See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301709853

A mode localization based resonant MEMS tilt sensor with a linear

measurement range of 360°

Conference Paper · January 2016

DOI: 10.1109/MEMSYS.2016.7421786

CITATIONS READS

30 227

4 authors, including:

Hemin Zhang
KU Leuven
62 PUBLICATIONS 1,118 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hemin Zhang on 18 October 2017.

The user has requested enhancement of the downloaded file.

 A MODE LOCALIZATION BASED RESONANT MEMS TILT SENSOR
WITH A LINEAR MEASUREMENT RANGE OF 360°
Boyang Li, Hemin Zhang, Jiming Zhong, and Honglong Chang*
Ministry of Education Key Laboratory of Micro and Nano Systems for Aerospace,
Northwestern Polytechnical University, Xi’an, China.

ABSTRACT STRUCTURE AND PRINCIPLE

This paper reports a resonant MEMS tilt sensor based The principle of the tilt sensor
on mode localization phenomenon for the first time. The
tilt sensor consists of two mode localization based
accelerometers which are perpendicular to each other. The
input tilt angle will cause dramatic changes of the mode
shape of both accelerometers. Furthermore, the linear
measurement range is extended to a full-range of [-90°, 90°]
by alternatively measuring the shifts of the two amplitude
ratios with a phase difference of 90°. Current test results
show that the relative sensitivity based on the amplitude
ratio is averagely 169 times higher than that of the Figure 1. (a) Schematic of the tilt sensor based on two
frequency with a non-linearity of less than 4.5% in the MLBAs. (b) Ideal output curves of the two MLBAs.
range of [-90°, 90°]. The outputs of the input tilt angle ( )
from [90°, 270°] and [-90°, 90°] are axial symmetry The schematic of the tilt sensor is shown in Figure 1(a).
about = 90° , so it is easy to extend the linear The two identical MLBAs (Ax and Ay) are mounted
measurement range to 360°. vertically to each other. When there is an input tilt angle (θ)
along the z-axis, the components of the gravity, =
INTRODUCTION and = , will be applied on the proof masses of Ax
MEMS tilt sensors are widely used in avionics, and Ay, respectively. Both = and =
automotive electronics and engineering machines. In the can be obtained by measuring the amplitude ratio (AR)
past decades, resonant tilt sensors based on measuring the shifts of the MLBAs. In the range of [-90°,-45°] and [45°,
frequency shift of the resonators emerged due to its 90°] the AR of Ay is used, while in the range of [-45°, 45°]
advantages such as quasi-digital output, high sensitivity, the AR of Ax is used as the output of the tilt sensor (Figure
and high dynamic range [1]. The state of the art of the 1(b)). Then the input tilt angle can be obtained by using
resolution for the resonant tilt sensor is 2.86 × 10 inverse trigonometric function calculations.
degrees [2]. However, its linear measurement range is only Sensitivity analysis of the tilt sensor
[-20°, 20°] which is restricted by the inherent output The SEM image of the tilt sensor and the detailed view
characteristic of the tilt sensor based on one uniaxial of Ay is shown in Figure 2.
accelerometer. The output curve of the frequency versus
the tilt angle is a sine shaped curve which has a short linear
segment around 0°.
In order to further improve both the sensitivity and the
linear measurement range of the resonant tilt sensors,
inspired by our previous work on mode localization based
accelerometers (MLBAs) [3], we proposed a new tilt angle
sensing method. The mode localization phenomenon can
be briefly described as that small introduced symmetry-
breaking perturbations to a periodical oscillating system
consisted of two nearly identical weakly coupled resonators
will inhibit the propagation of vibration and induce the
mode shapes to undergo dramatic changes to become
strongly localized [4]. Then the change of the amplitude Figure 2. SEM image of the tilt sensor on the left and the
ratio is much larger than that of the frequency. Therefore detailed view of Ay on the right.
the amplitude ratio instead of the frequency will be chosen
Taking the example of Ay, it consists of two proof
as the output metric in the proposed tilt sensor to enhance
masses and two identical weakly coupled resonators
the sensitivity. Furthermore, two MLBAs were mounted
(WCRs). The WCRs are driven by comb-finger capacitors
orthogonally to increase the linear measurement range. The
and sensed by the parallel-plate capacitors between the
outputs are selected alternatively per 90° as the output of
proof masses and the resonators. And there are potential
the tilt sensor. Thus the linear measurement range can be
differences between the proof masses and the WCRs,
extended to a full-range of [-90°, 90°].
which result in electrostatic negative stiffness to the WCRs.
When the component of the gravity ( ) acts on Ay, the two It is easy to know the smaller the coupling spring constant
gaps ( ) between the proof masses and resonators will kc is, the higher the sensitivity is.
change reversely, leading to differential stiffness changes Ideally, the two WCRs are mechanically identical.
of the WCRs. Then the mode localization occurs, and the When perturbation is 0, the amplitude ratio is 1 or -1, and
AR of the WCRs changes as a linear function of the input with the increase of the perturbation, the curve splits with
acceleration. a high slope. Actually, the inevitable fabrication tolerances
The occurrence of the mode localization phenomena often limit perfect matching of the resonators, leading to
mainly depends on two factors, i.e. the internal coupling asymmetries in the initial eigenstates which could make the
between the resonators of such nearly symmetrical 0 input working point not equal to 1 or -1.
structures and the induced perturbation (mass or stiffness The perturbation stiffness ∆ caused by the
changes of the WCRs) within the system. The block component of gravitational acceleration ( ) can be
diagram of the WCRs is shown in Figure 3. Assuming that described as equation (6):
the mass and stiffness of the two resonators are ideally
3
symmetric: = = , = = . The coupling  ms a y + ε AV 2 / 2G0 2 
spring constant is , the perturbation on resonator 1 is ∆ Δk y = −ε AV /  G0 −
2
 (6)
and ∆ on resonator 2.  k s − ε AV 2 / G03 


