Sensors and Actuators B 180 (2013) 50–59

Contents lists available at SciVerse ScienceDirect

Sensors and Actuators B: Chemical

journal homepage: www.elsevier.com/locate/snb

Integrated silicon-based chemical microsystem for portable sensing applications

K.S. Demirci a,b , L.A. Beardslee a , S. Truax a , J.-J. Su a , O. Brand a,∗
a
Georgia Institute of Technology, School of Electrical and Computer Engineering, Atlanta, GA, USA
b
Texas Instruments Inc., Dallas, TX, USA

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

Article history: This paper presents a chemical microsystem based on the integration of a silicon-based resonant
Available online 17 March 2012 microsensor and a CMOS ASIC for portable sensing applications in air. Two types of resonant microstruc-
tures are used as mass-sensitive microsensors: cantilever-based and disk-shape microresonators. Based
Keywords: on the characteristics of the microsensors, CMOS-integrated interface and control electronics have been
Chemical microsystem designed and implemented. The CMOS ASIC utilizes the self-oscillation method in which a main feedback
CMOS ASIC
loop starts and sustains oscillation at or close to the desired resonance frequency of the microresonators.
Silicon-based resonant sensor
For stable oscillation, an automatic gain control loop regulates the oscillation amplitude by controlling
Mass-sensitive sensor
the gain of the main feedback loop. In addition, an automatic phase control loop has been included to
adjust the phase of the main feedback loop to ensure an operating point as close as possible to the res-
onance frequency, which results in improved frequency stability. To improve the long-term stability,
a method to compensate for frequency drift has been proposed and implemented on the CMOS chip.
The performance of the implemented microsystem as a chemical sensor has been evaluated experimen-
tally by detecting different concentrations of volatile organic compounds such as benzene, toluene and
m-xylene with microresonators coated with chemically sensitive polymer films.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction cost are the key issues [1]; thus, reliable and low-cost hand-held
chemical sensor instruments targeting a finite number of analytes
Bench-top analytical instruments are widely used for chemi- of interest rather than a complete chemical analysis are needed.
cal analyses. These instruments are able to analyze even complex Among possible chemical sensing platforms, mass-sensitive
mixtures with high accuracy, high selectivity, and excellent repro- chemical sensors have captured substantial interest in recent
ducibility. However, their application areas are limited because of years in the form of quartz-crystal microbalances (QCMs), surface
their large sizes, high prices, and time consuming and labor inten- acoustic wave (SAW) devices and microcantilever sensors [2,3].
sive operation requirements. On the other hand, chemical sensors As application examples, detection of volatile organic compounds
offer fast response times, in-field sensing capabilities, portabil- (VOCs) using QCMs [4–7], SAW devices [8,9], and microcantilevers
ity and potentially low analysis cost allowing for widespread use, [10,11], and detection of warfare agents using SAW devices [12]
even if they show lower sensitivity and selectivity compared to and microcantilevers [13] have been demonstrated. Mass-sensitive
bench-top analytical instruments. Although chemical sensors and chemical sensors belong to the class of resonant sensors that are
bench-top analytical systems are not directly competing, chemi- covered with a chemically sensitive layer for chemical sensing and
cal sensors offer complementary solutions that should be chosen excited at one of their resonance modes. Their resonance char-
depending on the application requirements. In particular, micro- acteristics such as resonance frequency, quality factor (Q-factor),
fabricated chemical sensors and sensor arrays have tremendous phase, and vibration amplitude change upon mass loading on the
potential in applications requiring continuous on-site monitoring sensor caused by sorption of the chemical analytes in the sen-
of a range of chemical analytes, e.g. in environmental monitoring sitive layer. The resonance frequency is usually chosen as the
applications. In many of these applications, portability and system sensor output signal, since the frequency signal can be easily
digitized using a counter. This digitization eliminates the inaccu-
racies caused by analog-to-digital (A/D) conversion and provides
a wide dynamic range [14]. Moreover, the frequency output is
∗ Corresponding author at: Georgia Institute of Technology, School of Electrical
not sensitive to amplitude fluctuations associated with analog sig-
and Computer Engineering, 777 Atlantic Drive, Atlanta, GA 30332-0250, USA.
nals and can be measured with high resolution [15]. In addition,
Tel.: +1 404 894 9425; fax: +1 404 894 4700.
E-mail addresses: [email protected] (K.S. Demirci), [email protected] mass-sensitive chemical sensors, especially silicon-based ones, can
(O. Brand). be readily miniaturized using microfabrication techniques. This

0925-4005/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.snb.2012.01.061
 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59 51

miniaturization allows realization of array of sensors in which

each sensor is covered with a different sensitive layer and thus
offers partial selectivity that is generally required for discrimina-
tion of multiple analytes [16]. These advantages make these sensors
promising candidates for portable chemical sensor systems for con-
tinuous real-time monitoring applications [17].
Nevertheless, previous research efforts on mass-sensitive chem-
ical sensors have typically focused on the transducer and often
neglected important system aspects as well as the circuitry used to
condition the sensor and read out the sensor signal. As previously
shown in [18] and [19], proper circuitry can improve the chemical
sensor performance in critical areas, such as frequency stability,
and can also allow the integration of the entire system in a com-
pact package with reduced power consumption and cost. Therefore,
this work uses a system approach to develop interface and control
circuitry for a mass-sensitive chemical sensor array. The circuitry
automatically drives the sensors at their optimal working point and
compensates for temperature-induced long-term drift.

