.50b)
Input resistance
r 1t =
dVBE
d/B
vbe
lb
V Cf
V eE
=0
~0
gm
(7.50c)
Output conductance
IQ
die
1
-=
'o
dVcE
=
Q
(7.50d)
VAN
VB E
where it is understood that the collector current in the presence of the Early effect is
given by I c = gml B( 1 + V cEIV AN) . Furthermore, since we decided to operate in the
forward active mode, c1t denotes the diffusion capacitance, with the forward transit
time 'tbe of the base-emitter diode.
The parameter extraction for this simplified hybrid-1t model begins with the process of setting the desired Q-point, resulting in known /~, I~, and VAN. Thus, a measurement protocol would sequence through the following steps:
Transconductance g m
DC current gain ~ 0
= I~/ v T
for a given junction temperature
= I~/ I~
Input resistance r n = ~ 0 1 gm
Output resistance r 0
= V ANI I~
1
Input impedance Zin = ( 1I r n + jroCnf recorded at a particular angular frequency and then solved for the capacitance Cit
Instead of recording the input impedance and indirectly detennining C n, we can more
elegantly find the transition frequency and thus C1t. This is accomplished by noting that
the AC current gain at the transition frequency f r is unity:
389
licl _
libl Knowing that
Po 1 leads to f T
=z
Po
1 + jwrr1tc1t
=1
(7.51)
~ 0 1 (21tCrtr rt), from which it follows that
=z
~0
(7.52)
21tfrrrt
7t
This approach can be implemented quite easily with a network analyzer. Sweeping the
frequency until the base current is equal to the collector current would allow us to enforce
(7.51). The resulting transition frequency can then be substituted into (7.52) to find C1t.
---------------------------RF&uM~
Example 7-6: Small-signal hybrid-1t parameter extraction
without Miller effect
An npn-transistor is operated under DC bias of
Ii :
40 JlA,
and the
Early
voltage is
/g = 6 rnA,
recorded to be
= 30 V . Through a network analyzer measurement the transition frequency is determined to be f r = 37 GHz at room temperature. It is required to determine the hybrid-1t parameters: Po , r
VAN
1t ,
C1t , r0 , and gm .
Solution:
Neglecting feedback from the output to the input, we

can use the preceding equations directly and find
gm
lg
6xl0-3 A
= V T = 0.026 V
= 232 mS
The forward DC current gain Po of the transistor can be found simply as a ratio of the collector current to the base current:
Po = 1g11~ = 150
From the known Po and transconductance g m we find the input
resistance as r 1t = Pol gm = 647 Q. The output resistance is a
ratio of the forward Early voltage to the collector current
r 0 = V ANI
= 5 kQ. Finally, the capacitance is found from
(7.52):
lf:
390
- 1.00 pF
The small-signal parameter determination is almost a cookbook design process. However, the constant forward current gain
may not always reflect a realistic transistor behavior.
While Example 7-6 is applicable for low- and medium-range frequencies, the situation becomes more complicated for values approaching 1 GHz and beyond. Here we
cannot neglect the Miller effect, and our attempt must be directed toward finding a
strategy to obtain C ~ . As discussed in Chapter 4, electric measurements at high frequencies cannot rely on impedance, admittance, or h-parameter determinations because
of the difficulties associated with enforcing short- and open-circuit conditions. At these
frequencies we must resort to S-parameter measurements. How the S-parameters can be
utilized to find the feedback capacitor ell is explained in the following example.
------------------------------RF~~
Example 7-7: Small signal hybrid-1t parameter extraction
with Miller effect included
We re-examine the previous example, but this time use the network
analyzer to record the following S-parameters based on the characteristic impedance of 50 Q at 500 MHz:
[S] = [0.74e-i207 o 0.006ei82 5

9.78ei 1572 o 0.97 e-J7.8o
Our goal is to find the feedback capacitance Cll. In addition, we

would like to observe how the input and output impedances are
affected if C ~ is excluded.
Solution:
Since the DC measurements do not change, we will

not repeat them. For givenS-parameters we can easily compute the
Musurement of Active Devices
381
input impedance of the transistor using matrix transformations

described in Chapter 4:
c1 + s11 )( 1 - S22 ) + s12S21

.
Zin = Z 0 (1 S )(1 S ) S S = Rin + JXin = (77.5- j24.4) 0
-
11
22 -
12 21
Setting the input impedance equal to the circuit model yields
R. + jX. =
m
l
llrtt+jro(C7t+CM 1)
where CMt is the Miller-transformed capacitance. Rearranging this

equation leads to the form
(t)
r 7t Rin
r rt
CM1 =- ---2::::
1
-C1t
r;:]f"
ro
,.J ' 1t .n. in
where the real part of the input impedance is used and ro = 2rcf is
the angular frequency at which the S-parameters are recorded. Explicitly, we find C M 1 = 1.42 pF - 1.00 pF = 0.42 pF . To compute the
actual feedback capacitance CJ..l, we can use (7.25), where the ratio of
collector-emitter to base-emitter voltage is equal to the h 12 parame= 7.22 fF .
ter. This yields finally c~ = CMl/(1 +
To compute the frequency behavior of the input and output
impedances we can first calculate the h-parameters of the transistor
as given by (7.24) and then convert them into Z-parameter representation. Both input and output impedances are plotted with and without the feedback ( Cll = 0) in Figure 7-30.
lht2P
5.5.----..----~--.....------,
700,....----....---~---r----~
c 600
~ 500
=0
5.0 1---====--------~--i
4.5
;:.4.0
~.... 3.5
~ 400
~ 3.0
~ 2.5
c.
.5 2.0
..
.t300
! 200
100
0~---~--~--~~=-~
6
10
Frequency, Hz
Figure 7-30
~ 1.5
::s 1.0
0 0.5
0.0 L - -_ _...___ _......__ __.___ _--J
8
106
107
10
Frequency, Hz
Input and output impedances with and without feedback.
Chapter 7 Active RF Component Mode"ng
392
This example underscores the importance to include the feedback effect once the frequency begins to exceeds 100 MHz.
Although the preceding examples are simple extraction cases, they convey an appreciation of how difficult a realistic situation can become if the entire SPICE parameter set
is attempted to be extracted. For the nonlinear large signal circuit models, this is a
research task with no clear solution methodology. Many manufacturers have therefore
resorted to S-pararoeter characterization alone. This approach greatly simplifies the BIT
characterization by utilizing an appropriate test fixture or jig and relying on a network
analyzer to measure the S-parameters at certain bias conditions and operation frequencies.
7 .3.3
Measurements of Field Effect Transistor Parameters
Because the GaAs MESFET has gained such prominence in many RF circuits, it
is important to take a closer look at its parameter extraction. Since the circuit model is
the same for the HEMT. we can treat both cases in paral1el. The fundamental equation
for the drain current in the linear region is derived in Chapter 6 and is repeated here for
convemence:
(7.53)
The only difference between MESFET and HEMT lies in the definition of the threshold
voltage V ro. Specifically, with the Schottky barrier voltage V d, pinch-off voltage VP'
and energy difference t,. W c between the conduction bands of the heterostructure in a
HEMT, we obtain the following two expressions:
Vro
Vro =
= Vd- Vp
(MESFET)
Vd-~Wclq-Vp
(HEMT)
(7.54a)
(7.54b)
For the saturation region, when V DS = Vas- V TO' (7 .53) becomes the quadratic
equation
(7.55)
Using (7.55) we can easily extract values for conduction parameter P and threshold
voltage V ro by plotting the square root of the drain current versus the applied gate
393
Scattering Parameter Device Characterization
source voltage V GS A measurement arrangement of a MESFET for obtaining V TO and

~ is shown in Figure 7-31.
Vro
(a) Measurement arrangement

Figure 731
Vos
(b) 10 versus VGs transfer characteristic
Generic measurement arrangement and transfer characteristics in

saturation region.
The threshold voltage is detennined indirectly by setting two different gate-source

voltages V GSI and V csz while maintaining a constant drain-source voltage
VDS = const ~ V GS- V TO so that the transistor is operated in the saturation region.
The result of these two measurements gives
J1;;; = J1j(VGS1- V ro)
(7.56a)
ji;_
(7.56b)
jJ3( V GS2- V TO)
Here we assume that the channel length modulation effect is negligible; therefore, the
measured current is close to the saturation drain current as given by (7 .55). Taking the
ratio of (7.56 a) to (7.56 b) and solving for V TO, we obtain
V TO
= VGSI- (jl;;;l h ) VGS2

1- jl;;;!
(7.57)
Next we substitute (7 .57) into (7 .56a) and solve this equation for ~. The extraction process can further be simplified if we choose I v 2 = 41Dl so that (7.57) becomes
VTO = 2 Vest - V cs2 Upon substituting this expression in (7 .56a), we see that
~ = I Dl /( V GS2- VGSI) 2
7.4 Scattering Parameter Device Characterization

The S-parameter measurement approach greatly simplifies the device-under-test
(DUT) characterization by utilizing an appropriate test fixture or jig and relying on a
vector voltmeter or network analyzer to record the frequency and bias dependent four
S-parameters.
394
Although nowdays a vector voltmeter is seldom used for recording the S-parameters, it nonetheless allows us to gain valuable insight into the basic measurement procedure that is also at the heart of a network analyzer. We will therefore investigate this
approach first. It is generically depicted in Figure 7-32 and requires an RF signal generator, two dual-directional couplers, transistor biasing networks, the actual transistor fixture, and calibration kit to create short-circuit and through-line conditions.
---~ugh
I
----v.--d- - - 1...1...1 Short

;I
~
11
,.....------.....,
I
I : ..
I]
:
f-..........1
:
:I
50-0
load
coupler
.....................
Source
-- - '
Figure 7-32
Recording of $-parameters with a vector voltmeter.
The function of a dual-directional coupler in Figure 7-32 is to isolate the incident from the reflected power wave. How this is accomplished can be explained with :
reference to Figure 7-33, where a cross-sectional view of a coaxial coupler is shown.
For incident power coming from the left through the main arm, two slots, spaced 'A/ 4 ~
apart, couple the energy into an auxiliary path labeled 4. The incident wave does not .
produce any coupling into direction 3, since there is a 180 phase delay between sig-
nals coming from slot B and slot A, essentially canceling the entire wave. However, a
reflected wave from the OUT will enter the coupler at port 2 coming from the right and .
subsequently couple out the wave energy through the auxiliary path labeled 3, cancel- .
ing any wave leaving port 4. Therefore, port 3 provides an output for the reflected
power, whereas port 4 records the incident power. The two figures of merit for a direc-
Sclttertng Parameter Device Characterization
395
tiona! coupler are the coupling factor cf and its directivity factor df The factor cf is
defined as
cf =
!Olog(;J
(7.58a)
and denotes the logarithmic ratio of the power in the main port, either 1 or 2 (i = 1,2),
over the power in the auxiliary port, either 3 or 4 (n = 3, 4 ). The directivity df
df
= !Olog(~:)
(7.58b)
specifies the ratio of the powers in the auxiliary arm for the condition of equal forward
and reverse power levels applied to the main ports 1 and 2. For high signal discrimination we expect to see a large directivity value.
Port 3
Port 4
Auxiliary
arm
Port2
Port 1
A/4
Figure 7..33
Cross-sectional view of directional coupler and signal path

adjustment.
The actual signal propagation paths are observed in Figure 7-32. Here the vector
voltmeter records with channels A and B the incident and reflected powers from the
input port of the active device. Taking the ratio of the voltage magnitudes yields 11 j.
For recording the phase angle it is important to obtain an appropriate phase reference.
For this reason, the DUT is removed and a short circuit is inserted for phase reference.
To ensure equal path length (i.e., from the signal source to channel A, and from the
short to channel B), a line stretcher is used to perform the necessary adjustment to
achieve a zero phase difference.
The same test setup can also be utilized to find the forward gain S 21 . Switching
channel B to the directional coupler situated on the output side of the DUT yields the
ratio between the output and input voltages or jS21 j. The phase adjustment now calls
IS
396
for replacing the DUT with a through section element and again equalizing the signal
paths with the line stretcher.
The remaining two S-parameters, S22 and S12, are measured by reversing the DUT
jig and exchanging the biasing networks. As Figure 7-32 implies, the S-parameter measurements depend on the setting of an appropriate bias or Q-point and the signal source
frequency. As a result, a wide range of parametric curves could be generated.
Instead of employing a vector voltmeter, a more common approach involves the
use of the network analyzer. This instrument is capable of processing magnitude and
phase of a single or dual-port RF network. A simplified block diagram highlighting the
functionality is shown in Figure 7-34.
~----------------------~
Input-output
block
Display
Network
analyzer
I
Pulse
Generator
_: ___ I
Frequency
sweeping
oscillator
S-parameter
test set
Power
divider
Figure 7-34
Block diagram of a network analyzer with S-parameter test set.
The advantage of a network analyzer lies in the fact that all the separate functional
units associated with the vector voltmeter based measurement procedures are incorporated into one single instrument for an entirely automated testing of the RF or MW
device. The operation is such that a sweeping RF generator applies the RF signal to the
directional couplers. In fmward direction, the reference channel R records the incident
power wave and channel A provides the S 11 parameter via directional coupler 1 (DC 1).
Summary
397
At the same time, parameter S21 is recorded via directional coupler 2 (DC2). Switching
to reverse direction, the reference channel R records the incident power launched into
port 2 of the device under test, while channel B records S22 and channel A then yields
S12 This arrangement allows electronic switching between calibration and testing conditions, pennitting the recording of the entire S-parameter set without changing the test
jig. An interfrequency mixing and amplification stage feeds the signal into an analog-todigital conversion unit and subsequently into a microcomputer and display system. The
computer provides the user with the computed S-parameters (in magnitude and phase)
as well as such postprocessed parameters as group delay, return and insertion losses,
voltage standing wave ratio, input and output impedances, and many additional features.
The computer system allows for the software compensation of many imperfections introduced by the test arrangement. As a case in point, we recall the recording of
the Sparameters in Section 4.4.7 via the through-reftect-line (TRL) technique. This is
only one of a number of calibration schemes proposed to compensate for the various
error sources introduced by the measurement process.
7.5 Summary
Electric circuit models for active devices form the backbone of most CAD software packages. These circuits range from simple linear models to very sophisticated
large-signal models. Specifically, a large-scale BIT SPICE model that takes into
account temperature influences can involve over 40 adjustable parameters whose determination is a daunting task.
In this chapter we reviewed the basic large-scale diode model that is used for
modeling both the conventional pn-junction diode and the Schottky diode. Junction and
diffusion capacitances and the temperature-dependent saturation current are the key
ingredients constituting this model. By identifying a bias or Q-point and considering
only small-signal responses, we arrive at the linear diode model with the differential
conductance and diffusion capacitance
_ 1 _ d/D
Gd---Rd
dVA
VQ
The diode model is utilized as the basic building block to develop the static largescale BJT model as originally proposed by Ebers and Moll. Issues such as forward
active and reverse active modes are explained by simplifying the basic Ebers Moll equations. Starting from the injection model, we converted the Ebers-Moll BJT equations to
the transport representation and subsequently to the large-scale BJT model in forward
active mode. Additional refinements and modifications of the Ebers-Moll model have
398
Chapter 7 Active RF Component Modellnt
resulted in the more sophisticated Gummel-Poon model, whose large-signal normal

active mode circuit is shown in Figure 7-13. For the small-signal representation the
hybrid-1t is a popular linearization of the large-scale Ebers-Moll representation. The
hybrid-1t parameters are computed for a given collector current operating point:
gm
= I~IVr, r1t = ~0 /gm~ J3FIQ = flo~ and
l/r0
= 1g1vAN
For high-frequency operations the capacitive coupling between input and output ports
significantly influence the transistor operation. By taking into account the so-called
Miller effect, the collector-base capacitance is transformed into input and output capacitances, thus permitting us again to separate the two ports. Since lead inductances and
resistances also influence the high-frequency performance, we go through a detailed
design project to investigate, among other topics, how the input and output impedances
are affected as the frequency increases.
Attention is next directed toward the FET circuit models, specifically the high-frequency relevant types of MESFET and HEMT. Saturation, linear, reverse saturation,
and reverse linear regions are defined in close relation with Chapter 6. Specifically, the
drain currents in the saturation region
ID
= ~n(Vcs-Vro)
(l+A.VDs)
and in the linear region

ID
= J3n[2(V GS- V ro)V DS -
V DS](l +A V DS)
form the basis of the static and dynamic circuit models. Of particular interest are the
small-signal low- and high-frequency FET models. The cut-off frequency allows us to
quantify the frequency limitations of the device. For low to medium frequencies it is the
charging time of the capacitors that detennines the frequency performance, whereas for
very high speed operations it is the channel transit time that becomes the limiting factor.
Finally, we discuss some of the electric parameters of the active devices. For the
DC characterization of the BJT we can primarily rely on the collector and base currents
as a function of base-emitter voltage. From these curves, the saturation current, current
gain, and Early voltage are obtained. Measurement of the AC parameters is more of a
challenge, and only the linear hybrid-1t model allows a cook-book approach as outlined
by equations (7 .50). The FET model characterization follows a similar path as outlined
for the DC BJT model and involves the recording of the drain-current versus gatesource voltage.
In many cases, both for BIT and FET, the S-parameter representation is the most
common way to characterize an active device for a given bias and operating frequency.
For this purpose either a vector voltmeter or network analyzer is used to record the
Further Reading
399
input/output power waves of the device under test. Measurements with the vector voltmeter require directional couplers, signal sources, switches, and a forward and reverse
measurement protocol. This is all automated by connecting an S-parameter test set to
the three channels of a network analyzer. The recording of S 11 , S 22 , S 21 , and S 12 for
particular bias conditions and operating frequencies generally provides sufficient information for the circuit designer to characterize the device.
Further Reading
P. Antognetti and G. Massobrio, Semiconductor Device Modeling with SPICE,

J. J. Ebers and J. L. Moll, "Large-Scale Behaviour of Junction Transistors," Proc. of
IRE, Vol. 42, pp. 1761-1778, December 1954.
H. K. Gummel and H. C. Poon, "An Integral Charge Control Model of Bipolar Transistors," Bell System Tech. Journal, Vol. 49, pp. 827-851, 1970.
T.-H. Hsu and C. P. Snapp, "Low-Noise Microwave Bipolar Transistor with Sub-HalfMicrometer Emitter Width," IEEE Trans. on Electron Devices, Vol. ED-25, No. 6, June
1978.
E. S. Yang, Microelectronic Devices, McGraw-Hill. NY, 1988.

Problems
7.1
A silicon p~ ~unction diode has the following parameters at T = ~00K ,

1
Is = 5x 10 A, n = 1.2 , 'tr = 100 ps , and Rs = 10 .Q . Assuffilng that
the diode is operated under such biasing conditions that the applied junction
voltage is maintained at 0.7 V, find the differential resistance and the diode
capacitance for temperatures ranging from 200 to 450K.
7.2
The reverse saturation current of apn-diode is Is= 0.01 pA at 1j =25C and

has an emission coefficient of 1.6. For a junction temperature of 120C find
the reverse saturation current and the diode current I D at an applied diode
voltage of VA == 0.8 V.
7.3
The task for a process engineer is to obtain the model parameters for a
Schottky diode. From measurements it is determined that the saturation current is equal to Is = 2 pA. To obtain the remaining parameters (n and 'tr)
the engineer decides to use the differential capacitance of the diode. It is
assumed that the electric measurements at room temperature indicate a dif-
Chapter 7 AcUve RF Component Modeling
400
fusion capacitance of Cd = 0.329 pF at an applied junction voltage of

VA = 0.5 V, and C d = 0.371 nF at VA = 0.7 V. Find the emission coefficient n and the transit time tT .
7.4
A GaAs Schottky diode with gold contact is operated at 80 rnA. The following parameters are given at 300K: tT =40 ps, Rs =3 .Q, n 1.2,
Is= 10- 14 A. (a) Plot the magnitude of the small signal impedance behavior
in the frequency range from 1 MHz to 5 GHz. (b) Repeat the calculations for
a temperature of 400K.
7.5
For the PIN diode configuration shown below, compute the S-parameters of
the circuit when the control voltage equals either +1 V or -1 V and the frequency ranges from I MHz to 10 GHz. The diode model parameters are
F
I 5 = 5xl0-15 A, n = 1.2, tr =lOOps, m = 0.5, C 10 = 10 p,
V diff = 0.7 V, and Rs = 10 Q. The ambient operating temperature is
T =300K, and we set infinite values for the blocking capacitors and RFCs.
~ontrol o - - - - - - - - ,
RFC
Ca
RF'in
Ca
o---1t----1--+--......------1~
RF'out
RFC
7.6
Determine the change in the forward-bias voltage of an ideal Si pn-junction

diode with change in temperature from -20C to 80C. Assume that current
is kept constant and the initial bias voltage was 0. 7 V at T = 300K.
7.7
Find the maximum operation frequency of the ideal pn-junction diode

whose parameters are given in Example 7-1. The maximum frequency can
be estimated based on the RC constant of the diode.
7.8
Consider three ideal pn-junction diodes whose parameters are identical

except for the bandgap energy. Find the ratio of the forward-biased currents
for these diodes if the applied voltage is the same in each case and the diodes
are made of Ge, Si, and GaAs, respectively.
Problems
401
7.9
The terminal base current is constrained to be zero in an npn-BJT (opencircuit condition). Assuming that the device is operated at room temperature
and has aF == 0.99 and aR == 0.05, use the large-signal Ebers-Moll model
to find the base-emitter voltage as a function of the applied collector-emitter
voltage V CE.
7.10 Express the transconductance g m of a bipolar junction transistor in terms of

its collector current. Compare this expression with the expression for a differential resistance of a pn-diode.
7.11
Show that for a small-signal transistor model as depicted in Figure 7-16, the
input Miller capacitance can be written as C M 1 = (1 + g m r ce) C J.l.. In addition, obtain an upper frequency limit for which this formula is still applicable.
7.12 For a hybrid-1t BJT model plot the short-circuit current gain h1e in the frequency range from 10 MHz to 10 GHz. Assume the following parameters
are given at a collector bias point of 20 rnA and T =300K: ~ 0 == 140,
CJL = 0.1 pF, and Crc = 5 pF.
7.13
In Example 7-4 we discussed the relatively complicated case of a microwave

transistor analysis where we have taken into account effects associated with
parasitic elements such as lead inductances and resistances. In most practical
applications, the situation is even more complicated due to the presence of
internal matching and stability networks incorporated into the transistor
housing by the manufacturer.
For the internal circuit shown, compute the S-parameters in the frequency range of 100 MHz to 20 GHz.
Csc
402
The following component values are given: R 1 = 25 Q , R 2 = 20 n,

cl = c2 = 0.2 pF' CBE = CcE = 0.1 pF' and CBc = 10 tF. Assume
that the biasing conditions and the values for all inductances and components in the hybrid-1t model are the same as in Example 7-4.
7.14 An easy way to detennine the capacitance C ~ in the hybrid-1t BJT model is
to make a capacitance measurement between base and collector, as follows.
--------~
C-meter
(CEXT)
~------~
If the frequency is sufficiently low such that 1/ (roCf.l) r8 , we can directly

relate the externally recorded capacitance to the feedback capacitance C11
Show that this is true by proving that the voltage v1t is zero and that r7t, C7t,
and g m do not influence the measurement. If a precision instrument measures an external capacitance Cext = 0.6 pF at 1 MHz, can r 8 , which typi
cally ranges between 25 and 200 .Q, be neglected?
7.15
For the hybrid-1t model it is required to find the parameters r1t , r 8 , and gm
from low-frequency measurements (which allow us to neglect Cf.l and C1t).
The following measurement arrangement is given:
'....rii ........--.......--- ..---. --.....!
l---<>-...,
VBE-=-
E :..........................................................: E
At the operating point and at room temperature (25 C) we record a DC base

current of I 8 = 100 J.lA at a base emitter voltage of V BE = 11.8 V, and a
short-circuit collector current of I c = 25 rnA.
Problems
7.16
403
A small-signal BJT model has the following parameters: gm = 40 mS,

IT= 600 MHz, h21 = 100, 'ce = 2.5 kQ, 'bb'- 125 .n, and
C b' c = 2 pF . A load R L = 50 .Q is attached as shown.
Under the assumption that V L = -gm V b'eRL, find the Miller capacitance
C M such that the circuit can be approximated as
7.17
Neglecting all parasitic elements, including base, emitter, and collector

resistances in the transistor described in Example 7-4, find the maximum
frequency f 0 , the beta cut-off frequency f ~, and the transition frequency
fT
7.18
Obtain the h-parameter representation for a BJT in common-base configuration, neglecting base, emitter, and collector resistances ( r 8 , r E, and r c).
7.19
Derive the h-parameter representation for the following high-frequency FET

model:
404
7.20
Using the equivalent circuit shown in Problem 7.19, obtain the h-parameter
representation for a FET in conunon-gate configuration.
7.21
For the FET circuit model in Problem 7.19, find the equivalent input and
output impedances by replacing C gd with its equivalent ~iller capacitances.
Under what condition is this approximation valid?
7.22
For the simplified FET model shown, determine the capacitances Cgs and
C 8d as well as 8m.
Show that for low frequency operation it is sufficient to record the draincurrent and gate-source voltage under short-circuit output condition. Further,
design a measurement protocol to predict C gs and C gd .
7.23
PET models are often given in terms of Y-parameters, as the following

generic figure shows:
Convert this model into a

and D.
1t -network
and determine its coefficients A, B, C,
7.24
For the model parameters in Problem 7 .16, plot the cut -off frequency
a function of load resistance in the range 10 Q ::; R L ~ 200 Q .
fr
as
CHAPTER
Matching and Biasing

Networks
pointed out in Chapter 2, to achieve maximum

power transfer, we need to match the impedance of the load to that of the source. Usually this is accomplished by incorporating additional passive networks connected inbetween source and load. These networks are generically referred to as matching networks. However, their functionality is not simply limited to matching source and load
impedances for optimal power flow. In fact, for many practical circuits matching networks are not only designed to meet the requirement of minimum power loss but are
also based on additional constraints, such as minimizing the noise influence, maximizing power handling capabilities, and linearizing the frequency response. In a more
general context, the purpose of a matching network can be defined as a transformation
to convert a given impedance value to another, more suitable value.
In this chapter we restrict our coverage to the techniques of performing imped_ance transformation using passive matching networks. The emphasis is to ensure minimwn reflections between source and load. All remaining considerations, such as noise
:figure and linearity, are left for discussions in Chapter 9.
We commence with a study of networks based on discrete components. These networks are easy to analyze and can be used up to frequencies in the low GHz range.
i: Next, we continue with the analysis and design of matching networks using distributed
;elements, such as strip lines and stub sections. These networks are more suitable for
;operational frequencies exceeding 1 GHz, or for cases where vertical circuit dimen;:sions are of importance, as required in RF integrated circuit designs.
_;
To simplify our treatment and to gain clarity in the design methodology, the Smith
~Chart will be utilized extensively throughout as a primary design tool.
405
Chapter 8 Matching and Blaslng Networks
406
8.1 Impedance Matching Using Discrete Components

8.1.1
Two-Component Matching Networks
In a generic sense our engineering efforts primarily strive for two main goals:
first, to meet system specifications~ and second~ to find the most inexpensive and reliable way to accomplish this first task. The cheapest and most reliable matching networks are usually those that contain the least number of components.
The topic of this section is to analyze and design the simplest possible type of
matching networks: so-called two-component networks, also known as L-sections
due to their element arrangement. These networks use two reactive components to
transform the load impedance (ZL) to the desired input impedance (Zin). In conjunction with the load and source impedances, the components are alternatively connected/
in series and shunt configuration, as shown in Figure 8-1, which depicts eight possible
arrangements of capacitors and inductors.
r................... ...... .. :
:.. c2
.:.
:.........................................:
(a)
(b)
.
.l
(c)
,..........................................,
(d)
..l
.-------- ...............
(e)
Figure 8-1
(t)
(g)
(h)
Eight possible configurations of the discrete two-component matching

networks.
In designing a matching network we have two broad approaches at our disposal:

1. To derive the values of the elements analytically
2. To rely on the Smith Chart as a graphical design tool
The first approach yields very precise results and is suitable for computer synthesis. Alter-
natively, the second approach is more intuitive, easier to verify, and faster for an initial .
design, since it does not require complicated computations. The example below details
the use of the analytical approach to design a particular L-section matching network.
407
----------------------------~~~
Example 8-1: Analytical approach to the design of an L-section matching network
The output impedance of a transmitter operating at a frequency of

