You are on page 1of 20

J. Micromech. Microeng. 6 (1996) 157176.

Printed in the UK
Equivalent circuit representation of
electromechanical transducers: I.
Lumped-parameter systems
Harrie A C Tilmans
Katholieke Universiteit Leuven, Departement ESAT-MICAS, Kardinaal Mercierlaan
94, B-3001 Heverlee, Belgium
Received 7 December 1995, accepted for publication 28 December 1995
Abstract. Lumped-parameter electromechanical transducers are examined
theoretically with special regard to their dynamic electromechanical behaviour and
equivalent circuits used to represent them. The circuits are developed starting from
basic electromechanical transduction principles and the electrical and mechanical
equations of equilibrium. Within the limits of the assumptions on boundary
conditions, the theory presented is exact with no restrictions other than linearity.
Elementary electrostatic, electromagnetic, and electrodynamic transducers are
used to illustrate the basic theory. Exemplary devices include electro-acoustic
receivers (e.g., a microphone) and actuators (e.g., a loudspeaker),
electromechanical lters, vibration sensors, devices employing feedback, and force
and displacement sensors. This paper forms part I of a set of two papers. Part II
extends the theory and deals with distributed-parameter systems.
1. Introduction
Electromechanical transducers are used to convert electrical
energy into mechanical (or acoustical) energy, and vice
versa. They are utilized for electrical actuation and sensing
of mechanical displacements and forces in a wide variety
of applications (see e.g. [13]). An illustrative example of
a sensing device is a microphone in which a sound pressure
is converted into an electrical signal. In a microphone, the
pressure acts upon a spring-supported mass, which usually
consists of a stretched diaphragm. The generated mass
(diaphragm) displacement is next converted into an electric
output signal by means of an electromechanical transducer.
In a loudspeaker, on the other hand, an electromechanical
transducer is used to convert the electrical output signal
of an audio amplier into a force acting on the speaker
diaphragm. This results in a displacement of the diaphragm,
thereby generating sound waves.
The behaviour of electromechanical transducers can
be described by the differential equation(s) of motion of
the structural member(s), by the characteristic equations
of the transducer element(s), and by a set of boundary
conditions. A very explanatory and quick way of gaining a
deeper insight into the dynamic behaviour of the transducer
is the equivalent circuit approach, in which both the
electrical and mechanical portions of the transducer (or
system) are represented by electrical equivalents. The
approach is based on the analogy that exists between
Present address: CP Clare Corporation, Overhaamlaan 40, B-3700
Tongeren, Belgium. e-mail: tilmans@imec.be
electric and mechanical systems [14]. In this method, the
transducer is no longer described by complex differential
equations and boundary conditions, but by a lumped-
element electrical circuit in which the elements are
physically representatives of the transducers properties
such as its mass, stiffness, capacitance, and damping.
The circuits implicitly contain, because of the way they
are constructed, all the equations governing the system
represented. To the extent that the original assumptions
are valid, the equivalent circuit can be considered an exact
representation of the electromechanical transducer. The
practicality of the equivalent circuit approach stems from
the eld of electricity where it is unthinkable that the
design and analysis of electrical systems is carried out
on the basis of Maxwells equations. The applications
of lumped-element circuits are numerous nowadays, and
their use is strongly justied by modern electric network
theory which provides us with powerful mathematical
techniques and network analysis programs, such as SPICE.
Equivalent circuits are now also implemented for analysing
electromechanical systems, where one of their strengths is
that they provide a single representation of devices that
operate in more than one energy domain. Noteworthy is
their proven indispensable value in the development of
piezoelectrically driven resonators for use as mechanical
sensors, timebases, and electromechanical lters [3, 5, 6],
and further also in the eld of electroacoustics [2].
Equivalent circuits are particularly useful for the analysis
of systems consisting of complex structural members and
coupled subsystems with several electrical and mechanical
0960-1317/96/010157+20$19.50 c 1996 IOP Publishing Ltd 157
H A C Tilmans
ports. Not only is the strict use of differential equations
very difcult for these cases, but this method also often
obscures the solution [3]. The equivalent circuit method
lends itself to a better visualization of the system, and, once
the basic circuit is constructed, it may be used in further
analyses to investigate the effects of connecting subsystems
or of making modications to the structure.
The purpose of this paper is to lay a mathematical
foundation for developing equivalent circuit representations
of lumped-parameter electromechanical transducers and
for interconnecting the obtained circuits to the outside
world and further to illustrate their applicability in real
systems. The paper forms part I of a set of two papers.
Whereas part I deals with lumped-parameter systems, in
part II [7] the theory is extended to include distributed-
parameter systems as well. Only the steady-state small-
signal dynamic frequency response is considered, but the
circuits are equally well suited to analyse the transient
behaviour of the transducer system. Large-signal non-
linear behaviour cannot be analysed in a straightforward
manner using equivalent circuits and will not be dealt with
here. Therefore, the transducers that are intrinsically non-
linear are linearized around a bias point. The circuits
are constructed starting from basic electromechanical
transduction principles and the electrical and mechanical
equations of equilibrium. The focus will be on systems
with a single electrical and a single mechanical port.
Simple examples of how to account for more than one
port are given. The basic principles of electromechanical
transduction are not a subject of this paper. For a
description of this topic reference is made to the literature
(see e.g., the books by Woodson and Melcher [8] or Neubert
[1]). The analysis is limited to reversible or bilateral
transducers, i.e., those which give rise to mechanical
motion from electrical energy or the other way round
[1]. These include electrostatic, electromagnetic, and
electrodynamic and evidently exclude thermal transducers.
Further, only rectilinear or translational mechanical systems
are considered, although rotary systems can be described in
a similar way. Acoustic systems are briey introduced and
are accounted for in the analysis. Finally it is pointed out
that generally applicable assumptions, such as negligible
fringing elds, small velocities compared with the velocity
of light, and perfect conductors (see e.g., [8]), are implicitly
understood.
2. Lumped-parameter electromechanical systems
2.1. General
The essential characteristic of lumped-parameter (or
discrete) systems is that the physical properties of
the system, such as mass, stiffness, capacitance, and
inductance, are concentrated or lumped into single physical
elements. Thus, elements representing mass are perfectly
rigid, and conversely elastic elements have no mass. This
idealization is similar to electric circuit theory where
inductors are considered to have no capacitance, capacitors
no inductance, and resistors are purely ohmic. As such,
a lumped-parameter electromechanical system consists of a
nite number of interconnected masses, springs, capacitors,
inductors, resistors, etc. Lumped-parameter modelling is
valid as long as the wavelength of the signal is greater than
all dimensions of the system [2, 8]. The dynamic behaviour
of these systems can be described by ordinary differential
equations with time t being the only independent variable.
The analysis described below will focus on lumped-
parameter systems with a single degree of freedom (SDOF)
in the mechanical domain, implying that they display a
single mechanical resonance [9]. The study of SDOF
systems is of importance since (i) many real systems display
a behaviour sufciently close to the behaviour of an SDOF
system, (ii) they improve the understanding of real systems,
whilst not having to deal with tedious mathematics, and
(iii) very often in a limited frequency range distributed-
parameter systems can be treated as SDOF systems, as will
be further claried in part II [7].
2.2. An example
An example of a real electromechanical system that behaves
sufciently like an SDOF lumped-parameter system is
shown in gure 1(a). The structure shown illustrates the
basic conguration of an electrostatic transducer that can
be used as a force gauge, e.g., a microphone [1, 2, 8]. It
consists of a doubly clamped beam (or diaphragm) with
a rigid mass at the centre. The mass is electroded on
one face which denes one plate of the capacitor used for
the electrostatic transduction. The other plate is formed
by a stationary surface which is in close proximity to the
mass. The mass is effectively supported by two adjacent
beam elements that can be approximated by lumped springs,
each with a constant k/2. The total spring constant is
therefore equal to k. The mass moves in response to a
pure mechanical force F
mt
and/or a force of electric origin
F
et
which is due to the attractive electrostatic force of
the electrodes (nomenclature is explained below). The
circuit comprising the decoupling capacitor C
d
and the
resistance R
L
, which may represent the the input resistance
of an amplier, is used to isolate the output terminals
from the bias voltage v
0
, and is not intended to affect
the dynamic behaviour of the system in normal operation.
If the voltage v
e
is taken as the output signal and with
driving frequency this means that R
L
C
d
1 and
R
L
R
0
[8]. Assuming that the total mass of the two beam
segments is small relative to the rigid mass and, further, if
only incremental signal variations around a biasing point
(further explained below) are considered, the system can
be modelled as the lumped-parameter SDOF system of
gure 1(b). Here, m represents the rigid mass, k is the total
stiffness of the sections of the beam supporting the mass,
c is a parameter representing viscous energy losses, and
C
p
denotes a (parasitic) parallel capacitance, e.g., due to
the stray capacitance of the interconnecting wires. Further
analysis of the transducer, including the development of
its equivalent circuit, will be presented in the following
subsections.
158
Lumped-parameter electromechanical transducers
Figure 1. An example of an electromechanical transducer
consisting of a beam with a rigid mass at the centre, which
is subjected to a force F
et
of electric origin and a pure
mechanical force F
mt
. (a) A schematic representation; (b) a
lumped-parameter incremental model. The transducer can
be used as a force (F
m
) or displacement (x
m
) sensor, with
output voltage v
e
or current i
e
.
Figure 2. A schematic representation of a two-port
electromechanical transducer.
2.3. Energy exchange
Exchange of energy of a transducer and the outside world is
achieved through ports (depicted as a pair of terminals).
This is illustrated in gure 2, showing a block diagram of
the often encountered linear electromechanical transducer
with a single electrical and a single mechanical port. A
port is dened by a pair of conjugate dynamic variables
called effort or intensive variable and ow. The power
exchange through the port is given by the product of
effort and ow. The ow is given by the time derivative
of the corresponding state or extensive variable. The
transducer depicted in gure 2 is a two-port energy storage
element, emphasizing the number of ports and the fact that
transducers store energy. Electrical ports are dened by the
{voltage (v), current (i)} pair, and mechanical ports by the
{force (F), velocity (u)} pair. Two-port storage elements
are completely characterized by an energy function of the
two independent state variables [8]. The state variables
associated with the mechanical and the electrical ports are
displacement x and the electric charge q, respectively.
In describing transducers that are partly operating in the
magnetic domain it proves convenient to dene a magnetic
port, characterized by the {current (i), ow of ux linkage
(

