ULTRASONIC IMAGING 1, 189-209 (1979)

DESIGN OF SLOTTED TRANSDUCER ARRAYS WITH MATCHED BACKINGS

Gordon S. Kin0

E. L. Ginzton Laboratory
Stanford University
Stanford, California 94305

and

Charles S. DeSilets
Precision Acoustic Devices, Inc.
991 Commercial Street, # 1
Palo Alto, California 94303

A design procedure is described for acoustic transducer arrays suit-

able for focused acoustic imaging systems. The criteria governing the
choice of the width of the individual elements of the array and for
broadband matching are discussed. Experimental results on 2.5 MHz arrays
are given and are in good agreement with theory.

Key Words: Acoustic imaging; acoustic transducer arrays; cross coupling

of transducer elements; finite cross-section acoustic trans-
ducers; matched backing; PZT ceramics

Introduction

The design of broadband efficient linear acoustic transducer arrays

suitable for acoustic imaging in water is described in this paper. These
arrays, illustrated in figure 1 , are constructed of tall thin piezoelec-
tric elements mounted on a lossy backing medium and radiating into the
load medium (water). A face plate, acoustically similar to water, is
used to keep water from getting between the elements and introducing
acoustical and electrical cross-coupling. The choices of backing mater-
ials are based on the concepts described in a previous paper on single
element transducers [l]. In a later paper we will describe concepts and
results on slotted arrays with matching layers between the elements and
the load medium.

There are several problems, both theoretical and technological, which

are unique to the design of small element arrays. These include consider-
ation of the extensional mode of an element of nearly-square cross-section,
determination of the effective load impedance on each end of an element,
the design of acoustic matching layers of finite cross-section, and con-
sideration of the effects of cross-coupling between elements of the array.

Model of a Narrow Slotted Element

It is desirable in focused imaging systems to space array elements

one-half wavelength apart to avoid grating sidelobes in the image. The
width L andheight Hofan element, as illustrated in figure 2 , are usually
comparable in magnitude, although in the design we shall normally attempt
to make H > 2L . The boundary conditions for a piezoelectric resonator

0161-7346/79/030189-21$02.00/0
Copyright @ 1979 by Academic Press, Inc.
189 All rights of reproduction in any form reserved.
 KIN0 AND DESILETS

LOAD MEDIUM (WATER)

FACE PLATE

TRANSDUCER
ELEMENTS

BACKING MEDIUM

Fig. 1. Schematic drawing of transducer array elements.

--L +
2

Fig. 2. Schematic drawing of piezoelectric ceramic element showing

poling direction and coordinate system.

of these dimensions are obviously much different from the thin disc
resonator discussed in Reference 1 where L >> H .
Now the longitudinal
strains S1(Sxx) and S2(Syy) cannot be taken as zero. Consequently,
the simple theory of the thin disc resonator cannot be applied to this
structure without modification.

In a linear array where the transducer element is essentially a two-

dimensional structure as shown in figure 2 , there is strong coupling
between the desired zeroth-order extensional resonance (2-direction) and
the zeroth-order lateral extensional resonance (x-direction) when the
element has nearly-square cross-section [ 2 ] . In this case, the frequency of
the desired lower branch of the coupled modes is significantly down-
shifted. In addition, the relatively undamped upper branch of the coupled
modes is very close to the frequency of the lower branch which is damped
by the presence of the backing and matching layers. The former mode will
tend to be excited in any broadband system, yielding a peaky transducer
bandshape. For these reasons, transducer elements of nearly-square cross-
section should be avoided in broad bandwidth systems.

Some commercial unfocused acoustic imaging systems have been designed

in which several elements are excited simultaneously and the echoes
processed into an image. These systems work with elements spaced many

190
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

wavelengths apart, and so the elements have lateral dimensions much

greater than the thickness. These elements are operated in a mode similar
to the thin disc mode, and the acoustic response will include lateral
resonances and harmonics. Little discussion of this type of element will
be included in this work, however, since the emphasis is placed on tall,
thin elements suitable for electronically scanned and focused system with
a wide angle of acceptance. Some of the discussion of tall, thin elements
could easily be extended to wide elements.

If the height of the element is greater than about twice the width,
the lateral resonance is not strongly coupled to the extensional reson-
ance and can be ignored in calculation of the transducer response, as
will be shown. A simple one-dimensional theory can be derived and used
directly in the transducer transmission line model with good success.
Using the notation of Auld [3], the piezoelectric constitutive relations -
can be written in the form

E
T = - e * E + C : S

where the bars denoting vector and tensor qualities have been dropped for
brevity, TI is the stress in reduced coordinates, SI the strain, Ei
the electrical field, and D. the displacement density. The piezoelec-
tric stress matrix is denotea by eiJ in reduced coordinates, and
eiJ = eI. is the transpose of this matrix. The stiffness at constant
E is detloted by C& , and the permittivity at constant stress by E : ~ .
For a thin tall piezoelectric element poled in the z direction, the
transverse stress Tl is zero, so we can write

T1 = cE11S1 + E
C13S3 - ezlEz = 0

E
T~ = cE13s1 + c33~3- eZ3E

S
DZ = EzzEz + ezlSl + eZ3S3 *

It follows from Eqs. ( 3 ) , ( 4 ) , and (5) that

T~ = c;:s3 - e'23Ez

where
2
E

'i: =[3:'
1 -
'
(C13>
E E
1'
3
]
e cE
e' = e -- zl 13
23 23 E
c1 1

191
 KIN0 AND DESILETS

and

Eqs. (6) and (7) are identical in form to those for a thin disc trans-
ducer with the modified constants defined in Eqs. (8) - (10). Therefore,
all results for the thin disc transducer [l] can be applied to the case of
the narrow slotted transducer including quarter-wave matching and backing
schemes using these modified constants, with other considerations to be
discussed in detail later.

