See discussions, stats, and author profiles for this publication at: [Link]

net/publication/224351551

A Compact Angular Rate Sensor System Using a Fully Decoupled Silicon-on-

Glass MEMS Gyroscope

Article in Journal of Microelectromechanical Systems · January 2009

DOI: 10.1109/JMEMS.2008.2007274 · Source: IEEE Xplore

CITATIONS READS
82 2,766

3 authors, including:

Said Emre Alper Tayfun Akin

Mikrosistemler Co. Ltd. Middle East Technical University
41 PUBLICATIONS 1,381 CITATIONS 209 PUBLICATIONS 4,757 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Tayfun Akin on 27 August 2016.

The user has requested enhancement of the downloaded file.

 1418 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

A Compact Angular Rate Sensor System Using a

Fully Decoupled Silicon-on-Glass MEMS Gyroscope
Said Emre Alper, Yuksel Temiz, and Tayfun Akin, Member, IEEE

Abstract—This paper presents the development of a compact dominated by micromachined inertial sensors by year 2020 [5].
single-axis angular rate sensor system employing a 100-μm-thick One of the major limitations of the MEMS gyroscopes today
single-crystal silicon microelectromechanical systems gyroscope is the mechanical cross-coupling between the drive and sense
with an improved decoupling arrangement between the drive
and sense modes. The improved decoupling arrangement of the modes that arises due to poor manufacturing tolerances relative
gyroscope enhances the robustness of sensing frame against drive– to the minimum feature size in micromachining technologies
mode oscillations and therefore minimizes mechanical crosstalk [6]. The amount of this coupling can be as high as 1000 deg/s
between the drive and sense modes, yielding a small bias instabil- [7]. Although it is possible to suppress the offset caused by un-
ity. The gyroscope core element is fabricated by through-etching wanted mechanical coupling by more than two orders of mag-
a 100-μm-thick silicon substrate which is anodically bonded to
a recessed glass handling substrate. A patterned metal layer is nitude using phase-sensitive detection, it still cannot be nulled
included at the bottom of the silicon substrate, both as an etch-stop completely and is still a dominant factor limiting the operating
layer and a heat sink to prevent heating- and notching-based range, scale-factor linearity, and bias instability of MEMS
structural deformations encountered in deep dry etching in the gyroscopes [8]. As a numerical example, even a 1◦ phase
silicon-on-glass process. The fabricated-gyroscope core element error during the demodulation process causes 17-deg/s offset to
has capacitive actuation/sensing gaps of about 5 μm yielding
an aspect ratio close to 20, providing a large differential sense appear at the gyroscope output due to a 1000-deg/s quadrature
capacitance of 18.2 pF in a relatively small footprint of 4.6 mm × signal. A nondeterministic phase noise of 1 mdeg then causes an
4.2 mm. Excitation and sensing electronics of the gyroscope are offset instability over 100 deg/h. Therefore, it can be concluded
constructed using off-the-shelf integrated circuits and fit in a that any variation in the offset due to mechanical crosstalk
compact printed circuit board of size 54 mm × 24 mm. The directly exhibits a bias instability, and it is quite important to
complete angular rate sensor system is characterized in a vacuum
ambient at a pressure of 5 mtorr and demonstrates a turn-on bias keep these unwanted crosstalk signals as small as possible.
of less than 0.1 deg/s, bias instability of 14.3 deg/h, angle random There are various approaches reported for reducing mechan-
√
walk better than 0.115 deg/ h, and a scale-factor nonlinearity of ical coupling, including the use of electrostatic tuning [7]–[9],
±0.6% in full-scale range of ±50 deg/s. [2007-0158] the drive electrode with 1-DOF motion [10], the sense electrode
Index Terms—Angular rate sensor, gyroscopes, quadrature with 1-DOF motion [6], both the drive and sense electrodes
coupling, silicon-on-glass (SOG) micromachining. with 1 DOF, and only the proof mass with 2 DOF [11]–[14].
Ideally, the sense electrodes must remain completely stationary
during drive-mode oscillations for minimum bias instability,
I. I NTRODUCTION which is impractical since there should always be a physical
connection between the drive-mode oscillator and the rate
T HE PROGRESS in micromachined or microelectro-
mechanical systems (MEMS) vibratory gyroscopes en-
abled several newborn rate-grade applications starting in the
sensing frame in order to couple rotation-induced Coriolis
forces acting on the oscillating mass to the sensing frame. This
early 1990s, where low cost and small size are essential, such as physical connection typically couples a significant portion of
in the widespread automotive market [1], [2]. However, the per- drive-mode oscillations to the sense-mode frame, depending
formance of MEMS gyroscopes continuously increased within on the mechanical design and fabrication tolerances. Other
the past decade, approaching to the performance requirements than the limitations due to fabrication tolerances, which can be
of a number of military tactical-grade applications [3], [4]. It optimized during the process development, there is still a need
is expected that most of the tactical-grade applications will be for increasing the structural immunity of the sense-mode frame
against drive-mode oscillations in vibratory MEMS gyroscopes
by properly arranging the movable frames and the flexures
Manuscript received July 4, 2007; revised September 27, 2008. Current connecting those frames in order to minimize the amount of
version published December 4, 2008. This work was supported by the State mechanical coupling.
Planning Organization (DPT) of Turkey. Subject Editor F. Ayazi.
S. E. Alper and T. Akin are with the Department of Electrical and Electronics This paper reports a new fully decoupled MEMS gyro-
Engineering, Middle East Technical University, Ankara 06531, Turkey (e-mail: scope with increased robustness of the sense-mode frames
said@[Link]; tayfun-akin@[Link]). against drive-mode oscillations, and the construction of a high-
Y. Temiz was with the Department of Electrical and Electronics Engineering,
Middle East Technical University, Ankara 06531, Turkey. He is now with the performance yet compact angular rate sensor system using this
Institute of Electrical Engineering, Ecole Polytechnique Fédérale de Lausanne, new gyroscope as the core element. The MEMS gyroscope is
1015 Lausanne, Switzerland (e-mail: [Link]@[Link]). fabricated with a specially developed fabrication process, in
Color versions of one or more of the figures in this paper are available online
at [Link] which a thin silicon substrate with a metal etch-stop layer at the
Digital Object Identifier 10.1109/JMEMS.2008.2007274 bottom is anodically bonded to a glass substrate and patterned

