M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 1

Ultrasonic transducers: From analytical modelling

to design optimisation and validation
Álvaro Pérez Dı́az, Centre for Complexity Science, University of Warwick
Supervisor: Dr Nishal Ramadas, Department of Physics, University of Warwick
Co-supervisor: Dr Gareth Alexander, Centre for Complexity Science, University of Warwick

Abstract—The most important part in an ultrasonic transducer Symbol list

is a piezoelectric substrate, which will accumulate electric charge • ξ, particle displacement (m)
in response to an applied mechanical stress. In the most typical 2
• S, area (m )
configuration, one layer of such a piezoelectric material will
be present altogether with some passive backing and matching • F , force, (N )
2
layers, situated in the back and in the front of the transducer, • Γ = F/S, stress (N/m )
respectively. • ∂ξ/∂x, strain
Complexity in ultrasonic transducers design has increased in • Y , elastic constant under conditions of constant electrical
the last decades: even with the most simple configuration there
is a huge variety of possible designs: backing layers, matching displacement (N/m2 )
3
layers, electrical load, piezoelectric elements, etc. All this makes • ρ, density (kg/m )
intuitive design very difficult. When it comes to model this • h, piezoelectric constant (V /m)
kind of device, there are three different choices: Finite Element • E, electric field (V /m)
(FE), equivalent circuit or analytical models. The latter provide • D, electrical displacement (C/m )
2
physical insight into the processes taking place in the transducer,
• ε, permittivity under conditions of constant strain (F/m)
as well as being fast and suited for optimization purposes.
We will have a look at two typical analytical models available • s, Laplace complex variable
in literature, afterwards we shall consider a third one, which has
to be modified to fit our purposes. Then some comparisons will II. G OVERNING EQUATIONS
be performed among the three analytical models, real laboratory
data, and a FE commercial software.
W
or not:
E shall distinguish two different kinds of materials,
depending on whether they are piezoelectrically active

I. I NTRODUCTION
• For non-piezoelectric materials: the two equations gov-

W E will start by considering the physics behind

piezoelectric and non-piezoelectric materials, having
a look at their governing equations and including some
erning their behaviour are

Γ=Y
 ∂ξ
(1a)
∂x
assumptions that will let us derive our models. All the above 2
altogether with some technical details that will constitute the  ρ ∂ ξ = ∂Γ

(1b)
framework for the first two models will shape section 2. ∂t2 ∂x
[1], [2], [3]. where (1a) is Hooke’s law between stress and strain, and
(1b) is second Newton’s law for an infinitesimal volume
In the third and fourth sections we describe and analyze element.
the models given in [2]-[5] respectively. The first one is the If we now combine both equations after taking partial
simplest model and will provide clear distinction between derivative with respect to x in (1a):
receiving and transmitting modes, as well as being a very ∂Γ ∂2ξ ∂2ξ ∂2ξ ∂2ξ ∂2ξ
=Y =⇒ ρ 2 = Y =⇒ 2 = v 2 2
good introduction to more complicated setups. The second ∂x ∂x2 ∂t ∂x 2 ∂t ∂x
one introduces arbitrary layer possibilities providing better Y
behaviour. where v 2 = . This is the equation waves travelling in
ρ
this non-piezoelectric material must satisfy.
Section 4 introduces a model based on [6] that uses a • For piezoelectric materials: the equations are similar but
slightly different framework than the previous two ones, in this case we have to include the piezoelectric behaviour
all the required tools will be detailed there. Originally it of the material, represented by its piezoelectric constant
considered a different setup which is of no use for us, so h, obtaining the set of equations:
we had to modify and redo the calculations to fit our purposes. 
∂ξ

 Γ=Y − hD (2a)

 ∂x
To finish, several results and experiments are described, 

∂ξ
providing evidence of the capabilities that this analytical E = −h + D/ε (2b)
 ∂x
models have when it comes to model real world ultrasonic 
2
 ρ ∂ ξ = ∂Γ



transducers. (2c)
∂t2 ∂x
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 2

again, if we take partial derivative of (2a) with respect to combining this results we obtain
x and combine it with (2c), we get h i
F + hQ = sZc −Ae−sx/v + Besx/v
∂Γ ∂2ξ ∂D ∂2ξ Y ∂2ξ h ∂D
=Y 2
−h =⇒ 2 = −
∂x ∂x ∂x ∂t ρ ∂x2 ρ ∂x
If we are considering a thickness mode, this is, we assume We can also find the voltage across the piezoelectric trans-
the transducer to vibrate predominantly in one dimension ducer by integrating (2b) from x = 0 to x = L (all the
(thickness dimension), we can perform the following thickness of the transducer):
derivation:
∂D ∂D Z L Z L
ξ D

= =0 V = Edx = −h + dx =
∂y ∂z ∂x ε
0 0
and using Gauss’ law together with the fact that there Q
is no free charge inside the transducer (all the possible = −h [ξ(L) − ξ(0)] +
C0
charge would be located at the electrodes) yields to
F where C0 = Sε/L is the clamped or static capacitance of
∂D
∇ · D = ρinside = 0 =⇒ =0 the transducer. We can rewrite our expression as follows
∂x
hence in our previous equation we can get rid of the last sC0 Ze
V = −h [ξ(L) − ξ(0)] U , where U =
term: 1 + sC0 Ze
∂2ξ Y ∂2ξ Y
2
= , where v 2 = (3) being Ze the impedance of an arbitrary electrical load
∂t ρ ∂x2 ρ
connected across the transducer electrodes.
which is an identical wave equation to the one obtained
for non-piezoelectric materials. Note that if we had multiple piezoelectric layers, the overall
To solve the wave equation it is useful to use the Laplace voltage across the whole transducer would be the sum of the
transform, together with the assumption that all functions are correspondent voltages for each of the layers
zero at t = 0 Z b1 Z bn
Z ∞
L[f (t)] = f (t) = e−st f (t)dt V = E1 dx + . . . + En dx
0 a1 an

we shall apply it to our wave equation to obtain

III. L INEAR S YSTEMS M ODEL (LSM)
s2 ∂ 2 ξ(s, x)
2
ξ(s, x) =
v ∂x2
and its general solution will be Loading medium Piezoelectric Backing layer
−sx/v sx/v
ξ(s, x) = Ae + Be , A, B ∈ R A1 A A2
∞ ∞
From now on, for the rest of this chapter and in the two
following ones, we shall use Laplace transformed functions, B1 B
even if we omit the over line sign in some functions for the Z1 ,v1 Zc ,vc Z2 ,v2
sake of clarity. x=0 x=L
Fig. 1: Basic setup for a transducer with an infinite backing layer
We can now relate the total force exerted over an area S (at x = 0) and in contact with a loading medium at its front face
normal to the x direction, F = ΓS (x = L).
• For a non-piezolectric material, using (1a)

