sensors

Article
Development and Measurements of a Mid-Infrared
Multi-Gas Sensor System for CO, CO2 and CH4
Detection
Ming Dong 1 , Chuantao Zheng 1, *, Shuzhuo Miao 1 , Yu Zhang 1 , Qiaoling Du 1 , Yiding Wang 1
and Frank K. Tittel 2 ID
1 State Key Laboratory on Integrated Optoelectronics, College of Electronic Science and Engineering,
Jilin University, Changchun 130012, China; 1988dm@sina.com (M.D.); miaosz16@163.com (S.M.);
yuzhang@jlu.edu.cn (Y.Z.); duql@jlu.edu.cn (Q.D.); ydwang@jlu.edu.cn (Y.W.)
2 Electrical and Computer Engineering Department, Rice University, 6100 Main Street, Houston, TX 77005,
USA; fkt@rice.edu
* Correspondence: zhengchuantao@jlu.edu.cn; Tel.: +86-137-5609-0979

Received: 19 August 2017; Accepted: 25 September 2017; Published: 27 September 2017

Abstract: A multi-gas sensor system was developed that uses a single broadband light source and
multiple carbon monoxide (CO), carbon dioxide (CO2 ) and methane (CH4 ) pyroelectric detectors by
use of the time division multiplexing (TDM) technique. A stepper motor-based rotating system and a
single-reflection spherical optical mirror were designed and adopted to realize and enhance multi-gas
detection. Detailed measurements under static detection mode (without rotation) and dynamic mode
(with rotation) were performed to study the performance of the sensor system for the three gas
species. Effects of the motor rotating period on sensor performances were also investigated and a
rotation speed of 0.4π rad/s was required to obtain a stable sensing performance, corresponding to
a detection period of ~10 s to realize one round of detection. Based on an Allan deviation analysis,
the 1σ detection limits under static operation are 2.96, 4.54 and 2.84 parts per million in volume
(ppmv) for CO, CO2 and CH4 , respectively and the 1σ detection limits under dynamic operations are
8.83, 8.69 and 10.29 ppmv for the three gas species, respectively. The reported sensor has potential
applications in various fields requiring CO, CO2 and CH4 detection such as in coal mines.

Keywords: mid-infrared absorption spectroscopy; multi-gas detection; gas sensor

1. Introduction
The development of multi-gas detection technology plays an important role in many areas, such
as environmental atmospheric monitoring [1,2], medical diagnosis [3,4], industrial process control [5,6],
gas monitoring and fire alarm systems in coal mines [7–10]. So far, the most widely used technology
for multi-gas instruments is photo-acoustic spectroscopy (PAS) and catalytic reaction. PAS detects
gas concentration by acquiring acoustic signals in the gas chamber based on the photoacoustic
principle [11–13]. However, environment vibrations will interfere with normal sensor operation
and cause measurement errors of the PAS sensor. Additionally, a multi-gas sensor based on a catalytic
reaction has a much shorter service lifetime compared to infrared (IR) sensors [14–16]. Direct absorption
spectroscopy (DAS), however, has the advantage of a compact design [17]. In addition, a broadband
thermal source and pyroelectric detector in a DAS technique are less costly compared to a mid-infrared
laser and detector.
There are many ways to realize a multi-gas detector, which can be divided into three types: multiple
light sources with a single detector, single broadband light source with multiple detectors, and a
single tunable light source with a single detector. For example, Besson [18] demonstrated a multi-gas

Sensors 2017, 17, 2221; doi:10.3390/s17102221 www.mdpi.com/journal/sensors

Sensors 2017, 17, 2221 2 of 14

sensor using three near-infrared distributed feedback lasers (DFB) and a resonant photoacoustic cell.
Mukherjee [19] designed a non-dispersive optical multiplexer using a scanning galvanometer for
multispecies trace-gas detection, where five quantum cascade lasers (QCLs) were used, using time
division multiplexing (TDM). Tan [20] integrated four single-channel pyroelectric sensors in a miniature
optical gas chamber based on a non-dispersive infrared (NDIR) technique. Mao [21] reported an
all-optical photoacoustic spectrometer for multi-gas analysis based on a tunable erbium-doped fiber
laser and a fiber acoustic sensor. As the emission spectrum of a single laser cannot cover a wide
wavelength band, multiple lasers have to be used to achieve multi-gas detection. The lasers used in
tunable diode laser absorption spectroscopy (TDLAS) and PAS techniques are generally costlier than
an infrared thermal source or light emitting diode (LED).
The developed sensor is intended to be used in the mining industry, where the normal CO, CO2
and CH4 concentration are less than several ppm, hundreds to thousands of ppm and thousands of
ppm, respectively. So, the detection limits of CO, CO2 and CH4 are designed to be <10 ppm, <100 ppm
and <100 ppm, respectively, to meet the requirements of mining detection.
In order to reduce sensor cost, a single broadband light source and multiple detectors were used
in this work to realize a multi-gas sensor based on DAS. The emission spectrum of the wideband light
source covers the mid-infrared absorption band of CO (4.65 µm), CO2 (4.26 µm), CH4 (3.31 µm), which
are the fundamental absorption bands of these three gas species. A mechanical rotation module was
designed to switch between the three gas detectors and three gases were detected in sequence using
the time division multiplexing (TDM) technique. Both static (without rotation) and dynamic (with
rotation) detection characteristics were studied. The structure of this paper is organized as follows:
the sensor structure will first be described followed by detection theory, optical and mechanical design,
and the closed-loop control theory of the mechanical rotation module; sensor validation as well as
measurement results of the detection limit, response time and stability will then be proposed. Finally,
a discussion and conclusion section will be presented.

2. Multi-Gas Sensor Structure and Design

2.1. Sensor Structure

As shown in Figure 1, the proposed multi-gas sensor system can be divided into optical,
mechanical and electrical parts. In the optical part, the light emitted by a broadband light source
(IR-55, Boston Electronics, Brookline, MA, USA) is reflected by a spherical mirror with a focal length of
100 mm and converges at a pyroelectric detector. The detector has a reference channel and a detection
channel with two different filters installed in the window assembly. The detector generates a detection
signal and a reference signal. The three dual-channel pyroelectric detectors (Infra Tec GmbH, Dresden,
Germany) with the different wavelengths for CO, CO2 , CH4 sensing are mounted on a disc with an
interval angle of 45◦ . The disc is driven by a stepper motor. When a detector moves to a focal point,
it will remain at this point for the time required for gas detection.
The sensor system switches the three detectors continuously to detect the three gases, which can
then be processed by the TDM technique. In order to improve the accuracy and stability of the rotation
system, a 1/8 subdivided motor driver and an incremental encoder were used for feedback of the
current angle of the stepper motor to form a closed-loop rotation control system. In the electrical part,
a digital signal processor (DSP) reads the current angle from the encoder by means of an enhanced
quadrature encoder pulse module (eQEP).
A closed-loop algorithm was used to precisely control the rotation and switching of the three
detectors. A 4-Hz square-wave signal generated by the DSP by means of an enhanced capture module
(eCAP) was used to modulate the broadband light source through a constant current circuit. An analog
switch was then used to switch the sensing channels of the three detectors. When a detector starts to
operate, it will generate both a detection signal u1 (λ) and a reference signal u2 (λ). These two signals
are first subjected to an impedance match by a preamplifier (PA), which is followed by a differential
Sensors 2017, 17, 2221 3 of 14