1057-7157/$25.00 © 2008 IEEE

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1419

Fig. 2. Perspective view of the new fully decoupled gyroscope structure.

micromachined gyroscope structure are quasi-stationary dur-

ing drive-mode oscillations, i.e., they do not move along the
drive mode. However, drive-mode oscillations cause a spurious
motion of the sense electrode along the sense-mode direction
with a frequency that is twice that of drive-mode oscillations.
This spurious double-frequency motion arises due to small but
Fig. 1. Source for the frequency doubling at the sense mode. The sense-mode finite axial stresses induced in the flexible beams connecting
electrode is pulled down twice during the completion of a single cycle of drive- the sense electrodes to the proof mass. Fortunately, the double-
mode oscillations.
frequency motion causes a common-mode capacitance change
using deep reactive ion etching (DRIE). The process is opti- at differential sense electrodes, which is partially suppressed
mized so that the fabricated silicon structures do not suffer from by differential sensing. However, complete rejection of this
neither the notching nor heating effects experienced during double-frequency motion is only possible if the countermotion
DRIE, and the fabricated gyroscope has 100-μm-thick silicon of each sense frame is identical, which is not possible due
structural layer with approximately 5-μm capacitive gaps. The to mass imbalances during fabrication. Assuming that the dif-
gyroscope fabricated in this process is then used to construct a ferential sensing cannot reject even only 1% of this spurious
complete angular rate sensor system fitting into an area of only double-frequency motion, and assuming that this motion prop-
13 cm2 , which demonstrates exceptional performance, even agates through the sense-mode demodulator with a 1◦ phase
though discrete signal processing ICs are employed instead of a error (this phase error might be due to not only electron-
CMOS-ASIC. The measured performance of the angular rate ics but also the mechanical element itself even for a perfect
sensor system is 14.3 deg/h and 6.9 (deg/h)/Hz1/2 in terms second-order system), the resulting error at the output would
of bias instability and white noise, respectively, which are be 175 ppm of the initial double-frequency motion. In other
currently limited by the electronics and not by the gyroscope words, for a drive-mode vibration amplitude of 10 μm and
core element. Section II describes the electromechanical design a double-frequency motion of 0.2 μm, which is typical for
of the novel fully decoupled MEMS gyroscope, and Section III sense frames without a rigid connection between them [14],
illustrates the new fabrication process eliminating the notching the output offset signal due to the double-frequency motion
and heating effects associated with DRIE in silicon-on-glass would be 0.35 Å, corresponding to about 1 deg/s considering
(SOG) micromachining. Then, Section IV presents the char- the sensor specifications. Any phase noise during demodulation
acterization results of the MEMS gyroscope and the complete may cause this offset to vary in time and add to bias instability
angular rate sensor system, and finally, Section V summarizes of the gyroscope. Phase errors arise due to a number of factors,
the conclusions of this paper. including stray capacitances across preamplifiers, frequency
separation between the drive and sense modes, phase shift
in the drive-mode automatic gain control electronics, etc. For
II. E LECTROMECHANICAL D ESIGN
a transresistance preamplifier circuit, for example, a 1-MΩ
Electromechanical design is essential in minimizing the un- feedback resistance in parallel with a 100-fF stray capacitance
desired mechanical crosstalk signals experienced in MEMS results in 0.15◦ phase shift between the preamplifier input and
gyroscopes. One of the problems causing a nonzero mechanical output at 4 kHz. Phase errors can easily reach 1◦ with the
crosstalk in the symmetric and decoupled gyroscopes of 1-DOF contribution of the aforementioned factors and can increase
drive and 1-DOF sense frames is reported in [14], which states bias instability together with spurious mechanical crosstalk.
that there exists a time-varying signal at the sense electrode Therefore, minimizing the amount of any spurious motion, such
output with a frequency that is twice the drive-mode resonance as the double-frequency motion, has the potential to improve
frequency. Fig. 1 shows the source of this double-frequency the bias stability of MEMS gyroscopes, which is considered in
motion, in which the sense-mode electrodes of the decoupled the new structure given hereinafter.

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 1420 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

Fig. 3. Exaggerated visualization of the FEM simulation by CoventorWare,

where the maximum displacement of the sense-mode frame of the designed
MEMS gyroscope is less than 500 ppm of the drive-mode displacement.