F s h
T HE boundary conditions that apply for the wave equation
i
Γ= = Y −Ae−sx/v + Besx/v =⇒ solution are: continuity of particle displacements, ξ and
S v
h i continuity of forces across the interface, F , at the front and
=⇒ F = Zs −Ae−sx/v + Besx/v (4) rear face, resulting in four boundary equations:
where Z = ρvS is the characteristic acoustic impedance 
of the material (note that in some texts the cross sectional 
ξ1|x=0 = ξc|x=0

ξ
area S is omitted). c|x=L = ξ2|x=L
• For a piezoelectric material, using (2a) 
F1|x=0 = Fc|x=0
sh

Fc|x=L = F2|x=L
i 
Γ=Y −Ae−sx/v + Besx/v − hD
v
and applying again Gauss’ law to the surface of the where each of the expressions has the following form:
piezoelectric material (in our case, where the electrodes 
−sx/v1
are located, this is front and rear faces) ξ1 (s) = A1 e
 + B1 esx/v1
Z Z
Q ξc (s) = Ae−sx/vc + Besx/vc
D · ds = DS = ρcharge = Q =⇒ =D 

ξ2 (s) = A2 e−sx/v2
S V S
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 3

 Loading medium Piezoelectric Backing layer

−sx/v1
+ B1 esx/v1


F1 (s) = sZ1 −A1 e
 A1 A A2
Fc (s) + hQ(s) = sZc −Ae−sx/vc + Besx/vc


& ∞ ∞
F2 (s) = sZ2 −A2 e−sx/v2



B
hence the resulting equations are: Z1 ,v1 Zc ,vc Z2 ,v2
x=0 x=L
A1 + B 1 = A + B (5a)

 Fig. 2: Basic setup for a transducer in recevier mode with an infinite
Ae−sL/vc + BesL/vc = A2 e−sL/v2



 (5b) backing layer (at x = 0) and in contact with a loading medium at its
front face (x = L)


sZ1 (−A1 + B1 ) = sZc (−A + B) − hQ(s) (5c)

 
−sL/vc sL/vc


 sZ c −Ae + Be =
(5d) A. Receiver mode


  
 = sZ −A e−sL/v2 − hQ(s)

2 2
In receiver mode, there is no emitterd left going wave, we
altogether with the expression for the voltage will just consider the incident wave (A1 ).
h i Remember the equation for the force (4), together with the
V = −h A(e−sL/vc − 1) + B(esL/vc − 1) U (s) (6) previous assumptions:
h i
Now, combining (5a) and (5c) we obtain F1 = Z1 s −A1 e−sx/v1 + B1 esx/v1 =⇒
hQ(s) −F1
(1 − RF )A1 = A − BRF + (7) =⇒ F1 = −A1 Z1 s =⇒ A1 = −
s(Zc + Z1 ) Z1 s
Zc − Z1 Also replace Q = −V /sZe , which follows from Ohm’s law
where RF =
Zc + Z1 Z = V /I, to obtain for the voltage
and similarly from (5b) and (5d) hA1 KF (1 − RF )U (s)
sL/vc V =  
hQ(s)e KF KB U (s)
Ae−2sL/vc RB − B = − (8) 1 − h2 +
s(Z2 + Zc ) Zc + Z1 Zc + Z2 s2 Ze
Zc − Z2 −F1
where RB = and substituting A1 = − yields to:
Zc + Z2 Z1 s
Both RF and RB are the reflection coefficients for the
V −hKF TF U (s)/sZc
waves travelling into the piezoelectric medium. =   (12)
F1 KF TF KB TB U (s)
Now combine equations (7) and (8) and solve for A and B, 1 − h2 +
using for example Cramer’s rule, to obtain: 2 2 s2 Ze Zc

hQ(s)RF e−sL/vc which is the transfer function relating the transforms of

 
hQ(s)
+ (RF − 1)A1 − received voltage to incident force.
s(Zc + Z1 ) s(Z2 + Zc )
A= 
e−2sL/vc RB RF − 1
(9) We can represent this transfer function using a block
diagram with two positive feedback loops, see figure 3.
hQ(s)e−sL/vc
− − (1 − RF )A1 RB e−2sL/vc + If we replace RB by RF and vice versa, we obtain a
s(Zc + Z2 ) transfer function for the force applied in the rear face of the
hQ(s)RB e−2sL/vc transducer. However, as we assume a infinitely long backing
+
s(Zc + Z1 ) block and hence no returning wave, there will not be any rear
B=  (10)
e−2sL/vc RB RF − 1 face applied force.
We can now find the voltage across the transducer, plugging
(9) and (10) into (6) to obtain B. Transmitter mode
This is an analogous situation, where now there is no
  
hQ
V = h KF A1 (1 − RF ) − incident wave (see figure 4). Setting A1 = 0 into (11) yields
s(Zc + Z1 )
 (11) to
hQ hQTF hQTB RF e−sL/vc
−KB −
s(Zc + Z2 ) 2Zc s 2sZc
A=
We shall now develop the receiver and transmitter models, e−2sL/vc RB RF − 1
using the framework just detailed. This has the advantage of hQTB e−sL/vc hQTF RB e−2sL/vc
showing very clearly both functioning modes, which are quite − +
2sZc ) 2sZc )
independent from each other. B=
e−2sL/vc RB RF − 1
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 4

Force input at Z2
the front face
h/2sZE FB (s) = F|x=L = −hQ(s)KB (16)
Zc + Z2
F1
+
+
TF 1/sZc
The final step will be, as we did for the receiver mode,
to find a transfer function relating the force generated at the
KF Voltage
front face to the voltage across the transducer. Notice that we
output! have obtained forces in both rear and front faces, but we are
+ V
+
h U −1 primarily interested in the front face.