2. Silicon-based resonant microstructures

The CMOS ASIC was designed to interface with cantilever-based

[20] and disk-shape [21] microresonators with resonance frequen-
cies ranging from 100 kHz to 1 MHz, depending on the device
dimensions. Fig. 1 shows scanning electron micrograph (SEM)
images of these resonant microstructures, and Table 1 summarizes
their key mechanical and electrical parameters. The microres-
onators feature integrated resistive structures for electrothermal
excitation and piezoresistive detection of in-plane vibrations and
are fabricated on single crystal silicon substrates using a CMOS-
compatible fabrication process [21]. In addition, silicon dioxide
and silicon nitride films are used for insulation and passivation.
For array operation, eight microresonators are implemented on a
sensor chip.
To reduce the damping by the surrounding fluid, the microres-
onators are operated at their fundamental in-plane flexural
vibration mode, resulting in improved Q-factors and thus improved
short-term frequency stabilities. To this end, the excitation and
detection resistors have to be placed such that the desired in-plane Fig. 1. SEM images of (i) a cantilever-based “hammerhead” microresonator and (ii)
vibration mode is preferentially excited and detected. Therefore, a disk-shape microresonator.
four piezoresistors in a Wheatstone bridge arrangement are located
close to the clamped edge of the cantilever [20] and at the center of
the anchor beam of the disk-shape microresonator [21]. The partic- heating resistors, and R1–4 are the Wheatstone bridge resistors.
ular resistor arrangement is designed with the characteristic stress To actuate the disk-type microresonators at their fundamental
distributions of the different flexural mode shapes in mind and sup- in-plane resonance frequencies, out-of-phase AC signals superim-
presses unwanted mode shapes, such as the flexural out-of-plane posed on DC offsets (VDC and −VDC ) are applied to the heating
modes. resistors as excitation signals (VEX1 and VEX2 ) as shown in Fig. 2(ii).
Fig. 2(i) shows the schematic diagram of the excitation and In the case of cantilevers, only one heating resistor is used. With
detection elements of the microresonators. Here, REX are the the superimposed DC offset, the excitation signal and the heating
power generated in a heating resistor are, respectively, given by

Table 1 VEX = VDC + VAC cos ωt (1)

Key mechanical and electrical parameters of the microresonators used for chemical
2 2
VAC 2
VAC
sensing.
2
VEX VDC 2VDC VAC
P= = + + cos ωt + cos 2ωt (2)
Parameter Value
REX REX 2REX REX 2REX

Cantilever resonator Disk-type resonator The first two terms in Eq. (2) constitute the static heating power
resulting in a static temperature elevation in the microstructure.
Device dimensions
Radius 150–300 ␮m 120–150 ␮m The desired dynamic heating power component at the excitation
Thickness 5–20 ␮m 5–20 ␮m frequency ω is the third term in Eq. (2), which is directly propor-
Resonance frequency 100 kHz to 1 MHz 200 kHz to 1 MHz tional to the applied DC offset. Therefore, the effect of the unwanted
Q-factor (in air) 2000–5000 1500–5800 dynamic power component at the frequency of 2ω (fourth term in
Attenuation (in air) 40–80 dB
Heating resistance, REX 400–600 
Eq. (2)) can be reduced by increasing the DC offset (VDC > VAC ).
Wheatstone bridge resistance, R1–4 400–600  The Wheatstone bridge is biased with DC voltages VWH and
Wheatstone bridge bias, VWH 1–1.5 V −VWH as shown in Fig. 2. The deflection of the microresonator
DC offset, VDC 1.5–4.5 V results in an opposite-sign resistance change in the two diago-
AC excitation amplitude, VAC 0.2–1 V
nal resistor pairs of the Wheatstone bridge. In other words, when
52 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59

Fig. 3. Typical measured amplitude and phase transfer characteristics of a disk-type

microresonator.

3. Description of electronic system

Fig. 4 shows the schematic diagram of the implemented

microsystem. (The shaded blocks in this figure have been imple-
mented off-chip.) At a given time, one microresonator on the sensor
chip is connected to the CMOS ASIC using the input multiplexer.
The piezoresistive Wheatstone bridge of the selected resonator is
biased by an off-chip bias circuit composed of a buffer amplifier
and a digital-to-analog converter (DAC).
A self-oscillation technique is selected to operate the microres-
onator as a mass-sensitive sensor. This technique incorporates the
resonator as the frequency determining element in an amplifying
feedback loop, named the main feedback (MF) loop. If the loop
Fig. 2. (i) Schematic diagram of the excitation and detection elements of the
microresonators and (ii) out-of-phase excitation signals (VEX1 and VEX2 ). gain is greater than unity, and the total loop phase shift is 0◦ , the
overall system will act as a positive feedback loop, and the res-
onator will oscillate at its resonance frequency. Besides the MF loop,
the resistances of R1 and R3 increase, the resistances of R2 and R4
for improved short- and long-term frequency stabilities, the ASIC
decrease, or vice versa. Assuming matched resistors with a resis-
includes several closed-loop control circuits such as an automatic
tance of R, the resulting differential output voltage is given by
gain control (AGC) circuit, an automatic phase control (APC) circuit,
R and a frequency drift compensation (FDC) circuit. The AGC circuit
VDET = 2VWH (3)
R regulates the oscillation amplitude in the MF loop for stable oscil-
where R is the resistance change in each resistor as a result of the lation. The APC circuit adjusts the phase of the MF loop to ensure
deflection of the microstructure. The differential read-out mecha- an operating point as close as possible to the resonance frequency
nism doubles the output voltage compared to that of a single-ended of the resonator for improved short-term frequency stability. To
counterpart, while canceling out any common-mode resistance improve the long-term frequency stability, the FDC circuit com-
change originating from, e.g. out-of-plane and torsional vibrations pensates for temperature-induced frequency drift. The following
and environmental effects, such as temperature variations. On the sections give details of the implemented circuitry.
other hand, a perfect matching of the Wheatstone bridge resistors
is impossible; hence, an offset voltage at the output is inevitable 3.1. Gain stages and filters
and has to be taken into account in the design of the interface
electronics. To obtain an overall loop gain greater than or equal to unity
To properly design the interface electronics, open-loop transfer (and thus overcome the attenuation by the resonator), the MF
characteristics of the microresonators have been recorded using a loop includes a series of four gain stages that amplifies the low-
network analyzer. Fig. 3 shows typical measured amplitude and amplitude differential readout signal from the Wheatstone bridge
phase transfer characteristics of a disk-shape microresonator. The of the sensor. The resonance frequency is different for different
amplitude response varies with the device dimensions and Wheat- resonator geometries and varies between 100 kHz and 1 MHz (see
stone bridge bias voltage. In other words, the signal attenuation by Table 1). Therefore, the 3-dB bandwidths of the gain stages have
the microresonators is not constant. In addition, the phase response to be at least 1 MHz to make the amplification independent of
usually shows deviations from the theoretical value of −90◦ at the resonator used. Instrumentation amplifiers (in-amps) utiliz-
the resonance frequency and varies from resonator to resonator, ing closed-loop amplifier topology are typically used to interface
because of fabrication-induced variations and parasitic couplings sensors with Wheatstone bridge outputs. The closed-loop topol-
between the excitation and detection elements [22]. Both of these ogy improves the gain linearity of the in-amp and makes its gain
non-idealities are addressed by the interface electronics. less sensitive to variations in temperature, active device operating
 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59 53