Fig. 2 shows the perspective view of the new fully decoupled Fig. 4. First four mode shapes of the new gyroscope simulated by the
gyroscope structure. The new design preserves the major de- CoventorWare FEM tool, including the drive and sense modes as well as the
coupling idea of previous designs [12]–[14], i.e., the drive and closest higher order modes.
sense electrodes are restricted to 1-DOF motion for minimizing TABLE I
mechanical crosstalk, but now, the differential sense electrodes RESONANCE FREQUENCIES ASSOCIATED WITH THE
are mechanically connected to each other by rigid trusses FIRST FOUR MODES OF SOG GYROSCOPE DESIGN
constructing a single rectangular sense frame instead of two
isolated bar-shaped sense frames. This way, the stiffness of
the sense electrodes is highly increased against the spurious
deflections caused by the oscillations along the drive mode.
Constructing a rigid sense frame almost doubles the sense
mass, i.e., reduces the Coriolis response by a factor of two.
However, this reduction is acceptable considering the amount of
suppression of the double-frequency motion and the increased
immunity of the sense frames against rotary oscillations.
The amount of unwanted mechanical coupling of drive-mode
oscillations to the sense frame is also analyzed by FEM simula-
tions performed in the CoventorWare software. Fig. 3 shows Those modes can be excited by external physical inputs such
the exaggerated visualization of the FEM simulation, where as instant shocks and disturb the inherent operation. The high
the maximum displacement of the sense-mode frame of the thickness (100 μm) of the proposed gyroscope sufficiently pre-
designed MEMS gyroscope is less than 500 ppm of the drive- vents any out-of-plane oscillation modes, and there remain only
mode displacement. This result correspond to an improvement the in-plane rotary modes to be taken care of. For this reason,
by a factor of 40 compared to our previous work [13], [14], the flexures of the sense frame of the gyroscope are located in
which does not have rigid sense frames. This would cause such a way that the sense frame is also highly robust against ro-
a mechanical crosstalk signal much less than 1 deg/s, when tary oscillations that cause unwanted mechanical crosstalk be-
it is estimated for the 5.3-kHz matched-frequency operation, tween the operational modes and the higher order rotary modes.
a drive-mode oscillation amplitude of 10 μm, and a sense- Fig. 4 shows the first four mode shapes of the new gyroscope
mode quality factor of 1000 at vacuum. This small mechanical simulated by the CoventorWare FEM tool, including the drive
signal can be further suppressed by differential reading as well and sense modes as well as the closest higher order modes.
as by phase-sensitive demodulation of the sense-mode output, Table I provides the resonance frequencies associated with each
leading to much better gyroscope stability. mode. Increasing the rotational stiffness of the sense frame
The second problem that would increase mechanical sufficiently separates the resonance frequencies of the higher
crosstalk is the existence of unwanted oscillation modes with order modes by about 15 kHz from the 5.3-kHz operational
frequencies close to the operational frequency of the gyroscope. frequency of the gyroscope. In the predecessor [14] of this

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1421

Fig. 7. SEM photographs of a comb electrode array of a gyroscope fabricated

in the SOG process with the metal shield, showing the improvement achieved
compared to Fig. 5.

Fig. 5. (a) Mechanism of SOI notch and (b) resulting destruction at the bottom
of a comb electrode array, a flexure beam, and an etch hole.

Fig. 6. Method to prevent SOI notch using a metal shield laid on the glass Fig. 8. SEM photographs of (a) ion damage and (b) resulting sidewall
substrate under suspended silicon [17]. Here, the glass surface is not charged, destruction in a narrow beam structure fabricated with the SOG process with
and therefore, the directionality of ion bombardment does not change. the metal shield.

design, the frequency of the rotary mode was below the drive
and sense resonance frequencies, and that of the out-of-plane
mode was only 5 kHz above the operating frequency. With this
new design, the unwanted mechanical coupling between the
operational modes and the higher order modes is suppressed,
ensuring a more stable operation.

III. SOG F ABRICATION P ROCESS Fig. 9. New SOG process using a metal etch-stop layer to prevent SOI notch
and silicon overheating problems simultaneously. It also prevents the release of
SOG micromachining is based on fabricating microstructures microstructures during DRIE.
defined by through-etching of a thin silicon substrate anodically
bonded to a machined glass substrate [4], [15]. Through-etching very thick and high-aspect-ratio silicon microstructures on low-
of the silicon substrate is usually performed by DRIE, for loss insulating glass substrates, which yields highly sensitive
which aspect ratios higher than 30 are already reported [15]. capacitive inertial sensors with low mechanical–thermal noise
The SOG micromachining process is appropriate for fabricating and small parasitic capacitances.

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 1422 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

Fig. 10. New four-mask SOG fabrication process with the metal etch-stop
layer used to implement the proposed gyroscope. (a) Form anchors on glass
substrate. (b) Pattern Cr/Au shield at the bottom of the 100-μm-thick silicon
substrate. (c) Anodic bonding of silicon and glass substrates. (d) Pattern pad
metallization and the photoresist etch mask for DRIE. (e) Thoroughly etched
silicon substrate until reaching the metal etch-stop layer. (f) Remove the metal
etch-stop layer and then the photoresist etch mask.

Fig. 12. SEM photographs of the fabricated SOG gyroscope. (a) Half of the
symmetric structure. (b) Close-up view showing the varying-overlap-area-type
drive combs. (c) Close-up view showing the varying-gap-type sense combs.
The fabricated capacitive gaps are close to 5 μm, while the thickness of the
structural silicon layer is 100 μm, yielding an aspect ratio of 20.

sidewalls of silicon trench, destroying the anisotropic nature of

DRIE. This effect is called SOI notching and causes significant
Fig. 11. SEM photographs of microstructures fabricated with the new SOG destruction of the etched microstructures if not taken care of.
process with the metal etch-stop layer, (a) demonstrating no damage due to SOI Fig. 5 describes the mechanism of SOI notch and shows the
notch at the bottom surface of the etched silicon for a comb electrode array, a
flexure, and circular etch holes and (b) demonstrating no significant ion damage
resulting destruction at the bottom of a comb electrode array, a
at the top surface of the etched silicon around a narrow comb finger array. flexure beam, and an etch hole.
Fig. 6 describes a method to prevent SOI notch using a metal
A major problem for through-wafer etch with DRIE is known shield laid on the glass substrate under suspended silicon [17].
as the aspect-ratio-dependent etching. This effect refers to the The shield is extended inside the anchor regions, electrically
faster etch rate in a wide trench compared to the etch rate in shorting the metal shield to the etched silicon. As a result,
a narrow trench. The effect of the etch-rate variation becomes ions passing through the fast-etched trenches are collected by
essential for the through-etching of a silicon layer located on the shield, and the glass surface remains uncharged. Therefore,
top of an insulating base. The ions passing through a completely no significant damage is observed in the etched silicon during
etched cavity and reaching to the insulating base electrically excessive etch periods. Fig. 7 shows the SEM photographs
charge the insulator surface, which can increase the charged of a comb electrode array of a gyroscope fabricated in the
potential to several tens of volts [16]. The successive ions SOG process with the metal shield, showing the improvement
are then deflected from the charged insulator surface to the achieved compared to Fig. 5.