Other parameters of interest are derived below. The modified param-

eters are derived as follows, taking open circuit conditions or
Dz(x,z) = 0 in Eq. (7):
e'
EZ = - - 23 s3
€2

T3 =

Therefore, the stiffened elastic constant is

lD
c33
= Ci: (1 + KV2)
where

e' 2
23
&'S
33 zz

Since the constant K P L appears in the functional form of the one-

dimensional transducer input impedance in the form Kt2/(1 + KP2) [l], it is
convenient to introduce a simplified constant q2
, where

The stiffened velocity and acoustic impedance of the transducer can be

defined then as follows:
- D 112
v;, = (Ci3/P)
- -
z;, = v;,p

Various parameters for PZT-5A and PZT-5H are tabulated in table I

using the data published by Jaff6 and Berlincourt [4]. These materials have

192
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

Table I. Various parameters for PZT-5A and PZT-5H.

PZT-5A PZT-5H
PARAMETER

E
c1 1

been chosen since they have high coupling coefficients and high dielectric
constants. It will be noted that the parameters for the strip transducers
are very close to those for the extensional mode of a rod.

The high coupling coefficient is desired for the same reasons as

discussed in Reference 1, that is, broad bandwidth with high efficiency.
The high dielectric constant is most important for small transducer
elements in order to keep the electrical impedance as low as possible,
thus making electrical matching simpler. All the arrays reported in this
work were constructed of PZT-5A or its equivalent, depending on the manu-
facturer. PZT-5A has a smaller aging coefficient than PZT-5H and a much
higher Curie temperature, making it easier to work with than PZT-5H.

The most significant aspect of the changes in the transducer param-

mechanical coupling coefficient, 5s .
eters produced by slotting is the a proximate doubling of the electro-
The coupling constant for the
2
narrow, long element, q2, is close in value to k33 , the length
extensional coupling constant of a thin rod. as seen in table I. This

193
 K I N 0 AND DESILETS

G ( = L I H ) . L CONSTANT

Fig. 3 . Generalized coupled mode dispersion curves showing limiting

cases.

doubling is mostly due to the fact that C'E is decreased to about one-
half the unslotted value. Doubling k$ e83ectively halves the electri-
cal Q of a series tuned transducer, making the design of low-loss
broad bandwidth elements a simpler task.
2
In addition to the large increase in kT , it can be seen that there
is about a 14 percent decrease in the velocity and impedance of the PZT
materials and a slight increase in the effective dielectric constant.

Coupling to a Lateral Mode

A simple heuristic theory for coupling between the two dilatational

modes of a piezoelectric resonator has been given by Onoe and Tiersten[2]
for the short-circuited case, and we have adapted their approach to the
open-circuited case. This theory models the element as two one-degree-
of-freedom systems of resonant frequencies fa , fb , respectively,
coupled through a single mechanism. Assuming two degrees of freedom
and a mode coupling which is invariant with frequency, they obtained the
following biquadratic coupled mode equation.

where fa and fb are the eigenfreauencies of the uncoupled system, and

r is a dimensionless constant coupling coefficient. This equation is
plotted on figure 3 on a log (fH) versus log G where G = L/H coordinate
system. The dashed lines give the uncoupled resonant frequencies, and
the solid line the coupled frequencies. Following the arguments of Onoe
and Tiersten 121, a simple way of determining the coupling coefficient r
consists of selecting r to yield the known frequencies at the endpoints
of the frequency curves for large and small values of G .
A s f + fb ,
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

the limiting frequency is the well-known thickness dilatational mode of a

thin disc resonator. So if the transducer is open-circuited

(19)

where

As f -+ fd , the limiting frequency is that of the extensional mode

calculated in the previous section where

fd = -
2H

and ;,!C is defined in Eq. (13).

As f -+ fa , the limiting frequency is that of the lateral thickness

dilatational mode, which yields the result

fa = -
1 6
2L P

where

As f -+ fC , the limiting frequency is that of a lateral extensional mode,

with

- -

where

and

= cEz + eZ3/~33
2 E .

195
 K I N 0 AND DESILETS

-
EXPERIMENTAL DATA -
UNDAMPED LATERAL MODE -
A DAMPED EXTENSIONAL MODE -
-

-
-
-
-
-
-
-

0.I I I I I I l l l l
0.I 1.0 10.0
G ( = L / H ) . L CONSTANT

Fig. 4 . Experimental and theoretical dispersion curves for slotted

array elements.