2 GHz is Zr = ( 150 + j75).Q. Design an L-section matching network, as shown in Figure 8-2, such that maximum power is delivered to the antenna whose input impedance is Z A = (75 + jl5).Q.
fc 1
~~~~~
~-o~~
Transmitter
.... . ..............
.,.,
. . ... .........'
Figure 8-2 Transmitter to antenna matching circuit design.
Solution:
The condition of maximum power transfer from the

source to the load requires the source impedance to be equal to the
complex conjugate of the load impedance. In our case this implies that
the output impedance Z M of the matching network has to be equal to
thecomplexconjugateofZA [i.e.,ZM = Z~ = (75-jl5).Q ).
The impedance Z M can be computed as a series connection of
an inductor Land a parallel combination of C and Zr:
1
(8.1)
Z M = -1
+ JxL = z*A
Zr + jBc
where B e = roC is the susceptance of the capacitor and XL = roL

is the reactance of the inductor. Expressing transmitter and antenna
impedances in terms of their real and imaginary parts (i.e.,
Zr = Rr + jXr and ZA = RA + jXA ), we can rewrite (8.1) as
Rr+ jXT
.
.
1 + jBc (Rr + jXr) +}XL = RA- JXA
(8.2)
Separating real and imaginary parts in (8.2), a system of two equations is found:
Chapler 8 Matching and Blaolng
408
Rr = RA(l-BeXr)+(XA+XL)BeRr
..;J
(8.3a)
= RrRABe- ( 1- BcXr)(XA +XL)
(8.3b)
Solving (8.3a) for XL and substituting into (8.3b) results in a quadratic equation for B c whose solution is
Xr
RT 2
2
2
R(Rr + Xr)- Rr
Be
A
= ----~~2----2------
(8.4)
Rr+Xr
Since Rr > R , the argument of the square root is positive and

1
greater than X T Therefore, to ensure a positive B c we must choose
the "plus" sign in (8.4). Substituting (8.4) into (8.3a) yields XL as
XL
= --1 - RA( 1- BcXr) - X A

Be
BeRT
(8.5)
Inserting numerical values into (8.4) and (8.5), we find

Be
= 9.2 mS ~ C = Bclro = 0.73 pF
XL
= 76.9 Q
=}
= XL/ro
= 6.1 nH
This example shows the analytical approach of designing an Lsection matching network by solving a quadratic equation for C and
then a linear equation for L. The process is tedious but can be easily
implemented on a mathematical spreadsheet.
As we may anticipate from Example 8-1, the analytical approach of desi~

matching networks can become very complicated and computationally intensive e~
for simple L-sections. Instead of the preceding method, we can use the Smith Chart ~ ~
rapid and relatively precise designs of the matching circuits. The appeal of
approach is that its complexity remains almost the same independent of the number o~
components in the network. Moreover, by observing the impedance transformation
the Smith Chart we obtain a "feel" of how the individual circuit elements contribute,
achieving a particular matching condition. Any errors in component selection and val
assignment are observed immediately and the design engineer can directly interven :
With the help of a personal computer, this process is carried out in real time. That is,
parameter choice (Lor C) and its value assignment can be instantaneously displayed
part of the Smith Chart on the computer screen.
<1
409
The effect of connecting a single reactive cmnponent (either capacitor or inductor)

to a complex load is described in considerable detail in Section 3 .4. Here we just point
out the following:
The addition of a reactance connected in series with a complex impedance results
in motion along a constant-resistance circle in the combined Smith Chart
A shunt connection produces motion along a constant-conductance circle.
This is indicated in Figure 8-3 for the combined ZY Smith Chart. Concerning the direction of the rotation, the general rule of thumb is that whenever an inductor is involved,
we rotate in the direction that moves the impedance into the upper half of the Smith
Chart. In contrast, a capacitance results in the movement toward the lower half.
Figure 8-3
Impedance effect of series and shunt connections of Land C to a

complex load in the Smith Chart.
410
Having established the effect of connecting a single component to the load, we

can now develop suitable two-component matching networks that perform the transformation from any load impedance to any specified input impedance. In general, designing an L-type matching network, or for that matter any passive network, in the ZY
Smith Chart consists of moving along either constant resistance or constant conductance circles.
In the following example we illustrate this graphical design technique as an alternative to the analytical approach discussed in Example 8-1. Most modem CAD programs allow us to conduct this graphical approach interactively on the computer screen.
In fact, simulation packages such as MMICAD directly permit the placement of components with the corresponding impedance behavior displayed on the Smith Chart.
----------------------~~&MW4
Example 8-2: Graphical approach to the design of the L-section matching network
Design the L-type reactive matching network discussed in Example
8-1 by using the Smith Chart as a graphical design tool.
Solution:
The first step is to compute normalized transmitter and

antenna impedances. Since no characteristic impedance Z 0 is given,
we arbitrarily select Z 0 = 75 n. Therefore, the normalized transmitter and antenna impedances are zr = Zr1Z0 = 2 + jl and
zA = Z AIZ0 = 1 + j0.2, respectively. Since the first component
connected to the transmitter is a shunt capacitor, the total impedance
of this parallel combination is positioned somewhere on the circle of
constant conductance that passes through the point zr in the combined Smith Chart (see Figure 8-4).
Next, an inductor is added in series with the parallel combination of transmitter Zr and capacitor; the resulting impedance will
move along the circle of constant resistance. For maximum power
gain we require an output impedance of the matching network connected to the transmitter to be equal to the complex conjugate of the
antenna impedance. This circle has to pass through
zM = z~ = 1 - j0.2, as shown in Figure 8-4.
Impedance Matching Ualng Discrete Components
Figure 8-4
411
Design of the twoelement matching network as part of the ZYSmith

Chart.
The intersection of two circles in the Smith Chart determines the

normalized impedance fonned by the shunt connection of transmitter
and capacitor. Reading from the Smith Chart, we find that this impedance is approximately Zrc = 1 - j 1.22 with the corresponding
admittance of Yrc = 0.4 + j0.49. Therefore, the normalized susceptance of the shunt capacitor is jb c = Yrc - Yr = j0.69 and the normalized reactance of the inductor is jxL = zA- Zrc = j1.02.
Finally, the actual values for the inductor and capacitor are
= (xLZ0)/ro
= 6.09 nH
C = bc/ (roZ0 )
= 0.73 pF
412
Chapter 8 Matching and Blaslng Networkl
This example presents a simple and yet precise graphical

approach to design L-section matching networks. The method can
be readily extended to more complicated systems.
The design procedure described in Example 8-2 can be applied to any L-section
matching network shown in Figure 8-1. The generic solution procedure for optimal
power transfer includes the following six steps:
1. Find the normalized source and load impedances.
2. In the Smith Chart plot circles of constant resistance and conductance that pass
through the point denoting the source impedance.
3. Plot circles of constant resistance and conductance that pass through the point of
the complex conjugate of the load impedance.
4. Identify the intersection points between the circles in steps 2 and 3. The number
of intersection points determines the number of possible L-section matching
networks.
5. Find the values of the normalized reactances and susceptances of the inductors
and capacitors by tracing a path along the circles from the source impedance to
the intersection point and then to the complex conjugate of the load impedance.
6. Detennine the actual values of inductors and capacitors for a given frequency.
In the preceding steps it is not necessary to move from the source to the complex
conjugate load impedance. As a matter of fact, we can transform the load to the complex conjugate source impedance. The following example illustrates the first approach,
whereas Section 8 .1.2 discusses the second method.
----------------------------RF~~
Example 8-3: Design of general two-component matching
networks
Using the Smith Chart, design all possible configurations of discrete
two-element matching networks that match the source impedance
Zs = (50+ }25)0. to the load ZL = (25- }50)0.. Assume a characteristic impedance of Z 0 = 50 0 and an operating frequency of
/=2 GHz.
413
Solution:
We follow the six steps listed previously.
1. The normalized load and source impedances are:
zs
Z 8 / Z 0 = 1 + }0.5 or Ys
0.8- }0.4
zL
ZLIZ 0 = 0.5- jl or YL
3 + }0.8
2. We plot circles of constant resistance and constant conductance that

pass through the points of the normalized source impedance (dashed
line circles in Figure 8-5), and
Figure 8-5 Design of a matching network using the Smith Chart
3. Complex conjugate of the load impedance (solid line circles in Figure

8-5).
414
4. These circles intersect in four points denoted as A, B, C, and D, with

the normalized impedances and admittances being as follows:
ZA
z8
Zc
ZD
= 0.5+}0.6,
= 0.5- }0.6,
= 1 - j 1.2 ,
= 1 + }1.2,
= 0.8-jl
y 8 = 0.8 + jl
y c = 3 + }0.5
YD = 3- j0.5
YA
5. Since there are four intersection points, we expect four possible configurations of L-section matching networks. Indeed, if we move along the
Zs ~ zA ~ zi path we see that from point zs to zA the impedance is
transformed along the circle of constant conductance indicating shunt
connection. Moreover, we move toward the upper half of the Smith
Chart (see Figure 8-3), which indicates that the first component connected to the source should be a shunt inductor. From points zA to zi
the impedance is transformed along the circle of constant resistance,
with movement toward the upper half of the chart indicating series
connection of the inductance. Therefore, the zs --7 zA --7 zi path
results in a "shunt L, series L" matching network, as shown in Figure
8-l(f). If the Zs --7 z8 --7 zi path is chosen, we obtain a "shunt C,
series L" network [Figure 8-l(h)]. For zs --7 zc --7 z1 the matching
network is "series C, shunt L" [Figure 8-l(a)]. Finally, for the
zs --7 zD --7 zi path, a matching network is constructed by a "series L,
shunt L" combination, which is shown in Figure 8-1 (e).
6. We finally have to find the actual component values for the matching
networks identified in the previous step. If we direct our attention again
to the zs --7 zA --7 zi path, we see that from the source impedance to
the point zA the normalized admittance of the circuit is changed by
jbL2
= YA- Ys = (0.8- jl)- (0.8- }0.4) = -j0.6
From here the value of the shunt inductor is:
L2
= -bLZo-(i) = 6.63nH
2
Transformation from point zA to zi is done by adding an inductor

connected in series to the impedance zA Therefore,
}XL
= zi- ZA = (0.5 + jl)- (0.5 + j0.6) = }0.4
and the value of this inductor is
415
Z0
= 1.59 nH
ro
The values of the components for the remaining three matching networks are found in the same way. The results are shown in Figure 8-6.
L1 =
L,
6.37 nH
2.23 pF
0.94 pF
2.79 nH
z,_
Figure 8-6
Zr;
Matching networks for four different paths in the Sm ith Chart.
The Smith Chart allows us immediate observation whether or

not a particular impedance transformation is capable of achieving
the desired matching. Moreover, the total number of possible network connections can readily be seen.
8.1.2
Forbidden Regions, Frequency Response, and Quality Factor
Before continuing with the frequency analysis of L-type matching networks, let
us first note that not every network topology depicted in Figure 8-1 can perform the
required matching between arbitrary load and source impedances. For example~ if the
source is Zs = Z 0 = 50 .Q and if we use a matching network shown in Figure 8- l(h),
then the addition of the capacitor in parallel with the source produces motion in clockwise direction away from the circle of constant resistance that passes through the origin. This implies that all load impedances that fall into the shaded region in Figure
8-7(a) cannot be matched to the 50 Q source by this particular network.
Similar "forbidden regions" can be developed for all L-type matching network
topologies depicted in Figure 8-1. Examples of such regions for several other networks
416
Chapter 8 Matching and Blaslng Networi
a
0
1.()
II
1\f
II
f\.JU)
.c.
.....
::
~
.....
0
~r::::
C'l
r::::
:c
<.>
.....
ca
E
Q)
c..
c
I
....J
.....
( /J
c:
.Q
C'l
c:
Q)
"C
"C
:c
....
0
u.
417
based on a 50 n source impedance are shown in Figure 8-7. Here the shaded areas
denote values of the load impedance that cannot be matched to the 50 n source. It is
important to keep in mind that the forbidden regions in Figure 8-7 are applicable only
when dealing with a Z s = Z 0 = 50 n source impedance. The regions take on totally
different shapes for other source impedance values.
As explained in Example 8-3 and displayed in Figure 8-7, for any given load and
input impedances there are at least two possible configurations of L-type networks that
accomplish the required match. The question now is, what is the difference between
these realizations and which network should ultimately be chosen?
Besides the obvious reasons for selecting one network over another (for instance,
availability of components with required values), there are key technical considerations, including DC biasing, stability, and frequency response. In the remainder of this
section we concentrate primarily on the frequency response and quality factor of the Ltype matching networks, whereas DC biasing issues are covered later in Section 8.3.
Stability is deferred to Chapter 9.
Since any L-type matching network consists of series and shunt combinations of
capacitors and/or inductors, the frequency response of these networks can be classified
as either low-pass, high-pass, or bandpass filters. To demonstrate such behavior, let us
consider a matching network that transforms a complex load, consisting of resistance
RL = 80 n connected in series with capacitor C L = 2.65 pF, into a 50 .Q input
impedance. Let us further assume that the operational frequency for this circuit is
fo = 1 GHz.
At 1 GHz the normalized load impedance is zL = 1.6 - j 1.2 , and according to
Figure 8-7 we can use either one of the matching networks shown in Figure 8-7 (c) or
Figure 8-7 (d), following a similar design procedure as described in Example 8-2. However~ because the source impedance zs is real (zs =50 Q) it is easier to transform from
the load to the source impedance since z.S = Zs = 50 n . This is shown in Figure
8-8(a). The corresponding matching networks are shown in Figures 8-8(b) and 8-8(c).
The frequency responses of these two networks in terms of the input reflection
coefficient rin = (Zin- Z 5 )/ (Zin + Zs) and the transfer function H = V out I V s
(where the output voltage V out is measured across the load resistance R L = 80 n) are
shown in Figures 8-9(a) and (b), respectively.
It is apparent from Figure 8-9 that both networks exhibit perfect matching only at
a particular frequency f 0 = 1 GHz and begin to deviate quickly when moving away
from fo .
The previously developed matching networks can also be viewed as resonance
circuits with f 0 being the resonance frequency. As discussed in Section 5.1.1, these
Chapter 8 Matching and Blaalng Networka
418
(a) Impedance transformations displayed in Smith Chart
c~o.6pFI
(b)
Figure 8-8
(c)
Resulting matching networks

Two design realizations of an L-type matching network.
networks may be described by a loaded quality factor, Q L, which is equal to the ratio of
the resonance frequency f 0 over the 3 dB bandwidth B W
fo
QL = BW
(8.6)
'\
\ /
419
Circuit in
Figure 8-8(b)
''
'
I
'
~Circuit in
'
Figure 8-8(c) -
--- -----------------0.1
0~----~----~----~----~----~----~
1
1.5
2
Frequencyf, GHz
0.5
2.5
(a) Frequency response of input reflection coefficient

-3
Circuit in
/ F i g m e 8-8(b)
-3.5
-4
- - - - -- - - - - - - - - - - -
~ -4.5
~
0
.....
.....
-5.5
c.S
00
Circuit in
Figure 8-8(c)
a -6.5
7
-7.5
-8
0.5
1.5
2
Frequency,{, GHz
2.5
(b) Transfer function of the matching networks
Figure 8-9
Frequency response of the two matching network realizations.
Chapter 8 Matching and Blasing Networks
420
where both f 0 and B W are expressed in Hz. The question now is how to find the bandwidth of the matching network. To answer this, we will exploit the similarity between
the bell-shaped response of the matching network's transfer function near fo [see Figure 8-9(b)] and the frequency response of a bandpass filter.
For frequencies close to f 0 the matching network in Figure 8-8( c) can be redrawn
as a bandpass filter with a loaded quality factor calculated based on (8.6). The equivalent
bandpass filter is shown in Figure 8-lO(a). The equivalent capacitance C T in this circuit
is obtained by replacing the series combination of R L and C L in Figure 8-8(c) with an
equivalent parallel connection of R LP and C LP and then adding the capacitances C and
C LP : C T = C + C LP . The equivalent shunt inductance L LN is obtained by first replacing the series connection of the voltage source V s , resistance R 5 , and inductance L
with the Norton equivalent current source IN = V 5 1 ( R s + j ro0 L) connected to the parallel combination of conductance G SN and inductance L N , where the admittance is
given as follows: GsN + (jro0 LN)-1 = (R 5 + jro0 L)- 1 . Next, the current source IN and
conductance G SN are converted back into a Thevenin equivalent voltage source
Rs- jro0L
R
= Vs
= V s<1 -
j 1.2255)
(8.7)
and series resistance

Rsr
-1
Rs + ( rooL)
= GsN = --------Rs
(8.8)
The resonance circuit in Figure 8-10 is loaded by the combined resistance

RT = RL II Rsr = 62.54 n. Thus, the loaded quality factor QL of the equivalent
bandpass filter is given by
fo
QL = BW = ro0 R 7 C =
Rr
jXcl = 0.61
(8.9)
It is immediately noticed that the maximum gain for the equivalent bandpass filter
is higher than the gain of the original matching network. This is explained by the fact
that for the matching network we measure the output voltage on the load RL, while for
the equivalent filter we measure the output voltage at the equivalent load resistance RLP
which is connected in parallel with the capacitance CT . Therefore, the conversion from
V b to V out at the resonance frequency can be found through the voltage divider rule:
421
R.n-= 125.1 Q
,...---...--t---->------+--<1 ~
CT I
Luv
16.2 nH
1.55 pF
R,,, =125.1 Q
(a) Equivalent bandpass filter

-2
Equivalent
/filter
-3
~
~" -4
Circuit in
Figure 8-8(c)
0
......
.......
(.)
tE
-5
~
~
-6
f::
-7
-8
0.5
2
Frequency f, GHz
1.5
2.5
(b) Frequency response of the matching network compared to the equivalent

filter response
Figure 81 0 Comparison of the frequency response of the L -type matching

network and an equivalent bandpass filter.
which gives us
201og l~;~l = - 2.0382 + 20log:~:: = -3.9794 dB

a result that agrees very well with Figure 8-9(b).
From the known Q L we can directly find the bandwidth of the filter:
BW = f 0 1 QL = 1.63 GHz. The frequency response in Figure 8-9(b) shows that the
3 dB point for f < f 0 occurs at f min = 0.40 GHz and for f > f 0 the 3 dB point corresponds to f max = 2.19 GHz. Thus, the bandwidth of the matching network is
422
BW
= f max- f min = 1.79 GHz, which again agrees reasonably well with the result
obtained for the equivalent bandpass filter.

The equivalent bandpass filter analysis allows us to explain the bell-shaped
response of the matching network in the neighborhood of f 0 and provides us with a
good estimation of the bandwidth of the circuit. The only drawback to this approach is
its complexity. It would be desirable to develop a simpler method of estimating the
quality factor of the matching network without having first to develop an equivalent
bandpass filter or even computing the frequency response of the network. This is
accomplished through the use of a so-called nodal quality factor Qn.
Let us go back to Figure 8-8(a), where we illustrate the impedance transformation
as we move from one node of the circuit to another. We note that at each node of the
matching network the impedance can be expressed in terms of an equivalent series
impedance Z 5 = R 5 + jX5 or admittance Y p = Gp + jB p. Hence, at each node we
can find Qn as the ratio of the absolute value of the reactance X 5 to the corresponding
resistance R s
Qn
IXsl
= Rs
(8.10)
or as the ratio of the absolute value of susceptance B P to the conductance G p

Qn
8 PI
I
=G
(8.11)
Using (8.10) and (8.11) and the impedance transformations in Figure 8-8(a), we
can deduce that for the matching network shown in Figure 8-8(c) the maximum nodal
quality factor is obtained at point B where the normalized impedance is 1 - j 1.23 ,
resulting in
Qn
= 11.231/1 = 1.23
(8.12)
To relate the nodal quality factor Qn to QL, we compare the result of (8.12) with
(8.9) and find
(8.13)
This result is true for any L-type matching network. For more complicated configurations the loaded quality factor of the matching network is usually estimated as simply the maximum nodal quality factor. Even though this approach does not yield a
quantitative estimate of the circuit bandwidth, it nonetheless allows us to compare networks qualitatively and to select a network with higher or lower bandwidth.
Impedance Matching Ualng Discrete Components
423
To simplify the matching network design process even further we can draw constant- Qn contours in the Smith Chart. Figure 8-11 shows such contours for Qn valued
0.3, 1, 3, and 10.
Figure 8-11
Constant On contours displayed in the Smith Chart.
To obtain the equations for these contours we refer back to the general derivation
of the Smith Chart in Chapter 3. There it is shown in (3.6) and (3.7) that the normalized
impedance can be written as
= r + jx
1-r2r -r.2
2r.t
=
2
2+]
2
2
(1 - r ,) + r;
<1 - r,.) + r;
l
(8.14)
Thus, the nodal quality factor can be written as
Q =
n
lxl =
r
21ri1
1- r2- r~
r
Rearranging terms in (8 .15), it follows that a circle equation is found in the form
(8.15)
424
Chapter 8 Matching and Blasing Networt
2 (rr:r-Q1 )2 =1+21
ri+
(8.16
Qn
where the "plus" sign is taken for positive reactance x and the "minus" sign for nega
tive x.
With these constant Qn circles in the Smith Chart it is possible to find the loade1
quality factor of an L-type matching network by simply reading the corresponding Q,
and dividing it by 2. This procedure is discussed in Example 8-4.
----------------------------RF~~
Example 8-4: Design of narrow-band matching network
Using the forbidden regions in Figure 8-7, design two L-type networks that match a ZL = (25 + j20)Q load impedance to a 50 n
source at 1 GHz. Determine the loaded quality factors of these networks from the Smith Chart and compare them to the bandwidth
obtained from their frequency response. Assume that the load consists of a resistance and inductance connected in series.
As we see from Figure 8-7, the normalized load
Solution:
impedance zL = 0.5 + }0.4 lies inside of the constant conductance
circle g = 1 . There are two L-type matching networks that satisfy
our requirements. The first consists of a series inductor and shunt
capacitor, as shown in Figure 8-7(a), and the second is a series
capacitor with shunt inductor, as shown in Figure 8-7(b). Following
the same procedure as described in Example 8-2, we obtain the two
matching networks shown in Figure 8-12.
According to Figure 8-12(a), the nodal quality factor for both
networks is equal to Qn = 1. Thus, we can expect that the bandwidth should be equal to BW
f 0 /QL = 2f0 / Qn = 2 GHz.
This is checked by plotting the corresponding frequency responses
for the designed matching networks, as depicted in Figure 8-13.
We observe that the bandwidth for the network corresponding to
Figure 8-12(c) is approximately BWc = 2.4 GHz. Interestingly, the
matching network corresponding to Figure 8-12(b) does not possess a
lower cut-off frequency. However, if we assume that the frequency
425
(a) Impedance transformation in the Smith Chart
Rs= 50
0.8 nH LL= 3.18 nH
Rs =50 n C= 3.54pF LL=3.18 nH
fo"ut
Vou1
r C = 3.18 pF
--
(b)
(c)
Resulting matching networks
Figure 8-12 Two L-type matching networks for a 50 Q source and a

ZL = (25 + j20)il load impedance operated at a frequency of 1 GHz.
response is symmetric around the resonance frequency f 0 = 1 GHz,

then the bandwidth will be BWb = 2(/max- f 0 ) = 1.9 GHz, with
the upper cut-off frequency being f max = 1.95 GHz.
Chapter 8 Matching and Blasing Networkl
426
-8
Fg
-9
Figure 8-12(c)
:::.._v,
'1;
-10
:::.._o
3 dB
II
::t: -11
'I
1::::
..0. .
u
~
~
-12 --------
----
'
----- -- ---- ------------------- - -----------1
- 13
-14
'
- 15
L _ __ . . . J L . . __
Figure 8-13
0.5
__.___
__..J...
---'--....!....l...---L
1.5
2
2.5
Frequency f, GHz
3.5
Frequency responses for the two matching networks.
Despite their design for the same resonance frequency, certain

matching network configurations exhibit better high or low frequency rejection, as Figure 8-13 exemplifies.
In many practical applications the quality factor of the matching network is of

importance. For example, if we design a broadband amplifier we would like to utilize
networks with low Q in order to increase the bandwidth. However, for oscillator design
it is desirable to achieve high- Q networks to eliminate unwanted harmonics in the output signal. Unfortunately, as we have seen in the previous example, L-type matching
networks provide no control over the value of Qn and we must either accept or reject
the resulting quality factor. To gain the freedom of choosing the values of Q and thus
affect the bandwidth behavior of the circuit, we can introduce a third element in the
matching network. The addition of this third element results in either aT- or Pi-network, both of which are discussed next.
8.1.3
T and PI Matching Networks
As already pointed out, the loaded quality factor of the matching network can be
estimated from the maximum nodal Qn. The addition of the third element into the
matching network produces an additional node in the circuit and allows us to control
the value of Q L by choosing an appropriate impedance at that node.
427
The following two examples illustrate the design ofT- and Pi-type matching networks with specified Qn factor.
----------------------------~&)A~
'
Example
8-S: Design of a T matching network
Design aT-type matching network that transforms a load impedance
ZL = (60- j30)Q into a Zin = ( 10 + j20).Q input impedance
and that has a maximum nodal quality factor of 3. Compute the values for the matching network components, assuming that matching
is required at f = 1 GHz.
Solution:
There are several possible solutions that satisfy the

design specifications. In this example, we investigate only one design
since the rest can easily be obtained by using the same approach.
The general topology of the T-type matching network is shown
in Figure 8-14.
Figure 814
General topology of aT-type matching network.
The first element in this network is connected in series with the

load impedance. Because Z 1 is purely reactive, the combined
impedance Z A will reside somewhere on the constant resistance circle described by r = rL. Similarly, Z 3 is connected in series with
the input so that the combined impedance Z B (consisting of Z L,
Z 1 , and Z 2 ) is positioned somewhere on the constant resistance circle with r = rin. Because the network should have a nodal quality
factor Qn = 3, we can choose the impedance values in such a way
that Z B is located on the intersection of the constant resistance circle r = r in and the Qn = 3 circle (see point B in Figure 8-15).
428
Chapter 8 Matching and Blaelng Networb
Figure 8-15
Design of aT-type matching network for a specified Qn = 3.
We next find the intersection point A of the constant conductance circle that passes through the point B obtained from the previous step. The circle of constant resistance r = rL now allows us to
determine the required value of the remaining component of the network to reach the point Zin.
The complete T-type matching network with the actual component values is illustrated in Figure 8-16. The computed elements are
based on the required matching frequency off= I GHz.
I C,= 3.53pF
Figure 8-16
T-type matching network circuit schematics.
Impedance Matching Using Dlaerete Components
429
The extra degree offreedom to adjust the quality factor (bandwidth) of a matching network comes at the expense of an additional
circuit element.
In the following example the design of a Pi-type matching network is developed

with the intent to achieve a minimum nodal quality factor. A low quality factor design
directly translates into a wider bandwidth of the network, as required, for instance, in
broadband FET and BJT amplifiers.
----------------------------RF~~
Example 8-6: Design of a Pi-type matching network
For a broadband amplifier it is required to develop a Pi-type matching network that transforms a load impedance of Z L = ( 10 - j 10)Q
into an input impedance of zin = (20 + j40)Q. The design should
involve the lowest possible nodal quality factor. Find the component
values, assuming that matching should be achieved at a frequency of
f = 2.4 GHz.
Solution:
Since the load and input impedances are fixed, we cannot produce a matching network that has a quality factor lower than
the highest Qn computed at the locations Z L and Zin. Therefore,
the minimum value for Qn is determined at the input impedance
location as Qn = IXini/Rin = 40/20 = 2. The Smith Chart
design of the Pi-type matching network based on Qn = 2 is
In the design we employ a method very similar to the one used
m Example 8-5. First, we plot a constant conductance circle
g = gin and find its intersection with the Qn = 2 contour in the
Smith Chart. This intersection is denoted as point B. Next, we find
the intersection point of the constant conductance circle g = gL
with the constant resistance circle that passes through the point B .
The resulting point is denoted as A in Figure 8-17.
430
Chapter 8 Matching and Biasing Networkl
Figure 8-17
Design of a Pi-type matching network using a minimal On.
The network components can be determined based on converting the Smith Chart points into actual capacitances and inductances
as detailed in Example 8-2. The resulting circuit configuration is
c2= 1.65 pF
Figure 8-18
Pi-type matching network configuration.
It is interesting to note that unlike the situation discussed in

Example 8-5, the relative positions of Z L and Zin in this example
are such that only one possible Pi-type network configuration with
431
M'crostr'p Line Matchtng Networks
=2
exists. All other realizations of the Pi-type network will

result in an increased nodal quality factor. Furthennore, if we had a
lower load resistance, we would not be able to implement this Pitype network for lhe given Q n .
Qn
...
As this example shows, the bandwidth cannot be increased

arbitrarily by reducing the nodal quality factor. The limits are set by
the desired input and output impedances.
8.2 Microstrip Line Matching Networks

In the previous sections we have discussed the design of matching networks
involving discrete components. However, with increasing frequency and correspondingly reduced wavelength, the influence of parasitics in the discrete elements becomes
more noticeable. The design now requires us to take these parasitics into account, thus
significantly complicating the component value computations. This, along with the fact
that discrete components are only available for certain values, limits their use in highfrequency circuit applications. As an alternative to lumped elements, distributed components are widely used when the wavelength becomes sufficiently small compared
with the characteristic circuit component length, a fact already discussed in Chapter 2.
8.2.1
From Discrete Components to Microstrip Lines
In the mid-GHz range, design engineers often employ a mixed approach by combining lumped and distributed elements. These types of matching networks usually
contain a number of transmission lines connected in series and capacitors spaced in a
parallel configuration, as illustrated in Figure 8-19. The reader is also referred to Figure
1-2(a) for a practical example.
z,.
Figure 8-19 Mixed design of matching network involving transmission line

sections ( TL) and discrete capacitive elements.
432
Chapter 8 Matching and Blaslng Networ1tl
Inductors are usually avoided in such designs because they tend to have higher
resistive losses than capacitors. In general, only one shunt capacitor with two transmission lines connected in series on both sides is sufficient to transform any given load
impedance to any input impedance. Similar to the L-type matching networks, such configurations may involve the additional requirement of a fixed Qn, necessitating additional components to control the quality factor of the circuit.
The arrangement of components shown in Figure 8-19 is very attractive in practice,
since it permits tuning the circuit after it has been manufactured. Changing the values of
the capacitors as well as placing them at different locations along the transmission lines
offers a wide range of flexibility. The tuning capability makes these types of matching
networks very popular for prototyping. Usually, all transmission lines have the same
width (i.e., the same characteristic impedance) to simplify the actual tuning.
Example 8-7 discusses the Smith Chart approach to the design of a matching network containing two 50 Q transmission lines connected in series and a single shunt
capacitor placed in-between them.
----------------------------~~~
Example 8-7: Design of a matching network with lumped and
distributed components
Design a matching network that transforms the load
ZL = (30 + jlO)Q to an input impedance zin = (60 + j80)Q.
The matching network should contain only two series transmission
lines and a shunt capacitance. Both transmission lines have a 50
characteristic line impedance, and the frequency at which matching
is desired is f = 1.5 GHz.
Solution:
The first step involves identifying the normalized load
impedance Zr = 0.6 + j0.2 as a point in the Smith Chart. We can
then draw a SWR circle that indicates the combined impedance of
the load connected to the 50 .Q transmission line. The position on
the SWR circle is determined by the length of the transmission line,
as investigated in Chapter 3.
The second step requires plotting a SWR circle that passes
through the normalized input impedance point zin = 1.2 + jl.6
Mlcrostrlp Line Matching Networks
Figure 8-20
Design of the distributed matching network for Example 8-7.
The choice of the point from which we transition from the load
SWR circle to the input SWR circle can be made arbitrarily. In Figure 8-20 the point A is chosen, which approximately corresponds to
a normalized admittance value of y A = 1 - j0.6 . The addition of
the parallel capacitor results in the movement along the circle of
constant conductance g = 1 and transforms the impedance from
point A to point B on the input SWR circle of the Smith Chart.
From point B an impedance transformation is required along the
constant SWR circle by adding a series connected transmission line.
As a final step, the electrical length of the transmission lines
must be determined. This can be done by reading the two lengths 11 ,
433
Chapter 8 Matching and Bluing Networkl
434
12 from the so-called WTG (wavelength toward generator) scale

displayed on the outer perimeter of the Smith Chart (see Figure
8-20). The resulting circuit schematics for the matching network is
shown in Figure 8-21
U1
Figure 8-21
Matching network combining series transmission lines and shunt

capacitance.
3.5
c
10:
~
2
.5
1.5
"
CL)
.....
....::s~
c:l.l
2.5
c:l.l
g.
"0
4)
-s
0.5
0..
.s=
2
1.5
.::!
('d
~
-1
z
-0.5
0.5
Figure 8-22
0.15
0.2
Distance /, A.
0.25
Input impedance as a function of the position of the shunt capacitor

in Example 8-7.
...
It is interesting to investigate the tuning capability range for

this circuit configuration. Figure 8-22 shows the dependency of the
real rin and imaginary xin parts of the input impedance as a function of the distance l between the load and the capacitor location. In
other words, the total length 11 + 12 is kept fixed and the placement
of the capacitor is varied from the load end to the beginning of the
435
network (i.e., 0 ~ l ~ l 1 + 12 ). The dashed lines indicate the original

design. It is noticed that xin undergoes the expected inductive (positive values) to capacitive (negative values) transition.
In this example we have designed a combined matching network that involves both distributed (transmission lines) and a
lumped (capacitor) element. These types of networks have rather
large tuning capabilities. but are very sensitive to the placement of
the capacitor along the transmission line. Even small deviations
from the target location result in drastic changes in the input
impedance.
8.2.2
Single-Stub Matching Networks
The next logical step in the transition from lumped to distributed element networks is the complete elimination of all lumped components. This is accomplished by
employing open- and/or short-circuit stub lines.
In this section we consider matching networks that consist of a series transmission
line connected to a parallel open-circuit or short -circuit stub. Let us investigate two
topologies: The first one involves a series transmission line connected to the parallel
combination of load and stub, as shown in Figure 8-23(a), and the second involves a
parallel stub connected to the series combination of the load and transmission line, as
depicted in Figure 8-23(b).
ZOL, IL
Openor ~:
short circuit
(a)
Figure 8-23
: !----.... Open or -=-
short circuit
(b)
Two topologies of single-stub matching networks.
The matching networks in Figure 8-23 possess four adjustable parameters: length
ls and characteristic impedance Z 05 of the stub, and length lL and characteristic
impedance Z 0 L of the transmission line.
Chapter 8 Matching and Biasing Network
436
Example 8-8 demonstrates the design procedure for the matching network topol
ogy shown in Figure 8-23(a) with the characteristic impedances of both stub Z 05 anc
transmission line Z 0 L fixed to the same arbitrary value Z 0 and their electrical length:
variable to meet the particular input impedance requirement.
----------------------------~&uM~
Example 8-8: Single-stub matching network design with fixed

characteristic impedances
For a load impedance of ZL = (60- }45).0, design two single-stub
matching networks that transform the load to a Zin = (75 + j90)Q
input impedance. Assume that both stub and transmission line in
Figure 8-23(a) have a characteristic impedance of Z 0 = 75 Q.
Solution:
The basic concept is to select the length l s of the stub

such that it produces a susceptance Bs sufficient to move the load
admittance y L = 0.8 + j0.6 to the SWR circle that passes through
the normalized input impedance point zin = 1 + j 1.2 , as illustrated
in Figure 8-24.
We notice that the input SWR circle associated with
Zin = 1 + jl.2 intersects the constant conductance circle g = 0.8
in two points (at YA = 0.8 + }1.05 and at y 8 = 0.8 - }1.05) suggesting two possible solutions. The corresponding susceptance valj bSA = y A - y L = }0.45
and
ues
for
the
stub
are
jb58 = y 8 - YL = - }1.65, respectively. In the first case, the length
of an open-circuit stub can be found in the Smith Chart by measuring the length lsA' starting from the y = 0 point (open circuit) and
moving along the outer perimeter of the Smith Chart g = 0 toward
the generator (clockwise) to the point where y = j0.45. The length
for this case is l SA = 0.067 "A.. The open-circuit stub can be replaced ..
by a short-circuit stub if its length is increased by a quarter wavelength. Such a substitution may become necessary if a coaxial cable
is used because of excessive radiation losses due to the large crosssectional area. In printed circuit design, open-circuit stubs are sometimes preferred because they eliminate the deployment of a via,
Figure 8-24
437
Smith Chart design for the single-stub matching network based on

Example 8-8.
which is otherwise necessary to obtain the ground connection for a

short-circuit stub.
Similar to the first solution, bSB yields the length
lsB = 0.337A. for the open-circuit stub, and lsB = 0.087/.... for the
short-circuit stub. For this case we also notice that creating a shortcircuit stub requires a shorter length than an open-circuit stub. This
is due to the fact that the open-circuit stub models a negative
susceptance.
The length of the series transmission line segment is found in
the same way as described in Example 8-7 and is equal to
438
Chapter 8 Matching and Biasing Networks
ILA
= 0.266A.
for the first solution and
lLB
= 0.07A for the second
solution.
A circuit designer often has to minimize the size of the circuit

board and therefore must be concerned about employing the shortest possible transmission line segments. Depending on the impedance requirements, this can either be an open- or short-circuit stub
section.
In the next example we illustrate the generic design procedure for the matching
network topology shown in Figure 8-23(b). Unlike the previous example, we now fix
the lengths of both the stub and the transmission line segment but vary their characteristic impedances. In a microstrip line circuit design this is typically accomplished by
changing the width of the lines.
----------------------------~~~
Example 8-9: Design of a single-stub matching network using
transmission lines with different characteristic
impedances
Using the matching network topology shown in Figure 8-23(b),
choose the characteristic impedances of the stub and transmission
line such that the load impedance Z L = (120- j20)Q is transformed into the input impedance Zin = (40 + j30)Q. Assume that
the length of the transmission line is lL = 0.25A and the stub has
the length of ls = 0.375/.... Furthermore, determine whether a shortcircuit or an open-circuit stub is necessary for this circuit.
Solution:
The combined impedance Z 1 of the series connection

of the load impedance with the transmission line can be computed
using the formula for the quarter-wave transfonner:
2
Z 1 = Z 0 LIZL
(8.17)
..
439
The addition of the open-circuit stub results in an input admittance
of
(8.18)
where Y 1 = Z~ is the admittance of the previously computed
series combination of load impedance and transmission line and
jB5 = +jZ~1 is the susceptance of the stub. The "plus" or "minus"
signs correspond to either a short-circuit or an open-circuit stub.
Combining (8.17) and (8.18), we find
1
(8.19a)
(8.19b)
where we have used the input admittance and load impedance representation in terms of their real and imaginary components:
Yin
= Gin + j Bin ' Z L = RL + j XL
Using (8.19a), we find the characteristic impedance of the

transmission line to be
[T2o
ZoL = ~G: = ~D.Ol6 = 86.6
Substituting the obtained value into (8.19b), we find that the

"minus" sign should be used; that is, we need to implement an opencircuit stub with a characteristic impedance of
Z 05
I
2
= 107.1 fl
XLIZoL- Bin
This design approach is very easy to implement as long as the

characteristic impedance stays within reasonable limits ranging
approximately from 20 to 200 n.
In practical realizations single-sided unbalanced stubs are often replaced by the

balanced design, as shown in Figure 8-25.
Naturally, the combined susceptance of the parallel connection of stubs STl and
ST2 has to be equal to the susceptance of the unbalanced stub. Therefore, the susceptance of each side of the balanced stub must be equal to half of the susceptance of the
Chapter 8 Matching and Bluing Networks
440
ST2
TL 1
I(
o.2sA. )
sn
Zos
Figure 8-25
107.1 Q Z OL = 86.6 .Q
Balanced stub design for Example 8-9.
unbalanced stub. We note that the length lS B of each side does not scale linearly. In
other words, the length of the balanced stub is not half of the length of the unbalanced
stub l s . Rather, it has to be computed as
lsB
2rtl8)
= 2A1ttan_ ( 2tan--x-
(8.20)
lsB
21tls)
= 2A1ttan- I (12tan-r
(8.21)
for open-circuit stub, or
for short-circuit stub. This result can also be found graphically by using the Smith
Chart.
8.2.3
Double-Stub Matching Networks
The single-stub matching networks in the previous section are quite versatile and
allow matching between any input and load impedances, so long as they have a nonzero
real part. One of the main drawbacks of such matching networks is that they require a
variable-length transmission line between stub and input port, or between stub and load
impedance. Usually this does not pose a problem for fixed networks, but it may create
difficulties for variable tuners. In this section we examine matching networks that overcome this drawback by incorporating a second stub. The general topology of such a net- ,
work that matches an arbitrary load impedance to an input impedance Zin =Z 0 is
In double-stub matching networks two short- or open-circuit stubs are connected
in parallel with a fixed-length transmission line placed in between. The length 12 of this
line is usually chosen to be one-eighth, three-eighth, or five-eighth of a wavelength. The
441
Mlcrostrtp Line Matching Networks
ZA
Zin= Zo
Zc
ZB
Zv
12
[3
II
ls2
'
~~
ZL
lsi
'
Open or
/~
short circuit
Figure 8-26
Double-stub matching network arrangement.
three-eighth and five-eighth wavelength spacings are typically employed m highfrequency applications to simplify the tuner construction.
Let us assume for our subsequent discussion that the length of the line segment
between the two stubs is l 2 = (3/8)/..... To facilitate the analysis we start from the input
side of the tuner and work backward to the load end.
For a perfect match it is required that Z in = Z 0 and therefore YA = 1. Since it is
assumed that the lines are lossless, the normalized admittance y 8 = YA- jb82 is
located somewhere on the constant conductance circle g = 1 in the Smith Chart. Here
bs2 is the susceptance of the stub and ls2 is the associated length. For an 12 = (3/8)A.
line the g = 1 circle is rotated by 2J3l2 = 31t/2 radians or 270 toward the load (i.e., in
counter-clockwise direction, as depicted in Figure 8-27). The admittance Yc (being the
series connection of ZL with line / 1 in parallel to stub ls1) needs to reside on this rotated
g = 1 circle (called the Yc circle) in order to ensure matching.
By varying the length of the 15 1 stub we can transform point Yv in such a way
that the resulting y c is indeed located on the rotated g = 1 circle. This procedure can
be done for any load impedance except for the case when point yD (i.e., the series connection of Z L and line l 1 ) is located inside the g = 2 circle. This represents the forbidden region that has to be avoided. To overcome this problem in practical
applications, commercial double-stub tuners usually have input and output transmission
lines whose lengths are related according to 11 = / 3 A.I 4 . In this case, if a particular
load impedance cannot be matched, one simply connects the load to the opposite end of
the tuner, which moves y D out of the forbidden region.
The following example demonstrates the computation of the stub lengths to
achieve matching for a specific load impedance.
442
Chapter 8 Matching and Btasfng Networks
Figure 8-27
Smith Chart analysis of a double-stub matching network shown in

Figure 8-26.
----------------------------RF~~
Example 8-10: Design of a double-stub matching network
It is assumed that in the double-stub matching network shown in
Figure 8-26 the lengths of the transmission lines are
l 3 = l 2 = 3/.../8 and l 1 = A./8. Find the lengths of the short-circuit stubs that match the load impedance ZL = (50+ j50)Q to a
50 Q input impedance. The characteristic line impedance for all
components is Z 0 = 50 .Q.
443
Solution:
First the nonnalized admittance y D has to be determined and checked that it does not fall inside the forbidden region.
Using the Smith Chart (see Figure 8-28), we find YD = 0.4 + j0.2.
Since g D < 2 , we are assured that the admittance y D does not fall
into the forbidden region. Next we plot the rotated g = 1 circle as
explained previously. This allows us to fix the intersection of the
rotated g = 1 circle with the constant conductance circle that passes
through the point y D . The intersection point gives us the value of
y c . In fact, there are two intersection points that yield two possible
solutions. If we choose y c = 0.4 - j 1.8 , then the susceptance of the
first stub should be jb 51 = Yc- YD = -j2, which pennits us to
detennine the length of the first short-circuit stub: ls 1 = 0.074A.
Figure 8-28
Double-stub tuning network design for Example 8-10.
444
Rotating Yc by 12 = 3A./8 we find y 8 = 1 + )3, which

means that we have to make the susceptance of the second stub
equal to jb 52 = - }3 so that Yin = YA = 1. Using the Smith Chart,
we find that the length of the second stub is l s 2 = 0.051 A.
In some practical realizations the stubs are replaced by varactor diodes. This allows an electronic tuning of the diode capacitances and thus the shunt admittances.
8.3 Amplifier Classes of Operation and Biasing Networks

An indispensable building block in any RF circuit is the active or passive biasing
network. The purpose of biasing is to provide the appropriate quiescent point for the
active devices under specified operating conditions and maintain a constant setting irre~
spective of transistor parameter variations and temperature fluctuations.
In the following section we introduce a general analysis of the different classes of
amplifier operation. This will enable us to develop an understanding of how BJT and
FET need to be appropriately biased.
8.3.1
Classes of Operation and Efficiency of Amplifiers
Depending on the application for which the amplifier is designed, specific bias
conditions are required. There are several classes of amplifier operation that describe
the biasing of an active device in an RF circuit.
In Figure 8-29 the transfer function characteristic of an ideal transistor is displayed. It is assumed that the transistor does not reach saturation or breakdown regions
and in the linear operating region the output current is proportional to the input voltage.
The voltage V* corresponds either to the threshold voltage in case of PETs or the baseemitter built-in potential in case of BJTs.
The distinction between different classes of operation is made based upon the socalled conduction angle, which indicates the portion of the signal cycle when the cur~
rent is flowing through the load. As depicted in Figure 8-29(a), in Class A operation the
current is present during the entire output signal cycle. This corresponds to a
e A = 360 conduction angle. If the transfer characteristic of the transistor in the linear region is close to that of a linear function, then the output signal is an amplified replica of the input signal without suffering any distortion. In practical circuits, however,
Amplifier Classes of Operation and Blaslng Networks
445
Ideal transfer
k" function
Quiescent
point
/
Cut-off region
(b) Class B
(a) Class A
Quiescent
Quiescent
point
point
(c) Class AB
Figure 8-29
(d) Class C
Various classes of amplifier operation.
there is always a certain degree of nonlinearity present which results in a distorted output signal of the amplifier.
In Class B [Figure 8-29(b)] the current is present during only half of the cycle,
corresponding to a 8 8 = 180 conduction angle. During the second half of the cycle,
the transistor is in the cut-off region and no current flows through the device. Class AB
[Figure 8-29(c)] combines the properties of the classes A and B and has a conduction
angle 8 AB ranging from 180 to 360. This type of amplifier is typically employed
when a high-power "linear" amplification of the RF signal is required.
Chapter 8 Matchfng and Bluing Netwolb
446
In a Class C amplifier [Figure 8-29(d)l we have a nonzero current for less than
half of the cycle (i.e., the conduction angle is 0 < Elc < 180 ). This results in maximum
distortion of the output signal.
A logical question that arises is why are not all amplifiers operated in Class A
since this mode delivers the least signal distortion? The answer is directly linked to the
amplifier efficiency. Efficiency, 11 , is defined as the ratio of the average RF power PRF
delivered to the load over the average power P s supplied by the source, and is usually
measured in percent:
PRF
Tl- p-100%
(8.22)
The theoretical maximum efficiency of the Class A amplifier is only 50%, but the
efficiency of Class C can reach values close to 100%. Fifty percent efficiency of Class
A amplifiers means that half of the power supplied by the source is dissipated as heat.
This situation may not be acceptable in portable communication systems where most
devices are battery operated. In practical applications, designers usually choose the
class of operation that gives maximum efficiency but still preserves the informational
content of the RF signal.
In the following example we derive the maximum theoretical efficiency 11 of the
amplifier as a function of conduction angle.
----------------------------~~W4
Example 8-11: Amplifier efficiency computation
Derive the general expression for the amplifier efficiency 11 as a
function of conduction angle 8 0 . List the values of 11 for both
Class A and Class B amplifiers.
Solution:
The electrical current through the load for a conduction angle of 8 0 has a wavefonn shown in Figure 8-30(a), where
the cosine current amplitude is given by I 0 .
Similarly, the power supply current Is has a maximum value
of I 0 plus the quiescent current I Q :
Is
= I Q + 10 cose
(8.23)
447
1t
, ~1
\
I (
2 1t
r,
E>o ) \
\
\ I
\ I
I
\I
\J
31t
'
\
(a) Load current waveform at the output of the transistor
Is
I Q + /0
I Q .... ........... .............................. ..................
(b) Corresponding power supply current waveform
Figure 8-30 Load and power supply current waveforms as a function of

conduction angle.
The value of the quiescent current necessary to ensure the specified

conduction angle 8 0 can be found from (8.23) by setting Is to zero
at E> = E>0 / 2 :
(8.24)
The average power supply current is then computed as an integral
over the conduction angle ranging between the limits of
E> = -8 0 /2 and E> = 8 0 / 2 ; that is,
0012
(Is) =
_!_J lsdE>
21t -0 1 2
0
= -
[e cos( 2
21t
10
80
) - 2sin(
Thus, the average power from the power supply is
8 0
)]
(8.25)
Chapter 8 Matching and Blaslng Networkt
448
(8.26)
where V cc is the supply voltage.
Since the voltage on the load changes together with the current, the average RF power is computed as an integral of the product
of load current and load voltage:
1
PRF = 2 1t
9 12
- 0 0 /2
10 V c ccos 2 E>d8 ==
I oV cc
.
(80 - s1nE> 0 ) (8.27)
4 1t
Dividing (8.27) by (8.26), we find an amplifier efficiency

E>0 - sinE> 0
11 = 2[0 cos (E> / 2)- 2sin(E> /2)]
0
0
0
where the conduction angle 8 0 is measured in radians.
The graph of 11 as a function of the conduction angle
8 28
( )
e0
is
100~=-~--~--~----~--~--~----~
95
90
'$. 85
r:>-. 80 ..!J.~.?.~.:?.~--- ..
5 75
fE 70
l.t:l
65
60
55
eo= 180
50 ~--~--~----~--~--~--~--~~
Figure 831
50
100
150
200
250
Conduction angle, e0
300
350
Maximum theoretical efficiency of an ideal amplifier as a function of

conduction angle.
Substituting 8 0 = 27t into (8.28), we find that the efficiency

of a Class A amplifier is indeed 50%. To determine the efficiency of
a Class B amplifier, we simply use the conduction angle E> 0 = 7t in
(8.28), which yields
449
sin1t
1t
=
= 0.785
2 [1tcos (1t/ 2 ) - 2sin(1t/ 2)]
4
That is, Class B yields an efficiency of 78.5%.
llB =
1t-
Efficiency is an important design consideration when dealing

with low power consumption, as required, for instance, in personal
communication systems, where battery lifetime must be maximized.
8.3.2
Bipolar Transistor Biasing Networks
There are generally two types of biasing networks: passive and active. Passive (or
se]f..biased) networks are the simplest type of biasing circuits and usually incorporate a
resistive network, which provides the appropriate voltages and currents for the RF transistor. The main disadvantages of such networks are that they are very sensitive to
changes in transistor parameters and that they provide poor temperature stability. To
compensate for these drawbacks active biasing networks are often employed.
In this section we consider several network configurations for biasing RF BJTs.
1\vo possible topologies are shown in Figure 8-32.
The combination of the blocking capacitor C8 and the RFC connected to the base
and collector terminals of the transistor in Figure 8-32 serve the purpose to isolate the
RF.n
(a)
Figure 8-32
(b)
Passive biasing networks for an RF BJT in common-emitter

configuration.
450
RF signal from the DC power source. At high frequencies, the RFCs are usually
replaced by quarter wave transmission lines that convert the short-circuit condition on
the C 8 side to an open-circuit condition on the transistor side.
The following example discusses how to compute the resistors for the two biasing
networks shown in Figure 8-32.
----------------------~~&MW4
Example 8-12: Design of passive biasing networks for a BJT in

common-emitter configuration
Design biasing networks according to Figures 8-32(a) and (b) for
the BJT settings of I c = 10 rnA, V CE = 3 V, and V cc = 5 V.
Assume that the transistor has a J3 = 100 and V BE = 0.8 V.
Solution:
As seen in Figure 8-32(a), the current I 1 through

resistor R 1 is equal to the sum of the collector and base currents.
Since I 8 = I ciJ3, we obtain
I1
= lc+/ 8
= I c( l + p- 1)
= 10.1
rnA
The value of R 1 can be found as
VCC -VCE = 198 fl

I]
Similarly, the base resistor R2 is computed as
R2
V CE- V BE
I
B
V CE- V BE
I f A
C
J-1
= 22 kQ
For the circuit in Figure 8-32(b) the situation is slightly more complicated. Here we have the freedurn of choosing the value of the
voltage potential V x and the current I x through the voltage divider
resistor R 2 . Arbitrarily setting V x to 1.5 V, we determine the base
resistor R3 to be
The value of I x is usually chosen to be 10 times larger than I 8 .

Therefore, I x = 10/8 = 1 rnA and the values of the resistances
for the voltage divider are computed as
Amplifier Classes of Operation and Blasing Networks
451
- 1.5 kQ and R2
= V c c-V x
lx +Is
- 3.18 kQ
Finally, the collector resistor is found as

R4 = (Vcc-VcE)Ilc ::: 200 Q
The freedom of selecting particular voltages and currents is in

practice restricted by the need to choose electric settings that result
in standardized resistance values.
An example of an active biasing network for a BJT in common-emitter configuration is shown in Figure 8-33. Here we employ a low-frequency transistor Q 1 to provide
the necessary base current for the RF transistor Q 2 . The resistor RE1 connected to the
emitter of the transistor Q1 improves stability of the quiescent point. If transistors Q1
and Q 2 have the same thermal properties, then this biasing network also results in good
temperature stability.
Example 8-1 3 illustrates the determination of the component values for the active
biasing network depicted in Figure 8-3 3.
Vc:c
RF'in
Figure 8-33
Active biasing network for a commonemitter RF BJT.
452
-----------------------------~&uM~
Example 8-13: Design of an active biasing network for a BJT
transistor in common-emitter configuration
Design a biasing network as shown in Figure 8-33 for
lc2 = 10 rnA , V CEZ = 3 V , and V c c = 5 V. Assume that both
transistors have ~ = 100 and V BE = 0.8 V.
Solution:
Similar to the previous example we have several

degrees of freedom in this biasing network. First, we can pick the
value for a voltage potential V c 1 at the collector of transistor Q1
Second, we are free in our choice of the collector current through
Q 1 . Since I 82 should not be affected by current fluctuations in I c 1 ,
we choose I Cl such that I ct = 10I82 (i.e., I ct = 1 rnA). Then
the current I 1 through resistor Rc 1 is composed of collector current
I Cl and two base currents I 81 and I 82 ; that is,
I 1 = I c 1 + I 8 I + I 8 2 = I c 1( 1 +
Assuming V c 1
R82 =
~~ 1 ) + I C2 I ~2
== 1.11 rnA
= 3 V , we find
V C l - V 8E2
182
= 22 k.Q and Rc 1
- 1.8 kQ
Another degree of freedom is the choice of voltage V E 1 at the emitter terminal of the transistor Q1 . Setting V E 1 to 1 V, we find
R8 l
Vct-V8E1 -VE1
[8 1
== 120 kil
and
REt
V El
l e t - IBt
= 1.11 kO
Finally, the collector resistor Rc 2 is determined to be

Rc2 = (Vcc- v CE2)1 I c2 = 200 n
Although active biasing offers a number of performance
advantages over passive networks, certain disadvantages also arise:
Amplifier Claues of Operation and Blaslng Networks
453
specifically, additional circuit board space, possible layout complications, and added power requirements.
Another active biasing network for a BJT in common-emitter configuration is

shown in Figure 8-34. Here diodes D 1 and D 2 provide a fixed reference for the voltage
drop across the base-emitter junctions of both transistors. Resistor R 1 is used to adjust
the biasing current to the base of transistor Q1 and R2 limits the range of this adjustment. Ideally, for temperature compensation, transistor Q1 and one of the diodes
should remain at ambient temperature, whereas the second diode should be placed on
the same heat-sink as RF transistor Q2 .
Figure 834
Active biasing network containing low-frequency transistor and two

diodes.
As a final remark, it is important to point out that in all biasing networks the operational conditions (common-base, common-emitter, or common-collector) of the transistor at RF frequencies are entirely independent of the DC configuration. For instance,
we can take an active biasing network, shown in Figure 8-33, and modify it for common-base RF operation, as seen in Figure 8-35.
At DC all blocking capacitors represent an open circuit and all RFCs behave like
short circuits. Therefore, this biasing network can be redrawn as shown in Figure
8-36(a), indicating the common-emitter configuration. However, at RF frequency all
blocking capacitors become short circuits and all RFCs behave like open circuits. This
transforms the biasing network into a common-base mode, as depicted in Figure 8-36(b).
454
Figure 835
Modification of the active biasing network shown in Figure 8-33 for

a common-base RF operation.
::Cs
....
....... ~ f .. -.
~
::-~
::.'
0 ;:
.......
u::u
>.:::::
:::::>
r::::::;.RFC
~-::::: RFC
..
~~:
.---o
RFC
RFC
......, ....
Ca
1
.......
.............
(a) DC equivalent circuit
(b) RF equivalent circuit
Figure 836 DC and RF equivalent circuits for the active biasing network in
Figure 8-35.
Amplifier Classes of Operation and Blasing Networks
8.3.3
4!)5
Field Effect Transistor Biasing Networks
The biasing networks for field effect transistors are in many ways very similar to
the BJT networks covered in the previous section. One key distinction is that MESFET
usually require a negative gate voltage as part of the bias conditions.
The most basic passive bipolar biasing network for PETs is shown in Figure 8-37 .
...
RF;n
t----....,_-~RF'out
o----41~---t
Figure 8-37
t------,
Bipolar passive biasing network for FETs.
The main disadvantage of such a network is the need of a bipolar power supply
for V G < 0 and VD > 0 . If such a bipolar power supply is unavailable one can resort to
_a strategy where instead of the gate, the source terminal of the transistor is biased. The
gate in this case is grounded. Two examples of such networks are shown in Figure 8-38.
RF.n
o----
---..,...._--1
1------11---..
RFC
Vs
Figure 8-38 Unipolar passive biasing networks for FETs.
The temperature compensation of the FET biasing networks is typically accomplished through the use of thermistors.
456
Chapter 8 Matching and Staalng Networks
8.4 Summary
The material covered in this chapter is geared toward providing an understanding
of two key issues encountered in any RF/MW system: interfacing various components
of different impedance values, and suitably biasing the active devices depending on
their class of operation.
To ensure optimal power transfer between systems of different impedances, we
investigate at first two-element L-type matching configurations. In the context of twoport network analysis, the conjugate complex matching requirement at the input and
output ports results in optimal power transfer at a particular target frequency. The technique is simple and can be compared with the design of a bandpass or bandstop filter.
Care must be exercised in selecting a suitable L-type network to avoid the forbidden
regions for which a given load impedance cannot be matched to the desired input
impedance. From the knowledge of the network transfer function, the loaded quality
factor
fo
QL
= BW
and the simpler to compute nodal quality factor

Q =
fXsf
Rs
fBpf
Gp
can be utilized as a measure to assess the frequency behavior of the matching networks.
Unfortunately, L-type networks do not allow any flexibility in conditioning the frequency response and are therefore mostly used for narrow band RF designs. To affect
the frequency behavior, a third element must be added, resulting in T- and Pi-type networks. With these configurations a certain nodal quality factor, and indirectly a desired
bandwidth, can be implemented.
While the lumped element design is appropriate at low frequencies, distributed
transmission line elements must be employed when the frequency extends into the GHz
range. The hybrid configurations of using series connected transmission line elements
and shunt connected capacitors are very attractive for prototyping since the location and
value of the capacitors can easily be varied. If the capacitors are replaced by open- and
short-circuit transmission lines, one arrives at the single- and double-stub matching
networks.
Depending on the application (for instance, linear small signal or nonlinear large
signal amplification), various classes of transistor amplifiers are identified. The classification is done by computing the RF to supply power ratio, known as efficiency:
457
Further Reading
PRF
1l = -100%
Ps
which can be expressed in tenns of the conduction angle 8 0 quantifying the amount of
load current flow through the relation
8 0 - sinE> 0
2[80 cos(80 /2)- 2 sin(80 /2)]
For instance, Class A offers the highest linearity at the expense of the lowest efficiency
of 50%, whereas Class B co mpromises linearity but improves efficiency to 78.5%.
Once the class of operation is identified, a biasing network is chosen to set the
appropriate quiescent point of the transistor. Passive biasing networks are normally
easy to implement. However, they are not as flexible as biasing networks involving
active devices. The biasing not only sets the DC operating conditions but must also
ensure isolation of the RF signal through the use of RFCs and blocking capacitors.
Further Reading
W. A. Davis, Microwave Semiconductor Circuit Design, Van Nostrand Reinhold Company, New York, 1984.
N. Dye and H. Granberg, Radio Frequency Transistors: Principles and Practical Applications, Butterworth-Heinemann, 1993.
G. Gonzalez, Microwave Transistor Amplifiers: Analysis and Design, Prentice Hall,

P. Horowitz and W. Hill, The Art of Electronics, Cambridge University Press, Cambridge, UK, 1993.
D. Pozar, Microwave Engineering, John Wiley & Sons, New York, 1998.
,. P. Rizzi, Microwave Engineering: Passive Circuits, Prentice Hall, Englewood Cliffs,

NJ, 1988.
Problems
8.1
Obtain the "forbidden" regions for the two-element matching networks

shown in Figures 8-l(c)-(f). Assume that the load is matched to the normalized input impedance (i.e., Zin = 1 ).
458
8.2
Use the analytical approach and design a two-component matching network

that matches the ZL = (100 + j20)Q load impedance to a gtven
Z s = ( l 0 + j25 )0. source, at the frequency of f 0 = 960 MHz.
8.3
Develop a two-component matching network for a ZL = (30- j40)Q load

and a 50 n source. How many network topologies exist that can be used?
Find the values of the components if a perfect match is desired at
fo = 450 MHz.
8.4
Repeat Problem 8.3 for a Z L

quency of f 0 = 1.2 GHz.
8.5
Measurements indicate that the source impedance in Problem 8.3 is not

purely resistive but has a parasitic inductance of Ls = 2 nH. Recompute the
values for the matching network components that take into account the presence of Ls.
8.6
A load ZL = (20 + jlO)Q consisting of a series R-L combination is to be

matched to a 50 Q microstrip line at f 0 = 800 MHz. Design two two-ele-
= (40 + j 10) 0.
load and a matching fre-
ment matching networks and specify the values of their components. Plot a
frequency response for both networks and find the corresponding
bandwidths.
8.7
In Example 8-5 aT-network is discussed that matches a load impedance of

ZL = (60- j30)Q to an input impedance of Zin = (10 + j20)Q at
I GHz, under the constraint that Qn does not exceed the factor of 3. Stepby-step go through this design and identify each point in the Smith Chart in
terms of its impedance or admittance values. Verify the final results shown in
Figure 8-16.
8.8
Go through Example 8-6 and find each point in the Smith Chart shown in
Figure 8-17 and verify the final network components depicted in Figure
8-18.
8.9
Repeat the Pi-type matching network design in Example 8-6 for a nodal
quality factor of Qn = 2.5. Plot Zin(/) for this Qn value and compare it
against the Qn = 2 design in Example 8-6. As frequency range, choose
1GHz</<4GHz.
Problems
459
8.10
Design two T-type matching networks that transform a Z L

100 .Q load to
an Zin = (20- j40).Q input impedance at a nodal quality factor of
Qn = 4 . The matching should be achieved at f 0 = 600 MHz.
8.11
...
Design two Pi-type matching networks for the same conditions as in Problem 8.10.
8.12
To achieve matching conditions for a specified Qn, the circuit designer has
to use more than two or three elements in the matching network. Using a
graphical approach, design a multisection matching network that transforms
Z L = 10 0 into Z s = 250 Q at f 0 = 500 MHz while maintaining a
nodal quality factor of Qn = 1 . The multisection matching network should
consist of a series of two-element sections each of which is a "series inductor, shunt capacitor" combination [see Figure 8-l(h)].
8.13
For an increased frequency off0 = 1 GHz it was decided that the network
designed in Problem 8.12 should be replaced by a combined matching network shown in Figure 8-19. Detemrine the total number of capacitors and
transmission line sections necessary to achieve matching and find the values
of all components in the network.
- 8.14 Using the design from Example 8-7, find the length and width of each transmission line if an FR-4 substrate with dielectric constant of er = 4.6 and
height of h = 25 mil is used. Find the maximum deviation of the input
impedance of the matching network if the capacitor that is used in the circuit
has a +10% tolerance and the automatic component placement equipment
has a 2 mil precision (i.e., the capacitor can be placed within 2 mil of the
intended position).
8.15
In Example 8-7 it is argued that open-circuit stubs can be replaced by shortcircuit ones if the length is increased by a quarter wavelength. Matching is
achieved only for a single frequency, and over a broader frequency range the
network response can significantly differ from the target impedance values.
Design a single-stub matching network that transforms a ZL = (80 + j20)0
load impedance into a Zin = (30- jlO)Q input impedance. Compare the
frequency response over the +0.8 f 0 frequency range for two different realizations of the matching network: open-circuit stub, and using an equivalent
short-circuit stub. Assume that the matching frequency is f 0 = 1 GHz and
the load is a series combination of resistance and inductance.
460
Chapter 8 Matching and Biasing Networks
8.16 Using the matching network shown in Figure 8-23(b), find the stub length
l s, the characteristic line impedance ZoL, and the transmission line length
lL such that the ZL = (80- j40)Q load impedance is matched to 50 n
source. Assume that the characteristic impedance of the stub is
Z 0 s = 50 Q.
8.17
For a double-stub tuner shown in Figure 8-26 with parameters 11 = A./8,

12 = 5A./8, and 13 = 3A./8, detennine to which end of the tuner a
ZL = (20- j20)Q load has to be connected and find the length of the
short-circuited stubs such that the load is matched to a 50 Q line. Assume
that all stubs and transmission lines in the tuner have a 50 n characteristic
impedance.
8.18
Discuss a circuit configuration that replaces in the previous problem the stub
tuners with varactor diodes in series with inductors. Choose the appropriate
inductances if the varactor diodes can change their capacitances in the range
from 1 pF to 6 pF. For a frequency of 1.5 GHz discuss the tuning capabilities
in terms of possible load impedance variations.
8.19
An ideal amplifier has a transfer function given by the equation

30(Vin- V*),
vout = { 0
where V*
'
Vin 2: V*
v.m< V*
= 60 mV. Find the quiescent point ( VQ and IQ ) and the corre-
sponding maximum efficiency such that the amplifier is operated in the AB

class and has conductance angle of E>0 = 270. Assume that the input signal is a sinusoidal voltage wave of 100 mV amplitude.
Problems
461
8.20 Find the component values for a low-GHz range biasing network for a BIT
with bypassed emitter resistor R 3 , as shown below:
Vee
Assume that the power supply voltage is V cc = 12 V and the transistor has
the following parameters: I c = 20 rnA, V CE = 5 V, ~ = 125, and
VBE = 0.75
v.
8.21
For stability purposes a feedback resistor RF = 1 kQ has been added

between base and collector of the transistor in the biasing network shown in
Figure 8-32(b). Compute the values of all resistors in the biasing network if
the following biasing conditions must be satisfied: supply voltage of
V cc = 5 V, collector current of I c = 10 rnA, and collector-emitter voltage of V CE = 3 V. Assume that the transistor has a ~ = 100 and a
VBE = 0.8 V.
462
8.22 Design a biasing network (shown in the following figure) for I c 2 = 10 rnA,
V cE2 = 3 V, and V cc = 5 V. Assume that ~ 1 = 150, ~ 2 = 80, and both
transistors have V BE = 0.7 V.
Vee
8.23
Redraw the active biasing network shown in Figure 8-34 for a common-base
and a common-collector operating mode, respectively.
8.24
For the passive FET biasing network shown in Figure 8-38, find the value of
the source resistance Rs if V cs = -4 V, V DS = 10 V, and the drain current is given to be I D = 50 rnA.
,.
CHAPTER
RF Transistor Antplifier
Designs
Amplifier designs at RF differ significantly from

the conventional low-frequency circuit approaches and consequently require special
considerations. In particular, the fact that voltage and current waves impinge upon the
active device necessitates appropriate matching to reduce the VSWR and avoid undesirable oscillations. For this reason a stability analysis is usually the first step in the design
process and, in conjunction with gain and noise figure circles, is a basic ingredient
needed to develop amplifier circuits that meet the often competing requirements of
gain, gain flatness, output power, bandwidth, and bias conditions.
This chapter expands upon the material covered in Chapters 2 and 3, where power
relations of terminated transmission lines are investigated. However, unlike the passive
circuit presentations, Chapter 9 deals with active devices where gain and feedback considerations assume central importance. Issues such as power gain, unilateral and bilateral circuit designs and their graphical display in the Smith Chart constitute the starting
point of an extensive analysis into quantifying high-frequency transistor amplifier performance. The reader will note the flexibility of the Smith Chart, which allows constant
gain, VSWR, and stability circle displays to be superimposed over the reflection coefficient and impedance representation discussed in Chapter 3. Moreover, even a noise analysis can be conducted by converting the noise figure of an amplifier into circles that are
displayed in the Smith Chart.
After covering the basic design tools, Chapter 9 also investigates various types of
power amplifiers and their characteristics such as gain flatness, bandwidth, and intermodular distortion as well as the differences between single- and multistage amplifiers.
463
464
Chapter 9 RF Translator Amplifier Designs
9.1 Characteristics of Amplifiers

Perhaps the most important and complex task in analog circuit theory is the amplification of an input signal through either a single or multistage transistor circuit. A
generic single-stage amplifier configuration embedded between input and output
matching networks is shown in Figure 9-1 .
.:..
~
"":'
RF
source
~ncl11)
..:..
Input
Matching
Network
(IMN)
A.
~~~
.i
"!"
J7n
Figure 9-1
~
1'
~~~
l
Output
Matching
Network
(OMN)
"!"
DC bias
A.
"":'
~-!11 >
Load
"!"
rout
Generic amplifier system.
Input and output matching networks, discussed in Chapter 8, are needed to reduce
undesired reflections and thus improve the power flow capabilities. In Figure 9-1 the
amplifier is characterized through its S-matrix at a particular DC bias point. In tenns of
performance specifications, the following list constitutes a set of key amplifier
parameters:
Gain and gain flatness (in dB)
Operating frequency and bandwidth (in Hz)
Output power (in dBm)
Power supply requirements (in V and A)
Input and output reflection coefficients (VSWR)
Noise figure (in dB)
In addition, one often must consider such parameters as intermodular distortion (IMD)
products, harmonics, feedback, and heating effects, all of which can seriously affect the
amplifier performance.
To approach the amplifier design process systematically, we need first to establish
a number of definitions regarding various power relations. This is followed by several
important analysis tools required to define stability, gain, noise, and VSWR performance. The common denominator of all four topics is that they can be expressed as circle equations and displayed in the Smith Chart.
Ampltfler Power Relations
465
9.2 Amplifier Power Relations

9.2.1
RF Source
There are various power gain definitions that are critical to the understanding of
how an RF amplifier functions. For this reason, let us examine Figure 9-1 in terms of its
power flow relations under the assumption that the two matching networks are included
in the source and load impedances. This simplifies our system to the configuration
shown in Figure 9-2(a). The starting pojnt of our power analysis is the RF source connected to the amplifier network. For the convention depicted in Figure 9-2 we recall our
signal flow discussion in Section 4.4.5 fsee (4.82) and (4.83)] and write for the source
voltage
(9.1)
(a) Simplified schematics of a single-stage amplifier
""
,.....
,, s
rsJI\
s 22
II
'~
1
'
...
"'"
'"'
"'
'
(b) Signal flow graph
Figure 9-2
Source and load connected to a
single~stage
The incident power wave associated with b 1' is given as
amplifier network.
466
Chapter 9 RF Transistor Amplifier Designs
(9.2)
which is the power launched toward the amplifier. The actual input power Pin observed
at the input terminal of the amplifier is composed of the incident and reflected power
waves. With the aid of the input reflection coefficient rin we can therefore write:
2
Pin = Pinc(I-jrinl ) =
lbsl
II- rinfsl
(9.3)
2(1-jrinl )
The maximum power transfer from the source to the amplifier is achieved if the
* or, in terms of the reftecinput impedance is complex conjugate matched ( Z in = Z s)
* Under maximum power transfer condition, we define the
tion coefficients, if rin = r s.
available power P A as
PA
= p.mI
rin
= rs
Jbsl
=
11- rinrsl 2
-21
(9.4)
This expression makes clear the dependence on

2
from (9.2) and (9.4) that Pine = JbsJ /2.
9.2.2
r s. If rin
= 0 and
rs 0
it is seen
Transducer Power Gain
We can next investigate the transducer power gain Gr, which quantifies the
gain of the amplifier placed between source and load.
G
T
or with PL
power delivered to the load

_ P_L
available power from the source
PA
= ~Jb 2 j 2 ( 1- fr Lj 2 )
PL
Gr = -
pA
we obtain
jh2j
= -
Jbsl2
( 1 -I r Ll
)( 1 -jrsl 2)
(9.5)
In this expression, the ratio b 2 1 b5 has to be determined. With the help of our signal
flow discussion in Section 4.4.5 and based on Figure 9-2, we establish
Sztat
b2 = --~
1-Sz2rL
(9.6a)
467
bs = [ 1The required ratio
(su+ ~~~:~:)rs}1
(9 .6b)
is therefore given by
bz
s 21
bs
(l-Suf's)(l-S22rL) - S2tsl2rLrs
--
(9.7)
Inserting (9.7) into (9.5) results in
GT -
2
2
2
-1r
L) ))S21) 0 - 1rs) )
-------'----'----'--.:.....---'----'----
(1
(9.8)
j(l-S]JrS)(l-S22rL) - SzlS12r LrS] 2

which can be rearranged by defining the input and output reflection coefficients (see
Problem 9.2)
(9.9a)
s12s 21rs
r out =
s22 + 1 - s11r s
(9.9b)
With these two definitions, two more transducer power gain expressions can be derived.
First, by incorporating (9.9a) into (9.8), it is seen that
0-jrL\z)\S2I\z(l - j1s)z)
Gr =
jl - ['S['inl
jl - 522r Ll
(9.10)
Second, using (9.9b) in (9.8) results in the expression
- (1 - Jr L]2)JSz1J z(l- jr s]2)

Gr 2
2
Jl - r L ['outI Jl - S 11 r sl
(9.11)
An often employed approximation for the transducer power gain is the so-called unilateral power gain, Gru' which neglects the feedback effect of the amplifier ( S12 ::;;; 0 ).
This simplifies the form (9.11) to
Gru =
2
2
(1- )rL) )jS2J(\1 - Jrs) )
2
II - r L Szzl /1 - s11 r sl
(9.12)
As discussed in Section 9.4.1, equation (9 .12) is often used as a basis to develop approximate designs for an amplifier and its input and output matching networks.
468
9.2.3
Additional Power Relations
The transducer power gain is a fundamental expression from which additional important power relations can be derived. For instance, the available power gain for load
* ) is defined as
side matching ( r L = rout
= G
G
A
T 1
L
= 1 out
*
= power available from the amplifier

power available from the source
or, with the aid of (9 .11 ),
(9.13)
Further, the power gain (operating power gain) is defined as the ratio of the power
delivered to the load to the power supplied to the amplifier.
G =
power delivered to the load

PL
PL PA
PA
power supplied to the amplifier = Pin = P A . Pin = Gr Pin
Combining (9.3), (9.4), and (9.10), we find
G=
(I -[rL[2)[S21[2
----~--~~~--
(1
-[rinl )1l - S22rLl
(9.14)
It is interesting to note that (9 .14) can be obtained from (9 .1 0) by setting r s = fin*

since in this case Pin= Pk The following example goes through the computation of
some of these expressions for an amplifier with given S-parameters.
----------------------------RF~~
Example 9-1: Power relations for an RF amplifier
An RF amplifier has the followingS-parameters: S11 = 0.3L-70,
S21 = 3.5L85, S 12 = 0.2L-10, and S22 = OAL-45. Furthermore, the input side of the amplifier is connected to a voltage
source with V s = 5 V L0 and source impedance Z s = 40 n. The
output is utilized to drive an antenna which has an impedance of
Z L = 73 Q . Assuming that the S-parameters of the amplifier are
469
= 50 n
measured with reference to a Z0

find the following quantities:
characteristic impedance,
(a) transducer gain Gr, unilateral transducer gain Gru, available

gain G A , operating power gain G , and
(b) power delivered to the load P L , available power PA, and incident power to the amplifier Pine.
Solution:
First we find the source and load reflection coefficients
assuming a Z 0 = 50 Q characteristic impedance:
fs =
Zs-Zo
Zs + Zo
= -0.111
ZL -Zo
= ZL + zo =
and fL
0.187
Next, the input and output impedances, as given in (9.9a) and (9.9b),
are determined:
rin
s2tst2rL
r
= s)] + 1
r out = S 22
s22
s12s21rs
+I_ S
11
.
= 0.146- ]0.151
.
= 0.265- ]0.358
Substituting the obtained values along with the S-parameters into

(9.11), (9.12), (9.13), and (9. 14), the transducer gain Gr , unilateral
transducer gain Gru, available gain G A, and operating power gain
G are computed as follows:
= (l-lriJIS~1 1 < 1 -Irsl:l = 12.56

II - r L r outl [I - s11 r sl
2
GT
Gru =
(1-lrr[2)[S2t[2(1-[rsl2)
II - r Ls22! 2 [1 - s11 r sl2
2
GA =
= 12.67 or 11.03 dB
1 1 <l-[rsl
521
It - Ir oud 111- sII r sl

2
or 10.99dB
2
21
)[S 1
14.74 or 11.68 dB
= (1-[r2 L[
=
2
It -[rinl 1I1 - s22fd
13.74 or 11.38 dB
470
Chapter 9 RF Transistor Amplifier Design
Using (9.2) in conjunction with (9.1) allows us to find the incident

power flow into the amplifier:
2
Jbs]
1 Zo
]Vs]
Pine = 211- rinri = 2(Zs + Zo)2jl- rinri = 74.7 mW
Often Pine is expressed in dBm as
Pinc(dBm) = lOlog[Pinc/(1 mW)] = 18.73 dBm
Similarly, from (9.2) we find the available power to be
PA = 78.1 mW or PA = 18.93 dBm. Finally, the power delivered to
the load is the available power multiplied by the transducer gain.
This results in PL = P AGT = 981.4 mW, or, expressed in dBm,
PL(dBm) = PA(dBm) + Gr(dB) = 29.92 dBm
It is interesting to point out that the unilateral power gain

often matches the actual transducer power gain very closely. As discussed further, the use of the unilateral amplifier gain significantly
simplifies the amplifier design task.
9.3 Stability Considerations

9.3.1
Stability Circles
One of the first requirements that an amplifier circuit must meet is a stable performance in the frequency range of interest. This is a particular concern when dealing with
RF circuits, which tend to oscillate depending on operating frequency and termination.
The phenomenon of oscillations can be understood in the context of a voltage wave
along a transmission line. If 0 ] > 1 , then the return voltage increases in magnitude
(positive feedback) causing instability. Conversely, Ir 0 ] < 1 causes a (Hminished return
voltage wave (negative feedback).
Let us regard the amplifier as a two-port network characterized through its Sparameters and external terminations described by r L and r s. Stability then implies
that the magnitudes of the reflection coefficients are less than unity. Namely,
Jr
Ir L] < 1, Ir sl < 1
(9.15a)
Stability Considerations
471
lroutl
where~
Su-rv~ < 1
l-S22rL
(9.15b)
= Szz- fs~ < 1
(9.15c)
1-S 11 rs
= SuS22 - S 12S 21 has been used tore-express (9.9a) and (9.9b). Since the S-
parameters are fixed for a particular frequency, the only factors that have a parametric
effect on the Stability are r L and f s .
In terms of the amplifier's output port, we need to establish the condition for
which (9.15b) is satisfied. To this end the complex quantities
S 11 =
S~1 + jS~ 1 , S22 = S~2 + jS;2, d = AR + jd
1
,
rL
= ~ + jf~
(9.16)
are substituted into (9 .15b), resulting after some algebra in the output stability circle
equation
R
R2
(rL - Cout) + (rL- Cout)
= rout
(9.17)
where the circle radius is given by
IS12S211
rout :::
llsz2l
(9.18)
- IAI 1
and the center of this circle is located at

R
Cout
= Cout + jCout =
* *
(Szz- S11A)
2
ISzzl -IAI
(9.19)
as depicted in Figure 9-3(a). In terms of the input port, substituting (9.16) into (9.15c)
yields the input stability circle equation
(9.20)
where
T
:::
ISizSztl
I!Sul
and
-ldl 1
(9.21)
472
(9.22)
r's
(b) Input stability circle
(a) Output stability circle
Ir inI = 1 in the complex r L plane and stability circle

rout/ = t in the comptex rs plane.
Figure 93 Stability circle

1
When plotted in the r s -plane we obtain a response as schematically shown in Figure

9-3(b).
To interpret the meaning of Figure 9-3 correctly, a critical issue arises that is
investigated for the output circle [Figure 9-3(a)], although the same argument holds for
the input circle. If r L = 0, then 1rinl = 11 1and two cases have to be differentiated
depending on 11 1 < 1 or 11 1 > 1 . For 11 / < 1 , the origin (the point r L = 0) is part
of the stable region, see Figure 9-4(a). However, for jS 11 1 > 1 the matching condition
r L = 0 results in 1rinl
11 / > 1 , i.e. the origin is part of the unstable region. In this
case the only stable region is the shaded domain between the output stability circle
1rinl 1 and the
= 1 circle, see Figure 9-4(b).
For completeness, Figure 9-5 shows the two stability domains for the input stability circle. The rule-of-thumb is the inspection if jS22 < I, which leads to the conclusion that the center ( r s = 0 ) must be stable; otherwise the center becomes unstable for
IS
IS
IS
IS
= IS
1rLl
ISzzl >I.
Care has to be exercised in correctly interpreting the stability circles if the circle
radius is larger than C inI or CoutI Figure 9-6 depicts the input stability circles for
22 < 1 and the two possible stability domains depending on rin < ICinl or rin > ICin/
IS I
473
,.---
,,_- ....... ,
.......
''
'\
\
I
I
I
I
I
/
;'
;'
Unstable
(a) Shaded region is stable,

since ISul < 1
Figure 9-4
(a} Stable region excludes the origin,

rr 0, since IS1d> 1
Output stability circles denoting stable and unstable regions.
Unstable
(a) IS22I < 1
Figure 9-5
(b) IS22I > 1
Input stability circles denoting stable and unstable regions.
9.3.2 Unconditional Stability

As the name implies, unconditional stability refers to the situation where the
amplifier remains ~table throughout the entire domain of the Smith Chart at the selected
frequency and bias conditions. This applies to both the input and output ports. For
jS 11 1 < 1 and jS22 j < 1 it is stated as
474
!routl =
,. "
/
I
I
I
\
Figure g ..s
IS
Different input stability regions for 22 j < 1 depending on

rat10 between r 5 and [Cin .
II cinl - 'inI> 1
(9.23a)
IICout[- rout[ > 1
(9.23b)
sl
sl =
In other words, the stability circles have to reside completely outside the [r = 1 and
[rLl 1 circles. In the following discussion we concentrate on the jr
I circle
shown in Figure 9-7(a). It is shown in Example 9-2 that condition (9.23a) can be reexpressed in terms of the stability or Rollett factor k:
k = I
2
12
2
-ISn 1 -[Sz2 + 1~1 > 1
2IS 12IIS21i
(9.24)
Alternatively, unconditional stability can also be viewed in terms of the r s behavior in

the complex f out = r:ut + jf~ut plane. Here, the jfs[ :::;; 1 domain must reside completely within the Ifout! = I circle, as depicted in Figure 9-7(b). Plotting Irs[ = 1 in
the rout plane produces a circle whose center is located at
(9.25)
and which possesses a radius of
475
IS12S211
(9.26)
1 - 1Sul2
(a) 1
r outl
=1 circle must reside outside
(b) !rsl =1 circle must reside inside
Figure 9-7 Unconditional stability in the
r s and rout planes for ISnl < 1.
where the condi~ion ICsl + r 5 < 1 must hold. We note that (9.25) can be rewritten as
Cs = (S22 - LlS 11 ) / ( 1 -IS 11 j 2 ) . Employing ICsl + rs < 1 and (9.26) it is seen that
ls22 - LlS~ 1 1 +IS12S21

and since IS12S21 1 s
ls22 -LlS~ 1 / + fS 12S21
< 1 -IS 11 1
(9.27a)
we conclude
2
IS 12S21 I < 1 - IS11 1
(9.27b)
A similar analysis can be established for r L in the complex rin plane. From the
corresponding circle center C L and radius r L , we set C = 0 and rs < 1 . Thus,
Id
(9.28)
However, as long as llll < 1, (9.24) remains the sufficient requirement to ensure unconditional stability. This follows from the fact that when (9.27b) and (9.28) are added, it is
seen that
2IStzS21f < 2 -jS11I
Introducing the inequality !Lll =
IS11 S 22 -
2
-js22i
S21 IS IS 11 S22 + fS12S21
S 12
results in
Chapter 9 RF Tranel1tor Amplifier Dulgna
476
t
2
2
IAI < 1- 2<IS
11 J + JS 22
J
I l)
1
z
2JS 11 JIS22 j) = 1 - 2<IS 11 j- S 22
Since (1/2)(JS 11 J-JS22 j) < 1, it is seen that (9.27b) and (9.28) are equivalent to
IAI < 1
(9.29)
------------------------------~~~
Example 9-2: Stability factor derivation
Derive the stability factor k (Rollett factor) from (9.23a).
Solution:
Substituting (9.21) and (9.22) into (9.23a) gives
Is11- s;2AI-1s12s12! > 1

2
jS 11 - IL\1
(9.30a)
Squaring and rearranging (9.30a) results in
2!su - s;zA\ISizSzii < lsu -s;2AI
+ IS12Sz1l
22
-I!Sul -\Al l
(9.30b)
The term JS 11 - S22AJ in (9.30b) can be re-expressed as
ls11 -s;zAI
!St2Sztl 2 + (l - ISzzl 2)(jSul 2 -IAI 2 )
(9.30c)
Squaring (9.30b) again and rearranging tenns finally gives
The terms inside the curly brackets are recognized as the desired stability factor:
(9.30e)
A stability analysis starting from (9.23b) would have resulted

in exactly the same inequality. Thus, the stability factor k applies for
both input and output ports.
477
It is always prudent to determine that both the
1.11 <I and k > 1 conditions are
fulfilled to ensure an unconditionally stable design. The next example investigates a

transistor in common-emitter configuration in terms of its input and output stability
behavior.
----------------------------~~~
Example 9~3: Stability circles for a BJT at different operating

frequencies
Determine the stability regions of the bipolar junction transistor
BFG505W (Philips Semiconductors) biased at V CE = 6 V and
I c = 4 rnA . The corresponding S-parameters as a function of frequency are given in Table 9-1.
Table 9~1
BFG505W S-parameters as a function of frequency
Frequency
sll
sl2
s21
S22
500MHz
0.70L-57
0.04L47
10.5L136
0.79L-33
750MHz
0.56L-78
0.05L33
8.6L122
0.66L-42
lOOOMHz
0.46L-97
0.38L-115a
0.06L22
7.1Ll12
0.57 L-48
0.06L14
6.0Ll04
0.50L-52
1250 MHz
Based on the definitions fork, IL\1' cin ~ 'in' cout' and

rout, we compute the values via a MATLAB routine (see m-file
ex9_3.m). A summary of the results is given in Table 9-2 for the four
frequencies listed in Table 9-1.
Solution:
Table 9-2
Stability parameters for BFG505W for frequencies listed in

Table 9~1
1~1
cin
rin
cout
rout
0.41
0.69
39.04L108
38.62
3.56L70
3.03
0.60
0.56
62.21Lll9
61.60
4.12L70
3.44
0.81
0.45
206.23L131 o
205.42
4.39L69
3.54
1.02
0.37
42.42L143
41.40
4.24L68
3.22
Chapter 9 RF Translator Amplifier Dealgns
478
The example input and output stability circles for the frequencies off = 750 MHz and f = 1.25 GHz are shown in Figure 9-8.
We notice that 11 1 < 1 and 22 I < 1 in all cases. This implies that
the r L = 0 and r S = 0 points are stable, indicating that the interior domain of the Smith Chart up to the stability circles denotes the
stable region.
IS
IS
f= 1250 MHz
f= 1250 MHz
+1 .0
---...-----..'
Input
Output
/stability
2.o It(" / circles
stability~
circles
1.0
Figure 9-8
Input and output stability circles for BFG505W computed at f

750 MHz and f = 1.25 GHz.
Also, as can be seen from Figure 9-8 and Table 9-2, the transistor is unconditionally stable at f = 1.25 GHz and both input and
output stability circles are located completely outside of the lr! = 1
circle. At all other frequencies transistor is potentially unstable.
The stability circles are not only affected by frequency, but also
by the bias conditions. We recall that the S-parameters are given for
particular bias conditions. The entire stability analysis must be
repeated if biasing, or even temperature, changes.
479
Even though k can vary widely, most unstable practical designs fall into the range
0 :S k :S 1. Oscillators, discussed in Chapter 10, target the entire Smith Chart as the
unstable domain, resulting in negative values of k. It is also interesting to observe that
in the absence of any output to input feedback (S 12 = 0) the transistor is inherently
stable, since the stability factor yields k ~ oo In practice, one often examines k alone
without paying attention to the 1~1 < 1 condition. This can cause potential problems, as
the following example highlights.
----------------------------~~~
Example 9-4: Stable versus unstable region of a transistor
Investigate the stability regions of a transistor whose S-parameters
are recorded as follows: S 11 = 0.7L-70, S 12 = 0.2L-10,
Szl = 5.5L85 and s22 = 0.7 L-45
I
We again compute the values k, !ill, Cin, rin, Cout,

and rout The results are k = 1.15, 1~1 = 1.58, Cin = 0.21L52,
rin = 0.54' cout = 0.21L27 and rout = 0.54 (see Figure 9-9).
It is seen that even though k > 1 , the transistor is still potentially
unstable because 1~1 > 1 . This results in input and output stability
Solution:
Figure 9-9 Stability circles fork> 1 and 1~1 > 1.
480
IS
circles being located inside of the Smith Chart. Since both 11 1 and
jS22 are less than unity, the center of the Smith Chart is a stable
point. Therefore, since
< rin and
<rouP the area inside
of the stability circles represents the stable region, as shown in Figure 9-9.
ICinl
ICoutl
Usually manufacturers avoid producing transistors with both

k > 1 and 1.11> I by incorporating matching networks housed
inside the transistor casing.
9.3.3
Stabilization Methods
If the operation of a FET or BJT is found to be unstable in the desired frequency

range, an attempt can be made to stabilize the transistor. We recall that
> 1 and
jrout!> I can be written in terms of the input and output impedances:
1rinl
jrinl = z.m -Z0 > 1 and Ir out! = Z out -Z0 > 1

Zout + Zo
Zin + Zo
which imply Re{Zin} < 0 and Re {Zouc} < 0. One way to stabilize the active device is
to add a series resistance or a shunt conductance to the port. Figure 9-1 0 shows the configuration for the input port. This loading in conjunction with Re { Z s} must compensate the negative contribution of Re{Zin}. Thus, we require
Re{Zin + Rin' + Zs} > 0 or Re{ Yin+ Gin'+ Ys} > 0
Active device
Source
(BIT orFET)
zin
(a) Series resistance

Figure 9-10
(9.3la)
Active device
Source
(BIT orFET)
r;n
(b) Shunt conductance
Stabilization of input port through series resistance or shunt

conductance.
481
Following an identical argument, Figure 9-11 shows the stabilization of the output port.
The corresponding condition is
Re{Zout +Rout'+ ZL} > 0 or Re{ Yout +Gout'+ Y L} > 0
Active device
(BJ'T or FE'T)
Load
Active device G '

(BJ'T or FE'T) out
(a) Series resistance
Figure 9-11
(9.3lb)
Load
(b) Shunt conductance
Stabilization of output port through series resistance or shunt

conductance.
The next example explains the stabilization procedure for transistor.

----------------------------F~~
Example 9-5: Stabilization of a BJT

Using the transistor BFG505W from Example 9-3 operated at
f = 750 MHz (and with the S-pararneters given as follows:
S 11 = 0.56L-78, S21 = 0.05L33, S 12 = 8.64L122, and
S22 = 0.66L-42 ), attempt to stabilize the transistor by finding a
series resistor or shunt conductance for the input and output ports.
Solution:
With given S-parameters we can identify the input and

output stability circles by computing their radii and center positions:
Cin = 62.21Ll19,
rin = 61.60,
and
Cout = 4.12L70,
rout = 3.44 . The corresponding stability circles are shown in Figure
9-12. A constant resistance circle r' = 0.33 in the Z-chart indicates
the minimal series resistance that has to be connected to the input of
the transistor to make this port stable. If a passive network is connected in series to the resistor with the value of
Rin' = r'Z0 = 16.5 Q, then the combined impedance will be
located inside of the r' = 0.33 circle and therefore in the stable
Chapter 9 RF Transistor Ampltfler Deelgna
482
region. Similarly, by tracing a constant conductance circle g' = 2.8

we find the shunt admittance Gin' = g' /Z0 = 56 mS that stabilizes
the input of the transistor. This time any passive network connected to
Gin' will have the combined admittance residing inside of the
g' = 2.8 circle in the Y-chart, which is inside the stable region for the
input port of the transistor.
Figure 9-12
Input and output stability circles and circles for finding stabilizing
series resistance and shunt conductances.
Following an identical procedure we can find a series resistance of

Rout' = 40 Q and a shunt conductance Gout' = 6.2 mS, which
stabilize the output port of the transistor.
Due to the coupling between input and output ports of the

transistor it is usually sufficient to stabilize one port. The choice of
which port is generally up to the circuit designer. However, one
483
Constant Gain
attempts to avoid resistive elements at the input port since they

cause additional noise to be amplified.
Stabilization through the addition of resistors comes at a prize: the impedance

matching can suffer, there may be a loss in power flow, and the noise figure typically
worsens due to the additional thermal noise sources that the resistors present.
9.4 Constant Gain

9.4.1
Unilateral Design
Besides ensuring stability, the need to obtain a desired gain performance is

another important consideration in the amplifier design task. If, as sometimes done in
practice, the influence of the transistor's feedback is neglected (S 12 ~ 0 ), we can
employ the unilateral power gain GTu described by (9.12). This equation is rewritten
such that the individual contributions of the matching networks become identifiable.
With reference to Figure 9-13, we write
GTu =
-Ir sl2
II-Snrsl
2x
jS2 d x
- Ir dz
II - rLSzzl
= Gs x G 0 x GL
(9.32)
where the individual blocks are

(9.33)
r ...,' =o
Figure 9-13
r'=o
L
Unilateral power gain system arrangement.
Because most gain calculations are done in dB, (9.32) is frequently expressed as
Chapter 9 RF Transistor Amplifier Deelgns
484
(9.34)
where Gs and GLare gains associated with input and output matching networks and G0
is the insertion gain of the transistor. As seen from (9.33), the network gains can be
greater than unity which at first glance might appear stronger since they do not contain
any active devices. The reason for this seemingly contradictory behavior is that without
any matching a significant power loss can occur at the input and output sides of the
amplifier. The use of G s and GL attempts to reduce these inherent losses, which is considered a gain.
If jS11 1 and jS22 j are less than unity, the maximal unilateral power gain 9rumax
results when both input and output are matched (i.e., r s = S11 and r L = S22 ). For
this case it is seen that
Gsmax
GLmax
1
1 -ISnl
(9.35)
(9.36)
1-1s22l
The contributions from G 5 and GL can be normalized with respect to their maximum
values such that
gs =
gL
Gs
Gsmax
GL
GLmax
jr lz
s z< 1 -1Sllj2)
II- Sursl
1
lr 12
ll-Sz2rLl
2(
- IS22! 2 )
(9.37a)
(9.37b)
where the normalized gain is given in both cases as 0 ~ gi ~ 1 , with i = S, L .

Even though we have explicit gain equations for the input and output matching networks, they are not directly usable in tenns of providing parametric curves of constant
gain. The key question that must be answered is formulated as follows: For a given S 11
(or S22 ) and a desired normalized gain 8s (or gL), what is the range of values for fs
(orrL) that achieves a particulargain? The solution requires the inversion of (9.37)
gi
-lril
1- sll..r.jl
2<1 -jsiij2)
(9.38)
for the reflection coefficient ri. Here ii = 11, 22 depending on i

a set of circles with center locations at
= S, L. The result is
485
Constant Gain
(9.39)
and radii of size
(9.40)
Example 9-6 details the necessary steps to derive the unilateral constant gain circle
equations (9.39) and (9.40).
----------------------------~~~
Example 9-6: Derivation of the constant gain circles
Find the expressions for d 8 ; and r 8 ,. as given in (9.39) and (9.40).
Solution:
The derivation begins with (9.38), which is rewritten
as
g l< 1+
**
[s ..r.l 2 - s.r.s ..r.)
II
ll
The reflection coefficient
rlz-
II
r;
1-ISiil
2
gi JSul
0-ISiil
II
(9 41a)
*
2
r* +
(1- gJ
(9.4lb)
2 2
(1-gi)(1-ISiil )
2
2
0 -ISiii 0 - g;))
gj))
-lrl 2 + ISI 2 1rl 2
gisii
1-ISiil
(1- 8;)
2
can be factored out such that
(1 -
= 1 -ISI
ll
r.-
gisii
2
This equation is the complex form of a circle expression
(ri-dg,.)(ri -dg,.) =
rgi
(9.41c)
with
Multiplying out (9.41c) results in the more familiar from

R2
I
12
2
criR - dg)
+ crj- dg) = rgi
(9.41d)
486
Chapter 9 RF Transistor Amplifier Deslgna
where superscripts R and I denote real and imaginary parts of f;

and d 8 i .
Because of the unilateral assumption we are able to derive
separate gain circle equations for input and output ports.
The following observations can be made from the constant gain circle equations
(9.39) and (9.40):
Isiil 2
) is obtail!,ed for i = s;i ' which coin The maximum gain Gimax = 1I ( 1 cides with the gain circle whose center is at d 8 ; = Sii and of radius r E; = 0 .
The constant gain circles all have their centers on a line connecting the origin to
* The smaller the gain values, the closer the center d i moves to the origin and
Sii.
8
the larger the radius r8 i.
2
For the special case ri == 0, the normalized gain becomes gi = 1 -1Siil and
2
both d8 ; and r8 ; have the same value d 8 i = r8 ; = jSiij/(1 + jSiij ). This implies
that the G; = 1 (or 0 dB) circle always passes through the origin of the f;-plane.
Example 9-7 demonstrates the source gain circles for an amplifier design under
unilateral approximation.
----------------------------~~~
Example 9-7: Computation of the source gain circles for a unilateral design
A PET is operated at f = 4 GHz and is biased such that
S 11 = 0.7 L125 . It is assumed that the transistor is unconditionally stable so that the unilateral approximation can be applied. Find
the maximum source gain Gsmax and plot the constant source gain
circles for several values of G s .
Solution:
First we find the maximum source gain GSm ax usmg

(9.35). The result is
Constant Gain
487
Gsmax =
1
-Is 111
1
2
l - 0.7
- --
= 1.96 or Gsmax
= 2.92 dB
We can now plot the constant gain circles by using (9.39) and (9.40)
for the computation of circle centers d 85 and radii r 85 A summary
of several arbitrary source gains G5 is presented in Table 9-3.
Table 9-3
Parameters for constant source gain circles

in Example 9-7.
Gs
gs
dgs
'gs
2.6dB
0.93
0.67L-125
0.14
2dB
0.81
0.62L-125
0.25
1 dB
0.64
0.54L-125
0.37
OdB
0.51
0.47 L-125
0.47
-1 dB
0.41
OAOL-125
0.56
As seen from Table 9-3, the radius r85 of the Gs 0 dB circle

is equal to the magnitude of its center position d and the circle
85
indeed passes through the center of the Smith Chart. We also
observe that the centers for all Gs circles are located on the
e = LS II* = -125 line, and as Gs approaches GSmax ' the
radius of the corresponding circle reduces to zero and its center
position becomes S 11 * = 0.7 L-125.
Figure 9-14 illustrates the source gain circles based on the
computed numerical values given in Table 9-3. The figure points out
clearly that, despite the input matching network being passive, the
gain can be greater than 0 dB, indicating amplification. The physical
meaning for such a behavior lies in the fact that the matching network reduces the input reflection coefficient of the overall system,
thus effectively creating an "additional" gain.
488
Chapter 9 RF Transistor Amplifier Deslgnl
+1 .0
-1 .0
Figure 9..14 Constant source gain circles in the Smith Chart.
The underlying assumption of this example is that the gain

associated with the matched input port is not affected by the output
since the unilateral approximation neglects the reverse gain.
We next discuss a typical application that requires the use of the constant gain circle
approach. Specifically, let us develop a unilateral amplifier for a predetermined fixed
gain value.
-------------------------~&MW4
Example 9..8: Design of a 18 dB singJe..stage MESFET ampJi..

tier operated at 5.7 GHz
A MESFET operated at 5.7 GHz has the following S-parameters:
S 11 = 0.5L-60 , S 12 = 0.02L0, S 21 = 6.5L115, and
S 22 = 0.6L-35.
(a) Determine if the circuit is unconditionally stable.
Constant Gain
489
(b) Find the maximum power gain under optimal choice of the
reflection coefficients, assuming the unilateral design (S 12 =0).
(c) Adjust the load reflection coefficient such that the desired gain is
realized using the concept of constant gain circles.
Solution:
(a) The stability of the device is tested via (9.24) and
(9.29)~ with the results
2
1 -IS 11l -1s22l + 1~1
2IS12JIS211
= 2.17
and
1~1 = IS11S22-St2S2d
Because k > 1 and
= 0.42
ldl < 1 , the transistor is unconditionally stable.
(b) We next compute the maximum gain for the optimal choice of
the reflection coefficients (i.e., r L
S22 * and r S
S 11 *)
Gsmax
GLmax =
G0
1-JSul
1
1 -1s22l
= 1.33 or 1.25 dB
= 1.56 or 1.94 dB
= jS2 d2 = 42.25
or 16.26 dB
Therefore, the maximum unilateral transducer gain is given by
Grumax
= GsmaxGoGLmax
= 88.02 or 19.45 dB
(c) Since the source matching network ( r s = S 11 * ) and the transistor combined already provide a gain of 17.51 dB, we have to chose
r L in such a way that GL = 0.49 dB. This means that r L has to
reside on the rg ,_ = 0.38, dg L = 0.48L35 circle, as shown in Figure 9-15. If we chooserL = 0.03 + j0.17, the output matching network reduces to a single element (i.e., a series inductor with a value
of L = 0.49 nH) provided the load is equal to the characteristic
impedance (ZL = Z0).
Chapter 9 RF Transistor Amplifier Dealgns
490
+1 .0
1.0
Figure 9-15
Constant load gain circle in the Smith Chart.
If the amplifier is operated over a range of frequencies, the

gain has to be determined for a corresponding number of discrete
frequency points due to the changing S-parameters.
For the case where jS;il > 1 (ii = 11 for the input port and ii = 22 for the output port) it is possible for a passive network to produce an infinite value of Gi ( i = S
or L, respectively). This situation occurs when r; = Sfi1 , meaning that the real component of the impedance associated with r; is equal in magnitude to the negative resistance related to Sii. Thus, the two resistances cancel each other and oscillations will
result: the amplifier is unstable. To avoid this problem, we plot the constant gain circles
for jS;;j > 1 and the corresponding stability circle and choose r; in such a way that it is
located on the desired gain circle but also resides inside the stable region.
,/
9.4.2
Unilateral Figure of Merit
The unilateral design approach discussed in Example 9-8 involves the approximation that the feedback effect, or the reverse gain, of the amplifier is negligible
( S12 = 0 ). To estimate the error due to this assumption, the ratio between the transducer gain Gr, which takes into account S12 , and the unilateral transducer gain Gru
can be formed. Using definitions (9.8) and (9.12), we find
491
Constant Gain
(9.42)
where Gr ~ Gru .
The maximum value of Gru, and therefore the maximum error, is obtained for
* Therefore, (9.42)
the input and output matching conditions
= Su* and L = s22).
becomes
crs
~--------------------~
TUmax
1-
* *
Sl2S2l S22S 11
2
(9.43)
(1-ISul H 1- 1s22l )
This can be used to set bounds on the error fluctuation

G
( 1 + Uf < G r ~ ( 1 - Uf
2
(9.44)
TU
where U is known as the frequency-dependent unilateral figure of merit:
u=
jS12IIS2ti iS22j jSu l
(1-ISuj2)(1-js2212)
(9.45)
To justify a unilateral amplifier design approach, this figure of merit should be as small
as possible. In the limit, as Gr approaches Gru for the ideal case of S 12 = 0, we see
that the error does indeed vanish (i.e., U = 0 ).
--------------------------~RF~~
Example 9-9: Unilateral design applicability test
For the amplifier discussed in Example 9-8 estimate the error that is
introduced by making the unilateral design approximation.
Solution:
Substituting the S-parameter values into (9.45), the

unilateral figure of merit is found to be
U
= -~1s_t2.:..:...lls--:-21..:. :.11s_2.. .:. .21 s..........:ul_ = 0.0812

2
2
:..,_I
(I-1Sui H1-IS221 )
492
The maximum error can then be estimated from (9.44):

G
0. 86 < G
~ 1.18
TU
This implies that the theoretical value for the transducer gain can
deviate from its unilateral approximation by as much as 18%. Practically, however, the actual difference often is much smaller. This
becomes apparent if we substitute the values obtained in Example
9-8 into the transducer power gain definition (9 .8). It is found that
Gr = 62.86 or 17.98 dB, which compares favorably with
Gru = 63.10 or 18 dB. In other words, we introduced an error of
less than 1%.
The unilateral figure of merit computation constitutes a very

conservative, worst case error estimation.
9.4.3
Bilateral Design
For many practical situations the unilateral approach may not be appropriate
because the error committed by setting S 12 = 0 could result in an intolerably imprecise design. The bilateral design takes into account this feedback. Instead of the unilat
eral matching r s* = S 11 and r L* = S 22 , it deals with the complete equations [see
(9 .15b) and (9 .15c)] for the input and output reflection coefficients
sJI-rrL1
1- s22rL
*
s 12 s21 r s
r L = Sz2 + 1 - s11 r S -
s22 -
r s..1
I -Su r L
(9.46a)
(9.46b)
which require a simultaneous conjugate match. The meaning of simultaneous implies

that matched source and load reflection coefficients r MS and r ML have to be found
that satisfy both coupled equations. If the device is potentially unstable, then a simultaneous complex conjugate does not exist. The solution approach to obtain these optimal
coefficients is outlined in Example 9-10. The final results, for the matched source
reflection coefficient r MS is
493
Constant Gain
(9.47)
where
*
= Su-S22~
and B1
l-IS221 2 -1~1 2 + ISnl 2

Similarly, the matched load reflection coefficient r ML is
cl
(9.48)
(9.49)
where
(9.50)
The solutions (9.47) and (9.49) are derived under the assumption of unconditional
stability.
With ['ML and ['MS given by (9.47) and (9.49), the optimal matching can be
rewritten as
*
S12S21 r ML
r Ms = s11 + 1 - s22 r ML
(9.51a)
and
*
S12S21
r Ms
r M L = s22 + 1 - s11 r MS
(9.51b)
It is noted that the unilateral approach, which decouples input and output ports, is a
subset of the bilateral design approach.
---------------------------~~~
Example 9-10: Derivation of simultaneous conjugate matched
reflection coefficients
Derive the reflection coefficient expression (9.47).
Solution:
Starting from (9.46a) and (9.46b), we see that
* s11)
(1 - s22rL)(rs-
= r Ls 12s21
(9.52a)
494
Chapter 9 RF Tranatstor Amplifier Design
(1-Surs)(rL* - s22) = rsst2s21

Solving (9.52a) for
rL
(9.52b)
yields
rL=
sll - rs*
(9.52c)
*
.1-S22rs
Substituting (9.52c) into (9.52b) results, after some algebra, in

2
*
2
2
2
*
*
r sCS
u - s22.1) - r sC 1 + ISul - IS221 - ldl ) = - su + s22~
(9.52d)
2
2
2
Identifying c l = (Su- s;2~)' B l = (1 + IS111 -IS221 - 1~1 )
leads to the standard quadratic equation
2
rs -
B1
-rs
=
cl
-C1
(9.52e)
whose solution is
(9.520
The negative sign in front of the square root is picked to ensure stability (k > 1 ).
An identical analysis approach for the load side leads to a

quadratic equation for
rL
whose solution yields
r ML.
Example 9-11 demonstrates the use of simultaneously complex conjugate reflection coefficients for the design of an amplifier with maximum gain.
----------------------------F&JA~
Example 9-11: Amplifier design for maximum gain
A BJT with I c = 10 rnA and V CE = 6 V is operated at a frequency of f = 2.4 GHz. The corresponding S-parameters are:
S 11 = 0.3L30, S 12 = 0.2L- 60 , S21 = 2.5L-80, and
S 22 = 0.2L-15 o Determine whether the transistor is uncondition-
495
Constant Gain
ally stable and find the values for source and load reflection coefficients that provide maximum gain.
Solution:
The stability of the transistor is determined by computing k and IL\1 based on (9.24) and (9.29) with the explicit result
of k = 1.18, IL\1 = 0.56. Since k > 1 and IL\I < 1 , the transistor is
unconditionally stable.
As we see from the S-parameters of the transistor, S12 has a
relatively large magnitude and the use of the unilateral design
method for the amplifier does not appear appropriate, suggesting the
bilateral approach instead.
Using (9.48) and (9.50), we find the coefficients
B 1 = 0.74,
and
C 2 = 0.03 + j0.07,
C 1 = 0.19 + j0.06,
B2 = 0.64, which allow us to compute the simultaneously complex
conjugate
source
and
load
reflection
coefficients
r MS = 0.30L-18 and r ML = 0.12L69, respectively. It should
* ,
be noted that these values differ significantly from S *11 and S 22
which are the basis for the unilateral design.
Applying (9.8), with rL and rs replaced by r ML and r MS'
we find the transducer gain to be Gr = 8.42 dB. This also happens
to be the maximum transducer gain G rmax
The discrepancy between unilateral and bilateral gain is best
seen in the large differences in phase between S *11 and r MS as well
* and r ML
as S22
9.4.4
Operating and Available Power Gain Circles
For the situation where the reverse gain of S 12 cannot be neglected, the input
impedance is dependent on the load reflection coefficient. Conversely, the output
impedance becomes a function of the source reflection coefficient. Because of this
mutual coupling, the unilateral approach described in Section 9.4.1 is not appropriate to
design an amplifier for a predetennined gain.
In the bilateral case, which takes into account the mutual coupling between input
and output ports, there are two alternative design methods to develop amplifiers with a
specified gain.
Chapter 9 RF Translator Ampllfter Detlgn
496
The first method is based on the use of the operating power gain G given by
(9 .14). Here we attempt to find the load reflection coefficient r L, assuming that the
source is complex conjugate matched to the input reflection coefficient [i.e., r s = r~n,
where rin is computed based on (9.9a)]. This method yields an input voltage standing
wave ratio of VSWRin = 1.
The second method uses the available power gain G A definition of (9 .13 ). In this
*
case we assume perfect match on the output side of the amplifier ( r L = rout),
and the
load is chosen in such a way as to satisfy the gain requirement. This method is preferable if the output standing wave ratio should be unity (i.e., VSW Rout = 1 ).
Operating Power Gain

To develop the design procedure based on using the operating power gain (and
thus ensuring VSWRin = 1), we rewrite (9.14) in the form
2
( 1 -lrd )1Sztl
_
G -
(1
2 -
(1-
(I-Irinl )ll-S2zrLI
-lrLI )1Szii
S S r 2
s u + 121 s12 r L )11-S22r L 12
- 22 L
go S21 1
(9.53)
where we use (9.9a) for rin. The factor g 0 defines a proportionality factor given by
1 -l rLj
g =
o
s2ts12rL
I-
=
2
)
s u + 1 - s22 r L II - s22 r Ll
(9.54)
As shown in Example 9-12, (9.54) can be rewritten in terms of a circle equation

for the load reflection coefficient r L ; that is,
= r 8o
(9.55)
where the center position d 8 is

Q
d
8o
and the radius r
80
"' *
go(S22- ASll)
1 + go(ISzzl2 -IAI2)
(9.56)
is defined as
_Jt- 2kgojS12S2tl + g~IS12S21I

r so -
jt + goCIS2212 -IAI2)j
with k denoting the Roulette stability factor as defined in (9 .24).
(9.57)
497
Constant Gain
~-+4~
- - - - - - - - - - - - - - - R F & M 'VV"-+
Example 9-12: Operating power gain circle derivation
Starting from (9.54), derive the circle equation (9.55) in the complex
r L -plane.
Solution:
go
First we rewrite (9 .54) in the form

1-JrLJ2
2
(9.58)
1-JS11 j +jrLj (jS22 j -1.11 )-2Re[rL(S22 -.1S 11 )]
After multiplying both sides of (9 .58) by the denominator and rearranging terms, we see that
2
jrLj [1 + g 0 (jS22 J2 -IAI 2 )]- 2g0 Re{rL(S22 -
AS~ 1 )}
(9.59)
=1- go(l-jSllj2)
Dividing (9.59) by [1 + g 0 (IS22 j2 -IAI 2 )], we find

2
Ll -
2g 0 Re{rL(S22 -.1S~ 1 )}
Jr
+ g 0 (jS22 j 2 -IAI 2 )
1-g0 (1-jS 11 ()
1 + 8o(JSz21 2 -1.11 2)
This equation can already be recognized as a circle equation of the

2
L- dgol
= r;o, where the circle center dgo is given by
form
(9.56) and the radius rga is computed from
1r
r2
go
2
* *
1- go(l- ISll 1 ) + 8o(S22- ASn)
1 + go(js221z -IAI2)
1 + 8o<ISzzl2 -IAI2)
2
[ 1- go( 1-jSllj )] [1 + 8o(JSz21 2 -1.11 2 )]
+ fgols22- AS~~~ 1
--------~---~------~---~-----------~------~
-----------~~----~~-----------
[ 1 + go(JSzzl2 -IAI2)]2
1- go(l -JS11j2 -JSzzl2 + jA)2)- g~M
[1 + go(JSzzl 2 -1.11 2 )] 2
= l-2g
JS12S2dk-g;M
-------~--~-----
[1 + 8oCIS22I 2 -1.11 2 )1 2
where k is the stability factor defined in (9.24) and M is a constant

given by
498
= (l- ISui2H IS2zl2 -
ILllz) -ISzz- LlS~l l 2
= -IS12s2112
Thus, for the square of the circle radius we obtain

r
z
go
1-
2goiS12s21 lk + g;1s12s2tl 2
[
1 + go(ISzzl 2 -1~1 2 ) J2
which agrees with (9.57).
The following example demonstrates the design of an amplifier based on the bilat~
eral method. It targets a specified gain using the constant operating gain circle
approach.
-------------------------~&MW4
Example 9-13: Amplifier design using the constant operating
gain circles
Use the same BJT as described in Example 9-11, but instead of
Grmax = 8.42 dB, design an amplifier with 8 dB power gain. In
addition, ensure a perfect match on the input port of the amplifier.
Solution:
As shown in Example 9~11, the transistor is unconditionally stable. Because a perfect match on the input port must be
maintained, we employ the operating power gain circles In our
design.
First we compute the value of factor g 0 ; that is,
G
g0 = -,-,2 = 1.0095
Szl
where G = 6.31 is the required 8 dB operating gain. Substituting

g 0 into (9.56) and (9.57), we find center and radius of the constant
operating gain circle in the r L -plane. The corresponding values are
dgo = 0.11 L69 and r8 o = 0.35. The constant gain circle is
Constant Gain
499
Figure 9-16
Constant operating power circle in the
r L -plane.
There is a great variety of possible choices for the load reflection coefficient that ensures a G = 8 dB operating gain. To simplify
the output matching network, we pick L at the intersection of the
constant gain circle with the constant resistance circle r = I (see
Figure 9-16). The value obtained at that point is rL = 0.26L-75.
With r L known, we can next find the source reflection coefficient
that must be the conjugate to the input refection coefficient as given
in (9.9a):
su- ~r
L)* = 0.28L- 55o
rs = ( I- s22rL
Based on the previously computed values, we check the correctness

of our approach. Substituting rin and r L into (9.10), we find that
the transducer power gain is indeed 8 dB.
500
The complexity of the input matching network is directly

affected by the appropriate choice of r L because of the requirement
f s = fi~ where rin is a function offL.
In Example 9-13 we pick the value of r L arbitrarily (residing on the desired gain
circle) and compute a corresponding input impedance such that r s = ri:, assuming
that there are no restrictions imposed on the value of r s. Unfortunately, in many practical applications, r s has to satisfy certain constraints (for example, to stay within a
desired noise performance). Such additional conditions may therefore restrict our freedom in using r s and, as a consequence, limit the possible choices for r L. One way to
satisfy both requirements (rL residing within an appropriate gain circle, and r s sat-
isfy a particular noise requirement) is via trial-and-error, whereby we arbitrarily pick

r L and see whether the corresponding r s meets design specifications. This method is
simple but very tedious and time-consuming.
A more scientific approach relies upon mapping the constant gain circle (9.55) in
the L -plane into a circle in the s -plane, i.e.,
r 5 -dgs1 =
rgs
(9.60)
where the equations for the circle radius rgs and its center d gs are obtained from the
requirement that r s = ri~ . This can be written as
S 11 -arL
~~~~
1- s22rL
(9.61)
s 11 - r;
a- s22rs
(9.62)
rs*
rL
= ---
or
Substituting (9.62) into (9.55) gives us
sll- r;
a-s22rs
----* dg
2
Q
- rg2
(9.63)
which can be rewritten in the form of (9.60), where the circle radius is
gs
IS12S21I
Szzdgol 2 - riJS221 2I
rg
11 1 -
(9.64)
501
Constant Gain
and the center is given by
(I- S22 dg)(S 11