operation (DA) for generating a differential signal, i.e., ∆u(t) = u2 (t) − u1 (t). The amplitudes of u2 (t)
∆u(t)2017,
andSensors are 17, 2221
then extracted by a custom-made lock-in amplifier. The light source modulation 3signal of 14
is
used as a reference signal for the lock-in amplifier. The lock-in amplifier will produce a square wave
signal is used as a reference signal for the lock-in amplifier. The lock-in amplifier will produce a
which has the same frequency as the reference signal.
square wave which has the same frequency as the reference signal.
TheThe
square
squarewavewave is is
phase-shifted
phase-shiftedwith
with respect
respect toto the
thereference
referencesignal
signalbybythethe DSP.
DSP. The The signal
signal to to
be measured
be measured and thethe
and phase-shifted
phase-shiftedsignal
signalare
are then
then multiplied
multiplied by byanananalog
analogmultiplier
multiplier after
after passing
passing
through a sampling/holding circuit (S/H) and the output of the multiplier is then
through a sampling/holding circuit (S/H) and the output of the multiplier is then fed back to the DSP fed back to the
DSPforfor sampling.
sampling. TheTheDSPDSP adjusts
adjusts the phase
the phase difference
difference between between the reference
the reference and the phase-shift
and the phase-shift signals
signals according
according to thetofeedback
the feedback
and theand the optimum
optimum phase
phase shift shift obtained
obtained with a maximized
with a maximized feedback
feedback signal.
The The
signal. output of the
output of lock-in amplifier
the lock-in variesvaries
amplifier linearly with the
linearly withamplitude of the of
the amplitude sensing signal signal
the sensing to be to
measured. The maximized feedback signal is processed with a sliding average filtering
be measured. The maximized feedback signal is processed with a sliding average filtering algorithm algorithm and
finally the results are transmitted to a laptop for real-time monitoring.
and finally the results are transmitted to a laptop for real-time monitoring.

Figure
Figure 1. 1.Schematic
Schematic of
of the
the multi-gas
multi-gassensor system
sensor using
system a broadband
using light source
a broadband and a three
light source anddual-
a three
channel pyroelectric
dual-channel detector
pyroelectric system.
detector system.

2.2. 2.2. Detection

Detection Theory
Theory andand Consideration
Consideration
The Beer–Lambert law in Equation (1) shows the relationship between gas concentration C and
The Beer–Lambert law in Equation (1) shows the relationship between gas concentration C and
the received light intensity I, as
the received light intensity I, as
− KK( λ()λ
I (λI)(λ=) =I0I(0 (λλ))ee− CL)CL
(1) (1)

where I0 is
where emission
I0 is intensity, KKisisa amolecule
emissionintensity, molecule absorption
absorption coefficient
coefficient and Land L is optical
is optical pathWith
path length. length.
modulation, the time-dependent light intensity I(λ ) at λ = 4.65, 4.26, 3.31 µm
With modulation, the time-dependent light intensity I(λ1 ) at λ1 = 4.65, 4.26, 3.31 µm (corresponding
1 1 (corresponding to CO,
CO2CO
to CO, , CH24,)CH
and4 the time-dependent
) and light intensity
the time-dependent I(λ2) at λ2 =I(λ
light intensity 3.95
2 ) µm
at λwill be converted
2 = 3.95 µm willtobetwo voltage to
converted
twosignals,
voltagewhose
signals,amplitudes are U1 and are
whose amplitudes U2, U
respectively.
1 and U 2 , The final
respectively. gas concentration
The final gas can be expressed
concentration can be
by Equation (2)
expressed by Equation (2)
11 ∆U
C= C =−−KL ln (1 −−ΔU ) )
ln(1 (2)
(2)
KL UU2 2
where ∆U is a differential signal, i.e., ∆U = U1 − U2 . The signal-to-noise-ratio (SNR) of the sensor
where ΔU is a differential signal, i.e., ΔU = U1 − U2. The signal-to-noise-ratio (SNR) of the sensor
system can be improved by differential operation. Moreover, the light intensity change caused by
system can be improved by differential operation. Moreover, the light intensity change caused by
interference of dust on the mirrors and source current variations will result in changes of ∆U and U2
interference of dust on the mirrors and source current variations will result in changes of ΔU and U2
in ain
similar way,
a similar so so
way, that thethe
that ratio ofof∆U/U
ratio can be
ΔU/U22 can behelpful
helpfulininreducing
reducing the
the impact
impact of interferences.
of interferences.
Since the selected wavelength bands (3.31 µm for CH
Since the selected wavelength bands (3.31 µm for CH4, 4.25 4 , 4.25 for CO , 4.65
µm for CO2, 4.652 µm forµm
µm CO)forforCO)
the for
the three
threegases
gasesare
arefundamental
fundamental bands,
bands, there
there is interference
is no no interference
among among the target
the three three target gas species.
gas species. So,
Sensors 2017, 17, 2221 4 of 14

So, during gas detection, the main interference may come from H2 O, since H2 O has a lot of absorption
lines within the infrared wavelength band.
The developed sensor is intended to be used in CH4 detection in the mining industry, where
the normal CH4 concentration is several thousands of ppm. With reference to the high-resolution
transmission (HITRAN) molecular absorption database, one can get the detailed absorption spectra of
the 1000 ppm CH4 and the 2% H2 O (relative humidity is ~80% at 25 ◦ C) within the spectral range of
the two filter windows of the dual-channel CH4 detector (one filter was centered at 3.31 µm with a
half pass bandwidth of 167 nm, another filter was centered at 3.957 µm with a half pass bandwidth
of 91 nm). The 3.31 µm window covers the strong absorption lines of CH4 , which are helpful for
enhancing sensitivity. The CH4 absorption within the range of 3.9–4.0 µm is at least three times
smaller than the absorption at 3.31 µm. So, the 3.95-µm channel can be used as a reference channel for
background elimination. Under 1 atm and 298 K, the absorption of H2 O is at least 2–3 orders smaller
than CH4 absorption within the two sensing windows, so it has almost no effect on CH4 detection.
The normal CO2 concentration is around several hundreds of ppm in the mining industry.
With reference to the HITRAN database, one can get the detailed absorption spectra of the 300 ppm
CO2 and the 2% H2 O within the spectral range of the two filter windows of the dual-channel CO2
detector (one filter was centered at 4.26 µm with a half pass bandwidth of 198 nm, another filter
was centered at 3.957 µm with a half pass bandwidth of 91 nm). The absorption of H2 O is at least 7
orders smaller than CO2 absorption within the two sensing windows, so it has almost no effect on
CO2 detection.
The CO level, on the other hand, is ~10 ppm in the mining industry. Based on the detailed
absorption spectra of the 10 ppm CO and the 2% H2 O within the spectral range of the two filter
windows of the dual-channel CO detector (one filter was centered at 4.65 µm with a half pass bandwidth
of 175 nm, another filter was centered at 3.957 µm with a half pass bandwidth of 91 nm), the absorption
of H2 O is a bit smaller than CO absorption within the two sensing windows. So, some measures,
e.g., using driers, should be adopted to dry the air before entering the gas cell.

