applies, where r denotes the position vector of a general point

on the surface. The standard Helmholtz representation of the

scattered field becomes

E() = -

1{

aE

G(r, r) an (r)

+ E@) 7
an
ac (r, r)

ds

good agreement with a moment method solution. Also

shown is the OSRC solution without edge corrections. This
is clearly deficient close to grazing incidence.

(2)

where ds denotes the element of arc length on the boundary C

of the cylinders cross-section and a/an represents the
outward normal derivative on that contour. The twodimensional (2D) Greens function is given by

-Is -10

-301

G(r, r) = j/4 HL(k I r - r 1 )

where k = w(po
and p o and E,, are, respectively, the permeability and permittivity of free space. Eqn. 2 is an integral
equation for the scattered field. For the case of TM incidence
the (approximate) OSRC together with the (exact) boundary
condition eqn. 1 gives the normal derivative of this field on
the surface of the scatterer in terms of the known incident
field. In its B , form (see Reference 1) this condition is

aE

2[k

+ JS(s)] ds2 (I

where S(s) is the curvature at a point on the cylinder surface

at arc length s from some reference point. Thus the field
external to the scatterer is given as an integral of known
quantities.
Application to the case of scattering by an elliptic cylinder
is straightforward in the (U,U, z) co-ordinate system. Considerable simplification of the result follows for the particular case
of scattering by a conducting strip. Following a careful
analysis of the integral in the limit a -+ I, b -+ 0 we find
ES(r,4) = (1 - 3 cos a)E;(r, 4)

+ Ei(r, 4)

-40

1\

;
I,

;,

1860/21
Fig. 2 Normalised backscattering width of conducting strip for T M
incidence
kl=5
-moment method
. . . . . . . . OSRC
_ _ _ _ OSRC without edge term

Conclusions: The OSRC formulation of electromagnetic scattering has been used to analyse the scattering of an Epolarised plane wave by a perfectly conducting strip. The
existence of new contributions to the edge diffracted fields has
been shown. Their inclusion greatly enhances the quality of
the OSRC predictions.
Acknowledgments: The author wishes to thank Professor D. S.
Jones of The University, Dundee and Dr. I. Anderson of
RSRE for their encouragement.
21st November 1988
I. D. KING
Royal Signals & Radar Establishment
St. Andrews Road
Great Malvern, Worcs. W R 1 4 3PS, United Kingdom

Reference
where

1
I

E;(r, 4) = - 2

HL)[k(r

+ x

- 2rx cos

4)]

KRLEGSMANN,G . A., TAFLOVE, A., and UMASHANKAR, K. R.: A new

formulation of electromagnetic wave scattering using an onsurface radiation boundary condition approach, I E E E Trans.
Antennas Propag., 1987, AP-35,pp. 1 5 3 1 6 1

-I

2
cosa
and
3jn
32

E;@, 4) = - - {HL1[k(r2 I

ANALYSIS OF ANISOTROPIC MULTILAYER

BULK-ACOUSTIC-WAVE TRANSDUCERS

21r cos 4)1/z]dklc0sa

+ Ht)[k(r + 1 + 21r cos ~ ) 1 ~ 2 ] e - j k ~ c o r a }

It is readily verified that (1 - f cos a)E:P, 4) is the scattered
field obtained when the OSRC formulation is applied to the
two faces of the conducting strip I x I I I, y = 0 and the effect
of the edges is neglected. This contribution is consistent with
the complete result published in Reference 1.
The component &(r, 4) of the scattered field is the same as
that due to isotropic line sources of equal strength at the
edges x = fI, y = 0. The associated phase factors are those of
the incident field at the respective edges.
Results: Use of the Bz OSRC leads to the following result for
the scattering width of a perfectly conducting strip:
sin [kl(cos a - cos 4)]
(cos a - cos 4)

cos [kl(cos a - cos 4)]

Fig. 2 shows the normalised backscattering width 0 (n - .)/A

as a function of incidence angle U for the case kl = 5 . (A = 2x/k
is the wavelength of the incident radiation.) The result shows
ELECTRONICS LETTERS

5rh January 1989

Vol. 25

No. 1

Indexing terms: Acoustic waves, Ultrasonics, Transducers,

Multilayers, Superlattices, Piezoelectrics
A general algorithm is described for calculating the electroacoustic characteristics of anisotropic piezoelectric, arbitrarily oriented, multilayer bulk-acoustic-wave transducer
arrays. The algorithm is easily implemented on personal
computers using existing commercial software.

