You are on page 1of 13

Measurement 93 (2016) 327–339

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Design and Analysis of MEMS Comb Drive Capacitive Accelerometer


for SHM and Seismic Applications
S. Kavitha a, R. Joseph Daniel a,⇑, K. Sumangala b
a
NPMaSS National MEMS Design Centre, Department of Electronics and Instrumentation Engineering, Annamalai University, Annamalainagar – 608 002, Chidambaram, Tamil
Nadu, India
b
Department of Civil and Structural Engineering, Annamalai University, Annamalainagar – 608 002, Chidambaram, Tamil Nadu, India

a r t i c l e i n f o a b s t r a c t

Article history: This work presents the design of a MEMS accelerometer that is specifically intended for Structural Health
Received 5 March 2015 Monitoring (SHM) applications where sensing low frequency low amplitude accelerations with high res-
Received in revised form 7 February 2016 olution is essential. The surface micromachined comb drive capacitance accelerometer structure has been
Accepted 5 July 2016
considered in this design. The simulation experiments conducted on these devices using IntelliSuite
Available online 7 July 2016
MEMS design tool show that it has excellent displacement sensitivity of 21.39 lm/g, a capacitive sensi-
tivity of 1.22 pF/g and voltage sensitivity of 1783 mV/g/V when it is designed to measure 0–0.1 g. Further,
Keywords: p
it is seen that it has a very low noise floor of 1.32 lg/ Hz and therefore high resolution. Since the accel-
Structural Health Monitoring
MEMS
erations can be as low as 0.04 g in SHM applications, excellent resolution is the primary goal in this
Accelerometer design. Further, one more sensor specifically meant for strong motion seismic application has also been
p
Proof mass reported. This device has a bandwidth of 0–250 Hz and a noise floor of 5.612 lg/ Hz in addition to a sen-
COTS sor level voltage sensitivity of 97.9 mV/g/V. Finally, the comparison of these results with other similar
Noise floor devices reported in the past clearly illustrates the comparable performance of the present devices.
Further, these devices, unlike the commercial low frequency accelerometers and other similar devices
reported in the past can be fabricated by surface micromachining and CMOS compatible processes.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction sensitivity. This is due to the fact that the accelerometer sensitivity
is inversely proportional to the resonant frequency squared. Hence,
Structural Health Monitoring (SHM) systems collect and ana- these MEMS accelerometers largely depend on the complex signal
lyze online information about a civil structure so that indications conditioning electronic circuits for achieving larger sensitivity [3].
of structural distress can be identified early. The fundamental But the natural frequencies of civil structures are relatively small
building block of such SHM systems are the accelerometers which (<100 Hz) [1]. Most ambient vibrations in civil structures are char-
pick up structural response information for diagnosing structural acterized by low-amplitude vibrations (small ‘g’). Further,
safety conditions [1–4]. Hence, the quality and completeness of accelerometers used for SHM should be of low cost and consume
the data set collected for a given structure largely depends upon low power. So the MEMS accelerometers for Civil SHM can be
the capabilities and quality of the transducers used to record struc- designed for smaller band width. This will in turn result in high
tural responses. In the recent past, researchers have focused their sensitivity and ultra-low noise floor. This would also ensure con-
research on deployment of MEMS accelerometers for SHM and siderably larger sensitivity at sensor level itself thus making the
earthquake applications [1–11]. SHM using MEMS accelerometers signal conditioning electronics less complex. This will not only
based on different working principles such as piezoresistive, capac- bring down the cost but also makes the realization of low power
itive and piezoelectric accelerometers [1,10–13] have also been accelerometers a reality. Energy efficient accelerometer design is
reported. However, these accelerometers are designed for general important for the fast growing wireless sensor networks
purpose applications and have a bandwidth of few kHz. Because (WSN) [14].
they are designed for high frequency, they have poor sensor The authors have made an attempt in this work to design a
poly-silicon surface micromachined comb drive folded beam type
MEMS capacitive accelerometer that satisfies the requirements of
⇑ Corresponding author.
an accelerometer exclusively meant for civil SHM applications.
E-mail addresses: kaviraj_2003@rediffmail.com (S. Kavitha), josuma.au@gmail.
com (R. Joseph Daniel), josuma@rediffmail.com (K. Sumangala).
This device structure is considered in this work since standard

http://dx.doi.org/10.1016/j.measurement.2016.07.029
0263-2241/Ó 2016 Elsevier Ltd. All rights reserved.
328 S. Kavitha et al. / Measurement 93 (2016) 327–339