)} pair. The ux linkage denes the corresponding state


variable.
The purpose of this section is to derive equivalent
circuit representations of the two-port transducer in gure 2.
Although the transducer in gure 2 consists of only one
electrical and one mechanical port, the discussion below
can easily be generalized to any arbitrary number of ports
(see also [8]).
3. Elementary lumped-parameter transducers
3.1. Basic congurations
Four examples of elementary electromechanical lumped-
parameter transducers are shown in gure 3. Their op-
eration is respectively based on electrostatic (with out-of-
plane motion), electrostatic (with in-plane motion), elec-
tromagnetic, and electrodynamic transduction principles,
each of which is extensively described in the literature
(see e.g., [1, 2, 8, 10]). The transducer of gure 3(a) is
termed a transverse electrostatic transducer, emphasizing
that the plates move transverse or perpendicular to each
other, as opposed to the parallel electrostatic transducer of
gure 3(b) in which the plate surfaces stay parallel dur-
ing motion. Apart from the electrodynamic transducer, the
transducers shown all display non-linear behaviour. For
instance the electrostatic force of the transducers in g-
ure 3(a) and (b) shows a quadratic dependence on the
charge or the voltage (see appendix A). It is evident that
linear transducers are mathematically more tractable, but,
furthermore, linear transducers are also of great practical
importance. For instance, a condenser microphone is pur-
posely operated to behave as a linear device, since nonlin-
ear effects cause distortion and loss of delity [2]. Linear
behaviour is achieved for incremental or small-signal vari-
ations around bias or equilibrium levels. In fact, if only the
rst two terms in a Taylor series expansion about a static
equilibrium point are included, the total signal, which is in-
dicated with a subscript t , can be written as the sum of an
equilibrium signal, which is indicated with a subscript 0,
and the incremental signal, which is indicated without any
subscript, e.g., x
t
(t ) = x
0
+x(t ). Transducers can be biased
in several ways. For instance, an electrostatic transducer
can be electrically biased by applying a d.c. bias voltage
v
0
, or by introducing a bias charge q
0
, e.g., by means of an
electret. The electromagnetic transducer can be biased by
applying a bias current i
0
in the transducer coil or by plac-
ing a permanent magnet in the magnetic circuit. The type
of biasing will not change the analysis of the system signif-
icantly (see e.g., the book by Rossi [2]). Therefore, without
loss of generality, in this paper the analyses are limited to
voltage biasing for the electrostatic transducer and current
biasing for the electromagnetic transducer. These biasing
schemes are also most often employed in practice. Biasing
of the electrodynamic transducer is always done with a per-
manent magnet. It is pointed out that the electrodynamic
transducer is inherently linear and biasing is here imple-
mented to attain any electromechanical transduction at all.
It is further assumed that the incremental signals are sinu-
soidal with driving frequency , e.g., x(t ) = x exp(it ),
where x denotes a phasor [11]. Note that this is not a
real limitation, since for a linear system the steady-state
response to an arbitrary signal can be synthesized from the
response to sinusoidal driving signals using the techniques
159
H A C Tilmans
Figure 3. Schematic representations of elementary
electromechanical transducers: (a) parallel plate or
transverse electrostatic transducer, (b) in-plane or parallel
electrostatic transducer, (c) electromagnetic transducer,
and (d) electrodynamic (or moving-coil) transducer.
of Fourier transforms and Fourier series [11]. Also, if no
explicit time dependence is included for the incremental
signals, phasors are meant in this paper.
The transducers enclosed by the dashed box in gure 3
are assumed conservative or lossless. This means that
energy put into the system by the electric and mechanical
ports is stored and can be recovered completely through the
ports. All purely electrical elements (e.g., inductors and
capacitors not involved in the transduction and furthermore
all resistors) and mechanical elements (all masses, springs,
and dampers) are connected externally to the electric
and mechanical terminals, respectively. At this point
it is already noted that systems as dened above are
mechanically unstable. For instance, a slight displacement
Figure 4. A transverse electrostatic transducer completed
with a spring k connected to the mechanical port for
stabilization purposes.
of the movable plate from the equilibrium position would
cause the plates to collapse, or the movable plate would
disappear into innity (see also appendix A). Stability can
be attained by introducing an external mechanical spring k
and by making this spring an integral part of the transducer
as illustrated in gure 4. This will be further elucidated in
the following.
3.2. Characteristic equations
The characteristic equations of an electromechanical
transducer describe the linear relations that exist between
(dynamic) small-signal variations of the port variables
around the bias point. The equations of the two-port
electromechanical transducers of gure 3 are conveniently
represented with matrix algebra. Two variables are chosen
as the independent variable set, and the remaining two
variables form the dependent set. This implies six different
possible ways of formulating the characteristic equations
[11]. For instance, the constitutive equations, dened by
the matrix B, express the effort variables as a function
of the state variables. The transfer matrix T on the other
hand relates the effort-ow variables at the electrical port
directly to those at the mechanical port. It is beyond the
scope of this paper to go into too much detail regarding the
analytical development of the characteristic equations, but
in order to illustrate the underlying theory, in appendix A
a derivation is given of the constitutive equations of the
transverse electrostatic transducer of gure 3(a). For details
regarding the other transducers, reference is made to the
literature (see e.g., [1, 2, 8]). Table 1 summarizes the results
for the four types of transducer of gure 3. In each case,
a mechanical spring k is included to attain mechanical
stability in a similar way as illustrated in gure 4 for
the electrostatic transducer. More detailed expressions for
the static components (C
0
and L
0
) and the transduction
factors ( and ) are given in table 2. The bias force
F
0
affects the bias displacement, but usually F
0
is taken
to be equal to zero. The characteristic matrices for the
transducers without the spring k, as depicted in gure 3, can
easily be obtained from the matrices in table 1 by taking
k = 0. It is noted that the efforts are dened as the loads
which are applied externally to the transducer system via
the appropriate ports, and not as the loads exerted by the
transducer on the surroundings via the ports.
Note the similarity between the corresponding matrices
of the transducers operating in the electric and the magnetic
160
Lumped-parameter electromechanical transducers
Table 1. Constitutive and transfer equations of the lumped-parameter transducers shown in gure 3, completed with a
stabilization spring k at the mechanical port (as illustrated by gure 4). The matrix equations describe the relations between
the phasor quantities of the sinusoidal signals. The meaning of the symbols is explained in appendix B and in table 2.
Transducer State Efforts Flows
type q
1
, q
2
e
1
, e
2
f
1
, f
2
_
e
1
e
2
_
= B
_
q
1
q
2
_ _
e
1
f
1
_
= T
_
e
2
f
2
_
Electrostatic
(transverse, q(t ), x(t ) v(t ), F(t ) i (t ) = q(t ), x(t )
_
1
C
0

C
0

C
0
k
_
_
1

1
i
(k

2
C
0
)
i C
0


kC
0

_
gure 3(a))
Electrostatic
(parallel, q(t ), x(t ) v(t ), F(t ) i (t ) = q(t ), x(t )
_
1
C
0

C
0

C
0
k +

2
C
0
_
_
1

k
i
i C
0

_
C
0
+

2
k
_
_
gure 3(b))
Electromagnetic (t ), x(t ) i (t ), F(t )

(t ) = v(t ), x(t )
_
1
L
0

L
0

L
0
k
_
_
1

1
i
_
k

2
L
0
_
i L
0


kL
0

_
(gure 3(c))
Electrodynamic (t ), x(t ) i (t ), F(t )