2.3. Optical and Mechanical Design

Figure 2a shows a photograph of the optical part of the multi-gas sensor, where the blue line
indicates the light path. Figure 2b shows the Autodesk Computer Aided Design (CAD) model of
the optical support frame. A software tool named unigraphics (UG) was used for the CAD design.
The support frame and the disc were fabricated using 3-D printing technology.
As shown in Figure 2a, one of the outermost rays (AB) from the light source (A) is parallel to the
axis OC and the reflected light (BE) will pass through the focus (P). Next, the other outermost ray
(AD) from the light source passes through the focus (P) and the reflected light (DE) will be parallel
to the axis OC, BE and DE intersect at point E at which the detector is rotated. Points A and E are
symmetrical about the axis OC. The focal length of the spherical mirror is 100 mm and the specific
position parameters can be calculated by the following formula:

◦
 SP = AP · cos ∠ BAP = f · cos 15 = 96.6

SO = SP + PO = 96.6 + 100 = 196.6 (3)
 AS = AP · sin ∠ APC = f · sin 15◦ = 25.9


Finally, we obtain the coordinate of point A (196.6, −25.9), and the coordinate of point
E (196.6, 25.9) relative to point O (0, 0), which is the design parameters of the light path. The optical
path length is calculated to be 393.2 mm, i.e., twice the length of line SO.
The disc is fixed on the front axle of the stepper motor and an incremental encoder is attached
on the rear axle of the stepper motor to monitor the rotation. Two pairs of photodiodes are placed
on both sides of the motor. One pair is used to set the zero point of the rotation and the other pair is
used to stop the motor. The light source, detector and spherical mirror are placed according to the
Sensors 2017, 17, 2221 5 of 14

optical design parameters. The optical structure is held by an L-shape aluminum plate. A heat sink is
Sensors2017,
Sensors 2017,17,
17,2221
2221 55of
of14
14
firmly attached to the surface of the light source and fixed to the plate using thermal silica gel, which
stabilizes the temperature
Figure 3ashows
showsthe of
theCAD
CADthemodel
light source
ofthe andfor
thedisc
disc suppresses
fixingthe thermal
thethree noise. As shown in Figure 3b,
threedetectors.
detectors.
Figure 3a model of for fixing As shown in Figure 3b,
Figure
theCO,
CO,CO 3a
CO22andshows
andCH the CAD
CH44detectors model
detectorswere of
werefixed the
fixedwith disc
withan for fixing
anelectric the
electriccircuit three
circuitboard.detectors.
board.The As shown
Thefunction
function ofthe
theincircuit
Figure 3b,
the of circuit isis
thetoCO, CO
switchthe and CH
2the sensing 4 channels of the detectors and to filter and pre-amplify the sensing signals. circuit
detectors were fixed with an electric circuit board. The function of the The
to switch sensing channels of the detectors and to filter and pre-amplify the sensing signals. The
is to switch
curved hole theonsensing
the discchannels
allows of the
the opticaldetectors
signal and
to to filterbetween
transmit and pre-amplify
photodiodes. the The
sensing signals.
operating
curved hole on the disc allows the optical signal to transmit between photodiodes. The operating
Thesteps
curved holedisc
of the
the on thearedisc allows
follows.the
as follows. optical signal
Initially, the disc
discto transmit between photodiodes.
rotates clockwise
clockwise until the The operating
the receiver
receiver of the
the
steps of disc are as Initially, the rotates until of
steps of the disc are as follows. Initially, the disc rotates clockwise until
photodiodes acquires an optical signal through the hole in the disc. This position of the disc isis
photodiodes acquires an optical signal through the hole in the disc. the receiver
This of
position the
of photodiodes
the disc
acquires
consideredan optical
considered tobe
to bethesignal
the zerothrough
zero pointof
point the hole The
ofrotation.
rotation. in
Thethe disc.
disc
disc Thisθ°
rotates
rotates position
θ° clockwise
clockwise of so
the disc
sothat
that isCO
the
the considered
COsensor to be the
sensorreaches
reaches
zerothe focal
point of point of
rotation. the
The mirror
disc (point
rotates θ ◦E clockwise
in Figure so2a) for
that gas
the COdetection.
sensor The
reaches angle
the θ
focalis obtained
point of the
the focal point of the mirror (point E in Figure 2a) for gas detection. The angle θ is obtained
experimentally.
mirror (point E in The
experimentally. The
Figure sensors
2a) for
sensors are
aregas arranged
detection.
arranged at
at anan
Theangular interval
angle interval
angular of
θ is obtained 45° and the
of 45°experimentally.disc is rotated
The sensors
and the disc is rotated twice
twiceare
45°
arrangedcounterclockwise
at an angular to let the
interval CO
of 45 ◦ and the
2 sensor and CH
disc sensor
4is rotatedreach the
twice 45 ◦ counterclockwise
light focal point for gasto detection,
let the CO2
45° counterclockwise to let the CO2 sensor and CH4 sensor reach the light focal point for gas detection,
respectively.
sensor Finally, the disc turns clockwise back to the zero position
respectively. 4Finally, the disc turns clockwise back to the zero position and starts a new round turns
and CH sensor reach the light focal point for gas detection, and
respectively. starts a new
Finally, the round
disc of
of
multi-gas
clockwise
multi-gasback detection.
to the zero position and starts a new round of multi-gas detection.
detection.

Figure
Figure 2.2.2.
Figure (a)(a)
(a) Photographof
Photograph
Photograph ofthe
of theoptical
the opticalpart
optical partof
part the
ofthe multi-gas
themulti-gas sensor;
multi-gassensor;
sensor;(b)
(b)
(b) Computer
Computer
Computer Aided
Aided
Aided Design
Design
Design (CAD)
(CAD)
(CAD)
model
model of of
the the optical
optical support
support frame.
frame.
model of the optical support frame.

Figure3.3.(a)
Figure (a)CAD
CADmodel
modelofofthe
thedisc
discfor
formounting
mountingthree
threedetectors;
detectors;(b)
(b)Photograph
Photographof ofthe
thedisc
discwith
withaa
Figure
sealed (a) CAD
3. circuit model
board and of thepyroelectric
three disc for mounting three
detectors for detectors;
the detection(b)
of Photograph
CO, CO of the disc with a
2 and CH4.
sealed circuit board and three pyroelectric detectors for the detection of CO, CO2 and CH4.
sealed circuit board and three pyroelectric detectors for the detection of CO, CO2 and CH4 .
Sensors 2017, 17, 2221 6 of 14

