IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 67, NO.

5, MAY 2018 1161

An Efficient Readout Scheme for Simultaneous

Measurement From Multiple Wireless
Passive LC Sensors
Anish Babu and Boby George, Member, IEEE

Abstract— A new readout scheme for multiple passive changes with change in the value of the sensor capacitance.
LC sensors based on the impulse response is presented here. Such sensor systems are particularly useful in applications
A precharged capacitor is discharged through a readout coil where the physical access to the sensing element is limited. To
which is magnetically coupled to multiple sensor coils. Fast
Fourier transform of the resultant oscillatory current through the measure the sensor parameter, a readout circuit is required. The
readout coil is taken, from which the equivalent impedance of the readout circuit consists of a readout coil, which is magnetically
system at different frequencies is computed. The maximum value coupled to the sensor coil. The measurement circuit, which
of the imaginary parts of this impedance occurs at the resonant is connected to the readout coil, measures the resonance fre-
frequencies of the sensor coils that are coupled. The conventional quency f p of the sensor unit from which the sensor parameters
measurement schemes rely on a frequency sweep approach to
determine the resonance frequencies of the sensor coils; that (e.g., change in capacitance) can be calculated.
process is inherently time-consuming. A simple, low-cost readout Phase-dip method is one of the methods used to measure
scheme with a fast update rate is proposed. A method to measure the change in sensor capacitance in a wireless passive LC
the quality factor Q of the sensor coil system is also presented system. In [10], the phase dip of the input impedance at the
here. Output of the proposed scheme is independent of the resonance frequency is used to measure the sensor capacitance.
variation in the coupling factor k. The measurement resolution of
the proposed scheme has been improved using a Gaussian curve To detect the phase dip, the phase at different frequencies
fitting approach, which also reduces the effect of noise in the needs to be measured. The frequency at which the phase is
final output. A prototype of the proposed system has been built minimum is found to deviate from the resonance frequency,
with a readout coil and three sensors coupled to it and tested. when the coupling coefficient k is large. The method proposed
The worst case error observed in the test was less than 0.5%, in [11] uses the real part of the input impedance to detect the
when the prototype system was tested with a variable capacitor
(47 to 345 pF) in the sensor coil. The proposed readout scheme is resonance frequency. The real part of the input impedance is
useful in applications requiring simultaneous wireless monitoring found to be maximum at the resonance frequency, independent
of multiple physical parameters. of the value of k. In [12], the sensor is energized at a set
Index Terms— Fast Fourier transform (FFT), multiple of frequencies within a predetermined frequency range and
measurements, mutual inductance, passive LC sensor, Q factor. the sensor responds with its own magnetic field. From the
frequency, amplitude, and bandwidth of the magnetic field
I. I NTRODUCTION response the sensor parameters are determined. The methods
in [10]–[12] require a frequency sweep to detect the reso-
W IRELESS passive LC sensors are used for non-contact
measurement in several applications such as pH moni-
toring [1], monitoring of concrete structures [2], measurement
nance frequency. The measurement time associated with the
frequency sweep does not allow to achieve fast update rates.
of humidity [3], biopotential measurement [4], temperature [5], Further, for a measurement scheme using frequency sweep,
and strain [6]. They can also be used for measuring multiple a feedback loop is required to automatically track the sensor
physical quantities simultaneously [7], [8]. Simultaneous mea- capacitance, which is quiet complex [11]. Even though the
surement of multiple quantities such as temperature and pH update rate of the frequency sweep based method is low,
allows for temperature compensation in the measured value it offers high signal-to-noise ratio (SNR) and hence a large
of pH, which in turn helps to achieve higher accuracy in dynamic range. A time-domain method that operates on a
the measurement [9]. A passive LC sensor system typically single frequency offers a faster update rate [13].
consists of a sensor coil connected to a capacitive sensor. This In [13], the phase φ of the input current at a frequency in the
forms an LC tank circuit, whose resonance frequency ( f p ) region around the sensor resonance frequency is measured. It
was shown that the tangent of the phase, tan(φ), varies linearly
Manuscript received July 2, 2017; revised September 1, 2017; accepted with the change in sensor capacitance. A real-time capacitance
September 6, 2017. Date of publication November 22, 2017; date of current estimation methodology for batteryless wireless sensor sys-
version April 5, 2018. The Associate Editor coordinating the review process
was Dr. Salvatore Baglio. (Corresponding author: Anish Babu.) tems through cascaded filtering is presented in [14]. In [15],
The authors are with the Department of Electrical Engineering, IIT Madras, the circuit is excited at resonance frequency and the magnitude
Chennai 600036, India (e-mail: [email protected]; [email protected]). of the reader coil current is used to compute the sensor
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. capacitance. The magnitude of the current changes when C2
Digital Object Identifier 10.1109/TIM.2017.2770858 changes, but this current is also a function of the coupling
0018-9456 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
1162 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 67, NO. 5, MAY 2018