(t ) = v(t ), x(t )
_
1
L
0

L
0

L
0
k +

2
L
0
_
_
1

k
i
i L
0

_
L
0
+

2
k
_
_
(gure 3(d))
Table 2. Parameters used in table 1, expressed in terms of the dimensional parameters, the bias conditions, and physical
constants.
Transduction factors
[N V
1
= A (m s
1
)
1
] and Static ( 0)
Tranducer type Static components [N A
1
= V (m s
1
]
1
) coupling factor
Electrostatic C
0

q
0
v
0
=

0
Ae
d+x
0
=
q
0
d+x
0
=

0
Aev
0
(d+x
0
)
2
_

2
C
0
k
(transverse, gure 3(a))
Electrostatic C
q
0
v
0
=

0
(l
0
x
0
)h
d
=
q
0
l
0
x
0
=

0
hv
0
d
_
1
1+
C
0
k

2
(parallel, gure 3(b))
Electromagnetic L
0


0
i
0
=
N
2

0
Ae
d+x
0
=

0
d+x
0
=
N
2

0
Aei
0
(d+x
0
)
_

2
L
0
k
(gure 3(c))
Electrodynamic L
0
= L
s
= coil seriesconductance = B
0
l
_
1
1+
L
0
k

2
(gure 3(d))
domain. In fact the electromagnetic transducer and the
transverse electrostatic transducer are dual to each other
with respect to the electric domain. The same can
be said for the parallel electrostatic transducer and the
electrodynamic transducer. Dual systems are described by
equations of the same form, but in which the coefcients
(e.g., capacitance and inductance) and effort-ow variables
(e.g. voltage and current) are interchanged. The observed
electrical duality is physically described by Amp` eres
circuital law and Faradays law of magnetic induction,
conveniently expressed as
_
v
i
_
=
_
0 N
1/N 0
_ _
m.m.f.

_
=
_
0 1
1 0
_ _
i

_
or :
_
i

_
=
_
0 1
1 0
_ _
v
i
_
(1)
where m.m.f. denotes the magnetomotive force,

is the
magnetic ux ow in the magnetic circuit, and N is given
by the number of active turns of the transducer coil coupled
with the magnetic eld of the transducers in gure 3(c) and
(d). The above matrix equation clearly shows that the effort
variable in the electric domain becomes the ow variable
in the magnetic domain, and vice versa.
The coupling factor , also indicated in table 2, is
an important characteristic of electromechanical transducers
as it provides a measure for the electromechanical energy
conversion which takes place in the lossless transducer
[6, 10, 12]. For a two-port storage element the coupling
factor can be found from the constitutive matrix B
as the following ratio: (coefcients product of the
interaction terms)/(coefcients product of the principal
(diagonal) terms). A coupling factor of zero means no
electromechanical interaction. It can be shown that a
stable equilibrium exists for 0 < < 1 [10, 12]. Typical
values for are in between 0.05 and 0.25. Furthermore,
the coupling factor provides an elegant way to relate the
parameter values, e.g., the spring constant, measured at one
of the ports to the conditions, e.g., v = 0, at the other port.
This will be further explained in the next subsection.
161
H A C Tilmans
Table 3. Direct electromechanical analogies for lumped
translational systems [1, 2].
Mechanical quantity Electrical quantity
Force F Voltage v
Velocity u = x Current i = q
Displacement x Charge q
Momentum p Magnetic ux linkage
Mass m Inductance L
Compliance 1/k
a
Capacitance C
Viscous resistance c Resistance R
a
k represents the spring constant.
3.3. Equivalent circuit representations
3.3.1. Analogies. The development of equivalent
circuit representations is based on the analogy in the
mathemtical descriptions that exists between electric and
mechanical (including acoustical) phenomena [1, 2, 11].
The analogies are a result of the formal similarities of
the integrodifferential equations governing the behaviour of
electric and mechanical systems. For instance, Newtons
second law of motion relating the force F and velocity
u for a rigid mass m, F = mdu/dt = md
2
x/dt
2
, is
mathematically analogous to the constitutive equation of
an electric inductor, v = Ldi/dt = Ld
2
q/dt
2
. In this
analogy, the force F plays the same role as the voltage
v, the velocity u as the current i, and the displacement
x as the charge q. The mass m in mechanical systems
corresponds to the inductance L in electrical circuits. The
foregoing examples illustrate the so-called direct analogy,
summarized in table 3. It is pointed out that equivalent
systems that are constructed based on this type of analogy
display the duality property in the sense that across
or between variables are equated to through variables,
and, conversely, through variables are equated to across
variables. This means that force (a through variable) is
analogous to voltage (an across variable), and velocity (an
across variable) to current (a through variable). Hence, this
implies that the network topologies of the mechanical and
electrical circuits are not the same. A series connection in
the mechanical circuit becomes a parallel arrangement in
the equivalent electrical circuit, and vice versa. This will be
further elucidated by the examples presented in section 5.
The direct analogy was in fact implicitly understood in
the foregoing. The governing equations, however, can also
be written in a form that suggests an analogy between the
force and the current, between the velocity and the voltage,
between the mass and the capacitance, etc. This analogy
is called the inverse or mobility-type analogy (see e.g. [1]).
In order to avoid any confusion in this paper no further
reference will be made to this latter form of analogy.
3.3.2. Equivalent networks. The construction of the
equivalent networks starts with the transfer matrices given
in the last column of table 1. This becomes clear after the
matrices are split into their constituent transfer matrices.
For instance, the transfer matrix of the electrostatic
transducer of gure 4 can be split as follows:
_
v
i
_
=
_
1

1
i
(k k

)
iC0


kC0

_ _
F
u
_
=
_
1 0
iC
0
1
_ _
1/ 0
0
_ _
1
1
i
(k k

)
0 1
_ _
F
u
_
(2)
where k

=
2
/C
0
(see also table 4). In the matrix
equation above, the centre matrix represents the transducer,
anked by the matrices of the electrical admittance and the
mechanical impedance. Each of the constituent transfer
matrices can be represented by an equivalent network. The
overal equivalent network consists of a cascade connection
of these networks and is shown in gure 5(a). It can
easily be shown that the network in gure 5(a) forms an
exact representation of the transfer matrix equation (2).
According to the aforementioned analogy a spring is
represented by a capacitor. The impedance (force/velocity)
of the spring k in gure 5(a) therefore is equal to k/i.
The electromechanical coupling is modelled through an
ideal electromechanical transformer with a transformer ratio
given by , called the transduction factor, which was
introduced in tables 1 and 2. The transformer relations,
are given by F

= v

and i

= u

, conform to
the sign conventions indicated in gure 5(a). Note the
existence of a spring with a negative constant k

2
/C
0
=
0
A
e
v
2
0
/(d + x
0
)
3
. The spring is a result
of the electromechanical coupling and apparently leads to
a lowering of the overal dynamic spring constant. This
is easily seen by combining the two springs into a single
spring with constant k

= k k

= k(1
2
) (see also
table 4). As long as k

> 0, which is equivalent to the


condition < 1, the system is mechanically stable [10, 12].
Matrix algebra explains that there does not exist one
unique way of decomposing a 2 2 matrix into three
consituent 2 2 matrices. Equation (2) merely illustrates
one way of decomposing the matrix. Furthermore,
each decomposition corresponds to its own unique circuit
representation. It can easily be shown that the four
circuits shown in gure 5 all form equivalent network
representations of the same transducer as shown in gure 4.
The choice of which circuit to use is conveniently dictated
by the application. For instance, the circuit of gure 5(b) is
most suited to determine the value of the free capacitor (i.e.,
for F = 0), whereas the circuit of gure 5(a) is preferred
to determine the clamped capacitance (i.e., for u = 0). A
quick glance at the circuit diagrams reveals that the latter
is equal to the usual capacitance C
0
, whereas the former
is given by C

0
= C
0
/(1
2
). A special note is needed
for the diagram shown in gure 5(d). In the three other
diagrams the electric and mechanical domains are clearly
separated by a transformer. A dimensional test will conrm
this. For the last diagram, however, such a test will fail.
The domain separation has disappeared and the circuit only
serves an algebraic purpose.
In order to nd an equivalent network representation of
the electromagnetic and the electrodynamic transducer, it is
necessary to combine (1), representing the link between the
electric and magnetic domain, with the transfer matrices as
presented in table 1. For instance, for the electromagnetic
162
Lumped-parameter electromechanical transducers
Table 4. Circuit elements introduced as a result of electromechanical coupling. The elements are used in the equivalent
circuits of gures 58. Expressions for , , C
0
, and L
0
in terms of the physical parameters can be found in table 2.
Transducer type C

0
, L

0
C

0
, L

0
k

Electrostatic C

0
=
C
0
1
2
= C
0
+ C

0
C

0
=

2
1
2
C
0
=
1
1
2

2
k
k

= k(1
2
) = k k


2
k =

2
C
0
(transverse, gure 5)
Electrostatic C

0
=
C
0
1
2
= C
0
+ C

0
C

0
=

2
1
2
C
0
=

2
k
k

=
k
1
2
= k + k

2
1
2
k =

2
C
0
(parallel, gure 6)
Electromagnetic L

0
=
L
0
1
2
= L
0
+ L

0
L

0
=

2
1
2
L
0
=
1
1
2

2
k
k

= k(1
2
) = k k


2
k =

2
L
0
(gure 7)
Electrodynamic L

0
=
L
0
1
2
= L
0
+ L

0
L

0
=

2
1
2
L
0
=
1
1
2

2
k
k

=
k
1
2
= k + k

2
1
2
k =

2
L
0
(gure 8)
Figure 5. Possible equivalent circuit representations of the transverse electrostatic transducer with a single electric port and
a single mechanical port as depicted in gure 4. The meaning of the symbols is explained in tables 2 and 4. The
transformers model the electromechanical coupling. The transformer relations given by F