2.4. Closed-Loop Rotation Control

2.4. Closed-Loop Rotation Control
A stepper motor, which converts an electrical pulse signal into an angular displacement, was
used Atostepper
controlmotor, which converts
the rotation an electrical
of the disc. Since the pulse signal displacement
angular into an angularofdisplacement,
the stepper was motorusedis
to control the rotation of the disc. Since the angular displacement of the
proportional to the number of pulses and synchronized in time, the rotation angle of the motor can stepper motor is proportional
to the
be number
precisely of pulses and
controlled. synchronized
However, in time,
if the pulse the rotation
frequency is tooangle
high,ofthethemotor
motor cannot
can be precisely
correctly
respond to the change of a pulse, resulting in a difference between the actual position and theto
controlled. However, if the pulse frequency is too high, the motor cannot correctly respond the
ideal
change
position. of a pulse, resulting in a difference between the actual position and the ideal position.
Furthermore, as
Furthermore, asthe
themotor
motorswitches
switches between
between dynamic
dynamic andandstatic state,
static motor
state, control
motor may be
control maymorebe
difficult to achieve due to system inertia caused by an overshoot, especially
more difficult to achieve due to system inertia caused by an overshoot, especially when the motor when the motor suddenly
stops at high
suddenly stopsspeed.
at highIf the
speed.motor rotates
If the motorwith low with
rotates frequency, it will cause
low frequency, vibrations.
it will In order In
cause vibrations. to
overcome
order these problems,
to overcome a 1/8 subdivision
these problems, motor driving
a 1/8 subdivision circuit was
motor driving used.
circuit wasThe subdivision
used. driving
The subdivision
driving method results in a smooth driving current by allowing the phase current to step upvalue
method results in a smooth driving current by allowing the phase current to step up to a rated to a
or to zero.
rated valueAorreduction
to zero. Aof the current
reduction variation
of the currentrate leads to
variation a reduction
rate of a low-frequency
leads to a reduction vibration
of a low-frequency
caused bycaused
vibration an overshoot and improves
by an overshoot the rotation
and improves stability.stability.
the rotation The angular rotation
The angular resolution
rotation of the
resolution
stepper motor was also increased from 1.8 to 0.225 ◦ . As shown in Figure 4, the angle deviation e(t)
of the stepper motor was also increased from 1.8 to 0.225°. As shown in Figure 4, the angle deviation
is the
e(t) difference
is the differencebetween
between thethe
given
givenangle
angler(t)r(t)
andand thethefeedback
feedbackangle angley(t).
y(t). The
The control u(t)
output u(t)
control output
adjusted according to e(t). y(t)
is adjusted according to e(t). y(t) was measured by a photoelectric incremental encoder to form aa
is was measured by a photoelectric incremental encoder to form
closed-loop control.
closed-loop control. TheThe angle
angle value
value was
was decoded
decoded from from two two orthogonal
orthogonal squaresquare waves
waves generated
generated by by
the encoder
the encoder using
using the
the eQEP
eQEP module.
module. A A pulse
pulse signal
signal waswas generated
generated by by the
the eCAP
eCAP module
module and and the
the DSP
DSP
was also isolated from the motor by an opto-electrical coupler to increase
was also isolated from the motor by an opto-electrical coupler to increase the rotation stability. the rotation stability.

Figure
Figure 4.
4. Schematic
Schematic of
of the
the closed-loop
closed-loop control
control on
on the
the motor
motor rotation.
rotation.

3. Experimental Details and Results

3. Experimental Details and Results

3.1.
3.1. Gas
Gas Sample
Sample Preparation
Preparation
Figure
Figure 55 shows
showsthetheexperimental
experimentalsetup
setupofofthethe multi-gas
multi-gas sensor,
sensor, which
which waswas placed
placed into into an air-
an air-tight
tight chamber with a volume of 15,808 mL. In order to obtain stable and accurate
chamber with a volume of 15,808 mL. In order to obtain stable and accurate calibration results, calibration results,
aa dynamic
dynamic gas gas distribution
distribution method
method was was used
used instead
instead ofof aa static
static injection
injection distribution
distribution via
via aa needle.
needle.
Since N
Since N2 has no absorption in the infrared band, it was used as a carrier gas to mix with CO, CO22 and
2 has no absorption in the infrared band, it was used as a carrier gas to mix with CO, CO and
CH
CH44 to
to get
get different
different gas
gas samples
samples for
for sensor
sensor calibration.
calibration. ItIt is
is possible
possible to to use
use the
the sensor
sensor in
in air
air for
for various
various
applications,
applications, since
sinceother
othergases
gaseshave
havenonointerference
interference toto
thethedetection
detection of the three
of the target
three gasgas
target species. By
species.
mixing a 5000 parts per million in volume (ppmv) CO, CO
By mixing a 5000 parts per million in volume (ppmv) CO, CO2 , CH4 sample with a 99.999% pure N22,,
2 , CH 4 sample with a 99.999% pure N
respectively, gassamples
respectively, gas sampleswith
withdifferent
different concentration
concentration levels
levels were were prepared
prepared by means
by means of mass
of two two mass
flow
flow
meters (Horiba Metron, S49-33/MT, Beijing, China, with a 2% uncertainty). The preparation rulerule
meters (Horiba Metron, S49-33/MT, Beijing, China, with a 2% uncertainty). The preparation can
can be expressed
be expressed by: by:
V × 5000 ppmv
C = V11 × 5000 ppmv (4)
C= V1 + V2 (4)
V1 + V2
Sensors 2017, 17, 2221 7 of 14
Sensors 2017, 17, 2221 7 of 14

In Equation
In Equation(4), CC
(4), is isthe
therequired
requiredgas
gasconcentration
concentration in ppmv andVV1 1and
ppmv and andV2Vare
2 are
thethe required
required gasgas
flow
flow rates
rates of of
thethe standard
standard gassample
gas sampleand
andpure N22, respectively.
pure N respectively.

Figure
Figure 5. Measurementsetup
5. Measurement setupdepicting
depicting the
the electrical
electrical and
and optical
opticalparts.
parts.The
Theoptical
opticalpart was
part placed
was placed
inside an enclosure.
inside an enclosure.