dgs
Mg 0 )*-
r;o Ll*S22
jl- S22dg,J2- rjJSzzJ2
(9.65)
The derivation of (9.64) and (9.65) is left as a problem at the end of this chapter. The
example of constant gain circle mapping is discussed further in Section 9.5 t Example
9-14.
AvaUable Power Gain
In those cases where perfect matching on the output side of the amplifier is
required ( VSW Rout = 1 ), the available power gain approach should be used instead of
the previously presented operating gain method. For this situationt a constant available
gain circle equation can be derived in the same fashion as (9.55) is obtained. The result
of such a derivation is a circle equation which relates the source reflection coefficient to
the desired gain:
(9.66)
where the center position d g(J is
ga
ga(Su- Mzz*)*
1 + ga(!Slllz -ILl\2)
(9.67)
and the radius r ga is defined as
Jl- 2kgaJS12Sztl + g:IS12Szd2

=
~-------------------
--~~~----~--~~~
ga
II+ ga(ISllj2-IA!2)1
(9.68)
The proportionality factor ga is computed as

GA
ga =
!Szd2
(9.69)
where G A is the desired power leveL

Similar to the constant operating power circlest a constant available power circle
can be mapped into the r L -plane using
1r
L- dg,l
with the circle radius given by
= rg 1
(9.70)
502
r
8t
rgalS 12S21I
(9.71)
~---------------.
Ill- Slldgal2- riJSul21
and the center location defined by
( 1- S 11 d 8)(S22 - ild8 )*dgt
ria il*S
11
(9.72)
II-Sudgal2-riJSul2
We see that r81 and d 81 for VSW Rout = 1 have their correspondence to r8 s and
d 8 s for VSWRin = 1 with S 11 in (9.71) and (9.72) replaced by S
22.
9.5 Noise Figure Circles

In many RF amplifiers, the need for signal amplification at low noise level
becomes an essential system requirement. Unfortunately, designing a low-noise amplifier competes with such factors as stability and gain. For instance, a minimum noise
performance at maximum gain cannot be obtained. It is therefore important to develop
a method that allows us to display the influence of noise as part of the Smith Chart to
conduct comparisons and observe trade-offs between gain and stability.
From a practical perspective, the key ingredient of a noise analysis is the noise
figure of a two-port amplifier in the admittance form
Rn
= F min + G IYS s
Y optl
(9.73)
or in the equivalent impedance representation

F
Gn
= F min+ ]fiZs- zoptl
(9.74)
where Zs = l!Y5 is the source impedance.