KB Consider the following electrical setup:

TB 1/sZc
Z0

h/2sZE

Fig. 3: Receiver transfer function e ZE ZT

IT
Loading medium Piezoelectric Backing layer

A A2 Fig. 5: Circuit schematics for the transmitter mode

∞ B1 ∞
B where Zo is the output impedance of the non-ideal voltage
source used, and ZE is an arbitrary electrical load. Using
Z1 ,v1 Zc ,vc Z2 ,v2
circuit analysis we obtain the following relation:
x=0 x=L
IT (s) ZE ZE /(Zo + Ze )
Fig. 4: Basic setup for a transducer in transmit mode with an infinite = = =
backing layer (at x = 0) and in contact with a loading medium at its e(s) ZT (Zo + ZE ) + Zo ZE ZT + Zo ZE /(Zo + ZE )
front face (x = L) a(s) a(s)e(s)
= =⇒ IT (s) = =⇒
ZT + b(s) ZT + b(s)
IT (s) a(s)e(s)
which plugged into (6) result in =⇒ Q = =
s ZT + b(s)
h2 Co
 
Q
V = 1− (TF KF + TB KB ) (13) where
Co 2sZc ZE Zo ZE
a(s) = & b(s) =
From this expression we can determine the operational Zo + ZE Zo + ZE
impedance of the transducer, ZT : We can now use the new expression for Q(s) and (14) into
(15) to obtain
h2 Co
 
V (s) 1
ZT = = 1− (TF KF + TB KB ) (14)
sQ(s) sCo 2sZc FF (s) h a(s)Z1 /(Zc + Z1 )KF (s)Y (s)
=−
e(s) 1 − h2 Y (s)/(sZc ) (TF KF /2 + TB KB /2)
The next step is to find the force at any point within the
(17)
transducer. Recall the expression for the force in a piezoelec-
which is the desired transfer equation, with
tric layer
  Y (s) = (1 + sb(s)Co )/Co .
Fc (s) + hQ(s) = sZc −Ae−sx/vc + Besx/vc
In a completely similar fashion we can obtain the transfer
and using our values for A and B: function for the back face:
FB (s) h a(s)Z2 /(Zc + Z2 )KB (s)Y (s)
=−

1
F (s) = − hQ(s) 1 + · e(s) 1 − h2 Y (s)/(sZc ) (TF KF /2 + TB KB /2)
2s(e−2sL/vc RB RF − 1)
h (18)
· (TF − TB RF e−sL/vc )e−sx/vc + This transfer function can be pictured again in a block
io diagram structure, see figure 6.
+(TB e−sL/vc − TF RB e−2sL/vc )esx/vc

We are now particularly interested in the front and back IV. L ATTICE MODEL
faces of the transducer. Thus, set x = 0 and x = L ONSIDER the setup outlined in figure 7, where medium
respectively:
Z1
C 2, with a thickness l2 , is located between media
1 and 3, which are semi-infinite. We decompose forces
FF (s) = F|x=0 = −hQ(s)KF (15) into their backward Bn and frontward Fn components. The
Zc + Z1
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 5

the the first and the second coming from (20a) and the last
h/sZc TF /2
Front face
two ones from (20b). This can be written in matrix form:
force output!
    
FF F20 1 + R12 −R12 F10
KF AF /2 −1 =
Voltage input B10 R12 1 − R12 B20
e +
a + Y h     
+
F30 1 + R23 −R13 F2l2
=
B2l2 R13 1 − R13 B30
KB AB /2 −1
FB
Back face
where
force output! Z2 − Z1 Z3 − Z2
h/sZc TB /2 R12 = & R23 =
Z2 + Z1 Z3 + Z2

Fig. 6: Transmitter transfer function are the reflection coefficients for waves of force travelling into
the piezo at the 1-2 and 2-3 interfaces respectively.
Medium 1,Z1 Medium 2, Z2 Medium 3, Z3
F1 F2 F3 A. Transducer model
(Af 3 ) (Af 1 ) (Af 2 )
(Bb3 ) (Bb1 ) (Bb2 )
Now consider a similar disposition, placing the piezoelectric
layer between two passive layers.
B1 B2 B3
x1 x2 x3
∞ ∞
Back face medium, Z1 Transducer, Zt Front face medium, Z−1
Fig. 7: Basic setup for a number of passive layers. Displacements are F−1 Ft F1
bracketed. (Af −1 ) (Af t ) (Af 1 )
(Bb−1 ) (Bbt ) (Bb1 )
same for the displacements, backward Bbn and frontward Aan . B−1 Bt B1
x−1 xt x1
∞ ∞
Vt
Again, using complex Laplace transforms and the same It
analysis as for the systems model, we obtain: ZE ∼
( VE
ξ n = Af n e−sxn /vn + B bn esxn /vn (19a)
Fig. 8: Piezolectric system. Displacements are bracketed.
Γn = sZn [−Af n e−sxn /vn + B bn esxn /vn ] (19b)
Define now the force components at position xn to be The voltage across the transducer will be
F n = −sZn Af n & B n = sZn B bn h i Q
t
Vt = −h (ξt )|xt =lt − (ξt )|xt =0 + (21)
so we can rewrite (19a) and (19b) as Co

ξ n = [F n e−sxn /vn + B n esxn /vn ]/sZn ] (20a) where Qt is the total electrical charge in the electrodes.
Γn = F n e−sxn /vn + B n esxn /vn (20b)
The mechanical boundary conditions are
We shall drop the over lines to make notation clear, recall (
that for this model everything is done in complex Laplace ξ−1|x−1 =0 = ξt|xt =0 & ξt|xt =lt = ξ1|x1 =0
domain. Consider now the boundaries xn = 0 and xn = ln of Γ−1|x−1 =0 = Γt|xt =0 & Γt|xt =lt = Γ1|x1 =0
this n-th layer:
n−sx /vn
Fn0 = Fn e|xn =0
sx /v
& Bn0 = Bn e|xnn=0n that is

−sx /v
n n sx /v
Fnln = Fn e|xn =l & Bnln = Bn e|xnn=lnn 
n 
 (−F−10 + B−10 ) /Z−1 = (−Ft0 + Bt0 ) /Zt
hence we can write for the model in the figure

(−F + B ) /Z = (−F + B ) /Z
tlt −tt t 10 10 1
   −sT  
F2l2 e 2 0 F20 F−10 + B−10 = Ft0 + Bt0 − hQt
=


B2l2 0 esT2 B20

Ftlt + Btlt − hQt = F10 + B10 − hQt


where T2 = l2 /v2 is the transit time for waves travelling    −sT  

across the layer. Ftlt e t 0 Ft0
& =
Btlt 0 esTt Bt0
The boundary conditions result as follows
 and in matrix form we obtain
(−F10 + B10 )/Z1 = (−F20 + B20 )/Z2
    
 B−10 R−1 1 − R−1 F−10

(−F + B )/Z = (−F + B )/Z = +
2l2 2l2 2 30 30 3 Ft0 1 + R−1 −R−1 Bt0
F10 + B10 = F20 + B20   (22)
 h R−1 − 1
 + Qt
F2l2 + B2l2 = F30 + B30