coefficient k. It would be useful if a measurement system with

the following features can be developed for wireless passive
LC sensors. The features are: 1) low cost; 2) measurement
time as low as possible; 3) sensitivity of the output to the
variation in k is negligible; 4) hardware requirement is low;
5) ability to read from multiple sensors at the same time;
6) power required to measure is low; and 7) quality factor Q
can be measured in addition to the sensor capacitance.
A new method based on fast Fourier transform (FFT) of the
input current for an impulse input is proposed in [16]. A capac-
itor is initially charged to a predefined voltage. To measure
the sensor capacitance, this capacitor is discharged through Fig. 1. Equivalent circuit of the proposed readout scheme. The mutual
the readout coil. This is equivalent to giving an impulse input inductances M13 between L 1 and L 3 and M14 between L 1 and L 4 are not
to the system. The capacitor discharges quickly, resulting in shown to avoid complexity in the diagram.
an oscillating and exponentially decaying current flow through
the coil. The frequency of oscillation of the current depends
on the natural frequencies of the system, one of which is changed to position-1. This brings C1 in series with L 1 , Rs ,
the sensor parallel resonance frequency. A resistor connected and Z r2 . Here Z r2 is the secondary side impedance of sensor
in series with the readout coil gives a voltage proportional coil-2 expressed as a reflected impedance in the primary side
to the input current. The FFT of the voltage is taken from and is given by (1). The capacitor C1 discharges through L 1 ,
which the input impedance of the system is computed. For Rs , and Z r2 . This results in an oscillatory current flow, which
the proposed system, it has been found that the imaginary part decays exponentially. The oscillation frequencies correspond
of the input impedance is maximum at the sensor resonance to the natural frequencies of the system. The system, when
frequency. The imaginary part of the input impedance for S1 = 1, has two series resonance frequencies and one parallel
all the required frequencies is directly obtained from the resonance frequency [13]. The √ parallel resonance frequency
proposed method. Thus, no frequency sweep is required, f p2 is given by f p2 = 1/[2π L 2 C2 ]
allowing fast update rates. The proposed method can be used to −M122 C s3
2
read multiple LC sensor systems simultaneously. The quality Z r2 (s) = . (1)
factor Q of the sensor can also be measured using the proposed L 2 C 2 s + R2 C 2 s + 1
2

approach. The output of the measurement scheme is indepen- When S1 is at position-1, using Kirchhoff voltage law, we
dent of the coupling coefficient k. The mathematical model can write the following equation, which is written in the
of the proposed measurement scheme and the simulation and Laplace domain for ease of analysis
experimental results of the proposed system are presented as
I1 (s) Vin
follows. − + L 1 s I1 (s) + R1eq I1 (s) + Z r2 (s)I1 (s) = 0.
C1 s s
II. M ATHEMATICAL M ODEL (2)
The proposed wireless readout system can be represented In (2), R1eq is the sum of R1 , Rs , and the ON resistance of
by the equivalent circuit shown in Fig. 1. A capacitor C1 is switch S1 . Equation (2) can be simplified to (3), where Z (s)
connected to a single-pole-double-throw switch S1 as shown is given by (4). The impulse input results in a current flow I1 .
in Fig. 1. One terminal of the switch is connected to a dc The characteristics of the system, Z (s), can be determined
voltage Vin . The other terminal is connected to the readout from the quantity Vin /I1 (s)
coil. The self-inductance and resistance of the readout coil
are indicated as L 1 and R1 , respectively. The readout coil is Vin = Z (s)I1 (s) (3)
magnetically coupled to the sensor coils. The system shown in 2 
 
Fig. 1 has three sensor coils. The first sensor coil has a self- L n Cn s 2 + Rn Cn s + 1 − M12 2 C C s4
1 2
n=1
inductance L 2 and resistance R2 . The capacitive sensor C2 is Z (s) = . (4)
C1 (L 2 C2 s 2 + R2 C2 s + 1)
connected in parallel to the sensor coil. The mutual inductance
M12 between The imaginary part of Z (s), in the frequency domain, can
√ the sensor and the readout coils is given by
M12 = k12 L 1 L 2 , where k12 is the coupling coefficient. be written as
Similarly, L 3 and L 4 indicate the self-inductance of the sensor ω5 M12
2 C2 R
2 2
coils 3 and 4. The corresponding coil resistances are R3 and I m(Z ( j ω)) = R1 ω + . (5)
(1 − L 2 C2 ω )2 + (ω R2 C2 )2
2
R4 , and the capacitive sensors are C3 and C4 in order. The
resistor Rs , which is connected in series with the readout coil, The plot of the imaginary part of Z ( j ω) versus frequency
acts as a current sensing resistor. To explain the operation, is given in Fig. 2. The values of the parameters used for
initially, the coil-2 alone is considered. The other sensor coils this computation are given in Table I. The value of C2 used
are discussed at a later stage. was 300 pF. From Fig. 2, we can see that the imaginary
Initially, switch S1 is kept in position-0 and the capacitor C1 part of Z ( j ω) has a local maxima at the frequency where
is charged to a dc voltage Vin . To measure C2 the switch S1 is 1 − L 2 C2 ω2 = 0, which is the parallel resonance frequency

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
BABU AND GEORGE: EFFICIENT READOUT SCHEME FOR SIMULTANEOUS MEASUREMENT FROM MULTIPLE WIRELESS PASSIVE LC SENSORS 1163

Fig. 2. Imaginary part of the impedance Z ( jω) as a function of frequency. Fig. 3. Output voltage Vo obtained across the resistor Rs . This is obtained
The local maxima occur at the sensor parallel resonance frequency f p2 . from the simulation study.