For internal consistency, the relation fa/fc fb/fd should be satis-

fied. It follows directly from Eqs. ( 1 9 ) , ( 2 1 ) , ( 2 2 ) , and (24) that
fa/fc is algebraically identical to fb/fd so that internal consistency
is satisfied. Substitvting the parameters for PZT-5A and PZT-5H,
fa/fc = 1.16 in both cases.

To determine the coupling constant r2 , the ratio fd/fb must be

cast into the form

fd/fb
= (1 - r2)
112
= [(.-A)(%] ‘11‘33 k;3
112

(29)

Substituting the parameters for PZT-5A, r L is found to be 0.257 for the

open-circuited case. A plot of both branches of the frequency spectrum
for both the open-circuited and short-circuited cases is shown in figure
4 . From the figure, it can be seen that there is only a 9 . 3 percent fre-
quency shift from the open-circuited extensional mode for a configuration
ratio G = 0 . 7 and a 5 percent shift for G = 0 . 5 .
Note that the theoretical
curves are for the case of the resonator with free boundaries.

An important problem in designing elements where G is close to one

is the emergence of an undamped mode close to the desired passband of the
transducer element. This mode is clearly seen from the experimental
results in figures 4 and 5 where the measured real part of the electrical
input impedance of lead-backed (Z = 22) PZT-5A transducer elements of
varying configuration ratio is shown. A s the figuration ratio approaches
unity, this undamped lateral mode, characterized by the sharp peak in the
curves, moves into the passband of the element and decreases the bandwidth.

1%
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

8
. .
0.5
G = 0.60
I
. -

G = 0.50

.....
_.

O' 0:s I:O 3 & O

<
N O R l l I L l L E D FREOUENCY
3;O

Fig. 5. Real part of experimental electrical impedance of lead-

backed PZT-5A elements 0.0635 cm thick X 1.27 cm long
for varying values of configuration ratio. Data taken
with computer-controlled HP 4865 Vector Impedance meter.
The experimentally measured frequencies for lead-backed elements with the
G ratio between 0 . 3 7 and 1.03 are shown in figure 4. Note that the
experimental data are for elements with a high impedance backing while
the theory is for elements with free boundaries only. The lateral mode
frequencies are about 23 percent lower than that of the lateral extensional
mode defined in Eq. ( 2 2 ) , probably due to coupling to higher order modes and
to the fact that the theory is only an approximate one, and is worked out
for an undamped resonator. The extensional mode frequencies are slightly
less than the open-circuit frequencies calculated from an element with
free boundaries. The downshift in these frequencies when one side of the
element is heavily backed is expected from the one-dimensional transducer
model.
These data show that the configuration ratio should be 0.5 or less so
that the undamped lateral mode frequency is at least an octave higher
than the center frequency of the passband. At this point, there is little
coupling between the modes, and the undesired mode can be easily filtered
out if desired. Near G = 0.65 , coupling between the modes begins to
limit the bandwidth of the transducer element as seen in figure 5.

197
 KIN0 AND DESILETS

W
0
z 0-= 0.4

-=-r
C =0.333
0
l-

3.0
-__ ---:+i
:i-=++-F
4.0 5.0
kk

Fig. 6 . Normalized effective acoustic impedance seen by finite

width radiator for two values of backing Poisson's ratio.

Effective Acoustic Load Impedance of a Narrow Element

A transducer array element radiating into a semi-infinite loading or

backing medium excites waves in all directions in these media. If the
element is narrow enough, the power radiated into waves propagating at an
angle to the normal becomes large enough so that the plane wave longitu-
dinal mode impedance no longer adequately describes the load impedance
seen by the element. A variational theory is developed in this section
to calculate the effective load impedance of the element at the interface
taking into account the longitudinal, shear, and Rayleigh waves excited,
as illustrated in figure 6 . We assume, unlike much of the literature,
that the radiator is terminated by a semi-infinite medium and is contained
in an infinite baffle of zero impedance, i.e. the normal stress field
rather than the normal displacement is zero outside the element.

A similar but nonvariational calculation for acoustic radiators of

varying configuration ratio was carried out by Miller and Pursey[51 for a
radiator terminated by a semi-infinite solid. The approach used and the
expression they calculate gives results very close to the variational
expression derived in this paper for solids. The method we have used is
closely analogous to those employed in EM theory calculations of the
radiation impedance of a waveguide [6,7,8] and is described in detail
in a report by DeSilets[8].

The normalized load impedance for a slotted element calculated by

this method is shown in figure 6 as a function of (kaL) for various
values of Poisson's ratio. For (kaL) > 1 , the impedance is essentially
real and oscillates slowly about the longitudinal plane wave impedance.
Below (kaL) ^. 1 , the reactive part increases in value rapidly so that
the impedance becomes mainly capacitive. Since the assumed stress field
T3(x) was taken to be uniform, the impedance calculated here is identi-
cal to that of Miller and Pursey at the six values of ktL they eval-
uated. This complex-valued impedance was used as the backing impedance
in the standard program which computes the electrical impedance and
insertion loss of transducers, using the modified transducer constants
for the extensional mode [l]. The change in this impedance, due to finite
width, has little effect on the transducer elements of the widths used

