105

The Matlab/Simulink Modeling and Numerical Simulation of an Analogue

Capacitive Micro-Accelerometer. Part 1: open loop
Teodor Lucian Grigorie
Abstract - The paper deal with an open loop analogue until a few mm3 order or more ([5]). The initial prices of the
capacitive micro-accelerometer both from the point of view electronic sensors were high enough but the initiation of the
of mathematical modeling and through the achievement of series production had an important influence in their decreasing.
certain numerical simulations which must confirm the It exist much more types of acceleration sensors, but all of
mathematical models and in the same time help to establish these are based on the displacement of a proof mass because of
the optimal architecture of the system. The capacitive the inertia force generated by the applied acceleration. What
transducer is one of differential type, with two fixed plates makes them different is the method used to measure the proof
and one moving, which play the role of proof mass. mass displacement.
Structure of the system contains the proper transducer and The electronic detection of the acceleration can be made
an electronic device (charge amplifier + phase-sensitive using three principal methods: the piezoresistive method ([6-9]),
demodulator) to process the signal obtained from the tunneling current method ([10, 11, 12]) and the capacitive
transducer. The accelerometer is mathematical modeled, the method ([13-18]). Also, the proof mass of the accelerometer can
mathematical relation being expressed both in complex has the form of an elastic lamella embedded at one end or the
domain and in time. The mathematical modeling contains form of a plan plate attached in two or more points on the fixed
two distinct phases: the modeling of the proper structure of the accelerometer through elastic coupling. The
accelerometer, which includes the proof mass and elastic determination of the acceleration using first method is made
element, and the modeling of the differential capacitive using a piezoresistor fixed on the base of the embedded lamella;
transducer for the proof mass displacements. Using these the piezoresistor resistance is modified by the lamella deviation.
mathematical relations, the block schemas are realized for The method works successfully in case of relative high
each block separately and for the full system. Starting from deviations of the lamella, and she has some draw back if we
the numerical values of the accelerometer’s parameters the want a great resolution for the acceleration detection. More, the
numerical simulations with programming piezoresisitive elements have complex variations if the ambient
MATLAB/SIMULINK are made for each of the deduced temperature is changing what limiting their using at the
block schemas. manufacturing of the accelerometers with medium sensitivity.
Keywords: Capacitive Micro-Accelerometer, The second method supposes the presence of a sharp metallic
Mathematical modeling, Block schemas, Numerical tip, on the proof mass, and an electrode, micro-machined on the
simulation. fixed structure of the accelerometer. A tunneling current of
1. INTRODUCTION about 1nA can be measured, when a DC voltage bias is applied
between the sharp metallic tip and the electrode. The current
The acceleration sensors have a great importance in strap- depends exponential by the distance between the sharp metallic
down inertial navigation because the architecture of a strap- tip and the electrode. The proof mass displacement can be
down inertial system includes three such sensors in classic case measured with a 0,001 Å precision. The method is very precise,
at the level of the sensing function and six such sensors in case but very expensive because of the complicated manufacturing
of the substitution of the gyrometers with accelerometers ([1, 2, technology: the cavity that contains the proof mass must be
3]). vacuumed, the conductive layers obtained through deposition
As the strap-down navigator contains obligatory a navigation have a very small thickness and are currently made from noble
computer it is obvious that the data acquisition from the inertial metals, the control of the distance between the sharp metallic tip
sensors must be interfaced with specialized circuits for signals and the fixed electrode must be very good because the
conditioning ([4]). That supposes additional costs if mechanical transducer has a big sensitivity.
sensors are used. However, the discoveries in physics from the The capacitive method is developed much enough in the last
last years and the improvement of the manufacture technologies years because of the possibilities of signal processing with the
in electronics domain did possible the development of the technology of integrated circuits. This one presents the
electronic acceleration sensors simultaneously with the advantage of the miniaturization, but necessitates the presence
mechanical sensors. The obtained results were and are of specialized circuits for signal processing. So, the MEMS
extremely spectaculars, thing reflected also by the evolution of technology had a good start, leading to the realization on the
the strap-down inertial navigation systems. same substrate of an acceleration transducer which includes the
The researches were developed on three principal directions: afferent circuits for signals conditioning. The manufacturing
the improvement of the performances, the miniaturization of the technology and the adjustments which must be made in case of
electronic sensors and the obtaining of a price capable to the capacitive method are more accessible comparatively with
compete with mechanical sensors. The development of the those in the tunneling current method.
components and the appearance of some manufacture The principal objective of the paper is to model
technologies with good performances brought the electronic mathematically and to simulate numerically, using the
sensors in front of the mechanical sensors from the point of MATLAB/SIMULINK models, an analogue capacitive micro-
view of the performances, and were decreased their dimensions accelerometer ([14]) for its full characterization from the both