TABLE I to develop ways to improve and test; 3) assess the effect of

C IRCUIT PARAMETER VALUES noise in Vo , on the FFT output; 4) verify the functionality
and optimize the performance when reading multiple sensors
simultaneously; and 5) test the effectiveness of the proposed Q
measurement method.
of sensor coil-2. Thus, by measuring I m(Z ( j ω)), we can A. Transient Simulation Study
compute the resonance frequency and calculate C2 .
An FFT-based measurement scheme is used to determine A transient response of the system given in Fig. 1 was
I m(Z ( j ω)). In Fig. 1, the voltage Vo is given by Vo = Rs I1 . obtained, for the sensor coil-2 with L 2 and C2 , after imple-
Vo can be decomposed to its different frequency components menting it in a MATLAB Simulink environment. The values
by taking its FFT. The voltage Vo is sampled from the instance of the parameters used for the study are given in Table I.
S1 is changed to position-1. An n-point FFT of Vo is computed These values are taken from the prototype developed to
where n, the number of samples of Vo taken, is a power of 2. experimentally evaluate the proposed scheme. The coupling
The computed FFT consists of n/2 complex terms and each coefficient k was kept at 0.1. The value of C2 used was 330 pF.
term VK can be expressed in the form VK = α K + jβ K . The capacitor C1 was given an initial voltage of 5 V. At time
Here α K is the real part and β K is the imaginary part of the t = 0, the capacitor was discharged through the readout coil
K th term of the computed FFT output. The voltage Vo can be L 1 . The resulting current flow in the circuit was measured by
expressed in terms of Z (s) as sensing the voltage Vo across the resistor Rs . The voltage Vo
observed is given in Fig. 3. The sampling frequency was set
Rs Vin at 40 MHz.
Vo (s) = Rs I1 (s) = . (6)
Z (s) An 8192 point FFT of Vo was computed. As can be seen in
Assuming that Rs Vin = 1, the FFT components of the Fig. 3, the amplitude of voltage Vo decays exponentially with
impedance Z (s) can be computed as time. The FFT windowing function should not attenuate the
α K − jβ K initial portion of the signal. Therefore, a rectangular window
ZK = . (7) was chosen for the FFT computation. From the FFT output,
α 2K + β K2
the Z K was calculated using (7). The resulting imaginary part,
The imaginary part of Z K , denoted as Z i K , at different Z i K , at different frequencies obtained is given in Fig. 4. The
frequencies is given by −β K /(α 2K + β K2 ). Once the FFT is maximum value of Z i K is observed at the frequency very close
computed, the frequency corresponding the maximum value to f p2 . This was repeated for different values of k. The peak
of Z i K is identified and this is taken as f p2 . From f p2 and value of Z i K always occurred at the same frequency indicating
the known value of L 2 , the value of C2 can be computed. that the proposed scheme is independent of the value of the
Sensor coil-2 alone is considered in the above explanation. coupling factor k. The highest frequency component in the
In order to use the same readout coil for multiple sensors as FFT output is equal to half the sampling frequency Fs at which
illustrated in Fig. 1, the resonance frequencies fr3 and fr4 the signal is acquired. The resonance frequency of the sensor
of coil-3 and coil-4, respectively, should be selected in such coil should remain lower than Fs /2. Therefore, the sampling
a way that they are sufficiently apart—such that there is no rate of the analog to digital converter (ADC) determines the
cross coupling. In such a condition the Z K and Z i K for those range of the resonance frequency that can be determined by
coils can be computed as done above for sensor coil-2. The the proposed scheme. In the simulation study, the signal Vo
details are presented in the following section. was sampled at four times the maximum sensor resonance
frequency.
III. N UMERICAL S TUDY AND A NALYSIS
A numerical study was conducted to: 1) verify the mathe- B. Resolution of the Measurement Scheme
matical model and to study the effect of k and its variation; The FFT output Z i K gives the magnitude of the imagi-
2) determine the resolution of the measurement scheme and nary part for discrete frequency values (bins). From this, the

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
1164 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 67, NO. 5, MAY 2018

Fig. 5. Imaginary part of the FFT, Z i K , when noise was added to the input
Fig. 4. Imaginary part, Z i K , of the FFT for different values of coupling signal. The fit two-term Gaussian curve is also shown. The value of f p2
factor k. The peak value of Z i K occurs at the same frequency independent detected with and without curve fit were 5.057 and 5.051 MHz, respectively.
of the value of k. The true value of f p2 was 5.059 MHz.