Fig. 4. Schematic diagram of the integrated CMOS interface and control electronics when interfaced with the sensor chip.

conditions, and transistor parameters [23,24]. On the other hand, The attenuation of the microresonators is not constant but varies
the closed-loop topology typically limits the 3-dB bandwidth of from design to design between 40 dB and 80 dB depending on bias-
the in-amp at a given power consumption because of the multi- ing conditions and device dimensions (see Table 1). Hence, the gain
pole frequency response of the multi-stage op-amps generally provided by the gain stages has to be variable. In this respect, a
used to realize the in-amp. However, since in the present appli- variable-gain amplifier (VGA) is implemented as the 2nd gain stage.
cation the output signal of the implemented microsystem is a Moreover, the 3rd gain stage can be enabled and disabled to obtain
frequency and not a voltage, variations and nonlinearity in the additional gain tuning range. As a result, the total gain is adjustable
gain of the amplifiers are not very critical and are compensated between 41.6 and 79.2 dB.
with the automatic gain control (AGC) loop described later. There- First-order RC filters are implemented in between the gain
fore, open-loop instead of closed-loop amplifiers are implemented stages. Low-pass filters after the 1st and 3rd gain stages reduce
in the first three gain stages of the ASIC to obtain the necessary high-frequency signal coupling into the bandwidth of interest, and
3-dB bandwidth with relatively lower power consumption. In addi- high-pass filters after the 2nd and 4th gain stages remove any DC
tion, a fully differential topology is employed in the first three gain offset and flicker noise of these gain stages. The 3-dB cut-off fre-
stages, where the signal amplitude is low and thus more vulnerable quencies of the HPF and the LPF are measured approximately as
to noise. Fully differential amplifiers offer higher signal-to-noise 3.1 kHz and 14.2 MHz, respectively. Thus, the filters do not affect
ratio (SNR) because of their larger output voltage swing and are the amplitude and phase transfer characteristics of the feedback
less susceptible to common-mode noise than their single-ended loop in the desired operation bandwidth (100 kHz to 1 MHz).
counterparts [23,25].
Fabrication-induced mismatches between the Wheatstone
bridge resistors result in a DC-offset component added to the
low-amplitude differential AC readout signal of the resonator. In
addition, CMOS amplifiers suffer from input offset voltages orig-
inating from mismatches in their device parameters. As a result,
the offset voltage at the input of the 1st gain stage is typically on
the order of tens of millivolts, which reduces the dynamic range or
even can saturate the amplifier. Therefore, the 1st gain stage uti-
lizes an offset compensation method [26,27] as shown in the block
diagram in Fig. 5. In this method, the output signal of the main
amplifier is low-pass filtered to extract the low-frequency compo-
nent of the signal. This low-frequency signal is then amplified by
an auxiliary amplifier to generate an error signal proportional to
the total offset voltage at the output of the main amplifier. This
error signal is connected to the nulling port to cancel the offset
voltage at the main amplifier input. With this closed-loop opera-
tion, both amplifier and sensor offsets can be compensated without
disturbing the high-frequency sensor signal. Since this loop cannot
distinguish low-frequency (1/f) noise from offset, the implemented
offset compensation method also shapes the 1/f noise of the main Fig. 5. Block diagram of the 1st gain stage showing the concept of the offset com-
amplifier. pensation method used.
54 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59