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 106

points of view of stationary regime and dynamically regime. ε0 A v12 ε 0 A V12

II. BASIC OPERATIONS Fel1 = = (1 − cos 2ωt ),
2 (d 0 − x ) 4 (d 0 − x )
2 2

(5)
In principle the accelerometer operation relies on the action ε A v22 ε A V12
of the input acceleration of a proof mass that has a Fel 2 = 0 = 0 (1 − cos 2ωt ).
2 (d 0 + x ) 4 (d 0 + x )
2 2
parallelipipedic shape and is placed on a substrate using four
silicon flexible bars (fig. 1 [13]). On the same substrate there are
two electrodes, placed on opposed sides of the mobile plaque.
The couple of fixed electrodes and the mobile plaque form two The frequency of supplying voltages is very high ( f = 105 Hz ),
capacitors whose capacities vary with the application of an so that the components with sine variation from equations (5)
acceleration that has the ability to change the proof mass are strongly filtered by the sensitive element that behaves like a
position. low-pass mechanical filter. As a consequence the two
r electrostatic forces can be assumed as constant and their
The physical quantities from fig. 1 are: m - proof mass, a ,
r expressions are:
y acceleration and position of the carrying vehicle, a x , x
acceleration and proof mass position with respect to the carrying ε 0 A V12
r r r Fel1 = ,
vehicle, Fe , Fa , Fi - the elastic, the damping and respectively 4 (d 0 − x )
2

(6)
the inertial forces. The absolute values of the three forces are ε 0 A V12
provided by the following equations: Fel 2 = .
4 (d 0 + x )
2

Fe = kx,
v1
dx
Fa = b , (1) r
dt Fel1
2
d y . d0
Fi = ma = m , x
dt 2 r
Fel 2
where k is the resulting elasticity constant and b is the viscous d0
damping coefficient.
r r r
Fe Fa Fi
v2
r
x ax
m r
y a Fig. 2. Action of the electrostatic forces

The resulting electrostatic force will be

Silicon flexible bars
Fig. 1. Mechanical model of the accelerometer ε 0 AV12 ⎡ 1 1 ⎤
Fel = Fel1 − Fel 2 = ⎢ − 2 ⎥
, (7)
4 ⎣ (d 0 − x ) (d 0 + x ) ⎦
2

When the fixed electrodes are not supplied by electric

energy, according to fig. 1 the motion equation of the proof and the equation (2) becomes
mass is ([13, 17, 18])
d2 y d2 x dx
d2 y d2 x dx m 2
= m 2 + b(x ) + kx − Fel . (8)
m 2 = m 2 + b + kx. (2) dt dt dt
dt dt dt
Using (3) and (7), the equation (8) will be
The accelerometer is miniaturized, so the viscous damping is
provided by the air that is enclosed in its watertight carcass. For d2 x μA 2 ⎡ 1 1 ⎤ dx
=a− ⎢ + 3 ⎥
−
relatively large displacements of the mobile plaque (as high as 2m ⎣ (d 0 − x ) (d 0 + x ) ⎦ dt
2 3
dt
μm ) the viscous amortization is given by the expression ([14])
k ε AV 2 ⎡ 1 1 ⎤
− x+ 0 1 ⎢ − 2 ⎥
. (9)
1 2⎡ 1 1 ⎤ 4m ⎣ (d 0 − x ) (d 0 + x ) ⎦
2

b(x ) = μA ⎢ + , m
3 ⎥
(3)
⎣ (d 0 − x ) (d 0 + x ) ⎦
3
2
III. MATLAB/SIMULINK MODELS AND NUMERICAL
where μ is the air viscosity, A is the mobile plaque area and SIMULATION OF THE ACCELEROMETER BLOCKS
2d 0 is the distance between the fixed electrodes. Fig. 3 depicts the block diagram of the accelerometer
If the fixed electrodes are supplied by the voltages mechanical model. One denotes by C1 and C2 the lowest and
v1 = V1 sin ωt , highest capacities of the variable capacitors. They can be
(4) simulated as follows
v2 = −v1 = −V1 sin ωt ,

then the following electrostatic forces will act over the mobile
plaque (fig. 2)

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 107

ε 0 ε r A ε 0 ε r Ad 0 ε 0 ε r Ax sC3 R2
C1 = = + , Vi (s) = −V0 (s) . (14)
d 0 − x d 02 − x 2 d 02 − x 2 1 + sC4 R2
(10)
ε 0 ε r A ε 0 ε r Ad 0 ε 0 ε r Ax
C2 = = − . Using (11) and (13), the equation (14) becomes
d 0 + x d 02 − x 2 d 02 − x 2
2s 2C3 R2 ΔCV1 (s)
Vi (s) = − . (15)
(1 + sC4 R2 )[s(2C0 + C3 ) + 1 / R1 ]
2
d x dx
a + dt
2
1 dt 1 x
Σ - s s The block diagram of the detection device can be modeled as
+ -
depicted by fig. 5.
× μA2 ⎡ 1 1 ⎤ v1
⎢ + 3⎥
2m ⎣ (d 0 − x ) (d 0 + x ) ⎦
3