MEMS fabrication process [15] already exists for its manufacture. capacitance pairs C1 and C2 are equal and can be referred to as C0
Further, such small ‘g’ devices can also find applications in automo- as given in Eq. (4)
tive field to vehicle stability enhancement, roll-over detection and
e0 er Ns Ls h
inclination/theft detection. Low ‘g’ accelerometers [16] have also C0 ¼ ð4Þ
d0
been reported for medical applications like detection of hand tre-
mor or human health monitoring [17–19]. The earth quake sensing where e0 is the permittivity constant, er is the dielectric constant of
applications (strong motion) too need accelerometers of smaller air, h is the height of the finger, d0 is the air gap between each mov-
bandwidth, low noise floor on the same lines of the accelerometers able finger and its left and right fixed fingers, Ns is the number of
for SHM applications but an acceleration range of 1–2 g [20,21]. sensing fingers and Ls is the length of the movable or sensing finger.
Hence the design of an accelerometer for 0–2 g ranges and low When the acceleration is non-zero, the capacitances C1 and C2 are
pffiffiffiffiffiffi
noise floor (< 1 lg= HzÞ has also been presented in this paper. changed due to displacement (x) of the moving finger. The proof
The present paper is organized as follows: mass displacement (x) is caused by the acceleration and this situa-
First the authors discuss the device structure and brief the ana- tion can be depicted by Fig. 2(b) and (c). The capacitances C1 and C2
lytical model of comb drive folded beam type MEMS accelerome- can be written as [28–30]
ter. Subsequently, the design optimization of two devices  
e0 er Ns Ls h x
intended for SHM and earth quake applications is discussed. Fol- C1 ¼ 1þ ð5Þ
d0 d0
lowing this, the details and results of various performance analyses
on the device meant for SHM applications are presented to verify if  
e0 er Ns Ls h x
the performance satisfies the specification. Next, this paper C2 ¼ 1 ð6Þ
describes the performance analysis conducted on the accelerome- d0 d0
ter designed for earthquake sensing applications. Subsequently,
various performance factors of these devices obtained through for xnd0
simulation studies are compared with the performance of com- The displacement (x) must be very small compared to the air
mercially available accelerometers and few other accelerometers gap between the fixed and movable fingers (d0) at rest for ensured
reported in the recent past. Finally, the conclusion of these studies linearity. For acceleration in the +x direction, the displacement (x)
is presented. is positive and therefore C1 increases and C2 decreases. But, the dis-
placement (x) is negative for acceleration in the x direction. In
2. Device structure and background theory this case, C1 decreases and C2 increases. The maximum displace-
ment of the proof mass is generally restricted to 20% of d0 for
ADXL series MEMS accelerometers [15] use a comb finger type achieving high linearity. When the acceleration is non zero, the
differential capacitive MEMS structure as shown in Fig. 1. These capacitance of C1 and C2 become unequal and the difference
capacitive accelerometers have excellent sensitivity, low drift, between capacitances C1 and C2 is given by
good noise performance and simplicity [22,23]. They also show
2e0 er Ns Ls h
intrinsically insensitive transduction mechanism to temperature. DC ¼ C 1  C 2 ¼ ½x=d0  ¼ 2C 0 ½x=d0  ð7Þ
d0
Hence the present authors adopt the same structure in the design
of MEMS accelerometers for SHM applications. When an
accelerometer is subjected to acceleration, an external force is 3. Analytical model of MEMS comb drive capacitive
transferred to the proof mass through the suspension beam. The accelerometer
proof mass, together with movable fingers extruding from both
sides, moves along and against the forced direction, while the fixed The present accelerometer is built in consideration of mechan-
fingers in the right and left side of each movable finger remain sta- ical vibration principles. The principal component of this
tionary [24–27]. This movement changes the capacitance between accelerometer is the proof mass supported by four folded suspen-
each fixed and movable fingers. This change in capacitance can be sion beams, which can be modeled as springs as shown in Fig. 3.
estimated and calibrated with applied external force. The equivalent stiffness constant (ke) of this spring mass system
When there is no acceleration (a = 0), the movable fingers are can be written as [25,28–30]
resting in the middle of the air gap between left and right fixed fin-
gers as shown in Fig. 2(a). In this condition, the left and right 4EW 3b h
ke ¼ ð8Þ
L3b
where E is the Young’s modulus of polysilicon (160 GPa), Wb is the
beam width, Lb is the beam length and h is the height of the
beam. The natural frequency f0 of this spring-mass system is given
by [28–30]
1 pffiffiffiffiffi
f0 ¼ ke =ms ð9Þ
2p
The sensing mass (ms) of the accelerometer includes the proof
mass and all movable fingers. This can be expressed as
ms ¼ qh½ðW m Lm Þ þ ðNs W s Ls Þ ð10Þ
3 3
where q is the density of polysilicon (2.33  10 Kg/m ), h is the
height of the fingers and proof mass, Wm is the mass width, Lm is
the mass length and NS, WS, LS are the number of sensing fingers,
sensing finger width and sensing finger length respectively. The
displacement sensitivity (Sd) of this device along the sensitive direc-
Fig. 1. MEMS comb drive folded beam type capacitive accelerometer. tion is expressed as [25]
S. Kavitha et al. / Measurement 93 (2016) 327–339 329

Fig. 2. Single unit of differential capacitance in the comb.

Fig. 3. Free body diagram of the proposed accelerometer.

ms g qh½ðW m Lm Þ þ ðNs W s Ls ÞL3b ments Lb, Wb and h. Among these parameters, h depends on the
Sd ¼ ¼ 3
m=g ð11Þ maximum thickness of polysilicon layer that can be deposited by
ke 2EhW b
LPCVD system. Wb depends on the limit enforced by lithography
facility. Hence they are fixed first and therefore not available to
4. Accelerometers design optimization the designer. Out of other parameters, Wm from mass controlling
elements and Lb from stiffness controlling elements are chosen as
Since SHM applications demand the measurement of low design parameters.
amplitude and low frequency ambient vibrations, the specifica- Totally two devices namely Device-A (for SHM application) and
tions of proposed accelerometer for SHM applications (Device-A) Device-B (for earth quake application) have been considered and
have been fixed as given in Table 1 Specifications of accelerometer designed to satisfy the specifications given in Table 1. Displace-
for earth quake applications (Device-B) are also listed in Table 1. ment sensitivity of Device-A is calculated to be 24.86 lm/g for
The key differences between the specifications of accelerometers the specified f0 of 100 Hz whereas it is calculated to be
for SHM and earth quake applications lie in the range of accelera- 1.014 lm/g for Device-Bin which f0 is 500 Hz. This can be achieved
tion and natural frequencies (f0). The f0 of the accelerometer meant for various possible mass and stiffness combinations. Hence Wm
for earthquake application has been fixed at 500 Hz as seen in and Lb to achieve f0 = 100 Hz (Device-A) are calculated for different
Table 1. It could have been better if the f0 of this device is also cho- ms and ke and are plotted as shown in Fig. 4. Then optimum design
sen to be 100 Hz. But a higher frequency is chosen so that the values are chosen in such a way that the device size remains as
device size is optimized. small as possible so that the benefit of miniaturization is fully
A careful observation of Eq. (11) for displacement sensitivity achieved. Mass width (Wm) and beam length (Lb) and the other
shows that the device sensitivity is directly proportional to mass physical dimensions of the Device-A designed to achieve the spec-
controlling elements Wm, Lm, Ns, Ws, Ls and stiffness controlling ele- ifications (Table 1) are given in Table 2. The value of d0 has been set
330 S. Kavitha et al. / Measurement 93 (2016) 327–339

Table 1 Table 2
Specifications of the accelerometers for SHM and Seismic applications. Geometrices of the Devices-A and B.