3.2. Amplitude control the digital control code of the DCPS to shift the phase. At every
code change, the current oscillation amplitude is compared with
Even though gain stages are essential to start self-oscillation, the previous one associated with the previous code. If the current
they are not enough to sustain stable oscillation. A stable oscil- amplitude is greater than or equal to the previous one, the code
lation is obtained by controlling the amplitude of the oscillation. is changed again. This operation is repeated until the oscillation
Without an amplitude control, the positive feedback introduced by amplitude is maximized. Once the maximum is obtained, the digital
the MF loop will cause the oscillation amplitude either to increase control code is held constant.
to levels that saturate the amplifiers, resulting in distortion, or to Both the AGC and the APC loops have an influence on the oscil-
decrease until the oscillation disappears. To prevent this instabil- lation amplitude. The AGC loop directly adjusts the oscillation
ity, a gain control mechanism is needed to obtain a high loop gain amplitude, whereas the APC loop changes the oscillation ampli-
at the start-up, while maintaining a reduced loop gain during the tude through phase adjustment. Although the oscillation amplitude
quasi-stationary phase. Therefore, an automatic gain control (AGC) is a reference measure in both loops, they do not interfere with
loop has been implemented. each other since the APC loop monitors the 1st gain stage out-
AGC circuits have been employed with electrostatically trans- put, while the AGC loop tries to only keep the 2nd gain stage
duced MEMS resonators to keep their oscillation amplitude output amplitude constant by controlling the amplification of the
constant [28–30]. In this work, a similar architecture is used to 2nd gain stage. Therefore, the AGC loop does not affect the 1st
implement the AGC loop as shown in Fig. 4. Here, the output sig- gain stage output. (This is experimentally verified with Fig. 9 in
nal of the 4th gain stage is connected to a level detector composed Section 4.1.)
of a rectifier and an integrator. The input signal of the AGC is first
half-wave rectified by the rectifier. The integrator then low-pass fil-
ters this rectified signal to generate a DC voltage proportional to the 3.4. Frequency drift compensation
oscillation amplitude. The error amplifier compares this DC voltage
at the level detector output with a reference level (Vref ) to generate The resonance frequency of a microresonator is affected by
a control voltage, which is connected back to the 2nd gain stage to changes in its mass and its stiffness. If the microresonator is used
adjust the gain of the MF loop. This way, the AGC loop regulates the as a mass-sensitive chemical sensor, it is exposed to the environ-
oscillation amplitude. ment in order to detect the analytes of interest. In this case, not
only any changes in the environmental conditions, such as changes
3.3. Phase adjustment in temperature, humidity or pressure [32], but also microstruc-
ture aging [33] might affect the stiffness of the microresonator and
Another important point in the feedback loop operation is the hence cause stiffness-induced resonance frequency changes. As a
phase adjustment. The phase shift at the resonance frequency result, the resonance frequency varies in time and generates a sen-
varies from resonator to resonator. In addition, the circuit blocks in sor response that is very similar to the one generated by a slowly
the MF loop also add phase shifts and change the total loop phase. changing analyte concentration. In this situation, it becomes very
Therefore, a digitally controlled phase shifter (DCPS) is included difficult, if not impossible, to determine if the sensor response, i.e.
in the MF loop to make the phase in the loop close to 0◦ by over- frequency change, is the result of drift or the presence of a chem-
coming the phase shifts of the resonator and circuit blocks to start ical analyte. Therefore, frequency drift caused by environmental
and sustain self-oscillation. The DCPS is realized using a first-order disturbances limits the sensor resolution especially in long-term
all-pass filter [31] with digitally controlled capacitor and resistor monitoring applications with slowly changing analyte signatures
banks and introduces a frequency-dependent delay while keeping and has to be compensated for in order to obtain high sensor reso-
the amplitude of the input signal constant. lution.
The total phase shift in the MF loop in fact determines the To address the need for compensating drift effects, the ASIC
operating point, i.e. the oscillation frequency. Unless the phase of employs a frequency drift compensation (FDC) loop (see Fig. 4),
the feedback circuitry is exactly the same (with opposite sign) as which uses a controlled stiffness modulation scheme described in
the phase of the microresonator at the resonance frequency, the [34]. The FDC circuit shifts the phase of the main oscillation sig-
oscillation frequency will be slightly different from the resonance nal by −90◦ using a DCPS and hence generates a signal in-phase
frequency. More importantly, variations in the phase of the cir- with the deflection of the resonator. A fully differential VGA adjusts
cuitry will cause variations in the oscillation frequency, and thus the amplitude of this single-ended signal and, at the same time,
generate frequency noise. The effect of phase variations on oscil- converts it to fully differential out-of-phase compensation signals.
lation frequency can be reduced by putting the operating point as These compensation signals are added to the main oscillation signal
close as possible to the resonance frequency, where the slope of the by a summing amplifier in the MF loop and applied to the excitation
phase response is maximum (see Fig. 3). To this end, the phase shift elements of the resonator. As a result, the effective stiffness of the
introduced by the DCPS is controlled by an automatic phase con- resonator is modulated [35]. By controlling the polarity of the com-
trol (APC) loop as shown in Fig. 4 to adjust the operating point. The pensation signals using a cross-switch in the FDC loop, the effective
amplitude response of the resonators shows a resonance peak at stiffness of the resonator can be either reduced or increased. Con-
the resonance frequency, i.e. the oscillation amplitude is at its max- sequently, the resonance frequency of the harmonic oscillator is
imum at this point. Thus, to sustain oscillation as close as possible changed in response to the modulated effective stiffness. The FDC
to the resonance frequency, the APC loop monitors the oscillation loop is periodically enabled (e.g. for 2 s every 52 s in [34]) and result-
amplitude and tunes the phase of the DCPS until a maximum oscil- ing changes in resonance frequency are measured. By using the
lation amplitude is obtained. measured resonance frequency changes, the Q-factor (or the rela-
The APC circuit includes an amplifier, which converts the fully tive Q-factor change) of a microresonator in response to variations
differential signal at the output of the 1st gain stage into a single- in the environmental conditions can be extracted and tracked dur-
ended one with a gain of 20 dB. Then, the amplitude level of this ing closed-loop sensor operation. The measured Q-factor is then
amplified signal is detected by a level detector, the same as the used to compensate for the frequency drift using a relation between
one used in the AGC circuit. The detected level is interfaced to a Q-factor and resonance frequency obtained through an initial cali-
PC through a data acquisition (DAQ) card. A Labview program on bration step. (The details and experimental results of the FDC loop
the PC executes an iterative APC algorithm, in which the PC adjusts are not included in this paper.)
 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59 55

Fig. 6. Micrograph of the CMOS chip, which has been fabricated using Texas Instru-
ments’ 0.35 ␮m 4M2P BiCMOS technology and measures 2500 ␮m × 1500 ␮m.

3.5. Sensor drive

A cross-switch is implemented after the summing amplifier to

generate an additional 180◦ phase shift by changing the polarity of
the amplified and phase shifted signals in the MF loop, if needed.
Off-chip summing amplifiers then add DC-offset voltages to the sig-
nals at the cross-switch output to generate out-of-phase excitation
signals. The excitation signals are then applied to the heating resis-
tors through a multiplexer. Moreover, a comparator is included to
convert the sine-wave oscillation signal into a square-wave signal
so that it can be directly interfaced to a frequency counter.

4. Results and discussion

4.1. Circuit characterization

Fig. 6 shows the micrograph of the CMOS chip, which has been
fabricated using Texas Instruments’ 0.35 ␮m 4M2P BiCMOS tech- Fig. 7. DC-shifted out-of-phase excitation signals when the ASIC is interfaced to
nology and measures 2500 ␮m × 1500 ␮m. A custom-made printed resonators with resonance frequencies of (i) 382.5 kHz and (ii) 795 kHz.
circuit board (PCB) with the necessary off-chip electronics was
built to operate the CMOS ASIC together with a microresonator and
control it by a PC via a data acquisition (DAQ) card (PCI NI-6251, Wheatstone bridge is biased with ±1 V. With these bias voltages,
National Instruments). A Labview program controls the DAQ card the attenuations from the resonators with resonance frequencies of
to generate control signals for the ASIC and the off-chip electron- 382.5 kHz and 795 kHz have been measured with a network ana-
ics. The PCB is powered with ±6 V. The supply voltages ±1.65 V to lyzer approximately as 55 dB and 60 dB, respectively. Hence, the
power the ASIC are generated with regulators. Level translators are gain stages have to provide gains of at least 55 dB for the 382.5 kHz
incorporated to convert the signals from the DAQ card with CMOS operation and at least 60 dB for the 795 kHz operation. The refer-
logic voltage levels into ±1.65 V to be able to interface with the ence voltage of the AGC circuit (Vref ) is set to 1.4 V and 1.2 V, yielding
ASIC. The ASIC consumes a current of 5 mA, resulting in a power excitation amplitudes around 580 mVpp and 800 mVpp for the oper-
consumption of 16.5 mW. With the off-chip electronics on the PCB, ation at 382.5 kHz and 795 kHz, respectively. By decreasing Vref , the
the power consumption of the implemented system increases to excitation amplitude can be increased, or vice versa.
117.6 mW. Since commercial ICs with low current consumption are In addition, the start-up of the oscillation is demonstrated in
employed to realize the off-chip electronics (4.8 mA in total), this Fig. 8, which shows one of the excitation signals (top waveform)
power consumption increase is mostly because of the increase in and the detected oscillation level (bottom waveform) when the res-
the power supply voltages from ±1.65 V to ±6 V, which is necessary onator with a resonance frequency of 382.5 kHz is used. To start the
to be able to interface the system with the DAQ card. In a possible oscillation, the 3rd gain stage is enabled, which increases the total
implementation of the system on a hand-held device in the future, a loop gain by 20 dB. The start-up time in this case is below 15 ms.
microcontroller is going to be used instead of the DAQ card, and the After start-up, the oscillation amplitude is regulated to a constant
level translators are going to be removed. Therefore, the power con- level by the AGC loop.
sumption can be further reduced significantly. It should be noted The proper operation of the APC loop is verified by Fig. 9, which
that the current power consumption of the ASIC is of the order of shows the increase in the level detector output (bottom waveform)
the power consumption of the microresonators (5–20 mW) with and one of the excitation signals (top waveform) as the APC loop
their thermal excitation and piezoresistive detection resistors. adjusts the phase by changing the digital control code. The APC
The closed-loop operation of different microresonators in air by loop stops and holds the previous digital control code when the
the CMOS ASIC is verified by Fig. 7, which shows the DC-shifted maximum oscillation amplitude is obtained.
out-of-phase excitation signals when the ASIC is interfaced to res- The silicon-based cantilever- and disk-shape resonators have
onators with resonance frequencies of 382.5 kHz and 795 kHz. In been used for gas- and liquid-phase chemical measurements using
both cases, the DC-offset voltage is adjusted to ±1.5 V, and the off-chip electronics implemented on a PCB in our previous studies
56 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59