x ε 0ε r Ax ΔC − 2s 2C3 R2 vi
k d0 − x2
2 × (1 + sC4 R2 )[s(2C0 + C3 ) + 1 / R1 ]
m
Fig. 5. Block diagram of the detection device
ε 0 AV1 ⎡ ⎤
2
1 1 In order to simulate the system, one considers for it the
⎢ − ⎥
4m ⎣ (d 0 − x )2 (d 0 + x )2 ⎦
following parameters ([14]): m = 8,2mg , A = 12mm 2 ,
d 0 = 10μm , k = 83,3N/m , μ = 1,8 ⋅ 10 −5 Ns/m 2 ,
εr = 1 ,
Fig. 3. Block diagram of the mechanical model
If x << d 0 , one can consider the approximation f = 100kHz , V1 = 0,5V , R1 = 250MΩ , R2 = 820 kΩ ,
C3 = 1nF , C4 = 22pF , C0 = 10,6pF .
C1 ≅ C0 + ΔC , One represents the accelerometer block diagram in
(11)
C 2 ≅ C0 − ΔC , MATLAB/SIMULINK and determines the voltage vi with
respect to the input acceleration. The model from fig. 6 is
where
obtained.
ε0ε r A m Fi
C0 = , Gain4 valori Fi x
valori x
v1
d0 vezi x
(12) acceleratia + -2*R2*C3
ε 0 ε r Ax +
- 1/s 1/s f(u) *
C3s+1/R1
vi
ΔC = . - Integrator
Sum
Integrator1 Fcn2 Product1
Transfer Fcn1
valori vi

d 02 − x 2 Dc
* vezi Dc valori Dc
vezi vi
Product
To determine the proof mass displacement x and, implicit the m Gain1 f(u)
Fcn
applied acceleration a , the capacitive detection device must be Fa
valori Fa
coupled with a charge amplifier whose output voltage reflects vezi Fa

the acceleration modification (fig. 4 [14, 15]). -K-

Gain
C4 m Gain2
Fe
vezi Fe valori Fe
v1 R2
f(u)
R1 C1 C3
.v 0 _ m Gain3
Fel
Fcn1

C2 + vi
vezi Fel valori Fel

v2 Fig. 6. Simulink model of mechanical part with detection device

Step-like acceleration signals were applied at the model input
and there were obtained the variation of the mobile plaque
Fig. 4. Detection device with charge amplifier position x(t ) (fig. 7) and the variation of the capacity difference
The capacitor C3 is used in order to block the direct voltage ΔC (t ) (fig. 8).
signals that could possibly occur on the mobile plaque and the The values of the step-like acceleration signals used for
resistor R1 maintains the proof mass to a defined potential V0 simulation are 1g , 2 g , 4 g , 6 g and 8 g along a time-interval
and provides a time constant τ = C1 / 2 R1 that is much higher as of 0,1s. The step-like signals were applied after 0,02 s. From
compared to the excitation signal period. fig. 7 and fig. 8 one can notice an increase of the time constants
The proof mass potential V0 can be calculated using the once the input signal is increased. The explanation of this
phenomenon consists in the non-linear features of the reaction
equation: from the block-scheme of the mechanical element (the
V1 (s)sC1 + V2 (s)sC2 amortization and electrostatic forces). One can also notice that
V0 (s) = , (13) the time constants of x(t ) exhibit an approximately
s(C1 + C2 + C3 ) + 1 / R1
proportional increase starting from 5ms. By contrast, the
and the potential from the amplifier output can be calculated increase of ΔC (t ) time constants becomes more and more
using the equation significant starting from 10 ms.

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 108

x [m × 10 −6 ] From the graphic one can see that the electrostatic force Fel
8
8g is 3-order lower than the others but tends to a constant value.
7 During the transient regime the amortization and elastic force
6 change their weights from 100% to 0% and vice-versa during
5
6g
their balancing.
This way, the amortization force presents an initial jump
4
4g toward the maximum value, then falls slowly and reaches 0 after
3 the mobile plaque movement amortization, while the elastic
2 force raises from 0 and becomes steady at a value that equals the
2g
1 some of the other two forces. When the time-characteristics
1g
from fig. 9 there were considered the senses of forces that act
0
0 0.02 0.04 0.06 0.08 0.1 over the proof mass.
t [s ] x [m ×10−6 ]
Fig. 7. Variation of the mobile plaque position x(t ) 6
8g
ΔC [F × 10 −11 ]
4 6g
2
4g
8g
2
2g
1.5 1g
0