Parameter Device-A Device-B Parameters Dimensions Dimensions


in lm in lm
Vibration range 0–0.1 g 0–2 g
Device-A Device-B
Natural frequency 100 Hz 500 Hz
Displacement sensitivity 24 lm/g 1.01 lm/g Beam width (Wb) 4 4
pffiffiffiffiffiffi pffiffiffiffiffiffi
Noise floor 1 lg= Hz 1 lg= Hz Beam length (Lb) 1800 800
Cross-axis sensitivity <1% <1% Mass length (Lm) 6000 4900
Shock resistance <500 g <500 g Mass width (Wm) 1200 650
Movable or sensing finger length (Ls) 400 400
Fixed finger length (Lf) 400 400
Movable or sensing finger width (Ws) 4 4
Number of sensing or movable fingers (ns) or 290 238
number of capacitance units
Beam height (h) 4 4
Air gap between fixed and movable fingers (d0) 12 10
Air gap between proof mass and substrate (g0) 3 0.48

Device-B, air gap at rest condition (d0) is chosen to be 10 lm so


that maximum displacement at 2 g is less than 20% of d0. Had
the resonant frequency been chosen as 100 Hz, the required d0
would have been 250 lm which would result in unacceptable lar-
ger device size or less number of capacitance units.

4.1. Fabrication steps of proposed low-g accelerometer

Fabrication of accelerometers on Silicon substrate started with


deposition of Si3N4 insulation layer of 1 lm thickness. Following
this, BSG (Boron Silicate Glass) sacrificial layer is deposited for
Fig. 4. Design optimization of beam length and mass width of Device-A. 2 lm so that the movable microstructures such as folded-beams,
movable mass and movable fingers are to be released with BSG
as 12 lm so that the maximum displacement at 0.1 g is less than sacrificial layer technique. Then BSG is patterned to realize the
20% of the air gap between fixed and movable fingers (d0) at rest structure as shown in Fig. 5(d) so as to enable the beam and mass
condition. The criteria that is used to select optimum Lb and Wm structural layer to stand on the substrate. Subsequently the stand-
is 2Lb  W m þ 2Ls with the idea of minimizing the accelerometer ing structure polysilicon layer of 4 lm thicknesses is deposited by
size. In a similar way various geometries of Device-B for earth- LPCVD process and patterned to make the mass and beam
quake sensing were also obtained and reported in Table 2. In structure anchored on the Si3N4 layer as shown in Fig. 5(d). Finally,

Fig. 5. Fabrication process flow of proposed accelerometer.


S. Kavitha et al. / Measurement 93 (2016) 327–339 331

the sacrificial layer (BSG) is etched away to realize the surface


micromachined folded beam type accelerometer as shown in
Fig. 5(g). The detailed fabrication process flow is shown in Fig. 5
(a–g).

5. Performance analysis of Device-A for SHM application

Various analyses were conducted on the created structure of


Device-A using Intellisuite ÒCAD MEMS tool. Important perfor-
mance indicating factors were estimated by displacement sensitiv-
ity analysis, differential capacitance analysis, modal analysis,
voltage analysis, cross-axis sensitivity analysis, frequency response
analysis and noise analysis. The physical properties of this Device-
A are listed in Table 4 and the estimated results of these analyses
are summarized in Table 5. The details of these performance anal-
yses and the corresponding results are presented in this section.
Fig. 6. Displacement versus acceleration of Device-A.
5.1. Deflection sensitivity analysis

Estimated proof mass displacements at various acceleration for


Table 3 and it is seen that they match each other closely. It is
Device-A have been obtained through thermo mechanical analysis
important to note at this point that maximum displacement esti-
module in Intellisuite MEMS CAD tool. This FEM analysis uses 3D
mated at 0.1 g is 2.139 lm which is almost one fifth of the air
twenty node brick parabolic elements. Mesh size for FEA was
gap between movable and fixed fingers (d0). This is essential to
chosen to be 100. The temperature is fixed at 20 °C and fixed fin-
get linear change in capacitance with acceleration as explained in
gers and anchors of folded beams are fixed while defining the
Section 2.
boundary conditions. The displacement versus acceleration graph
as shown in Fig. 6 gives estimated displacements at various
accelerations obtained from IntelliSuite simulation experiments 5.2. Modal analysis
(Fig. 6). The displacement sensitivity of the Device-A is estimated
by finding the slope of the displacement versus acceleration curve Vibration mode analysis was also conducted by Intellisuite
shown in Fig. 6. The theoretical displacement sensitivity is also MEMS CAD tool. The first three mode natural frequencies of
calculated for Device-A using Eq. (11). These values are given in Device-A obtained through simulation can be seen in Fig. 7. The
first mode frequency estimated by simulation of Device-A is
99.46 Hz against the design value of 100 Hz. The second mode fre-
Table 3 quency is estimated to be 107 Hz. However, it does not affect the
Comparison between theoretical and simulation displacement response of Device-A.
displacement sensitivity if the frequency of estimated acceleration
Parameter Theoretical IntelliSuite is smaller compared with these second and third mode
Eq. (11) simulation frequencies.
Displacement sensitivity for Device-A (lm/g) 22.7 21.39

5.3. Capacitance analysis

Table 4 Capacitances C1 and C2 at various acceleration levels are com-


Physical properties of the Device-A. puted using the displacements obtained through simulation (Sec-
Physical parameter Device-A tion 5.1) and equations (5) and (6) respectively. The change in
Mass (ms) in lg 70.50
capacitance (DC) is obtained using Eq. (7) again by substituting
Stiffness (ke) in N/m 0.03 the displacement value (x) obtained from simulation. The rest
Natural frequency (f0) in Hz 104.46 capacitance for Device-A is 0.342 pF. These values obtained for
Displacement at lm/g 22.7 Device-A are plotted against acceleration as presented in Fig. 8.
Damping factor (n) 0.53
As it is expected, the capacitance C1 increases from rest capaci-
Damping force (b) in N/m/s 4.87  105
Quality factor (Q) 0.948 tance with acceleration in +x direction and C2 reduces from rest
Rest capacitance (C0) in pF 0.342 capacitance. However, it would be the other way if acceleration
is in the opposite direction. The DC changes linearly with acceler-
ation as seen in Fig. 8 and capacitive sensitivity for Device-A is
Table 5 1.22 pF/g as given in Table 5. This change in capacitance can be
Estimated performance parameters of the Device-A. converted into a voltage by a direction sensitive C–V converter
to get an electrical voltage signal that is proportional to applied
Parameter Device-A
acceleration.
Resonant frequency (f0) in Hz 99.46
Displacement sensitivity ðSd Þ in 21.39
lm/g 5.4. Cross-axis sensitivity analysis
Capacitance sensitivity in pF/g 1.22
Voltage sensitivity in mV/g/V 1783
It is essential that an ideal accelerometer exhibits large linear
Measurable frequency range in Hz 0–40
pffiffiffiffiffiffi displacement in the axis in which the acceleration is estimated
BNEA Noise floor in lg= Hz 1.31
pffiffiffiffiffiffi but nil displacement on other axes. Device-A is intended to mea-
CNEA Noise floor in lg= Hz 0.176
pffiffiffiffiffiffi sure vibration signals in its x and z axes. In order to evaluate its
TNEA Noise floor in lg= Hz 1.32
cross-axis performance, acceleration was applied in all the axes
332 S. Kavitha et al. / Measurement 93 (2016) 327–339