Both expressions are derived in Appendix H. When using transistors, typically
four noise parameters are known either through datasheets from the PET or BJT manufacturers or through direct measurements. They are:
The minimum (also called optimum) noise figure F min whose behavior depends
on biasing condition and operating frequency. If the device were noise free, we
would obtain F min = l .
The equivalent noise resistance Rn
= 1I G n of the device.
Noise FJgure Circles
503
The optimum source admittance Y opt = G opt + j B opt = 1I Z opt . Instead of the
impedance or admittance, the optimum reflection coefficient
The relationship between Yopt and
r opt
r opt is often listed.
is given by
1- ropt
y opt
= f ol + r opt
(9.75)
Since the S-parameter representation is a more suitable choice for high-frequency

designs~ we convert (9.73) into a form that replaces the admittances by reflection coefficients. Besides (9.75) we use
(9.76)
= Y 0 ( 1 - 1r sl 2) / jl + r sl 2 , the
in (9.73). Recognizing that Gs can be written as Gs

final result becomes
2
4 Rn
Irs- r opd
F = F min+ Zo (1 -/ri )/1+ ropt/2
(9.77)
In (9.77) the quantities F m in, Rn, and r opt are known. In general, the design engineer
has the freedom to adjust r s to affect the noise figure. For r s = f opt we see that the
lowest possible noise figure is achieved, F = F min . To answer the question of how a
particular noise figure, let us say F k , relates to r s , (9.77) is put into the form
k - F min)
lrs-ropd 2 = 0-lrsl2 ) il +foptl 2(F4Rn!
Z
0
(9.78)
which on the right-hand side already suggests the form of a circle equation. Introducing
a constant Qk such that
Qk =
jl + r opt(
2(FkF min)
4Rn1 Zo
(9.79)
and rearranging terms gives

(9.80)
Division by ( 1 + Qk) and forming a complete square yields, after some algebra,
2
Qk + Qk(l
(1
- Ifopt! 2)
+ Qk)2
(9.81)
504
This is the required circle equation in standard form that can be displayed as part of the
Smith Chart:
lrs-dFkl
R2
= (rs -dpk)
12
+ (rs-dF)
=rFk
(9.82)
with the circle center location d F denoted by the complex number

k
(9.83)
and the associated radius
2
= J< l -lr
opti
rp
"
)Qk +
Q;
1 + Qk
(9.84)
There are two noteworthy conclusions that can be drawn from (9.83) and (9.84):
The minimum noise figure is obtained for Fk = F min , which coincides with the
location d Fk = r opt and radius r Fk = 0 .
All constant noise circles have their centers located along a line drawn from the
origin to point r opt . The larger the noise figure, the closer the center d F" moves to
the origin and the larger the radius r F .
k
The following example points out the trade-offs between gain and noise figure for
a small-signal amplifier.
---------------------------~&)A~
Example 9-14: Design of a small-signal amplifier for minimum
noise figure and specified gain
Using the same transistor as in Example 9-13, design a low-noise
power amplifier with 8 dB gain and a noise figure that is less than
1.6 dB. Assume that the transistor has the following noise parameters: F min = 1.5 dB, Rn = 4 fl, and r opt = 0.5L45.
Solution:
The noise figure is independent of the load reflection
coefficient. However, it is a function of the source impedance. It is
therefore convenient to map the constant gain circle obtained in
Example 9-13 into the r s -plane. Applying equations (9 .64) and
505
Noise Figure Circles
(9.65) and values from Example 9-13, we find the center and radius
of the mapped constant gain circle: dg s
0.29L-l8 and
r gs = 0.18 . A r s residing anywhere on this circle will satisfy our
gain requirement. However. for the noise figure specifications to be
met we have to ensure that s resides inside the Fk = 2 dB constant noise circle.
The noise circle center and its radius are computed using
(9.83) and (9.84), respectively. They are listed below together with
the coefficient Qk, see (9.79):
Qk == 0.2, dp k = 0.42L45, rF k = 0.36

The obtained G == 8 dB and Fk = 1.6 dB circles are shown in Figure 9-17.
Figure 9-17
Constant noise figure circle and constant operating gain circle

mapped into the r s -plane.
Notice that the maximum power gain is obtained at the point

where MS = 0.30L-18 (see Example 9-11 for the detailed computations). However, the minimum noise figure is obtained at
5De
Chapter 9 RF Translator Amplifier Design
r s = r opt = 0.5 L45 o , which shows for this example that it is

impossible to achieve maximum gain and minimum noise figure
simultaneously. Clearly, some compromises have to be made.
To minimize the noise figure for a given gain, we should pick
the source reflection coefficient as close as possible to the location
of r opt while still residing on the constant gain circle. Arbitrarily
choosing r 5
0.29Ll9, the corresponding load reflection coefficient is found to be r L = 0.45L50 by applying (9.62). The
obtained amplifier noise figure is then computed using (9. 77):
4Rn
= F min + Z
lrs- roptl 2
o - Ir sl
) 11 + r opt!
2 -
1.54 dB
The requirements of maximum gain and minimum noise figure

will always be design trade-offs and cannot be met simultaneously.
9.6 Constant VSWR Circles

In many cases the amplifier has to stay below a specified VSWR as measured at
the input or output port of the amplifier. Typical values range between
1.5 :s; VSWR < 2.5. As we know from our discussion in Chapter 8, the purpose of
matching networks is primarily motivated by the desire to reduce the VSWR at the transistor. The complication arises from the fact that the input VSWR (or VSWRIMN) is
determined at the input matching network (IMN), which in turn is affected by the active
device, and, through feedback, by the output matching network (OMN). Conversely, the
output VSWR (or VSWRoMN) is determined by the OMN and, again through feedback,
by the IMN. This calls for a bilateral design approach, as discussed in Section 9.4.3.
To set the stage, let us consider the arrangement depicted in Figure 9-18. The two
VSWRs that are part of an RF amplifier specification are
1 + 1riMNI
VSWRIMN = 1 -lriMNI and VSWRoMN
1 + 1roMNI
1 -lrOMNI
(9.85)
The reflection coefficients r IMN , r OMN require further clarification. If we concentrate

on riMN, it is apparent from Section 9.2.1 that the input power Pin (under the assumption r~ = 0) can be expressed as a function of the available power P A :
(9.86)
507
Constant VSWR Circles
r's =o
r'=o
L
.,:...
~
RF
source
~ [~
A.
":'
Input
Matching
Network
(IMN)
)..
)..
T
~l ~
[>
~~~
Output
Matching
Network
(OMN)
;.
":'
)..
":'
~~~
Load
"'r
~n
System configuration for input and output VSWR.
Figure 9-18
Postulating that the matching network is lossless, the same power is also present at the
input terminal of the active device
pin
= pA
<1 -
Ir sl ><I - Ir in1)
\l- r Srinl
(9.87)
in the absence of any matching. Setting both equations equal and solving for
yields
\rrMNI
...
...
r . -rs = r.m -rs

(9.88)
1- rsrm
1 - rsrm
Equation (9 .88) can be converted into a circle equation for r s that is centered at locatn
tion dv IM N with radius rv IMN such that

R
(rs - dv,MN ) + (rs -dv,MN) =
'v,MN
(9.89)
where
(9.90)
and
2
VJMN -
(I
- lr inl
I -
>lriMNI
IriMNrs12
(9.91)
Here the subscript VlMN in d vIMN and r v JMN is used to denote the VSWR at the
IMN location.
508
Chapter 9 RF Tranststor Amplifier Deslgna
In an identical procedure, the circle equation for the output VSWR is found. The
voltage source is attached to the output side and impedance Z L is treated as source
impedance, whereas Zs is the load impedance. Therefore, in a perfectly analogous way
the output reflection coefficient becomes
rout-rL
*
rout- rL
1- rLrout
1- rLrout
(9.92)
We convert (9.92) into a circle equation for r L that is centered at location dv

with
OMN
radius r v
such that
OMN
(rL -dvOM) + (rL -dvOMN)
= TyOMN
(9.93)
where
(9.94)
and
( 1 -lroutl
'v
OMN
>lrOMNI
=
1 -lrOMNrLl2
(9.95)
The previous derivations allow us to draw the following conclusions regarding the constant VSWR circles:
For minimum VSWR (on the input side: VSWR 1MN
output side:
dv
IMN
VSWRoMN
IlriMNI:: o = rin*
= 1, 1rOMNI = 0)
(for the input) and dv
= 1,
jriMNI
= 0; on the
the circles are located at
oMN llroMNI
=o
*
= rout
(for the output)
with both radii equal to zero.

All VSWR circles reside on the line extending from the origin to
*
rout (output).
* (input) or
rin
It is important to be aware of the fact that under bilateral matching the input and
output reflection coefficients are functions of source and load reflection coefficients
(r s' r L ). Therefore, the input and output VSWR circles cannot be plotted simultaneously, but rather have to be considered one at a time in the iterative process of adjusting r S and L
Constant VSWR Circles
509
-----------------------------~&)A~
Example 9-15: Constant VSWR design for given gain and noise
figure
Using the results of Example 9-14, plot the VSWRrMN =1.5 circle in
the r s -plane as part of the Smith Chart. Plot the graph of
VSWRoMN as a function of the r s position for a VSWRIMN = 1.5.
Find r s that gives a minimum reflection on the output port of the
amplifier and compute its corresponding gain.
Solution:
In Example 9-14 we have found rs = 0.29Ll9 and
r L = 0.45L50 as source and load reflection coefficients that meet
specifications in terms of power gain and noise figure. Since we use
the design based on constant operating gain circles, we obtain a perfect match at the input port of the amplifier. However, the output
port is mismatched and the VSWRoMN can be computed from
jr OMNI, which is found from (9.92) in conjunction with (9.9b):
*
rout - rL
= 0.26
1 - r Lrout
The result is
1 + if'oMNi
VSWRoMN = 1
-lroMNI =
1.69
To improve the VSWRoMN' we can relax the requirements on

VSWR1MN and introduce some mismatch at the input. If we set
VSWR1MN = 1.5, the corresponding input VSWR circle can be plotted in the Smith Chart, as shown in Figure 9-19.
The center of the VSWRrMN = 1.5 circle and its radius are
found from (9.90) and (9.91), respectively. The numerical values
yield dv IMN = 0.28Ll9 and rv IMN = 0.18 .
Every point on the VSWRIMN = 1.5 circle can be expressed
in the polar form
r s = dv IMN + 'v IMN exp(ja)

. where the angle a changes from 0 to 360. As a changes, we
obtain a changing r s ' which in turn results in a corresponding rout
510
Figure 919 Constant operating power gain, noise figure, and input VSWR circle
in
r s -plane.
and a computed VSWRoMN The graph of such a dependence is

As can be observed in Figure 9-20, the VSWRoMN reaches its
minimum value of 1.37 at approximately a = 85 o . The corresponding source and output reflection coefficients, transducer gain,
and noise figure are as follows:
fs = 0.39L45,fout = 0.32L-52
GT = 7.82 dB, F = 1.51 dB
An improvement in VSWRoMN has been achieved at the expense of
a reduced gain. If the gain reduction becomes unacceptable, then
both source and load reflection coefficients have to be adjusted
simultaneously.
Many specifications explicitly prescribe a maximum tolerable

VSWR that the amplifier design must meet. This becomes particu-
Broadband, High-Power, and Multistage Amplifiers
2.3
511
--r-------. .. _
--r--------r-- - - - - - - - r - -
2.2
C/)
>'5
2.1
2
1.9
:<
= 1.8
0
'8
~
1.7
VSWRIMN = 1.5
- - - - ' - - - - ' - - - - - ' - - - _ _ _ _ , _ ,_ _L _ __
50
Figure 9-20
100
150
200
250
Angle a, deg.
_ _ . _ _ _LJ
300
350
Input and output VSWR as a function of angle <X
larly important when dealing with system integration issues where

several units are cascaded.
.../
9. 7 Broadband, High..Power, and Multistage Amplifiers

9.7.1
Broadband Amplifiers
Many modulation and coding circuits require amplifier with a wide or broad frequency band of operation. From the RP point of view, one of the major problems in
broadband amplifier design is the limitation imposed by the gain-bandwidth product of
the active device. As pointed out in Chapter 7, any active device has a gain roll-off at
higher frequencies due to the base-collector capacitance in the BJTs or the gate-source
and gate-drain capacitances in the PETs. Eventually, as the frequency reaches the transition frequency f r, the transistor stops functioning as an amplifier and turns attenuative.
Unfortunately 2 seldom remains constant over the wide frequency band of
operation, necessitating compensation measures. Besides forward gain 21 j degradation, other complications that arise in the design of broadband amplifiers include
!S d
IS
512
Increase in the reverse gain
IS 12
1,
which degrades the overall gain even further
and increases the possibility for a device to fall into oscillation

Frequency variation of S 11 and S22
Noise figure degradation at high frequencies
To account for these effects, two different amplifier design approaches are used:
frequency compensated matching networks and negative feedback. In the subsequent
sections we investigate both design techniques.
Frequency Compensated Matching Networks

Frequency compensated matching networks introduce a mismatch on either the
input or output port of the device to compensate for the frequency variation introduced
by the S-parameters. The difficulty with these types of matching networks is that they
are rather difficult to design and the procedures involved are more an art than a well
defined engineering approach that guarantees success. Frequency compensated matching networks have to be custom tailored for each particular case.
The following example demonstrates some of the key steps required to design a
frequency compensated matching network.
----------------------------~&)A~
Example 9-16: Design of a broadband amplifier using a frequency compensated matching network
Design a broadband amplifier with 7.5 dB nominal gain and 0.2 dB
gain flatness in the frequency range from 2 GHz to 4 GHz. For the
design use Hewlett-Packard's AT41410 BJT, which is biased with
I c = I 0 rnA collector current and V CE = 8 V collector-emitter
voltage. The corresponding S-parameters measured at frequencies of
2, 3, and 4 GHz under unilateral assumption are summarized in
Table 9-4.
Solution:
According to the data provided in Table 9-4 the insertion gain of the transistor is IS21 12 = 11.41 dB at f = 2 GHz,
8.16 dB at 3 GHz, and 5.85 dB at 4 GHz. To realize an amplifier
with a nominal gain of 8.7 dB, source and load matching networks
must be designed that decrease the gain by 2.71 dB at 2 GHz and
increase the gain by 0.54 dB at 3 GHz and 2.85 dB at 4 GHz.
513
Table 9-4
S.parameters of AT 4141 0 BJT (I c

V CE
8 V)
= 10 rnA ,
f,GHz
IS2d
Su
S22
3.72
0.61L165
0.45L-48
2.56
0.62Ll49
0.44L-58
1.96
0.62Ll30
0.48L-78
The maximum gain provided by the source and load are found
from (9 .35) and (9 .36) and are as follows:
f = 2 GHz: Gsmax = 2.02 dB,

f = 3 GHz : Gsmax = 2.11 dB,
f
./
= 4 GHz:
Gsmax
= 2.11
dB,
GLmax
= 0.98
dB
GLmax
= 0.93
dB
GLmax
= 1.14 dB
Although for the general case source and load matching networks
would have to be designed, in this example an additional gain G s
that can be produced by the source matching is already sufficient to
meet the amplifier specifications. Therefore, we concentrate on the
development of the source matching network and leave the output
port of the transistor without any matching network.
Since the output of the transistor is directly connected to the
load, we have GL = 0 dB. The input matching network should produce an additional gain of (- 3.9 + 0.2) dB at f = 2 GHz,
(- 0.7 + 0.2) dB at 3 GHz, and (1.7 + 0.2) dB at 4 GHz. The corre
sponding constant gain circles are shown in Figure 9-21 .
The required input matching network must be capable of transforming points on the constant gain circles in Figure 9-21 to the center of the Smith Chart. There are a number of networks that can
accomplish this task. One solution involves a combination of two
capacitors, one in shunt with the transistor and one in series with the
input port of the amplifier, as shown in Figure 9-22. From a known
r S we can COmpute the transducer gain by Setting r L = 0 in (9 .} 0).
We can next find the input and output VSWR. Since r L = 0, the
values for VSWRoMN is equal to VSWRout and is found as
514
./
Figure 9-21
Smith Chart design of a broadband amplifier in Example 9-16.
o.64pFI
Figure 9-22
Broadband amplifier with 8.7 dB gain and 0.2 dB gain flatness over
a frequency range from 2 to 4 GHz.
1+
VSWRout = 1
ISnl
-IS221
For the computation of the VSWR at the input port we use
1 + ifiMNI
vswRIMN
where
= I
-lrIMNI
1riMNI is computed based on (9.88):
515
The obtained values are summarized in Table 9-5.

Table 9-5
Parameters of a broadband amplifier
f,GHz
rs
Gp dB
VSWRIMN
YSWRoMN
0.74L-83
7.65
13.1
2.6
0.68L-101 a
7.57
5.3
2.6
0.66L-ll2
7.43
2.0
2.8
As seen from the values provided in Table 9-5, gain linearity is

achieved at the expense of significantly higher VSWR.
As demonstrated in Example 9-16, the addition of a frequency compensated

matching network to obtain an improved gain flatness may result in significant impedance mismatch, degrading the amplifier perfonnance. To circumvent this problem, a
balanced amplifier can be employed.
Balanced Amplifier Design
The typical balanced amplifier block-diagram using a 3 dB Lange or hybrid coupler and a 3 dB Wilkinson power divider and combiner are shown in Figures 9-23(a)
and (b), respectively. The input signal power is split into two, amplified, and combined
at the output. A complete discussion of the theory behind the operation of couplers and
power dividers is given in Appendix G.
Let us first discuss the operation of the balanced amplifier in Figure 9-23(a). Here
the input power launched into port 1 of the input coupler is equally divided in magnitude, but with a 90 phase shift between ports 2 and 3. No power is present at port 4.
The output coupler combines the output signals of amplifiers A and B by introducing an
additional 90 phase shift, thus bringing them in phase again. We denote the S-parameters of amplifier A as 1 , 2 , S~1 , S~2 , and the corresponding S-parameters of amplifier B with superscript B. The equations that relate the S-parameters of the entire
amplifier to the S-parameters of individual branches are as follows
s1 s1
516
port 1
port 2
port 1
port 2
Output
port 3
port 4
(a) Balanced amplifier using 3 dB coupler
A/4 transformer
3 dB Wilkinson
power divider
power divider
IJ4 transformer
(b) Balanced amplifier using 3 dB Wilkinson power divider and combiner
Figure 9..23
Block diagram of a balanced broadband amplifier.

jSil j
= 2IISlAl -SIIB I
fSztl
= 2IISzlA + SziB I
fS 12l ISzzl -
(9.96)
II
I
2IISzz - SzzI
A
B
2,Stz+S12
A
where coefficients 112 take into account the 3 dB attenuation, and the minus sign is due
to the 90 o phase shift at port 3 that is traversed twice, adding up to 180.
If the amplifiers in the two branches are identical, then ttl = 22 = 0 and the
forward and reverse gain of the balanced amplifier are equal to the corresponding gains
of each branch.
IS
IS
517
The operation of the balanced amplifier with Wilkinson power dividers [see Figure 9-23(b)] is identical. The only difference compared to the power divider is that the
signals are in phase, and we need to add additional A./ 4 transformers to produce a 90
phase shift between branches.
The main advantages of balanced amplifiers are that they possess very good
impedance match at the input and output ports (provided that the amplifiers in both
branches have similar characteristics), and one of the two amplifiers can continue operating even if the other branch should fail completely. The chief disadvantages of balanced amplifiers include increased circuit size and a reduction in frequency response
introduced by the bandwidth of the couplers.
Negative Feedback Circuits

The alternative to frequency compensating networks is the use of negative feedback. This allows a flat gain response and reduces the input and output VSWR over a
wide frequency range. An additional advantage of the negative feedback is that it makes
the circuit less sensitive to transistor-to-transistor parameter variations. The disadvantage of such circuits is that they tend to limit the maximum power gain of the transistor
and increase its noise figure.
__.-/ The term negative feedback implies that part of the signal from the output of the
transistor is coupled back to the input with opposite phase so that it subtracts from the
input signal, thereby reducing it. If the signals are added in phase, the resulting
response will grow and a positive feedback is obtained. The most general resistive feedback circuits for BJT and FET are shown in Figure 9-24, where resistor R 1 constitutes
a shunt feedback and resistor R2 a series feedback.
(a) Feedback in BJTs
Figure 9-24
(b) Feedback in FETs

Negative resistive feedback circuits.
518
As discussed in Chapter 7, both circuits in Figure 9-24 at low-frequencies can be

replaced by the equivalent 1t -models, as shown in Figure 9-25, where the input resistance r 7t is equal to infinity for FETs.
Figure 9-25
Low-frequency model of negative feedback circuit.
If we assume for the BJT that
(9.97)
rtt( 1 + gmR 2 ) R 1
./
then r tt in Figure 9-25 can be replaced by an open circuit and the h-parameter representation can be written as
RI
[h] -
(9.98)
gmRl
----1
1 + gmR2
Using the matrix conversion formula from Appendix D, we find the corresponding
S-parameter representation
2
[S]
(9.99)
where
(9.100)
519
Assuming ideal matching conditions S 11 = S 22 = 0 (i.e., the input and output

VSWRs equal unity) yields the following equation relating the value of the shunt feedback resistor to the series feedback resistor R 1 :
zz
R2 = _Q __
Rl gm
(9.101)
where the characteristic impedance Z 0 and transconductance gm are used.

Substituting (9.101) into (9.100) and (9.99) gives
Zo
0
[S]
R 1 +Z0
Rt
Zo
1--
(9.102)
As seen from (9.99) and (9.102), both gain flattening and perfect match can be
achieved by choosing appropriate values for the feedback resistors R 1 and R2 The
only limitation arises from the requirement that R2 in (9 .l 01) must be nonnegative; that
is, there exists a minimum value g m . that limits the range of g m to
nun
R1
Em
~gmmin = z2
0
l-S21
= z0
(9.103)
Any transistor with gm satisfying condition (9 .l 03) can be used in the negative feedback configuration shown in Figure 9-24.
The analysis of the feedback circuit is applicable only for ideal devices operated
in the low-frequency range where all reactances are neglected. In practical applications
the presence of the parasitic resistances in the transistor must be taken into account,
resulting in modified values of the feedback resistors. In addition, at RF and MW frequencies the influence of internal capacitances and inductances cannot be neglected,
and additional reactive components in the feedback loops enter the analysis. The most
common practice is to add an inductance in series with the feedback resistor R 1 This is
done to reduce the feedback from higher frequencies and thus compensate for S 21 related roll-off.
The following example demonstrates the use of negative feedback for a broadband amplifier design where the feedback resistors are first computed theoretically and
then adjusted using a CAD software package.
Chapter t
520
RF Translator Amplifier Dulgn
---------------------------~~~
E.x ample 9-17: Design of a negative feedback loop broadband
amplifier
The BJT BFG403W is biased with V CE = 3 V and I c = 3.3 rnA
(~ = 125 ). The corresponding S-parameters in common-emitter
configuration are listed in Table 9-6, where a 500 Q resistor has
been added to ensure stability.
Design a broad-band amplifier with Gr = 10 dB and a bandwidth ranging from 10 MHz to 2 GHz by using a negative feedback
loop.
Table 96
5-parameters for the transistor in Example 9-17
f,MHz
jSll j
LSt l
IS211
L S21
jS1 2I
LS12
jS22I
LS22
10
0.877
--0.3
7.035
179.6
1x 10-4
66.8
0.805
--0.1
./
100
0.876
-2.4
7.027
176.1
7x 10-
85.9
0.805
-1.4
250
0.870
-5.9
6.983
170.2
0.002
84.3
0.803
-3.4
500
0.850
-11.5
6.834
160.6
0.003
80.5
0.797
-6.6
750
0.820
-16.9
6.607
151.4
0.004
76.0
0.789
-9.8
1000
0.783
-21.7
6.327
142.8
0.005
68.2
0.777
-12.7
1500
0.700
-29.6
5.711
127.2
0.007
74.1
0.755
-18.1
2000
0.619
-35.7
5.119
113.8
0.007
74.1
0.735
-23.0
Solution:
As seen from Table 9-6, the minimum gain of 14.2dB
is attained at f = 2GHz, which is well above the required transducer power gain of Gr = IOdB.
Before continuing our approximate analysis, we have to ensure
that condition (9.103) is satisfied. The value of r 1t is found to be
r tt = J3! 8m = 984 0, where the transconductance 8m is computed as gm = I cl V T = 0.127 S . Thus, the negative feedback
Broadband, High-Power, and Multlatage AmpiHiers
521
analysis is applicable since condition (9.103) is satisfied even for

R2
= 0.
The next step involves an estimation of the resistances R 1 and

R2 . Because the desired gain is G
10 dB, the low-frequency S 21
coefficient should be equal to - 3.16 . Here the minus sign is due to
the 180 phase shift of the common-emitter configuration. Substituting this value into (9.103) yields
R1
= 2 0 (1 -
S 21 ) = 208
Applying (9.101), we compute the value for the series feedback

resistor R2 :
R2
= Z6
- - -1 = 4.1
R,
8m
The resulting insertion gain of the feedback network is listed in the
second column of Table 9-7. It is observed that the negative feedback makes the gain response of the amplifier more uniform at the
lower frequencies, unfortunately at too low a level. The discrepancy
between the expected gain of 10 dB and the obtained value of
JS 2112 = 7.5 dB is largely due to the fact that we neglected all parasitic resistances in the transistor. Such parasitics include the base
resistance that is connected in series with r 1t and thus reduces the
effective transconductance g m . Furthennore, the emitter resistance,
which is in series with R 2 , has to be subtracted from the obtained
value of R 2 .
Optimization of the circuit for frequencies up to 500 MHz
using CAD tools results in the following modified values of the
feedback resistances: R 1
276 Q and R 2 = 1.43 Q. The corresponding insertion gain is listed in the third column of Table 9-7.
As observed from Table 9-7, these new values for the feedback
resistances bring the transistor gain closer to the 10 dB specification
at lower frequency, but it degrades quickly as the frequency
increases. This indicates that a R 1 = 276 0 feedback resistor is too
small at those frequencies and has to be increased. This can be done
by connecting an additional L 1 == 4.5 nH inductor in series with
the resistor R1 (the value of L 1 is predicted by a separate CAD
optimization procedure).
522
Table 9-7
Insertion gain of the feedback amplifier
IS211
f,MHz
= 208 n,
R2 = 4.1 Q
R1
'dB
= 276 Q,
R2 = 1.4 Q
R1
= 276 n,
R 2 = 1.4 0,
L 1 = 4.5 nH
R1
10
7.50
10.01
10.01
100
7.50
10.01
10.01
250
7.50
10.00
10.01
500
7.50
9.97
10.00
750
7.50
9.93
10.00
1000
7.50
9.88
10.00
1500
7.51
9.75
9.99
2000
7.54
9.59
9.99
The resulting gain is listed in the last column in Table 9-7. As

seen from the values presented, the addition of an inductor flattens
the frequency response and improves the gain flatness to better than
0.1% over the entire bandwidth.
As the frequency increases, the negative feedback design

approach becomes increasingly prone to parasitic influences. Above
approximately 5 GHz, this lumped element method begins to break
down.
9.7.2
High-Power Amplifiers
Thus far we have discussed the design of amplifiers based on linear, small-signal
S-parameters. When dealing with high-power amplifiers, however, a small-signal
approximation is usually not valid because the amplifier operates in a nonlinear region
and large-signalS-parameters or impedances have to be obtained to conduct the appropriate design. Small-signal S-parameters can still be used when designing a Class A
523
Broadband, HlghPower, and Multistage Amplifiers
amplifier. Here the signal amplification is largely restricted to the linear region of the
transistor. However, the small-signalS-parameters become progressively unsuitable for
Class AB, B, or C amplifiers, which operate in the saturation region.
One of the important characteristics of a high-power amplifier is the so-called
gain compression. As the input signal to the amplifier approaches the saturation
region, the gain begins to fall off, or compress. The typical relationship between input
and output power can be plotted on a log-log scale, as shown in Figure 9-26.
~ut (dBm)
,
1 dB
;;..:.',/,
F:,ut, ! dB --~-:'
.
,
,,
...
"
Pout, mds ..
Rm,mds
Figure 9-26
--~- ..
~n.l dB
F:n (dBm)
Output power of the amplifier as a function of input power.
At low drive levels, the output is proportional to the input power. However, as the
power increases beyond a certain point, the gain of the transistor decreases, and eventually the output power reaches saturation. The point where the gain of the amplifier deviates from the linear, or small-signal gain by I dB is called the 1 dB compression point
and is used to characterize the power handling capabilities of the amplifier. The gain
corresponding to the 1 dB compression point is referred to as G 1dB and is computed as
G 1dB = G 0 - 1 dB , where G0 is the small-signal gain. If the output power Pout, 1dB at
the 1 dB compression point is expressed in dBm, it can be related to the corresponding
input power Pin, I dB as
Pout, 1dB(dBm)
= GldB(dB) + Pin,IdB(dBm)
= G0 (dB)- 1 dB+ Pin, Ids(dBm)
(9.104)
Another important characteristic of an amplifier is its dynamic range labeled d R

The dynamic range signifies the region where the amplifier has a linear power gain
524
Chapter 9 RF Translator Amplifier Dulgna
expressed as the difference between Pout 1dB and the output power of the minimum
'
detectable signal P out, mds . The quantity Pout, m d s is defined as a level X dB above the
output noise power Pn, out. In most of the specifications, X is chosen to be 3 dB. The
output noise power of an amplifier is given as
(9.105)
Pn,out = kTBG0 F
which, if expressed in dBm, can be cast in the fonn
Pn,out(dBm) = lOlog(kT) +101ogB+ G 0 (dB) + F(dB)
(9.106)
where IOlog(kT) = -173.8 dBm at T = 300K and B is the bandwidth.

As with any nonlinear circuit, high-power amplifiers create harmonic distortions
(multiples of the fundamental frequency) . They appear as a power loss in the fundamental frequency. In general, Class A operation produces the lowest distortion figures.
For higher-power applications where Class A operation is not feasible, due to low efficiency, Class AB push-pull amplifiers are employed to achieve nearly comparable distortion levels. Harmonic distortion is specified as the harmonic content of the overall
output expressed in dB below the output power at the fundamental frequency.
An undesirable property of power amplifiers is the occurrence of so-called inter
modulation distortion (IMD). Although present in any amplifier (like harmonic distortion) it is most prominent in the high-power region of an active device where the
nonlinear behavior has to be taken into account. Unlike harmonic distortions, IMD is
the result of applying two unmodulated harmonic signals of slightly different frequencies to the input of an amplifier and observing the output, as shown in Figure 9-27 .
......
......
[>
,..
-+---L-..L----+
J;J;
Figure 9-27
,..
-+--_.._......__.___.__
-+
fr h \
2J;- h
2/; - ft
Observing the intermodular distortion of an amplifier.
Due to third-order nonlinearities of the amplifier, the input signals Pin(/ 1) and
Pin(/2 ) create, besides the expected output signals Pout(f 1 ) and Pout(/2 ), additional
Broadband, HlghPower, and Multistage Amplifiers
525
frequencies Pout(2f 1 - f 2 ) and Pout< 2f2 - f 1 ). The additional frequency components

can serve a desirable purpose when dealing with mixer circuits (see Chapter 10). However, for an amplifier one would like to see these contributions to be as small as possible. The difference between the desired and the undesired power level (in dBm) at the
output port is typically defined as IMD in dB; that is,
IMD( dB) = P out(f2 )( dBm)- P out(2f2 -
f 1 ) ( dBm)
(9.107)
In Figure 9-28 the output powers P outCf2 ) and P out(2/2 - f 1) are plotted versus
the input power Pi0 ( /2 ) on a log-log scale. In the region of linear amplification, the
output power Pout<!2 ) increases proportionally to the input power Pin(!2 ), let us say
Pout ( f 2 ) = aPin ( f 2 ) . However, the third order product Pout ( 2 f 2 - f 1) increases pro3
portional to the third power [i.e., P 0 u1(2/2 - / 1 ) = o: Pin(/2 )]. Thus, the IMD is
reduced in proportion to the inverse square of the input power. Projecting the linear
region of P out(f2 ) and Pout( 2/2 - f 1 ) results in a fictitious point called the intercept
point (IP). In practice, if higher than third order products can be neglected, the IP
becomes a fixed point, independent of the particular power gain of the amplifier. This
allows us to us the IP as a single number to quantify the IMD behavior.
~lUI(dBm)
..
JP..1ut -----:~'
'
,.
/
/
,,''
~ut, ldb --------------;7'~ ,

'
,'
," /
p (f)
Ollt
:.
IMD
..
'
dfl~
4 l ',~(2!,- ft)
3
Pout, mds ----------------------------,------------------------+pin-,m-d-s--------~----------------+Rn(dBm)
Figure 9..28
Recording of IMD based on input-output power relation.
Also shown in Figure 9-28 is a quantity called spurious free dynamic range, d 1 ,
which is defined as
= 32 [IP( dBm)-G0 ( dB )-Pin, mds( dBm)]
(9.108)
Chapter t
526
Typical values for a MESFET are

d 1 = 85 dB.
9.7.3
Pin, mds
RF Tranalator Amplifier Deslgna
== -100 dBm. IP
= 40
dBm, and
Multistage Amplifiers
A multistage amplifier circuit should be considered if the power gain requirement

of the amplifier is so high that a single stage may not be able to achieve it. A typical
example of a dual-stage BJT amplifier is shown in Figure 9-29.
r Q; --Q~--- l
~ ~.-
i~
............................................................ -............... -............................................. -- - ......
Figure 9-29 Dual-stage transistor amplifier.

Besides the typical input and output matching networks ( M N 1 and M N 3 ), this
configuration features an additional so-called interstage matching network ( M N 2 )
for matching the output of stage 1 with the input of stage 2. In addition to providing
appropriate matching, M N 2 can also be used to condition the gain flatness.
Under the assumption of optimally matched and lossless networks, let us summarize the most important dual-stage perfonnance parameters. The total power gain G,0 ,
of a dual-stage amplifier under linear operating conditions results in a multiplication of
the individual gains G 1 and G 2 , or in dB
(9.109)
(9.110)
In addition, if the minimal detectable signal Pin, mds at 3 dB above thermal noise at the
input is given by Pin, mds = kTB + 3 dB+ F 1 , the minimal detectable output power
p o ut, mds becomes
Pout,mds(dBm) = kTB(dBm) + 3dB + F tot(dB) + G101 (dB)
(9.111)
527
The dynamic properties are also affected. For instance, Rhode and Bucher (see Further
Reading) have shown that the previously mentioned third-order intercept point changes to
(9.112)
where I P 1 and I P 2 are the third order intercept points associated with stages 1 and 2.
Finally, the total spurious-free dynamic range d ftot is approximately
d.ftot( dBm)
= I P 101 ( dBm)- Pout, mds( dBm)
(9.113)
Equation (9 .113) also reveals that the addition of a second stage reduces the total
dynamic range.
-----------------------------~~~
Example 9-18: Transistor choices for multistage amplifier
design
Design an amplifier with Pout, 1 dB = 18 dBm and a power gain not
less than 20 dB. Using the transistor choices listed in Table 9-8,
which shows pertinent characteristics at the operating frequency of
f = 2 GHz, determine the number of stages for the amplifier and discuss the choice of an appropriate transistor for each stage. In addition, estimate the noise figure Ftot and the third-order intercept point
IPtot of the amplifier.
Table 9-8
Transistor characteristics for Example 9-18.
Transistor
F[dB]
Gmax[dB]
pout I dB [dB m]
'
IP[dBm]
BFG505
1.9
10
10
BFG520
1.9
17
26
BFG540
21
34
Solution:
Since the output power should be 18 dBm, the only

transistor choice for the output stage of the amplifier is BFG540.
528
Chapter 9 RF Transistor Ampllfler Designs
Because the output power of the amplifier Pout,ldB = 18 dBm is

much lower than Pout,ldB of the BFG540, it can operate at maximum
gain of G =7 dB. This means that the remaining stages of the amplifier must be able to provide at least 20 dB - 7 dB = 13 dB of gain.
Thus, our amplifier should have at least three stages.
For the last stage to have 18 dBm output power, the secondstage transistor should be able to produce a power level of
Pout , 1dB = 18 dBm - 7 dBm = 11 dBm, which eliminates BFG505
2
from the list of possible candidates. Since the BFG540 has a much
higher power handling capability than necessary for the second
stage, we choose BFG520.
Due to the fact that Pout,! dB = 11 dBm is much lower than the
1-dB compression power of the BFG520, the second-stage transistor
will also operate well below the compression point and the maximum gain will be equal to G max = 9 dB . Therefore, the transistor
in the first stage has to have a minimum gain of G = 13 dB - 9 dB =
4 dB and be able to provide P out = 11 dBm - 9 dB =2 dBm. Thus,
I
the BFG505 is more than adequate for the task with Pout = 2 dBm
I
and G 1 = 4 dB. The input power to the amplifier is then
Pin= -2 dBm.
As shown in Appendix H, the noise figure of the entire amplifier is computed as
F tot = F I +
F 2 -l
G
F 3 -1
+G G
1
and is minimized if the gain of the first stage is high. The BFG505
cannot provide a gain higher than 6 dB because in this case (for a
given Pin) it reaches the compression point. This difficulty is avoided
if the BFG520 is used as the first stage. We can design the first stage
for maximum gain and the second stage for necessary power to drive
the output transistor. We can also adjust the gains of the individual
stages so that none of the transistors reaches the compression point.
The block diagram of the resulting amplifier is shown in Figure
9-30, where the gain of each stage is chosen according to the preceding discussion. The noise figure of this amplifier is predicted as
F tot = F 1 +
F 2 -1 F 3 - 1
G +G
= 2.13 dB
1
1 G2
summary
529
G l =9dB
Pout!=
Figure 9-30
7 dBm
~,u 1 2 =
14 dBm
Block diagram of a three-stage amplifier.
The output power at the third-order intercept point is calculated

using (9 .112) and modified for a three-stage amplifier
1
/Ptot =
= 28 dBm
l / IP3 + l/(G 3JP2 ) + l / (G3 G 2 /P 1)
where the preceding formula was obtained from (9 .112) by first
computing the I P of the first two stages and then resubstituting it
into (9.112).
The above analysis is actually one of the first steps required in

an amplifier design process. Here the cruciaL steps of picking suitable transistor types and deciding on the number of stages are
made. They then become the starting point of a detailed performance analysis.
9.8 Summary
This chapter deals with a broad spectrum of amplifier design concepts. First, the
various power relations are defined. Specifically, the transducer power gain
Gr =
2
2
( 1 - Ir L! )Is21! ( I -
Ir sl 2)
2
2
jl - r srinl II - S22rd
as well as the available and operating power gains are of key importance. We next
establish the various input and output stability circle equations and examine the meaning of unconditional stability. Specifically, the factor
530
is employed to assess the unconditional stability of an active device. If the transistor

turns out to be unstable, additional series or shunt resistances can be used to stabilize
the device. Next the constant unilateral gain circles are established and displayed in the
Smith Chart. The location and radius equations
d
g. I
g l.s~.ll
2
1 - ISiii (1 - g;)
an
rg .
Jl-g-(l-ISI2 )
2
1 - ISiii (1 - g i)
l
II
provide insight as to where certain constant gain values are located under unilateral
design conditions (inverse power gain is assumed negligible). The error committed by
using the unilateral design approach over the bilateral method is quantified through the
unilateral figure of merit. If the unilateral approach turns out to be too imprecise, a
bilateral design has to be pursued, leading to the simultaneous conjugate matched
reflection coefficients ( r M s , r M L) at the input and output ports. The optimal matching
r MS =
st2s 2IrML
S II+ 1- SzzrML
s ,zSz,rMs
and
r ML= Sz2 + 1- S II r MS
results in amplifier designs with maximum gain. Starting from the operating power gain
expression, circles of constant gain under optimal source matching are derived. Alter
natively, starting with the available power gain expression, circles of constant gain
under optimal load matching are derived.
We then investigate the influence of noise generated by an amplifier. Using the
noise figure of a generic two-port network
Rn
F = F min+ 0
1Ys- yoptl
circle equations for the Smith Chart are computed. The noise figure circles can be used
by the circuit designer to make trade-offs with the previously conducted constant gain
analysis.
An investigation into reducing the VSWR as part of various input and output
matching network strategies results in an addition set of circle equations that quantify
the VSWR at the matching network ports:
531
Further Reading
Combining the various circle representations permit the small-signal amplifier design
based on constant operating gain, noise figure, and VSWR circles, jointly displayed in
the Smith Chart.
For broadband design, we discuss the need to develop frequency compensated
matching networks in an effort to widen the operational frequency range. The use of
negative feedback loops is introduced as a way to flatten the power gain over the broadband frequency range.
In high-power amplifier applications issues related to the output power compression are of major concern since they limit the dynamic range of amplification. An
important figure of merit is the 1-dB compression point:
Pout, ldB(dBm)
= G 0 (dB) -1 dB+ Pin, ldB(dBm)
Furthermore, an additional undesirable property is the occurrence of intermodular distortion due to the presence of nonlinearities. Finally, the influences of power compression, noise figure, and gain are investigated in the context of a multistage amplifier
design.
Further Reading
I. Bahil and P. Bhartia, Microwave Solid State Circuit Design, John Wiley, New York,
1988.
G. Gonzalez, Microwave Transistor Amplifiers, Analysis and Design, Prentice Hall,
K. C. Gupta, R. Garg, and R. Chada, Computer-Aided Design of Microwave Circuits,

Artech, Dedham, MA, 1981.
Hewlett-Packard, RF Design and Measurement Seminar, Seminar Notes, Burlington,
MA, 1999.
Hewlett-Packard, S-Parameter Techniques for Faster and more Accurate Network
Design, Application Notes 95-1, 1968.
H. Krauss, C. Bostian, and F. Raab, Solid Radio Engineering, John Wiley, New York,
1980.
S. Y. Liao, Microwave Circuit Analysis and Amplifier Design, Prentice Hall, Englewood
Cliffs, NJ, 1987.
S. J. Mason, "Power Gain in Feedback Amplifiers, IRE Trans., Vol. 1, pp. 20-25, 1954.
532
Chapter 9 RF Transl1tor Amplifier Design
D. Pozar, Microwave Engineering, John Wiley, New York, 1998.