1 -2
6g
-4
0.5
4g
2g
-6
0 0.01 0.02 0.03 0.04 0.05
0
1g t [s]
0 0.02 0.04 0.06 0.08 0.1 Fig. 10. Displacement x of the proof mass for sine inputs
t [s ]
Fig. 8. Variation of the capacity difference ΔC (t ) The analysis of system response for sine inputs reveals
Based on these observations one can expect a linear interesting conclusions. The time evolution of x displacement
dependence of the steady values of x with respect to a . In the of the proof mass for sine inputs with various magnitudes is
same time the dependence of the steady values of ΔC with depicted by fig. 10. The simulation was accomplished for sine
respect to a is expected to exhibit a strong non-linear behavior input signals with frequency of 40Hz and magnitudes of 1g ,
starting with a certain value of the applied acceleration. The 2 g , 4 g , 6 g and respectively 8 g . The effect of the system
non-linear features are also increased by the non-linear block
non-linear features can be noticed in this case too. The system
placed between x(t ) and ΔC (t ) , where the weight of x 2 from
answer, x(t ) , is a sine wave who presents distortions that grow
the nominator is increased as a is increased.
with the input signal magnitude. Quite from the beginning (for
We calculated the forces Fi (inertial force), Fel (electrostatic time intervals comparable to the time constants), one can notice
force), Fa (damping force), Fe (elastic force) and we visible phase changes between the outputs with different
represented them with respect to time for an interval of 0,025s magnitudes, but this effect is amplified in the high magnitude
inputs domain. Based on these notices one can conclude that the
(fig. 9). The model simulation was performed for a step-like accelerometer in open loop can be also used as a sensor within a
signal with the value a = 3 g , applied at t = 0,005s . limited range of accelerations.
Forces [ N ×10−4 ] The curves from fig. 11 and fig. 12 represent the steady
3 values of x(t ) and ΔC (t ) for accelerations within the range
2
Fi −10 g ÷ 10 g . The dependency x(a) is approximately linear,
proving that the mechanic model behaves well as sensitive
1
103 × Fel element within the entire range −10 g ÷ 10 g . The situation is
0 different with the capacitive position transducer. The curve
Fa
ΔC (a) reveals the introdu-cing of some strong non-linearities
-1
for higher accelerations of 3 ÷ 4 g , so the measure range must
-2 be limited to −4 g ÷ 4 g .
Fe
-3
0 0.005 0.01 0.015 0.02 0.025
t [s]
Fig. 9. Graphics of the forces

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 109

that is modulated in magnitude with ΔC (t ) . Under the

x [m ×10−5 ]
conditions in which x << d 0 , the equation (18) becomes
1

2ε 0 ε r AV1
vi (t ) ≅ − x sin ωt (19)
0.5 C 4 d 02

and its magnitude will become direct proportional to x and

0
consequently to a . The curves from fig. 13 were obtained
through the simulation of the model from fig. 6 for a time of
-0.5 0,03 s at step-like inputs of acceleration with values of 0,5 g ,
2 g , 4 g and respectively 6 g . The curves exhibit an
-1 approximately linear dependency with the acceleration for input
-10g -5g 0 5g 10g
values under 4 g . For values higher than 4 g , the acceleration
a [ m/s 2 ]
results in an increase of the weight corresponding to the term
Fig. 11. Steady values of x(t ) d 02 − x 2 from equation (18) and equation (19) is no longer valid.

ΔC [F × 10 −11 ] vi [ V]
8 0.4
6g
6 0.3
4g
4 0.2
2 2g
0.1
0 0.5g
0
-2
-0.1
-4
-6 -0.2

-8 -0.3
-10g -5g 0 5g 10g
a [m/s ]2 -0.4
0.02998 0.029985 0.02999 0.029995 0.03
t [s]
Fig. 12. Steady values of ΔC (t )
Fig. 13. Accelerometer’s response at step-like inputs
Considering the model from fig. 6, if one separates the part
According to the model from fig. 6, the sine voltage v1 (t ) is
between ΔC and vi , and perform a simulation for constant
modulated in magnitude by the signal ΔC (t ) . From (15), if inputs of ΔC , one gets the curves from fig. 14 (for ΔC > 0 )
C3 >> 2C0 , 1/ R1 → 0 and sC 4 R2 >> 1 ([14]), we get and fig. 15 (for ΔC < 0 ). Owing to the high frequency of the
signal, it was extracted along a short interval, analogous to the
Vi (s) = −
2ΔC
V1 (s), (16) procedure used for fig. 13. The magnitudes of voltages vi (t ) are
C4 direct proportional to ΔC . This confirms both the passing from
the equation (15) to equation (16) through approximation and
that is
respectively the converting from complex domain to time
2ε 0 ε r A x domain reflected by equations (16), (17) and (18). Another
Vi (s) = − ⋅ 2 V1 (s ). (17) relative simply observation is that the switching of ΔC sign
C4 d0 − x2
results in a modification of the signal vi (t ) with a value of π .
From the previous simulations one can see that the values of the
mechanic model time constants are as high as ten of ms. Owing
to the high frequency of the sine signal v1 (t ) ( f = 105 Hz ) one
can consider that the signal x1 (t ) is approximately constant with
respect to v1 (t ) and the equation (17) can be translated in the
time-domain under the form
−2ΔC 2ε ε A x
vi (t ) = v1 (t ) = − 0 r v1 (t ). (18)
C4 C4 d 0 − x 2
2

Practically, one must wait for the output signal of the load
amplifier vi (t ) to get the same shape as the sine signal v1 (t )