Fig. 7. Modal analysis simulation result for Device-A.

Fig. 8. Acceleration versus capacitances C1, C2 and DC of Device-A.

Fig. 9. Cross axis sensitivity response of Device-A.

and the corresponding displacements at various g are found. These


displacements are plotted against acceleration as shown in Fig. 9. It
is obvious from these characteristics that the y-deflection is three
order magnitudes lower than x and z axes displacements. From
this, it is known that this accelerometer has excellent cross-axis
performance. Further it is seen that a displacement proportional
to the applied acceleration is estimated at z-axis for acceleration
in the z-direction and these values are almost equal to the response
obtained in x-direction. A closer look at the structure of the
accelerometer (Device-A) show that the z-axis [26] movement is
not hindered if it is subjected to acceleration along this axis. Hence,
it can be said that this is a dual axes accelerometer and
Fig. 10. Differential capacitive bridge.
z-axis acceleration could also be estimated with this accelerome-
ter. The air gap between proof mass and substrate should be esti- 5.5. Voltage sensitivity analysis
mated to calculate the Z acceleration. When Z-acceleration
changes the capacitance (Cz) between proof mass and Capacitors C1 and C2 are used to form a differential half bridge
substrate will change due to displacement of proof mass in the Z as shown in Fig. 10. Since electric charges Q1 and Q2 across capac-
direction. itors C1 and C2 should be equal,
S. Kavitha et al. / Measurement 93 (2016) 327–339 333

C 1 ðV m  V s Þ ¼ C 2 ðV s þ V m Þ ð12Þ is very small compared with d0. The voltage sensitivity of Device-
A is estimated to be 1.783 V/g/V as given in Table 5.
If the modulation voltages +Vm and Vm are equal in magnitude, the
voltage (Vs) at the output of the half- bridge can be written as
5.6. Noise analysis and shock performance
[23,24]

C1  C2 The accelerometer noise floor is usually dominated by displace-


Vs ¼ Vm ð13Þ ment noise but in micromechanical systems it is often set by
C1 þ C2
thermo-mechanical noise or Brownian noise equivalent accelera-
where Vs is the output voltage and Vm is the input modulation volt- tion (BNEA). The noise arises from thermal motion of atoms inside
age. When positive acceleration is applied C1 = C0 + DC and C2 = C0 - the structure and in the surrounding air. This noise is a tempera-
 DC. If modulation voltages (+Vm) and (Vm) are equal in ture dependent parameter. The thermal displacement noise for a
magnitude [23,24], the Eq. (13) can be simply rewritten as damped accelerometer outside the resonance (f0) is given by
[19,28–30]
DC
Vs ¼ Vm ð14Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
2C 0
jxn ðf Þj ¼ 4kB T=x3o ms Q m= Hz ð16Þ
When negative acceleration is applied, C1 = C0  DC and
C2 = C0 + DC. Now, the Eq. (13) can be simply rewritten as where kB is the Boltzmann’s constant, ms is the sensing mass, T is
the temperature, x0 is the resonance frequency and Q is the quality
DC factor when the temperature and pressure around the sensor is
Vs ¼  Vm ð15Þ
2C 0 300 K and 1 atm respectively. Since damping coefficient of air is
dependent on temperature and pressure, the operating temperature
Capacitance C1 and C2 at various acceleration levels are used to eval-
and pressure are important consideration in the design of micro
uate DC and Vs subsequently using Eqs. (14) and (15). The output
accelerometers. The quality factor can be represented as
voltage (Vs) is linearly varying with the acceleration in the range
of +0.1 g to 0.1 g as shown in Fig. 11. This linear performance Q ¼ x0 ms =b ð17Þ
could be achieved since the displacement (x) at any acceleration
where x0 = 2pf0 and b is the damping force [11]. The noise floor is
obtained by the expression

jX nf j pffiffiffiffiffiffi
jg n j ¼ g= Hz ð18Þ
Sd

where X nf is the mechanical noise in accelerometer and Sd is the dis-


placement sensitivity.
The quality factor (Q) for Device-A is calculated to be 0.948. The
BNEA noise level estimated using Eq. (18) for Device-A is 1.32 lg/
(Hz)0.5. This corresponds to a resolution of 8.22 lg in the DC-40 Hz
frequency range.
The other noise contributing factor in comb drive capacitive
MEMS accelerometer is the electronic noise otherwise called cir-
cuit noise equivalent of acceleration (CNEA) [35,36]. This is pro-
duced in signal conditioning circuit or interface circuit used for
transducing capacitance to measurable electrical voltage output.
CNEA depends on both the capacitive resolution of the interface
Fig. 11. Output voltage versus acceleration response of Device-A. circuit (DCmin) and capacitance sensitivity (S) of the accelerometer
and it is given by

Fig. 12. Interface circuit for a capacitance change of 0.122 pF.