3.2.3.2. Static Sensor Calibration

Static Sensor Calibration
For calibration, the motor was kept in a static state for only single-gas detection, i.e., the multi-
For calibration, the motor was kept in a static state for only single-gas detection, i.e., the multi-gas
gas sensor was being operated without rotation. Three groups of gas experiments were performed
sensor was being operated without rotation. Three groups of gas experiments were performed
with three kinds of gas samples. In each experimental group, a series of gas samples with different
with three kinds of gas samples. In each experimental group, a series of gas samples with different
concentration levels spaced apart by 100 ppmv were prepared and flushed through the chamber. The
concentration
amplitude oflevels spaced apart
the differential by 100
signal (ΔU)ppmv were
and the preparedfrom
amplitude and reference
flushed through
channel the
(U2)chamber.
were
Therecorded
amplitude of the differential signal (∆U) and the amplitude from reference channel
after sensor readings were stabilized. The measurement results of ΔU and U2 for a 2time (U ) were
recorded after sensor readings were stabilized. The measurement results of ∆U
period of 8 min for each concentration were recorded. In order to eliminate the influence2 of noiseand U for a time
period
fromofthe8 min
lightfor each and
source concentration
the detector were recorded.
of the Insystem,
detection order towe eliminate
obtainedthe influence ofbetween
a relationship noise from
theΔU/U
light 2source
and gasand the detectorC.
concentration of For
the detection
CO, CO2 andsystem,
CH4 we obtained
sensing, the aobtained
relationship between
relationship ∆U/U2
curves
andbetween
gas concentration
the averaged C. ΔU/U
For CO, COthe
2 and 2 and CH sensing,
concentration
4 C the
are obtained
shown in relationship
Figure 6a–c, curves between
respectively. Thethe
averaged ∆U/U2 and the concentration C are shown in Figure 6a–c, respectively. The fitting equations
fitting equations of CO, CO 2 and CH 4 are given by Equations (5)–(7), respectively, as
of CO, CO2 and CH4 are given by Equations (5)–(7), respectively,
0.47 − ΔU / U as
C = −755.52 ln( 2
) (5)
0.43
0.47 − ∆U/U 2
C = −755.52 ln( ) (5)
1.51 −0.43
ΔU / U 2
C = −912.67 ln( ) (6)
1.42
1.51 − ∆U/U2
C = −912.67 ln( ) (6)
0.96 −1.42
ΔU / U2
C = −1475.08 ln( ) (7)
− ∆U/U2
0.960.77
C = −1475.08 ln( ) (7)
0.77
3.3. Effects of Rotation Speed on Sensor Performance
3.3. Effects of Rotation Speed on Sensor Performance
Experiments were carried out to determine the impact of rotation speed on the detection stability
ofExperiments
the multi-gaswere carried
sensor out to
operating in determine the impact
a dynamic state of rotation
with different speed
speed CH4on the detection
concentration stability
results.
of the multi-gas sensor operating in a dynamic state with different speed CH 4 concentration
Figure 7a,b show the measured CH4 concentration and the deviation at different rotation speeds, results.
Figure 7a,b show the measured CH
where for each speed, the measurements 4 concentration and the deviation at different rotation speeds,
lasted for 100 s. The standard deviation continues to increase
where
afterfor each
0.4π speed,
rad/s, whichthemeans
measurements lasted for control
that the closed-loop 100 s. The standard
system deviation
becomes as stablecontinues to when
as possible, increase
after
the0.4π rad/s,
rotation which
speed meansrad/s.
is <0.4π that the closed-loop
In order control
to shorten systemresponse
the sensor becomestime,
as stable as possible,
the rotation speedwhen
is
theset
rotation
to 0.4πspeed
rad/s.is <0.4π rad/s. In order to shorten the sensor response time, the rotation speed is set
to 0.4π rad/s.
Sensors 2017,
Sensors 17,17,
2017, 2221
2221 8ofof1414
Sensors 2017, 17, 2221 88 of 14

0.5
0.5 (a) y=-0.43122*exp(-x/755.52084)+0.47027
(a) y=-0.43122*exp(-x/755.52084)+0.47027
0.4
0.4
2
R2 =0.99983
R =0.99983

U/U22
0.3
0.3

ΔΔU/U
0.2
0.2 CO ΔΔU/U
U/U2 mean
mean data
data
CO 2
0.1
0.1
0.0
0.0
00 500
500 1000
1000 1500
1500 2000
2000 2500
2500
Concentration (ppmv)
Concentration (ppmv)
1.6
1.6 (b) y=-1.42*exp(-x/912.67)+1.51
1.4 (b) y=-1.42*exp(-x/912.67)+1.51
1.4 2
1.2
1.2 R2 =0.99954
R =0.99954
U/U22

1.0
1.0
ΔΔU/U

0.8
0.8
0.6
0.6
0.4 CO2 ΔΔU/U
CO U/U2 mean
mean data
data
0.4 2 2
0.2
0.2
0.0
0.0
00 500
500 1000
1000 1500
1500 2000
2000 2500
2500
Concentration (ppmv)
Concentration (ppmv)
0.9
0.9
0.8 (c) y=-0.77*exp(-x/1475.08)+0.96
(c) y=-0.77*exp(-x/1475.08)+0.96
0.8 2
0.7
0.7
R2 =0.99957
R =0.99957
U/U22

0.6
0.6
ΔΔU/U

0.5
0.5
0.4
0.4
0.3 CH4 ΔΔU/U
CH U/U2 mean
mean data
data
0.3 4 2
0.2
0.2
0.1
0.1
00 500
500 1000
1000 1500
1500 2000
2000 2500
2500
Concentration (ppmv)
Concentration (ppmv)

Figure 6.Experimental
Experimental data
data dots
datadots and
dotsand
and fitting curve
fitting of the
curve the ratioratio
of ratio
the ΔU/U∆U/U
2 versus (a) CO; (b) CO2; (c) CH4
Figure 6. 6.
Figure Experimental fitting curve of ΔU/U 2 versus
2 versus (a) CO
(a) CO; (b) CO;2; (c)
(b)CH
CO4 2 ;
concentration.
(c)concentration.
CH4 concentration.

2100
2100 (a)
(a) rad/s)
v( rad/s)
v(
Concentrationsofof

2050
MeasuredConcentrations

2050
(ppm)

2000
CH4(ppm)

2000
4

1950
1950
CH
Measured

1900
1900
v=ππ
v= v=0.8π
v= 0.8π v=0.5π
v= 0.5π v=0.4π
v= 0.4π v=0.2π
v= 0.2π v=0.1π
v= 0.1π

1850
1850
00 100
100 200
200 300
300 400
400 500
500 600
600
Time(s)
Time(s)
(b)
(b)
40 σσ=41.33
=41.33
40
deviationσσ
Standarddeviation

30
30
σσ=25.62
=25.62

20
Standard

20
σσ=15.68
=15.68

σσ=10.14
=10.14
10
10 σσ=10.51
=10.51
σσ=9.95
=9.95

0.0
0.0 0.2
0.2 0.4
0.4 0.6
0.6 0.8
0.8 1.0
1.0 1.2
1.2
Speed ((ππ rad/s)
Speed rad/s)

Figure
Figure 7. (a)
(a)(a)
7. 7.
Figure MeasuredCH
Measured
Measured CH
CH 4 concentration and
4 4concentration
and (b)
concentration and (b) the
the standard
standard deviation
standard deviation as
asaaafunction
deviationas function ofofrotation
functionof rotation
rotation
speed,
speed, where,
where, thethe measurements
measurements lasted
lasted for
for 100
100 s s
forfor each
each
speed, where, the measurements lasted for 100 s for each speed. speed.
speed.
Sensors 2017, 17, 2221 9 of 14
Sensors 2017, 17, 2221 9 of 14