Fig. 10. Allan variance as a function of gate time of an uncoated resonator with a
resonance frequency of 377.1 kHz and a Q-factor of 2850 in air connected to the
off-chip electronics (solid line) and the CMOS ASIC (dashed line).

[36,37]. However, the power consumption and size of the off-chip

electronics have been preventing implementation of a portable
chemical sensor system. To estimate the obtainable frequency sta-
bility of the implemented microsystem in this work and compare
its performance with the off-chip electronics, the same uncoated
resonator with a resonance frequency of 377.1 kHz and a Q-factor
of 2850 in air was interfaced to both the ASIC and the off-chip elec-
tronics, and the closed-loop oscillation frequency was recorded
using a frequency counter with a gate time of 1 s. It should be
noted that the frequency stability is a measure of the system perfor-
mance. In other words, the noise of both the microresonator and the
interface electronics determine the frequency stability. Therefore,
Fig. 8. (i) One of the excitation signals (top waveform) and the detected oscillation a fair performance comparison between the interface electronics
level (bottom waveform) during start-up of the oscillation and (ii) close-up view. is attempted by incorporating the same resonator in the measure-
(For interpretation of the references to color in this figure legend, the reader is ments.
referred to the web version of this article.) The recorded data is then used to calculate the Allan variance
for different gate times as shown in Fig. 10. To obtain the Allan
variance for gate times larger than 1 s, neighboring data points from
the measured data have been averaged. The thermal noise of the
system limits the short-term frequency stability. In other words,
the Allan variance for small gate times, i.e. the portion of the Allan
variance curve in Fig. 10 up to a gate time of 5 s, depends on the
thermal noise performance. In contrast, the Allan variance for larger
gate times depends on both the flicker noise in the system and
environmental disturbances (e.g. temperature variations).
As can be observed from Fig. 10, the off-chip electronics show
the best short-term frequency stability (for a gate time of 1 s), since
the used discrete amplifiers have better noise performance because
of their BJT input stages and higher current dissipation. Conversely,
the integrated system outperforms the off-chip electronics for gate
times greater than 4 s due to the fact that environmental distur-
bances have less effect on the packaged ASIC than on the large PCB
containing the off-chip electronics.
Table 2 summarizes the frequency stability measurement
results mentioned above and gives a performance comparison
between the CMOS-integrated electronics and off-chip electron-
ics when interfaced with the same resonator. In summary, when
compared to off-chip electronics, the CMOS-integrated interface
circuits implemented in this work show comparable noise perfor-
mance with much less power consumption (more than an order
Fig. 9. Increase in the level detector output (bottom waveform) and one of the
excitation signals (top waveform) as the APC loop adjusts the phase by changing
of magnitude) and therefore allow the future integration of the
the digital control code. (For interpretation of the references to color in this figure entire system in a compact package resulting in reduced system
legend, the reader is referred to the web version of this article.) cost. In addition, with an Allan variance of 3.4 × 10−8 obtained with
 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59 57

Table 2
Performance comparison between CMOS-integrated electronics and off-chip electronics when interfaced with the same resonator (f = 377.1 kHz, Q = 2850 in air,  = gate time).

Off-Chip ASIC

Allan variance  =1s 1.67 × 10−8 3.42 × 10−8

 = 10 s 18.4 × 10−8 2.43 × 10−8
 = 100 s 55.2 × 10−8 3.45 × 10−8

Frequency stability  =1s 6.3 mHz 12.9 mHz

 = 10 s 69.5 mHz 9.2 mHz
 = 100 s 208.4 mHz 13 mHz

Power consumption Chip only N/A 16.5 mW (±1.65 V)

System 2.8 W (±5 V and ±10 V) 117.6 mW (±6 V)