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1423

Fig. 14. Measured resonance characteristics of the gyroscope at 5-mtorr

vacuum ambient, from which the quality factors of the (a) drive and (b) sense
modes are extracted as 9611 and 7284, respectively.

The sidewall destruction problem in the SOG process a with

metal shield is eliminated by using an advanced SOG process,
for which the metal shield is laid directly at the bottom surface
of the silicon substrate instead of the glass surface as in [18].
Fig. 13. Resonance characteristics for (a) the drive and (b) sense modes of Now, the metal shield acts as a notch-preventing etch-stop layer
the fabricated gyroscope measured using Agilent 4395A network analyzer and as well as an efficient heat sink during DRIE. Fig. 9 shows
Karl Suss probe station.
the advantage of the new SOG process with the metal etch-
Although the SOG process with the metal shield satisfacto- stop layer. The use of metal shield also prevents the release of
rily works in removing the SOI notch effect, it is experienced the suspended structures during DRIE; otherwise, the silicon
to be insufficient to prevent another detrimental effect observed structures are suspended and free to move once the silicon is
in SOG processes, particularly when there are large suspended thoroughly etched. This is also dangerous since the electrostatic
silicon masses with long thermal paths before reaching to the forces between charged silicon frames will cause random
glass substrate. This is important since the silicon heats up movement of the suspended silicon before the completion of
during DRIE due to ion bombardment, and if this heat cannot DRIE, which could easily destroy the narrow structures. In
be quickly transferred to the glass substrate under silicon, then order to eliminate these problems, a new fabrication process
both the etch rate and mask selectivity significantly reduce in is developed.
the overheated silicon regions. The glass substrate can properly Fig. 10 shows the new four-mask SOG fabrication process
be cooled during DRIE, but the heat induced on the large with the metal etch-stop layer used to implement the proposed
suspended silicon structures can only be transferred to the gyroscope. First, the anchor regions are defined by wet etch
cooled glass substrate through narrow silicon flexures, which on a 500-μm-thick Pyrex glass substrate. Next, a Cr/Au etch-
are usually insufficient in heat transfer. Excessive heating of stop layer is patterned at the bottom side of a 100-μm-thick
silicon causes photoresist mask erosion, and the sidewalls of highly doped (111) silicon substrate. Then, the glass and silicon
narrow features are significantly thinned, if not completely substrates are anodically bonded to each other. Finally, the
etched. Fig. 8 shows the SEM photographs of (a) ion damage silicon substrate is thoroughly etched with DRIE until the metal
and (b) resulting sidewall destruction in a narrow beam struc- etch-stop layer, which prevents the SOI notch effect as well as
ture fabricated with the SOG process with the metal shield. the mask-erosion-based ion damage at the sidewalls of narrow

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 1424 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

TABLE II
COMPARISON OF THE DESIGNED AND MEASURED ELECTROMECHANICAL PARAMETERS OF THE FABRICATED SOG GYROSCOPE

scope has a differential sense electrode capacitance of 18.2 pF

within a footprint less than 20 mm2 , owing to the 100-μm
structural layer thickness etched with an aspect ratio close to 20,
yielding an average of 5.5-μm capacitive gaps.

IV. C HARACTERIZATION R ESULTS

The capacitances of a single set (half of differential) of drive
and sense electrodes of the fabricated gyroscope are measured
using Agilent 4294A precision impedance analyzer as 1.2 and
9.1 pF, respectively. These values are slightly smaller than the
design values of 1.4 and 12.1 pF, respectively, which also in-
clude the simulated electrostatic fringe-field correction factors.
This small mismatch is possibly due to the enlarged capacitive
gaps of the fabricated gyroscope, which cause a reduction in
both the parallel-plate capacitors and the fringe-field correction
factors. The fabricated capacitive gaps are 5.3 and 5.6 μm for
the drive and sense electrodes, respectively, which are slightly
Fig. 15. Measured sense-mode raw output of the SOG gyroscope prior to different from each other and from the layout value of 5 μm,
demodulation, in response to a sinusoidal angular rate input having amplitude basically due to the mask-dependent etch variations in DRIE.
and frequency of 50 mrad/s (∼2.86 deg/s) and 10 Hz, respectively. Fig. 13 shows the resonance characteristics for the drive
and sense modes of the fabricated gyroscope measured with
structures. Once the silicon etch is completed, the etch-stop Agilent 4395A network analyzer and Karl Suss probe station.
metal layer is selectively etched in a wet etchant, while the pad The sense-mode resonance frequency can be tuned down to
metallization survives during this etch since it is still protected 50 Hz below the drive-mode frequency of 4.065 kHz with a
by the etch mask of DRIE, as shown in Fig. 10(e). Following dc polarization voltage of 31 V applied to the proof mass.
the removal of the metal etch-stop layer, the etch mask of DRIE The measured drive- and sense-mode mechanical resonance
is also removed in wet stripping solution, and the structures are frequencies and the polarization voltage required for frequency
released by drying on a hot plate after alcohol rinse. The high matching are both close to the expected values of 3.64 kHz and
stiffness of the 100-μm-thick silicon layer eliminates the risk of 28.8 V, respectively, estimated for a DRIE undercut of 2 μm
stiction during this release process. around narrow flexures during 100-μm-deep etch.
Fig. 11 shows the SEM photographs of microstructures fabri- The quality factors of the SOG gyroscope for the drive
cated with the new SOG process with the metal etch-stop layer, and sense modes at atmospheric pressure are measured to
(a) demonstrating no damage due to SOI notch at the bottom be 241 and 2.3, respectively. The large etch holes on the
surface of the etched silicon for a comb electrode array, a flex- perforated masses and the high-aspect-ratio sense fingers re-
ure, and circular etch holes and (b) demonstrating no significant sult in small quality factors at atmosphere; however, the gy-
ion damage at the top surface of the etched silicon around roscope performance significantly improves at vacuum for
a narrow comb finger array. Fig. 12 shows the SEM photo- which the air damping becomes negligible. Fig. 14 shows
graphs of the fabricated SOG gyroscope. The fabricated gyro- the measured resonance characteristics of the gyroscope at

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1425

Fig. 16. Block diagram of the complete angular rate sensor system.