TABLE II TABLE III

PARAMETER VALUES [3] E FFECT OF N OISE

frequency corresponding to the maximum value is taken

as f p2 . Therefore, the sensor capacitance values computed
using this method will also be discrete in nature. Thus, the
resolution of the measurement system depends on the size of
the frequency bin. Let the sampling frequency be Fs and the
number of samples taken for computing the FFT be Ns . Then,
the FFT output will have Ns /2 frequency bins and the size the measurement scheme. The measurement resolution is now
of each frequency bin is given by fbin = Fs /Ns . For a small limited only by the noise in the signal Vo .
change C2 in C2 , the change in resonance frequency  f p2 is
given by  f p2 = f p2 C2 /2C2 [13]. The frequency bin size D. Effect of Noise
f bin is the minimum change in f p2 that can be detected by the
The real-world signal Vo will have noise due to different
measurement scheme. Therefore, the measurement resolution
sources. Digitization of the signal will also introduce quan-
of C2 is given by 2 f bin C2 / f p2 .
tization noise in the digitized Vo . To study the effect of
Consider a typical wireless passive LC sensor system,
noise on the final output, an appropriate amount of noise was
whose parameter values are shown in Table II. Let the sam-
added to the voltage Vo obtained from the simulation, and
pling frequency be 40 MHz and the number of samples taken
the performance was studied. The ADC used in the prototype
for FFT is 8192. Therefore, the size of each frequency bin,
developed had an effective number of bits (ENOB) of 13.5 b.
f bin , will be 4.88 kHz. For these specifications and sensor
In the simulation study, the signal Vo was first quantized using
parameters, the resolution of the measurement scheme will be
a 13-b quantizer. Then, Gaussian white noise was added to the
0.56 pF (if the other errors are neglected). The total change
signal Vo , to make the SNR of the signal as 76 dB, such that
in f p2 for the full-scale change in C2 is 649 kHz. Therefore,
the standard deviation in the computed value of C2 from the
there will be 649/4.88 = 133 discrete output values over the
simulation study matched with the experimental results. The
full range of sensor C2 . A method to improve the resolution,
two-term curve fitting approach was not used in both these
without additional hardware is presented below.
studies. The FFT components Z i K after the noise was added
to the signal is given in Fig. 5. From Fig. 5, it can be seen
C. Improved Measurement Resolution that there are multiple minor peaks around the region near
As discussed above, the resolution of the proposed mea- f p2 compared to the smooth curve obtained from a noise-free
surement scheme is limited by the size of the frequency signal earlier. Therefore, the frequency bin corresponding to
bin of the FFT. As the shape of the Z i K versus frequency the maximum value of Z i K can be different from actual f p2 .
curve is known (see Fig. 4), the resolution can be improved To detect the peak accurately, a two-term Gaussian curve was
significantly by fitting the Z i K data to a two-term Gaussian fit to the plot of Z i K . As can be seen in Fig. 5, the same
curve as illustrated in Fig. 5. The Z i K data shown in Fig. 5 has approach used to improve the resolution of the measurement
intentionally included noise (the details are discussed below). helps to remove the multiple peaks in Z i K in the region
The fit curve had an R 2 value of 0.998. From this fit curve, around f p2 .
the frequency corresponding to the maximum value of Z i K The above-mentioned approach was simulated under differ-
can be determined at a much higher resolution as the curve ent conditions as given in Table III. The value of C2 was set as
is continuous. This significantly improves the resolution of 180 pF, and under each condition C2 was computed 12 times.

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
BABU AND GEORGE: EFFICIENT READOUT SCHEME FOR SIMULTANEOUS MEASUREMENT FROM MULTIPLE WIRELESS PASSIVE LC SENSORS 1165