x1 x2 We can get the ∆ caused by t in the same way:

k1 kc k2  m a + ε AV 2 / 2G0 2 
Δk x = −ε AV /  G0 − s x
2
3

m1 m2 

k s − ε AV 2 / G03 
(7)

where is the permittivity, is the area of the sense

Figure 3. Block diagram of the WCRs. capacitor, is the weight of the proof mass, is the
The dynamic equations of the system without damping potential difference between the proof mass and resonator,
consideration can be written as: is the stiffness suspension of the proof mass, =
and = are the components of
m x1   k + k c + Δk
0    − kc   x1   0  (1) gravitational acceleration. Therefore, the AR as a function
 + =
0

 x2  
m    − kc k + k c − Δk   x2   0  of the tilt angle θ could be derived by the equations (3), (4),
(5), (6) or (7).
where and represent the vibration displacements of Then we utilize equation (8) to calculate the angle
the resonator1 and resonator2, respectively. The resonant when θ ∈ [-90°,-45°] ∪ [45°, 90°], however, equation (9)
frequency ( ) and AR ( ) could be written respectively as will be utilized when θ ∈ [-45°, 45°].
equations (2) and (3):
θ = sin −1 (ax / g ) (8)
2 2
k + kc  Δk + k
ωi = c
(i=1, 2) (2)
m θ = cos −1 (a y / g ) (9)

Δk ± Δk 2 + kc2 DEVICE FABRACATION AND

ui = (i=1, 2) (3)
kc MEASUREMENT SETUP
The tilt sensor is fabricated in a standard SOI
where i represents the ith resonating mode. The coupled two technology [5] and packaged in vacuum environment
degree-of-freedom oscillating system has two vibration (~20mTorr) [6]. The fabrication process of the SOI
modes (out-of-phase mode is the 1st mode and in-phase technology as shown in Figure 4 contains the following
mode is the 2nd). The variation rate (relative sensitivity) steps: (a) photoresist spin coating and patterning; (b) DRIE
based on resonant frequency and AR can be written etching; (c) removing the photoresist; (d) dicing and (e)
respectively as: liquid HF release structures.

ωi − ωi0 kc  Δk 2 + kc2
≅ (4)
ωi0 2k

ui − ui0 Δk  Δk 
≅  1 ±  (5)
ui0 kc  2 kc 
Figure 4. The fabrication process of the SOI technology.
Comparing equation (4) with equation (5), the
sensitivity based on the amplitude ratio is ⁄2 times The tilt sensor was fixed on a dividing head and
higher than the sensitivity based on the resonant frequency. measured by an open loop circuit as shown in Figure 5. The
coupled resonators are driven by comb-finger capacitors
and sensed by gap-variation parallel capacitors. A 50mV It is obvious that the vibration amplitude and frequency
AC sweep signal is added to the AC drive ports to excite of the WCRs changed dramatically at the 1st mode (Out of
the resonators. A dynamic signal analyzer (Agilent 35670A) phase) as a function of tilt angle, while the changes of the
is used to detect the output current amplified by a trans- two WCRs at the 2nd mode were much smaller. Different
impedance amplifier with a gain of 10 MΩ. The from the traditional resonant sensors, the AR is chosen as
feedthrough signal is cancelled by using the method the output metric instead of frequency. Therefore the non-
reported in [7]. The proof masses and the substrates were dimensionalized AR of Ax is defined as the ratio of the
maintained at 0 V, while the DC voltage applied on the amplitude of resonator 2 divided by the amplitude of
resonators is 15V. resonator 1 in the 1st mode. The AR based sensitivity of
the tilt sensor including Ax and Ay are shown and
compared in Figure 8(a).