Fig. 18. Measured response of the angular rate sensor starting from zero rate,
first reaching up to 50 deg/s, next reaching down to −50 deg/s, and finally
reaching up to zero rate again, using 5-deg/s steps during the test.

instability problems in the drive mode due to the very high

(> 100-kN/m) axial stiffness of the drive-mode flexures along
the sense direction. Table II provides a comparison of the
designed and measured electromechanical parameters of the
fabricated SOG gyroscope.
Fig. 15 shows the sense-mode raw output of the SOG gy-
roscope prior to demodulation, in response to a sinusoidal
angular rate input having amplitude and frequency of 50 mrad/s
Fig. 17. Compact PCB of the complete angular rate sensor system, fitting into (∼2.86 deg/s) and 10 Hz, respectively. The central peak in this
an area less than 13 cm2 .
response corresponds to the sum of the quadrature and in-phase
5-mtorr vacuum ambient, from which the quality factors of coupling signals from the drive mode to the sense output. The
the drive and sense modes are extracted as 9611 and 7284, sidelobes correspond to the raw sense output signals, which
respectively. The high quality factor does not cause any comb- are separated from the central peak by the frequency of the

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 1426 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

Fig. 19. Measured response curve of the fabricated angular rate sensor system
specifying the scale factor, the zero-rate offset, and the nonlinearity. Fig. 20. Bias drift data collected from the rate output, while the sensor is
located in the vacuum chamber and kept stationary during the test.

applied angular rate input, i.e., 10 Hz. The separation between by identical transresistance preamplifier stages are converted
the drive- and sense-mode resonance frequencies is approxi- to a single-ended signal using an instrumentation amplifier,
mately 200 Hz for the near-match-mode operation, which is demodulated by using phase-sensitive demodulation with a
sufficiently high to prevent any imbalance between the am- carrier signal generated from drive-mode oscillations, and low-
plitudes of the sidelobes that are ±10 Hz separated from the pass filtered to provide a dc output that is proportional to the
drive-mode resonance frequency. The mechanical rate sensi- applied angular rate input signal. An optional gain stage pro-
tivity of the gyroscope is about 224 μVrms /(deg/s), while the vides the calibration of the scale factor, which is typically not
measured sum of quadrature and in-phase coupling signals is needed for the SOG gyroscope used in this prototype system.
approximately 3.6 mVrms , corresponding to a rate-equivalent Fig. 17 shows the compact PCB of the complete angular rate
coupling of only 16 deg/s. Quadrature coupling is believed to sensor system, fitting into an area less than 13 cm2 , which
dominate this value since the measured turn-on bias of the is comparable to similar high-performance MEMS gyroscopes
gyroscope, which has been directly affected by the in-phase with discrete IC electronics [19]. The lid of the surface-mount
coupling that cannot be filtered out by demodulation, is on the ceramic package is not hermetically sealed, allowing sensor
order of 0.1 deg/s. The measured quadrature signal is quite characterization in a vacuum chamber. In an application en-
small compared to the values in our previous work [14] and in vironment, the vacuum packaging of the gyroscope can be
some other studies reported in the literature [6]–[8]. It should achieved on ceramic package level or in wafer level [20].
be noted that the measured noise level in the raw sensor output The complete angular rate sensor system is tested in a vac-
has been observed to be significantly increased by the rate table uum chamber mounted on the Ideal Aerosmith 1280 VS single-
servomotor; therefore, it is not suitable to predict the noise- axis rate table. The chamber is vacuumed down to 5 mtorr
equivalent rate from Fig. 15. using a mechanical pump to characterize the zero-rate bias,
In order to construct a fully functional angular rate sensor scale factor, bias instability, and angle random walk of the
system, the fabricated MEMS gyroscope is first wirebonded sensor. For the characterization of the gyroscope at vacuum, it
to a surface-mount ceramic package and then integrated with is quite difficult to keep the drive- and sense-mode resonance
signal processing electronic circuitry on a compact printed frequencies matched against temperature variations without
circuit board. Fig. 16 shows the block diagram of the complete closed-loop control. Therefore, a dc polarization voltage of
angular rate sensor system. Most part of the system aims 30 V is applied to the gyroscope, operating it with near-
the generation and control of stable drive-mode oscillations match drive and sense frequencies. Fig. 18 shows the measured
using closed-loop control techniques. The preamplifier stage response of the angular rate sensor starting from zero rate, first
is a capacitance-to-voltage converter, and it is used together reaching up to 50 deg/s, next reaching down to −50 deg/s,
with the variable-gain amplifier (VGA) to satisfy the phase and finally reaching up to zero rate again, using 5-deg/s steps
requirements for starting self-oscillations along the drive mode during the test. Fig. 19 shows the measured input rate versus dc
of the gyroscope. The amplitude of this oscillation is controlled output voltage response of the fabricated angular rate sensor
by referencing the measured drive-mode oscillation amplitude system specifying the calibrated scale factor, the zero-rate
to a desired amplitude signal (Vset) and then by minimizing the offset, and the nonlinearity. The sensor has a scale factor of
error between these signals using a proportional–integral (PI) 22.2 mV/(deg/s) with a zero-rate offset of 2.2 mV, correspond-
controller. The output of the PI controller continuously adjusts ing to 0.1 deg/s. The offset is pretty small due to phase-sensitive
the gain of VGA in order to keep the drive-mode oscillation demodulation as well as the small mechanical coupling between
amplitude constant at the desired level. Once a stable drive- the drive and sense modes of the new MEMS gyroscope. The
mode oscillation is achieved, the system is ready for angular measured quadrature coupling of 16 deg/s and a null bias of
rate measurements. The open-loop rate sensing mechanism of 0.1 deg/s correspond to a demodulation phase error of 0.36◦ ,
the system is straightforward: Differential sense outputs picked which is expected to be reduced if a dedicated ASIC chip were