Introduction; Recent bulk-acoustic-wave (BAW) experiments

at millimetre-wave frequencies have demonstrated the practicability of transducers consisting of zinc oxide multilayers
with alternating crystal orientations. In this letter, an analysis
procedure is outlined for calculating the electromechanical
transduction characteristics of BAW transducers consisting of
anisotropic, piezoelectric, arbitrarily oriented, multilayer
arrays. The solution to the coupled electromechanical
problem which is given for a multilayer stack-on-a-substrate
with the other end of the stack free includes a full set of
electrical and mechanical variables and permits us to separate
the substrate acoustic power into shear and longitudinal mode
contributions. This is in contrast to the single-mode cascaded
Mason-model approximations current used.

57

Analysis: The structure analysed is shown in Fig. 1. It consists

of an N-layer stack on a substrate, with each layer of arbitrary
thickness and crystal orientation. The co-ordinate system sets
the x , axis normal to the interface planes. With the material
stiffness C, piezoelectricity e and permittivity E constants of
the layers rotated to the co-ordinate system, the constitutive
equations, the equations of motion and the electromagnetic
field equations for each material are: with i, j = 1 , 2, 3 and an
exp Gwt) time dependence:

dTj
dxj

-= j w p q
V.D=O

E = -V4

p is the material density,

(3)

T j ,S,, and uj are the components of

mechanical stresses, strains, and velocities, respectively, D j and

E j are the components of electric displacement and electric
field, and 4 is the electric potential. Choosing the variables
which must be continuous at interfaces: T = [T,, TZ3T,,]';

7,

(6)

= @fl

Thus, at a given frequency an entire stack of arbitrarily oriented layers is replaced by a single 8 x 8 transfer matrix relating
the values of the eight variables at the top to those at the
bottom. To complete our analysis of this problem requires
specifying the electrical and mechanical boundary conditions
at the end faces. We set the potential zero (datum) at the
substrate interface and consider the voltage V on the top as
unknown. With no charge carriers D is constant throughout
the entire stack and the current density is just jwD. In the
semi-infinite substrate only downward propagating modes are
allowed and the normal stresses and velocities (the 'lower'
values) are related by
VI

Y, TI

(7)

where Y,,,is the stiffened mechanical admittance matrix of the

substrate. We now interchange the 4th and 8th rows of CP and
express eqn. 6 as

4
electrode

hN

layer (N-1)

hN-l
~

material parameter matrix and thickness respectively. For

multilayers, the mapping of the vector z lower face to upper
surface is obtained with CP = CPN, .. . , CP1 and using subscripts
U and 1 to designate 'upper' and 'lower',

where Q is just CP with a 4 * 8 row interchange, and express

the first four equations of eqn. 8 as

where the Qd,Q,, matrices are the upper-left-half and upperright-half of Q. Using eqn. 7, the electrode potential conditions 4l = 0, 4, = V, and the stress-free upper-face condition
Ts= 0, results in

layer 3

h3

layer 2
I

electrode

hl
7--

where Y,, is Y , augmented by a 4th row and column of zeros

to reflect 4l = 0. In terms of the matrix elements of Y defined
by eqn. 10 we have
Tl =jwVCYl4