B. Razavi, RF Microelectronics, Prentice Hall, Upper Saddle River, NJ, 1998.
U. L. Rohde and T. T. N. Bucher, Communication Receivers, Principle and Design,

J. M. Rollett, "Stability and Power-Gain Invariants of Linear Two-Ports," IRE Trans.,

Vol.9,pp. 29-32, 1962.
G. D. Vendelin, Design ofAmplifiers and Oscillators by the S-Parameter Method, John

Wiley, New York, 1982.
Problems
9.1
The available power of an RF source driving an amplifier connected to load

Z L = 80 n can be represented as
pA =
! lbsl
21
- lfsl2
Based on the signal flow graph shown in Figure 9-2(b),

(a) Find the power to the load p L in terms of r L' r s' and b s.
(b) For Zs = 40 .Q, Z 0 = 50 .Q, V s = 5 V L0, find the available power
P A and the power at the load P L .
9.2
Use the signal flow graph in Figure 9-2(b) and establish the validity of equation (9.8) in Section 9.2.2.
9.3
An amplifier is characterized by the following S-parameters:

S 11
0.78L-65 ,S21
2.2L78 ,S 12 = 0.11 L-21 o , S22 = 0.9 L-29.
The input side of the amplifier is connected to a voltage source with
vs = 4 v L0 ' and impedance zs = 65 n . The output is utilized to drive
an antenna that has an impedance of L = 85
Assuming that the Sparameters of the amplifier are measured with reference to a Z 0 = 75 .Q
characteristic impedance, find the following quantities:
(a) transducer gain GT , unilateral transducer gain GTu, available gain G A,
operating power gain G
(b) power delivered to the load P L , available power P A , and incident power
to the amplifier Pine
n.
Problems
533
9.4
A FET is operated at f = 5.5 GHz and under bias conditions

V Ds = 3.2 V and I D = 24 rnA. The S-parameters are S 11 = 0.73Ll76,
S21 = 3.32L75, S 12 = 0.05L34, S22 = 0.26L-107. In the absence
of matching networks a load of ZL = 75 n and a source of Zs = 30 n are
attached. Assume Z0 = 50 0.
(a) Find Gru, Gr, GA, and plot the magnitude of Gru for
10 nszL~ 100 n .
(b) Match the input side for the unilateral case and find Gru.
(c) Match both input and output for the unilateral case and compute
Gru
9.5
= Grumax
Unconditional stability in the complex rout -plane requires that the

r sl = 1 domain resides completely within the lroutl = 1 circle, or
< 1, where
ICsl- rsl
Cs =
*
s
1zSz1 s"
Szz +
2
1 -IS 1d
(a) Derive these two equations.

(b) Find the circle equations
2
l 8 1zSzd < 1 -1Szzl
9.6
9.7
ls
for
CL
- ISI d
and
and show that
rL
Prove that 11 -s;2 ~\ = IS 12S2 d + (1-IS 1d )(IS22I -1~1

key identity in the stability factor derivation of Example 9.2.
2
).
This is a
A BIT has the followingS-parameters (see the table below) as a function of

four frequencies. Determine the stability regions and sketch them in the
Smith Chart.
s,l
s1z
Szt
Szz
500MHz
0.70L- 57
0.04L47
10.5Ll36
0.79L-33
750 MHz
0.56L-78
0.05L33
8.6L122
0.66L-42
lOOOMHz
0.46L-97
0.06L22
7.1Ll12
0.57 L-48
1250MHz
0.38L-ll5
0.06Ll4
6.0L104
0.50L-52
Frequency
9.8
IS tzSztl
and rs =
The S-parameters for a BJT at a particular bias point and operating frequency are as follows:
S 11 = 0.60L157 ,
S21 = 2.18L61 ,
534
S 12 = 0.09L77 , S 22 = 0.47 L-29. Check the transistor stability, stabilize it if necessary, and design an amplifier for maximum gain.
9.9
In this chapter we have derived the circle equations for constant operating
power gain. It can be concluded that the maximum gain is obtained when the
radius of the constant gain circle is equal to zero. Using this condition, prove
that the maximum achievable power gain in the unconditionally stable case is
_ jS2d
Grmax -
I 2
ISJzl (k- ,..;k
- 1)
where k is the stability factor ( k > 1 ).

9.10
A BJT is operated at f = 750 MHz (and with the S-parameters given as

follows: S 11 = 0.56L- 78, S 21 = 0.05L33 , S 12 = 8.64L122, and
S 22 = 0.66L-42 ). Attempt to stabilize the transistor by finding a series
resistor or shunt conductance for the input and output ports.
9.11
In Example 9-2 the stability factor k is derived based on the input stability
circle equation. Start with the output stability circle equation and show that
the same result (9 .24) is obtained.
9.12 A BJT is operated at f = 7.5 GHz and is biased such that the S-parameter
is given as S 11 = 0.85 L I05 . It is assumed that the transistor is unconditionally stable so that the unilateral approximation can be applied. Find the
maximum source gain and plot the constant source gain circles for several
appropriately chosen values of g s.
9.13
A MESFET is used as a single-stage amplifier at 2.25 GHz. The S-parameters at that frequency and under given bias conditions are reported as
S 11 = 0.83L-132 ,
S 12 = 0.03L22 ,
S21 = 4.9L71 o,
S 22 = 0.36L-82 . For a required 18-dB gain, use the unilateral assumption
by setting S12 = 0, and
(a) Determine if the circuit is unconditionally stable.
(b) Find the maximum power gain under the optimal choice of the reflection
coefficients.
(c) Adjust the load reflection coefficient such that the desired gain is realized using the concept of constant gain circles.
9.14 A BJT is used in an amplifier at 7.5 GHz. The S-parameters at that frequency
and under given bias conditions are reported as S 11 = 0.63L-140 t
535
Problems
S 12 = 0.08L35, S21 = 5.7 L98,
S22 = 0.47 L-57. The design
requires a 19 dB gain. Use the unilateral assumption and

(a) Find the maximum power gain under the optimal choice of the reflection
coefficients.
(b) Adjust the load reflection coefficient such that the desired gain under
stable operating conditions is realized.
9.15
A small-signal amplifier for a BJT operated at 4 GHz is appropriately biased

and
has
the
following
S-parameters:
S 11
0.57L-150,
S 12 = 0.12L45, S 21 = 2.0L56, S 22 = 0.35L-85. If a unilateral
design approach is pursued, estimate the error involved.
9.16 A BJT with I c = 10 rnA and V CE = 6 V is operated at a frequency of

f = 2.4 GHz. The corresponding S-parameters are S 11 = 0.54L-70,
S 12 = 0.017 L176, S 21 = 1.53L91 o, and S 22 = 0.93L-15. Determine
whether the transistor is unconditionally stable and find the values for source
and load reflection coefficients that provide maximum gain.
9.17 Using the same BJT discussed in the Problem 9.16, design an amplifier
whose transducer power gain is 60% of GTmax. In addition, ensure a perfect
match on the input port of the amplifier.
9.18 A MESFET operated at 9 GHz under appropriate bias conditions has the
S 11 = 1.2L-60,
S 12 = 0.02L0 ,
following
S-parameters:
S 21 = 6.5Lll5, and S 22 = 0.6L-35. Design an amplifier that stays
within 80% of GTUmax. Moreover, ensure that VSWRout = 1.
9.19
In Section 9.4.4 it is mentioned that the constant gain design for a matched
input results in the circle equation
sll-r;
~- S22r;-
2
Eo
= r2go
Show that the center d 80 and radius r 80 are given by

0
Es
and
rg IS12S21I
lll-S22dgol2-rfoiS22121
= ~----~---------
536
jrs- d8al
= r8 a [see (9.66)], show
9.20
For the constant available gain circle

that
9.21
A BFG 197X transistor is biased at V CE = 8 V and I c = 10 rnA and has

the following S-parameters measured at f = 1 GHz: S 11 = 0.73LI76
s12 = 0.07 L35 s21 = 3.32L75' and s22 = 0.26L107. Detennine
the unilateral figure of merit and compare the transducer gain of the amplifier designed under the unilateral and bilateral assumptions.
I
9.22 The BFG33 BJT is biased under V CE

following noise and S-parameters:
= 5 V and I c = 5
rnA and has the
F min' dB
ropt
R",n
0.72L-39 0.05L63 6.22Ll35 0.78L-32
2.3
0.64L5
58.5
lOOOMHz 0.45L-70 0.08L56 5.13L109 0.61L-43
2.5
0.56Ll3
67.5
2000MHz 0.18L-ll5 0.12L54 3.24L82 0.49L-54
3.0
0.52L39
49.7
S II
500MHz
sl2
S21
s22
Design a broadband low-noise amplifier with minimum gain of 10 dB and a

noise figure not exceeding 3.5 dB.
9.23
Design a microwave amplifier using a GaAs FET whose S-parameters at

f = 10
GHz
are
S 11 = 0.79Ll00,
S 12 = 0.20L-2l 0 ,
S21 = 6.5.L-73, S22 = 0.74L152. Analyze the trade-offs posed by stability, gain, and VSWRs.
9.24
A broadband amplifier with nominal characteristics of VSWRin = 4,

VSWRout == 2.8, and Gr = 10 dB is used as part of a balanced amplifier
design. Compute the worst input and output VSWR and the insertion gain of
the balanced amplifier if the values listed can vary by as much as 10%.
9.25
In Section 9.7.3 we have listed equation (9.112) for the IP definition of a

two-stage amplifier.
(a) Derive a generalized formula for the IP computation of an N-stage
amplifier.
537
Problems
(b) Compute the total IP and the noise figure of the N-stage amplifier
assuming that all stages are identical and have I Pout = 35 d.Bm ,
F = 2 dB, and G = 8 dB.
9.26
Design a 15-dB broadband amplifier using a BJT with feedback loop. Calculate the value of the feedback resistor and find the minimum collector current of the transistor. Assume that the amplifier is operated at T
300 K.
9.27
A transistor has the following S-parameters: S 11 = 0.61L152,

S 12 = 0.1 L79, S 21 = 1.89 L55, and S 22 = 0.47 L-30. Design an
amplifier for minimum noise figure if Fmin = 3 dB, ropt
0.52L-153,
and Rn = 9 0.
9.28
Prove equation (9.113), which states the total spurious-free dynamic range.
9.29
An amplifier has a transducer gain of Gr = 25 dB, and a 200 MHz bandwidth. The noise figure is given as F = 2.5 dB and the 1 dB gain compression
point is measured as Pout, IdB = 20 dBm. Calculate the dynamic range and
the spurious-free dynamic range of the amplifier if IPout
40 dBm.
Assume that the amplifier is operated at room temperature.
9.30
An amplifier has a gain of G =8 dB at 1 GHz and lists a 1 dB compression

point of Pout, 1 dB = 12 dBm and the third order intercept point at /Ptot = 25
dBm. Find the third order intercept points for the cascaded amplifier stages 2
and 3. What value of /Ptot is obtained in the limit of an infinite number of
stages?
9.31
Derive a formula for the noise figure of a balanced amplifier. Make the
assumption that the power gains and noise figures of the amplifiers in the
individual branches are G A , G8 , and FA , F 8 , respectively. Assume that the
balanced amplifier uses 3 dB hybrid couplers at the input and output ports.
CHAPTER
10
Oscillators and Mixers
W.th the advent of modern radio and radar systems came the need to provide stable harmonic oscillations at particular carrier frequencies to establish the required modulation and mixing conditions. While the carrier
frequencies in the early days mostly reached into the low to mid MHz range, today's
RF systems easily surpass the 1 GHz point. This has resulted in the need for specialized
oscillator circuits capable of providing stable and pure sinusoidal responses. What
makes the design of oscillators such a difficult task is that we exploit an inherently nonlinear circuit behavior that can only be described incompletely with linear system tools.
Specifically, the small-signallinear circuit models utilized to represent the active device
provide limited capabilities to handle the complicated feedback mechanism. Moreover,
since an oscillator has to provide power to subsequent circuits, frequency-dependent
output loading often plays an important role. It is for these reasons that the design process of oscillators remains more of an art than an exact engineering design task. This
holds particularly true for the high-frequency regime, where parasitic component influences can significantly impact the overall system performance. Affected in part by the
additional resonance effects of the passive circuit element, it is possible that the oscillator not only operates at the intended frequency but also at lower or higher harmonics.
Certain system realizations may even cease to oscillate completely.
In the first part of this chapter we concentrate on the negative resistance and feedback harmonic oscillators as well as a number of Schottky diode mixers. Once the fundamental idea is mastered of how to generate oscillations, we investigate the basic
Colpitts and Hartley resonators before moving to the modem RF circuit design
approaches involving the S-parameters of the active device in conjunction with the various network configurations.
539
Chapter 10 Oscillators and Mixers
540
In the second part of this chapter, we tum our attention to the basic frequency
translation tasks performed by mixers. Of the many different circuit implementations
for a wide range of applications, the main emphasis in this chapter is placed on downconverters. A typical application of a mixer in a receiver system is to convert the RF
input signal into a lower frequency intermediate signal that is generally more suitable
for subsequent signal conditioning and processing. This conversion is accomplished by
combining the RF input with a local oscillator signal as part of a multiplication operation that requires a nonlinear, at least quadratic transfer function. Primarily transistors
and diodes are nowadays in use where present FET technology permits the construction
of mixer circuits up to 50 GHz and with diode mixers already exceeding the 100 GHz
mark.
10.1 Basic Oscillator Model

At the core of any oscillator circuit is a loop that causes a positive feedback at a
selected frequency. Figure 10-l(a) illustrates the generic closed-loop system representation, while Figure 10-l (b) provides a two-port network description.
~(ro)
Vi
in
HA(ro)
r~
H,..(ro)
..._
(a) Closed-loop circuit model

Figure 10-1
t-
HF(oo)
~
(b) Network representation
Basic oscillator configuration.
The mathematical condition for a circuit to oscillate can be established by combining

the transfer functions of the amplification stage H A ( ro) with the feedback stage
H F( ro) to the closed-loop transfer function:
(10.1)
Since there is no input to an oscillator, V in = 0 , to obtain a nonzero output voltage,

V out , the denominator in ( 10.1) has to be zero. This requirement leads to the
Barkhausen criterion, which is also known as the loop gain equation:
(10.2)
541
Basic Oscillator Model
If the feedback transfer function H F( ro) is written as a complex quantity [that is,
H F( co) = H Fr( co) + jH Fi( co)] and the amplifier transfer function possesses a real valued gain, H A (co) = H A 0 , we can re-express ( 10.2) as
H Ao=
1
HF,.(ro)
(10.3a)
(10.3b)
The conditions (10.2) and (10.3) apply only for a steady-state situation. Initially, we
have to require that H AoHFr( ro) > 1 . In other words, the loop gain has to be larger than
unity to obtain an increasing output voltage. However, the voltage must reach a steady
state (i.e., the amplitude eventually must stabilize). This nonlinear behavior of the oscillator is shown in Figure 10-2.
HAO
HQ ............................................
Negative slope
(negative resistance)
Figure 10-2 Output voltage versus gain characteristic.
A negativ~ slope of the curve is needed to ensure a decrease in gain for increasing
voltage. At point j V out! = V Q for H AO = H Q = H Fr( ro) the stable operating point is
reached. A similar curve can be established for the frequency versus loop gain with a
stable resonance frequency f Q .
10.1.1 Negative Resistance Oscillator
To explain the idea behind an oscillator we need to understand the seemingly

impossible concept of creating a negative resistance. The requirement of a negative
resistance is best explained by investigating the series resonance circuit consisting of
resistance R, inductance L, and capacitance C. As an input we use a current-controlled
voltage source, as shown in Figure 10-3. The voltage source can represent the output of
an active device, e.g. BJT or FET.
Chapter 1o Oscillators and Mixers
542
v(i)~
Series resonance circuit wtth voltage-controlled source term.
Figure 10-3
The governing equation in terms of the current is written as

2
d i(t) Rdi(t) 1 '( ) _

L dt2 + --;{! + cz t -
dv(i)
-T
(10.4)
If we set the right-hand side to zero (i.e., we reach steady state and the voltage amplitude is stable), we obtain the standard solution
at
i(t) = e (I 1e
jmQt
- jmc;t
+12 e
(10.5)
where a = -R/(2L) and roQ = Jli(LC)- (R/(2L)) 2 . In general, because a is a

negative quantity, the harmonic response of the resonance circuit will reduce to zero as
time progresses. In the limit, as R reaches zero, an undamped sinusoidal response is
obtained. The goal of an active element in the oscillator is therefore to generate a source
response that compensates for the resistance in the circuit. This can only be achieved if
a negative resistance is provided. Thus, if we succeed in selecting a nonlinear device
2
whose voltage-current response is v(i) = v0 + R 1i + R 2 i + ... , then the terms may be
adjusted in such a way as to compensate for R. Indeed, substituting the first two terms
of this series expansion into (10.4), we see
2
1 .( ) = _ dv(t) = -R di(t)
1 dt
dr2 + --;{! + cz t
dt
L d i(t)
Rdi(t)
(10.6)
Combining the coefficients of the first derivative leads to
R + R1 = 0
(10.7)
as the requirement to set the attenuation coefficient to zero. It is now seen that (10.7)
implies a device with a negative differential resistance:
(10.8)
Moreover, to get the oscillations started, we require a positive attenuation coefficient,

which implies R 1 to be less than -R. This situation is equivalent to the transfer fuction
having poles in the right-hand side of the complex frequency domain.
S43
A direct way to implement such a negative resistance condition is via a tunnel

diode, whose electric circuit representation is discussed in Chapter 6 (see Figure 6-26).
Figure I0-4 depicts both the circuit of a tunnel diode oscillator and the corresponding
small-signal circuit model. Since the tunnel diode already possesses an inherent capacitance, an extra capacitor in the external circuit is not required.
Twmel
-g
diode
(a) Tunnel diode oscillator circuit
(b) Small-signal equivalent circuit
Figure 10-4 Tunnel diode oscillator circuit and its small-signal model.
Circuits like the one shown in Figure 10-4 can be used for oscillators with resonance frequencies up to 100 GHz.
10.1.2 Feedback Oscillator Design
Because of their fundamental importance in the development of low-frequency as

well as RF oscillators, let us next focus on the two-port feedback networks, shown
generically in Figure 10-5.
[>
(a) Pi-type feedback
[>
(b) T-type feedback
Figure 10-5 Feedback circuits with Pi- and T-type feedback loops.
544
It is straightforward to find the transfer function of the feedback loop. For instance,
for the Pi-network we obtain under high-impedance input and output assumptions
Hp(ro)
= -vl
v out
zl
zl + z3
(10.9)
More complicated is the computation of the transfer function H A ( ro) of the amplifier.
This depends on the chosen active element and its electric equivalent circuit model. To
demonstrate the concept, we use a simple, low-frequency FET model with voltage gain
J.lv and output resistance R8 . The corresponding loop equation for the circuit depicted
in Figure 10-6 is
J.lvV t +lsRs+IBZc = 0
(10.10)
where 1/Zc == Y c = l/Z2 + l/(Z 1 + Z 3 ).
Figure 10-6
Feedback oscillator with FET electric circuit model.
Solving (10.10) for I 8 and multiplying by Zc gives us the output voltage V out from
which the voltage gain is found to be
HA(ro)
= -VVoutl = y c-RBflv+ 1
(10.11)
The closed-loop transfer function is thus

Hp(ro)HA(ro)
= z2zt + z2z3-J.lvZ1Z2
=1
+ Rs(Zt + z2 + Z3)
(10.12)
This equation allows us to design various oscillator types depending on the choice of
the three impedances in the feed-back loop. To eliminate resistive losses, we choose
purely reactive components Z; = }X; (i= 1,2,3). This ensures that the numerator is
real. Further, to make the denominator real, it is necessary that X 1 + X2 + X3 = 0,
which implies that one of the reactances has to be the negative sum of the others. It is
understood that negative-valued reactances correspond to capacitors and positive-
545
valued reactances identify inductors. For instance, if we decide

X 3 = -(X 1 + X 2 ) , then, upon substitution into (10.12), the result is
JlvX1X2
-XzXI + Xz(Xt + Xz)
Jlv
= -X
1
X2
= 1
to use
(10.13)
It is apparent that X 1 and X2 must have the same sign but different values according to
( 10.13). In Table 10-1 a few possible configurations of the feedback loop are summarized.
Table 10-1
>~O
Vadous feedback configurations for oscillator designs based on

Figure 10-5(a)
XI, X z
o--1~
x3
o--1~
CJ
II
z3
Qz,:
~ ~3
L1~
')
~L2
')
~I
I . I ~2
Colpitts
Hartley
c~m:
.'I
I.'
Clapp
Two often used realizations are the Hartley oscillator, where X 1 = roL 1 ,
X2 = roL 2 , X 3 = 1/( roC3 ), and the Colpitts oscillator, where X 1 = I / ( roC 1),
X 2 = 1/ (roC2 ) , and X 3 = roL3 , as depicted in Figure 10-7, where a FET is
employed as active device. Here resistors RA, R 8 , Rv, and R 5 set the DC bias point.
C s is the RF bypass capacitor, and C8 denotes DC blocking capacitors.
The various choices of L and C element combinations are in practice limited by
the range of realizable values for a given frequency. Often hybrid configurations are
used; if, for instance, the inductance becomes very small, a capacitor connected in
series can yield a larger effective inductive reactance (Clapp oscillator).
Besides the standard conunon-source (or common-emitter for a BJT) configuration,
common-gate (common-base) and common-drain (common-collector) type oscillators
can be constructed, as shown in Figure 10-8 where all DC biasing elements are omitted.
546
(a) Hartley oscillator
(b) Colpitts oscillator
Figure 10-7 Hartley and Colpitts oscillators.
S ----.
D
G
z_,
(a) Common gate
(b) Common source
(c) Common drain
Figure 10-8 Common gate, source, and drain configurations.

10.1.3 Design Steps
What makes the oscillator design so complicated is that the nonlinear electric
equivalent circuit describing the active device (BJT, PET) becomes increasingly complicated as the frequency increases. Moreover, the oscillator has to drive additional circuits and must therefore provide a certain amount of power. This output loading affects
the oscillator in terms of frequency stability and waveform purity.
To provide the reader with a glimpse of the essential steps involved, we will at
first examine the design of a low-frequency Colpitts oscillator. The h-parameter configuration with the appropriate feedback loop is depicted in Figure 10-9. The corresponding Kirchhoff voltage mesh equations involving input, output, and feedback loops are
547
h---------------------------- ----------h-::r
II
22
c,
L:.
Figure 10-9
Colpitts oscillator design.
established by utilizing the output voltage V 2

matrix form for the unknown currents, we obtain
hl2
( h,t-jXcl - hl2h21)
h
22
h21
h22
jXc,
jXct
h 22
= V out = 121 h22 -I 1( h21 I h22 ).
J,
(-1
-JXcz)
h22
-)Xc2
-JXc2
)(XL3-XCI-Xc2)
[2
[,3
0
0
In
(10 .14)
Computing the determinant and setting its imaginary portion to zero results, after
lengthy algebra, in the fonn
! = 2-
~--~--~----~-
1tJc,c2
Jh221hlt + (C, + C2)1L3
(10.15)
Furthermore, setting the real part of the determinant of (10.14) to zero, and assuming
that h t 2 1 , yields a quadratic equation in terms of the capacitor ratio C tIC 2 :
2
cl
2
C
2
cl
ch11h22 -h 12h 2,>- c-hzt +

2
which, under the assumption that
C1 ~
hi
1 4(h'l 1h22 -
h21
(h11 h2z- h12h21)
=o
oo.16)
h 12h 21 ), can be simplified to
C2
(10.17)
The preceding treatment deals with the h-parameters as real quantities, an

assumption that generally may not be applicable. In fact, even for moderately high frequencies~ the h-parameters attain a significant phase angle. To incorporate the actual
Chapter 10 Oscillators and Mlxara
548
frequency-dependent behavior, we need to resort to the equations given in Section

4.3.2. For these situations, explicit formulas as (10.15) and (10.17) are impossible to
derive, and we must resort to a mathematical spreadsheet to find numerical results.
-----------------------------~&uM~
Example 10-1: Design of a Colpitts oscillator
For a 200 MHz oscillation frequency, a Colpitts BJT oscillator in
common-emitter configuration has to be designed. For the bias point
of V CE = 3 V and I c = 3 rnA, the following circuit parameters
are given at room temperature of 25 C: C8 c
0.1 fF,
r 8E
2 kQ, rcE
10 kQ, C 8 E
100 fF. If the inductance
should not exceed L 3 = L = 50 nH, find values for the capacitances in the feedback loop.
Solution:
The first step involves the determination of the hparameters. We compute the values for DC (i.e., f -7 0 ).
hu
= hie =
rBE
= 2000 Q
h21
= hfe =
(g -Jroc )
~E m
BC
1 + JOO(CBE+ CBc)rBE
= 233.32
h 22
= hoe = -rcE +
1 + jro(CBE + CBc)rBE
jroCBc(l +gmrBE+ jroCBErBE)

1 + jro(CBE + CBc)rBE
= 0.1 mS
At DC the h-parameters are real and we can find from ( 10.17) the
ratio between the capacitances c 1 and c2:
h21
ct = (h 11 h 22- h 12 h 21 )C2 = 1166.6C2

Introducing a proportionality factor K such that C 1
tion (10.15) is rewritten as
= K C 2 , equa-
Basic O.cUiator Modal
549
I
1
h
C
~+(l+K)~
= 21tc2JK h11
L
(10.18)
Solving the resonance condition (10.15) for C 2 , we obtain

1 ~K +
c~ )2 16K1t2j2~22
+
SK1t 2 f
Cz =
11
12.68pF
where the inductance L = 50nH has been used.

From the known C 2 we next find C 1 = 1166.6C2 , or
C 1 = 14.79 nF. In the preceding design, the transistor's h -parameters are given under DC conditions. In reality, however, the oscillator is operated at the resonance frequency of 200 MHz. Here the h parameters have the following values:
hll = hie = l
h12 = hre
. (C
+ ]00
'BE
BE+
BC 'B E
jroCBC'BE
.
= (1881- j473)Q
= 1 + ]00
C )
=
. (C
BE+ BC 'BE
-5
5.9x10
hzt
= hfe =
hzz
1
jroCBC(l + gmrBE + jroCBErBE)
= hoe = -rcE
+ - - - - - - - - - - -l+jro(CBE +CBc )rBE
'BE(gm-jroCBc)
1 + jro( C BE+ CBc)rBE
-4
+ J2.4xl0
= 219-j55
= (0.11 + j0.03) mS
As seen, the h-parameters at this frequency differ only slightly from
the DC conditions. Therefore, the analysis should equally apply for
this frequency setting and the oscillator will require only a minimal
amount of tuning.
In practice, the situation often arises where the h-parameters
at a given oscillation frequency differ significantly from their DC
values, necessitating substantial tuning. The difference becomes
more significant as the frequency increases.
Chapter 10 OsciiJators and Mixers
550
10.1.4 Quartz Oscillators

Unlike electric resonance circuits, quartz resonators can offer a number of advantages. A much higher quality factor (up to 105 .. 10 6 ), improved frequency stability,
and immunity to temperature fluctuations are among the chief benefits. Unfortunately,
because quartz crystals are mechanical systems, they cannot be constructed to exceed
approximately 250 MHz.
A quartz crystal exploits the piezoelectric effect whereby an applied electric field
causes a mechanical deformation of the crystal. Depending on the geometric configuration and crystal cut, the crystal performs either longitudinal or shear vibrations at distinct resonance frequencies.
A typical electric circuit representation for a quartz crystal is shown in Figure 1010. The circuit approximates the electric behavior at one of the resonance points for
which the quartz is designed for.
Figure 10-10 Quartz-resonator equivalent electric circuit representation.

The capacitor C q along with Rq and Lq describes the mechanical resonance
behavior while C 0 denotes the capacitance due to the external contacting of the crystal
through electrodes. Normally, the ratio between Cq and C 0 can reach values as high as
1000. Moreover, the inductance Lq is typically in the range from 0.1 mH to 100 H.
The admittance of this model can be stated as
= jroC0 + Rq + j[roLq ~ II( roCq) I = G + jB
(10.19)
The angular resonance frequency ro0 is found by setting the imaginary component B to
zero, or
<Oo C o-
<OoLq- 1/(rooCq)
Rq + [ro 0 Lq- l/(ro0 Cq)1
=0
(10.20)
Solution of this equation (see Problem 10.4) using a Taylor series expansion (and
retaining the first two terms) leads to approximate expressions for the series and parallel resonance frequencies:
Basic Oclllator Model
551
(10.21a)
ro0 =
where ffiso = 11 J LqCq and
is discussed next.
Olp z Olpo[ 1-~(~)]

ffipo
(l0.2lb)
= J( Cq + C 0 )/(LqCqC0 ). A representative model
------------------------~&MW4
Example 10-2: Prediction of resonance frequencies of quartz

crystal
A crystal is characterized by the parameters Lq = 0.1 H,
Rq = 25 .Q , Cq = 0.3 pF, and C 0 = 1 pF. Determine the series
and parallel resonance frequencies and compare them against the
imaginary component of the admittance given by ( l 0.19).
Solution:
As a first approach to compute series and parallel resonance frequencies of the quartz crystal we use (1 0.21 a) and
( 10.21b), respectively:
/p
= /po[l-~~(~;)] = 2~J~;;q~~[~-~(~;)] = 1.048 MHz
The second approach is graphical. At resonance reactance and susceptance of the circuit equal zero; thus we can plot the imaginary
portion of the admittance given by (10.19). Such a plot is shown in
Figure 10-11, where the absolute value of the suceptance is plotted
versus frequency.
Comparing the graphical results with the analytical approach
(10.21), we see that they are virtually the same.
552
10-J
~~--~--~--~--~--~--~--~~--~
10-2
Series
resonance
10-3
10
~ 10"5
N 1o
ft 1o
~
'J
00
10"8
Parallel
resonance
10-9
o- lO~.-..__..jJ.
0.9
_ ___._______.___--1-_ __..___ _.___ _....___ . . _____._ ____.

0.94
0.98
1.02
1.06
1.1
Frequency f, MHz
Figure 10-11
Susceptance response of a quartz element.
Care has to be exercised in selecting quartz crystals due to

their multiple resonances. Depending on the crystal these responses
can be very closely spaced and may result in an undesired oscillation frequency.
10.2 High-Frequency Oscillator Configuration
As the operating frequency approaches the GHz range, the wave nature of voltages and currents cannot be neglected. As outlined in previous chapters, reflection and
transmission coefficients and the associated S-parameter representation are required to
represent the circuit's functionality. This requires us to re-examine (10.1) from a transmission line point of view. The Barkhausen criterion has to be reformulated in the context of the reflection coefficients.
Toward re-expressing the loop gain in terms of transmission line principles, we
recall our signal flow chart representation in Section 4.4.5 (see Figure 10-12).
The input reflection coefficient for matched source impedance ( Z s = Z 0 ) is
r. =
m
bl
al
S11 +
s12s21
l - S22rL
rL =
Sn-~rL
l-S22rL
(10.22)
553
High-Frequency Osctllator Configuration
BJT
(FET)
Load
(a) Sourced and loaded transistor
(b) Equivalent signal flow graph
Figure 10-12
Sourced and loaded transistor and its flow chart model.
where Ll = S 11 S 22 - S12S 21 . This is consistent with definitions given in Example 4-8.

Conducting
the
computation
with
respect
to
the
source
term,
bs = V cJZoi(Zc + Z 0 ), we can define the 1oop gain:
bl
rin
I - fsrin
(10.23)
The equation implies that if
r .In r 5
= 1
(10.24)
at a particular frequency, the circuit is unstable and begins to oscillate.

The identical circuit situation applies if the output side is considered, implying the
condition
(10.25)
for oscillations to occur.
2
2
2
When the stability factor k = (l-jS 11 j - jS22 j +1Ll! ) / (2IS12IIS21 j), see
Chapter 9, is included, the preceding conditions for oscillation can be summarized as
follows:
k<1
(10.26a)
554
(10.26b)
f.m fs= 1
(10.26c)
Since the stability factor is dependent on the S-parameters of the active device we have
to ensure that condition (10.26a) is satisfied first and foremost. If the S-parameters at
the desired frequency do not ensure this requirement, we can switch to a common-base
or common-collector configuration or add a positive feedback to increase instability, as
the following example shows.
----------------------------~&uM~
Example 10..3: Adding a positive feedback element to initiate
oscillations
A BJT is operated at 2 GHz and has the following S-parameters
specified in common-base configuration: S11 = 0.94Ll74,
S 12 = 0.013L-98, S21 = l.9L-28, and S22 = l.OlL-17.
Determine how the Rollett stability factor is affected by adding an
inductance to the base of the transistor ranging from 0 to 2 nH.
Solution:
Using the definition for k gives us without inductance
the value
k
= (I-IS1112-IS2212+1dlz)/(21StziiSztP = -0.25
Accounting for the inductance can be accomplished by redrawing

the circuit in terms of two networks depicted in Figure I 0-13.
t~~-~
o----~...,.
.... .................. .
~
,--!-:---o
~Positive
feedback
loop
Figure 10-13 Network representation of the BJT with base inductance.
555
High-Frequency Oscillator Configuration
In this case the overallS-parameter representation can be found

by first converting the transistor's S-parameters into impedance representation, followed by adding the Z-parameters of the inductor, and
finally converting the result back into S-parameter form.
Using the conversion formulas described in Chapter 4, we find
the Z-representation of the transistor in common-base configuration:
=[
[Z]
-0.42 + j3.43
- 2.17- j0.097]
-95.23- j303.06 - 6.88- j321.03
tr
For the inductor the Z-matrix is given by
[Z].
md
= jroL[l 1] = ljroL jroLl

11
l;roL jroLj
Adding [ Z] tr and [ Z] ind results in the Z-parameters of the entire

circuit, which can then be converted into S-parameters.
To obtain the dependence of the Rollett stability factor as a
function of feedback inductance, we have to repeat the preceding
computations for each value of L. The result of such calculations is
shown in Figure 10-14 (see also file ex10_3.m).
-0.6
....
0
t)
~
-o.7
;.:::::l
.....
~ -0.8
Vl
~
......
0
~ -0.9
0.4
0.8
1.2
1.6
Feed-back inductance L, nH
2.0
Figure 10-14 Rollett stability factor (k) as a function of feedback inductance in

common-base configuration.
556
As seen from Figure 10-14, a maximum instability (minimum

value of k) is obtained by adding a 0.6-nH inductor to the base.
At frequencies in the GHz range, even the lengths of the leads

can be sufficient to create the desired inductance value in the base of
the transistor.
It is interesting to note that if the oscillation condition is met either at the input or
output port, the circuit is oscillating at both ports. This is directly seen by comparing
the reflection coefficients at the input and output ports. We know that
(10.27)
and solving for
r L yields
(10.28)
However, rout can also be written as
s22-ilrs
-sllrs
{10.29)
rout = 1
Therefore, we conclude that (10.28) is the inverse of (10.29), and thus

(10.30)
rL = 1/rout
as required by(10.26c).
10.2.1 Fixed-Frequency Oscillators
A very popular oscillator design approach involves the two-port design where the
transistor configuration is first chosen such that it meets the requirement of k < 1 (an
inductive feedback may have to be added). Next, we select L such that I ml > 1 or s
such that jrout!> 1 . Either case implies the other condition. For instance, if 1
r out!> 1
we conclude that
> 1 and vice versa. A proof is left as an exercise. The following
example details these steps.
1rinl
557
-----------------------------RI8UM~
Example 104: Design of a fixedfrequency lumped element
oscillator
A BFQ65 BJT manufactured by Philips Semiconductors is used in
the common-base configuration with biasing conditions specified by
V CE = 3 V and V BE = 0.9 V. For this case, the transistor has the
following S-parameters measured at 1.5 GHz: S 11 == 1.47 L125,
S 12 == 0.327 LJ 30, S21 = 2.2.L- 63 o, and S22 = l.23L-45.
Design a series feedback oscillator that satisfies conditions ( 10.26)
at/= 1.5 GHz.
Solution:
As the first step in the design process we have to
ensure that the transistor is at least potentially unstable. This can be
tested by computing the Rollett stability factor:
k == (1- \Sillz- ISzzl2+ 1a i2)/ (2 ISiz\ IS2d ) = -0.975

Since k is less than unity, the transistor is indeed potentially
unstable.
Next, we plot the input stability circle to choose a reflection
coefficient for the input matching network. The center and radius of
the input stability circle are computed base on the formulas provided
in Chapter 9:
Tm
S 12S2 1
Is111
Ii11
== 0.82
(S 11- ~S:iz)*
cin
= lsit 12 - I~12 = 0.27 L-57 0
Since [Cm[ < rin and [S22 > 1, the stable region is outside of the
shaded circle, as illustrated in Figure I 0-15.
According to Figure 10-15, we have a great deal of flexibility
in choosing the reflection coefficient for the input matching network. Theoretically, any r s residing inside of the stability circle
would satisfy our requirements. In practice, however, we would like
558
Chapter 10 Oscillators and Mhcert
+1.0
-1.0
Figure 1 G-15
to choose
cient:
rs
Input stability circle for the oscillator design.
such that it maximizes the output reflection coeffi-
rout
s22
s12s21
+1
sII r S rs
(10.31)
From (1 0.31) it is obvious that r out achieves its maximum value

when r s = S}: . In this case we obtain an infinite output reflection
L = 0
(i.e.,
coefficient, which from (10.26c) results in
Z L = Z 0 = 50 Q ). The problem with such an approach is that in
practical realizations it is almost impossible to achieve a perfect
50 Q matching. Moreover, as we approach r s = S}~ , the oscillator
becomes increasingly sensitive to changes in the load impedance. At
r s = S}: , the slightest deviation from the 50 Q value results in
ceasing all oscillations. Because of this phenomenon we choose r s
-I
somewhat close, but not exactly equal, to S11 .
After attempting several values for the source reflection coefficient, we finally select rs = 0.65L- 125. From the knowledge of
r s the source impedance is computed as zs = ( 13 - j25) n '
which is realized by a series combination of a 13 n resistor and a
4.3 pF capacitor, as shown in Figure 10-16.
Next the output reflection coefficient is computed using
(10.31) with the result rout = 14.67 L -36.85 . To determine the
559
HlghFrequency Osctllator Configuration
13
son
{l
'l
Figure 10-16
BFQ65
rr.
Series-feedback BJT oscillator circuit.
output matching network we utilize ( 10.26c) and obtain the value

r L = r~~l = 0.068L36.85 . This corresponds to the impedance
ZL = (55.6 + j4.57)Q = -Zout and can be realized as a series
combination of a 55.6 Q resistor with a 0.48 nH inductor.
The final point that has to be taken into account in our design
is the fact that as the output power of the osci11ator begins to build
up, the transistor's small-signal S-parameters become invalid. Usually, the power dependence of the transistor's S-parameters results in
a less negative Rout
Re{ZQut} for increasing output power. Thus,
it is necessary to choose RL == Re{ZL} such that RL +Rout< 0. In
practice, a value of RL = -Rout/3 is often used. However, we have
to be careful with such a choice because it is only applicable if r s is
sufficiently far away from s~!, as discussed previously. Another
implication of RL-:;::. -Rout is a shift in the oscillation frequency.
In our design we have chosen R L = 50 n , as shown in Figure
10-16. The output power for this oscillator at the fundamental frequency is Pout = 16 dBm, which corresponds to sinusoidal oscillations with 2 V amplitude. For our design the load resistance is very
close to -Rout so that the frequency shift is insignificant.
Although the component values assure that this oscillator

project meets design specifications and the electric behavior is successfully modeled, the final circuit implementation will pose additional problems. This is apparent when considering, for instance,
the 0.48 nH inductor; which is comparable with the inductance of
560
PCB through hole connections (vias) and parasitics of the individual components.
For high-frequency applications, a more realistic design requires the use of distributed elements. A typical oscillator example involving a FET with connection to a 50
Q load is seen in Figure 10-17. Here, T L; (i 1, ... , 6) represent microstrip lines.
V.ss
0.65 V
RFC
TL3A
TL6
50Q
Figure 10-17 GaAs FET oscillator implementation with microstrip lines.
The design approach is presented in the following example, which provides

details of how to increase the instability through a microstrip line attached to the common gate and how to select appropriate microstrip lines to match the load impedance.
----------------------------~&JA~
Example 10-5: Microstrip design of a GaAs FET oscillator
The S-parameters of the GaAs FET (Hewlett-Packard ATF13100) in
common-gate configuration are measured at 10 GHz and have the
following values:
S 11 = 0.37 L-176, S 12 = 0.17 L19.8,
S21 = 1.37 L-20.7, and S22 = 0.90L-25.6. Design an oscillator with 10 GHz fundamental frequency. Furthermore, match the
oscillator to a 50 Q load impedance.
561
Solution:
Similar to Example 10-4, we first check the stability of
the transistor by computing the Rollett stability factor:
k
= (1-!Sid2-IS2212+1.11 2)/ (2ISI211S2tl) = 0.776
Even though k < 1 indicates that the transistor is potentially unstable,

we can attempt to increase the instability by connecting a feedback
inductor to the gate of the transistor. Following the same approach as
discussed in Example 10-3, we plot the dependence of the stability
factor as a function of inductance (see Figure 10-18).
0.8
. - --
r -- - - , - -,-------,. - - . - -- -,--
----.---
0.6
..lor!
0.4
1-;
8(.)
0.2
-0.2
:-3
.J::J
:tl
~ -0.4
0
~ -0.6
-0.8
-1
0.4
0.8
1.2
1.6
2.0
Feedback inductance L, nH
Figure 10-18
Stability factor for FET in common-gate mode as a function of gate

inductance.
It is seen that maximum instability is achieved for L = 0.9 nH.