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1427

Fig. 21. Allan variance plot of the fabricated gyroscope, obtained by processing the measured drift data of Fig. 20.

used instead of PCB-level electronics. Another important per- averaging time of 15 s for reaching the bias instability limit.
formance figure of the characterized angular rate sensor system The bias stability of the sensor is believed to be limited by
is the linearity of the scale factor. The R2 -nonlinearity of the open-loop sense electronics, the measurement setup inside the
data of Fig. 19 is measured to be less than ±0.01%, whereas the vacuum chamber under continuous pumping, and the stability
composite nonlinearity of the scale factor (including hysteresis, of the proof mass voltage.
systematic nonlinearity, and other uncertainties in the measure- The plot of Fig. 21 can also be used to specify the angle
ment) is measured to be as low as ±0.6%, i.e., compatible with random walk (due to white noise) of the developed angular rate
several commercial rate sensors [19]–[21]. The measurement sensor system. Rate noise density due to white noise can be
range of the angular rate sensor system is currently limited to represented by fitting a line with slope −1/2 to the region of the
±50 deg/s, above which the scale-factor nonlinearity increases Allan variance plot, for which the measured data can be well
significantly due to open-loop rate sensing. represented by the fit line [22]. The rate noise density can then
The Allan variance analysis is used in order to determine be directly obtained by reading this fit line at τ = 1 s, which
the bias instability and angle random walk of the fabricated yields 6.9(deg/h)/Hz1/2 or, equivalently, an angle random walk
angular rate sensor system [22]. The output of the sensor is of 0.115 deg/h1/2 . The accuracy of this measurement is also
collected for a period of more than 1.5 h, which is observed examined against the fact that whether the correlated noise
to be sufficient for determining the bias instability and angle is affecting the correct measurement of angle random walk
random walk of the gyroscope. Fig. 20 shows the bias drift data or not. The correlated noise can be modeled by fitting two
collected from the rate output, while the sensor is located in the asymptotes to the Allan variance graph with slopes of ±1/2,
vacuum chamber and kept stationary during the test. The initial as described in [22]. The contributions of correlated noise and
bias trend shows an overshoot at the beginning of the test, but white noise in the total measured noise are then found to be 8
it drops to 0.025 (deg/s)/h after about 1 h. We believe that this and 12.7 deg/h, respectively, for a sampling time of τ = 0.2 s
characteristic is due to the thermal stabilization of the system (or a sampling frequency of 5 Hz), verifying that the selected
and can be easily modeled and compensated by external elec- sampling time is dominated by the white-noise power. The
tronics. Fig. 21 shows the Allan variance plot of the fabricated white-noise density in 1-Hz bandwidth can then be calculated
gyroscope, which is obtained by processing the measured drift as 5.7 deg/h or, equivalently, 5.7(deg/h)/Hz1/2 . This value is
data of Fig. 20. The possible reason for having a plot with two 18% less and even better than the value that is directly obtained
dips is due to the exponentially correlated noise, as explained in from the fit line that describes white noise. This analysis shows
[22]. The potential source of the exponentially correlated noise that the existence of correlated noise in the Allan variance
is reported to be the randomized dither vibrations in ring-laser plot does not prevent the determination of angle random walk
gyros [23] or similar phenomena observed in spinning-wheel with an acceptable accuracy. The measured angle random walk
gyros [22]. We believe that randomized environmental vibra- is believed to be sufficiently low, considering that the whole
tions caused by the vacuum pump located in close vicinity of electronic stages, including preamplifiers, are constructed by
the rate table also generate a similar exponentially correlated discrete ICs and passive components.
noise (as in Fig. 21). The bias stability (due to flicker noise) The developed angular rate sensor system demonstrates an
of the gyroscope is found by dividing the measured dip level exceptional performance by integrating a novel fully decou-
of the Allan variance plot of Fig. 21 by a factor of 0.664, pled MEMS gyroscope with high-quality electronics gener-
as described in [22], which yields 14.3 deg/h. The measured ating amplitude-controlled stable drive-mode oscillations and
amount of bias stability is not affected by the initial thermal extracting the rate signal by phase-sensitive detection. The
stabilization characteristics, since the time constant of thermal integrated system occupies a compact size, which improves
stabilization is about 1000 s, which is much higher than the reliability and repeatability.