When the signal Vo had no noise, the standard deviation in

the measurement of C2 was zero. But when noise was added
to the signal, a standard deviation of 0.65 pF was observed in
the computed value of C2 . The seed for the Gaussian noise
generator was varied each time so that the noise signal is not
repeated in the study. The standard deviation in the computed
C2 reduced significantly when the Gaussian fitting was used.
Further, the performance of the proposed measurement scheme
was not affected when the FFT size was reduced by a factor
of 4. Thus, fitting a two-term Gaussian curve to the plot of
Z i K gave the following advantages: 1) the standard deviation Fig. 6. Imaginary part Z i K of the FFT when the readout coil is coupled to
of the measurement improved which in turn improved the three LC sensors as illustrated in Fig. 1.
measurement resolution and 2) the FFT size can be reduced
without affecting the performance of the measurement scheme.
same can be noticed in the plot of Z i K given in Fig. 6. Here
fr = 6.4 MHz. The plot of Z i K in the region near fr is a
E. Reading Multiple Sensors smooth curve and the measurement scheme gave good results
The numerical study has been extended to the case of when the sensor coil frequency was near fr . This limits the
reading from multiple numbers of passive LC sensors as effective frequency range over which the resonance frequency
illustrated in Fig. 1. Each of the sensor coils has a frequency of the sensor coils can vary, which in turn determines the
range over which its resonance frequency varies as the value number of sensor coils that can be measured simultaneously.
of the sensing parameter changes. There should be a sufficient
gap between the frequency ranges (resonant frequency and F. Measurement of Quality Factor Q
its change) of the sensor coils to avoid any cross coupling.
Simulation studies showed that, to fit a two-term Gaussian In a number of wireless LC sensing applications, the
curve to the plot of Z i K , the values of Z i K over a frequency measurement of the quality factor Q of the sensor coil part
range f p ± 0.15 MHz is required, for the prototype designed. is required [9]. The proposed method can be used to measure
The resonance frequency of sensor coil-3 is f p3 and let the Q factor of the sensor system. The Q factor of a passive
f p3 > f p2 as in Fig. 6. For the readout circuit, to be able LC sensor system with coil-2 (i.e., L 2 , C2 , and R2 ) is given
to measure the entire range of both the sensor coils, there by Q = (1/R2 )(L 2 /C2 )(1/2). Thus, once Q is measured, the
should be a gap of 0.3 MHz between the maximum value of value of the resistive element in the sensor can be computed.
f p2 and the minimum value of f p3 . A resistive sensor, say, a thermistor can be connected in
Three sensor coils were considered for the simulation study. parallel with C2 . This resistor comes in parallel with the
In addition to the parameters given in Table I, C2 = 270 pF, leakage resistance of the sensor capacitor C2 . The effective
C3 = 200 pF, C4 = 150 pF, L 3 = L 4 = 3 μH, and parallel resistance, R p2 , can be expressed as an equivalent
k12 = k13 = k14 = 0.1 were used for the study. To measure series resistance R p2 /(1 + (ω R p2 C2 )2 ). This is in series with
the sensor capacitor values, S1 is changed to position-1. the sensor coil resistance R2 , and can be estimated from the
Quantization noise and Gaussian white noise was added to measured value of Q factor.
the voltage Vo . The SNR value of Vo is now 76 dB. From 1
Z so ( j ω) = (8)
the FFT of Vo , the FFT components Z i K was computed. The (1−L 2 C2 ω2 )+ j ω R2 C2
plot of Z i K when three sensor-coil combinations are placed 1
near the readout coil is shown in Fig. 6. As can be seen from |Z so ( j ω)| =  (9)
(1−L 2 C2 ω )2 +(ω R2 C2 )2
2
the figure, there are three peaks corresponding to the three
ω R2 C 2
resonance frequencies that are easily distinguishable, correctly. I m(Z so ( j ω)) = . (10)
When multiple sensor coils are used, the coupling between (1−L 2 C2 ω2 )2 +(ω R2 C2 )2
the sensor coils slightly shifts the resonance frequency of Consider a second order impedance Z so given by (8). The
the LC sensor. The sensor coils should be placed such that magnitude of Z so can be written as in (9). Fig. 7 shows
the coupling between them is very small. In the simulation the plot of |Z so ( j ω)|. Here f 1 and f 2 are the frequencies
study, for the value of the parameters used (equivalent to the corresponding to the half-power points (−3 dB points). The
prototype), a coupling coefficient of less than 0.02 between bandwidth of the system is given by fbw = f2 − f 1 . The
the sensor coils was found to have negligible effect on the Q factor of the system can be calculated as Q = f p2 / f bw .
output. The imaginary part of Z so can be expressed as in (10).
The L 1 and C1 in the readout circuit acts as an equivalent The denominator term of I m(Z so ( j ω)) is the square of the
bandpass
√ filter with center frequency fr given by fr = 1/ denominator term of Z so given in (9). Therefore, the bode
L 1 C1 . Due of this, in Vo , the magnitude of the frequency plot of I m(Z so ( j ω)), in the logarithmic scale, will be the plot
components near fr is high, while that for the frequency of Z so ( j ω) scaled by a factor of two and then multiplied by
components away from fr is low. Therefore, the effect of the numerator term. The two frequencies f 1 and f 2 will now
noise will be more in the frequency range away from fr . The appear at the −6 dB points in the graph of I m(Z so ( j ω)).

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
1166 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 67, NO. 5, MAY 2018

Fig. 8. Equivalent circuit of the modified readout scheme.

Fig. 7. Magnitude plot of Z so ( jω) and the plot of I m(Z so ( jω)). The value
of f bw computed from both the plots are the same.

Here, it is assumed that the change in the numerator term

of (10) over the frequency range f bw is small. The frequency
corresponding to the peak value remains the same. Therefore,
the Q factor can be measured from the graph of I m(Z so ( j ω)),
Fig. 9. Photograph of the experimental setup.
using the values of f1 and f 2 measured at the −6 dB points.
I m(Z so ( j ω)) can be estimated from the plot of Z i K . To verify
this, a simulation study was conducted. The Q factor of the
LC sensor was varied and from the plot of Z i K , the Q was
measured. The measured value of Q matched with the true
value of Q.