334 S. Kavitha et al. / Measurement 93 (2016) 327–339

DC min
CNEA ¼ ð19Þ
S
The interface or readout circuit considered in this study for dif-
ferential capacitance to voltage conversion is shown in Fig. 12. This
circuit is used to just explain the effect of readout circuit on the
noise performance of the proposed accelerometer. SPICE software
has been used to design and simulate the interface circuit. This cir-
cuit is used to convert the differential capacitance formed by the
accelerometer structure into a proportional voltage. The circuit is
driven with two square wave signals VA and VB that are in 180°
out of phase with each other. The voltage at the output terminal
of the differential capacitance bridge (VAB) and the output voltage
of the low noise op-amp (LF411) used in the readout circuit at 0.1 g
are shown in Fig. 13. The peak-to-peak output voltage from the
readout circuit is plotted against the acceleration and shown in
Fig. 14. Fig. 14. Peak to Peak output voltage to acceleration.

Subsequently, the noise spectral of this readout electronic cir-


cuit is obtained using SPICE simulation and is presented in
Fig. 15. The total circuit noise floor at 100 Hz is estimated to be tion is limited by the BNEA alone. This has been achieved by the
pffiffiffiffiffiffi
2:666lV= Hz and this corresponds to a capacitive resolution of large capacitance sensitivity achieved when the accelerometer is
pffiffiffiffiffiffi
0:214 aF= Hz. The circuit noise equivalent acceleration (CNEA) is designed for low frequencies.
pffiffiffiffiffiffi
also calculated from these values and it is 0:176 lg= Hz at Further, the V-Mises stresses obtained through simulations
100 Hz. This quantity is much smaller than the Brownian noise show that the Mises stress does not exceed the rupture stress of
pffiffiffiffiffiffi 7 GPa for accelerations more than 250 g [31]. This fall in the shock
equivalent acceleration (BNEA) of 1:31 lg= Hz at 100 Hz.
The Total Noise Equivalent Acceleration (TNEA) is calculated resistance limit is indeed due to larger proof mass needed to max-
using the Eq. [35] imize the sensitivity and achieve very small noise floor. Also, end
stops are provided to limit the larger deflections in cases of shocks
and these end stops resist any possible contact between the mov-
ing and fixed electrodes.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
TNEA ¼ BNEA2 þ CNEA2 ð20Þ
pffiffiffiffiffiffi 5.7. Frequency response analysis
The TNEA of Device-A is estimated to be 1:32 lg= Hz. There-
fore, the noise analysis clearly shows that the CNEA at 100 Hz is Accelerometer for SHM need not be designed to sense a broad
much smaller than the BNEA thus demonstrating that the resolu- vibration frequency range since the civil structures generally have

Fig. 13. Output of the differential capacitance bridge and readout circuit for 0.1 g [C1 = 0.403 pF, C2 = 0.281 pF and Vm = 1 V].
S. Kavitha et al. / Measurement 93 (2016) 327–339 335

Fig. 15. Total circuit noise spectral density of the readout circuit.

smaller natural frequency. However the sensor resonant frequency


of COTS (commercial off-the-shelf sensing technologies)
accelerometer is typically a few kHz (5.5 kHz for ADXL 103/203).
Sensitivity of an accelerometer is inversely proportional to 4p2f20.
So, COTS accelerometers [32] with larger resonant frequencies will
have lower sensitivity and the signals obtained from these
accelerometers are raised to readable value by high gain amplifiers
and CMOS readout circuits. But, the civil condition monitoring
needs very low band width [1,2,9–12]. Therefore, it is a favourable
situation because it is possible to enhance the sensitivity consider-
ably in addition to bringing down noise appreciably if accelerome-
ter is designed for low frequencies. This larger sensitivity is
obtained from the basic sensing section itself thus reducing the
burden of integrating high gain amplifier that follows the sensor.
Fig. 16. Frequency response of Device-A.
In order to validate this argument, the frequency response of
Device-A has been obtained. The frequency response of this simple
mass-spring system obeying the second order differential equation
can be written as [28–30] design has chosen the air gap between the capacitor electrodes
much larger than the proof mass displacement (x <<< d0) to ensure
F=k
d ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð21Þ linearity as explained in Section 2. However, this section is devoted
2 2 2 2 to study the extent to which this has been achieved. The percent-
½1  ðf =f 0 Þ þ ½2nðf =f 0 Þ
age of non-linearity of the present capacitive accelerometer at any
The damping ratio (n) for Device-A is calculated to be 0.53 [11]. intermediate acceleration is defined with reference to the straight
The frequency response characteristics of the proposed low line drawn between the end points of the actual capacitance curve
frequency-small ‘g’ MEMS accelerometer for SHM applications over the desired acceleration range as shown in Fig. 17a–c. It is
(Device-A) obtained using Eq. (21) is shown in Fig. 16. The applied expressed as
acceleration has been kept at 0.1 g and the sensitivity in dB has
been obtained using the relationship C reference  C actual
%NL ¼  100 ð23Þ
C reference  C Offset
Sd ¼ 20 logðSd =Sf ÞdB ð22Þ
where C reference is the reference capacitance at the mid range, C actual is
where Sd is the displacement sensitivity at any given frequency and the actual capacitance at the mid range and C offset is the capacitance
Sf is the sensitivity at a reference frequency that is kept as 0.01 Hz in at g = 0. The actual capacitance over the range 0–0.1 g is plotted for
this calculation. The response shows that the sensor ensures linear both capacitor C1 and capacitor C2. While drawing these curves, C1
behavior in 0–40 Hz thus satisfying the requirement of an and C2 are not estimated using the Eqs. (5) and (6) but using the fol-
accelerometer for civil SHM. lowing equations:

5.8. Non-linearity analysis e0 er Ns Ls h


C1 ¼ ð24Þ
d0 þ x
This work aims at designing a MEMS comb drive folded beam
type capacitive accelerometer with excellent linearity, even though e0 er Ns Ls h
C2 ¼ ð25Þ
the capacitive sensors are inherently non-linear (NL). The C actual d0  x
336 S. Kavitha et al. / Measurement 93 (2016) 327–339

Table 6
Physical properties of the Device-B.

