Proceedings of IEEE International Conference on Applied System Innovation 2018

IEEE ICASI 2018- Meen, Prior & Lam (Eds)

A Multi-Gas Sensing System for Air Quality Monitoring

Chun-Ming Huang, Yi-Jun Liu, Yi-Jie Hsieh, Wei-Lin Lai, Chun-Ying Juan, Ssu-Ying Chen,
Chih-Chyau Yang, and Chien-Ming Wu
National Chip Implementation Center
National Applied Research Laboratories
7F, No.26, Prosperity Rd. 1, Science Park
Hsinchu City, Taiwan
E-mail:[email protected]

Abstract in closed spaces [1]. Therefore, most of the air-quality

It is widely known that the atmosphere has been highly monitoring products aim to detect harmful gases which are
polluted over the recent years, and thus more and more people common seen pollution resources, such as CO, CO2, CH2O,
start to take this subject seriously. The air pollutant are emitted total volatile organic compounds (TVOC) and smog (PM2.5).
from both man-made and natural sources, such as burning of As people started to understand the influence of air pollution
fossil fuels in electricity generation, industrial process and on certain disease, they want to understand the ambient air
solvent use, and windblown dust. Therefore, most of the quality in real time. In response to this request, researches and
existing air-quality monitoring products aim to detect harmful commercial products about air quality monitoring have
gases which are common seen pollution resources, such as become more and more popular. There are various products on
carbon monoxide (CO), carbon dioxide (CO2) and the market now, and the prices fall approximately between
formaldehyde (CH2O) as well as a mixture of smoke and fog, 100~250 USD due to different functions and different
i.e. smog (PM2.5). Most of the current affordable air monitor accuracies [2-3]. However, most of the affordable air monitor
system can detect only one kind of gas. When multiple gases system can detect only one kind of gas. When multiple gases
are of interest, users have to buy many products. are of interest, user has to buy many products.
In this paper, we propose a wireless multi-gas sensing On the other hand, the improvement of the sensor
system which is able to detect the concentration of three developing technology has greatly increased the portability of
harmful compounds, i.e. CO, CO 2 and CH2O as well as the air monitor system, since the state-of-the-art gas sensors
PM2.5. A smartphone App is developed to show real-time are of small size and having low power consumption, yet with
data and alert users when the corresponding gas concentration acceptable performance. A portable monitor system allows
is too high, so that users can take actions to eliminate those users to deploy sensor system in any location, which would
factors in time. The hardware sharing technique is utilized to realize more flexible applications. For example, [4] proposed a
maximize the hardware utilization and thus reduce the portable particulate monitoring system that greatly improves
hardware cost. As far as power consumption is concerned, the the mobility when comparing with traditional ones. However,
energy-hungry Wi-Fi module will enter the deep sleep mode the data is only displayed on the screen of the system; users
while the Bluetooth mode is selected. The power consumption wouldn’t be able to know the reading from a distance.
saving report is given. In addition, the sensor calibration Besides, it can only detect one item, i.e. the particulate. To be
schemes for these four gas sensors are also presented in this able to detect different kinds of gases, [5] integrates a
paper. commercial development board, different sensors and a
wireless communication module to build a system. This
Keywords— multi-gas sensing system, air quality monitoring prototype can realize reading multiple gas concentration data
system, air pollutant at one time, but the overall size and power consumption are
large, therefore not easy to carry around and apply in real life.
Introduction In our work, we integrate four gas sensors which detect the
Environmental problem has always been an important issue common indoor air pollutants. We also include wireless
around the world. One of the most serious problems related to modules so that we can move the bulky user interfaces from
our daily lives is air pollution, and there are more and more the hardware itself to the smartphone or a remote server;
people who start to take this subject seriously. Human inhales hence the hardware is small and compact. The wireless
approximately 10,000 liter of air a day. Since we spent most of communication also brings flexibility for future applications’
the time indoor, the indoor air quality (IAQ) has naturally development. Besides, the resource sharing technique is
become an important factor affecting our health and live adopted to further minimize the overall system size. The rest
quality. Poor indoor air quality makes people wearing and of this paper is organized as follows. In Section II, we present
sick. The most common resources of indoor air pollution are the hardware architecture of our proposed multi-gas sensing
chemical substances used in decoration, human activities, system. Then, the calibration schemes for all the gas sensors
contaminated ventilation system and insufficient air exchange are described in Section III. Next, the firmware and App
development of the system is provided in Section IV and
Section V. Finally, the conclusions are given in Section VI.