IV. E XPERIMENTAL S ETUP AND R ESULTS

A. Prototype Unit
A prototype of the proposed scheme was developed and
tested experimentally to validate its performance. A planar
coil of dimension 20 mm × 20 mm with 10 turns was used
as the sensor coil. A coil of dimension 70 mm × 60 mm Fig. 10. FFT components Z i K obtained from the experimental setup. The
with 10 turns was selected as the readout coil. Both the coils fit two-term Gaussian curve is also shown. The plot of Z i K for a lower value
were fabricated on a printed circuit board. The inductance and of Q factor is also given.

resistance values of these coils are given in Table I. The switch

S1 in the circuit given in Fig. 1 was realized using MAX4624
IC. The switch parameters had a significant effect on the tested. The sensor coil was kept on the top of the readout coil
transient characteristics of the system. The switch had a charge and the coupling coefficient k was measured as 0.08. Initially
injection of 65 pC, which was comparable to the initial charge the capacitor C was charged to 5 V. Initial value of C2 was
stored in C1 . Therefore, the circuit was modified (functionally 344.7 pF. To measure C2 , S1 was changed to position-1. The
equivalent to the circuit in Fig. 1) as shown in Fig. 8. The voltage Vo was sampled from the instant when S1 was changed
capacitance C was realized by a capacitor of value 100 μF. to position-1. The sampling frequency was set at 40 MHz
This comes in series with C1 and the effective capacitance and 2048 samples of Vo were recorded. A 2048 point FFT of
is close to C1 (assuming that the resistance R is very large the sampled data was computed using MATLAB. From the
compared to the opposition to current flow introduced by L 1 , FFT output, the components Z i K were calculated. The plot
C1 , R1 , Z r , and Rs . A resistor R = 330 k was connected of Z i K versus frequency obtained is given in Fig. 10. After
from the terminal 1 of S1 to ground. This provides a path fitting a two-term Gaussian curve, the frequency corresponding
for the C1 to discharge when S1 is changed to position-0. To to the peak value of Z i K was obtained. This was taken as
digitize the voltage Vo , a 16-b ADC, ADS5560 from Texas the resonance frequency f p2 from which the value of C2 was
Instruments Incorporated, Dallas, TX, USA, was used. A data calculated. For the selected value of sampling frequency and
capture card TSW1405EVM was used to read the data from number of samples, the size of frequency bin is 19.52 kHz.
the ADC through a low-voltage differential signaling interface The resolution of the measurement setup for this set of L 2
and transfer it to a personal computer. A photograph of the and C2 is 1.2 pF. This is the resolution without the fitting
experimental setup is given in Fig. 9. approach. After implementing the Gaussian fit, the resolution
is limited only by the noise in the input signal Vo .
The value of C2 was then varied from 47 to 345 pF.
B. Functionality Test The actual values of C2 were measured using an HP 4724
In this test, the accuracy of the measurement scheme while LCR meter, which has an accuracy of 0.1%. The plot of the
reading the change in capacitance of a sensor capacitor C2 was measured value of C2 versus the actual value of C2 is given

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
BABU AND GEORGE: EFFICIENT READOUT SCHEME FOR SIMULTANEOUS MEASUREMENT FROM MULTIPLE WIRELESS PASSIVE LC SENSORS 1167

Fig. 11. Measured value of C2 connected in the sensor coil versus the actual
value of C2 . Fig. 12. Imaginary part of the FFT, Z i K when reading three LC sensors
simultaneously. The three fit two-term Gaussian curves are also shown.
TABLE IV
TABLE V
E FFECT OF C HANGE IN k
Q UALITY FACTOR M EASUREMENT

in Fig. 11. The worst case error was found to be less than
0.5%. To test the repeatability of the measurement scheme,
each value of C2 was measured 12 times and the maximum
value of standard deviation in the measurement was found to
be 0.24 pF.