Due to the high operating frequency of the oscillator, the use of
lumped elements is undesirable and we have to replace the inductor
by its distributed equivalent. One of the ways to realize an inductance
is to replace it with a short-circuit transmission line stub. Referring
back to Chapter 2, we can calculate the electrical length of the transmission line assuming 50 n characteristic line impedance:
(roL) = 48.5
A
E> = tJI
= tan-1 Zo
562
The resulting S-parameters for the FET with a short-circuited stub

connected to the gate contact are as follows:
[S]
= [1.01 L l69 0.29Ll48l

2.04L- 33 1.36L-34oj
The next step in the design procedure is the development of an

input matching network. As mentioned in Example 10-4, for a realizable oscillator we should choose a source reflection coefficient
close to the inverse of the S 11 parameter of the transistor. In our
design we have selected r 5 = IL-160, which corresponds to a
source impedance of Z 5 = - j8.8 Q and which can be realized as
an open-circuit stub with a 50 n characteristic impedance and 80
electrical length.
The output reflection coefficient is computed as
rout= Szz+ l
s 12s 21
S r rs = 4.18L26.7
11
which is equivalent to Zout = (- 74.8 + j 17.1)Q . To satisfy

(10.26c), we would have to choose a load impedance of
ZL = - Zout' but due to the power dependence of the transistor's Sparameters (see Example 10-4) we choose the real portion of the
load impedance to be slightly smaller than -Rout:
ZL
= (70- }17.1)!1
The transformation of the 50 Q load impedance to Z L is done

through a matching network consisting of a 50 n transmission line
with an electrical length of 67 and a short-circuit stub of 66
length.
The conversion of the electric parameters of the transmission
lines into physical dimensions is done using the same approach as
described in Example 2-5 in Chapter 2. The dimensions of the lines
computed for a FR-4 substrate of 40 mil thickness are summarized
in Table 10-2.
Based on the oscillator circuit diagram shown in Figure 10-17,
the TL3 line is cut into two halves, TL3A and TL3B, to accommodate the blocking capacitor. The lines TLS and TL6 can have any
length since they are connected to a 50 n load.
563
High-Frequency Osclllator Configuration
Table 102
Dimensions of the transmission lines in the FET oscillator
Transmission line
Electrical length, deg.
Width, mil
Length, mil
TLI
80
74
141
TL2
48.5
74
86
TL3
67
74
118
TL4
66
74
116
The microstrip line design allows for an extremely small

circuit board implementation as seen by the individual line lengths.
10.2.2 Dielectric Resonator Oscillators
When dealing with mircostrip line realizations, a dielectric resonator (DR) can
5
be added to provide a very high-Q oscillator design (up to 10 ) with extraordinary temperature stability of better than 10 ppm/C. This resonator, simply called a puck, can
either be placed on top or next to the microstrip line in a metallic enclosure. The electric
field coupling between the strip line and the cylindrical resonator (see Figure I0-19),
can be modeled near resonance as a paraHel RLC circuit. The tuning screw pennits a
geometric adjustment which translates into a change of the resonance frequency.
Field coupling reg
Figure 10-19 Dielectric resonator (DR) placed in proximity to a microstrip line.
We will not investigate the various waveguide modes (TE and TM modes) that are
established inside the resonator, but rather concentrate on the use of DRs under TEM
conditions.
564
In general, the circuit model of a DR (see Figure 10-20) is specified in terms of

the unloaded Q, or Qu ,
Q
= - = ro 0 RC
ro0 L
(10.32)
and the coupling coefficient, ~ ,
~ =-
Rext
R
= 2Zo
=
(10.33)
at the desired angular resonance frequency ro0 = 1I ( Ji;C) . The value of the external
resistance Rext is equal to twice the line impedance because of the symmetric termination into Z 0 . Similar to a transformer, the coupling coefficient quantifies the electromagnetic linkage between the resonator and the microstrip line, with typical values in
the range of 2 to 20. Additionally, p is also employed to describe the relationship
between the unloaded ( Qu ), loaded ( QL ), and external ( QE) quality factors:
(10.34)
For the oscillator design it is required to specify the DR behavior in terms of the Sparameters. The modified transmission line configuration is illustrated in Figure
10-20(b).
500
d~~
ez
Z0 = 50 .Q
Zo
son
(a) Terminated microstrip line with DR

Figure 10-20
Zo
(b) Transmission line model
Placement of DR along a transmission line and equivalent circuit

representation for S-parameter computation.
Recalling our discussion of parallel resonance circuits in Section 5.1.4, we can

compute the impedance ZvR as
=
DR
1 + jroRC- jRI(roL) - 1 + JQu(ro/ro0 ) - JQu(ro0 /ro)
which simplifies to
(10.35)
565
R
------- - -R- - 2 2J 1 + j2Qullf I J0
1 + iQu ro - Wo
( CO COo
~
(10.36)
where llf =f- fo is the deviation from the center frequency. The last equation is only
valid around the resonance point, where ro + ro0 ~ 2ro0 . Normalized with respect to Z 0
near resonance, it is seen that
RI Z 0
zvR ~ 1 + j2Qu(llf If o)
= 2~
(10.37)
The transmission line segments on either side can now be included, leading to
(10.38)
Depending on the direction, we can determine the reflection coefficient as either s~t
or
2R . If the electric line length is equal on both sides of the DR, we obtain
eI = e2 = e = (21ti A)(ll2) ' and therefore
sf
fin ( COo)
-j29
= ~ + 1e
= f
out ( COo)
(10.39)
The selection and purchase of a DR can be carried out quickly and efficiently,
often over the manufacturers' websites. The design engineer specifies a particular resonance frequency and board material (thickness, dielectric constants) and the manufacturer will provide a particular DR in terms of diameter, length, tuning screw extension,
distance d from the microstrip line, and cavity material. In addition, the coupling
parameter and the unloaded Q are given as well as the lumped parallel resonant circuit
elements needed in the CAD simulation programs.
------------------------Rf&MW4
Example 10..6: Dielectric resonator oscillator design
Design an 8 GHz dielectric resonator oscillator (ORO) using a GaAs
FET whose S-parameters at / 0 = 8 GHz are S 11 = 1.1L170 ,
566
S 12 = 0.4L-98, S 21 = 1.5L-163, and S22 = 0.9L-170. A

dielectric resonator that is used in the design has the following
parameters at resonant frequency f res = f 0 : ~ = 7 , Qu = 5000 .
Find the length of the 50 0 microstrip line at the input port side of
the FET, if the DR is located in the middle. Assume the DR is terminated with a 50 n resistor. Examine the difference in the DRO
response to the frequency fluctuations as compared to the conventional designs discussed previously.
Solution:
The input stability circle of the FET at fo = 8 GHz is
shown in Figure I0-21.
+1.0
-1.0
Figure 10-21
Input stability circle of the FET in the ORO design example.
To satisfy the oscillation conditions we have to chose a source

reflection coefficient somewhere in the non-shaded area of Figure
10-21. Since the termination resistance for the dielectric resonator is
equal to the characteristic line impedance, the output reflection coefficient of the DR is computed according to ( 10.39):
fs =
~~
e-j
1
20
= 0.875e-j 20
567
As discussed in the previous examples, to maximize the output

reflection of the transistor, we have to chose r s close to the inverse
of the S 11 parameter. Since the absolute value of r s is fixed, the
best we can do is to select
such that the phase angle of s is
equal to the phase angle of S}; , or -28 = LS!~ = -LS 11 , leading
to e = 85 o The resulting electric circuit for the input matching
network of the oscillator is shown in Figure 10-22.
f......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . '"i
:
jS
c
1
Dielectric resonator
...... .......... ...... .... .. ...... ......... ................................ .. ............................... ~ .. ........ .J
Figure 10-22
OR-based input matching network of the FET oscillator.
If the DR is not used in the input matching network of the transistor, then the simplest network that yields the same
rs = 0.875L-170 at the oscillation frequency fo would be a
series combination of a 3.35 Q resistor and a 4.57 pF capacitor. A
comparison of Ir outl for the DR versus no DR realization as a function of frequency is shown in Figure 10-23, where the FET S-parameters are assumed to be frequency independent and the DR is
approximated by its equivalent circuit shown in Figure 10-20 with
parameters computed using (10.32) and (10.33) :
R = 2 ~Z 0 = 700
L == Rl(ro0 QJ
C
= 2.79
= ro-02 L- 1 = 14.2
pH
nF
1r
As clearly seen in Figure 10-23, the DRO design has a outl > 1 in
a much narrower frequency band than the conventional oscillator
without the DR. This approach generally results in high selectivity
and reduced drifts of the oscillation frequency. With the tuning
screw small frequency adjustments can be done, typically in the
range +O. 01 f 0 around the target frequency.
568
12
~~
.... 10
()
IS
~
8 8
'{~
g.
6
-
1'--------~)
QL-~--~~---=~~~--~--~~~
7.98 7.985 7.99 7.995
8.005 8.01 8.015 8.02 8.025
Frequency f, GHz
(a) Oscillator design with DR

:;13
~ 12
~
II
8
Cl
0
'J:)
10
9
g. 7
6
5L-~--~~--~--~~--~--~~~
5.5
6.5
7.5
8.5
9.5
10
10.5
Frequencyf, GHz
(b) Conventional oscillator without DR

Figure 1023 Frequency response of the output reflection coefficient for an
oscillator design with and without DR.
The dielectric resonator is an inexpensive and easy way to

improve the quality factor of an oscillator. Unfortunately. its geometric size depends on the resonance frequency and typically gets
too large at low frequencies.
569
10.2.3 YIG-Tuned Oscillator
The dielectric resonator allows tuning over a very narrow band around the resonance frequency, typically between 0.01 and 1%. As an alternative, a magnetic element
offers a wideband tunable oscillator design with a tuning range of more than a decade
of bandwidth. Such a tunable element, often of spherical shape, derives its name from
yittrium iron garnet (YIG), a ferrimagnetic material whose effective permeability can
be externally controlled through a static magnetic bias field H 0 . This applied field
directly influences the Q of the equivalent parallel resonant circuit consisting of conductance G 0 , inductance L0 , and capacitance C 0 . Figure 10-24 depicts a typical YIG
element oscillator circuit.
Matching
network
Figure 10-24 Oscillator design based on a YIG tuning element.
(10.40)
(10.41)
where y is the gyromagnetic ratio recorded in 2.8 MH:zJOe. The resonance frequency
follows from the bias field:
(10.42)
From these equations the circuit elements of the parallel resonance circuit can be quantified. Specifically, the inductance is found to be
Lo -_ J.!oO>m(~
2 31ta
ro0d
3)
(10.42a)
570
with a being the radius of the YIG sphere. This also determines C 0 from the resonance
condition ro~ = 1/(L0 C0 ); that is,
(10.42b)
Finally, the conductance is
(10.42c)
In ( 10.42a)-( 10.42c) d is the diameter of the coupling loop.
10.2.4 Voltage-Controlled Oscillator

It is mentioned in Chapter 6 that certain diodes exhibit a large change in capacitance in response to an applied bias voltage. A typical example is the varactor diode,
12
with its variable capacitance C v = C vo< 1 - V Q/ V diff) -t that can be affected by the
reverse bias V Q . Figure 10-25 illustrates how the feedback loop for the Clapp oscillator
can be modified, by replacing C 3 in Figure 10-25(a) with the varactor diode and an
appropriate DC isolation. The modified circuit is shown in Figure I0-25(b). This circuit
can readily be analyzed if a simplified BJT model ( RL h 22 ) is employed.
,. ....
.. .................. -- .. -.- ... .

~
--~~----~
RFC
(a) Pi-type feedback loop
(b) Redrawn circuit with DC isolation
Figure 1025 Varactor diode oscillator.

In Figure 10-26 the varactor diode and a transmission line element, whose length
is adjusted to be inductive, form the termination circuit connected to the input of the
oscillator. If the varactor diode and the transmission line segment is disconnected, the
input impedance ZIN can be computed from two loop equations:
571
viN- iiNXCl-
iiNXc2 + iBXCl - ~iBXC2 = 0
(10.43a)
(10.43b)
hu iB + isXct - irNXcl = 0
Zr
L3
Figure 10-26
h"~
c,
CJ
RL
~is
f'-r
RL
Z IN
Circuit analysis of varactor diode oscillator.
Rearranging leads to
ZIN = h
1
II+
[hll (Xct + Xc2) + Xct XczO + Jl)]
(10.44)
Cl
The equation can be simplified by noting that ( 1 + Jl) :::: ~ and assuming that
h 11 X c 1 , which results in
ziN = 1~[~~ + ~J- ~Ll;1 cJ
(l0.4SJ
As expected from our previous discussion, the input resistance is negative. Therefore,
with gm = ~lh 11 ,
RIN
(10.46a)
and
1
XIN = jroC
(10.46b)
IN
where C1N = C 1C2 / ( C 1 + C 2 ). The resonance frequency follows from the previously
established condition X 1 + X 2 + X 3 = 0 (see Section 10.1.2), or
1(ro 0L 3 -
with the result
1 ) - -.1- [ - 1 + -1
JWo CI Cz
rooC3
J= 0
(10.47)
572
Chapter 10 Oscillators and Mixer
f0
..!.(_!_
+ _!_)
c + _!_
C
C
= 21t L 3
(10.48)
It can be concluded from (10.46a) that the combined resistance of the varactor diode
must be equal to or less than jRINI in order to create sustained oscillations.
----------------------------RF&JA~
Example 10-7: Design of a varactor-controlled osciUator
A typical varactor diode has an equivalent series resistance of 45 n
and a capacitance ranging from 10 pF to 30 pF for reverse voltages
between 30 V and 2 V. Design a voltage controlled Clapp-type oscillator with center frequency of 300 MHz and 10% tuning capability. Assume that the transconductance of the transistor is constant
and equal to gm = 115 mS.
Solution:
To create sustained oscillations, we have to ensure that
the series resistance of the varactor diode is smaller or equal to
1NI over the entire frequency range as computed in (10.46a). From
achieves its minimum value at
(10.46a) we can conclude that
maximum frequency of operation. Substituting romax = 21tjmax
(with f max = 1.1 f 0 = 330 MHz being the maximum oscillation
frequency) into (10.46a), it is found that the capacitances C 1 and
C2 are related as
IR
IRrNI
c,
romaxRsC2
1
= ----kCz
21
1.68xl0 C 2
(10.49)
where Rs = 45 n is the varactor's equivalent series resistance.

Since the maximum oscillation is obtained when the varactor
capacitance has its minimum value, and the minimum frequency
corresponds to the maximum C 3 , we can rewrite ( 10.48) as
1
f min -- 27t
fmax
_!_(
L 3 C 3 max
+ __!_ +
C2
kc2)
(10.50)
(10.51)
573
where relation (10.49) is used to eliminate C 1 Dividing (10.50) by

(1 0. 51) and taking the square of the result, the following quadratic
equation is obtained for c 2:
k( 1 -
a2 )C~ + (c 1
3max
ca
2
_
3mm
)c + ( 1- a
2
(10.52)
where a = f min/ f max. Solving (10.52) and substituting the result

in (10.49) and (10.50) or (10.51), we find C 1 = 12.4 pF,
C2 = 48 pF, and L 3 = 46.9 nH as our desired values.
Unlike a mechanically adjustable dielectric resonator, the varactor diode permits dynamic tuning over a substantial frequency
range.
10.2.5 Gunn Element Oscillator

The Gunn element can be employed to create oscillators from 1 to 100 GHz at
low power outputs of roughly up to 1 W. It exploits a unique negative resistance phenomenon first discovered by Gunn in 1963. When certain semiconductor structures are
subjected to an increasing electric field , they begin to shift, or transfer, electrons from
the main valley to side valleys in the energy band structure. The accumulation of up to
90-95% of the electron concentration into these valleys results in a substantial decrease
in effective canier mobility and produces a technologically interesting /- V characteristic. Semiconductors with these band structures are primarily GaAs and InP. Figure
10-27 depicts a Gunn element and its current versus applied voltage response.
We notice that in the presence of an applied DC voltage to the Gunn element it
behaves like a normal ohmic contact resistor for low field strength. However, if a certain
threshold voltage V 0 is exceeded, dipole domains begin to be created below the cathode
triggered by doping fluctuations. The formation of these domains lowers the current, as
indicated in Figure 10-27 (b). The current then remains constant while the domains
travel from cathode to anode. After collection, the process repeats itself. The frequency
5
can be estimated from the drift velocity of the domain motion vd = 10 m/s and the
travel length L of the active zone of the Gunn element. For a length of 10 J.!m, we obtain
- vd-
1-L
10 m/s
-6
10 x 10
= 10 GHz
(10.53)
Chapter 10 Oscillators and Mlxera
574
max
........
--active layer
(10 ... 20 fim)
Anode
-4----------~~----------.v
Vo
(a) Gunn element structure
(b) Current vs. applied voltage response
Figure 1D-27 Gunn element and current versus voJtage response.
If an external DC voltage is applied, the domain motion can be influenced and thus the
resonance frequency is varied. The tuning range is approximately within 1% of the res
onance frequency.
Figure 10-28 shows a microstrip line implementation of a Gunn element oscillator. Here the Gunn element is connected to a 'A./ 4 microstrip line, which in tum is coupled to a dielectric resonator. The bias voltage for the Gunn element is fed through an
RFC onto the microstrip line.
Vso--+-......,
bias
voltage
Gwm
element
Figure 1028 Gunn element oscillator circuit with dielectric resonator (DR).
10.3 Basic Characteristics of Mixers
Mixers are commonly used to multiply signals of different frequencies in an effort

to achieve frequency translation. The motivation for this translation stems from the fact
that filtering out a particular RF signal channel centered among many densely populated, narrowly spaced neighboring channels would require extremely high Q filters.
Basic Characteristics of Mixers
575
The task, however, becomes much more manageable if the RF signal carrier frequency
can be reduced or downc.onverted within the communication system. Perhaps one of the
best known systems is the downconversion in a heterodyne receiver, schematically
.,. .....
''
Mixer
--.-- . . ---.- .. -..... --- ............ - ... --.... -.... -.. -..-.. --- ........................... ----------------
LNA
Combiner
Detector
'
'
,....----LP Filter
fw
Figure 10-29
Heterodyne receiver system incorporating a mixer.
Here the received RF signal is, after preamplification in a low-noise amplifier

(LNA), supplied to a mixer whose task is to multiply the input signal of center frequency f RF with a local oscillator (LO) frequency fL O . The signal obtained after the
mixer contains the frequencies f RF ILO, of which, after low-pass (LP) filtering, the
lower frequency component I RF- fLO, known as the intermediate frequency (IF), is
selected for further processing.
The two key ingredients constituting a mixer are the combiner and detector. The
combiner can be implemented through the use of a 90 (or 180) directional coupler.
A discussion of couplers and hybrids is found in Appendix G. The detector traditionally
employs a single diode as a nonlinear device. However, anti parallel dual diode and double-balanced quadrupole diode configurations are also utilized, as discussed later. In
addition to diodes, BJT and MESFET mixers with low noise figure and high conversion
gain, have been designed up to the X-band.
10.3.1 Basic Concepts
Before going into details of the circuit design, let us briefly review how a mixer is
capable of taking two frequencies at its input and producing multiple frequency components at the output. Clearly a linear system cannot achieve such a task, and we need to
select a nonlinear device such as a diode, FET, or BJT that can generate multiple harmonics. Figure 10-30 depicts the basic system arrangement of a mixer connected to an
RF signal, V RF(t), and local oscillator signal, V Lo(t) , which is also known as the
pump signal.
576
Figure 1030
Basic mixer concept: two input frequencies are used to create new
frequencies at the output of the system.
It is seen that the RF input voltage signal is combined with the LO signal and supplied to a semiconductor device with a nonlinear transfer characteristic at its output side
driving a current into the load. Both diode and BJT have an exponential transfer characteristic, as expressed for instance by the Shockley diode equation discussed in Chapter 6:
I = I 0 (e
VIV T
(10.54a)
- 1)
Alternatively, for a MESFET we have approximately a square behavior:

I(V) = I ossO - V I V ro)
(10.54b)
where the subscripts denoting drain current and gate-source voltage are omitted for
simplicity. The input voltage is represented as the sum of the RF signal
V RF = V RFCOS ( 00R pf) and the LO signal v LO = V LOCOS ( W LOt) and a bias V Q; that
IS,
(10.55)
This voltage is applied to the nonlinear device whose current output characteristic can
be found via a Taylor series expansion around the Q-point:
d/) v +2:V
I 2(d
l(V) = IQ+V (dV
dV J
2
/
2
+ ... = IQ+VA+V 2 B+ ...
(10.56)
VQ
where the constants A and B refer to (dlldV)Iv and (l/2)(d 2 //dV 2 )Iv , respecsubstitution of (10.55) in~o (10.56)
tively. Neglecting the constant bias VQ and IQ,
yields
thi
/(V) = A{ V RFcos(O>Rpt) + V LOcos(roL 0 t)}

2
+ B { V RFCOS
( (J)RFt)
+ V LOCOS
( O>Lot)}
+ 2BVRF V LOcos ( OORFt) COS (())Lot)+
...
(10.57)
sn
The factors containing the cosine square terms can be rewritten, via the trigono2
metric identity cos ( rot) = (1/2){ 1 -cos (2rot)}, into DC terms and terms involving
2roRFt and 2roL 0 t. The key lies in the last term of (10.57), which becomes
!(V)
= ... +BVRFVL0 {cos[(roRF+roL0 )t] + cos[(roRF-O>Lo)t]}
(10.58)
This expression makes clear that the nonlinear action of a diode or transistor can generate new frequency components of the form roRF roLo. It is also noted that the amplitudes are multiplied by V RF V LO , and B is a device-dependent factor.
Equation (10.58) is the Taylor series representation up to the third term, and thus
2
up to second-order intermodular product ( V B ). Any higher-order products, such as
3
third-order intermodular product ( V C), are neglected. For diodes and BJTs these
higher-order harmonic terms can significantly affect the performance of a mixer. However, the second-order intermodular product is the only surviving term if a FET with
quadratic transfer characteristic is utilized. Thus, a FET is less prone to generate undesired higher-order intermodular products.
The following example discusses the down conversion process from a given RF
signal frequency to a desired intermediate frequency.
-------------RF&MW-+
Example 10-8: Local oscillator frequency selection
An RF channel with a center frequency of 1.89 GHz and bandwidth
of 20 MHz is to be downconverted to an IF of 200 MHz. Select an
appropriate fLo Find the quality factor Q of a bandpass filter to
select this channel if no downconversion is involved, and determine
the Q of the bandpass filter after downconversion.
Solution:
As seen in (10.58), by mixing RF and LO frequencies
through a nonlinear device we produce an IF frequency that is equal
to either fiF = f RF- fLO or /IF = j LO- j RF, depending on
whether fRF or fLo is higher. Thus, to produce a f IF = 200 MHz
from f RF = 1.89 GHz we can use either
fLo= !RF- ! 1F
= 1.69 GHzor fLo=
/RF+ !IF= 2.09 GHz
These two choices are equally valid and are both used in practice.
When f RF >fLO is chosen, the mixer is said to have low-side
Chapter 10 Oscillators and Mlxen
578
injection, whereas when f RF < f LO the design is called high-side

injection. The first approach is generally preferred since lower LO
frequencies are easier to generate and process.
Before down conversion, the signal has a bandwidth of
B W = 20 MHz at a center frequency of f R F = 1.89 GHz. Therefore, if we attempted to filter out the desired signal we would have to
use a filter with Q = f RF/ BW = 94.5. However, after downconversion, the bandwidth of the signal does not change but the center
frequency shifts to f 1F = 200 MHz, thus requiring a bandpass filter with a quality factor of only Q = f 1FI BW = 10.
This example shows that less selective filtering
once the mixer has downconverted the RF signal.
lS
required
10.3.2 Frequency Domain Considerations
It is important to place the previous section into a frequency domain perspective.

To this end it is assumed that the angular RF signal is centered at roRF with two extra
frequency components situated row above and below roRF. The LO signal contains one
single component at roLo. After performing mixing, according to (10.58), the resulting
spectral representation contains both upconverted and downconverted frequency
components. Figure 10-31 graphically explains this process.
Typically the upconversion process is associated with the modulation in a trans
mitter, whereas the downconversion is encountered in a receiver. Specifically, when
dealing with modulation, the following tenninology is common:
Lower sideband, or LSB ( roRF - roLo)
Upper sideband, or USB ( roRF + roLO)
Double sideband, or DSB ( roRF + roL0 , roRP - roLo)
A critical question to answer is the choice of an LO frequency that shifts the RF frequency to a suitable IF level.
An interrelated issue is the problem of image frequencies mapping into the same
downconverted frequency range. To understand this problem, assume an RF signal is
downconverted with a given LO frequency. In addition to the desired signal, we have
579
~F (ffi) I
~------------------~~~----------------~~00
_,./'!
(ORF
OORF - ffiw
OOw + ffiw
(a) RF signal
~----------------~----------------------~(0
ffiw
(b) LO signal
V:.ut (ro)
1\
(I)
ffiruo - OOw
OORF + C.Ow
(c) Down and upconverted spectral products
Figure 10-31
Spectral representation of mixing process.
placed symmetrically an interferer about IF (see Figure 10-32). The desired RF signal
transforms as expected:
00RF - 00 LO
= 00IF
(10.59a)
However, the image frequency roiM transforms as

(J)IM - (J)LO
= (<OLO -
(l)IF) - (J)LO
= -rolF
(10.59b)
580
Chapter 10 Oscillator and Mlxera
Since cos (-ro1pt) = cos ( ro1Ft) , we see that both frequency spectra are shifted to the
same frequency location, as Figure 10-32 illustrates.
RF
Undesired
image signal
co
If
LO
Figure 10-32
Problem of image frequency mapping.
To avoid the presence of undesired image signals that can be greater in magnitude
than the RF signal, a so-called image filter is placed before the mixer circuit to suppress this influence, provided sufficient spectral separation is assured. More sophisticated measures involve an image rejection mixer.
10.3.3 Single-Ended Mixer Design
The simplest and least efficient mixer is the single-ended design involving a
Schottky diode, as shown in Figure 10-33(a). The RF and LO sources are supplied to an
appropriately biased diode followed by a resonator circuit tuned to the desired IF. In
contrast, Figure 10-33(b) shows an improved design involving a PET, which, unlike the
diode, is able to provide a gain to the incoming RF and LO signals.
In both cases the combined RF and LO signal is subjected to a nonlinear device
with exponential (diode) or nearly quadratic (FET) transfer characteristic followed by a
bandpass filter whose task is to isolate the IF signaL The two very different mixer realizations allow us to contrast a number of parameters important when developing suitable designs:
Conversion loss or gain between the RF and IF signal powers
Noise figure
Isolation between LO and RF signal ports
Nonlinearity
581
(a) Diode mixer
_j_
(b) FET mixer
Figure 1033
Two single-ended mixer types.
Since LO and RF signals are not electrically separated in Figure 10-33(a), there is the
potential danger that the LO signal can interfere with the RF reception, possibly even
reradiating portions of the LO energy through the receiving antenna. The FET realization
in Figure 10-33(b) allows not only for LO and RF isolation but also provides signal gain
and thus minimizes conversion loss. The conversion loss (CL) of a mixer is generally
defined in dB as the ratio of supplied input power PRF over the obtained IF power PIF :
CL =
JO!og(::)
(10.60)
When dealing with BJTs and PETs, it is preferable to specify a conversion gain (CG)
defined as the inverse of the power ratio.
Additionally, the noise figure of a mixer is generically defined as
F
p
noul
CGPn.on
(10.61)
582
with CG again being the conversion gain, and P n nut , P n-m the noise power at the output
due to the RF signal input (at RF) and the total noise power at the output (at IF). The
FET generally has a lower noise figure than a BJT, and because of a nearly quadratic
transfer characteristic (see Section 7.2) the influence of higher-order nonlinear terms is
minimized. Instead of the FET design, a BIT finds application when high conversion
gain and low voltage bias conditions are needed (for instance, for systems relying on
battery operation).
Nonlinearities are customarily quantified in terms of conversion compression
and intermodular distortion (IMD). Conversion compression relates to the fact that
the IF output power as a function of RF input power begins to deviate from the linear
curve at a certain point. The point where the deviation reaches 1 dB is a typical mixer
performance specification. As already encountered in the amplifier discussion, the
intennodulation distortion is related to the influence of a second frequency component
in the RF input signal, giving rise to distortion. To quantify this influence, a two-tone
test is typically employed. If f RF is the desired signal and f 2 is a second input frequency, then the mixing process produces a frequency component at 2/ 2 - f RF fLO,
where the +1- sign denotes up- or downconversion. The influence of this intermodulation product can be plotted in the same graph as the conversion compression (see Figure 10-34).
;:~, (=~). .. . .~.~~~~'. :~?--
Thlrd order intercept point
l dB~----<___j
/f_ __
p l dB .... ...... ...... .... ..... .... ......... ,.-.~ '
,'
:
'
Small-signal
gain in dB
_,/
'
Real P.,., vs.
P.,
curve
1 dB compression point
~ Third harmonic
I'
''
~----------------~p----------+~ n(dBm)
OdBm
Figure 10-34
f/PJ
Conversion compression and intermodulation product of a mixer.
The intercept point between the desired linear output response and the undesirable
third-order IMD response is a common figure of merit, indicating the ability of a mixer
to suppress this influence.
583
Additional mixer definitions include distortion generated inside the mixer which
is defined as harmonic IMD; isolation between RF and IF ports, which is directly
linked to the influence of the combiner (hybrid coupler; see Appendix G); and dynamic
range~ which specifies the amplitude range over which no performance degradation
occurs.
The circuit design of an RF mixer follows a similar approach as discussed when
dealing with an RF amplifier. The RF and LO signals are supplied to the input of an
appropriately biased transistor or diode. The matching techniques of the input and output side are presented in Chapter 8 and directly apply for mixers as well. However, one
has to pay special attention to the fact that there is a large difference in frequencies
between RF, LO on the input side, and IF on the output side. Since both sides have to be
matched to the typical 50 .Q line impedance, the transistor port impedances (or Sparameter representation) at these two different frequencies have to be specified. Furthermore, to minimize interference at the output side of the device, it is important to
short circuit the input to IF, and conversely short circuit the output to RF (see Figure
10-35). Including these requirements as part of the matching networks is not always an
easy task.
Input
matching
network
fRI'
Short circuit
for IF
Output
matching
network
Active
device
j;F
~ort
circuit
forRF
DC biasing
network
Figure 1035
General single-ended mixer design approach.
These short-circuit conditions in general affect the transistor's behavior through

internal feedback mechanism. Ideally, r 10 (roRF) should be known based on the shortcircuit output condition and similarly rout( ro1F) requires a short-circuit input condition. Typically, an additional load resistance is added to the output port to adjust the
conversion gain. In the following example, the salient design steps are explained.
584
Chapter 1 0 Oscillators and Mixers
----------------------------RF&uM~
Example 10-9: Design of a single-ended BJT mixer
For the DC-biasing topology shown in Figure 10-36, compute the
values of the resistors R 1 and R2 such that biasing conditions are
satisfied. Using this network as a starting point, design a low-side
injection mixer for fRF == 1900 MHz and f 1F = 200 MHz. The
BJT is measured at IF to have an output impedance of
zout ;:;::: (677.7- )2324 ).Q for short-circuit input, and an input
impedance of Zin = (77.9- j130.6).Q for short-circuit output at
RF frequency. Attempt to minimize the component count in this
design.
V"cr:
= 3V
VBH
lc
18
RFC,
0.89 v
2.2 rnA
=30~
Zout (frF) = (677.7 - j2324)Q
Zin(fRF) =(77.9- j130.6)Q
Figure 10-36 DC-biasing network for BJT mixer design.
Solution:
Since the voltage drop across resistor R2 is equal to
the difference between V cc and V CE and the current is the sum of
the base and collector currents, R 2 is computed as
R2
= v cc- v cE
lc+ IB
= 448
585
Similarly, the base resistor R 1 is computed as a ratio of V CE - V BE

over the base current:
Vc -V
E/
= 70.3 k.Q
Before beginning the design of an input matching network, we

have to decide on how to supply the LO signal. The simplest
arrangement is to connect the LO source directly to the base of the
transistor via a decoupling capacitor, as shown in Figure 10-37.
R2 =448 n
Zru: =50 .Q
l
Figure 10-37 Connection of RF and LO sources to the BJT.
The value of this capacitor C LO has to be chosen small enough

so as to prevent RF signal coupling into the LO source. We arbitrary
pick CLo = 0.2 pF. In this case the series combination of CLo and
ZLo creates a return loss RLRF of only 0.24 dB , since
RLRF
= -20logjrLol
= -20log(0.9727)
= 0.24 dB
fRF
Unfortunately, the LO frequency is very close to f RF so that the

same capacitance will attenuate not only the RF signal but the LO as
well. We can compute the insertion loss I LRF due to this capacitor
at fLo = fRF- fiF
Chapter 10 Oaclllatora and Mlxera
586
Thus, if the LO source pumps at -20 dBm, only - 33.6 dBm reaches
the transistor. This seemingly high power loss is still tolerable since
we can adjust the power provided by the local oscillator.
The presence of CLo and ZLo modifies the value of the input
impedance. A new total input impedance Z~n can be computed as a
parallel combination of CLo and ZLo, and the input impedance of
the transistor connected to the LO source is
z;n = (zLO + .
]~
c ) II zin = (47.2 LO
jl03.5)Q
The output impedance does not change since the input is shorted
during the measurement of Z out.
Knowing z;n, we can next design an input matching network
using any of the methods described in Chapter 8. One of the possible topologies consists of a shunt inductor followed by a series
capacitor, as shown in Figure 10-38, where we added the blocking
capacitor C 81 to prevent DC short circuit to ground.
Input matching
network
RFC1
\,
:......... ... ..... .... :
: cl
! .. ~
i
Ll
ICBI!
.
:............................~............. :
Figure 1o-38
Input matching network for a single-ended BJT mixer.
587
There are several modifications that can be made to the circuit

in Figure 10-38. First we notice that instead of biasing the base of
the transistor through an RFC, we can connect R 1 directly to the
contact between L 1 and C 81 . In this case we still bias the base of
the transistor through L 1 and maintain isolation of the RF signal
from the DC supply by grounding the RF through C B 1 . One more
task of this matching network is to provide a short-circuit condition
for the IF signal. Even though the impedance of the inductor L 1 is
rather small at f lF, we still can lower it by choosing the value of
C B 1 such that L 1 and C B 1 exhibit a series resonance at IF. For
example, if we choose C 8 1 = 120 pF, we still maintain a solid
short circuit for the RF signal and we improve the path to ground for
the f IF signal. The modified input matching network is shown in
Figure 10-39.
ro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .io
CB l
!_ci~
: -
! -
Input matching
network
hF ~
;
:........................ ;
Cw
Figure 1039
Modified input matching network.
The output matching network is developed using a similar

approach. The original matching network again consisted of a shunt
inductance L 2 followed by a series capacitance C 2 The values are
L2 = 416 nH and C 2 = 1.21 pF. This topology allows us to eliminate the RFC at the collector terminal of the transistor. However, the
problem with this topology is that it does not provide a short circuit
588
to ground for the RF signal that may interfere with the output. To
remedy this drawback we replace L 2 with an equivalent LC combination where the additional capacitance C 3 = 120 pF is chosen to
provide solid ground condition for the f RF signal and L 2 is
adjusted to L 2 = 5.2 nH. The complete circuit of the designed single-ended BJT mixer is shown in Figure 10-40.
fc~~------1
Output matching
~ i/
network
/ :.........!
Input matching
network
l
~
Ct
+-J
./;F "TT
;.
:
Tc,,
----!
........................ ....... J
J;_()
~ J;,
Ic,
!
.
: -=- ~ kr: blockmg

capacitor
Figure 10-40 Complete electrical circuit of the low-side injection, single-ended

BJT mixer with fRF = 1900 MHz and "F = 200 MHZ.
This design shows the multiple purposes that a matching network can perform. At first glance they are often difficult to understand. Specifically, the dual network purposes of matching and
isolation provide challenges for the circuit designer.
10.3.4 Single-Balanced Mixer
From the previous section it is seen that the single-ended mixers are rather easy to
construct circuits. The main disadvantage of these designs is the difficulty associated
with providing LO energy while maintaining separation between LO, RF, and IF signals for broadband applications. The balanced dual-diode or dual-transistor mixer in
conjunction with a hybrid coupler offers the ability to conduct such broadband operations. Moreover, it provides further advantages related to noise suppression and spurious mode rejection. Spurs arise in oscillators and amplifiers due to parasitic resonances
589
and nonlinearities and are only partially suppressed by the front end. Thermal noise can
critically raise the noise floor in the receiver. Figure 10-41 shows the basic mixer design
featuring a quadrature coupler and a dual-diode detector followed by a capacitor acting
as summation point.
90 branch line
coupler
Figure 10-41
Balanced mixer involving a hybrid coupler.
Besides an excellent VSWR (see Appendix G), it can be shown that this design is
capable of suppressing a considerable amount of noise because the opposite diode
arrangement in conjunction with the 90 phase shift provides a good degree of noise
cancellation. The proof is left as an exercise, see Problem 10.22.
A more sophisticated design, involving two MESFETs and 90 and 180 hybrid
couplers is shown in Figure 10-42. The 180 phase shift is needed since the second
MESFET cannot easily be reversed as done in the anti-parallel diode configuration seen
in Figure 10-41. It is also important to point out that this circuit exhibits LO to RF as
well as LO to IF signal isolation, but no RF to IF signal isolation. For this reason, a lowpass filter is typically incorporated into the output matching networks of each of the
transistors in Figure 10-42.
Input
matching
network
Input
matching
network
Figure 10-42
Output
matching t - - - - - - - ,
network
Wilkinson
power
combiner
Output
180
matching phase 1 - - -.....
network
shift
fw
Single-balanced MESFET mixer with coupler and power combiner.
590
Chapter 10 Oscillator. and Mixers
10.3.5 Double-Balanced Mixer
The double-balanced mixer can be constructed by using four diodes arranged in a

rectifier configuration. The additional diodes provided better isolation and an improved
suppression of spurious modes. Unlike the single-balanced approach, the double-balanced design eliminates all even harmonics of both the LO and RF signals. However,
the disadvantages are a considerably higher LO drive power and increased conversion
loss. Figure 10-43 depicts a typical circuit of the double-balanced design. All three signal paths are decoupled, and the input and output transformers enable a symmetric mixing with the LO signal.
fw
Figure 10-43 Double-balanced mixer design.
For design details of double-balanced mixers the reader is referred to the books by
Vendelin and Mass listed at the end of this chapter.
10.4 Summary
Oscillators and mixers require a nonlinear transfer characteristic and are therefore
more difficult to design than standard linear amplifiers. It is not uncommon to encounter circuits that perform as desired, but the design engineer does not understand exactly
why they behave this way. Today's extensive reliance on CAD tools has often reduced
our thinking to trial-and-error approaches. This certainly applies both to oscillators and
mixer RF circuits.
One of the key design requirements of an oscillator is the negative resistance condition as a result of the feedback loop equation, which can be formulated as the
Barkhausen criterion:
H F( ro )HA ( ro) = 1
For instance, the feedback Pi-type network results in a host of different oscillator types,
of which we discussed the Hartley, Colpitts, and Clapp designs. At frequencies up to
approximately 250 MHz one of the passive feedback elements can be replaced by a
591
Summary
quartz crystal whose. mechanical vibrations allow substantial improvements in frequency and temperature stability.
For higher frequencies the S-parameters again become the preferred design procedure. For a two-port oscillator, the stability condition and the input and output oscillation conditions assume primary importance:
k < 1,
fin
r s==
1,
routr L =
A typical approach would start with the test of the stability circle k. Next, from a particular output loading condition, the output reflection coefficient is determined from the
knowledge of the input reflection coefficient. Conversely, the design can also be conducted from the input side. To enhance the Q-factor of the high-frequency performance
a dielectric resonator can be added whose behavior is that of a parallel resonance circuit
with normalized line impedance:
R!Z0
ZoR::::
1 + j2Qi~f I f
Instead of the dielectric resonator a magnetically induced resonance condition can be

established with the help of a YIG element. A Gunn diode finds applications in very
high frequency osciUators. To add frequency tuning flexibility, a varactor diode is often
employed to adjust the resonant circuit capacitively.
Besides oscillators, mixers are the second group of practical circuits directly
exploiting the nonlinear transfer characteristic of active solid-state elements such as
diodes and bi- and monopolar devices. The ability of a mixer to achieve frequency translation finds applications in heterodyne receiver and transmitter circuits. An RF signal
roRF mixed with a local oscillator frequency roLo results in a main current product of
/(V) = ... + BVRF VL0 { cos [{~+ roL0 )t] + cos[(roRF- roL0 )t]}
where the first teffil signifies upconversion and the second term downconversion. This
second-term response can, for instance, be utilized as the required intermediate output
signal in a receiver. To isolate the desired signal frequency, extensive filtering is
required on the input (image filter) and output (low-pass) sides of the mixer. Singleended, single-balanced, and double-balanced designs can be constructed by appropriate
impedance matching of the source and load to the active device. One additional complication over the amplifier matching network design arises because of the need to isolate
the RF and LO inputs from the IF output and, conversely, to isolate the IF output from
both RF and LO input signals. While balanced mixers offer improved signal performance through partial cancellation of undesired harmonic responses, they require the
additional complexity of couplers to accomplish the required phase shifts.
592
Chapter 10 Oeclllatora and Mlxera
Further Reading
Y. Anand and W. J. Moroney, "Microwave Mixer and Detector Diodes," Proceedings of

IEEE, Vol. 59, pp. 1182- 1190, 1971.
R. J. Gilmore and F. J. Rosenbaum, "An Analytical Approach to Optimum Oscillator
Design Using S-Parameters," IEEE Trans. on Microwave Theory and Techniques, Vol.
31,pp. 633-639, 1983.
G. Gonzalez, Microwave Transistor Amplifiers, Analysis and Design, Prentice-Hall,
J. B. Gunn, "Effect of Domain and Circuit Properties on Oscillations on GaAs," IBM
Journal of Res. Development, Vol. 10, pp. 310-320, 1966.
J. M. Manley and H. E. Rowe, "Some General Properties of Nonlinear Elements," Proceedings ofIRE, Vol. 44, pp. 904-913, 1956.
S. A. Mass, Microwave Mixers, Artech House, Dedham, MA, 1986
M.A. Smith, K. J. Anderson, and A.M. Pavio, "Decade-Band Mixer Covers 3.5 to 35
GHz," Microwave Journal, pp. 163-171, Feb. 1986.
G. Vendelin, A. Pavio, and U. L. Rhode, Microwave Circuit Design Using Linear and
Nonlinear Techniques, John Wiley, New York, 1990.
G. Vendelin, Design ofAmplifiers and Oscillators by the S-Parameter Method, John
Wiley, New York, 1982.
P. C. Wade, "Novel FET Power Oscillators," Electronics Letters, September 1978.
Problems
10.1 Derive the transfer function for the series resonance circuit depicted in Figure 10-3 with v(i) = 0 and zero initial conditions. In particular, show that
-112
2
1/2
roQ = (LC)
(1- R C / (4L))
and a = -R/ (2L). For the values
R = 5 n , L = 50 nH, and C
270 pF, plot the frequency response.
10.2 In Section 10.1.3 the Colpitts oscillator centered around the h-parameter
description of the BIT in emitter configuration is derived. Follow similar
steps and derive the Hartley oscillator. Specifically, find the oscillator frequency in terms of L 1 , L2 , C 3 , and the h-parameters. Further, establish the
ratio of L 2 to L 1 .
Problems
10.3
593
A Colpitts oscillator is to be designed for 250 MHz. At the bias point

V CE = 2.7 V and I c = 2 rnA, the following circuit parameters are given at
room temperature of 25 o C: C BC = 0.2 fF, r BE = 3 kQ , r CE = 12 kQ,
C BE = 80 fF. If the inductance is fixed at 47 nH, find values for the capacitances in the feedback loop. Examine whether it is appropriate to use the hparameters obtained under DC conditions.
10.4 In Section 10.1.4 the quartz element is discussed. Show that solving (10.20)
results in the approximate series and parallel resonance conditions of equations (l 0.2la) and (10.21b). Hint: Use the Taylor Series expansion and retain
the first two terms.
10.5
Quartz resonators are typically specified in terms of their series and parallel
resonance frequency. For the electric equivalent circuit parameter of
Rq = 50 .Q, Lq = 50 mH, Cq = 0.4 pF, and C0 = 0.8 pF, find the seriP,s
and parallel resonance frequencies based on (l0.2la) and (10.2lb). Plot the
reactance of this quartz resonator over a suitable frequency range.
10.6 A particular crystal oscillator operates with the crystal in the parallel resonance mode. Then a lossless inductor is added in parallel with the crystal. If
the combination of inductor and crystal is required to have the same reactance as the crystal originally did, will the oscillator frequency go up or
down? Explain your answer.
10.7 In an oscillator design it is often required that we need the S-parameters of
the transistor operating in common-base (CB) mode. Unfortunately, the
manufacturer typically supplies the S-parameters for the transistor measured
in conunon-emitter (CE) mode. We therefore have to convert them into CB
S-parameters. The usual practice is that the S-parameters are first converted
into Y-parameters, then the CE Y-parameters are converted into CB mode,
and the result is finally converted into S-parameter representation. Derive CE
to CB conversion formulas for the Y-parameter representation.
10.8
A GaAs MESFET chip has the followingS-parameters in common-source

configuration measured at 4 GHz: S 11 = 0.83L-67, S21 = 2.16Lll9,
S 12 = 0.17 L61 o, S22 = 0.66L-23. Using conversion formulas derived
in the previous problem, compute the transistor S-parameters in commonbase mode. Determine the stability circles for both configurations without
and with a positive feedback of L = 0.5 nH at 4 GHz.
594
10.9 In Section 10.2.1 we discussed the case where r L is chosen such that 1rinl
becomes greater than unity. Show that jr inl > I implies 1
r oud > 1 and vice
versa.
10.10 In designing oscillators based on the S-parameter approach, it was stated that
the following conditions must be satisfied: k < I and r Sr in = r L r out = I .
By representing the input impedance as Z in = Rin + jXin and output impedance Zout = R out+ jXout as well as the source Zs = R 5 + jX5 and load
ZL = RL + jXL impedances, show that Rin = -Rs, xin = - Xs,
Rout = -RL' and X out = -XL . This proves that the S-parameter design is
equivalent to the negative resistance design.
IO.l1 An oscillator has to be designed for 3.5 GHz. The S-parameters of the BJT
in common-base configuration are determined to be S 11 = 1.1 L 127 ,
S 12 = 0.86Ll28, S2 1 = 0.94L-61 o, and S 22 = 0.9L-44 . By adding
an inductance to the base, the instability can be enhanced. Determine the
inductance for which the instability of the BJT is maximized.
10.12 In Section 10.2.2 the dielectric resonator is introduced and the S-parameters
for the angular resonance frequency ro0 are derived in (1 0.38). Show that
near resonance (1 0.38) has to be modified to the form
~
[S]
1 +j2(Qu!::t.fl fo)
1 + 13 + )2(Qu!::t.fI fo) 1 + 13 + j2(Qu!lf I fo)

1 +j2(Qu!::t.flf0 )
~
1 + J3 + j2(Qutlf I f 0 ) 1 + J3 + j2(Q1/l.f I f
0)
10.13 Since jrinl > I and jr outl > I , they cannot be displayed in a conventional
Smith Chart. Extend the Smith Chart in such a way as to be able to display
these quantities. What happens with the circles of constant resistance when
the reflection coefficients are larger than unity?
10.14 Design a 7.5 GHz oscillator in common-emitter BJT configuration. The Sparameters at V CE = 5.0 V and I c = 20 mA are as follows:
S 11 = 0.87 L-40,
S 12 = 0.25L-32 ,
S21 ;::; 0.6Ll00 ,
and
S22 = 1.21L165 . Sketch the circuit, including the DC biasing network
(~ = 80).
Problems
5115
10.15 A BJT is used in common-base configuration with biasing conditions specified at V C E = 3 V and V BE = 0.9 V. For this case, the transistor has the
at
2.5
GHz:
S 11 = 1.41L125,
following
S-parameters
S 12 = 0.389LI30 , S 21 = l.SL- 63, and S 22 = 1.89L-45. Design a
series feedback oscillator that satisfies the three conditions ( 10.26).
10.16 The S-parameters of a GaAs PET in common-source configuration are mea
sured at 9 GHz and have the following values: S 11 = 0.30L-167 ,
S 12 = 0.15L21.3 , S21 :::: 1.12L-23.5, and S22 = 0.90L-25.6 .
Design an oscillator with 9 GHz fundamental frequency and match the circuit to a 50 Q load impedance. Use microstrip lines for a substrate FR-4 with
40 mil thickness (Er = 3.6) and determine the widths and lengths of the
elements.
10.17 A tunable oscillator involving a varactor diode has to be designed. For the
varactor diode, the following data is known: equivalent series resistance of
35 Q and a capacitance ranging from 15 pF to 35 pF for reverse voltages
between 30 V and 2 V. Design a voltage controlled Clapp-type oscillator
with center frequency of 300 MHz and 10% tuning capability. Assume that
transconductance of the transistor is constant and equal to gm = 115 mS.
10.18 The output power of an oscillator can be approximated by
GoPin)J
p out = p sat [ 1 - exp ( P sat
where Psat is the saturated output power, G0

jS21 j is the small signal
power gain, and Pin is the input power. For maximum output power we
obtain
dPout
d(Pout- Pin) = 0 or dP
tn
Show that this leads to the maximum oscillator output power
Pout(max) =
P sat ( 1-
lnG0)
Go- Go
For a typical MESFET at 7 GHz, with G0 = 7 dB and Psat

maximum oscillator power.
= 2 W, find the
598
10.19 The basic downconverting receiver system is shown in Figure 10-21. Draw a
similar block diagram describing an upconversion transmitter system and
explain its functionality.
10.20 When building BJT and diode-based mixers, the third-order intermodulation
distortion (IMD) is an important design criterion. Ideally, over the entire
range of RF input signal magnitudes the mixer should not generate any intermodulation. In reality, however, there may be a significant influence. Follow
the same derivation as discussed in Section 10.3.1 and derive the first-, second-, and third-order harmonics for the combined mixer input signal
V = V RF cos ( roRFt) + V LOcos ( roL0 t) . If the RF signal is 1.9 G Hz and the
output IF is 2 MHz, determine all frequencies up to the third-order harmonics that are generated by this mixer.
10.21 Design a single-ended BJT mixer as shown in Figure 10-36. Compute values
for the resistors R 1 and R 2 such that biasing conditions V CE = 2.5 V,
V BE = 0.8 V, I c = 2.5 rnA, and I B = 40 J.lA are satisfied based on a
supply voltage of V cc = 3.2 V. RF and IF frequencies are
f RF = 2.5 GHz and f IF = 250 MHz. The BJT is measured at IF to have
an output impedance of zout = (650- }2400)0 for short-circuit input and
an input impedance of Zin = (80- j136)Q for short-circuit output at RF
frequency.
10.22 For the balanced diode mixer in Figure 10-41 assume the following voltages:
vRF(t) = VRFcos(wRFt) and vL 0 (t) = [VLO + vn(t)]cos(wL 0 t)
where the constant amplitudes are such that V RF V LO and where the noise
voltage vn is much smaller than Vw
(a) Find the currents through the upper diode i 1(t) and lower diode i2(t) if
the transfer characteristic is
in= C(-l)n+l, (n::: 1, 2)

where Cis a constant, and v1, v2 is the respective diode voltage.
(b) Explain how some of the noise cancellation occurs and show that the IF
current, after suitable low-pass filtering (behind each diode), can be
written as
i 1p = -2CVRF(VLo+ vn)sin[(wRF-wL 0 )t]
::: -2CVRF VLosin(w1pt)
APPENDIX
Useful Physical Quantities

and Units
Table A-1
Physical constants
Value
Symbol
Units
Permittivity in vacuum
F/m
8.85418xl0- 12
Permeability in vacuum
llo
Him
4xl0- 7
Speed of light in vacuum
mls
2.99792xl0
Boltzmann's constant
J/K
1.38066xl0-23
Electron charge
Coulomb
1.60218xl0- 19
Electron rest mass
mo
kg
0.91095xl0- 30
Electon volt
eV
1.60218xl0-19
Quantity
597
598
Appendix A Useful Physical Quantities and Units
Table A-2
Quantity
Relevant quantities, units, and symbols

Symbol
Unit
Value
to-15
femto
plCO
nano
mtcro
milli
Io-3
kilo
103
mega
106
giga
Mil
Mil
0.001 inch= 25.4J!m
Conductivity
(J
S/m
Resistivity
n-m
to-12
1o-9
10-6
109
International System of Units