a gate time of 1 s, the ASIC shows almost 3 times better short- with analyte and generates saturated analyte vapor at the output
term frequency stability than an oscillator microsystem utilizing of the bubbler. The concentration of the analyte in the carrier gas
a cantilever-type microresonator [18]. stream depends on the temperature-dependent saturation vapor
Finally, the influence of the phase adjustment on the short-term pressure of the analyte, which is given by the Antoine equation [39].
frequency stability is investigated for the same resonator and pre- The initial analyte concentration is then diluted with carrier gas
sented in Fig. 11. Here, the digital code of the DCPS in the MF loop is piped through line L2 to obtain a final analyte concentration. The
changed to operate the uncoated resonator at different oscillation carrier gas line L3 provides the reference gas to the gas-tight mea-
frequencies. The relative phase shift represents the phase change surement chamber in the purge mode. Three mass flow controllers
with respect to the optimal phase adjustment in the MF loop. At the (MFCs) (SFC4100, Sensirion) adjust the rate of gas flows through the
optimum phase adjustment, the oscillation frequency is as close lines.
as possible to the resonance frequency of the resonator; therefore, During chemical measurements, a 4-way pneumatic valve
the best short-term frequency stability is obtained at this operating sequentially connects the analyte-loaded gas stream and the refer-
point. As the oscillation frequency moves away from the resonance ence gas stream to the measurement chamber. During the analyte
frequency, the frequency stability deteriorates. This measurement mode, the analyte-loaded gas is applied to the chamber, while the
demonstrates the importance of the phase adjustment to achieve carrier gas is exhausted. During the purge mode, the 4-way valve
better frequency stability and thus better sensor resolution. switches, and reference gas purges the measurement chamber. The
4-way valve provides sharp transients in the analyte concentrations
4.2. Gas-phase chemical sensing and makes the setup suitable to study transient effects in the tested
chemical sensors [40]. A solenoid valve controls the pressured air
The performance of the implemented microsystem (consist- to switch the 4-way valve. A flow meter is connected to the output
ing of the resonant sensor array, the described CMOS ASIC and of the setup to measure the overall flow through the measurement
the external components as shown in Fig. 4) as a chemical sensor chamber. The setup is controlled by a PC through a DAQ card and a
has been evaluated experimentally by detecting volatile organic Labview program, which allows automated measurements.
compounds (VOCs) in the gas phase. Fig. 12 presents a schematic To perform gas-phase chemical tests, microresonators were
diagram of the implemented measurement setup, which is similar spray-coated with 1.5 ␮m poly-isobutylene (PIB) as the chemically
to the one built in a previous study [38]. The implemented setup is sensitive polymer film. The PIB-coated resonator was mounted
composed of an analyte line, L1, and two carrier gas lines, L2 and in a 28-pin DIL package and then attached to the measurement
L3. Nitrogen (ultra high purity grade, Airgas) is used as the carrier chamber. The Wheatstone bridge bias and DC-offset voltages were
gas, which is supplied by a gas cylinder. The analyte line L1 employs adjusted to ±1 V and ±2 V, respectively. AC excitation signal ampli-
a glass bubbler to create saturated vapor of the analyte. The bub- tudes of approximately 800 mVpp were used. During the chemical
bler contains quartz sand soaked with liquid analyte. When the measurements, different concentrations of VOCs, such as ben-
carrier gas is streamed through the quartz sand, it becomes loaded zene, toluene and m-xylene have been detected. Reference gas
(nitrogen) and analyte-loaded gas were alternately applied to the
measurement chamber in approximately 5 min intervals. The ana-
lyte concentration was ramped up or down after every analyte
injection by adjusting analyte and reference gas flow rates while
keeping the total flow rate constant at 80 ml/min. The measure-
ments were performed at a temperature around 23 ◦ C.
Fig. 13 shows an example measurement where a cantilever-
based microresonator with a resonance frequency of 368.5 kHz
is exposed to different concentrations of benzene, resulting in a
decrease in the oscillation frequency because of the sorbed analyte.
Fig. 14 shows the measured frequency change of the PIB-coated
resonator as a function of analyte concentration when the con-
centration is ramped up (red circles) and ramped down (blue
squares). The frequency change shows almost a linear increase with
increasing analyte concentration and good reproducibility. From
the linear fits (lines) to the measured data, the sensitivity of the
microsystem for benzene, toluene, and m-xylene is extracted as
16.1 mHz/ppm, 70.7 mHz/ppm and 244.1 mHz/ppm, respectively.
The sensitivity increases with increasing partition coefficient of
Fig. 11. Influence of the phase adjustment on short-term frequency stability and
oscillation frequency. the analyte-polymer combination, which generally increases with
58 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59

Fig. 12. Schematic diagram of the chemical measurement setup.

decreasing saturation vapor pressure of the analyte at a given

temperature for structurally similar analytes. For mass-sensitive
resonant sensors, the limit of detection (LOD) is expressed by [10]

fmin fn
LOD = =3 (4)
S S
where S is the sensor sensitivity, fmin is the minimum detectable
frequency change, and fn is the frequency output noise estimated
by the Allan variance method. Therefore, with a short-term fre-
quency stability of 28.5 mHz extracted with the Allan variance
method, the LOD of the sensor system for benzene, toluene and
m-xylene is calculated as 5.3 ppm, 1.2 ppm and 0.35 ppm, respec-
tively.

5. Conclusion

In realizing hand-held sensing instruments, CMOS-integrated

interface electronics can play a crucial role. When compared to
off-chip electronics, CMOS interface circuits enable the integration
of the entire sensing system in a compact package with reduced
Fig. 13. Frequency shift of PIB-coated resonator as a function of time; the microsen- power consumption and system cost. Therefore, in this work, a
sor is subsequently exposed to different benzene concentrations. CMOS ASIC has been developed and integrated with silicon-based
resonant microsensors to implement a chemical microsystem,
which can be incorporated in a hand-held platform.
The development of a microsystem requires the assessment of
not only the appropriate CMOS interface circuit architectures for
signal conditioning but also of the characteristics and operation
principles of the sensor. Therefore, the resonant microstructures
previously developed in our research group have been carefully
examined to understand their characteristics pertinent to the cir-
cuit and system design. Accordingly, CMOS-integrated interface
and control electronics have been designed and implemented. The
ASIC includes interface and control circuitries to sustain oscillation
of the resonant sensor at its optimal working point and to compen-
sate for temperature-induced long-term drift. The circuit blocks
have been designed to obtain low noise and adequate bandwidth
with low power consumption, which is especially important for
portable devices. The implemented microsystem has been used for
detection of VOCs in the gas-phase showing detection limits with
high ppb to low ppm range for the analytes tested.