C. Effect of Variation in k
The effect of variation in k in the measured output was measured value of C2 was found to match with the actual
studied in this test. A 148.8-pF fixed value capacitor was used value. The worst case error in the measurement of the sensor
as C2 . Initially, the sensor coil was kept on the top of the capacitance was less than 0.5%, similar to the results obtained
readout coil as shown in Fig. 9. The coupling coefficient k was when sensor coil-2 alone was measured.
then varied by changing the position of the sensor coil with
respect to the readout coil. The capacitance C2 was measured E. Measurement of Quality Factor Q
for different values of k and is given in Table IV. As can
The Q factor of the sensor was measured using the method
be seen, the measured value of C2 was not affected when
described in Section III-F. A 344.7-pF capacitor was used as
k was changed from 0.03 to 0.125. Therefore, the proposed
C2 . The coupling factor k was measured as 0.08. A resistor R L
measurement scheme is independent of the value of k.
was connected across C2 , and the value of R L was changed
to vary Q. The plot of Z i K for two different values of Q
D. Measuring Multiple Sensors Simultaneously recorded is given in Fig. 10. The change in Q is clearly
To test the performance of the measurement scheme while visible in Fig. 10. The true value of Q was estimated using
measuring multiple LC sensors, three LC sensors were kept a network analyzer. The measured value of Q and its true
magnetically coupled to the readout coil. Each sensor coil had value for different values of R L is given in Table V. The
a self-inductance of about 3 μH. Three capacitors of values measurement of Q involved the estimation of three frequencies
273.3, 201.7, and 148.8 pF were used as sensor capacitors. The f 1 , f 2 , and f p2 . The errors in the estimation of each of those
readout circuit is the same as the one used in Section IV-B frequencies resulted in additional error in the measurement of
when a single sensor capacitance was measured. As usual, Q compared to the measurement of sensor capacitance where
S1 was changed to position-1, and from the voltage Vo the only single frequency was measured. The Q measurement
FFT components Z i K were computed. The plot of Z i K versus results showed that the scheme is useful, for example, in
the frequency obtained is given in Fig. 12. As can be seen in applications where temperature compensation of the measured
Fig. 12, there are three peaks in the plot of Z i K corresponding quantity is desired [9], [17].
to the three resonance frequencies. As explained earlier, to In the proposed scheme, to measure C2 , S1 is changed to
estimate the resonance frequency with a higher resolution, a position-1 and 2048 samples of Vo are taken. When sampled
two-term Gaussian curve was fit to the section of the plot at 40 MHz, this takes about 50 μs. A 2048 point FFT can be
around the peak value. This was done for all the three resonant computed in about 150 μs [18]. The time taken for the compu-
peaks. The fit curves are shown in Fig. 12. From the fit curve, tation of C2 and the Q factor from the FFT output depends on
the individual resonance frequencies were measured as 5.498, the speed of the processing unit. These computations can take
6.392, and 7.438 MHz, respectively. The measured values of place as a parallel process while the measurement unit samples
the sensor capacitors were 272.5, 202.2, and 149.3 pF. Next, the voltage Vo again. Thus very fast update rates are possible
the value of C2 was varied, keeping C3 and C4 constant. The using the proposed scheme. In the frequency sweep approach,

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.
1168 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 67, NO. 5, MAY 2018