198
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

-.-.__
1.0 - -- ----
I-
K
2
-\-----

4 A .
/.", I I I
0.0
gd 1.0 2.0 3.0 4.0 5.0
2: -u = 0.40
W
a
g -1.0
-1.0-
--- u = 0.49
--U = 0.50 (liquid)
uI-
I-
v)

Fig. 7. Comparison of effective load impedances of finite fini.te width

radiator for Poisson's ratios of 0 . 4 , 0 . 4 9 , and 0 ..so.

in the arrays reported here but must be accounted for in narrower elements
where kkL << 1 .
For the case where the load medium is water, a similar calculation
has been carried out by taking into account that there are no radiated
shear or Rayleigh waves in a liquid. The results obtained are somewhat
different than those for a solid because a liquid is unable to support a
discontinuity in the stress field at the edges of the transducer and
cannot propagate the shear waves normally excited at these edges. To
show that this difference is real, a case was.calculated where the load
was a solid with u = 0.49 , that is, almost a liquid. This case is
shown in figure 7 compared to a water load and a solid load with = 0.4.
As can be seen in the figure, the o = 0.49 solid exhibits a more induc-
tive nature than the u = 0.4 solid, and the impedance characteristic is
intermediate between the two extremes. Although we have not carried out
the calculations, it might be expected that a rubbery solid like RTV, as
it exhibits high shear wave loss, would give an input impedance much
like that for u = 0.4 given here. But the shear wave that was excited
would be rapidly attenuated, and thus the efficiency of excitation of
longitudinal waves in the medium would be comparable to that of a liquid
with the same longitudinal wave impedance. However, when such a layer
is used at the front of the transducer to excite waves in water, it
would make the matching problem simpler, despite the losses introduced by
it. The effect of this variation in load impedance in the transducer
characteristics is shown in the next section. However, for the width of
elements used in the experimental arrays, there is little effect on the
bandshape of the transducer elements.

199
 K I N 0 AND DESILETS

Experimental Results with Simple Backed Arrays

The construction of broad bandwidth linear arrays of these elements

mounted on a high impedance, lossy backing will be discussed in this
section. The experimental electrical impedance and insertion loss of
these elements will be compared to the theory. The radiation pattern of
the elements will also be briefly described.

The easiest technique for achieving broad bandwidth in a water-loaded

transducer is to mount the transducer onto a matched backing. Although
most of the power generated is radiated uselessly into the backing in
this case, bandwidths of 80 - 100 percent are achievable with ferroelectric
ceramics. The bandshape is nearly Gaussian-shaped so that an excellent
impulse response can be achieved as well. Our first arrays were built
using PZT-5A ceramic (ZQ = 34(10)6) soldered to a lead (ZQ = 22(10)6)
backing. The PZT-5A was lapped to the correct thickness, surface fin-
ished with a 5 pm grit, and slightly polished. Chrome-nickel electrodes
of 2000 f thickness were rf sputtered on and the ceramic was repoled at
150° C at an applied voltage of 2kV/m to compensate for any depoling
which occurred during the sputtering process.

Because of the background noise associated with lead backings and

other metals like soft iron and brass, tungsten-loaded epoxy backings
were manufactured which yielded not only the high impedance desired, but
also the high loss and fast background decay rates. Tungsten is the
obvious choice as the filler material because its impedance
(ZQ = 101(10)6) is large. Composites made of tungsten and epoxy [9] and
tungsten and vinyl powder [9] are described in the literature. One tech-
nique uses tungsten powder mixed with excess epoxy. The excess epoxy
was then squeezed out in a press. Backings were readily made using this
technique, but the highest impedance achieved was only ZQ = 14(10)6 .
In addition, the material contained many voids which would be disastrous
mechanically and acoustically if they occurred at the surface where the
ceramic is bonded. A second method is to mix tungsten and vinyl powders
in known ratios and combine the matrix with standard hot-pressing tech-
niques. Lack of facilities precluded using this technique. A third
method, suggested by L . Zitelli [lo], utilized vacuum impregnation of epoxy
into a matrix of tungsten powder. This method was easily implemented.
By compressing the powder in a mold to a given density and then vacuum
impregnatin the backing impedance could be easily varied from
ZQ = 12(10)t' to ZQ = 40(10)6 . However, as the impedance rose, the
loss dropped. A compromise choice was used for the arrays with
ZQ = 25-28(10)6 with a loss of approximately 8 dB/cm .
All the epoxy-backed arrays reported in this work were built using
a cast-in-place process developed jointly with J. Fraser. In this method,
the ceramic was bonded to the backing in the same step where the epoxy
was impregnated into the tungsten. This whole process was carried out
in a mold. The mold was made in two parts, a bottom plate containing
a vacuum tight "O-ring" groove which is then bolted onto the mold itself.
The mold is open top and bottom and conformed to the shape of the backing.
The ceramic is placed in the bottom of the mold with the bottom plate in
position and held flat against the bottom with spring-loaded rods to
prevent particles from getting underneath the ceramic. Tungsten powder
is poured intothe desired thickness and compressed with a ram to the
desired density. Epoxy, Dow Chemical resin DER 332 cured with a 14 percent
metaphenylene diamine by weight[13]in this case, is poured on top of the
tungsten. The whole mold is then placed into a vacuum station and pumped
down to 100 - 300 pm. This process evacuates all of the air from the
mold. When the vacuum is released, air pressure simply forces the epoxy
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

into the evacuated tungsten powder. A bar is clamped onto the back of
the tungsten-epoxy matrix until the epoxy cures to keep the tungsten
particles in intimate contact with the ceramic. This last step is
vitally important as a fairly thick epoxy bonding layer develops if
omitted. This process eliminates the need for a separate, elaborate
bonding process to attach the ceramic to the backing and yields consis-
tent, uniform results. The "equivalent" epoxy bond thickness is less
than 1 pm as measured in angle-cut samples and in comparison to theor-
etical results.