834
ISBN 978-1-5386-4342-6
 Proceedings of IEEE International Conference on Applied System Innovation 2018
IEEE ICASI 2018- Meen, Prior & Lam (Eds)

Hardware Architecture SPECIFICATIONS OF THE SELECTED SENSORS

The proposed multi-gas sensing system integrates multiple Sensors Model Nominal Sensor
gas sensors, i.e. CO, CO2, CH2O and PM2.5. Besides, Range Type
wireless communication is adopted for data transmission in CO ME2-CO 0~5000ppm Electro-
our design to achieve more possibility in applications and chemical
CO 2 MG811 350~10kppm Solid electrolyte
reduce the form factor as well. The hardware block diagram is cell
as shown in Fig. 1, where the sensor modules are colored in CH2O ME2-CH2O 0~10 ppm Electro-
gray. The corresponding controllers for each sensor or chemical
communication module are illustrated as well. PM2.5 PMS3003 > 0.3um Laser
diffraction
Temp./ HTU21D -40~125°C/ MEMS
Humi. 0~100 %RH

(a) Upside (b) Downside

Fig. 1 Block diagram of our proposed platform Fig. 2 The picture of our proposed platform without a cover

A. Cost-efficient Readout Circuit Design for Sensors Sensor Calibration Schemes

The specifications of the selected sensors are listed in Table In this section, the calibration schemes for all the gas sensors
I. For those electrochemical type current-mode output sensors in our system are presented. To downsize the multi-gas
whose output currents are relative to the gas concentration sensing system, the CO and CH2O sensors in this system are
they detect, the sensor outputs have to be amplified before we electrochemical sensors. To achieve better performance and
quantify the result, because the output current is too small reliability, some proper post processing and data modeling are
(micro-A level) to be detected. We use the LMP91000 analog necessary [6].
front-end (AFE) to translate current signal to voltage output.
LMP91000 is a single channel trans-impedance amplifier A. CO2
(TIA) and the gain can be adjustable through I2C interface. To The most common method of gas sensor data modeling is to
achieve enough resolution, a 16-bit analog-to-digital converter expose the sensor in gas with known concentration, then
(ADC) ADC161S626 is chosen for converting the AFE’s acquire the sensor reading and do the following comparison.
voltage output into digital data. Consider that we have two To produce specified gas concentration, we use high CO2
current-mode gas sensors (ME2-CO and ME2-CH2O) while concentration gas (5000 ppm), a zero gas generator which
the LMP91000 chip has only one analog input, we use the would purify atmosphere air, and a gas dilution system to mix
NX3L2467 dual double-pole double-throw (DPDT) analog CO 2 gas and purified air. The experiment environment was
switches to minimize the design cost. The channel resistance simplified as block diagram shown Fig. 3. The CO 2 cylinder
of the analog switch will not cause signal attenuation because and the zero air generator are connected to the gas dilution
the sensors output in current, other than in voltage. To system with tubes. The dilution system can adjust the mix
minimize the cross talk effect between two analog channels, ratio of the two gases and generate CO2 of different
the input/output pins of other unused channel of the concentration. The mixed air will first flow through a
NX3L2467 chip are all tie to ground. commercial CO2 sensor, Gascard NG [7], then flow to the CO2
sensor of our system. After acquiring the reading results of the
B. Wireless Communication Gascard NG sensor and our CO2 sensor, we can get the
There are two communication modules in our design, i.e. mapping table for our CO 2 sensor.
Bluetooth and Wi-Fi. The nRF51822 is a Bluetooth 4.0
Since the property of each CO2 sensor is not exactly the
compatible device. We make it work as an SPI-slave module
same, we have to individually calibrate each sensor. The test
in full charge of the BLE protocol, so that MCU can easily
result of the data modeling experiment is as shown in Fig. 4.
manipulate data to be transmitted. The ESP07 is a UART-
The horizontal axis represents the output voltage of the CO2
interface 802.11 B/G compatible Wi-Fi module. The baud-rate
sensor, and the vertical axis shows the corresponding CO2
is set as 921600 bps to maximize the data rate. It is mainly
concentrations obtain from the Gascard NG sensor. There are
used to communicate with mobile devices and can also realize
two different sensor data in the chart, it can be observed that
more advanced network topology in the future. Our proposed
sensor #1 and sensor #2 have similar trends, but there is a
multi-gas platform without a cover is as shown in Fig. 2.
voltage bias and slightly slope difference. The reading of each
CO 2 sensor can be modeled as a quadratic equation, and can
TABLE I