2 R−1 + 1
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 6

The coefficients are found to be

     P11 ={(R1 − R−1 e−2sT ) + U (1 − e−sT )[(e−sT − 1)·

F10 R1 1 − R1 B10
= + · (1 + R1 R−1 ) + 2(R−1 e−sT − R1 )]}/PD
Btlt 1 + R1 −R1 Ftlt
  (23) P12 ={(1 − R1 )(1 + R1 )e−sT + U (1 − e−sT )(1 + e−sT )·
h R1 − 1
+ Qt · (R1 − 1)(R−1 + 1)}/PD
2 R1 + 1
P13 = G(1 − e−sT )(R1 − 1)(1 − R−1 e−sT )/PD
where P21 ={(1 − R−1 )(1 + R−1 )e−sT + U (1 − e−sT )·
· (1 + e−sT )(R−1 − 1)(R1 + 1)}/PD
Zt − Z1 Zt − Z−1
R1 = & R−1 = P22 ={(R−1 − R1 e−2sT ) + U (1 − e−sT )[(e−sT − 1)·
Zt + Z1 Zt + Z−1
· (1 + R1 R−1 ) + 2(R1 e−sT − R−1 )]}/PD
P23 = G(1 − e−sT )(R−1 − 1)(1 − R1 e−sT )/PD
are the usual reflection coefficients for waves of force coming
into the piezoelectric medium. P31 = G(−sZE /Zt )(1 + R1 )(1 − e−sT )(1 − R−1 e−sT )/PD
We need to obtain now a relationship between the source P32 = G(−sZE /Zt )(1 + R−1 )(1 − e−sT )(1 − R1 e−sT )/PD
voltage and charge, so we can then relate the applied voltage to P33 ={(1 − R1 R−1 e−2sT )/(1 + sZE Co ) − U (1 − e−st )·
the one across the transducer. Looking at the electric diagram
· [(1 + R1 )(1 − R−1 e−sT ) + (1 + R−1 )(1 − R1 e−sT )]
in figure 8:
/PD }

Vt = VE − It ZE and It = sQt (24) where

h2 Co
U= &
2sZt (1 + sCo ZE )
which plugged into (21) gives
PD = {(1 − R1 R−1 e−2sT ) − U (1 − e−sT )[(1 + R1 )·
Co

h
 ·(1 − R−1 e−sT ) + (1 + R−1 )(1 − R1 e−sT )]}
−sTt


Q(t) = VE + 1−e (Ft0 + Btlt )
1 + sCo ZE sZt
(25)
B. The multilayered acoustic lattice
Now using (24) ad (25) we can obtain
Consider n + 1 pasive layers as pictured in the figure:

VE − Vt VE Layer 1 Layer 2 Layer i Layer i + 1 Layer n Layer n + 1

= Qt =⇒ Vt = −
sZE 1 + sCo ZE F1 F2 Fi Fi+1 Fn Fn+1
··· ···
B1 B2 Bi Bi+1 Bn Bn+1
0 x1 l1 0 x2 l2 0 xi li 0 x li+1 0 xn ln 0 x li+1
i+1 n+1

Co sZE 1
− h(1 − e−sTt )(Ft0 + Btlt ) (26) Fig. 9: Pasive multilayer setup, forces F & B.
1 + Co sZE sZt
Consider the interface between layers i & i+1, the boundary
conditions are
Equations (22), (23) and (26) completely control the be-
(
(−Fili + Bili )/Zi = (−Fi+1,0 + Bi+1,0 )/Zi+1
haviour of the transducer: we have three input parameters:
Fili + Bili = −Fi+1,0 + Bi+1,0
F−10 , B10 and VE , altogether with three output parameters:
F10 , B−10 and Vt . Then we can develop a transfer function hence
from input to output (
− Fi+1,0 + Bi+1,0 = Zi+1 /Zi (−Fili + Bili )

F10

B10
  Fi+1,0 + Bi+1,0 = Fili + Bili
B−10  = (Pi,j ) F−10 
if we add and subtract both equations we obtain the following
Vt Ve
    
Fi+1,0 1 1 + Zi+1 /Zi 1 − Zi+1 /Zi Fili
=
where
Bi+1,0 2 1 − Zi+1 /Zi 1 + Zi+1 /Zi Bili

  together with
P11 P12 P13
e−sTi
    
P = P21 P22 P23  Fili 0 Fi,0
=
P31 P32 P33 Bili 0 esTi Bi,0
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 7

This allows us to rewrite the expression as where the coefficients are:

B F F F F

Fi+1,0

1 (1 + Zi+1 /Zi ) e−sTi (1 − Zi+1 /Zi )esTi
  W11 = {U21 (U11 U22 − U12 U21 )(P12 P21 − P11 P22 )+
= · F F F B F
Bi+1,0 2 (1 − Zi+1 /Zi )e−sTi (1 + Zi+1 /Zi )esTi + P11 U11 U22 + U12 (1 − P22 U21 − P11 U21 )}/WD
  F B
Fi,0 W12 = U11 U22 P12 /WD
·
Bi,0 F
 B
W13 = U11 U21 (P12 P33 − P13 P22 ) + P13 /WD
B F
We shall call the previous matrix Ti+1 /i the transmission W21 = U11 U22 P21 /WD
F B B B B
matrix for the i : i + 1 interface: W22 = {U21 (U11 U22 − U12 U21 )(P12 P21 − P11 P22 )+
B B B F B
Ti+1 1 (1 + Zi+1 /Zi ) e−sTi (1 − Zi+1 /Zi )esTi
  + P22 U11 U22 + U12 (1 − P11 U21 − P22 U21 )}/WD
=
i 2 (1 − Zi+1 /Zi )e−sTi (1 + Zi+1 /Zi )esTi B
W23 = U11 F
{U21 (P21 P13 − P11 P23 ) + P23 }/WD
F B B F

W31 = P31 U22 (1 − U21 P22 ) + P31 U21 P21 U22 /WD
Hence we can represent a multilayered structure using
B F F B

transmission matrices: W32 = P32 U22 (1 − U21 P11 ) + P32 U21 P12 U22 /WD
F B
Tn+1 Tn+1 Tn T3 T2 W33 = {P33 + U21 (P31 P13 − P11 P33 ) + U21 (P23 P32 −
= ··· F B
1 n n−1 2 1 − P22 P33 ) + U21 U21 (P31 (P12 P23 − P13 P22 )+
where the matrix Tn+1 /1 relates the force components F(n+1)0 + P32 (P21 P13 − P23 P11 )P33 (P11 P22 − P12 P21 ))}/WD
& B(n+1)0 to F10 & B10 . F
with WD = (1 − U21 B
P11 )(1 − U21 F B
P22 ) − U21 U21 P12 P21 .
This can be expressed in matrix form as follows:
     We have then reduced all our multilayer system to a
F(n+1)0 T11 T12 F10
= three inputs-outputs system. We can see that we recover the
B(n+1)0 T21 T22 B10
receiver and transmitter behaviour: for the receiver we will be
or interested in W31 , as we receive the input force B(n+1),0 on
     the front face and all other inputs are set to zero. Similarly, for
F(n+1)0 U11 U12 F10
= = the transmitter we will look at W13 , as we assume no input
B10 U21 U22 B(n+1)0
   (27) forces and we drive our device applying a voltage VE across
T11 − (T21 T12 )/T22 T12 /T22 F10 its electrodes. Similarly, W33 relates the applied voltage to the
=
−T21 /T22 1/T22 B(n+1)0 voltage found across the transducer.