= v

and i

= u

conform to the
sign conventions given in (a).
transducer (including a spring k), the decomposition can be
expressed as
_
v
i
_
=
_
0 1
1 0
_ _
1 0
iL
0
1
_ _
1/ 0
0
_

_
1
1
i
(k k

)
0 1
_ _
F
u
_
(3)
where k

=
2
/L
0
(see also table 4). The rst constituent
matrix represents (1) and can be modelled using an ideal
gyrator with gyrator resistance equal to unity. As a result
of the aforementioned duality (see subsection 3.2), the
equivalent circuit representing the cascade of the remaining
three matrices can easily be derived from the circuit in
gure 5(a). The overall equivalent circuit is obtained
by placing the gyrator in cascade with this circuit. The
result is shown in gure 6(a). It can easily be shown that
the circuits shown in gure 6(b)(d) also form equivalent
circuits of the same electromagnetic transducer. In the
latter two representations the gyrator and the transformer
are combined into a single gyrator, dened by the following
relations: F

= i

and v

= u

, conforming to the sign


conventions indicated in gure 6(c). The circuit parameters
indicated with a are claried in table 4.
A similar approach can be used to construct the
equivalent circuits for the parallel electrostatic and the
electrodynamic transducer. The results are shown in
gures 7 and 8, respectively. The two transducers are
clearly dual to each other. Also note the small, but
important, differences between the electromagnetic and the
electrodynamic transducer on the one hand, and between
the transverse and the parallel electrostatic transducer
on the other hand. For instance, the equivalent circuit
representation of the electrodynamic transducer shown in
gure 8(a) can be obtained from the equivalent circuit of
the electromagnetic transducer of gure 6(c) by replacing
the compliance 1/k

by a compliance 1/k.
The parameters indicated with a are a result of
the electromechanical coupling. The circuit parameters
indicated without a are the usually observed parameters,
i.e., in the absence of electromechanical coupling ( = 0).
For zero coupling, there will be only one spring constant k,
163
H A C Tilmans
Figure 6. Possible equivalent circuit representations of the electromagnetic transducer of gure 3(c) completed with an
external spring k connnected to the mechanical node (similar to gure 4). The meaning of the symbols is explained in
tables 2 and 4. The gyrator relations given by F

= i

and v

= u

conform to the sign conventions given in (c).


Figure 7. Possible equivalent circuit representations of the parallel electrostatic transducer of gure 3(b) completed with an
external spring k connnected to the mechanical node (similar to gure 4). The meaning of the symbols is explained in
tables 2 and 4. The transformers model the electromechanical coupling. The transformer relations are given by F

= v

and
i

= u

, conforming to the sign conventions given in (a).


Figure 8. Possible equivalent circuit representations of the electrodynamic transducer of gure 3(d) completed with an
external spring k connnected to the mechanical node (similar to gure 4). The meaning of the symbols is explained in
tables 2 and 4. The gyrators model the electromechanical coupling. The gyrator relations are given by F

= i

and
v

= u

, conforming to the sign conventions given in (a).


one capacitor C
0
, and one inductor L
0
. Coupling causes
a dependence of the properties in one domain on the
conditions in the other domain. For instance, the stiffness
k

of the transverse electrostatic transducer is the measured


mechanical stiffness if the electrical port is short-circuited,
i.e., v = 0, whereas the stiffness k is measured for open-
circuit conditions (i = 0). A summary of the results is
given in table 4.
4. Coupling of the transducers to the outside
world
As indicated in subsection 2.3, the exchange of energy of
the transducer and the outside world is done through ports.
External loads or sources can be connected to either one
of the ports for instance as illustrated by the system of
gure 1. Laws of equilibrium link the transducers via their
port relations (e.g., the transfer matrices) to the external
elements. Only linearly operating systems are considered,
164
Lumped-parameter electromechanical transducers
meaning that the relation between the incremental variables
can be described by linear (differential) equations with
constant coefcients. The equilibrium conditions at the
electrical site are governed by Kirchhoffs voltage (KVL)
and current (KCL) laws. At the mechanical site the
governing laws are Newtons second law of motion or,
more appropriately dAlemberts principle, expressed as
F
i
= 0, and the geometric compatibility or the continuity
of space, formulated as u
i
= 0 [8]. The latter
condition is seldom used for the analysis of mechanical
systems, but must at all times be satised. Note that
the latter two conditions are directly obtained by invoking
Kirchhoffs laws to the mechanical part in the equivalent
circuit representation, thereby illustrating once more the
analogy between electric and mechanical systems. As a
matter of fact, this mathematical correspondence of laws
is an essential requirement for using equivalent circuit
representations for the analysis of mechanical systems.
The foregoing forms the fundamental basis for devel-
oping equivalent circuits for electromechanical transducers
with arbitrary electric and mechanical loads and/or sources.
As an illustration, consider the system shown in gure 1(a)
and its lumped-parameter model of gure 1(b). The elec-
trical part is in fact already represented by an equivalent
network and therefore needs no further explanation. For
the mechanical part the conditions may be derived as fol-
lows. Applying dAlemberts principle to the mass m leads
to
F
m
(t ) F
e
(t ) kx(t ) c x(t ) m x(t ) = 0
F
m
F
e

k
i
u cu imu = 0 (4)
where the second of the above equations is given in terms
of the phasors of the respective quantities. Recalling that
dAlemberts principle is the electromechanical analogue
of KVL, it is easy to show that the circuit shown in
gure 9 denes a possible equivalent circuit of the system
in gure 1. The equivalent circuit of gure 5(a) (enclosed
by the dashed box in gure 9) is chosen to represent
the transducer (plus stabilization spring k), because of its
direct physical signicance. Note, however, that the circuit
enclosed by the box can be replaced by any of the other
circuits presented in gure 5. This may appear a little
awkward. For instance if the circuit of gure 5(b) is
used, the velocity through the compliance (capacitor) 1/k
as predicted by the equivalent circuit is not the same as
the velocity of the mechanical resistance c and the mass
(inductor) m. In practice however (see gure 1(b)) the
velocities through the aforementioned components are the
same. The answer to this apparent paradox is simple: the
compliance 1/k in the equivalent circuit is not the same
as the compliance 1/k in gure 1(b). They only happen
to be numerically equal. The internal conguration of the
circuit in gure 5(b) is reorganized at the expense of giving
up the practical meaning. It is evident that the physical
link is lost even further for the circuit in gure 5(d).
This emphasizes once more that the equivalent circuits
very often only serve an algebraic purpose. Based on the
foregoing, it is concluded that the circuit representations of
gures 5(a), 6(c), 7(a) and 8(a) have a true physical link,
thus minimizing the chance for misinterpretation as much
as possible. It is for this reason that emphasis will be on
these circuit representations while discussing the examples
presented in section 5.
5. Examples of lumped-parameter
electromechanical systems
5.1. Force and displacement transducers (microphone)
The basic operating principles and conguration of an
electrostatic force or displacement transducer have already
been presented in subsection 2.2. A schematic diagram
is shown in gure 1. In operation, the force F
m
(t ) to be
measured (dened as positive in the positive x direction) is
exerted on the mass, or the mass is displaced by an amount
x
m
(t ) in the case where the displacement must be measured.
As explained before a motion of the mass is converted
into an electrical signal, e.g., a current, which ows in
part through the resistors R
0
and R
L
, thereby producing
an output voltage v
e
(t ). This voltage is a measure for
the applied force or displacement. Using the equivalent
circuit shown in gure 9, extended with the appropriate
mechanical sources (F
m
or x
m
) which have to be connected
to the mechanical terminals, the steady-state analysis for
sinusoidal signal operation becomes very straightforward.
The transfer function describing the relation between the
input variable and the output voltage is easily obtained from
the equivalent circuit. For the force transducer this results
in
v
e
F
m
=
iR
p