Figure 8. Relative sensitivity comparison in the full

measurement range of [-90°, 90°]. (a) AR based sensitivity.
(b) Frequency based sensitivity.
Figure 5. The open loop measurement setup. As shown in Figure 8(a), Ax（black square points）
shows better linearity and sensitivity in the range of [-45°,
TEST RESULTS AND ANALYSIS 45°], while Ay (red circle points） behaves better in[-90°,-
Regarding Ax, the amplitude-frequency responses of 45°] and [45°, 90°]. And the slope near 90° and -90° of the
the resonator 1 and resonator 2 are shown in Figure 6 in AR curve of Ax and that near 0° of the AR curve of Ay is
zero tilt angle input condition. almost zero. That means the sensitivities of the two
MLBAs near those points were zero, respectively. Thus,
the selected three blue linear segments own not only better
linearity but also better sensitivity than the rest segments of
the curves.
The frequency based sensitivity is about 1.24Hz/degree
averagely, which was calculated from figure 8(b). It is
shown from Figure 8 (a) that the sensitivity based on AR is
averagely 0.00089/degree. Because the AR is non-
dimensionalized, so using relative sensitivity to compare
the AR based sensitivity with the corresponding frequency
based sensitivity is a favorable method [8]. The comparison
results are listed in Table I.
Figure 6. Amplitude-frequency responses of resonator1
and resonator2 from Ax in zero tilt angle input condition Table I. Test Results
(Q-factor was approximately 2380).
Sensitivity
Non- Sensitivity Sensitivity
The out-of-phase mode (27551Hz) is at a lower Linear
Linearity of AR
of
Enhance-
frequency than the in-phase mode (27733Hz). When the tilt range Frequency
(%) （ppm/°） ment
（ppm/°）
angle changed from -90° to 90° in anticlockwise, the
amplitude-frequency responses (in linear scale) of the two -90°,-45° 0.86% 7022 39.9 176
WCRs are shown in Figure 7. -45°,45° 2.82% 9377 49.9 188
45°,90° 2.36% 4789 40.2 119
Average 7641 45 169

The test results indicate that the relative sensitivity of

the amplitude ratio is averagely 7641 ppm/degree while
that of the resonant frequency is 45 ppm/degree. In the full
measurement range of [-90°, 90°] the AR based sensitivity
is averagely 169 times higher than that based on frequency.
In order to demonstrate the linearity of the tilt sensor,
Figure 7. Amplitude-frequency responses of Ax when tilt we made a compensation on the output AR curves (Figure
angle changes from -90° to 90°. (a) Resonator 1. (b) 8 (a)) manually. We offset the output value at 0° to 0 and
Resonator 2.
 the non-linearity of the output curve is less than 4.5% by a SOI process with selective roughening under
(Figure 9). structures”, Micro & Nano Letters, vol.6, no.7, pp. 486-
489, July 2011.
[6] Q. Shen, H. Li, Y. Hao, W. Yuan, and H. Chang (2015,
October), “Bias Contribution Modeling for a
Symmetrical Micromachined Coriolis Vibratory
Gyroscope”, IEEE Sensors Journal, [Online].
[7] H. Zhang, W. Yuan, Y. Hao, H. Chang, “Influences of
the feedthrough capacitance on the frequency
synchronization of the weakly coupled resonators”,
IEEE Sensors Journal, vol.15, no.11, pp. 6081-6088,
Nov. 2015.
[8] M. Spletzer, A. Raman, H. Sumali and J. P. Sullivan,
“Highly sensitive mass detection and identification
using vibration localization in coupled microcantilever
arrays”, Appl. Phys. Lett. 92, 114102, 2008.

CONTACT
Figure 9. Compensated output amplitude ratio shift *Honglong Chang, [email protected]
response versus tilt angle in the full measurement range of
[-90°, 90°].

CONCLUSIONS
In this paper, the mode localization based resonant
MEMS tilt sensor with a linear measurement range of [-90°,
90°] is experimentally demonstrated. The test results show
that the sensitivity based on AR is averagely 169 times
higher than that based on resonant frequency. The non-
linearity of its output in the full measurement range is less
than 4.5%. More specifications of the tilt sensor will be
investigated based on a future closed loop measurement
and compensation circuit.

ACKNOWLEDGMENT
This work was supported by the Chinese National
Science Foundation under Grant no. 51575454. And also
grateful acknowledgement is made to Yongcun Hao and He
Li for their kind discussions on fabrication process.

REFERENCES
[1] S. X. P. Su, H. S. Yang, and A. M. Agogino, “A
resonant accelerometer with two-stage microleverage
mechanisms fabricated by SOI-MEMS technology”,
IEEE Sensors Journal, vol. 5, no. 6, pp. 1214–1223,
Dec. 2005.
[2] X. Zou, P. Thiruvenkatanathan and A.A. Seshia, “A
high-resolution micro-electro-mechanical resonant tilt
sensor”, Sensors and Actuators A: Physical, vol.220,
pp.168–177, Oct. 2014.
[3] H. Zhang, W. Yuan, B. Li, Y. Hao, M. Kraft and H.
Chang, “A novel resonant accelerometer based on
mode localization of weakly coupled resonators”, In the
18th International Conference on Transducers,
Anchorage, AK, USA, June 21-25, 2015, pp. 1073-
1076.
[4] C. Pierre, "Mode localization and eigenvalue loci
veering phenomena in disordered structures", Journal
of Sound and Vibration, vol.126, no.3, pp. 485-502,
May 1988.
[5] H. Chang, J. Xie, Q. Fu, Q. Shen, and W. Yuan,
“Micromachined inertial measurement unit fabricated

View publication stats