C. The multilayered transducer model V. H UANG ’ S MODEL OF MULTILAYERED ULTRASONIC

TRANSDUCERS , MODIFIED WITOUTH THE INVERSION
Now we shall put all the previous derivations together in LAYER
the following setup, see figure 10.
Backing layers Piezoelectric Matching layers
m backface layers Piezoelectric n frontface layers db1 da da1
F−(m+1),0 F−1,0 F1,0 F(n+1),0 Load
... ...
medium
··· ··· bN b1 a a1 aM
B−(m+1),0 B−1,0 B1,0 B(n+1),0
Zc ,vc
Vt
Fig. 11: Setup for an arbitrary number of pasive layers.
ZE ∼
VE
ONSIDER the multilayered setup detailed in figure 11.
Fig. 10: The multilayer transducer model, putting all the previous
situations together. C As explained in the introduction, this model uses a
different framework, we no longer use Laplace transformed
In this case, using (27) for the front and back faces: functions. Consider the interface between a and b1, keeping in
  F mind that b1 is a passive layer and a is the active piezoelectric
F
  
F(n+1)0 U11 U12 F10 layer. The expressions for the particle velocities and pressures
= F F
B10 U21 U22 B(n+1),0 are:


B−(m+1)0
  B
U11 B
U12

B−10
 At the left boundary of layer a:
= B B
F−10 U21 U22 F−(m+1),0
(
ua|LB = Aa eika da /2 + Ba e−ika da /2
what can be reexpressed as: pa|LB = za Aa eika da /2 − za Ba e−ika da /2 + hD
    At the right boundary of layer b1:
F(n+1)0
B(n+1),0 (
B−(m+1)0  = Wij F−(m+1),0  ub1|RB = Ab1 e−ikb1 db1 /2 + Bb1 eikb1 db1 /2
Vt VE pb1|RB = zb1 Ab1 e−ikb1 db1 /2 − zb1 Bb1 eikb1 db1 /2
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 8

And the boundary conditions impose that both velocities In the very front and back layers there is no incoming
and pressures must be the equal: and returning wave (respectively), this is AbN = BaM = 0,
resulting in:
z0 d
 
Aa eika da /2 + Ba e−ika da /2 =

  A  0 
(28a) HaM  aM hV  AaM


=Ab1 e−ikb1 db1 /2 + Bb1 eikb1 db1 /2 xaM = = =

A0aM

FaM z0 d

 

ika da /2 −ika da /2
 AaM
 za A a e − B a e + hD = hV


   (28b) 
z 0d

−ikb1 db1 /2
− Bb1 eikb1 db1 /2 B

 =z
b1 Ab1 e
   0 
HbN bN BbN
xbN = =
 hV 
= 0
FbN z0 d  −BbN
we will denote from now on αj = kj dj . −BbN
hV
Now, the voltage across the only active layer of the trans-
Due to the form of the equations, we can now reduce them
ducer looks like
using transmission matrices, the procedure is detailed below.
Z da /2

We shall start with the front layers: consider first equation
V = E dx = −h ξda /2 − ξ−da /2 + D/ε da = (31d):
−da /2
Ma G− +
a xa + s = Ma1 Ga1 xa1
h  +iαa /2 
= e − e−iαa /2 (Aa − Ba ) + D/ε da
iω from the next equation we can write
(29)
xa1 = (Ma1 G−
a1 )
−1
(Ma2 G+
a2 )xa2
and rearranging we can find the expression for D:
Similarly, from the following one
εV 2hε
D= − sin (αa /2) (Aa − Ba ) xa2 = (Ma2 G−
a2 )
−1
(Ma3 G+
a3 )xa3
da ωda
Thus, following Huang’s calculations, we can now represent and so on. Substituting the obtained expressions we get to:
− −1
equations (28a) and (28b) in the following matrix form Ma G− +
a xa + s = Ma1 Ga1 [(Ma1 Ga1 ) (Ma2 G+
a2 )]·

1 0

cos (−αb1 /2) i sin (−αb1 /2)

Hb1
 ·[(Ma2 G−
a2 )
−1
(Ma3 G+ −
a3 )] . . . [(Ma(M−1) Ga(M−1) )
−1
·
0 = − −1
0 zb1 i sin (−αb1 /2) cos (−αb1 /2) Fb1 ·(MaM G+ +
aM )]xaM = [(Ma1 Ga1 )(Ma1 Ga1 ) ]·
− −1
·[(Ma2 G+ ] . . . [(Ma(M−1) G+
   
1 0 cos (−αa /2) i sin (−αa /2) Ha a2 )(Ma2 Ga2 ) a(M−1) )
= +
0 za0 i sin (−αa /2) cos (−αa /2) Fa
  (Ma(M−1) G−
a(M−1) )
−1
]MaM G+
aM xaM =
0
+
1 − kT2 µa Fa (T+ + +
a1 . . . Ta(M−1) )MaM GaM xaM =
+
(30) = T+ +
a MaM GaM xaM = Ma Ga xa + s

where T+ a is the transmission matrix from the front layer to the

h2 
where kT2 = , µa = 2/αa sin (αa /2) , transducer front face. Each of the single transmission matrices
Y is as follows
Ha = z0 da /hV (Aa + Ba ) and Fa = z0 da /hV (Aa − Ba )
−

−1
T+ +
am = Mam Gam Mam Gam =
It is easy to check that we recover equations (28a) & (28b)    
1 0 cos(αm /2) i sin(αm /2)
if z0 = Y /va . Rewrite now equation (30) as 0 · ·
0 zam i sin(αm /2) cos(αm /2)
Mb1 G− + −1  −1
b1 xb1 = Ma Ga xa + s

cos(−αm /2) i sin(−αm /2) 1 0
· 0 =
i sin(−αm /2) cos(−αm /2) 0 zam
and repeating this calculations for all the interfaces we
cos2 (αm /2) − sin2 (−αm /2) 2i/zam
0
 
obtain cos(αm /2) sin(αm /2)
= =
0
2izam cos(αm /2) sin(αm /2) cos2 (αm /2) − sin2 (−αm /2)
MbN G− +
bN xbN = Mb(N−1) Gb(N−1) xb(N−1) (31a)

0
 
 cos(αm /2) i/zam sin(αm /2)
=

0

 . i/zam sin(αm /2) cos(αm /2)
 ..Mb2 G− +

b2 xb2 = Mb1 Gb1 xb1 (31b)


 (32)
− +
Mb1 Gb1 xb1 = Ma Ga xa + s (31c)
 − + where we have used the following trigonometric identities
Ma Ga xa + s = Ma1 Ga1 xa1 (31d)

cos2 x − sin2 x = cos(2x) and sin x cos x = sin(2x)/2.