3.4.
3.4. Dynamic
Dynamic Response
Response Characteristics
Characteristics
Figure
Figure 88 shows
shows the
the working
working steps
steps ofof the
the multi-gas
multi-gas sensor
sensor in in one
one detection
detection cycle.
cycle. First,
First, the
the disc
disc
rotates
rotates (90
(90 −−θ)° ◦
θ) clockwise
clockwisewithin
withinaatimetime tt11to toswitch
switchthethedisc
disc from
from the the CH
CH44 detection
detection position
position toto the
the
zero point of rotation. Then, the disc rotates θ°◦ clockwise within a time t to let
zero point of rotation. Then, the disc rotates θ clockwise within a time t2 to let the CO detector reach
2 the CO detector reach
the
the light
light convergence
convergence point
point ofof the
the mirror.
mirror. After
After aa delay
delay time
time (t
(t33)) of
of 0.5
0.5 ss for
for stabilizing,
stabilizing, the
the detector
detector
indicates
indicates that
that aa measurement
measurement time time (t
(tDD)) of
of 22 ss is
is required
required for
for signal
signal processing
processing and and concentration
concentration
determination.
determination. Finally,
Finally, the
the disc
disc will
will rotate
rotate to to the
the CO
CO22 and
and CHCH44 detection
detection positions
positions and
and detect
detect the
the
corresponding gases,respectively.
corresponding gases, respectively.AsAs thethe optimal
optimal rotation
rotation speedspeed
is 0.4πisrad/s
0.4π and
rad/sθ isand
31 ,θthe
◦ is detection
31°, the
detection
period (T)period
of the (T) of the multi-gas
multi-gas sensor cansensor can be calculated
be calculated by the following
by the following formula formula
90 − 31 31 45
T = t1T+=t2t1++ 3t
t2 ++3t2t+ 2+t43t+ 3t= ( − 31 +
= 90 + 31++3 ×
3 0.5 + 2+
× 0.5 × 2 ×+45
3 ×+ = 10
2)3s× 2) ss = 10 s (8)
3 3
4 D D ( 72 72 72 (8)
72 72 72
This
This means
means that
that the
the multi-gas
multi-gas sensor
sensor will
will complete
complete the
the detection
detection of
of three
three gases
gases by
by scanning
scanning the
the
detectors in 10 s with a detection time of 2 s for each gas.
detectors in 10 s with a detection time of 2 s for each gas.

2000
CH4
3 gases in Dynamic Detection(ppm)

Detecting
Measured Concentrations of

CO2
1500 tD
Detecting
rotate (90-θ)°
clockwise
tD
t1
1000 detection
CO delay t3
detecting t3

tD
500 t3 t4
t2 t4 rotate 45°
rotate θ° rotate 45° counterclockwise
clockwise counterclockwise
0
0 2 4 6 8 10
Time(s)
Thedetection
Figure 8.8.The
Figure detectionofof three
three gasgas species,
species, CO,CO,
CO2CO 2 and
and CH4 CH 4 within
within a period
a period of T on
of T based based on the
the multi-
multi-gas sensor
gas sensor system. system.

3.5. Comparison
3.5. Comparison of
of Detection
Detection Stability
Stability for
for Dynamic
Dynamic and
and Static
Static Operations
Operations
Initially, the chamber was flushed by N22 for
Initially, for 20
20 min,
min, which
which was
was followed
followedby byinjecting
injectingwith
withaa2%2%
uncertainty 10
uncertainty 10 mL
mL pure
pure CO, 20 mL pure CO22,, and and 3030 mL
mL pure
pure CHCH44 into
into the
the enclosure.
enclosure. A A mixing
mixing time
time
of 20 min
of min isisrequired
requiredbefore
beforedetection.
detection.Then,
Then,thethe
concentrations
concentrations of the three
of the gases
three in the
gases inchamber were
the chamber
were calculated
calculated to be to be1263
632, 632, and
12631891
and ppmv,
1891 ppmv, respectively.
respectively.
Subsequently,
Subsequently, the themulti-gas
multi-gas sensor
sensor waswas operated
operated in astate
in a static static state rotation.
without withoutAn rotation.
experimentAn
experiment
was conducted was conducted
for ~5.5 hfor for~5.5
eachh for
gaseach gas measurements
measurements and the andresults
the results are shown
are shown in Figure
in Figure 9a.
9a.
TheThe measured
measured COCO concentration
concentration variedfrom
varied from651.47
651.47toto664.01
664.01ppmvppmv withwith an
an average value of of
657.55 ±±2.35
657.55 2.35ppmv
ppmv(1σ);(1σ);the measured
the measured COCO
2 concentration
2 concentrationvaried fromfrom
varied 1303.74 to 1318.56
1303.74 ppmv,ppmv,
to 1318.56 with
an
withaverage valuevalue
an average of 1311.33
of 1311.33 ± 3.12
± 3.12 ppmvppmv(1σ), andand
(1σ), thethe
measured
measuredCH CH concentration
4 4 concentrationvaried
varied from
from
1948.28 to 1963.10 ppmv, with an average value of 1957.47 ±
1948.28 ± 2.98
2.98ppmv
ppmv(1σ).
(1σ).
Next, the sensor was operated in a dynamic state with rotation and ~5.5 h long measurements
were performed. The results are shown in Figure 9b. The measured CO concentration varied from
625.09 to 675.26 ppmv with an average value of 649.43 ± 9.39 ppmv (1σ); the measured CO2
Sensors 2017, 17, 2221 10 of 14

Next, the sensor was operated in a dynamic state with rotation and ~5.5 h long measurements
were performed.
Sensors The results are shown in Figure 9b. The measured CO concentration varied
2017, 17, 2221 10 from
of 14
625.09 to 675.26 ppmv with an average value of 649.43 ± 9.39 ppmv (1σ); the measured CO2
concentration
concentration varied
varied from
from 1299.93toto1356.26
1299.93 1356.26ppmv
ppmvwith
withananaverage
averagevalue
value of
of 1328.75
1328.75 ±
± 11.88
11.88ppmv
ppmv
(1σ)(1σ)
andand
thethe measured
measured CHCH4 4 concentration
concentration varied
varied from
from 1892.06
1892.06 to
to 1943.94
1943.94 ppmv with
with an
an average
average value
value
of 1924.17 ± 10.39 ppmv
of 1924.17 ± 10.39 ppmv (1σ). (1σ).
Table
Table 1 shows
1 shows a comparison
a comparison ofof
thethestandard
standarddeviations
deviationsbetween
between static
static and
and dynamic
dynamicoperations
operations
of the multi-gas sensor system. For dynamic operation, due to unavoidable
of the multi-gas sensor system. For dynamic operation, due to unavoidable motor vibrations motor vibrations andanda
rotation control error, the dynamic deviations were 3.99, 3.81, 3.49 times higher
a rotation control error, the dynamic deviations were 3.99, 3.81, 3.49 times higher than the static than the static
deviations,
deviations, respectively.
respectively.
3 gases in Static Detection(ppm)

(a) Static stability of CH4

Measured Concentrations of

2000

1500 Static stability of CO2

1000
Static stability of CO
500

0
0 5000 10000 15000 20000
Time(s)
3 gases in Dynamic Detection(ppm)

(b) Dynamic stability of CH4

Measured Concentrations of

2000

1500 Dynamic stability of CO2

1000
Dynamic stability of CO
500

0
0 5000 10000 15000 20000
Time(s)

Figure 9. Long-term (a) static and (b) dynamic measurements of a gas mixture composed of CO, CO2
Figure 9. Long-term (a) static and (b) dynamic measurements of a gas mixture composed of CO, CO2
and CH4 concentration levels of 632, 1263 and 1891 ppmv, respectively.
and CH4 concentration levels of 632, 1263 and 1891 ppmv, respectively.

Table 1. Comparison
Table 1. Comparison of of
standard
standarddeviations
deviationsbetween
betweenstatic
static and
and dynamic
dynamic operations of the
operations of the sensor
sensor
system, where σ is the deviation in static state and σ is the deviation in dynamic state.
system, wheres σs is the deviation in static state and σdd is the deviation in dynamic state.