The impedance of composite backing materials depends strongly on the

size, size mixture, and shape of the particles used. A powder contain-
ing a range of particles of spherical shape will pack more densely than
single-sized or jagged particles. Backings made with flame spray tungsten
powder sieved to -325 mesh (less than 4 4 microns) with no pressure packing
came out with an impedance of 23(10)6 when simply vacuum impregnated.
However, with shape and size variability between powders, the impedance
prediction remains an empirical one.

The loss in these high impedance tungsten epoxy backings is high, but
reflections from the back of the backing still limit the dynamic range of
the array. The backing may be made as long as desired with consequent
size and weight penalties or cut into a wedge-shape [11].

Several arrays with 100 to 128 individual elements were built using
the cast-in-place, tungsten-epoxy backings just described. The PZT-5A
ceramic plate with chrome nickel electrodes was first cut into 10 cm X
1.27 cm x 0.0635 cm slabs so that the free boundary half-wave resonance
of narrow slotted elements was 2.97 MHz. The backing impedance was
25(10)6 so that the center frequency was downshifted to about 2.5 MHz.
Ground connection to the elements was made by laying a 0.001 inch thick
brass strip next to the ceramic before pouring the tungsten powder into
the mold. Direct soldering of the brass to the nickel electrode was
unsuccessful as the high pressure (4000 lb/in2) necessary to get the
impedance desired also tended to crack the ceramic in this case. As the
backing was only slightly conductive, the brass was capacitively coupled
to the nickel. The individual elements were cut one at a time on 0.020"
or 0.025" centers using a 0.006" diamond saw. The kerf of such saws
varied from 0.008" to 0.010". This left elements 0.010" to 0.015" wide
although each array had element widths varying only by at most 0.001".
A gold wire 0.002" in diameter was spot-welded onto the top of each
element to make the hot connections to circuit boards glued to the sides
of the backing. Impedance transformers were used to match the high
impedance of elements, typically 800 ohms, down to the 50 ohms of coaxial
cable.

The arrays were glued into an aluminum housing, and a protective

material placed on top of the array. The design of this protective layer
is important. Ideally, the spaces between the elements should be air
filled to prevent acoustic coupling between elements. Epoxy and epoxy
filled with glass microballoons were unsuccessfully tried as filler
materials. The ideal material would be "hard air" which might be made
from various hard-setting foams. To protect the top of the array from
water and mechanical damage, "hard water" would be the ideal material so
as not to disturb the characteristics of the transducers. In our arrays,
a plastic tape was used which was vacuum suctioned onto the array. This
tape was 0.005" thick, but had an acoustic impedance of less than
2.0(10)6 and was very pliable. This pliability was important so as to
reduce the cross coupling between elements through shear waves propagating
in the tap. This tape worked adequately for these arrays.

201
 K I N 0 AND DESILETS

1200r

-2000 THEORY
EXPERIMENT
-2400 1,. 2.97 M H z

Fig. 8. Comparison of theoretical and experimental electrical

impedance of Z = 28(10)6 tungsten-epoxy backed PZT-5A
transducer element 0.0292 cm X 0.0635 cm high X 1.27 cm
long.

A comparison of the experimental and theoretical electrical impedance

of an array element is shown in figure 8. This 128 element array was
used in a reflection mode imaging system [14] where two individual elements
were connected in parallel to form one transmitter or receiver element.
The individual elements are 0.292 cm wide X 0.0635 cm high X 1.27 cm long
backed with Z = 28(10)6 tungsten epoxy. Excellent agreement between
theory and experiment is seen in figure 8. The theory does include
200 pf series capacitance arising from ground lead coupling through the
backing material. The theory does not include the complex load or
backing impedance. A comparison of the theoretical electrical impedances
of a slotted array element (G = 0.46) is shown in figure 9 where in one
case the plane wave load impedances are used and in the other, the complex
effective load impedances are used. The difference between the two cases
is small except at low frequencies where a small peak in the real part of
the impedance is observed at f/fo = 0.25 . The measured electrical
impedance, shown in figure 8, is shown in figure 9 as well and has a
similar peak at f/fo = 0.47 . This is consistent with the measured
electrical impedances of lead-backed elements shown in figure 5 with
G = 0.50 and G = 0.40
f/fo = 0.40 .
. In these cases there is a peak approximately
It is very possible that these observed effects are in
fact the effect of the complex load impedances. The discrepancy between
the measured and predicted peaks could be accounted for by the approxi-
mations used in the variational theory. It is clear that for narrow
radiators like that in figure 5 where G = 0.40 , the peak in the imped-
ance on the low edge of the passband could yield some undesirable
resonances in the frequency response and concommitant ringing in the
impulse response.