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 110
ΔC=1 pF of v1 and decoupled on the other. Both interrupters operate in a
0.05
0 conjugate manner so as the circuit performs a function that is
-0.05
3 3.2 3.4 3.6 3.8 4
similar to the rectification. The voltage sign vd is then filtered
ΔC=2 pF using a low-pass filter. In this way the output signal v f will
0.1
0 contain information concerning both the magnitude of vi (t ) and
-0.1
3 3.2 3.4 3.6 3.8 4 respectively its phase (through its sign).
vi [V]
ΔC=4 pF
0.2
R4 R6
0 C5
R3 R5
-0.2 - v′i vi′′ vin vd R7 R8
3 3.2 3.4 3.6 3.8 4 vi +
- -
+ + vf
ΔC=6 pF
Amplifier Inverter R10
0.2 vid C6
R9
0
-0.2
v1 1
3 3.2 3.4 3.6 3.8 4 -1 LPF
t [s x 10-4] sgn

Fig. 14. vi (t ) dependences for ΔC = constant ( ΔC > 0 )

ΔC=-1 pF
Fig. 17. Demodulator electronic scheme
0.05
0 The demodulator electronic scheme can be modeled as
-0.05
3 3.2 3.4 3.6 3.8 4
depicted by fig. 17 [14]. The transfer functions of the three
ΔC=-2 pF
blocks from fig. 17 are
0.1
Vi ' (s) R
0 =− 4 ,
Vi (s) R3
-0.1
3 3.2 3.4 3.6 3.8 4 Vi ' ' (s) R (20)
vi [V]
ΔC=-4 pF = − 6 = −1,
0.2
V 'i (s) R5
0 V f (s) R9 + R10
= .
-0.2 Vd (s) s 2 R7 R8 R9 C5 C 6 + s[C 6 R9 (R7 + R8 ) − C5 R7 R10 ] + R9
3 3.2 3.4 3.6 3.8 4
ΔC=-6 pF The following notations are used
0.2
0
-0.2
k a = − R4 / R3 ,
3 3.2 3.4 3.6 3.8 4 k A = 1 + R10 / R9 ,
t [s x 10-4] (21)
Fig. 15. vi (t ) dependences for ΔC = constant ( ΔC > 0 ) a2 = R7 R8 R9C5C6 ,
a1 = C6 ( R7 + R8 ) − C5 R7 R10 / R9 ,
Considering the above conclusions, one can deduce that the
The equations (20) become
acceleration detection can be performed through the
determination of vi (t ) magnitude. Still a very serious problem Vi ' (s)
= ka ,
occurs here. If the signal vi (t ) is demodulated through the Vi (s)
extraction of its magnitude, one cannot precisely know if the Vi ' ' (s)
determined acceleration is positive or negative. As one could = −1, (22)
Vi ' (s)
see, the acceleration sign is influenced by the phase of vi (t ) . As
V f (s) kA
a consequence, during the detection stage one must use a phase- = .
sensitive demodulator as the one depicted by fig. 16 [14]. Vd (s) a2s 2 + a1s + 1

Inverter The model MATLAB/SIMULINK of the demodulator is

vin
depicted by fig. 18. In order to simulate the model, the following
vid vd
values of parameters were used [14]: R3 = 1kΩ ,
Amplifier LPF
vi vf R4 = R5 = R6 = R9 = 10kΩ , R7 = R8 = 180 kΩ , R10 = 5,6
v1 1
kΩ , C5 = C6 = 1,5nF . Using them, the following variables
-1 were computed (21): k A = 1,56 , a1 = 0,3888 ⋅ 10 −3 s ,
sgn
a2 = 72,9 ⋅ 10 −9 s 2 , k a = −10 . The cutting frequency of the low-
Fig. 16. Phase-sensitive demodulator pass filter become, under these conditions,
f t = 401,97Hz ≅ 402Hz .
The control of both interrupters is performed by means of the
alternative voltage sign v1 . They are coupled for an alternation

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 111

vin
valori vin
vi [V]
vezi vin
1
vi -K- 0
valori vi
vezi vi Gain2 -1
Switch vezi vf
0 1 2 3 4 5
+ kA vin [V] t [s x 10-5]
-K- + vf
a2.s2 +a1.s+1 valori vf 10
intrare Gain1 Sum1
sinusoidala Transfer Fcn2
5
vd 0
0 1 2 3 4 5
Switch1 valori vd vezi vd vid [V] t [s x 10-5]
10
v1 Sign vid
valori vid
5
vezi vid
0
Fig. 18. MATLAB/SIMULINK model of the demodulator 0 1 2 3 4 5
t [s x 10-5]
vd [V]
Considering that the input is a sine function with the 10
magnitude Vi = 1V and the frequency f = 10 5 Hz (equal to 5
0
that of the voltage v1 (t ) ), the characteristics from fig. 19 and 0 1 2 3 4 5
t [s x 10-5]
fig. 20 are obtained. In both figures there are outlined the
voltage waveforms up to the input in the low-pass filter. In both Fig. 20. Voltage waveforms up to the input in
cases we considered the influence of acceleration sign over the the low-pass filter ( a < 0 )
signal phase of the input vi (t ) . Fig. 19 represents the voltages
due to a positive acceleration and fig. 20 depicts the voltages Considering the acceleration lower than 4 g , the proof mass
owing to a negative one. One can notice that the demodulator
operates as a rectifier: in the case a > 0 it rectifies along a displacements can be assumed as very low ( x << d 0 ) and
negative alternation and in the case a < 0 it rectifies along a equation (25) becomes
positive alternation. 2 2ε ε AV
The output of the phase-sensitive demodulator can be vf ≅ k a k A 0 r 2 1 x. (26)
calculated with π C4 d 0
T
1 Consequently the output signal v f is approximately linear
T ∫0
v f = k A vd (t ) = k A vd (t )dt =
with the displacement x of the proof mass and consequently
with the applied acceleration, but under the condition that the
1⎡ ⎤
T /2 T