Acknowledgments

The authors thank Dr. Jeanne Pitz at Texas Instruments Inc. for
Fig. 14. Measured frequency change of the PIB-coated resonator as a function of
the analyte concentration when the concentration is ramped up (red circles) and providing IC fabrication support and the staff of the Georgia Tech
ramped down (blue squares), and linear fits to the measurement data (lines). (For Nanotechnology Research Center for their assistance during device
interpretation of the references to color in this figure legend, the reader is referred fabrication. This work has been funded in part by the National Sci-
to the web version of this article.) ence Foundation under awards ECCS-0601467 and ECCS-0824017.
 K.S. Demirci et al. / Sensors and Actuators B 180 (2013) 50–59 59

References [32] F.L. Walls, J.J. Gagnepain, Environmental sensitivities of quartz oscillators, IEEE
Transactions on Ultrasonics, Ferroelectrics and Frequency Control 39 (1992)
[1] P.A. Lieberzeit, F.L. Dickert, Sensor technology and its application in environ- 241–249.
mental analysis, Analytical and Bioanalytical Chemistry 387 (2007) 237–247. [33] M. Koskenvuori, T. Mattila, A. Haara, J. Kiihamaki, I. Tittonen, A. Oja, H. Seppa,
[2] R. Lucklum, P. Hauptmann, Acoustic microsensors – the challenge behind Long-term stability of single-crystal silicon microresonators, Sensors and Actu-
microgravimetry, Analytical and Bioanalytical Chemistry 384 (2006) 667–682. ators A: Physical 115 (2004) 23–27.
[3] J. Janata, M. Josowicz, P. Vanysek, D.M. DeVaney, Chemical sensors, Analytical [34] J.H. Seo, K.S. Demirci, A. Byun, S. Truax, O. Brand, Temperature compensation
Chemistry 70 (1998) 179r–208r. method for resonant microsensors based on a controlled stiffness modulation,
[4] Z.H. Ying, Y.D. Jiang, X.S. Du, G.Z. Xie, J.S. Yu, H.L. Tai, Polymer coated sensor array Journal of Applied Physics 104 (2008), 014911-1-9.
based on quartz crystal microbalance for chemical agent analysis, European [35] R. Sunier, T. Vancura, Y. Li, K.-U. Kirstein, H. Baltes, O. Brand, Resonant magnetic
Polymer Journal 44 (2008) 1157–1164. field sensor with frequency output, Journal of Microelectromechanical Systems
[5] P.A. Lieberzeit, A. Rehman, B. Najafi, F.L. Dickert, Real-life application of a 15 (2006) 1098–1107.
QCM-based e-nose: quantitative characterization different plant-degradation [36] S. Truax, K.S. Demirci, J.H. Seo, P. Kurzawski, Y. Luzinova, A. Hierlemann, B.
processes, Analytical and Bioanalytical Chemistry 391 (2008) 2897–2903. Mizaikoff, O. Brand, Gas and liquid phase sensing of volatile organics with
[6] X.X. Jin, Y. Huang, A. Mason, X.Q. Zeng, Multichannel monolithic quartz crystal disk microresonator, in: 21st IEEE International Conference on Micro Electro
microbalance gas sensor array, Analytical Chemistry 81 (2009) 595–603. Mechanical Systems – MEMS’ 08, 2008, pp. 220–223.
[7] J. Rabe, S. Buttgenbach, J. Schroder, P. Hauptmann, Monolithic miniaturized [37] S. Truax, K.S. Demirci, A. Hierlemann, O. Brand, Exploring the resolution of
quartz microbalance array and its application to chemical sensor systems for different disk-type chemical sensors, in: 15th International Conference on
liquids, IEEE Sensors Journal 3 (2003) 361–368. Solid-State Sensors, Actuators and Microsystems. Transducers 2009, 2009, pp.
[8] H.B. Lin, J.S. Shih, Fullerene C60-cryptand coated surface acoustic wave quartz 1838–1841.
crystal sensor for organic vapors, Sensors and Actuators B: Chemical 92 (2003) [38] A. Koll, CMOS capacitive chemical microsystems for volatile organic com-
243–254. pounds, PhD Thesis, ETH Zurich, 1999.
[9] Z.H. Li, Y. Jones, J. Hossenlopp, R. Cernosek, F. Josse, Analysis of liquid-phase [39] J.A. Riddick, W.B. Bunger, T.K. Sakano, Organic Solvents – Physical Proper-
chemical detection using guided shear horizontal-surface acoustic wave sen- ties and Methods of Purification, 4th ed., John Wiley & Sons, New York,
sors, Analytical Chemistry 77 (2005) 4595–4603. 1986.
[10] D. Lange, C. Hagleitner, A. Hierlemann, O. Brand, H. Baltes, Complementary [40] J.-J. Su, C. Carron, S. Truax, K.S. Demirci, L.A. Beardslee, O. Brand, Assessing poly-
metal oxide semiconductor cantilever arrays on a single chip: mass-sensitive mer sorption kinetics using micromachined resonators, in: 16th International
detection of volatile organic compounds, Analytical Chemistry 74 (2002) Conference on Solid-State Sensors, Actuators and Microsystems. Transducers
3084–3095. 2011, 2011, pp. 1420–1423.
[11] D. Then, A. Vidic, C. Ziegler, A highly sensitive self-oscillating cantilever array
for the quantitative and qualitative analysis of organic vapor mixtures, Sensors Biographies
and Actuators B: Chemical 117 (2006) 1–9.
[12] B.S. Joo, J.S. Huh, D.D. Lee, Fabrication of polymer SAW sensor array to classify
chemical warfare agents, Sensors and Actuators B: Chemical 121 (2007) 47–53.
K.S. Demirci received the B.S. and M.S. degrees in electrical and electronics engineer-
[13] I. Voiculescu, M.E. Zaghloul, R.A. McGill, E.J. Houser, G.K. Fedder, Electrostat-
ing from Middle East Technical University (METU), Ankara, Turkey in 2003 and 2005,
ically actuated resonant microcantilever beam in CMOS technology for the
respectively, and the Ph.D. degree in electrical engineering from Georgia Institute of
detection of chemical weapons, IEEE Sensors Journal 5 (2005) 641–647.
Technology, Atlanta in 2010. From 2003 to 2005, he was a Research Engineer with the
[14] E.R. Benes, R. Groschl, F. Seifert, A. Pohl, Comparison between BAW and SAW
METU-MEMS Research Group, Department of Electrical and Electronics Engineer-
sensor principles, IEEE Transactions on Ultrasonics Ferroelectrics and Fre-
ing, METU, where he was involved in the development and design of CMOS-based
quency Control 45 (1998) 1314–1330.
low-cost uncooled infrared detector focal plane arrays and their read-out circuits.
[15] H.A.C. Tilmans, M. Elwenspoek, J.H.J. Fluitman, Micro resonant force gauges,