(k

+ic
2
m)(1 +iR
p
(C
0
+C
p
)) +iR
p

2
(5)
and for the displacement transducer
v
e
x
m
=
iR
p

1 +iR
p
(C
0
+C
p
)
=
v
0
d +x
0
C
0
C
0
+C
p
iR
p
(C
0
+C
p
)
1 +iR
p
(C
0
+C
p
)
(6)
where R
p
R
0
/R
L
. It is noted that the current i
e
can
also be taken as the output signal. The respective transfer
functions for this case are easily obtained from the above
equations by noting that i
e
= v
e
/R
L
.
5.1.1. The condenser microphone. If the applied force
is the result of an acoustic pressure p
m
(t ) the transducer can
be used as an electrostatic or condenser microphone. For
instance, consider the condenser microphone as depicted
in gure 10(a). The electric terminals are biased with
a voltage v
0
and a bias resistor R
0
in the same way
as shown in gure 1(a.) The input signal is the sound
generator pressure p
g
(t ) which is applied to a movable
rigid front plate with area A and mass m. The plate
is mounted on a peripheral spring with an equivalent
constant k. The pressure produces a small signal voltage
v
out
(t ) at the electric terminals that can be connected to
a (pre-)amplier (not shown). A Th evenin equivalent
circuit [11] of the transducer is shown in gure 10(b).
The Th evenin equivalent circuit is very simple and its
elements can be determined (experimentally) as follows.
165
H A C Tilmans
Figure 9. An equivalent circuit representation of the electromechanical transducer system shown in gure 1. The circuit
enclosed by the dashed box represents the transducer of gure 4 (see also gure 5(a)).
Figure 10. The condenser microphone. (a) A schematic diagram; (b) the incremental signal Th evenin equivalent circuit; (c)
a detailed equivalent circuit illustrating the interaction between the variables in the three signal domains of interest, i.e.,
electric, mechanical, and acoustic.
The Th evein voltage v
T h
is given by the open-circuit
voltage and the Th evenin impedance Z
T h
is the impedance
seen at the electric terminals if all the independent
sources, here only the pressure generator, are set to
zero. However, the circuit is not at all convenient for
attaining a better understanding of, and for analysing and
optimizing, the microphone behaviour, as the shape of the
frequency response and the microphones sensitivity are
determined by the bias conditions, geometrical parameters,
and damping and dynamic characteristics of the specic
microphone structure, including the electrical, mechanical,
and acoustical parts. Damping for instance is due to the
air-streaming resistance of the air gap, and can be strongly
reduced by introducing acoustic holes in the backplate.
Also, a narrow gap is of importance to attain a high
sensitivity. The stray capacitance C
p
between the metal
case (which is electrically connected to the front plate)
and the back plate results in a capacitive attenuation of
the transducer signal and thus a lowering of the sensitivity
[2]. None of these individual components or others are
reected in the Th evenin equivalent circuit. Therefore, a
more detailed equivalent circuit similar to the one shown
in gure 9 is required. For this purpose, the acoustic
signal domain must be included. The effort, ow, and state
variables in the acoustic domain are respectively given by
the acoustic pressure p
a
(N m
2
), the volume velocity

(m
3
s
1
), and the volume displacement (m
3
) [2]. It can
be shown [2] that, for a system in equilibrium, equality
of the acoustic pressures p
i
at either side of an acoustic
junction applies, mathematically formulated as p
i
= 0.
Further, the continuity law states that for the incident
volume velocities

i
at a junction the following relation
applies:

i
= 0 [2]. At this point it is evident that
the latter two relations form the analogies to KVL and
KCL, respectively. In the example discussed here, the
mechanical and the acoustic domain are linked via the
166
Lumped-parameter electromechanical transducers
movable front plate and are related as follows: F
m
= Ap
a
and u
m
= (1/A)

, where A denotes the area of the front


plate. In an equivalent circuit this relation is represented by
a mechano-acoustical transformer with ratio A. Note that
the effective plate area A
e
for the electrostatic transduction
is reduced compared with A by an amount equal to the
total area of the acoustic holes. The resulting equivalent
circuit, reecting the inuence of all three signal domains,
is shown in gure 10(c). The radiation impedance, due to
the radiation of sound back into the surrounding medium, is
modelled as a series combination of a radiation resistance
R
ar
(representing frictional forces) and a radiation mass
(inductance) m
ar
(representing the mass of the air close to
the front plate that is vibrating in phase with the plate). The
series combination of the acoustic resistance R
ag
and the
acoustic mass m
ag
represents the air gap and the acoustic
holes, and C
bc
represents the acoustic compliance of the
back chamber. Other elements that are often encountered
in the acoustical part of a microphone but that have not
been indicated in gure 10(a,c) are for instance the pressure
equilization holes or the effects of a exible backplate
[2, 13]. Finally, it is pointed out that by replacing the
equivalent circuit of the electrostatic transducer (enclosed
by the dashed box) in gure 10(c) by another appropriate
transducer circuit, e.g., any of the circuits in gures 6
8, the equivalent circuit of any other type of microphone,
e.g., an electrodynamic or an electromagnetic microphone,
is readily obtained.
It is evident at this point that the overall microphone
behaviour is determined by a complex interplay between
the elements in the different signal domains. With respect
to this it goes without saying that the equivalent circuit is
of great help in analysing the microphone behaviour. It
is beyond the scope of this paper to go into more detail
regarding the design and operation of microphones. For
this, reference is made to the literature (see e.g., [2, 13]).
5.2. Electromechanical and electroacoustical actuators
In an electromechanical actuator, electrical energy is
converted into mechanical energy. Actuators are for
instance used in force-balanced transducers [1] (see also
subsection 5.6), or as an acoustic transducer or loudspeaker,
in which the output signal consists of pressure sound waves
in air [2]. As an example, consider the equivalent circuit
representation of an electrodynamic loudspeaker as shown
in gure 11. To a rst approximation, the output of an
audio amplier behaves like a voltage source v
g
with an
internal impedance Z
g
. The resistor R
s
represents the series
resistance of the moving coil with self-inductance L
0
. The
mechanical part consists of the mass m of the moving
parts (e.g., coil and diaphragm), the compliance 1/k of
the coil suspension, and the mechanical loss resistance
c. A transformer with ratio 1/A, where A corresponds
to the (projected) area of the vibrating diaphragm of the
speaker, represents the mechano-acoustical coupling. The
equivalent acoustic impedance Z
a
represents the different
acoustic elements, e.g., the radiation impedance and the
acoustic compliance of the speaker box [2]. Similar to
the equivalent circuit for the microphone of gure 10,
the circuit in gure 11 can be reduced to its Th evenin
equivalent, consisting of a Th evenin acoustic pressure
source p
T h
and a Th evenin acoustic impedance Z
T h
in
series. Furthermore, by substituting any of the circuits
of gure 5 for the circuit enclosed by the dashed box
in gure 11, the equivalent circuit representation of an
electrostatic loudspeaker is obtained.
5.3. In-plane parallel microresonators using
comb-shaped driving elements
In-plane parallel microresonators using electrostatic inter-
digitated or comb structures for excitation and detection
of the vibrational motion are used as transducing elements
in a wide variety of applications (see e.g. [14]). Lumped-
parameter modelling is particularly adequate for these kinds
of system. Figure 12(a) shows the layout of a basic con-
guration of an in-plane resonator consisting of a movable
plate of mass m, a folded beam suspension of stiffness k,
and two comb structures, according to Tang et al [14]. The
structure behaves like an SDOF system displaying a sin-
gle mechanical resonance. The comb structures serve as
parallel electrostatic transducers of the type shown in g-
ure 3(b). The movable plate is biased with a voltage v
0
as
indicated. The only difference between the basic parallel
transducer in gure 3(b) and the combs is the number of
active gaps. Each of the combs in gure 12(a) consists
of 16 active gaps or in other words of 16 basic parallel
transducers. The system also illustrates the implementa-
tion of two electrical ports (and a single mechanical port).
Ignoring the mass of the spring suspension, assuming vis-
cous damping with resistance c, and in the absence of an
external mechanical force F
m
, leads to the equivalent cir-
cuit shown in gure 12(b). The numerical subscripts refer
to comb 1 or comb 2. For a symmetric design, x
0
= 0,
C
01
= C
02
= 2nC
0
, and
1
=
2
= 2n, where n denotes
the number of teeth of the movable plate for a single comb
(equal to eight for the design in gure 11), and C
0
and
are given in table 2.
Using the equivalent circuit it becomes very straight-
forward to determine the frequency response of the system.
For instance, the transadmittance Y(i) = i
2
/v
1
for a
short-circuited output (v
2
= 0) is easily derived, yielding
Y(i) =
i
2
v
1