= kA ⎢ ∫ (− k a )vi (t )dt + ∫ k a vi (t )dt ⎥, (23) acceleration should not exceed 3 ÷ 4 g .

T⎣0 ⎦
T /2
IV. NUMERICAL SIMULATION OF THE ENTIRE
where T represents the period of the signal vi (t ) that ACCELEROMETER AND CONCLUSIONS
generates vd (t ) . After the substitution vi (t ) = Vi sin ωt we get If we merge the block diagram of the mechanic model (fig.
3), the block diagram of the detection device (fig. 5) and the
k a k AVi ⎡ T /2 T
⎤ 4k k V block diagram of the demodulator, we get the block diagram of
vf =
T
⎢ −
⎣ 0
∫sin ωt dt +
T /2 ⎦
∫
sin ωtdt ⎥ = − a A i ,
ωT
(24)
the entire accelerometer, depicted by fig. 21.
2
d x dx
that is a + dt
2
1 dt 1 x
∑ -
- s s
2 2 2ε ε AV x +
vf = − k a k AVi ≅ k a k A 0 r 1 2 . (25)
π π C4 d0 − x2 × μA2 ⎡ 1
+
1 ⎤
⎢ 3⎥
2m ⎣ (d 0 − x ) (d 0 + x ) ⎦
3

vi [V]
1 k
0 m
-1
0 1 2 3 4 5
ε 0 AV1 ⎡ 1 ⎤
2
t [s x 10-5] 1
vin [V] ⎢ − 2 ⎥
4 m ⎣ (d 0 − x ) (d 0 + x ) ⎦
2
0
-5
-10 v1
0 1 2 3 4 5
vid [V] t [s x 10-5] − 2s 2C3 R2 ΔC ε 0ε r Ax
0 (1 + sC4 R2 )[s(2C0 + C3 ) + 1 / R1 ] × d0 − x 2
2

-5
-10
0 1 2 3 4 5 vin
vd [V] t [s x 10-5] -1
0 vi + vd kA vf
ka ∑
-5
vid + a2s 2 + a1s + 1
-10
0 1 2 3 4 5
t [s x 10-5] v1 1

Fig. 19. Voltage waveforms up to the input in -1

sgn
the low-pass filter ( a > 0 )
Fig. 21. Block diagram of the entire accelerometer

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 112

In the above the blocks that form the accelerometer were values of the step-like inputs used for simulation were chosen as
simulated one at a time. The operation of accelerometer as a follows: positive and negative values, lower and higher
whole must be also considered, especially because in the case of magnitudes and respectively a combination between step inputs
intermediate and final blocks, the time-variations of the input that should outline the accelerator ability to follow be means of
quantities were not considered. These quantities actually come its output the input quantity. In this way, for simple step-like
from the previous blocks. inputs with the values a = 1g , a = −1g , a = 4 g and
The MATLAB/SIMULINK model that was used for the a = −4 g , the characteristics from fig. 23, fig. 24, fig. 25 and
entire simulation of the accelerometer is that depicted by fig. 22.
respectively fig. 26 were obtained.
In order to emphasize all the possible aspects of the problem, the
vin
vi valori vin vezi vin
valori vi vezi vi
x v1 -K-
valori x vezi vf
vezi x Gain2
Switch
+ -2*R2*C3 kA
+ f(u) * -K- + vf
- 1/s 1/s C3.s+1/R1 +
acceleratia - Product1 Gain1 a2.s2 +a1.s+1 valori vf
Integrator Integrator1 Fcn2 Transfer Fcn1 Sum1
Sum Transfer Fcn2

* f(u) Dc
vd
Product valori Dc
Fcn Vezi Dc Switch1 valori vd
vezi vd

-K-
Gain v1 vid
Sign
demodulare valori vid vezi vid
f(u)
Fcn1

Fig. 22. MATLAB/SIMULINK model of the entire accelerometer

x [m x 10-6]
displacement x of the proof mass, the variation ΔC of the
1
variable capacitors capacities and the accelerometer output
0.5
voltage v f ). Other quantities were
0
0 0.005 0.01 0.015 0.02 0.025 0.03 x [m x 10-6]
ΔC [F x 10-12] t [s] 0
2
-0.5
1
-1
0 0.005 0.01 0.015 0.02 0.025 0.03
0 t [s]
0 0.005 0.01 0.015 0.02 0.025 0.03 ΔC [F x 10-12]
vi [V] t [s] 0
0.05
-1
0
-2
0 0.005 0.01 0.015 0.02 0.025 0.03
-0.05 vi [V] t [s]
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vin [V] t [s] 0.05