From 2005 to 2010, he was a Graduate Research Assistant with the Georgia Insti-
Sensors and Actuators A: Physical 30 (1992) 35–53.
tute of Technology, where he focused on interface circuitry and system aspects of
[16] J.J. Lavigne, E.V. Anslyn, Sensing a paradigm shift in the field of molecular
resonant chemical sensors. Since 2010 he has been working as an analog IC design
recognition: from selective to differential receptors, Angewandte Chemie-
engineer at Texas Instruments, Inc., Dallas, TX.
International Edition 40 (2001) 3119–3130.
[17] C.K. Ho, M.T. Itamura, M. Kelley, R.C. Hughes, Review of chemical sensors for L.A. Beardslee received a B.S. and Master of Engineering in Electrical Engineering
in situ monitoring of volatile contaminants, Sandia National Laboratories, Albu- from Cornell University in 2006 and 2007 and a Ph.D. in Electrical Engineering from
querque, Report SAND2001-0643, 2001. the Georgia Institute of Technology in 2011. He is currently pursuing an MD degree
[18] D. Lange, O. Brand, H. Baltes, CMOS Cantilever Sensor Systems: Atomic Force at Albany Medical College in Albany, NY. His research interests include the applica-
Microscopy and Gas Sensing Applications, Springer, Berlin, 2002. tion of micro/nanotechnology to biomedical problems, in particular biosensors and
[19] Y. Li, C. Vancura, D. Barrettino, M. Graf, C. Hagleitner, A. Kummer, M. Zimmer- implantable devices.
mann, K.-U. Kirstein, A. Hierlemann, Monolithic CMOS multi-transducer gas
sensor microsystem for organic and inorganic analytes, Sensors and Actuators S. Truax received the B.S., M.S., and Ph.D. degrees in electrical engineering from
B: Chemical 126 (2007) 431–440. the Georgia Institute of Technology in 2005, 2006, and 2011, respectively. His Ph.D.
[20] L.A. Beardslee, S.M. Heinrich, F. Josse, O. Brand, Thermal excitation and thesis work focused on resonant MEMS for chemical sensing. He joined the Micro
piezoresistive detection of cantilever in-plane resonance modes for sensing and Nanosystems Group at the ETH Zurich in 2011 as a research scientist. His cur-
applications, Journal of Microelectromechanical Systems 19 (2010) 1015–1017. rent research interests include radio frequency nanosystems and nano-structured
[21] J.H. Seo, O. Brand, High Q-factor in-plane-mode resonant microsensor platform sensors.
for gaseous/liquid environment, Journal of Microelectromechanical Systems 17
(2008) 483–493. J.-J. Su was born in Tainan, Taiwan, in 1982. He received B.S. degree in electro-
[22] J.H. Seo, Silicon-based resonant microsensor platform for chemical and biolog- physics from National Chiao-Tung University, Hsinchu, Taiwan, in 2005. He began
ical applications, PhD Thesis, Georgia Institute of Technology, 2007. his graduate study in 2007 at Georgia Institute of Technology, Atlanta, GA where he
[23] P.R. Gray, P.J. Hurst, S.H. Lewis, R.G. Meyer, Analysis and Design of Analog is pursuing Ph.D. degree and received M.S. degree in electrical engineering in 2009.
Integrated Circuits, John Wiley & Sons, Inc., New York, 2001. From 2007 to 2009, he worked with GTAP group at Georgia Institute of Technology
[24] B. Razavi, Design of Analog CMOS Integrated Circuits, McGraw-Hill, New York, in designing power converter IC for energy harvesting applications. He is currently
2001. working toward Ph.D. thesis with Dr. Oliver Brand in analog interface circuits design
[25] P.E. Allen, D.R. Holberg, CMOS Analog Circuit Design, Oxford University Press, for MEMS-based chemical sensors. His research interests include design of mixed-
Inc., New York, 2002. signal IC sensor systems, low power electronic circuits, and development of CMOS
[26] S. Galal, B. Razavi, 10-Gb/s limiting amplifier and laser/modulator driver in compatible chemical sensors.
0.18-␮m CMOS technology, IEEE Journal of Solid-State Circuits 38 (2003)
2138–2146. O. Brand received his diploma degree in Physics from Technical University Karl-
[27] E. Sackinger, W.C. Fischer, A 3-GHz 32-dB CMOS limiting amplifier for SONET sruhe, Germany in 1990, and his Ph.D. degree from ETH Zurich, Switzerland in
OC-48 receivers, IEEE Journal of Solid-State Circuits 35 (2000) 1884–1888. 1994. From 1995 to 1997, he worked as a postdoctoral fellow at Georgia Tech.
[28] L. Seungbae, C.T.C. Nguyen, Influence of automatic level control on microme- From 1997 to 2002, he was a lecturer at ETH Zurich, Switzerland and deputy
chanical resonator oscillator phase noise, in: Proceedings of the 2003 IEEE director of the Physical Electronics Laboratory. In 2003, he joined the Electri-
International Frequency Control Symposium and PDA Exhibition Jointly with cal and Computer Engineering faculty at Georgia Institute of Technology where
the 17th European Frequency and Time Forum, 2003, pp. 341–349. he is currently Professor. He has co-authored more than 150 publications in sci-
[29] L. Yu-Wei, L. Seungbae, L. Sheng-Shian, X. Yuan, R. Zeying, C.T.C. Nguyen, Series- entific journals and conference proceedings. He is a co-editor of the Wiley-VCH
resonant VHF micromechanical resonator reference oscillators, IEEE Journal of book series Advanced Micro and Nanosystems, a member of the editorial board
Solid-State Circuits 39 (2004) 2477–2491. of Sensors and Materials, and has served as General Co-Chair of the 2008 IEEE
[30] H. Lin, X. Yong Ping, M. Palaniapan, A CMOS readout circuit for SOI resonant International Conference on Micro Electro Mechanical Systems (MEMS 2008). Dr.
accelerometer with 4-g bias stability and 20-g/Hz resolution, IEEE Journal of Brand is a senior member of the IEEE and a co-recipient of the 2005 IEEE Donald
Solid-State Circuits 43 (2008) 1480–1490. G. Fink Prize Paper Award. His research interests are in the areas of silicon-based
[31] T. Kugelstadt, Active filter design techniques, in: R. Mancini (Ed.), Op Amps for microsystems, microsensors, MEMS fabrication technologies, and microsystem
Everyone: Design Reference, Elsevier, Burlington, MA, 2003. packaging.