1865/11
Fig. 1 Geometry of multilayer transducers
U = [u1u2u,]'; D = D,, jar$ as system variables (for convenience vectors are written as transposed row vectors), eliminating all others, and assuming no variations of field
quantities in the x i or x 2 directions of Fig. 1, eqns. 1 to 3
reduce to

y24

Y3,1'

= jwVY,,

(11)

from Fig. 1. The electrical driving point or input admittance

for an electrode area a and input current I is yin = I/V =
-jwDa/V; hence using the second statement in eqn. 11, the
input admittance is Yin= w 2 Y44 a and using eqn. 11 and 7 the
shear and longitudinal mechanical power propagating into the
substrate can be calculated separately. We conclude this letter
by showing, in Fig. 2, our calculated one-way transducer
insertion loss, for the 32 GHz device geometry described in
Reference 1. The device consists of two gold electrodes and

dT

-= jwpv
dx,

X is a 4 x 4 matrix of material constants rotated with respect

to the x l x 2x , co-ordinate system., Defining r =
[T,, T', T,, D U , ~u,, jofl' as the vector of variables which are
continuous at interfaces (eqn. 4) can be cast into the vectormatrix differential equation form3
dr
dx,

-= j w

[x o'1%

= jwAr

where the only nonzero elements of y are yii = p for i = 1, 2,

3. Eqn. 5 is the eight dimensional (8D) differential equation
which each material must satisfy and has the well-known
matrix exponential solution 7(x3 h) = exp GwAhMx,) =
@(hMx,). The matrix exponential is a transfer matrix mapping
values of r within a layer separated by a distance h. In particular for mapping across the entire ith layer, the transfer matrix
is Qi = exp (jwAihi) where the A i and hi are the ith layer

58

01
26

28

30

32

frequency, GH z

3L

36

38

40

Fig. 2 One-way transducer insertion loss

-calculated longitudinal loss
- _ _ _ calculated shear loss
estimated from data reported in Reference 1

***

ELECTRONICS LElTERS

5th Januarv 1989 Vol. 25 No. 1

fifteen 950 A thick ZnO layers with nominally alternating

C-axis orientations with respect to the xg axis of o", 4o",0",
-40" etc.; the substrate is sapphire and the transducer radius
is 20pm. In Fig. 2 the solid curve is the calculated insertion
loss for the longitudinal power and the dashed curve is that
for the shear power. Estimates from measured data reported
in Reference 1 are shown by asterisks; although the fit is not
good, the general features of a number of loss minima are in
agreement with the simulations. Note that the shear contribution is not negligible in some parts of the response and hence
Mason-model based simulations would be inaccurate. The
calculations were carried out on a PC AT clone using PCMATLAB: a commercially available matrix-computation
oriented software package.
Summary: A general formulation has been described for calculating the characteristics of anisotropic multilayer BAW transducers. Use of the algorithm has been illustrated by an
example showing the shear and longitudinal insertion losses
for a 15 layer ZnO transducer. In contrast to the Masonmodel, single-mode simulations, the algorithm described in
this letter allows the calculation of all the substrate modes
excited by the transducer.
Acknowledgment: The support of the Natural Sciences and
Engineering Research Council of Canada is gratefully
acknowledged.

E. L. ADLER
2lst November 1988
Department of Electrical Engineering
McGill University
3480 University Street, Montreal, Quebec, Canada H3A 2A7

References
et al. : 'High efficiency multilayer ZnO transducers
at millimeter-wave frequencies', Appl. Phys. Lett., 1987, 50, pp.
1642-1644
SITTIC, E. K. : 'Transmission parameters of thickness-driven
piezoelectric transducers arranged in multilayer configuration',
IEEE Trans. Sonics and Ultrasonics, 1967, SU-14, pp. 167-174
FAHMY, A. H., and ADLFX, E. L.: 'Propagation of acoustic surface
waves in multilayers: a matrix description', Appl. Phys. Lett., 1973,
22, pp. 495-497
PC-MATLAB is a trademark of the Mathworks, Inc., Sherborn,
MA 01770. U.S.A.