Physical parameter Device-B


Mass (ms) in lg 32.80
Stiffness (ke) in N/m 0.32
Natural frequency (f0) in Hz 497.32
Displacement at lm/g 1.004
Damping factor (n) 0.65
Damping force (b) in N/m/s 1.33  104
Quality factor (Q) 0.767
Rest capacitance (C0) in pF 0.337

the case with the sensors for civil SHM. However, it is required to
sense strong motion condition caused by earthquake. Hence, they
should be capable of sensing acceleration up to 2 g [20,21]. But
the accelerometer design for the maximum acceleration of 2 g
and low frequencies (<100 Hz) would need large air gap
(250 lm) between the moving and fixed fingers to ensure linearity.
This will in turn make the device size larger or will restrict the
number of fingers thus resulting in poor sensor capacitance which
is undesirable. Hence, the device is designed for a frequency of
500 Hz as given in Table 1. The physical properties of this
Device-B are listed in Table 6.

6.1. Displacement sensitivity and Modal analyses

Simulation experiments have been conducted on Device-B to


assess its performance. The displacement and modal frequency
analyses of Device-B are shown in Fig. 18. The displacement sensi-
tivity of the Device-B is estimated by finding the slope of the dis-
placement versus acceleration curve (Fig. 19) obtained as
explained in Section 5.1. It is estimated to be 1.075 lm/g for Device
B. The modal analyses on the Device-B shows that its first mode
frequency is 503.80 Hz and it is closely matching the specified
value of 500 Hz. The second and third mode frequencies obtained
through modal analysis are 960.91 Hz and 1255.02 Hz respectively.

6.2. Capacitance and voltage sensitivity analyses

The rest capacitance (C0) of the Device-B is 0.337pF. This capac-


itance changes from rest capacitance to 0.404 pF for a positive
acceleration of +2 g and to 0.269 pF for a negative acceleration of
2 g. This amounts to 0.0675 pF/g capacitive sensitivity. When
capacitors C1 and C2 are connected to form a half bridge as shown
in Fig. 10, the output voltage swings from 0.2002 V to +0.2002 V
when the acceleration is changed from 2 g to +2 g and Vm = 1 V.
Fig. 17. (a) Non-linearity analysis of C1. (b) Non-linearity analysis of C2. (c) Non- This gives the voltage sensitivity to be 0.1001 V/g/V as given in
linearity analysis of DC. Table 7.

Similarly, actual DC is plotted by taking the difference between C1 6.3. Noise and frequency analyses
and C2 calculated using Eqs. (24) and (25). The results show that
non-linearity is still about 20% for C1 and C2. But the non-linearity As discussed in the Section 5.6, the noise spectrum obtained
is only 3% as for as DC is concerned. It is the advantage of using dif- using SPICE for the readout electronics and presented in Fig. 15
ferential capacitance structure where high level of linearity is pos- is used to calculate the output total circuit noise floor at 500 Hz.
sible. This can further be improved more stringent design Assuming that the same read out circuit is used for capacitance
consideration in the choice of air gap between the electrodes. to voltage conversion in Device-B, the circuit noise equivalent
pffiffiffiffiffiffi
acceleration (CNEA) is calculated to be 3:166 lg= Hz at 500 Hz.
This quantity is comparable to than the Brownian noise equivalent
pffiffiffiffiffiffi
6. Performance analysis of Device-B for seismic applications acceleration (BNEA) of 4:64 lg= Hz at 500 Hz obtained for Device-
B. This corresponds to a total noise equivalent acceleration (TNEA)
pffiffiffiffiffiffi
This section describes the various performance analyses on the of 5:612 lg= Hz for Device-B. Table 7 lists the various perfor-
Device-B that is designed for earthquake sensing applications. The mance factors of Device-B. Here the CNEA is not negligible as in
MEMS accelerometer for earthquake sensing has to be designed for the case of Device-A and the use of a low-noise low-power
p
low frequency and it should have low noise floor <1 lg/ Hz as it is reference-capacitor-less switched-capacitor circuit which gives a
S. Kavitha et al. / Measurement 93 (2016) 327–339 337

Fig. 18. Displacement and modal analyses of Device-B.

Fig. 20. Frequency response of Device-B.

7. Comparison with commercial MEMS accelerometers


Fig. 19. Displacement versus acceleration of Device-B.
In order to further bring out the satisfactory performance of
designed accelerometers (Devices-A and B), the performance
Table 7 parameters have been compared with the performances of some
Estimated performance parameters of the Device-B.
widely used commercial MEMS accelerometers and other MEMS
Parameter Device-B accelerometers reported by Kavitha et al. and Junseok Chae et al.
Resonant frequency (f0) in Hz 503.803 [11,33]. The Devices-A and B reported in this work give high volt-
Displacement sensitivity (Sd) in lm/g 1.075 age sensitivity at the sensor level itself as it is seen from Tables 5
Capacitance sensitivity in pF/g 0.0667 and 7. Further, it is important to note that the noise performance
Voltage sensitivity in mV/g/V 97.9
of Device-A is better than all other devices and very much satisfies
Measurable frequency range in Hz 0–250
pffiffiffiffiffiffi the requirement for SHM applications. This work also shows that it
BNEA Noise floor in lg= Hz 4.64
pffiffiffiffiffiffi is possible to achieve low noise floor even in in-plane surface
CNEA Noise floor in lg= Hz 3.166
pffiffiffiffiffiffi micromachined devices [33]. The noise performance of Device-B
TNEA Noise floor in lg= Hz 5.612
is comparable to the commercial sensors available for the band-
pffiffiffiffiffiffi width in the range of 0–500 Hz. The noise performance of
CNES as low as 0:2 lg= Hz can solve this problem [35]. The fre- Device-B can be further improved by designing the sensor for 0–
quency response of the Device-B is shown in Fig. 20. Device-B gives 100 Hz bandwidths. But that is possible only at the cost of capaci-
linear performance over 0–250 Hz. tance sensitivity imposed by the condition x <<< d0 for better lin-
338 S. Kavitha et al. / Measurement 93 (2016) 327–339

Table 8
Comparison of Devices-A and B performance with other MEMS accelerometers.