0 0
-0.2
-0.05
-0.4 0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vin [V] t [s]
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vid [V] t [s] 0.4

0 0.2

-0.2 0
-0.4 0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vid [V] t [s]
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vd [V] t [s] 0.4

0 0.2
0
-0.2
-0.4 0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vd [V] t [s]
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vf [V] t [s] 0.4
0 0.2
0
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
-0.5 vf [V] t [s]
0 0.005 0.01 0.015 0.02 0.025 0.03
0.5
t [s]

Fig. 23. The time-evolution of all quantities from the block-

scheme for a = 1g 0
0 0.005 0.01 0.015 0.02 0.025 0.03
In the figures 23, 24, 25 and fig. 26, one can notice the time- t [s]

evolution of all quantities from the block-scheme. Some of the Fig. 24. The time-evolution of all quantities from the block-
quantities were represented for the entire simulation period (the scheme for a = −1g

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 113

represented in the final stages of the simulation. It was the case x [m x 10-6]
0
when only the final stage of the signal and its modifications of
-2
phase and amplitude (the voltage vi from the detection circuit
-4
output, the output voltage vin across the inverting side of the 0
ΔC [F x 10-12]
0.005 0.01 0.015 0.02 0.025 0.03
t [s]
demodulator and respectively the demodulator output voltage 0

vd ) presented interest.
x [m x 10-6]
-5
4 0 0.005 0.01 0.015 0.02 0.025 0.03
vi [V] t [s]
2 0.5

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03
ΔC [F x 10-12] t [s]
-0.5
5 0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vin [V] t [s]

2
0 1
0 0.005 0.01 0.015 0.02 0.025 0.03
vi [V] t [s] 0
0.5 0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vid [V] t [s]
0
2
-0.5 1
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vin [V] t [s] 0
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
0 vd [V] t [s]
-1
2
-2
1
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vid [V] t [s] 0
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
0 vf [V] t [s]
-1 4
-2
2
0.02995 0.02996 0.02997 0.02998 0.02999 0.03
vd [V] t [s]
0
0 0.005 0.01 0.015 0.02 0.025 0.03
0 t [s]
-1
-2
Fig. 26. The time-evolution of all quantities from the block-
0.02995 0.02996 0.02997 0.02998 0.02999 0.03 scheme for a = −4 g
vf [V] t [s]
0 − when the input value is increased, a higher difference can be
-2 noticed between the magnitude of vi and that of vd .
-4 One can notice certain symmetry of quantities, appearing for
0 0.005 0.01 0.015 0.02 0.025 0.03
t [s] the same values of inputs with opposite signs. Therefore one can
Fig. 25. The time-evolution of all quantities from the block- conclude that there is a symmetrical static characteristic v f (a)
scheme for a = 4 g for the studied accelerometer. The first 3 notices assert the
tendency of obtaining a nonlinear static characteristic but, on the
After the analysis of the four figures one can notice the other side, the 7-th notice asserts that within the range
following: −4 g ÷ 4 g the nonlinear influences of the mechanical model
− the time constants become higher as the scheme from input and of the detection circuit cannot be neglected, the dependency
toward output; Vi = f (a ) being approximately nonlinear. One cannot
− the time constants corresponding to the same quantity raise
with the input increase; guarantee the same thing for the relation v f = f (a ) . The 4-th,
− the time constants rest constant for the same quantity, for the 5-th , 6-th and 8-th notices prove the existence of a nonlinear
case of inputs with the same magnitude but opposite signs; dependency v f = f (Vi ) within the range −4 g ÷ 4 g . The
− a phase-difference occurs between the voltages vi and v1 ; it deviations from linearity of this dependency are not very high,
is reflected by the voltages ν in and vid , which present but are visible as compared to those of the dependency
passing with an observable value, on the other side of the null Vi = f (a ) . As a consequence one expects to obtain a static
axis; nonlinear charac-teristic v f = f (a ) , even within the domain
− phase-difference between the voltages vi and v1 raises when
−4 g ÷ 4 g .
the output quantity is increased and it becomes constant for
The graphical representation of the steady values of the
inputs with the same magnitude and opposite signs;
voltage v f obtained for various values of the acceleration
− the useful sign vd is lost, before its entrance in the low-pass
filter, through periodic passes, with considerable values, on within the range −4 g ÷ 4 g is depicted by fig. 27, representing
the other side of the null axis; a static characteristic. The shape of the characteristic comes to
− there is a certain proportionality between the magnitude of vi certify its symmetry with respect to the positive and negative
accelerations and some visible non-linearity are revealed, even
and the applied acceleration;
if the measuring domain was restricted to −4 g ÷ 4 g . The
accelerometer can operate within this range, because the