..
.Ma(M−1) G− +

a(M−1) xa(M−1) = MaM GaM xaM(31e)
We can perform the same calculations for the backing
which are 2·(N +M ) linear equations, and G− +
bN = GaM = Id layers, starting with (31c)
& AbN = BaM = 0.
Mb1 G− +
b1 xb1 = Ma Ga xa + s
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 9

from the previous equation we can write a32 = −[Tb− (1, 1) + Tb− (1, 2)zbN
0
]
xb1 = (Mb1 G+
b1 )
−1
(Mb2 G−
b2 )xb2 a42 = −[Tb− (2, ) + Ta+ (2, 2)zbN
0
]
Similarly, from the previous one
xb2 = (Mb2 G+ −1
(Mb3 G− Thus we can finally write
b2 ) b3 )xb3

−c− −is−
   0   
and so on. Substituting the obtained expressions we get to: a11 0 a a AaM 0
a21 0 −is− a −c+ 2   0  1
a + kT µa   BbN 
c+ is+  ·  Ha  = 0
  
−
 0 a32
Ma G+ + −1
a xa + s = Mb1 Gb1 [(Mb1 Gb1 ) (Mb2 G−
b2 )]·
a a
0 a42 is+
a c+ 2
a − kT µa Fa 1
· [(Mb2 G+
b2 )
−1
(Mb3 G− +
b3 )] . . . [(Mb(N−1) Gb(N−1) )
−1
·
· (MbN G− − + −1 where c± ±
a = cos(±αa /2) and sa = sin(±αa /2).
bN )] · xbN = [Mb1 Gb1 (Mb1 Gb1 ) ]·
· [Mb2 G− + −1
b2 (Mb2 Gb2 ) ] . . . [(Mb(N−1) G−
b(N−1) ) We can easily solve this system using for example Cramer’s
(Mb(N−1) G+
b(N−1) )
−1
]MbN G−
bN xbN = rule.
= (T− − −
b1 . . . Tb(N−1) )MbN GbN xbN =
− To finish with this model, we can write the expression for
= T+ +
b MbN GbN xbN = Ma Ga xa + s the transducer impedance as follows: consider equation (29)
where T− bn is the transmission matrix from the back layer h  +iαa /2 
to the transducer rear face. Each of the single transmission V = e − e−iαa /2 (Aa − Ba ) + D/ε da =⇒
iω
matrices is as follows, in a completely analogous fashion as Dda

h2 ε 2 zo da

we did before: =⇒ =V 1− sin(αa /2) (Aa − Ba )
ε Y αa hV ε
T− −
b = (Mbn Gbn )(Mbn Gbn )
+ −1
=
Dda εV 
= V 1 − kT2 µa Fa =⇒ D = 1 − kT2 µa Fa
  
0 =⇒
 
cos(−αn /2) i/zan sin(−αn /2) (33)
= 0
ε da
i/zan sin(−αn /2) cos(−αn /2)
thus the electrical current through the transducer is
εV
Using (32) and (33) we can rewrite equations (31a)-(31e) I = iωDS = iωS (1 − kT2 µa Fa ) = iωCo (1 − kT2 µa Fa )V
da
in the more compact matrix form
( and the impedance
T− − +
b MbN GbN xbN = Ma Ga xa + s
Ma G− + + V 1 1
a xa + s = Ta MaM GaM xaM Z= = 2 (35)
I iωCo 1 − kT µa Fa
−
but Ma , G+ aM and GbN can be set to the identity matrix,
yielding to VI. A NALYTICAL RESULTS
(
T− +
b MbN xbN = Ga xa + s
− +
Ga xa + s = Ta MaM xaM
(34)
I N the three previous sections we have introduced and
defined three different analytical models that varied in
capabilities and modelling techniques. For a considered
where transducer, all will provide analyical results for the
A0aM 0
   
BbN
xaM = & xbN = transmitter and receiver force-voltage relations and the
A0aM −BbN0
electrical impedance behaviour.

As a last step, we can write (34) as a single linear system To test our models against FEM and real data from the
Ax = b. From (34) we obtain: laboratory, we have considered three test cases whose specifi-
 cations can be found in the following table:
0
−a32 BbN − [cos(αa /2)Ha + i sin(αa /2)Fa ] − 0 = 0

 0 Test Case 1 Test Case 2 Test Case 3
 −a42 BbN − [i sin(αa /2)Ha + cos(αa /2)Fa ] −


 Material PZT PZT PZT
− 1 − kT2 µa Fa = 0

  
Dim 25x25mm 20x20mm 20x20mm

 a11 A0aM − [cos(−αa /2)Ha + i sin(−αa /2)Fa ] − 0 = 0 Thickness 0.5mm 1mm 2mm
−a21 A0aM − [i sin(−αa /2)Ha + cos(αa /2)Fa ] −




 The three transducers are made of Lead zirconate titanate
− 1 − kT2 µa Fa = 0

  
(PZT), one of the most common piezoelectric materials. Notice
where the lateral dimensions-thickness ratio for the three of them,
beginning with a reasonably high value and decreasing af-
0
a11 = [Ta+ (1, 1) + Ta+ (1, 2)zaM ] terwards. As the considered analytical models are 1D, we
will check how well this approximation holds for real world
0
a21 = [Ta+ (2, 1) + Ta+ (2, 2)zaM ] devices.
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 10

A. Transducer impedance Test Case 3 (PZT5A) - 20mmx20mm, 2mm thickness

1
Lab data
The results for the impedance modulus are shown in figures 0.9 FEM
12,13 & 14. The spike on the right hand side of the plot cor- Lattice

Normalized impedance modulus (Ohms)

0.8 Huang
responds to the main resonance region, while the descending LSM
curve on the left hand side corresponds to the lateral resonance 0.7
behaviour. 0.6

0.5
Test Case 1 (PZT5A) 25mmx25mm, 0.5mm thickness
1
0.4
Lab data
0.9 FEM
Lattice 0.3
Normalized impedance modulus (Ohms)

0.8 Huang
LSM 0.2
0.7
0.1
0.6
0
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
0.5 Frequency (Hz) ·106

0.4 Fig. 14: Electrical impedance modulus for test case 3

0.3

0.2 higher frequencies, agreeing as well with the laboratory data.