Device/Manufacturer Range (g) Specified sensitivity Resonance Noise floor Micro machining MEMS type
p
frequency (Hz) (lg/ Hz) process
Silicon Designs, INC. (1221) 0–2 2 (V/g) 400 5 BM Non-Silicon Nickel
Colibrys (MS9002) 0–2 1000 (mV/g) 100 18 BM Silicon
Endevco (7290A) 1000 (mV/g) 1300 BM Silicon
Li and Tseng [31] 0–2 0.49 (V/g) 500 1.6 BM and SM combined Silicon
Kavitha et al. [11] 0–2 4 (mV/g/V) 100 4.53 BM Silicon
a
Device-A (SHM applications) 0–0.1 1915 (mV/g/V) 100 1.3 SM Polysilicon
a
Device-B (Strong motion applications) 0–2 106.8 (mV/g/V) 500 4.64 SM Polysilicon

SM – Surface micro-machined.
BM – Bulk micro-machined.
a
Present paper.

earity. Above all it is important to note that the Devices-A and B are The voltage sensitivity of these Devices-A and B have been esti-
the only surface micromachined silicon devices which can be fab- mated to be 1783 mV/g/V and 97.9 mV/g/V respectively. The
ricated by CMOS compatible processes since their structure resem- results thus clearly bring out the superior performance of the pro-
ble the ADXL capacitive sensors. Further, the sensing finger length posed MEMS comb drive accelerometers. The Device-A also shows
has been restricted to 400 lm as recommended by Fedder and excellent noise immunity and the noise floor has been found to be
p
Mukherjee [34] to avoid any curling of moving fingers in a CMOS 1.32 lg/ Hz. This corresponds to a resolution of 8.22 lg over the
compatible fabrication process. 0–40 Hz frequency range in which Device-A exhibits good
Also, the Devices-A and B can be made to outperform the COTS linearity. The noise floor of Device-B has been found to be
p
sensor with the inclusion of signal conditioning electronics. More- 5.612 lg/ Hz and this corresponds to a resolution of 73.36 lg over
over, it is imperative to note that this high performance obtained the 0–250 Hz frequency range. The excellent noise performance of
from Device-A leads to a less complex signal conditioning circuits Device-A makes it suitable for SHM applications and Device-B for
thus resulting in less power consumption and per chip cost. Finally, seismic applications. Finally, this work has brought out new device
it is obvious from the Table 8 that the noise floor can be exceed- designs that can be fabricated using silicon surface micromachin-
ingly small if the accelerometers are designed for smaller band- ing techniques unlike the commercially available low frequency
widths (Eqs. (21) and (22) irrespective of the maximum MEMS accelerometers. Also, these designs are CMOS process
measureable acceleration for which the sensor is designed. How- compatible.
ever, fabrication and characterization of these devices could only
give a clear picture about the performance since the micro struc-
tures when fabricated can show large variations in their perfor- Acknowledgements
mance compared with the performance predicted in simulation
studies. The authors are presently focusing their effort to fabricate The authors acknowledge the support received from National
and characterize these devices. Program on Micro and Smart Systems (NPMaSS) and financial sup-
port from University Grant Commission (UGC), New Delhi, India
through MRP (Major Research Project) scheme. Also the authors
8. Conclusions record their sincere gratitude to the reviewer for his useful sugges-
tions for improvement.
Structural health monitoring of civil structures requires the
measurement of vibration signals that are low in magnitude and
frequency. The commercially available MEMS accelerometers have References
been designed for general purpose use and therefore generally
[1] A. Emin Aktan, F. Necati Catbas, Development of a Model Health Monitoring
have larger bandwidth. So the natural frequencies (f0) of these sen-
Guide for Major Bridges, Federal Highway Administration Research and
sors are high. But the displacement sensitivity is inversely propor- Development, Drexel Intelligent infrastructure and Transportation Safety
tional to the square of natural frequency in a MEMS accelerometer. Institute, July 2003.
Therefore, the displacement sensitivity of such sensors is relatively [2] J.P. Lynch, Aaron Partridge, K.H. Law, T.W. Kenny, A.S. Kiremidjian, Ed Carryer,
Design of piezoresistive MEMS-based accelerometer for integration with
low and these sensors do not ensure accurate measurement of very wireless sensing unit for structural monitoring, J. Aerospace Eng. 16 (2003)
low magnitude acceleration signals from huge civil structures. Fur- 108–114.
ther it is important to know that these COTS sensors suffer due to [3] C.R. Farrar, Gyuhae Park, D.W. Allen, M.D. Todd, Sensor network paradigms for
structural health monitoring, J. Stru. Control Health Monit. 3 (2006) 210–225.
large Brownian noise thus reducing the resolution of the accelera- [4] J.P. Lynch, Yang Wang, K.J. Loh, Jun Yi, C.B. Yun, Wireless structural monitoring
tion measurement. The sensitivity can be considerably improved of the geumdang bridge using resolution enhancing signal conditioning, in:
with noise immunity by designing the accelerometer for low natu- Procd. 24th Int. Modal Analysis Conf. (IMAC XXIV), St. Louis, MO, 2006. Jan. 30
– Feb. 2.
ral frequency which is an essential requirement for SHM applica-
p [5] J.P. Lynch, K.J. Loh, A summary review of wireless sensors and sensor networks
tion (noise floor less than1 lg/ Hz is insisted). Hence the for structural health monitoring, J. Shock Vib. Digest. (2006) 91–128.
authors have designed and analyzed a surface micromachined [6] R.O. Curadelli, J.D. Riera, D. Ambrosini, G. Amani, Damage detection by means
of structural damping identification, J. Eng Stru. 30 (2008) 3497–3504.
comb type capacitive accelerometers and investigated their perfor- [7] Takeshi Miyashita, Masatsugu Nagai, Vibration-based structural health
mances for the acceleration ranges of 0.1 g and 2 g. The displace- monitoring for bridges using laser doppler vibrometers and MEMS-based
ment sensitivity (Sd) for Device-A is estimated to be 21.39 lm/g technologies, J. Steel Strut. 8 (2008) 325–331.
and for Device-B is 1.075 lm/g and these values closely match [8] S. Neylon Colibrys, S.A. Maladière, Neuchâtel, MEMS based seismic and
vibration sensors in building and structural health monitoring systems, in:
the theoretical displacement sensitivity. The modal analysis gives MEMSCON Workshop in Bucharest on, 2010. 7 Oct..
the natural frequencies for Devices-A and Bas 99.46 Hz and [9] Junhee Kim, K.J. Loh, J.P. Lynch, Implementation of wireless monitoring
503.803 Hz slightly less than the design specifications of 100 and systems for modal analysis of bridges along a Korean Test Road, in: Proc. of NSF
Engg Research and Innovation Conf., Honolulu, Hawaii, 2009.
500 Hz respectively. Also it is estimated that the capacitance sen- [10] S. Kavitha, R. Joseph Daniel, C. Antony Jeyasehar, K. Sumangala, Computer
sitivity for Device-A is 1.22 pF/g and for Device-B is 0.0667 pF/g. modelling design and analysis of piezoresistive MEMS accelerometer for
S. Kavitha et al. / Measurement 93 (2016) 327–339 339