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 1428 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 17, NO. 6, DECEMBER 2008

V. C ONCLUSION [8] M. Palaniapan, R. T. Howe, and J. Yasaitis, “Performance comparison

of integrated Z-axis frame microgyroscopes,” in Proc. 16th IEEE MEMS
This paper presents the design and fabrication of a 100-μm- Workshop, Kyoto, Japan, Jan. 2003, pp. 482–485.
thick single-crystal silicon MEMS gyroscope with an improved [9] F. Ayazi and K. Najafi, “A HARPSS polysilicon vibrating ring
gyroscope,” J. Microelectromech. Syst., vol. 10, no. 2, pp. 169–179,
decoupling arrangement between the drive and sense modes, Jun. 2001.
where the gyroscope is used in the development of a com- [10] W. Geiger, W. U. Butt, A. Gaißer, J. Frech, M. Braxmaier, T. Link,
pact single-axis angular rate sensor system. The improved A. Kohne, P. Nommensen, H. Sandmaier, and W. Lang, “Decoupled
microgyros and the design principle DAVED,” Sens. Actuators A, Phys.,
decoupling arrangement is achieved by a special arrangement vol. 95, no. 2/3, pp. 239–249, Jan. 2002.
of sense-mode frame and flexures, minimizing both double- [11] M. S. Kranz and G. K. Fedder, “Micromechanical vibratory rate gyro-
scopes fabricated in conventional CMOS,” in Proc. Symp. Gyro Technol.,
frequency mechanical crosstalk and quadrature signal, which 1997, pp. 3.0–3.8.
are the two important effects limiting the stability of MEMS [12] S. E. Alper and T. Akin, “Symmetrical and decoupled nickel microgyro-
gyroscopes. The gyroscope is fabricated by using DRIE and scope on insulating substrate,” Sens. Actuators A, Phys., vol. 115, no. 2/3,
pp. 336–350, Sep. 2004.
through-wafer etching of a 100-μm-thick silicon substrate [13] S. E. Alper and T. Akin, “A single-crystal silicon symmetrical and de-
which is anodically bonded to a glass substrate; the notching coupled MEMS gyroscope on an insulating substrate,” J. Microelectro-
and heating effects associated with DRIE of silicon are elimi- mech. Syst., vol. 14, no. 4, pp. 707–717, Aug. 2005.
[14] S. E. Alper, K. Azgin, and T. Akin, “High-performance SOI-MEMS
nated by using a sacrificial metal etch-stop layer patterned at the gyroscope with decoupled oscillation modes,” in Proc. 19th IEEE Int.
bottom of the thin silicon substrate. The fabricated gyroscope Conf. MEMS, Istanbul, Turkey, Jan. 22–26, 2005, pp. 70–73.
has capacitive gaps of about 5 μm yielding an aspect ratio close [15] J. Chae, H. Kulah, and K. Najafi, “A hybrid silicon-on-glass (SOG) lateral
micro-accelerometer with CMOS readout circuitry,” in Proc. 15th IEEE
to 20, which results in a total differential sense capacitance of MEMS Workshop, Las Vegas, NV, Jan. 2002, pp. 623–626.
18.2 pF in a footprint of 4.2 mm × 4.6 mm. The fabricated [16] T. Kinoshita, M. Hane, and J. P. McVittie, “Notching as an example
MEMS gyroscope is then integrated with drive-mode automatic of charging in uniform high density plasmas,” J. Vac. Sci. Tech-
nol. B, Microelectron. Process. Phenom., vol. 14, no. 1, pp. 560–565,
gain control and sense-mode signal processing electronics, all Jan./Feb. 1996.
fitting on a PCB of area less than 13 cm2 . The complete [17] T. Matsuura, M. Chabloz, J. Jiao, Y. Yoshida, and K. Tsutsumi, “A method
angular rate sensor system is then characterized at vacuum, to evade silicon backside damage in deep reactive ion etching for anodi-
cally bonded glass–silicon structures,” Sens. Actuators A, Phys., vol. 89,
and it is shown that the system demonstrates a scale factor no. 1/2, pp. 71–75, Mar. 2001.
of 22.2 mV/(deg/s), with a composite nonlinearity as small as [18] C.-H. Kim, J. Park, N. Park, and Y.-K. Kim, “MEMS fiber-optic variable
±0.6% within the ±50-deg/s measurement range. The zero-rate optical attenuator using collimating lensed fiber,” in Proc. IEEE/LEOS
Int. Conf. Opt. MEMS, Waikoloa, HI, Aug. 2003, pp. 145–146.
bias of the sensor is less than 0.1 deg/s after turn-on, while the [19] SiRRS01 Datasheet, Silicon Sens., Plymouth, U.K., 2004. [Online].
bias instability is measured to be only 14.3 deg/h. The rate- Available: [Link]
[20] SMG066 Product Information, Robert Bosch GmbH, Gerlingen,
equivalent white-noise density of the gyroscope is measured to Germany, 2001. [Online]. Available: [Link]
be better than 6.9(deg/h)/Hz1/2 , although all of the electronic [21] IDG-300 Product Information, Invensense, Sunnyvale, CA, 2006.
stages are constructed by discrete ICs and passive components. [Online]. Available: [Link]
[22] IEEE Standard Specification Format Guide and Test Procedure for
The performance of the developed sensor system can be ex- Single-Axis Interferometric Fiber Optic Gyros, IEEE Std 952-1997, 1998.
tended further by using closed-loop rate sensing electronics and [23] IEEE Standard Specification Format Guide and Test Procedure for
hermetically sealing the sensor in a package with stable vacuum Single-Axis Laser Gyros, IEEE Std 647-1995, 1996.
inside.