v2=0
=
1
k
i
1

2
_
1 +
1
Q
_
i
0
_
+
_
i
0
_
2
_
(7)
where
0
=

k/m and Q = m
0
/c denote the undamped
resonant frequency and quality factor of the spring
dampermass system, respectively. The above given
transadmittance has the standard second-order form in i
of a series-resonant circuit [10]. For high Q it can be
used for bandpass ltering, whereby the bandwidth BW
is determined by the mechanical quality factor Q of the
resonator: BW =
0
/Q, and whereby
0
establishes the
centre frequency of the lter. Typical amplitude response
plots are shown in gure 12(c).
167
H A C Tilmans
Figure 11. An equivalent circuit representation of an electrodynamic (moving-coil) loudspeaker. The acoustic power is
transmitted into an acoustic impedance Z
a
; details of Z
a
depend on the specic speaker construction and can be found in [2].
Figure 12. A schematic diagram of an electrostatic comb-driven micromechanical resonator according to Tang et al [14]. (a)
A typical layout of a linear resonant plate. (b) An equivalent circuit representation for a zero externally applied mechanical
force, F
m
= 0. The system behaves like an electrical two-port network. (c) Typical amplitude response plots of the
transadmittance i
2
/v
1
evaluated for a short-circuited output (v
2
= 0), as expressed by (7).
5.4. Electromechanical series lters employing in-plane
parallel microresonators
Electromechanical lters are used for signal processing
in for instance telecommunication systems which require
narrow bandwidth (high Q), low loss, good signal-to-
noise ratio, and stable temperature and aging characteristics
[3]. Lin et al [15] have recently described a new
class of passive bandpass electromechanical lters that
employ laterally driven in-plane resonators (of the type
that were introduced in the previous subsection), linked
through coupling springs. The passband characteristics
168
Lumped-parameter electromechanical transducers
Figure 13. (a) A schematic diagram of a series two-resonator electromechanical lter using electrostatic comb-driven
micromechanical resonators, according to Lin et al [15]. (b) An equivalent circuit representation. C
b
represents a parasitic
feedthrough capacitor and/or an intentionally added bridging capacitor. (c) Typical amplitude response plots of the
transadmittance i
2
/v
1
evaluated for a short-circuited output (v
2
= 0) and for different values of the coupling spring, as
expressed by (8).
(centre frequency, shape, and bandwidth) are determined
by the specic design of the individual lter components.
Very important in this respect are also the number of
resonators and coupling springs used [3, 15]. In fact, a
lter conguration consisting of a single resonator and no
coupling springs has already been described in the previous
subsection. A typical response is given by (7) and is
graphically displayed in gure 12(c). The graph clearly
shows that the passband characteristics are far from being
the ideal square shape. Using several resonators that are
linked through coupling springs is a well known method
for obtaining better passband characteristics [3].
An example of a so-called series two-resonator
lter, with a square truss coupling spring, is shown in
gure 13(a). The lter consists of two spring-suspended
masses m
1
and m
2
that are coupled through a relatively
weak coupling spring k
c
. Electromechanical coupling to
each of the masses is accomplished using comb-shaped
electrostatic transducer elements. The system denes an
electrical two-port network, that behaves like an electrical
bandpass lter. The mechanical part of the lter behaves
like a two-degrees-of-freedom lumped-parameter system,
having two (closely separated) mode frequencies that are
determined by the masses m
1
and m
2
and by the three
169
H A C Tilmans
springs k
1
, k
2
, and k
c
. Usually, the structure is designed
such that k
1
= k
2
= k and m
1
= m
2
= m. The
two mode frequencies are then given by

(k/m) and

[(k + 2k
c
)/m]. An equivalent circuit representation of
the lter is shown in gure 13(b). The transduction factors
and the static capacitors of the transducer elements are the
same as explained in section 5.3. Damping is represented
by the viscous resistors c
1
and c
2
, each one associated
with one of the resonators. Note that the coupling spring
experiences a displacement (and therefore a velocity) that
is given by the difference in displacements (velocities)
of the two masses. This example also illustrates the
duality that exists between the equivalent circuit diagram
of gure 13(b) and the mechanical diagram of gure 13(a)
(see also subsection 3.3). For instance the series connection
of the two springmass systems and the coupling spring in
the mechanical diagram becomes a parallel arrangement in
the equivalent circuit. The equivalent circuit also includes
a capacitor C
b
which may either represent an unwanted
feedthrough capacitance between the two electrical ports or
it may represent an intentionally added bridging capacitor
used to further shape the passband characteristics [3].
The lter action is reected in the transadmittance
Y(i), dened as i
2
/v
1
, which can easily be obtained
from the equivalent circuit, either from a direct circuit
analysis or by using circuit simulation software. For a
symmetrical design (m
1
= m
2
= m, k
1
= k
2
= k,
c
1
= c
2
= c, and
1
=
2
= ), assuming short-circuited
conditions at the output (i.e., v
2
= 0) and further in the
absence of capacitor C
b
, the transfer admittance can easily
be obtained directly from gure 13(b), resulting in
Y(i)
i
out
v
in
=
i
2
v
1
=
i
2
k
k
c
/k
1 +2k
c
/k

1
_
1 +
ic
k
+
(i)
2
m
k
__
1 +
ic
k+2kc
+
(i)
2
m
k+2kc
_. (8)
The two mode frequencies (

(k/m) and

[(k + 2k
c
)/m)
indicated before are clearly reected in the above equation.
The frequency response Y(i) of this symmetrical design
is determined by the magnitude of the coupling spring as
illustrated by the graph of gure 13(c) showing typical
amplitude response plots. For k
c
>
0
c, the response
displays two distinct resonant peaks, whereas for k
c
<
0
c,
only a single resonance is observed. For the transition
value, k
c
=
0
c, a at passband is obtained. The latter
condition is generally preferred in practical lter designs.
The response achieved for k
0
c =
c
clearly shows that the
passband more closely resembles the ideal square shape as
compared to the passband obtained for a single resonator as
shown in gure 12(c). For more details, reference is made
to the literature, e.g. [3, 15].
5.5. Vibration sensors
Vibration sensors are employed for measurements on
moving vehicles, on buildings, or on machinery or
as seismic pickups [1]. The basic principle of
vibration measurements is simply to measure the relative
displacement of a mass connected by a (soft) spring to
the vibrating body. An example of a vibration sensor
employing an electrodynamic displacement transducer is
shown in gure 14(a). The transducer detects the mass
displacement x
m
relative to the displacement x
in
of the
vibrating body. In the equivalent circuit the input motion
is modelled using an ideal velocity source u
in
. The
displacement x
in
and acceleration a
in
can be directly
obtained from the velocity as follows: x
in
= u
in
/i
and a
in
= iu
in
. An equivalent circuit representation
of the system in gure 14(a) is shown in gure 14(b).
The resistor R
s
represents the total series resistance of
the coil and interconnecting wires. Note that the input
velocity source, the mass m, and the network consisting
of the series connection of the damper c, the spring k,
and the mechanical port of the transducer, are subject to
the same force. In the equivalent circuit this means that
these three networks are placed in parallel. It is evident
that the velocity (and thus the displacement) experienced
by the spring, the damper, and the mechanical port of
the transducer is given by the difference of the mass
displacement and the vibrational input, u
m
u
in
. From
the equivalent circuit, the frequency response for velocity
measurements can now easily be obtained:
v
e
u
in
=
imR
L

2
+
k
i
_
1 +
1
Q
_
i
0
_
+
_
i
0
_
2
_
_
R
L
+R
s
+iL
0
_

_
i
0
_
2
_
1 +
1
Q
_
i
0
_
+
_
i
0
_
2
_
= B
0
l (9)
where the rst approximation applies for very large load
resistors R
L
and the second approximation is valid at
high frequencies,
0
. Further,
0
=

k/m and
Q = m
0
/c denote the undamped resonant frequency
and quality factor of the springdampermass system,
respectively. Typical amplitudefrequency response plots
are graphically depicted in gure 14(c). The equation above
illustrates that at very high frequencies (
0
) accurate
velocity measurements are possible since the output voltage
is directly proportional to the input velocity. At these high
frequencies the mass practically stays at rest, u
m
0.
For a good design thereof displaying a large bandwidth,
a low resonant frequency is desired, which can be achieved
by choosing a soft spring and/or a large (seismic) mass.
Signals from such transducers can be readily integrated
electrically to obtain displacement information.
Finally, it is pointed out that the electrodynamic
transducer can be replaced by any of the other transducers
shown in gure 3. The equivalent circuit of gure 14(b) is
easily adapted to the new conguration by carrying out the
appropriate replacements of the transducer circuit enclosed
by the dashed box in gure 14(b).
5.6. Systems employing electromechanical feedback
Electromechanical feedback (or force balancing) is often
employed for applications requiring a great accuracy [1].
Instruments employing feedback are even often considered
170
Lumped-parameter electromechanical transducers
Figure 14. (a) A schematic diagram of a vibration sensor using an electrodynamic relative displacement sensor. (b) An
equivalent circuit representation. R
s
represents the series resistance of the moving coil. (c) Typical amplitude response
curves, displaying high-pass second-order lter characteristics.
for use as secondary standards. Examples presented in the
literature of micromechanical devices employing feedback
techniques include capacitive accelerometers [16, 17] and
condenser microphones [18, 19]. An arbitrary example of
a system employing electrostatic feedback, that will be
used to illustrate the principle of feedback and the use of
equivalent circuits herein, is shown in gure 15(a). The
system can be used for measuring the force F
m
exerted
directly on the mass (e.g. as in a microphone [18, 19])
or it can be congured to measure the acceleration a
in
of the case which for this application is rigidly xed
to the vibrating body [16, 17] (compare subsection 5.5).
The upper capacitor senses the induced mass displacement
resulting in a change of the plate charge, which can be
detected by means of a charge amplier. The output
voltage of the charge amplier v
a
is next amplied by a
high-gain (servo)amplier yielding the output voltage v
out
which is fed back to the lower capacitor. This generates
an electrostatic force which is proportional to the relative
mass displacement and which always opposes motion of
the mass from the rest position. This way, the mass itself
is kept very close to the zero-displacement position, i.e.,
u
m
u
in
. In this system, the mechanical spring force is
effectively being replaced by an electrostatic restoring force
which is proportional to the displacement of the mass.
The transfer characteristics of the system in gure 15(a)
can be obtained from its equivalent circuit representation,
which is shown in gure 15(b). At this point, it is
emphasized once more that the following analysis is a
small-signal analysis which does not account for non-
linear behaviour. Including non-linearities in feedback
systems is somewhat more involved (see e.g., [1, 16]).
The charge amplier is modelled as an ideal current- (or
better, charge-) controlled voltage source with sensitivity
171
H A C Tilmans
Figure 15. (a) A schematic diagram of a electrostatic force-balanced transducer for the measurement of forces (including
pressures) F
m
or accelerations a
in
. (b) An equivalent circuit representation. The subscripts s and f refer to the sensing
capacitor (the upper one) and the feedback or actuating capacitor (the lower one), respectively. (c) Typical amplitude
response curves, clearly displaying low-pass second-order lter characteristics.
parameter K
s
= 1/C
F
. Note that the output voltage
of the charge amplier, v
a
=
s
K
s
(x
m
x
in
), is directly
proportional to the relative mass displacement. The analysis
does not change signicantly if the upper-capacitor/charge
amplier combination is replaced by another displacement
sensor, e.g., a differential capacitive detector [1, 16, 17, 19].
The frequency response function as obtained from the
equivalent circuit can be expressed as
H(i)
v
out
F
m
ma
in
=
A
s
K
s
k