The insertion loss of ten array elements connected in parallel is

shown in figure 10. This 100-element array was used as a receiver in a
transmission mode phase contrast imaging system [15,16]. The elements,
connected singly, are 0.0366 cm wide X 0.0635 cm high X 0.10 cm long.
A 4:l impedance transformer was used to bring the input impedance of the
elements down to the reference 50 ohms; 3.5 dB was added to the measured
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

- PLANE WAVE IMPEDANCES

-- SLOTTED
= I
0 * - IMPEDANCES
* EXPERIMENTAL DATA

W
0
z
a
a
I 0.5 I .o I .5 2.0
3 t NORMALIZED FREQUENCY

Fig. 9. Theoretical electrical impedance of 2 = 25(10)6 backed

PZT-5A transducer element with (kOkL)backing = 1 . 7 5 and
(kO&)load = 3.16 showing effect of calculated effective
backing and load impedance.

NOTE: f o = 2 . 9 7 M H ~
G = 0.58
0 ~ 1 1 - 1 ~ 1 . * 1 . 1 . 1 1 . * 1 1

data to compensate for the power reflected into the spaces between the
elements. Diffraction was neglected as the air-water interface was only
1.0 cm from the array. Good agreement between theory and experiment is
seen; the measured bandwidth is 96 percent and the insertion loss is 25.5 dB.
Because G is rather large (G = 0 . 5 8 ) in this case, a strong resonant
peak is noted at 4.2 MHz corresponding to the undamped lateral resonance.
This was of no concern in the cw imaging system for which the array was
used, but in the case where the element is excited by a broadband pulse,
this resonance would cause long decay-time ringing in the transducer
response.

203
 KIN0 AND DESILETS

Cross-Coupling Between Array Elements

Acoustical and electrical cross-coupling between the elements of an

array can produce undesirable artifacts in the spatial frequency response
of an individual element. It can be easily shown that the frequency
response of an element with no cross-coupling to other elements is of the
form

F(8) = sin (kL sin 8/2)/(kL sin 8 / 2 ) (30)

where k is the wavenumber in the load medium, L the width of the

element, and 8 is the angle from the surface normal. If the cross-
coupling is large, the effective width of the element is larger, and the
angular beam width will be reduced significantly. Elements adjacent to
the transducer element can be excited acoustically or electrically.
Acoustic cross-coupling can be due to propagation of waves along the
face plate, typically shear waves, through a filler between the elements,
or along the backing. As far as the backing is concerned, there is
some theoretical and experimental evidence to support the hypothesis
that a slow wave can propagate along its top surface. This wave can
only exist because of the periodic surface loading due to the elements.
For example, at the center frequency, the impedance backing due to an
element 1/2 long is essentially that of water. But at a lower fre-
quency its impedance is inductive, and at a higher frequency capacitive.
This effect tends to lead to a resonance or cut off for a wave propagating
along the periodically-loaded structure at frequencies near the resonant
frequency, and hence to slow wave propagation.

Electrical coupling can be severe when there is much capacity between

the elements. Thus long unshielded lead lengths and water between the
elements must be avoided. A simple theory can be constructed to account
for this type of coupling by assuming the coupling amplitude to be of the
form aei0 where a is a coupling coefficient and 0 is a phase shift.
The coupling to the nth nearest neighbor then will be aneinO The .
angular response is easily shown to be

F(e) = A sinc (kL sin 9/2)

0 1
cinein0 cos (nkll sin 8 ) (31)

where ll is the center-to-center spacing and N is the number of neigh-

boring element pairs. As can be seen in Eq. (31), the first term in just
the uncoupled sin x/x response of a single element. The terms in
brackets are the modulation due to excitation of neighboring elements.
Normally it is only the n = 1 term which is of major importance.

Some of the effects of cross-coupling can be demonstrated by examining

lots of the spatial frequency response given by Eq. (31). In figure 11,
pF(e)I is shown for several values of coupling coefficient including only
nearest neighbor coupling and assuming 4 = 0 or zero phase shift, a
center-to-center spacing-to-width ratio ( l l / L ) of 2, and kll = 1.68 .
It can be seen that coupling of -20 dB introduces fairly significant
modulation of the beam pattern and a reduction in the 3 dB half beam width
from 40' to 15O. It follows from this theory that cross-coupling should
be kept below 30 dB in order to achieve broad acceptance angle transducers.
The effect of phase shift, which occurs with acoustic cross-coupling can
drastically alter the form of the spatial frequency response as seen in
figure 12. In this figure, I F ( 8 ) l is plotted for Oo and 90' phase

204
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

NO COUPLING
10 dB COUPLING
20 dB COUPLING

I-\
I \ ! '.
;/- iJ.4
. 0.8

-80 -60 -40 -20 0 20 40 60 80

ANGLE OF ACCEPTANCE, DEGREES

Fig. 11. Effect of cross-coupling on angular response of an array

element.