0.1
Similarly, we show below the results for the impedance
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 phase (figures 15, 16 & 17), we can see good agreement
Frequency (Hz) ·106
predicting the peak area.
Fig. 12: Electrical impedance modulus for test case 1
Test Case 1 (PZT5A) 25mmx25mm, 0.5mm thickness
1
Lab data
0.8 FEM
Test Case 2 (PZT5A) 20mmx20mm, 1mm thickness
Lattice
Normalized impedance phase (degrees)

1 0.6 Huang
Lab data LSM
0.9 FEM
Normalized impedance modulus (Ohms)

0.4
LSM
0.8 Lattice 0.2
Huang
0.7
0
0.6
−0.2
0.5
−0.4
0.4
−0.6
0.3
−0.8
0.2
−1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
0.1 Frequency (Hz) ·106

0 Fig. 15: Electrical impedance phase for test case 1

0 0.5
Fig. 13: Electrical impedance modulus for test case 2
We can see good agreement with the real data for the
three cases, although the FE model’s predictions are closer
to reality. Due to the lack of lateral dimensions and any kind
of attenuation or loss mechanism we can see our models
predicting a narrower and earlier peak. Due to a laboratory
problem, the data for the second test case is incomplete,
but we can still appreciate its behaviour. All the plots are
normalized to the unit due to some magnitude discordances
among the models.
It is not shown in the pictures for the sake clarity, but we
are also able to predict the impedance peak’s harmonics in
1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz) ·106
We have then tested some realistic devices and found that
we can successfully predict their behaviour in a fast and clean
way using analytical models. Let’s see now how to interpret
the impedance results: when designing or analyzing a given
transducer, we want to locate its impedance peaks, which will
be the desired operating frequencies, as we will get the best
behaviour there. This is then the first step in the design process.
B. Transducer bandwidth
We will define bandwidth as the range of frequencies for
which the signal power has dropped to half of its maximum
value (-3 dB). In the case of ultrasonic transducers, the signal
will be the force output when transmitting and the voltage
output when receiving (completely analogous results using
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 11

Test Case 2 (PZT5A) 20mmx20mm, 1mm thickness Fractional BW: 0.073993 . Test Case 1
1
0.8
Lab data 0.9
0.6 FEM
Normalized impedance phase (degrees)

LSM 0.8
0.4 Lattice

Normalized force modulus (N)

Huang 0.7
0.2
0.6
0
0.5
−0.2
0.4
−0.4
0.3

−0.6 0.2

−0.8 0.1

−1 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 4.05 4.06 4.07 4.08 4.09 4.1 4.11 4.12 4.13 4.14
Frequency (Hz) 6 Frequency (Hz) ·106
·10

Fig. 16: Electrical impedance phase for test case 2 Fig. 18: Fractional bandwidth for test case 1

Test Case 3 (PZT5A) - 20mmx20mm, 2mm thickness Fractional BW: 0.587257 . Test Case 2
1
1
Lab data
FEM 0.9
0.8
Lattice
Normalized impedance phase (degrees)

Huang 0.8
0.6
Normalized force modulus (N)
LSM
0.7
0.4

0.2
0.6

0 0.5

−0.2 0.4

−0.4 0.3

−0.6 0.2

−0.8 0.1

−1 0
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.9 1.92 1.94 1.96 1.98 2 2.02 2.04 2.06 2.08 2.1 2.12 2.14 2.16 2.18 2.2
Frequency (Hz) ·106 Frequency (Hz) ·106

Fig. 17: Electrical impedance phase for test case 3 Fig. 19: Fractional bandwidth for test case 2

Fractional BW: 2.217252 . Test Case 3

1
either of them). Similarly, the fractional bandwidth is defined
as bandwidth divided by the maximum frequency. Higher 0.9
bandwidth means a wider curve and smaller bandwidth a 0.8
Normalized force modulus (N)

narrower one. We will want to design our transducers so

0.7
that they get higher fractional bandwidth as possible, so that
we can efficiently operate the device in a wider range of 0.6

frequencies. In time domain this is equivalent to short ringing 0.5

signal and consequently better response.
0.4

Our results for the three transducers are presented below 0.3

(figures 18,19 & 20). The obtained fractional bandwidth is 0.2

probably not very accurate as our analytical models predict 0.1
narrower peaks, but as we will see later it still provides good
0
qualitative behaviour. 0.860.88 0.9 0.920.940.960.98 1 1.021.041.061.08 1.1 1.121.141.161.18 1.2
Frequency (Hz) ·106

A mismatch between transducer and loading medium Fig. 20: Fractional bandwidth for test case 3
acoustic impedances will result in poor bandwidth and energy
transfer, being them increased as both impedances get closer.
Hence, a preliminary test for our models will consist of shown below (figure 21).
varying the theoretical loading medium impedance and find
the maximum in the fractional bandwidth curve, which is The black vertical line shows the transducer impedance,
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 12

ideally we should get the maximum exactly at that point, but the first one, the model predicts as well high bandwidth peaks
due to the tendency of finding the maximum impedance peaks at the successive λ/2 jumps. The second one seems to have
a bit shifted, we find a maximum point which is close to the destructive wave interaction for early jumps, being maybe
theoretical result. The third test case, the one further from 1D stabilized at a later stage. However it is difficult to judge
approximation, presents some drift for higher impedances. because the dashed lines cut the bandwidth curve at very
steep points. The third case is more difficult to interpret: we
FBW optimization for a varying loading medium impedance definitely see the spike behaviour located around the dashed
120
TC1 lines, but it does not agree with the expected behaviour. This
110 TC2 may be caused by numerical instabilities or by a too small
TC3
100
lateral dimensions-thickness ratio, making theh 1D model
insuficient for this kind of analysis. In this case, further
90
analysis using laboratory instrumentation or FEM should be
Fractional bandwidth

80 used.
70
60 Summarizing, nice agreement with theory is found using the
50
two first test cases, while the third one needs further testing. It
seems to work fine for transducers thoroughly satisfying our
40
assumptions. Deeper and broader analysis must be performed
30 for other transducers to check the validity of this models.
20
·10−2 Test Case 1 - Optimum matching impedance
10 5

0 4.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Front medium impedance Z1 (Rayl) ·108 4

Fig. 21: Acoustical impedance optimization 3.5

Fractional bandwidth

2.5
C. Bandwidth and matching layers
2
When designing an application specific transducer with a
1.5
given loading medium, the matching layers come into play:
their purpose it to minimize the acoustic impedance mismatch 1

between the transducer and the loading medium. 0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Literature gives the ideal impedance value for a single Matching layer thickness (m) ·10−3
matching layer as the geometric mean √ of the transducer Fig. 22: Matching layer optimization for TC1
and loading medium impedances: Z = Zt Zl . The other
important aspect of the matching layers is their thickness:
optimal thickness for a single matching layer is a quarter Test Case 2 - Optimum matching impedance
of the operating wavelength (λ = c/f , being c the speed of 0.35

sound in the selected material and f the frequency). When

the thickness of the piezoelectric layer is λ/2 this guarantees 0.3

that waves that are reflected within the matching layer are
Fractional bandwidth

in phase when exiting the layer, hence maximum energy 0.25

transmission is achieved. [7]