R EFERENCES
[1] K. Funk, H. Emmerich, A. Schilp, M. Offenberg, R. Neul, and F. Larmer,
“A surface micromachined silicon gyroscope using a thick polysilicon
layer,” in Proc. 12th IEEE Int. Conf. MEMS, Orlando, FL, Jan. 1999, Said Emre Alper was born in Ankara, Turkey,
pp. 57–60. in 1976. He received the B.S., [Link]., and Ph.D.
[2] S. Chang, M. Chia, P. Castillo-Borelley, W. Hidgon, Q. Jiang, J. Johnson, degrees in electrical and electronics engineering
L. Obedier, M. Putty, Q. Shi, D. Sparks, and S. Zarabadi, “An electro- (with high honors) from Middle East Technical Uni-
formed CMOS integrated angular rate sensor,” Sens. Actuators A, Phys., versity (METU), Ankara, in 1998, 2000, and 2005,
vol. 66, no. 1–3, pp. 138–143, Apr. 1998. respectively.
[3] M. Weinberg, J. Connely, A. Kourepenis, and D. Sargent, “Microelectro- From 1998 to 2005, he was a Research Assistant
mechanical instrument and systems development at the Charles Stark with the MEMS-VLSI Research Group, Department
Draper Laboratory,” in Proc. AIAA/IEEE Digital Avionics Syst. Conf., of Electrical and Electronics Engineering, METU,
Oct. 1997, pp. 8.5.33–8.5.40. where he has been a Senior Research Scientist since
[4] G. He and K. Najafi, “A single-crystal silicon vibrating ring gyro- 2006. His major research interests include capacitive
scope,” in Proc. 15th IEEE Int. Conf. MEMS, Las Vegas, NV, Jan. 2002, inertial sensors, micromachined resonators and actuators, capacitive interface
pp. 718–721. circuits, and microfabrication technologies.
[5] G. T. Schmidt, “INS/GPS technology trends,” in Proc. NATO RTO Sens. Dr. Alper received the “METU Thesis of the Year Award” in 2000 and 2005
Electron. Technol. (RTO-SET) Panel: Symp. Emerging Mil. Capabilities for his [Link]. thesis and Ph.D. dissertation, respectively, which were awarded
Enabled Advances Navigat. Sens. (SET), Istanbul, Turkey, Oct. 2002, by the Prof. Mustafa N. Parlar Education and Research Foundation. He is the
pp. 55–62. first author of the symmetric and decoupled gyroscope design, which won the
[6] J. A. Geen, S. J. Sherman, J. F. Chang, and S. R. Lewis, “Single-chip first-prize award in the operational designs category of the International Design
surface micromachined integrated gyroscope with 50◦ /h Allan deviation,” Contest organized by DATE and CMP in March 2001. He is also the first author
IEEE J. Solid-State Circuits, vol. 37, no. 12, pp. 1860–1866, Dec. 2002. of the tactical-grade symmetrical and decoupled microgyroscope design, which
[7] W. A. Clark, R. T. Howe, and R. Horowitz, “Surface micromachined won the third-prize award, among 132 MEMS designs from 24 countries and
Z-axis vibratory rate gyroscope,” in Proc. Solid-State Sens. Actuator 25 states across the U.S., in the international 3-D MEMS Design Challenge
Workshop, Hilton Head, CA, Jun. 2–6, 1996, pp. 283–289. organized by MEMGEN Corporation (currently Microfabrica), in June 2003.

Authorized licensed use limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.
 ALPER et al.: COMPACT ANGULAR RATE SENSOR SYSTEM USING A FULLY DECOUPLED SOG MEMS GYROSCOPE 1429

Yuksel Temiz was born in Erzurum, Turkey, in 1982. Tayfun Akin (S’90–M’97) was born in Van, Turkey,
He received the B.S. and M.S. degrees in electrical in 1966. He received the B.S. degree (with high
and electronics engineering from Middle East Tech- honors) in electrical engineering from Middle East
nical University (METU), Ankara, Turkey, in 2005 Technical University (METU), Ankara, Turkey, in
and 2007, respectively, with the highest GPA among 1987 and the M.S. and Ph.D. degrees in electrical
all graduates of the METU Engineering Faculty. He engineering from the University of Michigan, Ann
is currently working toward the Ph.D. degree with Arbor, in 1989 and 1994, respectively, with a gradu-
the Institute of Electrical Engineering, Ecole Poly- ate fellowship provided by the NATO Science Schol-
technique Fédérale de Lausanne (EPFL), Lausanne, arship Program through the Scientific and Technical
Switzerland. Research Council of Turkey (TUBITAK).
Between 2005 and 2007, he was a Research As- Beginning in 1995 and 1998, he was an Assistant
sistant with the MEMS-VLSI Research Group, METU, with a full graduate Professor and an Associate Professor, respectively, with the Department of
fellowship provided by the Scientific and Technical Research Council of Turkey Electrical and Electronics Engineering, Middle East Technical University,
(TUBITAK). He is currently a Research/Teaching Assistant with the Microelec- where he has been a Professor since 2004. He is also the Director of the METU-
tronic Systems Laboratory, EPFL. His research interests include readout and MEMS Research Center, which has a 1300-m2 clean-room area for MEMS
control electronics for micromachined inertial sensors, interface electronics for process and testing. His research interests include MEMS, microsystems
biosensors, and fabrication technologies for BioMEMS. technologies, infrared detectors and readout circuits, silicon-based integrated
sensors and transducers, and analog and digital integrated-circuit design.
Dr. Akin has served various MEMS, EUROSENSORS, and TRANSDUC-
ERS conferences as a Technical Program Committee Member. He was the
Cochair of the 19th IEEE International Conference of Micro Electro Me-
chanical Systems (MEMS 2006) held in Istanbul. He was the winner of the
First Prize in the Experienced Analog/Digital Mixed-Signal Design Category
at the 1994 Student VLSI Circuit Design Contest organized and sponsored
by Mentor Graphics, Texas Instruments, Hewlett-Packard, Sun Microsystems,
and Electronic Design Magazine. He is the coauthor of the symmetric and
decoupled-gyroscope project that won the first-prize award in the operational
designs category of the international design contest organized by the DATE
Conference and CMP in March 2001. He is also the coauthor of the gyroscope
project that won the third-prize award of the 3-D MEMS Design Challenge
organized by MEMGen Corporation (currently Microfabrica).

Authorized licensed use stats

View publication limited to: ULAKBIM UASL - MIDDLE EAST TECHNICAL UNIVERSITY. Downloaded on December 14, 2008 at 09:40 from IEEE Xplore. Restrictions apply.