ANGLE OF ACCEPTANCE,DEGREES
Fig. 12. Effect of phase shift in the cross-coupling on the angular
response of an array element.

shifts for -3 dB element-to-element coupling including 10 pairs of

neighboring elements. The sample parameters are assumed as in figure 11.
In this case, when a 90° phase shift is introduced, there is a minimum
in the response at 8 = Oo instead of a maximum. As can be seen, inter-
pretation of the angular response function can be complicated when the
coupling is strong.
 KIN0 AND DESILETS

I I 1 I I 1

ANGLE OF ACCEPTANCE, DEGREES

Fig. 13. Comparison of theoretical and experimental angular response

of a backed array with 0.5 mm wide elements.

A plot of acceptance angle for a lead-backed array with 1 mm wide

elements and a thin Mylar face plate (for which the cross-coupling
through the face plate should be very small) is shown in figure 13.
Reasonable agreement between experiment and theory for the basic angular
pattern is shown in this plot with small cross-coupling of the order of
-20 dB introducing small peaks at f 23' and narrowing the response. How-
ever it is not possible to predict from the theory how the angular res-
ponse is affected by cross-coupling, because the phase shift of the
coupling is not known. Certainly a theory based on zero phase shift
between elements would not agree with the experimental results. In later
arrays, a commercial brand (3M) plastic self-adhesive clear plastic tape
was also found to be acceptable as a face plate and had the advantage of
being very easy to apply. This tape is 0.005" thick, is very flexible,
and stretches easily. The tape was applied by vacuum suctioning the
tape down to the face of the array. The angular acceptance of the
tungsten-epoxy backed array described in figure 10 which uses this tape
as a face plate is shown in figure 14 at frequencies of 2.5 MHz, 3 MHz,
and 4 MHz, respectively. The measured acceptance angles of 25O, 28O,
and 13O, respectively, are somewhat less than the theoretical angles of
48O, 3 8 O , and28O predicted from Eq. (2). However, there are no large
peaks or valleys in the overall beam pattern. A best solution would be
to use a material with acoustic properties similar to water for the face
plate. As mentioned earlier, some polyurethane materials meet this
requirement.

Conclusions

Broadband heavily backed transducer arrays can be constructed employ-

ing either tungsten epoxy backings or lead backings. The tungsten epoxy
or other types of filled plastic material is most suitable because slowly
decayingringing occurs due to scattering from grain boundaries in metals.
A relatively rigid backing material is required due to the fact that small
thin elements must be employed.
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

11.2

0 f = 3 MHZ

z 0.4
W
n
0
0 .
: 0.2
W
N

-60 -40 -20 0 20 40 60

ANGULAR ACCEPTANCE, DEGREES

Fig. 14. Comparison of theoretical and experimental angle of

acceptance of tungsten-epoxy backed array element. Same
array as shown in figure 10.

The requirement for an array with a wide angle of acceptance is that

the individual elements must be thin and the elements themselves must be
designed so that their width is much less than their height so as to
avoid transverse resonances. Under these conditions the design of. the
array can be carried out by either using simple extensional mode theory
or by modifying it in a simple manner to take account of the coupling
between the transverse and extensional modes. Such considerations pre-
clude the use of almost square elements, although reasonable designs
can be carried out with elements which are wider than their height. In
this latter case, the design procedure would be very similar to that of
conventional piston transducers.

The problem of cross-coupling between elements has not been completely

solved. It is apparent that in order to obtain a wide angle of acceptance,
the cross-coupling should be below 30 dB, although reasonably satisfactory
results can be obtained with cross-coupling in the range of 20 to 25 dB.
Acoustic cross-coupling can occur through the front face of the transducer,
through coupling between the elements if there is a filling medium between
them and through the backing. The front face cross-coupling is best

207
 K I N 0 AND DESILETS

e l i m i n a t e d by e i t h e r u s i n g a v e r y t h i n l a y e r of f l e x i b l e material, such
a s mylar o r a f i l m of p l a s t i c ; a l t e r n a t i v e l y , t h e u s e of u r e t h a n e o r o t h e r
p l a s t i c media w i t h a n impedance v e r y c l o s e t o t h a t of water i s a s u i t a b l e
approach. A s f a r a s t h e backing i s concerned, t h e problem i s more d i f f i -
c u l t because p r o p a g a t i o n can occur a l o n g t h e backing l a y e r i n a mode of
p r o p a g a t i o n which i s l i k e t h a t of a p e r i o d i c system, i . e . , t h e backing
l a y e r loaded by t h e elements.

F i n a l l y , t h e r e are problems a s s o c i a t e d w i t h matching of t h e elements

i n t o t h e back l o a d i n g medium a s w e l l as i n t o t h e f r o n t l o a d i n g medium due
t o t h e f a c t t h a t t h e element i t s e l f i s much narrower t h a n a wavelength.
T h i s i m p l i e s t h a t t h e l o a d on t h e element i s n o t t h a t of a n i n f i n t e medium
and t h e e f f e c t must b e t a k e n i n t o a c c o u n t . F o r t u n a t e l y , i n most c a s e s of
p r a c t i c a l i n t e r e s t , t h i s d o e s n o t seem t o b e a major e f f e c t . But due t o
t h i s d i f f i c u l t y , s u c h e f f e c t s probably a l s o a c c o u n t f o r p a r t of t h e few
dB e x c e s s l o s s of t h e s e t r a n s d u c e r s o v e r t h e t h e o r e t i c a l v a l u e s .