Quantity
Unit
Symbol
Dimensions
Electric Charge
Coulomb
A s
Current
Ampere
Cis
Voltage
Volts
J/C
Frequency
Hertz = cycles per second
1/s
Electric field
Hz
E
Magnetic field
Aim
Magnetic flux
Wb
Weber
V s
Energy
Joule
Nm
Power
Watt
J/s
Capacitance
Farad
C/V
Inductance
Henry
Wb/A
Resistance
Ohm
VIA
Conductance
Siemens
A/V
V/m
Appendix A UM1'ul Physical Quantities and Units
Table A-3
Loss angle tangent for different dielectric materials
Material
Aluminum oxide
Barium titanate
Porcelain
Silicon dioxide
Araldite CN-501
Epoxy resin RN-48
Foamed polystyrene
Bakelite BM120
Polyethylene
Polystyrene
Teflon
Sodium chloride
Water (distilled)
f= 1kHz
0.00057
(AWG)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Loss Tangent
f= 1 MHz
f= 100 MHZ
0.00033
0.0003
0.00044
0.0140
0.00075
0.0024
0.0038
<0.0002
0.0220
0.0002
0.0075
0.0001
0.0190
0.0142
<0.0001
0.0280
<0.0002
<0.00005
<0.0003
<0.0001
<0.0002
0.00007
<0.0002
<0.0002
0.0400
Table A-4
Wire Size
599
Diameter In
Mils
289.3
257.6
229.4
204.3
181.9
162.0
144.3
128.5
114.4
101.9
90.7
80.8
72.0
64.1
57.1
50.8
45.3
f= 3GHz
0.001
0.0023
0.0078
0.0002
0.0340
0.0264
<0.0002
0.0380
0.00006
0.0270
0.0210
0.0001
0.0438
0.0002
<0.0001
<0.0002
0.00031
0.00033
0.00015
0.0050
<0.0005
0.1570
American wire gauge chart

Diameter In
Mllll meters
7.34822
6.54304
5.82676
5.18922
4.62026
4.1148
3.66522
3.2639
2.90576
2.58826
2.30378
2.05232
1.8288
1.62814
1.45034
1.29032
1.15062
Area In
Square Mils
262934
208469
165324
131125
103948
82448.0
65415.8
51874.8
41 115.2
32621.1
25844.2
20510.3
16286.0
12908.2
10242.9
8107.32
6446.83
Area In
Square
Millimeters
169.6345
134.4959
106.6606
84.59682
67.06296
53.19212
42.20364
33.46752
26.52585
21.04581
16.67370
13.23244
10.50709
8.327859
6.608296
5.230518
4.159237
600
Appendix A Useful Physical Quantities and Units
Table A-4
American wire gauge chart (Continued)
Wire Size
(AWG)
Diameter In
Mils
Diameter in
Millimeters
Area in
Square Mils
Aream
Square
Millimeters
IM
19
46
47
48
40.3
35.9
32.0
28.5
25.3
22.6
20.1
17.9
15.9
14.2
12.6
11.3
10.0
8.9
8.0
7.1
6.3
5.6
5.0
4.5
4.0
3.5
3.1
2.8
2.5
2.2
2.0
1.76
1.57
1.40
1.24
49
1.11
50
0.99
1.02362
0.91186
0.8128
0.7239
0.64262
0.57404
0.51054
0.45466
0.40386
0.36068
0.32004
0.28702
0.254
0.22606
0.2032
0.18034
0.16002
0.14224
0.127
0.1143
0.1016
0.0889
0.07874
0.07112
0.0635
0.05588
0.0508
0.044704
0.039878
0.03556
0.031496
0.028194
0.025146
5102.22
4048.92
3216.99
2551.76
2010.90
1604.60
1269.23
1006.60
794.226
633.470
498.759
401.150
314.159
248.846
201.062
158.368
124.690
98.5203
78.5398
63.6173
50.2654
38.4845
30.1907
24.6301
19.6350
15.2053
12.5664
9.73140
7.74371
6.15752
4.83051
3.87076
3.07907
3.291754
2.612199
2.075474
1.646293
1.297354
1.035224
0.818860
0.649417
0.512403
0.408690
0.321780
0.258806
0.202683
0.160545
0.129717
0.102172
0.080445
0.063561
0.050671
0.041043
0.032429
0.024829
0.019478
0.015890
0.012668
0.009810
0.008107
0.006278
0.004996
0.003973
0.003116
0.002497
0.001986
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
APPENDIX
Skin Equation for a

Cylindrical Conductor
The starting point of the skin effect analysis is
Maxwell's equations expressed by the laws of Ampere and Faraday in differential form:
VxH
= J = crE
(B.la)
VxE
=-~(a:)
(B.lb)
where the displacement current density (dEidt) in (B.la) is neglected inside a conductor. This is pennissible since the electric field in conjunction with the dielectric constant is very small, even for rapidly changing fields, when compared with the
conduction current. We evaluate these equations in a cylindrical coordinate system
where Ez , E,, and H~ are the only non-zero components. Carrying out the curl in
cylindrical coordinates, results in
(B.2a)
dH
--~
dz
= crEr
= 0
(B.2b)
(B.2c)
The second equation is zero because H' does not depend on the z-coordinate. Consequently, E, is also zero. Differentiating the last equation with respect to r, and then
substituting the first into it, yields a second-order differential equation:
2
d Ez +
()r2
!(aEz)_~cr(aEz)
= 0
dt
r
or
601
(B.3)
602
Appendix B Skin Equation for a Cylindrical Conductor
For time hannonic fields, the time derivative can be replaced by jro and combined with
2
J..LO to form the new parameter p = - jroJ..La. The final form
2
d Ez
2
+t(dEz)
- - +pE
=0
dr2
r dr
(B.4)
is the standard Bessel equation with the solution Ez = AJ0 (pr), where A is a constant
and J 0 is the Bessel function of zeroth order. Substituting this solution into the time
hannonic form of (B.2c) gives us
jroJ.LH<p
= ApJ0'(pr)
(B.5)
with the prime denoting differentiation with respect to the argument. The current is
related to the line integral of H ~ along the outer perimeter, r
a , of the conductor:
H<P21ta = I. Thus, we can write
= A(~)J '(pa) = _I_

(B.6)
"'
JO>J..l
21ta
which allows us to determine the constant A. Substituting A into the solution of the
Bessel equation leads to
H.._
E - jroJ..L I(-l.....,....o(_pr_))
z - 21tpa
(B.7)
J 0 '(pa)
An interesting property of the Bessel function is the fact that J 0 ' (pa) = -J 1(pa),
which gives us, after a small algebraic manipulation, the final result
E p I(-lo__(p
__r__))
z - 21taa J 1(pa)
(B.8)
This equation is used in Chapter 1. The validity of (B.8) for the case of zero frequency,
or DC condition, can be proved easily. For low frequency we see that
2
lo(pr) = 1- pr + (pr) - (pr)

+ ... ~ 1
( 2 )
2
2
(2. 4 ) (2. 4. 6)
J_2pa
_ pa [
(pa)
J 1(pa) - 2 I - 2 4
+ ... -
(B.9a)
(B.9b)
Substituting (B.9) into (B.8) yields Ohm's law for uniform current density Jz:
E _
Ip (~) _
z - 21taa pa
I
a1ta 2
= Jz
a
(B.lO)
APPENDIX
Co01plex Nuntbers
This appendix provides a brief summary of several

useful concepts and definitions regarding complex numbers and their manipulations as
repeatedly used throughout this textbook. Emphasis is placed on the basic definition of
a complex number, its use in the magnitude computations, and its meaning in terms of
the circle equation.
C.1 Basic Definition

A complex number z, such as the normalized impedance, can be represented in
rectangular and polar forms as
= x + jy =
lzl e1
(C. I)
where the magnitude is given by
lzl - ~
J(x+ jy) (x- jy)
Jx2 + y 2
(C.2)
and the phase is
e = tan- 1(y/x)
(C.3)
The star notation denotes the complex conjugate (i.e., z*
=x-
jy ).
C.2 Magnitude Computations

Let us apply the preceding definition to a typical computation involving the magnitude of two complex numbers such as
lz + w*l2
where w is another complex number of the form w
603
= u + jv . Substituting w yields
Appendix C Complex Numbert
804
(C.4)
lz + w*l = (z + w*) (z* + w ) = lzl + lwl + 2Re{ z w}
where we used the fact that the terms z w = ux- vy + j(uy + vx) and
z* w* = ux - vy - j(uy + vx) can be combined to 2Re{z w}. Here Re{ ... } represents the real part.
C.3 Circle Equation

Perhaps one of the most useful equations involving complex numbers in RF circuits is the circle equation
2
(C.5)
lz - wl = r or lz- wl = r
which forms the foundation of the Smith Chart. We can verify that this is indeed a circle
equation by going through the magnitude computation
2
*
2
2
2
(C.6)
lz-wl = (z -w)(z-w) =(x-u) +(y- v) = r
It is seen that u and v are the coordinates of the circle center in the complex z-plane and
r is its radius, as depicted in Figure C-1.
Figure C-1
Circle representation in the complex z-plane.
APPENDIX
Matrix Conversions
Conversion between Z, Y, h, and ABCD representations
[Z]
[Y]
Z22
-~
zll z,2
[Z]
Z21 Z22
Zzt
-~
[YJ
-tYZ -zl2
Z11 tYZ
-
Z22
-Zu
tlZ
Z21
Z22
Z22 Z22
y12
.1Y
.1Y
YII YI2
I
- -y12
-r,,
yll
y21 y22
y21
ar
-aY
M
h12
h22 h22
h12
-l -hll
h21 1
-hzz h22
h21 llh
-hu
-hi!
-Ac
MBCD
-c1
-c
1!.2 = ZuZ22-Zl2Z2t' 6.Y =

MBCD:::: AD - BC
yll
yll
[ABCD]
[ABCD]
y22
y21
[h]
Z12
--
[h]
h,,
hll
D
B
--B1
l Z22
Z21 Z21
I
-y22
--y21
6.h
-DB
MBCD
D
1
-D
= h11h22 -h12h21
y 21
yll
6.Y
---y 21
y 21
6.h hll
---h21 h21
h22 1
---h21
605
hl2
h2l h22
MBCD
YllY2z -Yt2Y2I
-6.Y
r,,
Z21 Z21
h21
AB
CD
606
Appendix D Matrix Conversions
Conversion from S-parameters to Z, Y, h, or ABCD representations
[Z]
Zu
= Zo
Zzz
= Zo
where
[Y]
Yu
Y22
where
[h]
(1 + u )( 1 - Szz) + 12Szt
'PI
2Stz
Z12
= Zo1
Y,2
12
=Zo'Pz
Z21
2Szt
= Zo
'PI
{1- u )( 1 + s22> + S12s21
'P 1
'P
1
= (1- S11 )(1- S22 ) - S 12S 21
( 1 - S u )( 1 + S22 ) + S 12S 21
-2S
Zo'Pz
y 21
-2Szt
--Zo'Pz
(1 +S 11 )(1-S22 )+S 12S 21

Zo'Pz
'2
h11
= Zo
h22
= (1-S11 )(1-S22 ) -S12S 21
(1 + Su)( 1 + Szz)- StzSzt
2Stz
hl2 = -
'P3
h21
'3
-2S21
-
(1-Su)(l-S22 ) -S 12S21
2 o'P3
where '3 = (1-S 11 )(1 +S22 )+S 12S 21

A=
[ABCD]
C=
( 1 + S 11 )(1- S 22 ) + S 12 S 21
2Szt
(1-S,t)0-Sz2)-S,zSzt
2SztZo
= Zo
D=
(1 +
11 )( 1
+ s22>- 12s21
2S
21
(1-Su)(l +Szz) +S1zSz1

2Szt
'P3
607
Appendix D Matrix Conversions
Conversion from Z, Y, h, and ABCD representations to Si)arameters.
S II =
[Z]
Szz =
(Zu - 2o)(Z22 + Zo) - 212221
SJz =
'1'4
2Z 12Z 0
'1'4
Szt
2ZztZo
'1'4
(Zu + Zo)(2zz- Zo)- Zt2Z21

''4
where ' 4 = (Z 11 + Z 0)(Z 22 + 2 0 ) -Z 12 Z 21
s"
[Y]
(1-ZoYJ,)(l + ZoY2z) + YI2YztZ5
Szz =
'Ps
[h]
s"
s22
[ABCD ]
Szz
'l's
Szt =
-2Y21 Z0
'f's
'Ps
= ( 1 +Z0 Y11 )(1 +Z0 Y22 ) - Y12 Y21 Z5
(h 11/Z0 -l)(h 22 Z0 + l)-h 12 h 21
'1'6
sl2
2h12
=-
Sz,
''6
-2h21
- -'6-
(h ll/Z 0 + 1)(h 22 Z 0 - 1) + h 12 h21
'6
where ' 6
sll
12 Z 0
( 1 + ZoYII)( 1 -2oYzz ) + Y12Yz1Zij
where ' 5
=
-2 f
S12 =
= (h 11 /Z0 + 1)(h 22Z 0 + 1) -
A+BIZ0 - CZ 0 -D
'1
-A+BIZ0 -CZ0 +D
'7
where '1' 7 = A+ B/20 + CZ 0 + D
h 12h 21
_ 2(AD-BC)
12 -
''7
S21
= '7
APPENDIX
Physical Parameters of
Semiconductors
Table E-1
Properties of Ge, Si, and GaAs at 300K
Properties
Ge
Sl
GaAs
Dielectric constant
16.0
11.9
13.1
Energy gap (e V)
0.66
1.12
1.424
Intrinsic carrier concentration (cm-3 )
2.4x10
13
1.45x10
10
1.79x10
8
Intrinsic resistivity ( Q em)
47
2.3xl0
Minority carrier lifetime (s)
10-3
2.5xl0-
Electron mobility (drift) (em IV S)
3900
1350
8500
Normalized effective mass of the electron (m:lme)
0.55
1.08
0.067
Hole mobility (drift) (cm IVs)
1900
480
400
Normalized effective mass of the hole ( m; I me)
0.37
0.56
0.48
Electron affinity, X(V)
4.0
4.05
4.07
Specific heat (J/(g. K))
0.31
0.7
0.35
Thermal conductivity :WI ( c m K))
0.6
1.5
0.46
0.36
0.9
0.24
Thermal diffusivity (em Is)
608
10
10-8
APPENDIX
Long and Short Diode

Models
l : e current flow through a diode under an applied

forward bias voltage (see Chapter 6) can be evaluated based on the concentration of the
injected excess charge carriers in each semiconductor region. Depending on the
length of the semiconductor layers, we need to differentiate between a long and short
diode model. In the following discussion the current flow is derived for both cases.
With reference to Figure F-1, let us examine the pn-junction under forward bias
voltage VA.
I
-.
...,
II
Ij
-.
...,
Pn(dn)
np(-dp) I
npO
it
. . . ...............
Pno
--~--------+-~-r------~~~x
-(dp + Jp)
Figure F-1
-dp
dn
dn + W,
pn-junction under forward bias.
Under this applied voltage the junction is no longer in thennal equilibrium, and
minority concentrations are created that exceed the equilibrium condition n po in the player and Pno in then-layer. Indeed, thennodynamic considerations predict the minority concentrations in each layer as
Pn(dn) = Pnoe
VAIVr
and np(-dp) = np0 e
The corresponding excess charge concentrations
609
VAIVr
(F. I)
Appendix F Long and Short Diode Modell
610
!l.pn
= Pn- Pno and !l.nP = nP - npo
(F.2)
begin to diffuse into the semiconductor layers, a process governed by the steady-state
diffusion equation. For then-layer, the equation reads
2
d (!l.pn)
!l.pn
dx2
DPtP
(F.3)
where DP, 'tP are the diffusion constant of holes in then-layer and excess charge car7
6
rier lifetime (on the order of 10- 10- s), respectively. It is the so-called diffusion
length
(F.4)
with respect to the length of each semiconductor layer that determines whether we have
to deal with a long or short diode model. The general solution to (F.3) is
!l.pn = C 1ex/ LP + C 2 e-x/LP , with two unknown constants to be determined through
the boundary conditions on either end of the semiconductor layer. The following two
cases are considered:
F.1
Long Diode (Wn > Lp, ~n ~ 0 as X~ oo)
Since the excess carriers completely decay to zero before reaching the end of the
layer, only C 2 has to be specified and C 1 = 0 . Applying (F.l) as a boundary condition, we can find C 2 and insert it into the general solution, with the result
!l.pn = Pno(e
VAIVr
- l)[e
-(x-d11 )1 L1,
(F.5)
In an identical way we can find for the p-layer ( W P > Ln, !l.n P ~ 0 as x
11nP = np0(e
VAIV 7
- 1 )[e
(x+dp )IL11
~ -oo)
(F.6)
F.2 Short Diode (Wn < LP' !l.pn ~ 0 as x ~ dn + Wn)

Here the situation is more complicated since the decay takes place over a finite
distance. As as result, both coefficients have to be retained. The additional boundary
condition on the right-hand layer now reads Pn(dn + W n> = Pno. Going through the
mathematics eventually leads to
_
11pn - Pno(e
v A I v7
(sinh [ ( d n + W n - x) I L
-1)
sinh[Wn/LP]
pl)
(F.7)
which can be further simplified by approximating the hyperbolic sine function, sinh, by
its argument. This is permissible as long as the layer length is less than the diffusion
Appendix F Long and Short Diode Models
611
length ( W n < LP ). The final result is
Llpn = Pno(e
VA I Vr
- 1)
(dn + Wn -X)
L
(F.8)
Similarly for the p-layer (W P < Ln, dnP

_
flnP - nP 0(e
VA IV r _
0 as x
1)
~ -(dP
+ W P))
(X - (d p + W p ))
(F.9)
Similar to (6.14), equations (F.5), (F.6) or (F.8), (F.9) can be used to find the total current through the diode:
I
dllpn)
= A[J p(d.) + J.(dp)l
= A [ (-q)DP ( dx
d.,+ qD.
(dllnp)
dx
-d,,
(F.lO)
Inserting (F.5), (F.6) or (F.8), (F.9) into (F.l 0) finally results in the Shockley equation:
I= J0 (e
VA I VT
(F.ll)
-1)
where the reverse saturation current is for the long diode
lo-A
[qDpPno
L
qDnnpo]
(F.12)
and for the short diode
_ [qDWpPno + qDnnpo]
W
lo-A
(F.13)
A typical numerical example for a short Si diode involves the following

parameters:
-5
= 22 em 2 Is , D P = 9 em2 Is , N A = 1.5 x 10 16 em -3 ,
=2x
ni
= 1.5 X 10 10 em-3 , N D = 3 X 10 16 em-3 , 'tP = 'tn
10
em , D n
= 10
-7
s,
Wn = W P = 25 )lm .
With these data we can compute the minority carrier electron and hole concentrations
in thermal equilibrium:
2
3
-3
2
3
-3
p, 0 = ni/ND = 7.5x10 em ,npo = n .INA = 15x10 em
Inserting into (F.13) results in a reverse saturation current of 0.5 fA.
APPENDIX
Couplers
ranchline couplers and power dividers play
important roles in RF circuits and measurement arrangements since they allow the separation and combination of RF signals under fixed phase references. Notably, in the
mixer section of Chapter 10 and the measurement protocol of characterizing a device
under test in Chapter 4, we see their usefulness. The purpose of this appendix is to discuss some of the couplers and dividers encountered most often in terms of their Sparameters and figures of merit.
G.1 Wilkinson Divider

The transmission line configuration and its microstip line implementation of this
power divider are shown in Figure G-1. The S-parameters for such a three-port network
are given by the matrix
[S]
(G.l)
The figures of merit are the return loss at ports 1 and 2
RL 1
= -20logJSuJ
and RL2 = -20logJS22 J
(0.2)
the coupling between ports 1 and 2

CP 12 = -20logJS21 J
(0.3)
and isolation between ports 2 and 3

IL 23
= -20logiS23 1
(G.4)
Figure 0-2 provides a typical frequency response of RL 1 , C P 12 , and I L 23 for a center

frequency of f 0 = 1 GHz.
612
613
Appendix G Couplers
Port 2
Port 1
2Z0
Z0
Port 3
(a) Transmission line model
(b) Microstrip line realization

Figure G-1
3 dB Wilkinson power divider.
- 10
- 15
..,"
N
~ -20
- 3.2
- 3.3
~--25
l...---
0.5
_ _ I . . --
0.75
- - L . . - --
1.0
...1....--- - . J
1.25
Normalized frequency, fifo
Figure G-2
1.5
- 30
0.5
0.75
1.0
1.25
Normalized frequency, fifo
Frequency response of Wilkinson power divider.
1.5
614
Appendix G Couplers
Ideally, return loss and isolation should approach negative infinity at the center
frequency and the coupling should be as close to the 3 dB value as possible. We also
note that a coupler is not a broadband device. 'TYpical frequency bandwidths do not
exceed 20% of the center frequency.
The derivation of matrix (G.l) is most conveniently carried out by an even and
odd mode analysis, as depicted in Figure G-3 for the computation of the S 12 coefficient. We attach a source V s to port 2 and tenninated the other two ports with a Z 0
load. To make the circuit symmetric the source V s at port 2 is divided into a series
combination of two V 5!2 sources operating in phase. At port 3 two V5 !2 sources
have a 180 phase shift and their sum is equal to zero. Also, the Z 0 load impedance
connected to port 1 is replaced by the parallel combination of two 220 impedances.
Vs/2
evenrnode
o.c.
&
Vs/2
~odd mode
o.c.
(a) Even mode

(b) Odd mode
Figure G-3 Even and odd mode representation of Wilkinson divider
(o.c. = open circuit).
The reason for choosing the odd and even mode decomposition becomes immediately apparent. Let us consider at first the circuit in Figure G-3(a), which is driven by an
even mode, meaning that the drive signals at ports 2 and 3 are in phase. In this case both
ends of the 2Z0 cross impedance have the same potential. Thus, there is no current flow
615
Appendix G Couplers
and the impedance can be neglected. The input impedance seen at port 2 for this case is
the impedance of a J2z0 quarter-wave transformer terminated with a 2Z0 load [i.e.,
Z 2 = CJ2Z0 ) 2 /(2Z0 ) = Z 0 ]. Consequently, in the even mode excitation, port 2 is
perfectly matched and the voltage at port 2 is V~ = 0.5 ( V sl 2) = V sl 4 . The corresponding voltage at port 1 can be found based on our discussion regarding the voltage
distribution along a transmission line (see Chapter 2):
+
V 1 = V (1 + r 0 )
(0.5)
where r~ = (2Z0 - J2Z0 )/(2Z0 + J2Z0 ) is the even mode reflection coefficient at
port 1. Therefore, the even mode voltage at port 1 is
ve1
v+ (1 + r e0 )
= 1
v2er~+1
r~-1
-jJ2 vs
4
(0.6)
and where the factor j is due to the /J4 transmission line. For the odd mode excitation
voltages at ports 2 and 3 have opposite polarities and there is a zero potential along the
middle of the circuit. This means that the middle is shorted to ground. Since the input
impedance seen from port 2 is again Z 0 and port 1 is grounded, we find that V~ = 0
and V~ = Vs/ 4.
The total voltage at ports 1 and 2 is found by adding the even and odd mode voltages. The corresponding S 12 parameter is then computed as
e
S1z
= vl
Vz
= vt +vi = _...L
v; + v~
J2
(G.7)
An identical analysis for the port 3 to 1 configuration results in S 13 = - j I J2 .

Furthermore, because the divider is a linear, passive network we conclude that
S 21 = S 12 and S 31 = S 13 Also, both in the even and odd mode analysis port 2 is isolated from port 3 by either an open circuit or ground, we find that s23 = s32 = 0.
Thus, all off-diagonal tenns in (G.l) are verified.
In addition, S 22 = S 33 = 0 is due to the matching of the odd and even modes.
This leaves us only to prove that S 11 = 0 . We notice that when port 1 is driven, the
current through the 2Z0 resistor between ports 2 and 3 is again zero and has no influence on the circuit. Thus, the impedance Z 1 seen at port 1 is a parallel combination of
two Z 0 terminations connected through J2z 0 quarter-wave transformers
Zt
1(J2Zo)2
= 2-
Z0
= Zo
This proves that port 1 is matched (i.e., S 11 = 0 ).
(0.8)
616
Appendix G Couplere
G.2 Branch Line Coupler

There are two 3 dB branch line couplers of importance. According to their phase
shifts, they are either referred to as 90 (quadrature) or 180 couplers. The S-parameter
representation for the 90 coupler is
0 j 10
[S ] = -1 j 0 0 1
90
(G.9)
J2tOOj
0 l j 0
and a circuit schematic is shown in Figure G-4.
Port 3
Z0 /J2
Figure G4 Microstrip line realization of quadrature hybrid.
Besides return loss, isolation, and coupling definitions given in (G.2)-(G.4), the directivity of a branch coupler is a key parameter and defined as
(G.lO)
where D 34 ideally approaches infinity at f 0

In our derivation of (G.9) we begin by using an even and odd mode analysis, as
depicted in Figure G-5. We drive the hybrid at port 1 with an RF source V s and terminate the remaining ports into the characteristic line impedance Z 0 An equivalent circuit results if the source voltage at port I is written as the sum of an even ( V 1e ) and odd
(V 10 ) voltage such that V 1 = Vs = V 1e + V 10 with V 1e = Vs/2 and V 10 = Vs/2.
At port 4 we can enforce zero voltage condition by setting V 4
0
V 4e + V 40 ,
where V 4 e = Vs/2 and V40 = -Vs/2.
The total transmitted voltage at port 2 due to the input voltage at port 1 can be
established as
= =
(G. II)
Appendix G Couplers
617
"fs /2
"fs/2
Aft
~./2
Vs /2
Port 4
vt
Zo
Zo
Zo
vt
Zo
Z0 !./2
Zo
+-- IJ4----+ Port 3
Zo Port 1
Port 2
Z 0 1/2
Aft
~ Odd mode
Even mode &

Zo Port 1
Port 2
Z0 1./2
Zo
Alt
Zo
Port 2
Z0 /./2
Alt
Zu
Zo
Zo
(a) even mode

(b) odd mode
Figure GS Building blocks of a branch line coupler.
and similarly
(G.l2)
V4
Vs
= ( T e-To)2
(G.l3)
= S41Vs
The reflected signal at port 1 is
(G.l4)
We must next tum our attention toward finding T e , T 0 , r e , and r 0 The transmission
line circuits in Figure G-5(a) and (b) can be represented as a three-element model
involving either short or open-ended stub lines A./8 in length.
The even mode and odd mode stub lines have an admittance of
Ye
= Yoc = Z-1 tan (1t)

-4
0
and
(1t)
4
Y 0 = Y sc = -- I cot -
z0
(G.15)
Appendix G Couplers
618
The three-component circuit in ABCD network representation is then
_ [ 1 ol [ cos(f3l) jYA sin(f3l)l [ 1 ol{ Vu2}

{Vui}
lu1
jYe,o tj jYAsin(f3l) cos(f3/) J jYe,o tj -lu2
1
(0.16)
ABl{ Yu2}
- [c vj -Iu2
where Y A = 1I Z A is the admittance of the AI 4 line element. Multiplying the three
matrices and converting the result into S-parameter form yields, after some rather tedious
computations, the following nonzero coefficients: S 21 = S 12 = - j(ZAIZ0 ),
.
2 1/2
= S34 = -J(ZAIZ
),
and
S
=
S
=
-[1-(ZAIZ
)]
= S42 = S24 . Set0
31
13
0
ting Z A = Z 0 1 J2 gives the desired matrix listed in (0.9). Again, it is noted that all four
S43
ports are matched into Z 0

The 180 coupler can be constructed by adjusting the lengths of the four transmission line segments and arranging the impedances in a ring configuration, as shown in
Figure 0-6.
Port 2
Port I
Port4
Port 3
Figure G6 A 180 ring coupler.
The S-parameter matrix for this configuration, also known as "rat race," is given by
.
-}
1 1 0
l 0 0 -1
-,J21 0 0 1
0 -1 1 0
(0.17)
Appendix G Couplers
619
G.3 Lange Coupler

Port2
Port I
Port 3
Figure G-7 A 3 dB lange coupler.
A popular implementation of the quadrature hybid in microstrip line form is the

so-called Lange coupler shown in Figure G-7 for a four-strip configuration. Additional
variations involve six- and eight-strip realizations. The interdigital form of the microstrips permits a very compact geometric size and provides for tight coupling.
Typical coupling values range between -5 and -1 dB. By choosing the length of
the microstrip elements appropriately, a very broadband realization of up to 40% bandwidth can be achieved.
Further Reading
P. Karmel, G. Colef, and R. Camisa, Introduction to Electromagnetic and Microwave

Engineering, John Wiley, New York, 1998.
J. Lange, "Interdigitated Stripline Quadrature Hybrid," IEEE Trans. on MTI; Vol. 17,
pp.ll50-1151, 1969.
APPENDIX
Noise Analysis
' I : e intent of this appendix is to provide an overview of the most important noise definitions and concepts as related to the noise figure
analysis conducted in Chapter 9.
H.1 Basic Definitions

In a broad sense noise can be characterized as any undesired signal that interfers
with the main signal to be processed. Examples of noisy signals are AC power coupling, crosstalk between circuits and electromagnetic (EM) radiation to name but a few
sources. Mathematically we use random variables of Gaussian distribution and zero
mean to describe the noise behavior. Although the mean is zero, the root mean square
(RMS) value of a noisy voltage signal vn(t) is not. This can be expressed as
(H.l)
where T 1 is an arbitrary point in the time and T M is the measurement interval.

In 1928 Johnson first observed the fact that a resistor in the absence of any exter"
nal current flow generates noise due to the random motion of charge carriers in the conductor. The noise power in a conductor is quantified as
(H.2)
Pn = kT&f = kTB
where k is Boltzmann's constant, Tis the absolute temperature in K, and !l.f = B is
the noise bandwidth of the measurement system. The noise bandwidth is defined as the
integration of the instrument's gain G(/) over all frequencies normalized with the
respect to the maximum gain G max :
B
= c!-[
G(f)df
max 0
(H.3)
We next turn our attention to the noise voltage. Let us consider the simple circuit
shown in Figure H -1.
620
121
Appendix H Noise Analysis
Figure H-1
Noise voltage of a circuit.
According to this circuit, the noise power is treated as if a noise voltage source
drives a noiseless resistor R s . Under matching condition R s = R L, the noise power of
the resistor is given as
2
pn
VnRMS
4R
= kTB
(H.4)
from which the RMS noise voltage is found

(H.5)
V nRMS = J4kTBRs
To keep the notation simple (and since no ambiguity will arise) the subscript RMS is
dropped (i.e., V nRMS Vn ). In general, we represent a noisy resistor R as a noise voltage source in series with the noise free resistor R (Thevenin equivalent circuit) or as a
noise current source In = J4kT Bl R in shunt with a noise free resistor, as shown in
Figure H-2.
R
(noise free)
~Jn
R
(noisy)
Figure H-2
Equivalent voltage and current models for noisy resistor.
If the the bandwidth is eliminated from (H.5) we can define a so-called spectral
noise voltage and a spectral noise current:
Vn = Vnl JiJ
whose units are given in VI
and
In
JHZ and A/ JHZ .
= In / JiJ
(H.6)
622
Appendix H Noise Analyale
Frequently, the spectral density S(j) is used to quantify the noise content in a unit
bandwidth of 1 Hz. For the thermal noise source associated with resistor R, it is given
by
vz
S(f)
= 2B = 4kTR
(H.7)
If S(f) is a constant (i.e., independent of frequency), we speak of white noise. Care is

required when noisy elements are added in a circuit. For instance, if two noisy resistors
R 1 and R 2 are added, the associated noise sources V n 1 and V n 2 cannot be linearly
summed. Instead, the resultant noise source V n is
Vn =
Jv;
1+
V~ 2
(H.8)
provided both noise sources are uncorrelated. This is equivalent to saying that only
power proportional voltage square quantities can be added because of their random distribution of amplitudes and phases as well as different nonharmonic frequencies.
If the noise sources are correlated, a correlation coefficient Cnl, n 2 enters (H.8)
such that
2
V n = V n I + V n2 + 2 Cn 1, n2 V n 1 V n2
(H.9)
where -1 $; C n 1 n 2 $; 1 . It is interesting to observe that if V n 1 and V n 2 are 100% corre,
2
2
2
2
lated (Cnl,n 2 = I), then Vn = Vnt + Vn 2 + 2Vn 1Vn 2 = (Vn 1 + Vn 2 ) and the voltages can again be added, in agreement with Kirchhoff's linear circuit theory.
The thermal noise of a resistor is also referred to as an internal noise source,
since no external current has to be impressed to observe the noise voltage. However,
many noise mechanisms only occur due to externally impressed current flow through
the device. They are collectively known as excess noise. Chief among them are the 1/f
noise (also known as flicker noise, semiconductor noise, pink noise) and shot noise.
The 1/f noise is most prominent at low frequencies and exhibits, as the name implies,
an inverse frequency-dependent spectral distribution. It was first encountered in vaccum
tubes as a result of "flickering" noticed on the plates. The shot noise is most important
in semiconductor devices and can be attributed to the discontinuous current flow across
junction potential barriers. As an example, in a semiconductor diode the reverse bias
noise current I Sn is given as
Isn = J4qlsB
where Is is the reverse saturation current and q is the electron charge.
(H.lO)
Appendix H Noise Analysle
623
H.2 Noisy Two-Port Networks

The previous analysis can be expanded to two-port networks. Figure H-3 shows a
noisy network and the equivalent noise-free network augmented by two current noise
sources In 1 and I n 2 .
~,..
Jl
n.
v1
]v, ~ vJ
Noisy
network
/I
Noise free
network
1,,
Figure H-3 Noisy two-port network and its equivalent representation.

In Y-parameter matrix representation we can write
{lz/1} = [Yil Y12l{ V
Vl
Y2 1 f22J
}+{ /nl}
(H.ll)
ln2
A more useful representation is obtained when rearranging (H.ll) as follows:
Yzz
1
1
VI= --Vz+-12--y ln2
Yzl
Y21
(H.l2a)
21
and
I,=
Yll Y22- YrzYzt

y
21
Yn
Yu
Vz+y-12+/ni-ylnz
21
(H.l2b)
21
Defining the transformed voltage and current noise sources

1
Vn = -y-ln2 and In =
Yu
/nt-
y-Inz
21
we arrive at the network model shown in Figure H -4.
21
(H.l3)
624
Appendix H Nofse Analysis
v,
VI
In
Noise free
network
Ji2
~~----~L-----~~~
~- ~~ isy network
Figure H-4 Transformed network model with noise sources at the input.
To apply the various noise definitions and concepts to a practical example, let us
consider a simplified BJT amplifier.
-------------------------------~~~
Example H-1: Noise analysis of a low-frequency BJT amplifier
.
"
;-.
In Figure H-5 a simplified BJT amplifier is treated as a two-port network consisting of the following parameters: V s = 25 m V,
Rs = 50 Q, R in = 200 n, voltage gain gv = 50, and measurement bandwidth B = 1 MHz. The spectral noise voltage and current
of the a~lifier are ~en by the manufacturer as V n = 9 n V / JHZ
and
In = 9fAJHz.
Find
the
signal-to-noise
ratio
SNR = 20log(V2 /Vn 2 ) at the output.
....
-~
. ............ . ....... .
~ --- --
... -- ............ ---- ... ---- -........ . .... .
........ . ........................... -- ........... ............ ............ -- ------ ...
Figure H..
Amplifier model and network representation with noise sources.
The
output
voltage
V2
is
directly
found
from
V 2 = gvRin / (Rin + Rs)Vs = 1 V. The spectral noise sources of
the network are next expressed in RMS noise voltage and current:
Vn
= VnJB = 9J.!V
and
111
=I
11
JB
= 9 pA
625
The voltage source creates through the voltage divider rule the following noise voltage across Rin:
Rin
---Vn
Rin + Rs
= 7.2 nV
The noise current source is responsible for the noise voltage of

R- Rs
m
Rin + Rs
In = 0.36 nV
Finally, the source resistor contributes the voltage

Rin
R in+ R S
where Vns
= J4kTBRs
V ns
= 728 nV
= 910 nV, assuming T= 300K.
Therefore, the total noise voltage at the output is
v.2 = gv
.r (R~:R~/Y (R,:~R/Y
(R,.R~"R/
= 36.4 "v
Finally, the signal to noise ratio is
SNR = 201og(::J = 122.8 dB

We notice that the noise voltage is dominated by the source.
The example makes clear how the noise voltages are individually computed, added, and amplified to provide the output noise
voltage. This is in stark contrast to linear circuit theory.
H.3 Noise Figure for Two-Port Network

The noise figure is defined as the ratio between the SNR at the input to the SNR
at the output port of a network. Specifically, Figure H-6 depicts the relevant power flow
conventions, including the noise representation of the source Z s.
The noise figure F can be cast into several equivalent representations. The first
form involves the ratios of the signal to noise power at the input and output ports:
F=
P1/Pnl
Pn2/P2
=--P2/Pn2
Pn1IP1
(H.14)
Appendix H Noise Anlysls
626
Noise free
network
Figure H6
Generic noise model for noise figure computation.
Employing the available power gain G A from Section 9 .2.3 to express P2

and PnZ = GApn 1 + P,.;, (H.l4) is re-expressed as
= 1+
GAP 1
p .
m
GAPnl
where Pni is the internally generated noise power within the amplifier.
Based on Figure H-6, we see that the signal power P 1 ts
P1 =
1 Re { Zin}
IZs+ zinl
Vs
2
(H.15)
(H.16)
which is less than the power under source matching ( Z s = Zi~ ):

2
p.
= l2 (4Re{Zin}
Ivsl J
(H. l7)
Zs = z in
The thermal noise at the input side is with Z s = R s + j X s:
Pnl
= 4kTR S B
Re{Z }
m
IZs + zinl
(H. IS)
The power ratio is therefore
vzs
vns2
(H.l9)
The signal power P2 is simply P 2 = GAP 1 , where P 1 is given by (H.16). For the
noise power P112 we set Pnz = GAPn 1 + Pni' where the internally generated noise
power Pni takes into account the noise sources associated with the two-port network
V n and I 11 Thus, V~s in (H.18) has to be replaced by all three noise sources:
827
V~s + V~ + (/nRi0 )
where Rin = Re{Zin} is the input resistance of the network.

Since the gain applies equally to signal and noise, it cancels and we arrive at
,
Pz1Pn2 =
v2s
(H.20)
vns + v n + UnRin)
The noise figure therefore takes the form
2
Vn s + Vn + (InRin)
F =
vns
Vn + (InRin)
l + 4kTBR.
(H. 2 l)
The preceding treatment does not take into account the fact that the same noise
mechanisms are usually responsibe for both V n and In . Thus, these sources are, to a
certain degree, correlated. This can be incorporated into the noise model by splitting In
into an uncorrelated, I nu , and a correlated current, Inc, contribution, respectively. The
correlated current contribution is related to the noise voltage V n via a complex correlation factor Y c = G c + j B c , such that Inc = Y c V n . Since it is more convenient to
deal with noise currents than voltages for our network, we convert the source into an
equivalent Norton representation, as seen in Figure H-7.
Figure H-7 Noise sources modeled at network input.
The total RMS noise current I ntot under short circuit input conditions can be expressed
as
(H.22)
where Inc = Y c V n and In
now rewrite (H.21) as
= V nY s are combined because of their correlation. We can

2
F =
Ins+ Vn(Ys + y c ) + Inu

2
(H.23)
Ins
Under the assumption that all noise sources are represented by an equivalent thermal
noise source, we identify in (H.23)
628
Appendix H Nolte Analysis
l~s == 4kTBG5 :noiseduetothesource Y 5 = Gs+iBs
(H.24)
l~u
(H.25)
= 4kTBGu : noise due to the equivalent noise conductance Gu
V~ = 4kTBRn: noise due to the equivalent noise resistance Rn
(H.26)
Inserting (H.24)-(H.26) into (H.23) gives

(H.27)
The circuit designer can minimize (H.27) through an appropriate choice of source
admittance Y s. This process is accomplished by first observing that the imaginary part
2
can be chosen such that Bs =-Be . This eliminates the (B5 +Bc) term in (H.27).
Next, the remaining expression is minimized with respect to Gs; that is,
dF(Bs= -Be)
dG
=
- 2-{Rn[2GsoptCGsopt
+ Ge) - (Gsopt + Ge) ]} = 0
(H.28)
Gsopt
which yields the explicit optimum value

Gsopt
= ~JRnG~+ Gu
(H.29)
~jRnG~+ Gu)-iBc
(H.30)
;.JRn
The minimum noise figure is thus obtained by the optimal source admittance
Ysopt = (
;.JRn
Substituting (H.29) into (H.27) results in the expression
F min
Gu
Rn
= I+~+ c;-CGsopt + Ge)

Sopt
(H.31)
Sopt
Eliminating Gu in (H.31) by using Gu = RnG~opt- RnG~ from (H.29) gives

F min= 1 + 2Rn(Gsopt + Gc )
(H.32)
This number is typically provided by the device manufacturer. It is dependent on frequency and bias conditions. Equation (H.32) can be incorporated into (H.27) with the
result
(H.33)
Replacing Gu by Gu
result
F
629
2
2
= RnGSoptRnGc
Rn
= Fmin+(f[(Gs-Gsopt)
and rearranging terms provides the final
+(Bs-Bsopt)
Rn
J = Fmin+(fiYs-Ysoptl
s
(H.34)
This is the starting point of our noise circle analysis in Section 9.5. Based on the characteristic line impedance Z 0 = 1I Y 0 , (H.34) is often expressed in terms of normalized
noise resistance r n = Rnl Z 0 , conductance g s = G slY0 , and admittances
Ys = Y 8 /Y0 , Ysopt = Ysopt/Y0 in the form
rn
2
F = F min+ )Ys- Ysoptl
8
(H.35)
H.4 Noise Figure for Cascaded Multipart Network

The previous noise figure discussion for a single two-port network, with Pnl
being the input noise and Pn 2 = G APn 1 + Pni being the output noise, can be extended
to multiple cascaded networks, as shown in Figure H-8.
1
P2 l
GAl
-r
pnil
f!l
GA2
P,i2
GAk
pnik
I'nk
pnl
Figure H-8
Pxi
Cascaded network representation.
In accordance with Figure H-8, we adopt a suitable notation such that GAk and
P nik denote power gain and internal noise generated by amplifier block k = 1, 2, ...
Thus, for the noise power at the second amplifier section it is seen that
Pn3
GA2(GA1pnl
(H.36)
+ Pnil) + Pni2
or for the total noise figure Ftot we see

Ft0 t =
p 3
n
pn1GA1GA2
= 1+
p '1
m
pnlGAI
p "2
nt
(H.37)
pn1GAIGA2
It is customary to retain the same noise figure expression for the individual blocks as
derived for the single network; that is,
630
F1 = 1 +
1
nt
pniGAl'
F2 = 1
Pn,2
+ pn1GA2' ... ,
1+
p 'k
m
pnkGAk
(H.38)
For two networks this concept leads to the expression

F 2 -1
(H.39)
F tot = Fl + -G-Al
or for multiple cascaded networks

Ftot
= F 1+
F 2 -1
GAl
F 3 -1
Fk-1
+ ... +
GA1GA2
+ ...
(H.40)
GAIGA2 .. GA (k - I)
The preceding considerations have important practical implications. For instance, if

two amplifier stages with different gains and noise figures (F 1, G A 1 and F 2, G A 2 ) are
to be cascaded, which sequence of these stages results in the lowest noise figure? To
answer this question, let us hypothetically assume amlifier block 1 (F 1, G A 1 ) is followed by amplifier block 2 ( F 2 , GA 2 ). The total noise figure for this configuration is
F 2 -1
F tot ( I , 2) = F 1 + G
(H.41)
AI
On the other hand, if block 2 is followed by block I , we obtain
F 1- I
F 101(2, 1) = F 2 + G
(H.42)
A2
Under the assumption that F 101 ( 1, 2) has a lower noise figure than F 101(2, I) , the following inequality must hold:
(H.43)
Rewriting (H.43)
(F 1 -
1
1
t)(t--- ) <(F2 - 1)( 1--- )
GA2
(H.44)
GAI
allows us to define
NM 1 < NM2
-1
(H.45)
-1
where NM 1 = (F 1 -l)/(1-GA 1 ) and NM 2 = (F2 - l)/(l-GA2 ) are the nmse

measures of amplifiers 1 and 2, respectively. In other words, it is a combination of
noise figure and gain that determines the noise measure as a basis of an overall noise
performance comparison.
APPENDIX
Introduction to MATLAB
considerable number of MATLAB simulations

have been created to enable the reader to reproduce the results presented in the examples. Moreover, it is hoped that these so-called M-files will stimulate and encourage the
reader to develop code on his or her own relative to the RF topics covered in the ten
chapters. This appendix is neither a tutorial of MATLAB nor a detailed discussion of the
software written in support of this textbook. Rather, it is hoped that sufficient background is provided to understand how MATLAB routines are created, and how code can
be written to reproduce some of the results and graphs presented in the text. Being a general-purpose mathematical spreadsheet tool, MATLAB does not replace specifically
developed RF and MW CAD programs, such as MMICAD and ADS, with their powerful circuit analysis, optimization, and even layout utilities. Such dedicated simulation
packages cannot be expected to be available to the general reader. For this reason, the
authors have attempted to use MATLAB as a package that is widely available to students
and at very reasonable cost. For more information regarding the use of MATLAB the
reader should refer to the following Web site: http://www.mathworks.com.
This appendix first provides some general background as to how we created the
M-files, followed by a brief example of how they are used in the context of a stability
analysis, as done in Chapter 9. All M-files can be downloaded from our Web site:
http://www. wpi.edu/ece/EM_RF_LAB/book.
1.1
Background
MATLAB is an easy-to-use mathematical spreadsheet with the capabilities to write

special routines for mathematically evaluating the equations discussed in the main text
and to display the results graphically. The authors have MATLAB implemented on a PC
with a 450 MHz Pentium II processor, 128MB RAM and 8GB disk space. This does
not constitute a mimimal configuration; it merely reflects the environment used during
the writing of this textbook.
631
Appendix I Introduction to MATLAB
632
Upon executing MATLAB, a window is opened with a command line indicator>>.

The appropriate directory can be checked with the command pwd, which yields
>>pwd
ans =
d: \ RF\s imulations
indicating that the directory is located on d-drive under subdirectories RF\simulations.