required is in excess of 600 MHz. Any phase distortion introduced has the effect of degrading the range sidelobe level of
the radar point target response. Amplitude distortion is considered less important, since this can be corrected if necessary
using limiting stages, provided these introduce no phase distortion of their own.
voltagecontrolled

delay T

n
signal

voltage
ramp
qenerator

[N~I

oscilloscope

phase
detector
trigger

rn

b F . i I e

frequency
divider

oscillator

Fig. 1 Block diagram of method of measuring phase errors in linear FM

waveforms (adaptedfrom Reference 2)

The method is based on a technique2 for measuring the

phase distortion on linear FM signals, shown in Fig. 1. If an
undistorted linear FM pulse is mixed with a delayed version
of itself, the difference frequency is a pure sinusoidal beat note,
whose frequency is proportional to the delay and the frequency sweep rate. If this beat note is phase-detected against a
coherent sinusoidal signal of the same frequency, the result
will be a constant DC level. However, if there is any phase
distortion present on the linear FM pulse, the beat note is not
a pure sinusoid, and the corresponding output from the phase
detector is proportional to the phase distortion of the original
pulse. This characteristic can be displayed directly on an
oscilloscope and (if necessary) optimised in real time.

HALXMIOGLU, B.,

Description of method: The principle of the method to be

described is to derive beat signals, as above, on the signal at
the input to the multiplier, using a delay T , and also at the
output, using a delay T / N (where N is the frequency multiplication factor). The frequencies of the beat signals will be identical, so by comparing them in a phase detector a direct
indication of the phase distortion introduced by the frequency
multiplier is obtained (Fig. 2). Since the method is based on a
comparison of beat signals, no coherent reference is necessary.
delay t

METHOD OF MEASURING PHASE ERRORS

INTRODUCED BY FREQUENCY MULTIPLIER
STAGES
Indexing terms: Signal processing, Frequency multipliers, Distortion, Radar
A technique is described for measuring the phase distortion
introduced by frequency multiplier stages. The distortion
characteristic is displayed on an oscilloscope, which allows
the response to be optimised in real time. The technique has
been demonstrated on several types of frequency multiplier.

Introduction: Frequency multiplier stages are often used to

increase the frequency and bandwidth of signals in radar and
communications systems, especially where it is more convenient or necessary to generate the signal at lower bandwidth
than ultimately required. For many such systems it is important that no distortion, either in amplitude or phase, is introduced by the frequency multiplier. Yet this is a difficult
parameter to measure directly, because the frequency multiplier is by nature a nonlinear device, and since the output
frequency is different to that of its input, conventional
network analyser techniques cannot be used.
The method to be described was developed to measure the
phase distortion introduced by frequency multipliers for a
wide-bandwidth radar altimeter system.' The signal to be
multiplied is a linear FM chirp, and the final bandwidth
ELECTRONICS LETTERS

5th January 1989

Vol. 25

No. 1

voltagecontrolled
oscillator

I
voltage
ramp
generator

trigger
I

1911:21

Fig. 2 Block diagram of method of measuring phase distortion introduced b y frequency multiplier stage

There are several practical considerations to be taken into

account. The number of cycles of the beat signal within the
pulse is given by T T AfAt, where T is the pulse duration, T the
delay and AfAt the FM sweep rate. For maximum precision
the values of T , T and AfAt should be chosen so that there are
several cycles of beat signal within the pulse duration. The
delays may be realised using lengths of cable, but any variation of cable loss over the sweep bandwidth will cause a corresponding amplitude variation in the beat note. For these
problems to be minimised the FM sweep rate should be as
large as practicable; however, it is also important that the
cable dispersion should be negligible over the sweep bandwidth, since otherwise this would introduce phase errors of its
own. In practice, the delay T at the input is fixed, while the
delay (nominally T / N )at the output allows for a small degree
of adjustment by means of a line-stretcher, so that the beat
frequencies can be exactly equalised.

59