Standard Deviation CO CO2 CH4

Standard Deviation CO CO 2 CH4
σs (ppmv) 2.35 3.12 2.98
σs σ(ppmv)
d (ppmv)
2.35
9.39 3.12
11.88 2.98
10.39
σd (ppmv) 9.39 11.88 10.39
3.6. Allan Deviation
3.6. Allan Deviation
A long-term evaluation of the multi-gas sensor was performed in a pure N2 environment inside
Achamber.
the long-term Inevaluation of thethe
static detection, multi-gas sensor
sampling wasi.e.,
interval, performed in a pure
the detection N2was
period, environment inside
set to 1 s, and in
thedynamic
chamber.detection,
In static detection,
the samplingtheinterval
samplingwasinterval,
set to 10i.e.,
s. AntheAllan
detection period,
deviation was set
analysis wastoperformed
1 s, and in
to study both stability and detection limit. In static detection, 3600 data points were collected for each
gas in one hour to calculate the Allan variance. In the dynamic detection, 4320 data points for three
gas species, i.e., 1440 data points for each gas, were collected in 4 h.
Sensors 2017, 17, 2221 11 of 14

dynamic detection, the sampling interval was set to 10 s. An Allan deviation analysis was performed
Sensors 2017, 17, 2221 11 of 14
to study both stability and detection limit. In static detection, 3600 data points were collected for each
gas in one hour to calculate the Allan variance. In the dynamic detection, 4320 data points for three
Based on Figure 10a,b, the Allan deviation for static CO detection is 2.96 ppmv with a 1-s
gas species, i.e., 1440 data points for each gas, were collected in 4 h.
averaging time, and that for dynamic CO detection is 8.83 ppmv with a 10-s averaging time. Based
Based on Figure 10a,b, the Allan deviation for static CO detection is 2.96 ppmv with a 1-s averaging
on Figure 11a,b, the Allan deviation for static CO2 detection is 4.54 ppmv with a 1-s averaging time
time, and that for dynamic CO detection is 8.83 ppmv with a 10-s averaging time. Based on Figure 11a,b,
and that for dynamic CO2 detection is 8.69 ppmv with a 10-s averaging time. Based on Figure 12a,b,
the Allan deviation for static CO2 detection is 4.54 ppmv with a 1-s averaging time and that for dynamic
the Allan deviation for static CH4 detection is 2.84 ppmv with a 1-s averaging time and that for
CO2 detection is 8.69 ppmv with a 10-s averaging time. Based on Figure 12a,b, the Allan deviation for
dynamic CH4 detection is 10.29 ppmv with a 10-s averaging time. So it can be estimated that the 1σ
static CH4 detection is 2.84 ppmv with a 1-s averaging time and that for dynamic CH4 detection is
detection limits under static operations are 2.96, 4.54 and 2.84 ppmv for CO, CO2 and CH4,
10.29 ppmv with a 10-s averaging time. So it can be estimated that the 1σ detection limits under static
respectively and the 1σ detection limits under dynamic operations are 8.83, 8.69 and 10.29 ppmv for
operations are 2.96, 4.54 and 2.84 ppmv for CO, CO2 and CH4 , respectively and the 1σ detection limits
the three gas species, respectively. The variation trend of the Allan plot indicates the noise type of the
under dynamic operations are 8.83, 8.69 and 10.29 ppmv for the three gas species, respectively. The
sensor system, which relates to both hardware and software. Since most of the hardware and the
variation trend of the Allan plot indicates the noise type of the sensor system, which relates to both
software for detecting the three gas species are the same, the Allan-plot difference among the three
hardware and software. Since most of the hardware and the software for detecting the three gas species
gases probably results from the differences among the three pyroelectric detectors. The stability of
are the same, the Allan-plot difference among the three gases probably results from the differences
the newly purchased CH4 detector is better than the old CO and CO2 detectors, which may account
among the three pyroelectric detectors. The stability of the newly purchased CH4 detector is better
for the observed phenomenon, i.e., a good White–Gaussian noise dominated sensor operation for
than the old CO and CO2 detectors, which may account for the observed phenomenon, i.e., a good
CH4.
White–Gaussian noise dominated sensor operation for CH4 .

1
10
(a)
CO Allan deviation in static detection
~1/sqrt(τ)
Allan deviation σ (ppm)

0
10

-1
10 0 1 2 3
10 10 Time τ (s) 10 10

(b)
CO Allan deviation in dynamic detection
1
~1/sqrt(τ)
10
Allan deviation σ (ppm)

0
10

2 3 4
10 10 10
Time τ (s)

Figure 10. Allan deviation plot of the CO sensor in (a) static state with a sampling interval of 1 s and
Figure 10. Allan deviation plot of the CO sensor in (a) static state with a sampling interval of 1 s and
(b) dynamic state with a sampling interval of 10 s.
(b) dynamic state with a sampling interval of 10 s.
Sensors 2017,
Sensors 2017, 17,
17, 2221
2221 12 of
12 of 14
14
Sensors 2017, 17, 2221 12 of 14

Figure 11.
Figure 11. Allan
Allan deviation
deviation plot
plot of
of the
the CO
CO22 sensor
sensor inin (a)
(a) static
static state
state with
with aa sampling
sampling interval
interval of
of 11 ss and
and
Figure 11. Allan
(b) dynamic
dynamic statedeviation plot of the
with aa sampling
sampling CO2 sensor
interval of 10 in (a) static state with a sampling interval of 1 s and
10s.s.
(b) state with interval of
(b) dynamic state with a sampling interval of 10 s.
1
10
(a)
10
1
CH4 Allan deviation in static detection
(a) ~1/sqrt (τ) deviation in static detection
CH 4
Allan
~1/sqrt(τ)
σ (ppm
deviation
deviation ) )
σ (ppm

0
10
0
10
Allan
Allan

0 1 2 3
10 10 10 10
0 1 Time τ (s) 2 3
10 10 10 10
Time τ (s)
(b)
CH4 Allan deviation in dynamic detection
(b)
~1/sqrt
CH Allan
(τ)deviation in dynamic detection
) )

4
σ (ppm

~1/sqrt(τ)
1
σ (ppm

10
1
10
deviation
deviation
Allan

0
10
Allan

0
10

2 3 4
10 10 10
2 Time τ (s) 3 4
10 10 10
Time τ (s)
Figure 12. Allan
Figure 12. Allan deviation
deviation plot
plot of
of the
the CH sensor in
CH44 sensor in (a)
(a) static
static state
state with
with aa sampling
sampling interval
interval of
of 11 ss and
and
Figure
(b) 12.
(b) dynamicAllan
dynamic statedeviation
state with plot of
with aa sampling the CH
sampling interval sensor
interval of
4
of 10 in
10s.s. (a) static state with a sampling interval of 1 s and
(b) dynamic state with a sampling interval of 10 s.
Sensors 2017, 17, 2221 13 of 14

4. Conclusions
Finally, we were able to develop a multi-gas sensor system based on a single broadband light
source and three pyroelectric detectors. A stepper motor driven rotation system was designed to
switch to CO, CO2 and CH4 detection channels and a single-reflection spherical optical mirror was
used to enhance gas absorption. A dual-channel differential detection method was used to suppress
noise and improve the detection performance. Both static and dynamic measurements were performed
to investigate the sensing characteristics of the sensor system. The Allan deviations for static CO, CO2
and CH4 detection are 2.96, 4.54 and 2.84 ppmv, respectively, with a 1-s averaging time and that for
dynamic CO, CO2 and CH4 detection is 8.83, 8.69 and 10.29 ppmv, respectively, with a 10-s averaging
time. The reported detection method has various applications such as coal mine gas detection.