Many of t h e s e i d e a s c a n b e e x t r a p o l a t e d t o t h e u s e of quarter-wave-
l e n g t h matching l a y e r s on t h e f r o n t of t h e e l e m e n t s . I n t h i s case, broad-
band and f a r more e f f i c i e n t o p e r a t i o n can b e o b t a i n e d . T h i s s u b j e c t w i l l
be d e a l t w i t h i n a l a t e r paper.

Acknowledgment

T h i s work w a s supported by t h e Advanced Research P r o j e c t s Agency

under Rockwell I n t e r n a t i o n a l S u b c o n t r a c t RI74-20773 and by t h e O f f i c e
of Naval Research under C o n t r a c t N00014-75-C-0632.

REFERENCES

[ l ] D e S i l e t s , C . S., F r a s e r , J . D., and Kino, G . S., The d e s i g n of

e f f i c i e n t broadband p i e z o e l e c t r i c t r a n s d u c e r s . IEEE T r a n s . S o n i c s
and U l t r a s o n . SU-25, 115-125 (1978).

[ Z ] Onoe, M., and T i e r s t e n , H. F., Resonant f r e q u e n c i e s of f i n i t e piezo-

e l e c t r i c ceramic v i b r a t o r s w i t h h i g h e l e c t r o m e c h e n i c a l c o u p l i n g ,
IEEE T r a n s . U l t r a s o n . Engin. 10, 32-39 (1973).

[ 3 ] Auld, B. A., A c o u s t i c F i e l d s and Waves i n S o l i d s , Vol. 1 (John Wiley

and Sons, New York 1973).

[ 4 ] J a f f e , H., and B e r l i n c o u r t , D. A., P i e z o e l e c t r i c t r a n s d u c e r materials,

Proc. IEEE 53, 1372-1386 (1965).

I51 Miller, G. F., and Pursey, H., The f i e l d and r a d i a t i o n impedance of

mechanical r a d i a t o r s on t h e f r e e s u r f a c e of a s e m i - i n f i n i t e i s o t r o p i c
s o l i d , Proc. Royal SOC. London 223, 521-541 (1954).

161 Auld, B. A., A c o u s t i c F i e l d s and Waves i n S o l i d s , V o l . 2, pp. 177-190

(John Wiley and Sons, New York, 1973).

[ 7 ] H a r r i n g t o n , R . F., Time-Harmonic E l e c t r o m a g n e t i c F i e l d s , pp. 180-186

(McGraw-Hill, New York, 1961).

[8] D e S i l e t s , C . S . , Transducer A r r a y s S u i t a b l e for A c o u s t i c Imaging,

Ginzton Laboratory Report 2833, S t a n f o r d U n i v e r s i t y , S t a n f o r d , C a l i f . ,
June 1978.
 TRANSDUCER ARRAYS WITH MATCHED BACKINGS

[9] Lees, S., Gilmore, R . S., and Kranz, P. R., A c o u s t i c p r o p e r t i e s of

t u n g s t e n v i n y l c o m p o s i t e s , IEEE T r a n s . S o n i c s and U l t r a s o n . SU-20,
1-2 (1973).

[ l o ] Z i t e l l i , L. T., S t a n f o r d U n i v e r s i t y , p r i v a t e communication.

1111 Swartz, R . G., Plummer, J . D., and Meindl, J . D . , Improved wedge t y p e

backing f o r p i e z o e l e c t r i c t r a n s d u c e r s , IEEE T r a n s . S o n i c s and U l t r a s o n .
SU-26, 140-142 (1979).

1121 Mason, W. P., P r o p e r t i e s of Polymers and N o n l i n e a r A c o u s t i c s , i n

P h y s i c a l A c o u s t i c s , V o l . 11, P a r t B , C h a p t e r s 8-9 (Academic P r e s s ,
New York, 1965).

[13] Dow L i q u i d Epoxy R e s i n s , The Dow Chemical Company, Midland, Michigan,

1969.

1141 Kino, G. S., D e S i l e t s , C. S., F r a s e r , J . D., and Waugh, T . M . , New

A c o u s t i c Imaging Systems f o r N o n d e s t r u c t i v e T e s t i n g , i n 1975 IEEE
U l t r a s o n i c s Symp. Proc. pp. 94-101 (IEEE C a t . No. 75 CHO994-4SU).

1151 Leung, W . , Shaw, H. J . , Kino, G . S., Winslow, D. K., and Z i t e l l i , L. T . ,

A c o u s t i c P h a s e C o n t r a s t Imaging w i t h E l e c t r o n i c Scanning, i n 1976
U l t r a s o n i c s Symp. P r o c . , pp. 151-156 (IEEE C a t . No. 76 CH1120-5SU).

[16] Leung, W. P., Shaw, H. J . , Kino, G. S., and Z i t e l l i , I.,. T., A New
Technique f o r High Speed S t r e s s D e t e c t i o n and Imaging Using a
100-Element A c o u s t i c Phased Array, i n 1977 IEEE U l t r a s o n i c s Symp.
P r o c . , pp. 255-258 (IEEE C a t . No. 77 CH1264-1SU).