concrete SHM applications, in: Procd 7th Stru Eng Convention (SEC-2010), [23] Qinglong Zheng, Yunfeng Zhang, Ying Lei, Jiakun Song, Yong Xu, Haircell-
Allied Publishers Pvt. Ltd, India, 2010, pp. 367–376. vol. 1. inspired capacitive accelerometer with both high sensitivity and broad
[11] S. Kavitha, R. Joseph Daniel, K. Sumangala, Design and analysis of bulk dynamic range, in: IEEE Sensors (2010) Conf., China, 2010.
micromachined piezoresistive MEMS accelerometer for concrete SHM [24] Chih-Ming Sun, Ming-Han Tsai, Yu-Chia Liu, Weileun Fang, Implementation of
applications, Int. Conf. Smart Mater. Str. Syst. (2012). Jan. 04-07 (Bangalore, a monilithic single proof-mass tri-axis accelerometer using CMOS – MEMS
India). technique, IEEE Trans. Electron Devices 57 (2010) 7.
[12] S. Kavitha, R. Joseph Daniel, K. Sumangala, A simple analytical design approach [25] W. Wai-chi, A. Azid, B.Y. Majlis, Formulation of stiffness constant and effective
based on computer aided analysis of bulk micromachined piezoresistive mass for a folded beam, J. Arch. Mech. 62 (2010) 405–418.
MEMS accelerometer for concrete SHM applications, J. Measur. 46 (2013) [26] O. Sidek, M. Afif, M.A. Miskam, Design and simulation of SOI-MEMS Z-axis
3372–3388. capacitive accelerometer, Inter. J. Eng. Tech. 10 (2010) 7–13.
[13] S. Kavitha, R. Joseph Daniel, K. Sumangala, Design and analysis of bulk [27] Yi Ma, Xu Limei, Hui Li, Optimization design of a Capacitive
micromachined piezoresistive MEMS accelerometer for concrete SHM microaccelerometer, International conf. mechatronics and automation Proc.
applications, J. Sens. Trans. 144 (2012) 62–75. IEEE 4–7 (2010).
[14] Andreas Vogl, Dag T. Wanga, Preben Storasa, Thor Bakkea, Maaike, M.V. Taklo, [28] M.H. Bao, Micro Mechanical Transducers, Pressure sensors, Accelerometers
Allan Thomson, Lennart Balgard, Design, process and characterisation of a and Gyroscopes, Handbook of Sensors and Actuators, vol. 8, Elsevier,
high-performance vibration sensor for wireless condition monitoring, Sens. Amsterdam, 2000.
Actuators A 153 (2009) 155–161. [29] D. Stephen, Senturia, Microsystem Design, Kluwer Academic Publishers, 2009.
[15] ADXL 150/250, Specification Data Sheets, Analog Devices Inc., 1998. [30] G.K. Ananthasuresh, K.J. Vinoy, K.N. Bhat, Micro and Smart Systems, Wiley
[16] Robert G. Walmsley, Lennie K. Kiyama, Don M. Milligan, Rod L. Alley, David L Publishers, New Delhi, India, 2010.
Erickson, Peter G. Hartwell, Micro-G Silicon Accelerometer Using Surface [31] G. Li, A.A. Tseng, Transient and impact dynamics of a Micro-accelerometer, J.
Electrodes, in: IEEE SENSORS, Conf., 2009. Mater. Process. Manuf. Sci. 9 (2000) 143–156.
[17] B.B. Graham, Using an Accelerometer Sensor to Measure Human Hand Motion [32] Cenk Acar, Andrei M. Shkel, Experimental evaluation and comparative analysis
PhD. Thesis, Massachusetts Institute of Tech., 2000. of commercial variable-capacitance MEMS accelerometers, J. Micromech.
[18] V. Biefeld, B. Clasbrummel, J. Binder, Implantable low-g accelerometer for the Microeng. 13 (2003) 634–645.
telemetric monitoring of micro-movements in fracture zones, Proc. 1st Annual [33] Junseok Chae, Haluk Kulahand Khalil Najafi, An In-planehigh-sensitivity, low-
Int. IEEE-EMBS Special Topic Conf. Microtech. Medicine and Biology, 2000, pp. noise micro-g silicon accelerometer with CMOS readout circuitry, J.
497–501. Microelectromech. Syst. 13 (4) (2004).
[19] J.C. Lotters, W. Olthuis, P.H. Veltink, P. Bergveld, Design, realization and [34] Gary K. Fedder, Tamal Mukherjee, Physical design for surface-micromachined
characterization of a symmetrical triaxial capacitive accelerometer for medical MEMS, in: In Proceedings of the 5th ACM/SIGDA Physical Design Workshop,
applications, Sens. Actuators, A 61 (1997) 303–308. Reston, VA USA, 1996, pp. 53–60. April 15–17.
[20] B. John Merchant, Technical Staff Sandia National Lab., MEMS Applications in [35] Babak Vakili Amini, Farrokh Ayazi, Micro-gravity capacitive silicon-on-
Seismology, Seismic Instrumentation Tech. Symp., Nov 11th, 2009. insulator accelerometers, J. Micromech. Microeng. 15 (2005) 2113–2120.
[21] Silicon Designs Inc., Advanced accelerometer Solutions Model 1221 Low Noise [36] Kalyan Kumar Mistry, K.B.M. Swamy, Siddhartha Sen, Design of an SOI-MEMS
Analog Accelerometer. high resolution capacitive type single axis accelerometer, Microsyst. Technol.
[22] E. Ismail Gonenil, Zeynep Celik-Butler, P. Donald, Butler, Surface 16 (2010) 2057–2066.
micromachined MEMS accelerometer on flexible polymide substrate, IEEE
Sens. J. 11 (2011) 2318–2326.

You might also like