Changing to a different directory can be initiated througth the command cd, and a listing of the files within a directory is done with the commands ls or dir.
By way of an example from Chapter 2, let us consider the following command
lines, which can be executed sequentially, each line terminated by pressing ENTER.
I=5
a=O.OOS
N=lOO
M=lO
r=(O:N) / N*(M*a)
for k=l:N+l
if(r(k)<=a)
H(k)=I*r(k)/(2*pi*a*a)
else
H(k)=I/(2*pi*r(k))
end
end
plot(r*lOOO,H, 'k')
In the first line of the program we specify a current through the wire. The second
line defines the wire radius. Variables Nand M specify the number of points and the maximum distance from the center of the wire at which the magnetic field will be computed.
In our case M=lO, which means that we will look at distances ranging from 0 to 10 wire
radii, and the number of points is set to N=lOO. The fifth line of the code defines a onedimentional array of points that determine the actual position from the center of the
wire. The command (O:N) creates an array of N + 1 elements with values of 0,1,2,3, etc.
After dividing this array by N, the values range between o to 1. Next, the array is scaled
so that the distance changes from o to M*a. An alternative way to define this array would
be r =(O:M*a/N:M*a), where the parameter between the two colons defines the step size.
The next line of the code starts a for loop cycle for k ranging from 1 to N+ 1. For
each k we take the corresponding radius from the array r and check whether it is less or
greater than the wire radius. As discussed in Chapter 2, the field inside the wire is linear
H=
Ir
27ta
633
with respect to the radial distance, whereas outside the wire we observe
I
H=-
21tr
The last line of the code instructs the program to plot a graph of the magnetic field
H versus radius r . The graph is shown in black, which is specified in the last parameter
of the plot instruction. Some of the possible choices for color include 'k' -black, 'r'red, 'y' -yellow, 'b' -blue, and 'g' -green. Other usefull options for creating graphs
include the following:
semilogx-logarithmic scale along x-axis, linear scale on y-axis
semilogy-logarithmic scale along y-axis, linear scale on x-axis
loglog- logarithmic scale on both axes
polar-polar plot
The entire list of commands can be entered in an interactive mode by using the
command line. Alternatively, the commands can be placed in a file for batch-mode execution. For example, we can save this program in a file with name field.m; then to execute this program we simply type >>field on the MATLAB command line. Note that .m
is a file extension resevered for use by MATLAB.
1.2
Brief Example of Stability Evaluation
Another useful capability of MATLAB is the creation of functions. For example,

the following listing is a function that takes an array of S-parameter data (s_param) and
computes two output parameters: the stability factor k and 1~1, denoted asK and delta.
function [K,delta] = K_factor(s_param)
% Usage: [K,delta) = K_factor(s_param)
%
% Purpose: returns K factor for a given a- parameter matrix
% if K>l and delta<l then circuit is unconditionally stable
% otherwise circuit might be unstable
sll=s_param(l,l);
s12=s__param{1 , 2);
s21=s_param(2,1);
s22=s_param(2,2);
delta=abs(det(s_param));
K=(l-abs(sll). A2-abs(s22).A2+del ta. A2). / (2*abs(sl2.*s21));
634
The first line in the listing defines a function K_factor that takes one input parameter s_param and returns two values as a result: K and delta, which are computed inside
the function. Unlike program scripts, files containing functions must have the same
name as the function name. Therefore, this function is stored in the file K factor .m.
If the user does not know or forgot how to use the function, he or she can type
help K_factor in the command line of MATLAB, and the comments that follow the first
line in the function will be displayed.
The program file for creating the S-parameter matrix of a particular transistor and
the stability check as well as the display of the stability circles is shown in the next file,
entitled test.m
% a-parameters for hypothetical transistor
close all;
sll=0.7*exp(j*(-70)/160*pi);
s12=0.2*exp(j*(-10)/180*pi);
s21=5.S*exp(j*(+85)/180*pi);
s22=0.7*exp(j*(-45)/180*pi);
s_param={sll,sl2;s2l,s22];
check stability
[K,delta] = K_factor(s_param)
% create a Smith Chart

smith_chart;
% plot input and output stability circles
input_stability(s_param, 'r');
output_stability(s_param, 'b');
% create
PostScript copy of the figure

print -deps 'fig9_8.eps'
This file is not a function; it is a collection of commands (program script) and
therefore can have any name. In our case we use the name test .m.
We notice that the S-parameters are given in magnitude and phase and stored in an
array called s_param. Next, a stability check is performed by passing on the s_param
array into theM-file K_factor.m, whose task is to find the stability factor and IAI based
on equations (9.24) and (9.29). After this we call three user-defined functions:
amith_chart~reates a figure containing the Z-Smith Chart.
input_stability-draws the input stability circles computed from the supplied Sparameters. Circles are drawn in the currently active figure (Smith Chart) and use
a specified color (red in our case).
output_stability-draws the output stability circles in the currently active
figure.
The last line of the script creates a file called fig9_8.eps, which contains the figure
stored in PostScript format. This is the format employed to produce most of the simulation results throughout the book.
1.3
Simulation Software on Compact Disk

1.3.1
Overview
The intent of the software contained on this CD is to provide support for the material covered in the textbook. All programs have been developed and tested using MATLAB Versjon 5.2. Although the authors believe that all routines should be compatible
with earlier versions of MATLAB, this may not be the case. The software is maintained
and regularly updated through our Web-site at www.wpi.edu/ece!EM_RF_lab/book.
1.3.2
Software Installation
The installation of the RF software involves the following steps:

1. Copy the entire directory rf_matlab onto the desired harddrive location.
2. Invoke MATLAB.
3. At the command prompt in the main MATLAB window type: cd
c: \rf_matlab, (here it is assumed that all files are copied to disc-drive C)
4. At the MATLAB command line type set__path. This will add all necessary paths
to the search tree of the MATLAB. If you do not wish to save this infonnation for
future use you can stop the installation process now. In this case you will be able
to run all programs but all path information will be deleted after closing MATLAB.
If you decide to store the path information for future use, continue with the next
step.
5. In the MATLAB command window go to the file\set path option. This will
launch the path browser.
6. In the path browser go to the file\save path option shown the main window.
7. At this step all path information is stored and you can begin to run theM-tile routines from the command line.
636
1.3.3
File Organization
All files are organized in the directory structure shown below and the content of
each folder is described in the table.
Wt
RF_matlab
j.... flil ch01
~....(iii ch02
i.. fil ch03
lfil ch04
:....Qj ch05
Folder name
Description
RF _matlab
Root directory
chOI-chlO
Selected examples and figures for chapters

1 through 10.
tools
Common files for simulations
amplifiers
Programs for computation of stability factor and simultaneous complex-conjugate

matching for the bilateral design
circles
Various circle equations
gam
Constant gain circles
nmse
Constant noise circles
quality
Constant Qn circles
stability
Stability circles
conversiOn
Conversion routines between different twoport network representations
global
Some useful routines for the computation

of the input and output reflection coefficients, VSWR, etc.
networks
Routines for the definition of matching network circuit topologies
smith
Programs related to the construction and

plotting of various arcs in the Smith Chart
!flit
ch06
ch07
[....(ij ch08
J.... alJ ch09
j.... ijQ ch10
EJ . UiJ tools
j. . !iiJ amplifiers
~.. (ij circles
. ;.... {U) gain
fU noise
~....GiiJ quality
L... iiJ stability
[....GiJ conversion
j....ifi global
!. . W)
!. .
!. . i.J
networks
L...{ij smith
Additional infonnation for each of the programs can be obtained by executing the
command help <program_name>, where <program_name> is the name of the
particular m-file. For example, to obtain help about the program smith_chart .m,
you execute the command help smith_chart in MATLAB's main window.
Further Reading
Student Edition of MATLAB, The MathWorks, Inc., 1995.
Index
A
ABCD-matrix, 148, 194,244--45
ABCD network representation, 156-{)1
of an impedance element, 157-58
of aT-network, 158-59
of a transmission line section, 159-61
Active biasing networks, 449
design of, for BJT in common-emitter
configuration, 452-53
Active RF component modeling, 351~04
diode models, 3.:52- 57
linear diode model, 354-57
nonlinear diode model, 352-54
measurements, 385- 97
of AC parameters of bipolar transistors, 387- 92
of DC parameters of bipolar transistors, 385-87
of field effect transistor parameters,
392-93
scattering parameter device charactedzlltion, 393-97
transistor models, 357-85
large-signal BJT models, 357-66
large-signal FET models, 378- !! I
small-signal BJT models, 366- 78
small-signal FET models, 382-85
Active RF components, 271-350
bipolar-junction transistor (BJT). 312- 28
brief history ot; 3 12
construction of, 312-14
frequency response, 321-23
functionalily of, 314-21
limiting values, 327-28
temperature behavior, 323-27
lypd of. 312
high electron mobility transistors,
338~3
computation of HEM1'-related electri~: characteristics, 341--42

construction of, 339
frequency response, 343
functionaUty of, 339~2
pseudomorphic: HEMTs (pHEMTs),
339
RF diodes, 293-311
BARRITT diodes. 311
Gunn diodes, 311
fMPATT diodes, 304-7
PIN diodes, 296-301
Schollky diode, 293-96

TRAPATI diodes, 31 1
tunnel diodes, 307-10
varactor diodes, 302--4
RF field effccl transistors, 328-38
fr<:quency response, 337
functiorutlily, 331-37
semiconductors, 272-93
See also Active RF component modeling;
Bipolar-junction transistor (BJT): Field
effect transistors
Admittance matrix, 145
Admittance Smith Chart, See Y-Smith Chv.rt
Admittance transformation. 121-26

parametric admittance equation, 121- 25
Y-Smith Chart, 124-25
Z-Smith Chart, 121- 24
ZY-Smith Chart, 125-26
Admittance/Y-matrix, 145--48
ADS, 351
AJm, 598
American Wire Gauge (AWG):
chan,599-X>
system, 13
Ampere, 598
Ampere's law, 47,48-51, 59,61-62
in differential (point) form, 48-49

Amplifiers:
balance<! amplifier design, 515-17
broadband, S 11- 22
characteristics of, 464
class A. 444
class AB, 445
class B.445
class C, 446
classes of operation, 444--49
dynamic range, 523-24
efficiency computation, 44~9
gain compression, 523
high-power, 522- 26
intennodulation distortion (IMD),
524-25
multistage, 526-29
power amplifier (PA), 4-6
power relations, 465-70
RF source, 46:5--66
RF transistor amplifier designs, 463- 538
AT41410 BJT (Hewlett-Packard), 512-13
ATFI3100 (Hewleii-Packard), 560
Attenuation behavior vs. normalized frequency,223
Available power, 466
Available power gain, 468, 501-2
Avalanche breakdown, 327
Avalanche effect, 304
B
Balanced amplifier design, 515--17
BBndgt~p energy,
353
Bandpass filters, 201,210-17

with coupled line tran~mision line segments, 261-62
frequency transformation, 235-37

Bandstop filters, 201,210-17,249
response, 212- 13
Bandwidth, 204
noise, 620
Bandwidth factor, 249
Barkhausen criterion, 540-41, 552
BARR!lT diodes, 311
Base-width modulation, 367
BFG403W BJT, 520-22
BFG540 BJT, 527- 28
BFQ65 BJT, 557
Biasing networks, 444-55
ampIifter classe~ of operation and,
reverse active mode, 314-15,319-20

saturation mode, 315, 32Q-21
S-parameters. computing for. 372- 78
stabilization of, 481-83
temperature behavior, 323- 27
types of, 312
Bipolar transistor biasing networks, 449-54
BIT, See Bipolar-junction transistor (Bin
Black bolt methodology, 143
Blocking capacitors, 298-99
Boltzmann's constant. 597

Branches, 178
Branch line coupler. 616-18
Breakdowns, 327-28
Broadband amplifiers, 511-22
balanced amplifier design, 515-17
design of, using frequency compensated
matching networks, 512- IS
negative feedback circuits, 517-22
parameters of, 5 15
Butterworth filters, 203, 220, 221- 24
coefficients for, 223
compared to linear phasse Butterworth
and Chebyshev filters, 229-31
c
Calibration procedure. 190
Capacitance, 598
Capacitors:
chip, 25-26
high-frequency, 17-21
RF impedance response of, 18-20
Carbon-composite resistor&, 14
Cascading networks. 155-56
Cauer low-pass filters, 203
Cellulllr phon~s. 3
Center frequency, 202
Chain/ABCD-matri1(, 148
Chain scattering matrix, 175-77
Channellen~th modulation parameter, 335
Characteristic line impedance, 63
Charge:
electric, 598
electron. :597
eJtcess charge carrier*, 609
OLebyshev bandpass filter design, 239~1
Chebyshev filters, 203.220,224-31
coefficients for, 228-29
compared to Butterworth and linear
phasse Butterworth filters, 229- 31
Chebyshev polynomials, 2.03, 224-25
Chip capacitors, 25-26
444-49
bipolar transistor biasing networks,
449-54
.field effect transistor biasjng nelwor.ks..
455
Bilateral matching, 508-9
Binomial low-pass filters, 203
Bipolar-junction transistor (BIT), 149- 52,
312-28
bias conditions, setting for, 372-78
brief history of, 312
forward active mode, 317-19
frequency response, 321-23
functionality of. 314-21
GaAs field effect transistors (GaAs
FETs}, 3!2-14
high electron mobility transistol1)
(HEMTs), 3 12
input/output impedances, determining
for, 372- 78
internal resistances/current gain, J 5 I - 52
low-frequency hybrid network, description of, l49- 5J
parameter nomenclature, 316
parameters, 373
637
Chip resistors, 24--25

Circle equation, 604
Clapp osc.ill!tor. 545
Cla" A amplifier, 444
Class AB amplifier, 445
Class B amplifier, 445
Class C amplifier, 446
Coaxial cable, 41-42
Colpitts oscillator, 545--46
design, 547-49
Commensurate line length, 243
Complex numbers, 603-4
basic definition. 603
circle equation. 604
magnitude computations, 603~
Complex propagation constant, 62, 71-72
Compressed Smith Charts, 109
Computer-aided design (CAD), 119-21. 35 I
ConducWICe, 598
Conduction angle, 444
Conduction parametef, 380
Conductiviry, 598
thermal. 608
Conformal mapping, 65
Constant current ftow in a conductor, magnetic field generated by, 50-52
Index
638
Constant gain, 483- 502

bilateral design, 492-95
operating and available power gain circles, 49.5- 502
unilateral design, 483-92
applicability test, 491-92
unilateral figure of merit, 490-92
Constant gain circles, derivation of, 485--86
Constant VSWR circles, 506-11
design for given gain/noise figure,
509-11
Conversion compression, mixers, 582
Conversion gain (CO), mixers, 581
Conversion loss (CL), mixers, 581
Correlated noise sources, 622
Coulomb, 597, 598
Coupled filter, 253- 62
bandpass filter section, 257-58
cascading bandpass filter elements,
259-60
design example, 260--62
odd/even mode excitation, 2.54-56
Couplers, 612- 19
branch line coupler, 616-18
dual-directional, 394-95
Lange coupler, 619
Wilkinson power divider, 6 I 2-15
Coupllng,612
Coupling coefficient, 564
Coupling factor, 395
Current, 598
Cut-off frequency, 201, 202,321
Cylindrical conductor, skin equation for,
601-2
D
Device under test (OUT), 168, 189, 195, 393,
395-96
Dielectric constant, 608
Dielectric resonator oscillators. 563~8
design of, 565-68
Diffusion length, 610
Digital-to-analog converter (DAC), 4
Dimensions, 6-8
Diode models, 352- 57
linear diode model, 354-57
nonlinear diode model, 352-54
Diodes:
Gunn, 311
IMPATT, 304-7
leakage, 354-55
PIN, 296-301
RF, 293-311
Schottky, 293- 96
TRAPATT, 311
tunnel, 307-10
vamctor, 302-4
Dipole domains, 573
Direct current (DC), transition from high-frequency modes of operation to, 1-6
Directivity factor, 395
Dispersion-free transmission, 72
Dissipation factor d, 213
Double-balanced mixer, 590
Double-stub matching networks, 440-44
design of, 442-44
Downconverted frequency components, 578
Drain saturation voltage, 332
Dual-directional coupler, 394-95
Dual-port network, ftowchart analysis of,
181-84
Dynamic Ebers-Moll chip-model, 359-61
Dynamic range:
amplifiers, 523-24
mixers, 583
spuriousfree,525- 26
E
Early effect, 363
Ebers-Moll BJT model, 351, 357-66, 387,
398
dynamic Ebers-Moll chip-model, 359-61
large-signal, transport vs. injection fonn
of,361~
popularity/acceptance of, 363

Edison, Thomas, 2
Effective dielectric constant, 65
Effective width, 69
Electrical length, 110
Electric charge, 598
Electric field, 598
Electron affinity, 608
Electron ch3!'-e, S97
Electron mob1lity, 608
Electron rest mass, 597
Electron volt, 597
Elliptic low-pass filters, 203
Emission coefficient, 352
Emitter efficiency, 313
Energy, 598
Energy gap, 608
Equivalent circuit representation, 45-47
Equivalent noise resistance, 502
Equivalent series resistance (ESR), 20
Excess charge carriers, 609
Excess minority carrier lifetime. 297
Excess noise, 622
External Q, 206
F
Farad, 598
Faraday's induction law, 2
Faraday's law, 47, 51-53,59
Feedback loop, 180
Feedback oscillator design, 543-46
Femto,598
Field effect transistor biasing networks, 455
Field effect transistors (FETs), 328- 38
functionality, 331-37
hetero FET, 329
junction FET (JFET), 329
metal insulator semiconductor FET
(MISFET), 329
metal oxide semiconductor FET (MOSFET), 329
MEtal Semiconductor FET (MESFET),
329
Filter implementation, 241-53
Kuroda's identities, 243-45
unit elements (UEs), 243
Filter Q, 206
Filters:
bandpass,201 , 210-17,261~2
binomial low-pass, 203

Butterworth, 203, 220, 221-24, 229-31
Cauer low-pass, 203
Chebyshev,203,220,224-31
coupled, 253~2
elliptic low-pass, 203
high-pass, 201, 209-10, 233-35
image, 580
linear phase low-pass, 224
low-pass, 201 , 203,206-9,233
RF filter design, 201~9
Fixed-frequency oscillators, .556-63
design of, 557-60
GaAs FET oscillator, microstrip design
of, 560-63
Flat coils, 27- 28
Flow charts, 178-84
branches, 178
nodes, 178
Forbidden rgions, 415- 17
Forward active mode, 317-19
Forward Early voltage, 365
Forward power gain, 172
Forward voltage gain, 172
Frequency, 598
center, 202
cut-off, 201
Frequency compensated matching networts,
512-15
Frequency domain considerations, 578-80
Frequency response, 417- 22
bipolar-junction transistor (BIT), 321-23
field effect transistors (FETs), 337

high electron mobility transistors
(HEMTs), 343
input reftection coefficient, 419
matching networks, 417-22, 426
RF field effect transistors, 337
Wilkinson power divider, 613
Frequency spectrum, 8--9
Frequency transformation, 232-38
bandpass filters, 23>-37
bandstop filters, 237-38
G
GaAs field effect transistors (GaAs FETs),
312-14
GaAs MESFET, 379
determination of cut-off frequency of,
384-85
Gain-bandwidth product, 372
Gain compression, 523
Generic RF system, block diagram of, 3
Giga, 598
Global po.sitioning systems (GPS), I, 2
Gradual-channel approximaiton, 332
Gummei-Poon BJT model, 351, 357, 36~5,
387, 398
Gunn diodes, 311
Gunn element oscillator, 573- 74
H
Harmonic disconions, .524
Hannonic JMD, mixers, 583
Hartley oscillator, 54.5-46
Henry, .598
Hertz, 598
Hertz, Heinrich, 2
Heterodyne receiver, 575
Hetero FET, 329
Heterojunction bipolar transistors (HOTs),
312-13
High electron mobility transistors (HEMTs).
312,338-43
computadon of HEMT-related electric
characteriscics, 341-42
construction of, 339
functionality of, 339-42
pseudomorphic HEMTs (pHEMTs), 339
High-frequency capacitors. 17-21
series loss tangent, 18
Hig~frequency inductors, 21-24
RF coils (RFCs), 21
High-frequency resistors, 14-17
surface mounted devices (SMOs), 14
High-pass filters, 201, 209-10
High-power amplifiers, 522-26
High-side injection, 578
H-matrix, 148-49, 194
Hole mobility, 608
Hybrid circuits, 27-28
Hybridlh-matrb., 148-49, 194
I
IEEE frequency spectrum, 8--9
1-layer, 296
Image filter, 580
Image frequencies mapping, 578- 80
Image impedance, 257
Impact ionization, 304
IMPATI diodes, 304-7
Impedance:
characteristic line, 63
general definition, 63~
input impedance matching, 90-91
intrinsic, 6, 7-8
normalized impedance equation, 104
wave, 64
Impedance matching, using discrete components, 406-31
Impedance matrix, 144--45
Impedance transformation, tt0-21, 232,
239-41
computer simulations, 119-21
Index
for general load. 11 0-12

special transformation conditions,
115-19
open-cirellit transformtioos, \ \ f>-17
shan-circuit transformations,
Il7-19
standing wave ratio (SWR). ll3-! 4
lmpcdance/Z-matrix, 144-45. 148
Induced voilage, in a sta!ionary wire loop,
52-53
Inductance, 598
Inductors:
high-frequency, 2 1- 24
surface-mounted. 26-28
Injection version, 3.58
lnpul impedance matching, 90--91
Input matching network (IMN), 4, 506
Input reflection coefficient, frequency
response of, 419
Input resistaoce, 367
Input stability circle equation. 471-72
Input VSWR, 506
ln&ertion loss (!L}, 91- 93. 2{)4, 217-20
Institute of 61ectricaf and Electronic Engineers (IEEE), 8-9
Intercept poiot (IP), 525
Interconnecting networks, 1~3-61
ABCD network represenations, 156-61
cascading networks, I 55-56
parallel connection of networks, 154- 55
series connection of networks, 153-54
Intermediate frequency (IF), 575
lntermodular distortion (!MD), 524
lntermodulation distortion (!MD):
amplifiers, 524-25
mixers, ~82
Internal noise source, 622
International System of Units, 598
1ntentage matching network. 4, 526
Intrinsic carrier concentration, 608
Intrinsic impedance, 6, 7-8
Intrinsic resistivity, 608
Inverse Early voltage. 365
Iso lation, 298,583, 612
J
Joule. 598
Junction FET (JFET). 329
Junction grading coefficient, 352
K
Kilo, 598
Kirchoff's voltage and circuli laws
(KVUKCL), 39-41, 45, 58-62
Kuroda's identities, 242, 243--45
L
Lambda-quarter transfonner. 8l
Lange coupler, 619
Large-scale diode model, 352
Large-signal BIT models, 357--66
Large-signal Ebers-Moll circuit model, 358
large-signal FET models, 378-81
FET-related benefits, 378-79
linear region, 31!0
noninsulated-gate FET. 379
reverse linear region, 381
reverse s:~turation region, 380
saturation region, 379-80
Large-signal S-parameters. 522- 23
Lead resistance, I 5
Leakage diodes, 354--55
Libra, 120
Linear active mode, 3 14
Linear diode model, 354-57
Linear phase behavior. 222
Une&f phase lQw-pass filter&, coefficients for.
224
Loaded Q, 205-6
Local oscillator (LO) frequency, 575
selection of, 577- 78
Long and short diode models, 609-11
Loop gain equation, 540
Loss factor, 218
639
Loss tangent, 599

Low- to high-frequency circuit operations,
evolution of, J- 36
L-ow-pass filters, 2(}1, '206-9

actual attenuation profiles for, 203
frequency transformation, 233
Low-side injection, 57?-18
L-section matching networks, 406-15
analytical approch to design of, 407-8
graphical approach to design of, 41Q- 12
M
Magnetic field, 598
Magnetic flux, 598
Magnitude computations, 603-4
Mapping;
conformal, 65
image frequencies, 578-80
Matched load reflection coefficient, 493
Matched source reftection coefficient, 492
Matching networks:
design approaches, 406
double-stub matching networks, 44()-44
forbidden regions, 415-17
ftequency compensated matching network~. 512- 15
frequency response, 417-22, 426
impedance matching using discrete components. 406-31
interStage matching network. 4, 526
with lumped and distributed components,
design of. 432-35
microstrip line matching networks,
431-44
narrow-band matching network, design
of, 424-26
quality factor. 422-26
series C, shunt L matching network, 414
series L, shunt L matching network, 414
shunt C, series L matching network, 414
shunt L, series L matchin& network, 414
single-stub matching networks, 435-40
T and Pi matching networks, 426-31
transfer function of, 219
two-component ma.lching networks,
406--15
MathCad, 134, 167
Mathematica, 134, 167
MATLAB, 120--21, 132, 134, 167,631-35
background,631-33
stability evaluation, brief example of,
633-35
Matrix conversions, 605-7
Maximally flat filters, See Butterworth fllten;
Maximum gain, amplifier design for, 494--95
Maximum power hyperbola. 327
Maximum power transfer, 466
Maxwell, James, '2
Mega,598
Mesa processing technology, 296
Metal-film resi$IOI'S, L4
Metal film resistors, RF impedance response
of, 16-17
Metal insulator semiconductor PET (MIS
FET), 329
Metal oxide semiconductor FET (MOSFET),
329
MEtal Semiconductor FET (MESFET), 329
drain saturation cur:rent in, 333-35
GaAs MESFET, determina!Wn of c ut-off
frequency of, 384-85
idealized MESFET device structure, 382
1-V characteristic of, 335-37
small-signal MESFE'T model, 382
SPICE modeling parameters for, 381
m.files, 134,631
Micro, 59S
Microstrip filter design, examples of, 245-52
Mlcrostrip line matching networks, 431-44
from discrete components to microstrip
lines, 431-35
Microstrip transmission lirtes, 42-44, 64-69
desigrt of, 67-68
Microwave (MW), 2
Mil, 598
Miller effect, 369-72
Milli, 598
Minimum detectable signal, 524
Minilllum noise figure, 502, 628
design of small-signal amplifier for,
S04-6
Minority carrier lifetime, 608
Mixers, 574-91
basic characteristics of, 574-90
basic concepts, 575-78
conversion compression, 582
conversion gain (CG). 581
conversion loss (CL ), 581
double-balanced mixer, 590
dynamic range, 583
frequency domain considerations.
578-80
httnnonlc IMD, 583
intennodular distonion (IMD), 582
local oscillator (LO) frequency, 575
selection of, 577-78
noise figure, 58\-82

single-balanced mixer, 588-89
single-ended mixer design, 580-88
BJI mixer, 5&4--S&
See also Oscillators
Mixing process. spectral representation of,
579
Modulation-doped field effect transistors
(MODFETs). See High electron mobility
transistors
Monolilhic and Microwave Integrated C ircuit
Analysis and Design (MMICAD), 120,
131, 134.252.351,410
Monoplar devices, 328
Multistage amplifiers, 526-29
transistor choices for desing of, 527-29
N
Nano, 598
Narrow-band mMching network, design of,
424--26
N-channel. 328
Negative filedb&dr., defined, Sl 7
Negative feedback circuits, 517-22
Negative feedback loop broadband amplifier,
design of, 520-22
Negative resistance oscillator, 541-43
Nodal quality factor, 422-24
Nodes, 178
Noise analysis, 620--30
basic definitions, 620--22
correlated noise sources, 622
excess noise, 622
internal noise source, 622
noise bandwidth, 620
noise figure:
for cascaded multipart netwrok,
629- 30
fOt IW<>-port network., 625-29
noise power, 620-21
noisy two-pon networks, 623- 25
l/fnoise, 622
random variables, 620
root mean square (RMS), 620
shot noise, 622
spectral density, 622
spectral noise current, 62 I
spectral noise voltage, 621
uncon-elated noise sources, 622
white noise, 622
Noise figure, ~02
for cascaded multipart network, 629-30
equivalent noise resistance, 502
minimum, 502, 628
design of small-signal amplifier for,
504-6
mixers, 581-82
OJ)(imum rd'\ec\ion coefficient, 503
optimum source admittance, 503
for two-port network, 625-29
Noise figure c:lrcles, 502-6
Noise measures, 630
640
Noisy two-port networks, 623-25

Noninsulated-gate FET. 379-81
Nonlinear diode model, 352-54
Normalized effective mass of the electron,
608
Normalized impedance equation, 104
Ohm, 598
I dB compression point, 523

lifnoise. 622
Open-circuit transformations, 116-17
Open-circuit transmission, 79--80
input impedance of, as a function of frequency, SO
Operating point. 295
Operating power gain, 468,496-501
circle derivation, 497-98
constant operating power gain circles,
amplifier design using, 498-500
Optimum reflection coefficient, 503
Optimum source admittance, 503
Oscillations, adding positive feedback element to iniliate, 554-56
Oscillators, 479, 539-74
basic model, 540-52
Clapp os.cillator, 545
Colpitts oscillator, 545
design, 547-49
design steps, 546-49
dielectric resonator oscillators, 563-{)8
design of, 565--68
feedback oscillator design, 543-46
fixed-frequency oscillators, 556--63
Gunn element oscillator, 573- 74
H11rtley usci llatur, 545
high-frequency oscillator configuration,
552-74
negative resistance oscillator, 541-43
output vollll.ge vs. gain characteristic, 541
quartz oscillators. 550-52
varactor diode oscillator, 570-73
design of, 572-73
voltage-controlled oscillator, 570-73
YIG (yittrium iron garnet) oscillator,
569- 70
Su also Mixers
Output characteristic. 332
Output conductance, 367
Output matching network (OMN), 506
Output stability circle equation, 471
Output VSWR, 506
p
Parallel connection of networks, 154-55

Parallel-plate transmission line:
drucit parameters for, 53-57
derivation of equations, 59--62
line paramters of, 56-57
Parametric admittance equation, 121-25
Parametric reflection coefficient equation,
106--8
Passive networks, 449
design of, for BJT in common-emitter
configuration, 450-51
P-channel, 328
Permeability in vacuum, 5'17
Pennittivity in vacuum, 597
Phase velocity, 7-8
Physical constants, 597
Pico, 598
Pinch-off voltage, 333
PIN diodes, 296-301
computation of transducer loss of,
300-301
Pi-network, matrix representation of, 146-48
Pi-type matching network, 426-31
design of, 429-31
Planar printed circuit board (PCBs), 42-43
Power, 598
Power amplifier (PA), 4--6
circuit, knowledge requred to build, 5--6
Power derating curve, 324
Power gain. 468
Index
available, 468.501-2
forward. 172
operating,468,496-501
reverse, 173
transducer, 466--67
unilateral, 467
Printed circuit boards (PCBs), 10, 42-43
Pseudomorphic HEMTs (pHEMTs}, 339
Puck,563
Quality factor Q. 24, 205

Quarter-wave transmission line. 80--84
Quartz crystal:
and piezoelectric effect, 550
prediction of resonance frequencies of,
551-52
Quartz oscillators, 550-S2
Quiescent point, 295
R
Radio frequency coils (RFCs), 5, 298
Radio frequency design, importance of, 2--6
Radio frequency (RF), 2
Random variables, 620
Receiver, 578
Reflection coefficient:
input, 419
matched load, 493
matched source, 492
normalized impedance equation, 104
optimum, 503
parametric reflection coefficient equation, 106-8
in phasor form, 102-3
representations, 102-3
simultaneous conjusate matched, 493-94
voltage, 69-71
Rejection. 204
Resistance, 598
Resistivity, 598
Resistors, 14
chip. 24-25
high-frequency, 14-17
metal film, 16-17
Return loss (RL), 91-93,612
Reverse active mode, 314-15, 319-20
Reverse bias, 298
Reverse collector and emitter saturation currents, 359
Reverse power gain, 173
Reverse saturation current temperature coefficient, 353
Reverse voltage gain, 173
RF amplitier, power relations of, 468--70
RF behavior of passive oomponents, 10-24
high-frequency capacitors, 17-21
high-frequency inductors, 21-24
high-frequency resistors, 14-17
RF coils (RFCs), 21
RF impedance response of, 22-24
RF diodes, 293-311
BARRIIT diodes, 311
Gunn diodes, 311
IMPAIT diodes, 304-7
PIN diodes, 296-301
Schottky diode, 293-96
TRAPATI diodes, 311
tunneld~s.307-IO
varactor diodes, 302-4

Rf filter design, 20 1--69
bandpass filters, 201,210-17
bandstop filters, 201,210--17, 249
response, 212-13
basic resonator/filler configurations,
202-20
coupled filter, 253-62
bandpass filter section, 257-58
cascading bandpass filter elements,
259--60
design example, 260--62
odd/even mode excitation, 254-56
denonnalization of standard low-pass
design, 231-41

impedance transformation, 232,
239-41
filler implementation. 241-53
Kuroda's identities. 242,243-45
unit elements, 243
filter types/parameters, 202-6
high-pass filters, 201, 209-10
insertion loss, 217-20
low-pass filters, 201, 206-9
acrual attenuation profiles for. 203
microstrip filter design, examples of,
245-52
special filter realizations, 220-41
Butterworth filters, 220, 221-24
Chebyshev filters, 220, 224-31
RF filters, calculation of quality factors for,
219-20
RF impedance response:
of capacitors, 18-20
of metal film resistors, 16- I 7
of RFCs. 22- 24
RF transistor amplifier designs, 463-538
amplifiers:
characteristics of, 464
power relations. 465-70
broadband amplifiers, 511-22
constant gain, 483-502
bilateral design, 492-95
operating and available power gain
circles, 495-502
unilateral design, 483-92
constant VSWR circles, 506-1 I
high-power amplifiers, 522-26
multistage amplifiers. 526-29
noise figure circles, 502--6
power gain (operating power gain), 468
stability, 470--83
circles, 4 70-73
stabilization mehtods, 479-83
unconditional, 473-79
transducer power gain, 466--67
Richardson constant, 294
Richards transformation, 242-43, 263
Ripple, 204
Rollett stability factor, 474-76, 553-55, 557
Root mean square (RMS). 620
Safe operating area (SOAR), 327

Saturation mode, 315,320-21
Saturation voltage, 314
Scattering parameter device characterization,
393-97
Scattering parameters (S-parameters), 168-95
chain scattering matrix, 175-77
conversion between Z-parameters and,
177- 78
definition of, 168--71
generalization of, 184-88
meaning of, 171-75
practical measurements of, 188--93
signal flow chart modeling, 178-84
Schottky diode, 293-96
Second-order intermodular proudcts, 577
Self-biased networks, 449
Self-loop, 180-82
Semiconductors, physical parameters. of. 608
Series connection of networks, 153-54
Series C, shunt L matching network, 414
Series loss tangent, IS
Series L, shunt L matching network, 414
Shape factor, 204
Short-circuit transfonnations, 117-19
Short-circuit transmission line, 76-78
input imedance of, as a funciton of frequency, 77-78
Shot noise, 622
Shunt C, series L matching network, 414
Shunt L, series L matching network, 414
Siemens, 598
Siemens, Werner, 2
Signal How chart:
building blocks, 181
Index
input impedance com(mlat\on of a tre.nr.mission line bBsed on use of, 187-88

modeling, 178-84
Simultane<:>us oonjugate match, 492
Simultaneous conjugate matehed reflection
coefficients, derivation of, 493-94
Sittgle-balaru:ed mixer, 588-89
Single-ended mixer design, 58~88
SJT mixer, 584-88
Single-lmuJti-pon networks, 143-99
;1dmittance/Y-mattix, 145-48
basic definitions. 144--52
chain!ABCD-mlltrix, 148
hybrid/h-matrix, 148--49
lmpedaoce/Z-matrix, 144-45, 148
interconnecting nenvorks, 153-61
ABCD nttwork rept~enations,
156-61
cascading networks, 155-56
pll!'a\\el COI\neel\()n of network&,
154-5.5
serie~; connection of networks,
153- 54
network properties/applications, 161-68
analysis of microwave amplifier,
164-08
interrelaLions between parameter
sets, 161-64
scauering parnmeter.~ ($-parameters),
168-95
See alscJ Scattering parameters ($-parameters)
Single-stub matching networks, 4'.)5-40
with fixed chaJliCCeritic imp~ances,
design of, 43b-38
using transmission lines with characteristic impedance.~. de~;ign of, 4JS- J9
Skin depth, 11- 13,54-57
S\l.in equa\ion f<>r cyHndrical conductor,
601-2
Small-signal BJT models, 366-78
Small-s(gnal current gain, 367
Small-signal diode model, 355
derivation of. 355- 57
Srnttllsigoal equi~alent circuit model, 294
Small-signal FET models, 382.-85
Smith, P. H.. l 0 I
Smith Chan, 101-42.4-36, 437, 44.Z, 443,
463,464,487- 88.490,505,513-\4
admittance transformation, I Z1-26
parametric admittJmce equation.
121 -25
Y-Srniclt Chart, I 24-25
Z-Smith Chart, 121-24
ZY-Smith Chart, 125-16
compressed, 109
design of matching network using, 413
impeoan~ traMformation, lJ()-21, 418,
4.25
computer simulations, 1!9-2.1
(or general load, I 10-12
special transformation conditions,
115-19
standing wave ratio (SWR). 113-14
key ingredient. I 02
one-to-one mapping, l09
parallel cormecllons, !26-27
of R and C elements, 12'7
of K and L eleme1\ts, 127
reflection coefficient, 102.-8
in phasor form, 101-3
series oonn<Xcions, I 27-33
of R and C elements, 129-30
of Rand I.. elements, 121-29
T-type network, 130...33
Z-Smith Chart, 121
Sourced a.ttd loaded transmission line, 84--93
input impedlloce matching, 9o-9l
phasor representation of source, 85-87
power considcracions for, 87-90
return loss and insertion loss, 91-93
Source-field relations, 47-48
841
Source gain circles, computation of, for unilateral design, 486-88

Specific heat, 608
Spectral density, 622
Spectral noise current, 621
Spectral noise voltage, 621
Speed of light in Yacuum, 597
SPICE, 351, 354,365
Spurious free dynamic range, 525-26
StabjJiJy, 470-83
circles, 4 7()...-'B
for a BJT at different operating frequencies, 477-78
input, 41\- 72
OUlpU!,
471
RoUett stabilitv factor, 474-76, 5.S3- 55,

557
stabilily factor deri vJttion, 476
stabiliution of a BJT, 481-83
stabilitation mehtOd5, 479-83
stable vs. unstable region of a transistgor,
479
UO{;octdicionsl, 473-19
Standing wave ratio (SWR), 74-'75, l I1-\4
voltage (VSWR), 75
Standing wave.~. 72-75
Steinmet:z, Charles, 2
Stopband width, 250
Surface-mounted devices (SMDs), 14
Surface-mounled indu~\Ol'S, 26-l&
"Switch off," 298
1'
Terminated lossleSG transmission line, 69-75

input impedance of, 7'5-76, 105
opeo-c:ircuit traosmission, 79-80

prop_aaal.ion constant and phase velocity,
71-72
quarter-wave transmission line, 80-84
short <::ircuit transmission line, 76- 78
special termination conditions, 75-84
standing waves. 72~15
voltage reflection coeffient, 69- 7 I
Te~~Ja, Nikolas, 2
Thermal analy~>iS, involving BJT moumed on
heat sink, '32&-21
Thermal conductivity, 608
Thermal diffusivity, 608
ihin-film ch\p rc~>islors, l4
Third-o!'der intermodular producu. 577
Threshold voltage, J33
Througb-Rellect-LiCte (TRL), 191-93, 195
T matchingnetW()(k, 42>-31
design of, 427-29
T-network .. 130-33, 134
derennination of e.Jements, 173- 74
Transceiver, 3
Transconductance, 367
Transducer power gain, 466-67
Transfer function, 208
Transistor models, 35?--85
large-signal BlT models, 351--66
large-signal FET models, 378-81
small-signs! BJT models, 366-78
srna\1-signn\ FET models, 'S2-SS
Transition frequency, 32 I
Transition time, 353
Transmission coefficient, <letennination of, 86
Transmission line anll!ysis, 37-l 00
Transmission line&:
c.Qar.ial cable, 41-42
terminated lossless ttansmission Jine,

69-75
two-wire Jines, 41
TransnUssion line theory, 37--41
Transrniner, 5'1?.
Transverse electric (TE} field W!lve propagation, 7,44
Transverse electromagnetic mode (TE.M), 6-'1
TransveJ:Se magnetic (TM} field wave propagation, 7, 44
TRAPATI diodes, 311
Traveling voltage and CJJ.m:nf waves, 62-63
T-type matching network, design of, 427-29
Tunnel diodes, 307-10
Tunneling. 30 8
Two-component matching networks, 406-15
analytical approclt to design of, 40"1- 8
general, design of, 412-15
graphical approach to design of, 41 ~12
Two-wire transmission Jines, 4-1
u
Unc.onditional stability, 473-79
Uncortelated noise 30\ltcet., 621
Unilateral figure of merit, 490-92
Unilateral power gain. 467
Unit elements (UEs), 243
Units, 6-8
Unloaded Q, 564
Upconverted frequency components, 578
v
Varactor diode oscillator:
circuit analysis of, 570...71
design of, 572-73
Varactor diodes, 302-4
V/m, 598
Volta, A., 2
Voltage, 598
drain saturafaon \'Ohage, 332
forward Early voltage, 365
forward voltage sain. 172
induced, in a s\ationary wire loop. 52-53
inverse Early volcage, 365
Kirchoff's voltage law, 39-40. 45, 58-62
pincl\-Qf{ vot~age, 333
reflecrioB coefficient, 69-7 J
teverse voltage gain, 173
saturation voltage, 3 I4
spectral noise voltage. 6.21
voltage standing wave 111tio (VSWR), 75
Voltage-controlled osciUator, 57~73
Voll88e reflection coefficiem, 69-71
Voltage standing wave ratio (VSWR). 75
Volcaic cells, 2
Volts, 598
Wan,5%
Wave impedance, 64
Wavelength, 6, 7-8
Weber, 59S
White noise, 622

Wilkinson power divider, 612-15
frequency response of, 613
Wire diamecer, conversion between AWG size.
and. 13- 14
Wirele~s local area networks (WLANs). 3
Wire size sumd.erdizaJjon, J3- I 4
Wire-wound resistors, 14-15
WTG (wavelength toward generator) scale,
434
CQIIfigurations, 57
general equation, .SS-64

general impedance definition, 63-64
Kirchhoff voltage and current law
representations, 58-62
lo~less transffli$sion line model, 64
traveling voltage and current waves,
62-63
microstrip lines, 42-44
power considerations for, S/-9()
sourced and loaded transmission line,
84-93
y
YlG (yittriurn iron garnet) osci\\ator. SW-10
Y-mlltrix, I 45, 194
Y-Smilh Cbart. 124-25, 134
ZmatriJL, 144-45, 194

Z-Stnitl\ Chart, 121-24, l34
ZY-Smith ChArt, 125-26, 134, 409-JO

use of, 126
YOU SHOULD CAREFULLY READ THE FOLLOWING TERMS AND CONDITIONS BEFORE OPENING THIS CD-ROM PACKAGE. OPENING THIS CD-ROM PACKAGE INDICATES YOUR ACCEPTANCE OF TH ESE TERMS AND CONDITIONS. IF YOU 00 NOT AGREE WITH
THEM, YOU SHOULD PROMPTLY RETURN THE PACKAGE UNOPENED.
Prentice-Hall , Inc. provides this program and licenses its use. You assume responsibility for the selection of the
program to achieve your intended results, and for the installation, use, and results obtained from the program. This
license extends only to use of the program in the United States or countries in which the program is marketed by duly
authorized distributors.
LICENSE
You may:
a. use the program;
b. copy the program into any machine-readable form without limit;
c. modify the program and/or merge it into another program in support of your use of the program.
LlMITED WARRANTY
THE PROGRAM IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT
NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRB
RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE,
YOU (AND NOT PRENTICE-HALL, INC. OR ANY AUTHORIZED DISTRIBUTOR) ASSUME THE ENTIRE COST OF ALL NECESSARY
SERVICING, REPAIR, OR CORRECTION.
SOME STATES DO NOT ALLOW THE EXCLUSIO N OF IMPLIED WARRANT IES, SO THE ABOV E EXCL USION MAY NOT APPLY
TO YOU. THIS WARRANT Y GIVES YOU SPECIFIC LEGAL RIGHTS AND YOU MAY ALSO HAVE OTHER RIGHTS THAT VARY FROM
STATE TO STATE.
Prentice-Hall , Inc. does not warrant that the function s contained in the program will meet your requirements or
that the operation of the program will be uninterrupted or error free.
However, Prentice-Hall, Inc., warrants the CD-ROM(s) on which the program is furnished to be free from
defec ts in materials and workmanship under normal use for a period of ninety (90) days from the date of delivery to you
s evidenced by a copy of your receipt.
LIMITATIONS OF REMEDIES
Prentice-Hall 's entjre liability and your exclusive remedy shall be:
1. the replacement of any CD-ROM not meeting Prentice-Hall's "Limited Warranty" and that is returned to
Prentice-HaJJ, or
2. if Prentice-Hall is unable to deliver a replacement CD-ROM or cassette that is free of defects in materials or
workmanship, you may tenninate this Agreement by returning the program.
IN NO EVENT WILL PRENTICE-HALL BE LIABLE TO YOU FOR ANY DAMAGES, INCLUDING ANY LOST PROFITS, LOST SAVINGS,
OR OTHER INCIDENTAL OR CONSEQUENTlAL DAMAORS ARISING OUT OF THE US E OR INABILITY TO USE SUCH PROGRAM
BVBN IF PRENTICE-HALL, OR AN AUTHORIZED DISTRIBUTOR HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES, OR
FOR ANY CLAIM BY ANY OTHER PARTY.
SOME STATES DO NOT ALLOW THE LIMITATION OR EXCLUSION OF LIABILIT Y FOR INCIDENTAL OR CONSEQUENTIAL
DAMAGES, SO TH E ABOVE LIMITATION OR EXCLUSION MAY NOT APPLY TO YOU.
GENERAL
You may not sublicense, assign, or transfer the license or the program except as expressly provided in this Agreement.
Any attempt otherwise to sublicense, assign, or transfer any of the rights, duties, or obligations hereunder is void.
This Agreement will be governed by the laws of the State of New York.
Should you have any questions concerning this Agreement, you may contact Prentice-Hall, Inc., by writing to:
Prentice Hall
College Division
Upper Saddle River, NJ 07458
YOU ACKNOWLEDGE THAT YOU HAVE READ THIS AGREEMENT, UNDERSTAND IT, AND AGREE TO BE BOUND BY ITS TERMS
AND CONDITIONS. YOU FURTHER AGREE THAT IT IS THE COMPLETE AND EXCLUSIVE STATEMENT OF THE AGREEMENT
BETWEEN US THAT SUPERSEDES ANY PROPOSAL OR PRIOR AGREEMENT, ORAL OR WRITTEN, AND ANY OTHER COMMU NICATIONS BETWEEN US RELATING TO THE SUBJECT MATTER OF THIS AGREEMENT.
ISBN: 0-1 3-095323-7

642

Rfcircuitsdesigntheoryandapplicationsludwig 140506103907 Phpapp02

Uploaded by

Copyright:

Available Formats

Rfcircuitsdesigntheoryandapplicationsludwig 140506103907 Phpapp02

Uploaded by

Document Information

Original Title

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

Rfcircuitsdesigntheoryandapplicationsludwig 140506103907 Phpapp02

Uploaded by

Copyright:

Available Formats

What is the objective of the textbook?

What is the objective of the textbook?

What simulation software is discussed?

What simulation software is discussed?

RF Circuit Design

Theory and Applications

PRENTICE-HALL OF JAPAN, INc., Tokyo

design is receivsignificant industrial attention

'11.., .... "'......

This book is used in

4. Single- and Multi-Port Networks

Chapter 9. RF Transistor Amplifier Designs

Chapter 6, Active RF Components

apter 10, Oscillators and Mixers

the entire course u:rgamzat:lun

he provided the funding

students, and pntcti.CU:lLR

help. Linda Gu,

Lai, Joe Plunkett, Dr. Funan

and Josh Resnik

' V U ....... ..., . . .

Importance of Radio frequency Design

2.10.2 Short Circuit Transmission Line

4.4.2 Meaning of S-Parameters

Chapter 5. An Overview of RF Filter Design

Basic Resonator and Filter Configurations

Chapter 6. Active RF Components

Chapter 7. Active RF Component Modeling

Chapter 8. Matching and Biasing Networks

Impedance Matching Using Discrete Components

Chapter 9. RF Transistor Amplifier Designs

Chapter 10. Oscillators and Mixers

10.1 Basic Oscillator Model

Appendix A. Useful Physical Quantities and Units

Appendix B. Skin Equation for a Cylindrical Conductor

Appendix C. Complex Numbers

Appendix E. Physical Parameters of Semiconductors

Appendix F. Long and Short Diode Models

Appendix H. Noise Analysis

Appendix I. Introduction to MATLAB

1.1 Importance of Radiofrequency Design

Importance of Radlofrequency Design

What characteristics make the high-frequency behavior of electric components so

t ........ ------ - -- - -- --- - ------ ............

Analog Signal Circuits

Block diagram of a generic RF system.

8.2 pF-=RFC ;----

Simplified circuit diagram of the first stage of a 2 GHz power

Importance of Radlofrequency Design

Printed circuit board layout of the power amplifier.

To understand, analyze, and ultimately build such a PA circuit requires knowledge

High-frequency bipolar junction and field effect transistors as ~ell as RF'diodes

1.2 Dimensions and Units

!X = Zo = JIUE = J<JloJl,)/(eoe,) = 377 n.JJl/E,

Dimensions and Units

Example 1-1: Intrinsic wave impedance, phase velocity, and

8.85xl0The phase velocity according to (1.3) is equal to

which happens to be the speed of light v P = c . The wavelength is

where f is the operating frequency. Using equation (1.4), we find