+A
s

f
K
s
1
_
1 +
1
Q
_
i

0
_
+
_
i

0
_
2
_

f
1
_
1 +
1
Q

_
i

0
_
+
_
i

0
_
2
_

1

f
(10)
where the rst approximation applies for very high closed-
loop gains, resulting in a large electrical spring constant
k
e
A
s

f
K
s
k

, and the second approximation is


172
Lumped-parameter electromechanical transducers
Figure 16. (a) A schematic diagram showing the implementation of Q-control for a three-port electrostatic comb-driven
micromechanical resonator according to Nguyen and Howe [20]. (b) An equivalent circuit representation. The subscripts s,
f , and in refer to the sense electrode, the feedback electrode, and the driving or input electrode, respectively.
valid at low frequencies,

0
. Further,

0
=
0
_
1 +
A
s

f
K
s
k


0
_
A
s

f
K
s
k

=
0
_
k
e
k

=
_
k
3
m
(11)
and
Q

= Q
_
1 +
A
s

f
K
s
k

Q
_
A
s

f
K
s
k

= Q
_
k
e
k

=
_
mk
e
c
(12)
where
0
=

k

/m and Q = m
0
/c =

m/c denote
the undamped resonant frequency and quality factor of the
springdampermass system, respectively. Note that for
a symmetrical design the bias displacement x
0
= 0, and
that the transduction factors
s
and
f
as well as the static
capacitors C
0s
and C
0f
are equal, respectively given by

s
=
f
= =
0
A
e
v
0
/d
2
and C
0s
= C
0f
= C
0
=

0
A
e
/d, as presented in table 2. Moreover, as indicated
in the equivalent circuit of gure 15(b), it follows that k

is now given by k

= k k

s
k

f
= k 2k

, where the
latter equality applies for a symmetric design with k

2
/C
0
(compare table 4). Typical amplitudefrequency
response plots are graphically depicted in gure 15(c). The
system displays second-order low-pass lter characteristics.
Equation (11) clearly shows that, due to the feedback,
the bandwidth, compared to a system without feedback,
is increased by a factor

k
e
/k

, where k
e
denotes the
electrical spring constant. Thus, the mechanical spring k

may be considered to be replaced by an electrical spring


k
e
, which offers a greater linearity and accuracy and lack
of hysteresis. Other advantages of employing feedback
are the already indicated increased bandwidth, and further
173
H A C Tilmans
the lower distorsion and lower sensitivity to instabilities
due to environmental factors (e.g. temperature) of some
parameters (e.g., the mechanical spring) [1]. Disadvantages
of feedback systems are usually the greater complexity and
higher costs.
Equation (12) indicates that also the quality factor
increases by a factor

k
e
/k

due to the feedback. This


increase is generally unwanted however, and care must
be taken to ensure a low Q such that Q

will still
be low enough. For critical damping, and thus a
at response, the quality factor Q

should be equal to
1/

2. The observation that feedback may also affect


the damping properties, as reected in the quality factor,
has been utilized by Nguyen and Howe [20] in parallel
microresonators to precisely control the quality factor and
moreover to make it independent of the ambient operating
pressure of the micromechanical system. Quality factor
control is considered necessary for the bandpass lters
that were described in subsections 5.3 and 5.4. This
is easily understood realizing that the passband shape
strongly depends upon the quality factor of the constituent
resonators. The structure proposed by Nguyen and Howe
[20] to accomplish this is shown in gure 16(a). It
comprises an electrical three-port. One port is used as the
drive or input port, the second as the sense or output port,
and the third is used for feedback purposes as indicated
in the gure. The corresponding equivalent circuit is
shown in gure 16(b). Circuit analysis yields the following
expression for the transfer function:
v
out
v
in
=
i
s

in
R
a
/k
_
1 +
1
Q

_
i
0
_
+
_
i
0
_
2
_
(13)
where the undamped resonant frequency
0
=

k/m and
the quality factor Q

is given by
Q

=
Q
1 +
s f Ra
c

Q
s f Ra
c
=
m
0

f
R
a
=

km

f
R
a
. (14)
The latter equation indicates that the quality factor can be
made independent of the viscous damping parameter c and
thus independent of ambient pressure (and temperature)
uctuations as these are reected in c. Furthermore the
quality factor will be lowered through the application of
the negative feedback. By using positive feedback, on
the other hand, it is possible to raise the quality factor.
Positive feedback can be realized by merely changing R
a
from positive to negative, or by feeding back the output
signal to the driving site instead of the sensing site. For
more details, reference is made to [20].
6. Conclusions
The underlying theory describing the electromechanical be-
haviour of electrostatic, electromagnetic, and electrody-
namic lumped-parameter transducers has been presented
and has been illustrated with a wide range of practical de-
vices. The results are summarized in the form of small-
signal equivalent circuits that form an exact representation
of the linear electromechanical behaviour of the transduc-
ers. Irrespective of the type of transducer employed, it
turned out that the circuits all display the same principal
arrangement, consisting of three main parts: an electrical
part, an electromechanical coupling part, and a mechanical
(including acoustical) part.
The equivalent circuits can be used to determine
the frequency and transient response of the transducer,
thereby conveniently making use of widely available circuit
simulation software such as SPICE. The theory as presented
in this paper denes a fundamental basis for further
investigation of the transducer characteristics. For instance,
the equivalent circuits can be used as a starting point to
study the effects of connecting subsystems or of making
modications to the structure.
Acknowledgment
The author greatly acknowledges Professor Jan Fluitman
of the Mesa Research InstituteUniversity of Twente
for the many fruitful discussions on the principles of
electromechanical transduction.
Appendix A. Lumped-parameter systems: the
constitutive equations
In order to illustrate the underlying theory, consider the
ideal electrostatic transducer of gure 3(a). The derivations
are based on the principle of conservation of system energy
[8]. The state variables are the displacement x
t
and the
capacitor charge q
t
. In a dynamic analysis of the system all
variables are functions of time t , but explicit indication of
this dependence is omitted as a matter of convenience. The
electric energy W
e
contained in the electrostatic transducer
is given by [8]
W
e
= W
e
_
q
t
, x
t
_
=
q
2
t
2C(x
t
)
=
q
2
t
(d +x
t
)
2
0
A
e
(A1)
where C(x
t
) =
0
A
e
/(d +x
t
) denotes the capacitance as a
function of x
t
, A
e
is the area of the capacitor plates, and
d the spacing of the uncharged plates. Taking the total
differential of W
e
results in
dW
e
=
_
W
e
q
t
_
xt =const ant
dq
t
+
_
W
e
x
t
_
qt =const ant
dx
t
.
(A2)
For a conservative transducer in thermodynamic equilib-
rium, the energy put into the transducer through the electric
and mechanical ports is given by
dW
e
= v
t
dq
t
+F
t
dx
t
(A3)
where v
t
is the voltage between the plates and F
t
is the
mechanical force acting on the movable plate. Subtracting
(A2) and (A3) and noting that dq
t
and dx
t
can have arbitrary
values, the resulting equation must be satised by requiring
that the coefcients of dq
t
and dx
t
must be zero. Hence,
v
t
_
q
t
, x
t
_

W
e
(q
t
, x
t
)
q
t

xt =const ant
=
q
t
(d +x
t
)

0
A
e
(A4a)
and
F
t
_
q
t
, x
t
_

W
e
(q
t
, x
t
)
x
t

qt =const ant
=
q
2
t
2
0
A
e
. (A4b)
174
Lumped-parameter electromechanical transducers
The above equations dene the terminal voltage and the
force as being the effort variables at the respective ports.
The equilibrium values are given by the partial derivatives
of W
e
with respect to the corresponding state variable [8].
Note that F
t
is the externally applied force necessary to
achieve equilibrium. It is evident that in magnitude it
is equal to the electrostatic Coulomb force between the
charged plates of the capacitor, but has opposite direction.
Also note that the force displays a quadratic dependence
on the charge, which makes the system non-linear.
Linearization is commonly accomplished by introducing
bias or equilibrium conditions. The equations which
describe the linear relations between the incremental or
small-signal effort variables and state variables, determined
at the bias point dened by a displacement x
0
and charge q
0
,
are called the constitutive equations. For the electrostatic
transducer these are given by
v(q, x) =
v
t
q
t