0.2

We have tested this results analytically using lattice

model, the results plotted below: we have considered a 0.15

single matching layer with optimum impdance, and different

thicknesses varying from zero to twice the thickness of the 0.1

piezoelectric layer. The first dashed line corresponds to λ/4

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
thickness, and the gap between any two consecutive lines
Matching layer thickness (m) ·10−3
is λ/2, this will also keep reflected waves in phase. It is
worth mentioning that our piezoelectric thicknesses are way Fig. 23: Matching layer optimization for TC2
larger than the optimum λ/2 explained before, so we may
see discrepances with the optimum matching thickness as well.
D. Communication systems: pitch-catch
We can see very good agreement in the two first test cases: As a last setup for our models we will consider simulating a
higher bandwidth is obtained as predicted theoretically. For complete transducer system, that is typically seen in practical
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 13

Test Case 3 - Optimum matching impedance considering the new index notation specified in the figure.
2.5

Then, for the receiver, using (12)

2
−F · h · KF 2 TF 2 U (s)/sZc2
V = =
Fractional bandwidth

 
KF 2 TF 2 KB2 TB2 U (s)
1.5 1 − h2 +
2 2 s2 Ze Zc2
h · KF 2 TF 2 U (s)/sZc2
1 =   ·
K F 2 TF 2 KB2 TB2 U (s)
1 − h2 +
2 2 s2 Ze Zc2
0.5
e · h · a(s)Z2 /(Zc1 + Z2 ) KF 1 (s)Y (s)
·
1 − h2 Y (s)/(sZc ) (TF 1 KF 1 /2 + TB1 KB1 /2)
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Matching layer thickness (m)
To test the results we have used the first test case transducer
·10−3
as transmitter and receiver. The in-between medium is chosen
Fig. 24: Matching layer optimization for TC3 to be steel (Z2 = 45MRayl) with large enough thickness. The
results are presented below.
applications: we place two ultrasonic transducers at both sides
of an in-between medium, this can be a wall or other kind of Pitch - Catch system for TC1
object (see figure 18). 1 1

Air Piezo In-between medium Piezo Air Normalized force output - transmitter
0.8 0.8

Normalized voltage output - receiver

0.6 0.6
Z1 .v1 Zc1 .vc1 Z2 .v2 Zc2 .vc2 Z3 .v3

Fig. 25: Pitch-catch transmission system 0.4 0.4

One of them will act as a transmitter and the other one

will receive the transmitted wave. This kind of systems 0.2 0.2
have started being used in places where conventional wired
communication is not possible, for example in submarines or
in any kind of pressurized enclosure. This kind of ultrasonic 0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
communication problem is usually approached using an Frequency (Hz) ·107
ad-hoc procedure, get the transducers in place and test them Fig. 26: Pitch-catch transmission system i/o using TC1
using a short pulse trying to obtain the frequency signature
that would be used to model more complex waves.
The thick curve is the generated force in the transmitter,
and the thin one is the generated voltage in the receiver. We
Hence we can try to apply analytical modelling to this
can see that we receive a similar but attenuated signal. Similar
problem: we shall use the linear systems model for both
results are obtained using the other two test cases. The system
transmitter and receiver. To use this model we have to assume
can be improved introducing matching layers and chosing a
that the in-between medium’s thickness is much larger that
better backing than air. It would also be interesting to obtain
the transducer, this shoud not be a problem though. We will
an analytical model comprising both transmitter and receiver
hit the transmitter with an ideal pulse and obtain the force
in a compact way, i.e. solve the wave equations for the present
output in frequency domain, finally we will plug it into
boundary conditions (see figure 10).
the receiver, obtaining the final voltage output. We could
use an inverse Fourier transform to get the time domain
respresentation and then use convolution to obtain the final VII. C ONCLUSIONS
output.

The analytical model gives the following: applying a voltage

W E have explained and tested three different analytical
models for ultrasonic transducers in a variety of
experiments to check their behaviour, obtaining typically
e to the transmitter, the transfer function for the output force good results, although when the transducer gets a bit far from
is, using equation (17) the models’ assumptions, predictions break and one has to be
more careful.
e · h · a(s)Z2 /(Zc1 + Z2 ) KF 1 (s)Y (s)
FF 1 = −
1 − h2 Y (s)/(sZc ) (TF 1 KF 1 /2 + TB1 KB1 /2) Our models allow us to find the optimal operating
where the coefficients are detailed in the third section, frequencies for each considered transducer. Once this is
M1 MINIPROJECT, ERASMUS MUNDUS MASTER IN COMPLEX SYSTEMS SCIENCE, JUNE 2014 14

achieved, our interest will be maximizing the bandwidth and

the energy transmission. This will be done using matching
layers, whose optimum impedance and thickness can be
obtained analytically.

Thus analytical models are a good tool for design and

optimization purposes, providing an insight into the undelying
physics. This kind of models are also very fast and can be
easily coded, in contraposition with FE models which are
much slower and computationally expensive, and usually come
in the form of commercial software packages.

VIII. ACKNOWLEDGEMENT
I would like to thank Nishal Ramadas for all the help and
training provided.

R EFERENCES
[1] M. Willatzen, Ultrasound transducer modelling – general theory and
applications to ultrasound reciprocal systems, IEEE Transactions on
Ultrasonics, Ferroelectrics and Frequency Control, vol. 48, No 1, January
2001.
[2] M. Redwood, Transient performance of a piezoelectric trasnducer, J.
Acoust. Soc. Am. 33 (4), April 1961.
[3] G. Hayward, C. J. MacLeod, A systems model of the thickness mode
piezoelectric transducer, J. Acoust. Soc. Am. 76 (2), August 1984.
[4] N. Ramadas, G. Hayward, C. J. C. Barbenel, Y. Estanbouli, A linear
systems model of the thickness mode piezoelectric transducer containing
dual piezoelectric zones , Ultrasonics Symposium, 2004 IEEE, Vol. 3.
[5] G. Hayward, M. N. Jackson, A lalttice model of the thickness-mode
piezoelectric transducer, IEEE Transactions on Ultrasonics, Ferroelectrics
and Frequency Control, vol. UFFC-33, No 1, January 1986.
[6] C Huang, M. V. Z. Marmarelis, Q. Zhou, K. K. Shung An analytical
model of multilayer ultrasonic transducers with an inversion layer, IEEE
Transactions on Ultrasonics, Ferroelectrics and Frequency Control, vol.52,
No 3, March 2005.
[7] T. E. Gómez Álvarez-Arenas and Luis Dı́ez, Novel impedance matching
materials and strategies for air-coupled piezoelectric transducers, SEN-
SORS, 2013 IEEE Conference, November 2013.