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE

 114

nonlinearities are small, but if one intends to measure categories of signals.

accelerations higher than 4 g , it is strictly necessary to close its REFERENCES
loop, in order to avoid the overlapping of the two characteristics [1] J. Farrell, M. Barth, The Global Positioning System and
that present strong nonlinear features ( Vi = f (a ) and Inertial Navigation, Mc-Graw-Hill, New-York, 1999
v f = f (Vi ) ) for a > 4 g . [2] C.W. Tan, K. Mostov, P. Varaiya, Feasibility of a gyroscope-
free inertial navigation system for tracking rigid body motion.
vf [V]
California PATH Research Report, UCB-ITS-PRR-2000-9
2.5
[3] J.C. Radix, Systemes inertiels a composants lies <<Strap-
2 down>>. Cepadues Editions, Ecole Nationale Superieure de
1.5 l’Aeronautique et de l’Espace SUP’AERO, Toulouse, 1993
1 [4] O.S. Salychev, Inertial Systems in Navigation and Geophysics.
Bauman MSTU Press, Moscow, 1998
0.5
[5] A. Lawrence, Modern Inertial Technology: Navigation,
0 Guidance and Control. Springer Verlag, New York, 1993
-0.5 [6] C. Hong, B. Minhang, Z. Haijun, S. Shaoqun, A piezorezistive
-1 accelerometer with a novel vertical beam structure. Sensors
and Actuators A, Vol. A. 63, (No. 1), Elsevier, Sept. 1997, p.p.
-1.5
19-25
-2 [7] E.J. Kruglick, B.A. Warneke, K.S.J. Pister, CMOS 3-Axis
-2.5
-4g -3g -2g -1g 0 1g 2g 3g 4g
Accelerometers with Integrated Amplifier. Berkeley Sensor
a [m/s2]
and Actuator Center, University of California, Berkeley
[8] S. Marco, J. Samitier, O. Ruiz, A. Herms, J. Morante, Analysis
Fig. 27. Steady values of the voltage v f of Electrostatic damped piezorezistive silicon accelerometer.
In order to test the accelerometer answer, one determined the Sensors and Actuators, A 37-38, p.p. 317-322, 1993
characteristic from fig. 29 corresponding to the input from fig. [9] O. Tsutomu, Development of Piezoelectric Accelerometer of
28. One can see that the accelerometer triggers the signals with a Charge Shear Type having Temperature Independent
rapid time-variation, but with specific time constants. Sensitivity. Japanese Journal of Applied Phisics, Part. 1, No.
a [m/s2] 4A, April 1998
40 [10] E. Boyden, A High-Performance Tunneling Accelerometer.
4g Term project 6.777, Introduction to Microelectro-mechanical
30 Systems (MEMS), Spring 1999
[11] T.W. Kenny, H.K. Rockstad, C.H. Liu, Micromachined
20 Tunneling Accelerometer. http://www. Stan-
1g ford.edu~chliu/MTA.html
10 [12] C.H. Liu, Modeling and Control Analysis for Tunneling
Sensors. http://cdr.stanford.edu/DD/SM-SL/Tunneling
0 Sensors/Basic Research/ Mode
[13] B. E. Boser, R. T. Howe, Surface Micromachined
-10
Accelerometers, IEEE J. Solid-State Circuits, vol. SC-31, pp.
-20 366–375, March 1996
[14] M. Kraft, Closed-loop accelerometer employing oversampling
-3g
-30 conversion. PhD Thesis, Coventry University, 1997
0 0.05 0.1 0.15 0.2 [15] M. Kraft, C.P. Lewis, T.G. Hesketh, J. Tulmann, Signal Pick-
t [s]
off in micromachined accelerometers. 5th IEEE International
Fig. 28. Input signal for accelerometer testing Conference on Optimisation of Electric and Electronic
vf [V]
1.5
Equipment, Vol. III, p.p. 391-399, Brasov, Romania, 1996
-3g [16] M. Van Paemel, Interface circuit for capacitive accelerometer.
1 Sensors and Actuators, A17, p.p. 629-637, 1989
0.5 [17] T. Veijola, Accelerometer Model in APLAC. CT-18, Helsinki
University of Technology, February 28, 1994
0
1g [18] T. Veijola, H. Kuisma, J. Lahdenpera, J. Compact Large-
-0.5 Displace-ment Model for a Capacitive Accelerometer.
-1 Proceedings of MSM’99, San Juan, p.p. 218-221, April 19-21,
1999
-1.5

-2
4g
-2.5
0 0.05 0.1 0.15 0.2
t [s]
Fig. 29. Accelerometer answer corresponding to the input from
fig. 28
Another reason for which the loop must be closed is due to
the time constants variation at the increase of a produces the
modification of the accelerometer cutting frequency, which
sometimes leading to an undesired filtering or passing of some

MEMSTECH’2008, May 21-24, 2008, Polyana, UKRAINE