0
q +
v
t
x
t

0
x
=
(d +x
0
)

0
A
e
q +
q
0

0
A
e
x =
1
C
0
q +
v
0
x
0
x (A5a)
and
F(q, x) =
F
t
q
t

0
q +
F
t
x
t

0
x
=
q
0

0
A
e
q +0 x =
v
0
x
0
q +0 x. (A5b)
To obtain the very last expressions on the right hand side,
the following equality was used: q
0
= C
0
v
0
=
0
A
e
v
0
/x
0
,
where v
0
denotes the bias voltage and C
0
denotes the static
or bias capacitance. It is noted that the bias signals are
independent of time since they dene the static equilibrium
condition.
The ideal transducer described above is not of great
practical use since it is not stable. This can be reasoned as
follows. Assume the system is in equilibrium, meaning that
the externally applied mechanical force is counterbalanced
by the electrostatic force. Now if, due to some disturbance,
the movable plate is displaced a little in the direction of the
xed plate (i.e., the negative x direction) while keeping
the applied voltage constant, the attractive electrostatic
force will increase a little and becomes somewhat larger
than the external mechanical force. This will increase
the displacement of the plate even further, which in its
turn results in a further increase of the electrostatic force.
This will go on until the gap spacing is reduced to zero.
Similarly, if the movable plate were initially displaced a
little further away from the xed plate (i.e., in the positive
x direction), force equilibrium is again destroyed. Now, the
mechanical force exceeds the ever decreasing electrostatic
force and the plate will eventually disappear into innity.
Stability is easily attained by including a mechanical spring
with constant k, e.g., as shown in gure 4. Analytically this
means that a mechanical energy term must be added to the
energy function of (A1), resulting in
W
em
= W
em
_
q
t
, x
t
_
=
q
2
t
2C(x
t
)
+
1
2
k
_
x
t
x
r
_
2
=
q
2
t
(d +x
t
)
2
0
A
e
+
1
2
k
_
x
t
x
r
_
2
(A6)
where x
r
denotes the rest position of the spring.
Equation (A4a) is not affected this way, but (A4b) is:
F
t
_
q
t
, x
t
_

W
em
(q
t
, x
t
)
x
t

qt =const ant
=
q
2
t
2
0
A
e
+kx
t
.
(A7)
Finally, the second of the constitutive equations, (A5b),
must be replaced by
F(q, x) =
F
t
q
t

0
q +
F
t
x
t

0
x =
q
0

0
A
e
q +kx =
v
0
x
0
q +kx.
(A8)
The constitutive equations (A5a) and (A8) are expressed
as a relation of the (q, x) type, whereby the state variables
are chosen as the independent variables. Sometimes it is
more convenient to choose the voltage and the displacement
as the independent variables. The electromechanical
interactions are now described by equations of the (v, x)
type, given by
q(v, x) =

0
A
e
d +x
0
v
q
0
d +x
0
x
=

0
A
e
d +x
0
v

0
A
e
v
0
(d +x
0
)
2
x (A9a)
F(q, x) =
q
0
d +x
0
v +
_
k
q
2
0

0
A
e
(d +x
0
)
_
x
=

0
A
e
v
0
(d +x
0
)
2
v +
_
k

0
A
e
v
2
0
(d +x
0
)
3
_
x. (A9b)
It can easily be shown that the equilibrium position of the
transducer, including the spring k, is, apart from excessive
bias loads [10], stable. In fact the system is stable as long
as k > k

, where k

=
0
A
e
v
2
0
/(d + x
0
)
3
, i.e., the second
term within parentheses in (A9b).
Appendix B. Nomenclature
A beam or diaphragm area (subjected to
acoustic pressure) [m
2
]
A
e
effective electrode area (also for the
electromagnetic transducer) [m
2
]
B
0
applied bias magnetic induction [T]
c viscous drag parameter [N s m
1
]
C
0
static or bias capacitance of the
electrostatic transducers [F]
C
p
parasitic capacitance [F]
d gap spacing at rest, also called the
zero-voltage gap spacing [m]
F, F
0
, F
t
incremental, bias (static) and total
applied mechanical force [N]
i, i
0
, i
t
incremental, bias (static) and total
current [A]
k mechanical spring constant [N m
1
]
k

spring induced by electromechanical


coupling effects [N m
1
]
k

effective dynamic spring constant


as a result of electromechanical coupling
[N m
1
]
175
H A C Tilmans
l active wire length of the moving
coil coupled with the magnetic eld [m]
L
0
static or bias self-inductance of
the electromagnetic transducer [H]
L
s
self-inductance of the current loop
of the electrodynamic transducer [H]
N number of windings of the driving
coil for the electromagnetic (and
moving coil) transducer []
p
g
generator acoustic pressure [N m
2
]
p, p
0
, p
t
incremental, bias (static), and total
acoustic pressure [N m
2
]
q, q
0
, q
t
incremental, bias (static), and total
charge [C]
Q, Q

quality factors []
u incremental velocity [m s
1
]
v, v
0
, v
t
incremental, bias (static), and total
terminal voltage [V]
x, x
0
, x
t
incremental phasor, bias (static),
and total displacement [m]
Z
a
acoustic impedance [N s m
5
]
transduction factor of the electrostatic
transducer [N V
1
]

0
permittivity of free space [F m
1
]
,

(acoustic) volume displacement [m
3
]
and volume velocity [m
3
s
1
]
electromechanical coupling factor []
,
0
,
t
incremental, bias (static), and total
ux linkage [Wb]

incremental ow of total ux linkage


[Wb s
1
]

0
permeability of free space [H m
1
]
transduction factor of the lumped-
parameter electromagnetic and
electrodynamic transducers [N A
1
]
radial frequency [rad s
1
]

0
undamped resonant frequency [rad s
1
]
References
[1] Neubert H K P 1975 Instrument Transducers 2nd edn
(Oxford: Clarendon)
[2] Rossi M 1988 Acoustics and Electroacoustics (Norwood,
MA: Artech)
[3] Johnson R A 1983 Mechanical Filters in Electronics (New
York: Wiley)
[4] Skudrzyk E J 1958 Vibrations of a systems with a nite or
an innite number of resonances J. Acoust. Soc. Am. 30
114052
[5] Mason W P 1935 An electromechanical representation of a
piezoelectric crystal used as a transducer Proc. Inst.
Radio Eng. 125263
[6] Berlincourt D A, Curran D R and Jaffe H 1964
Piezoelectric and piezomagnetic materials and their
function in transducers Physical Acoustics vol 1A, ed W
P Mason (New York: Academic) pp 169270
[7] Tilmans H A C 1996 Equivalent circuit representation of
electromechanical transducers: II. Distributed-parameter
systems J. Micromech. Microeng. submitted
[8] Woodson H H and Melcher J R 1968 Electromechanical
Dynamics Part I, II and III (New York: Wiley)
[9] Meirovitch L 1975 Elements of Vibration Analysis (New
York: McGraw-Hill)
[10] Tilmans H A C 1993 Micromechanical sensors using
encapsulated built-in resonant strain gauges PhD
Dissertation University of Twente
[11] Huelsman L P 1984 Basic Circuit Theory 2nd edn
(Englewood Cliffs, NJ: Prentice Hall)
[12] Prak A, Lammerink T S J and Fluitman J H J 1993 Review
of excitation and detection mechanisms for
micromechanical resonators Sensors Mater. 5 14381
[13] Scheeper P R, van der Donk A G H, Olthuis W and
Bergveld P 1994 A review of silicon microphones
Sensors Actuators A 44 111
[14] Tang W C, Nguyen T-C H and Howe R T 1989 Laterally
driven polysilicon resonant microstructures Sensors
Actuators 20 2532
[15] Lin L, Nguyen C T-C, Howe R T and Pisano A P 1992
Micro electromechanical lters for signal processing
Proc. Micro Electro Mechanical Systems (MEMS92)
(Travem unde, 1992) pp 22631
[16] Henrion W, DiSanza L, Ip M, Terry S and Jerman H 1990
Wide dynamic range direct digital accelerometer Tech.
Dig. IEEE Solid-State Sensor and Actuator Workshop
(Hilton Head Island, SC, 1990) pp 1537
[17] Kuehnel W and Sherman S 1994 A surface micromachined
silicon accelerometer with on-chip detection circuitry
Sensors Actuators A 45 716
[18] Van der Donk A G H, Scheeper P R, Olthuis W and
Bergveld P 1992 Amplitude-modulated
electro-mechanical feedback system for silicon
condenser microphone J. Micromech. Microeng. 2 2114
[19] Bay J, Hansen O and Bouwstra S 1995 Modelling of
micromachined microphones with differential capacitive
read-out and force balancing Tech. Dig. 6th Workshop on
Micromachining, Micromechanics and Microsystems
(MME95) (Copenhagen, 1995) pp 2214
[20] Nguyen C T and Howe R T 1992 Quality factor control for
micromechanical resonators Tech. Dig. IEEE Int.
Electronic Devices Meet. (IEDM92) (1992) pp 5058
176