Acknowledgments: The authors wish to express their gratitude to the National Key Research and Development
(R&D) Program of China (Grant Nos. 2017YFB0405300, 2016YFD0700101, 2016YFC0303902), National Natural
Science Foundation of China (Grant Nos. 61775079, 61627823, 61307124), Science and Technology Department of
Jilin Province of China (Grant No. 20140307014SF), and Changchun Municipal Science and Technology Bureau
(Grant No. 14KG022) for their generous support to this work. Frank K. Tittel acknowledges support by the Robert
Welch Foundation (Grant No. C-0586).
Author Contributions: C.Z. and Y.W. conceived and designed the experiments; Q.D. and Y.Z. were responsible
for the simulations; S.M. and M.D. performed the reported experiments; C.Z. and F.K.T. wrote the manuscript
based on the input from the other authors.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Wang, W.Z.; Wang, Y.M.; Song, W.J.; Li, X.Q. Multiband infrared inversion for low-concentration methane
monitoring in a confined dust-polluted atmosphere. Appl. Opt. 2017, 56, 2548–2555. [CrossRef] [PubMed]
2. Zheng, C.T.; Ye, W.L.; Sanchez, N.P.; Li, C.G.; Dong, L.; Wang, Y.D.; Griffin, E.J.; Tittel, F.K. Development and
field deployment of a mid-infrared methane sensor without pressure control using interband cascade laser
absorption spectroscopy. Sens. Actuators B Chem. 2017, 244, 365–372. [CrossRef]
3. Pakmanesh, N.; Cristescu, S.M.; Ghorbanzadeh, A.; Harren, F.J.M.; Mandon, J. Quantum cascade laser-based
sensors for the detection of exhaled carbon monoxide. Appl. Phys. B 2016, 122, 10. [CrossRef]
4. Cho, H.J.; Kim, S.J.; Choi, S.J.; Jang, J.S.; Kim, I.D. Facile synthetic method of catalyst-loaded ZnO nanofibers
composite sensor arrays using bio-inspired protein cages for pattern recognition of exhaled breath. Sens.
Actuators B Chem. 2017, 243, 166–175. [CrossRef]
5. Wang, J.N.; Zheng, L.J.; Niu, X.T.; Zheng, C.T.; Wang, Y.D.; Tittel, F.K. Mid-infrared
absorption-spectroscopy-based carbon dioxide sensor network in greenhouse agriculture: Development and
deployment. Appl. Opt. 2016, 55, 7029–7036. [CrossRef] [PubMed]
6. Jones, A.Y.; Lam, P.K. End-expiratory carbon monoxide levels in healthy subjects living in a densely
populated urban environment. Sci. Total Environ. 2006, 354, 150–156. [CrossRef] [PubMed]
7. Karacan, C.Ö.; Ruiz, F.A.; Cotè, M.; Phipps, S. Coal mine methane: A review of capture and utilization
practices with benefits to mining safety and to greenhouse gas reduction. Int. J. Coal Geol. 2011, 86, 121–156.
[CrossRef]
8. Yuan, L.; Zhou, L.; Smith, A.C. Modeling carbon monoxide spread in underground mine fires. Appl. Therm.
Eng. 2016, 100, 1319–1326. [CrossRef] [PubMed]
9. Jiang, Y.; Li, G.; Wang, J. Photoacoustic compound fire alarm system for detecting particles and carbon
monoxide in smoke. Fire Technol. 2016, 52, 1255–1269. [CrossRef]
10. Hu, L.H.; Fong, N.K.; Yang, L.Z.; Chow, W.K.; Li, Y.Z.; Huo, R. Modeling fire-induced smoke spread and
carbon monoxide transportation in a long channel: Fire Dynamics Simulator comparisons with measured
data. J. Hazard. Mater. 2007, 140, 293–298. [CrossRef] [PubMed]
11. Bidaux, Y.; Bismuto, A.; Patimisco, P.; Sampaolo, A.; Gresch, T.; Strubi, G.; Blaser, S.; Tittel, F.K.; Spagnolo, V.;
Muller, A. Mid infrared quantum cascade laser operating in pure amplitude modulation for background-free
trace gas spectroscopy. Opt. Express 2016, 24, 26464–26471. [CrossRef] [PubMed]
Sensors 2017, 17, 2221 14 of 14

12. Miklós, A.; Hess, P.; Bozóki, Z. Application of acoustic resonators in photoacoustic trace gas analysis and
metrology. Rev. Sci. Instrum. 2001, 72, 1937–1955. [CrossRef]
13. Dumitras, D.C.; Dutu, D.C.; Matei, C.; Magureanu, A.M.; Petrus, M.; Popa, C. Laser photoacoustic
spectroscopy: Principles, instrumentation, and characterization. J. Optoelectron. Adv. Mater. 2007, 9, 3655–3701.
14. Bogue, R. Detecting gases with light: A review of optical gas sensor technologies. Sens. Rev. 2015, 35, 133–140.
[CrossRef]
15. Han, C.H.; Hong, D.W.; Han, S.D.; Gwak, J.; Singh, K.C. Catalytic combustion type hydrogen gas sensor
using TiO2 and UV-LED. Sens. Actuators B Chem. 2007, 125, 224–228. [CrossRef]
16. Schwarz, H.; Dong, Y.; Horn, R. Catalytic methane combustion on a Pt Gauze: Laser-Induced fluorescence
spectroscopy, species profiles, and simulations. Chem. Eng. Technol. 2016, 39, 2011–2019. [CrossRef]
17. Li, G.L.; Sui, Y.; Dong, M.; Ye, W.L.; Zheng, C.T.; Wang, Y.D. A carbon monoxide detection device based on
mid-infrared absorption spectroscopy at 4.6 µm. Appl. Phys. B 2015, 119, 287–296. [CrossRef]
18. Besson, J.P.; Schilt, S.; Thévenaz, L. Multi-gas sensing based on photoacoustic spectroscopy using tunable
laser diodes. Spectrochim. Acta Part A 2004, 60, 3449–3456. [CrossRef] [PubMed]
19. Mukherjee, A.; Prasanna, M.; Lane, M.; Go, R.; Dunayevskiy, I.; Tsekoun, A.; Patel, C.K.N. Optically
multiplexed multi-gas detection using quantum cascade laser photoacoustic spectroscopy. Appl. Opt. 2008,
47, 4884–4887. [CrossRef] [PubMed]
20. Tan, Q.; Tang, L.; Yang, M.; Xue, C.; Zhang, W.; Liu, J.; Xiong, J. Three-gas detection system with IR optical
sensor based on NDIR technology. Opt. Lasers Eng. 2015, 74, 103–108. [CrossRef]
21. Mao, X.; Zhou, X.; Gong, Z.; Yu, Q. An all-optical photoacoustic spectrometer for multi-gas analysis. Sens.
Actuators B Chem. 2016, 232, 251–256. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).