the circuit is excited at one frequency and measurements are [6] M. Bona, M. Serpelloni, E. Sardini, C. O. Lombardo, and B. Andò,
taken. This process is repeated at different frequencies from “Telemetric technique for wireless strain measurement from an inkjet-
printed resistive sensor,” IEEE Trans. Instrum. Meas., vol. 66, no. 4,
which f p is computed. A sudden modification in excitation pp. 583–591, Apr. 2017.
frequency results in a transient response which can influence [7] Q. Y. Ren, L. F. Wang, J. Q. Huang, C. Zhang, and Q. A. Huang,
the measurement of f p . The transient period can be as high “Simultaneous remote sensing of temperature and humidity by LC-type
passive wireless sensors,” J. Microelectromech. Syst., vol. 24, no. 4,
as 150 μs [19]. In such a case, the time taken to measure pp. 1117–1123, Aug. 2015.
C2 = (range of the frequency sweep/frequency change step [8] Q. Tan, T. Luo, T. Wei, J. Liu, L. Lin, and J. Xiong, “A wireless passive
size) × 150 μs. To compute Q factor, three frequencies f 1 , pressure and temperature sensor via a dual LC resonant circuit in harsh
environments,” J. Microelectromech. Syst., vol. 26, no. 2, pp. 351–356,
f 2 , and f p need to be identified for which the range of the Apr. 2017.
frequency sweep has to be extended to include all the three [9] S. Bhadra, D. S. Y. Tan, D. J. Thomson, M. S. Freund, and G. E. Bridges,
frequencies. The time required to read from multiple sensors “A wireless passive sensor for temperature compensated remote pH
monitoring,” IEEE Sensors J., vol. 13, no. 6, pp. 2428–2436, Jun. 2013.
will be much higher. In the proposed scheme, multiple sensors [10] K. Bao et al., “A readout circuit for wireless passive LC sensors and its
can be read in a single measurement cycle. application for gastrointestinal monitoring,” Meas. Sci. Technol., vol. 25,
no. 8, p. 085104, 2014.
V. D ISCUSSION AND C ONCLUSION [11] R. Nopper, R. Has, and L. Reindl, “A wireless sensor readout system—
Circuit concept, simulation, and accuracy,” IEEE Trans. Instrum. Meas.,
A simple, low-cost measurement scheme capable of mea- vol. 60, no. 8, pp. 2976–2983, Aug. 2011.
[12] S. E. Woodard and B. D. Taylor, “Measurement of multiple unrelated
suring from multiple passive LC sensors is presented. Three physical quantities using a single magnetic field response sensor,” Meas.
LC sensors were measured simultaneously using the proposed Sci. Technol., vol. 18, no. 5, pp. 1603–1613, 2007.
method. The change in the capacitance of the sensor connected [13] A. Babu and B. George, “A linear and high sensitive interfacing scheme
for wireless passive LC sensors,” IEEE Sensors J., vol. 16, no. 23,
to the sensor coil can be measured in a single impulse pp. 8608–8616, Dec. 2016.
measurement excitation, thereby reducing the measurement [14] H. S. Kim, S. Sivaramakrishnan, A. S. Sezen, and R. Rajamani, “A novel
time. The readout circuit is simple as no complex feedback real-time capacitance estimation methodology for battery-less wireless
sensor systems,” IEEE Sensors J., vol. 10, no. 10, pp. 1647–1657,
loops are required to track the sensor capacitance continuously Oct. 2010.
compared to measurement schemes that rely on frequency [15] H. Zhang, Y. Hong, B. Ge, T. Liang, and J. Xiong, “A novel readout
sweep approach [11]. The measured output was not affected system for wireless passive pressure sensors,” Photon. Sens., vol. 4,
no. 1, pp. 70–76, 2014.
when the coupling coefficient k was varied by about ±50% [16] A. Babu and B. George, “An FFT based readout scheme for passive LC
from its initial value of 0.08. A method to measure the Q factor sensors,” in Proc. IEEE I2MTC, Turin, Italy, May 2017, pp. 669–673.
of the sensor is also proposed which is useful for temperature [17] M. Rocznik, F. Henrici, and R. Has, “ASIC for a resonant wireless
pressure-sensing system for harsh environments achieving ±2% error
compensated measurement of physical quantities such as pH, between −40 and 150 °C using Q-based temperature compensation,”
humidity, and pressure. A prototype of the proposed scheme in IEEE Int. Solid-State Circuits Conf. (ISSCC) Dig. Tech. Papers,
has been developed and tested. The worst case error was San Francisco, CA, USA, Feb. 2012, pp. 202–204.
[18] M. McKeown, (2013). “FFT implementation on the TMS320VC5505,
found to be less than 0.5%. The excitation circuit sends a TMS320C5505, and TMS320C5515 DSPs,” Texas Instrum.
nearly impulse-like current through the readout coil, with an Dallas, TX, USA. Appl. Rep. SPRABB6B. [Online]. Available:
energy per measurement cycle in the range of 2.5 μJ. The http://www.ti.com/lit/an/sprabb6b/sprabb6b.pdf
[19] R. J. F. de Oliveira and P. J. Abatti, “Analysis of telemetric system based
hardware required in the readout is low as no sweep frequency on remote resonant sensing circuit,” Electron. Lett., vol. 42, no. 13,
source is required, unlike the conventional schemes. The total pp. 750–752, Jun. 2006.
energy requirement can be kept low by using low power
ADCs and FFT computing processors. A 2048 point FFT can
be completed in about 150 μs consuming about 6 μJ [18]. Anish Babu was born in Kottayam, India, in 1987.
The transient current which flows through the readout coil He received the B.Tech. degree in electrical and
dies down within tens of microseconds and the update rate is electronics engineering from the National Institute of
Technology, Calicut, India, in 2010, and the M.Tech.
limited only by the FFT computation time. degree in electrical engineering from IIT Madras,
Chennai, India, in 2014, where he is currently pursu-
R EFERENCES ing the Ph.D. degree in sensors and instrumentation.
His current research interests include sensors,
[1] S. Bhadra et al., “Fluid embeddable coupled coil sensor for wireless pH measurements, and instrumentation.
monitoring in a bioreactor,” IEEE Trans. Instrum. Meas., vol. 63, no. 5,
pp. 1337–1346, May 2014.
[2] K. Perveen, G. E. Bridges, S. Bhadra, and D. J. Thomson, “Corrosion
potential sensor for remote monitoring of civil structure based on printed
circuit board sensor,” IEEE Trans. Instrum. Meas., vol. 63, no. 10,
pp. 2422–2431, Oct. 2014. Boby George (M’07) was born in Kannur, India, in
[3] T. J. Harpster, S. Hauvespre, M. R. Dokmeci, and K. Najafi, “A passive 1977. He received the M.Tech. and Ph.D. degrees
humidity monitoring system for in situ remote wireless testing of in electrical engineering from IIT Madras, Chennai,
micropackages,” J. Microelectromech. Syst., vol. 11, no. 1, pp. 61–67, India, in 2003 and 2007, respectively.
Feb. 2002. He was a Post-Doctoral Fellow with the Institute
[4] J. Riistama, E. Aittokallio, J. Verho, and J. Lekkala, “Totally passive of Electrical Measurement and Measurement Sig-
wireless biopotential measurement sensor by utilizing inductively cou- nal Processing, Technical University of Graz, Graz,
pled resonance circuits,” Sens. Actuators A, Phys., vol. 157, no. 2, Austria, from 2007 to 2010. He joined the faculty
pp. 313–321, 2010. of the Department of Electrical Engineering, IIT
[5] D. Marioli, E. Sardini, and M. Serpelloni, “Passive hybrid MEMS Madras, in 2010, where he is currently an Asso-
for high-temperature telemetric measurements,” IEEE Trans. Instrum. ciate Professor. His current research interests include
Meas., vol. 59, no. 5, pp. 1353–1361, May 2010. measurements, sensors, and instrumentation.

Authorized licensed use limited to: Indian Institute of Technology Palakkad. Downloaded on June 21,2024 at 16:22:04 UTC from IEEE Xplore. Restrictions apply.