ISBN 978-1-5386-4342-6
835
 Proceedings of IEEE International Conference on Applied System Innovation 2018
IEEE ICASI 2018- Meen, Prior & Lam (Eds)

be saved in the App to obtain the CO2 concentration based on The measurement result of the PM2.5 concentration is shown
the sensor reading. in Fig. 6. It is clearly that the concentration trend is very alike.
We make some post processing on the sensor’s reading and
make it more accurate base on the test result.

Fig. 3 CO2 data modeling experimental environment

Fig. 6 Measure result of PM2.5 concentration

D. CH 2O
Fig. 4 Relationship between the commercial CO2 sensor and system The CH2O sensor is a sensitive current-mode
sensor electrochemical sensor. The CH2O’s sensitivity is about
1.1uA/ppm and the zero-drift is 0.2ppm [11]. The output
B. CO
current variation is linear but may vary among different
We adopt the CO calibration scheme proposed in our sensors. In order to get accurate transfer function between the
previous CO sensor system work [8]. The calibration method
output current value and the concentration, we built a
of the CO sensor is similar to that of the CO2 sensor. The
calibration environment. A CH 2O generator outputs the
experiment environment is the same as Fig. 3, except
formaldehyde gas at a given concentration. We make a gas
replacing the CO2 gas cylinder with 100 ppm CO gas. The
pump intermittently operate to ensure the gas is evenly
sensor calibration method adopts has been made. We used a
covered on our sensor. We measured with three different
high-precision CO sensor, LPT-A-COB [9] as the reference
formaldehyde concentrations, i.e. 0, 0.275 and 0.501 ppm.
device. The calibration experiment was performed with the
Each condition is hold at least thirty minutes to get a stable
CO gas concentration varying from 25 to 100 ppm, and the
environment and output data. Therefor we can get a transfer
comparison result was shown in Fig. 5. The reading of LPT-
function to calculate the gas concentration with the measured
A-COB and the CO sensor of the multi-gas sensing system
current.
present in a linear function, therefore we can confirm the
accuracy of our sensor and further calculate the transfer
Firmware Development
equation for our CO sensor.
In our proposed wireless multi-gas sensing system, MCU is
in charge of initializing all sensors, fetching those sensor’s
data periodically and realizing wireless communication with
the smartphone. It uses different interfaces to configure and
communicate with different sensors, as depicted in Fig. 2. In
the system, CO/CH2O’s sensor data are amplified by an SPI-
interfaced AFE chip, while the BLE chip interacts with MCU
with the same interface. Since data acquisition and wireless
transmission is not operated at the same time, MCU enables
the corresponding chip select (CS) signals for AFE chip and
Fig. 5 Relationship between the LPT-A-COB and the CO sensor BLE chip one at a time. Besides, AFE and BLE share the
same SCK pin of SPI but with different clock rate, we re-
C. PM2.5 initialize the SPI controller when we change the enabled CS
The PM2.5 sensor in the multi-gas sensing system is signal.
designed with laser scattering principle. We properly designed
an air duct that let the particles in air goes through the A. Interaction between the Firmware and App
detecting area, and to allow the detector to receive the The smartphone and the platform can interact with each
scattered light correctly. Some field test has been done to other by a robust handshaking mechanism that utilizes the
make sure the system’s construct wouldn’t affect the inlet air instruction set we made. The App will always get a response
flow and cause inaccurate reading. A high-accuracy laser after issuing a command to the platform. Even the error cases
PM2.5 monitor [10] was used as a reference device. We put are encoded with different IDs that the App will always be
the multi-gas sensing system and the reference device closely aware of the status of the platform.
to each other on the windowsill, with the window wide open.

836
ISBN 978-1-5386-4342-6
 Proceedings of IEEE International Conference on Applied System Innovation 2018
IEEE ICASI 2018- Meen, Prior & Lam (Eds)

The smartphone triggers the wireless linking process with

the platform. For BLE connection, pairing-up is quite
straightforward, while the Wi-Fi connection needs some
configuration first. To make the platform connect with the
smartphone through Wi-Fi, user has to inform the platform the
Access Port’s (AP’s) SSID/Password first. When the platform
and the smartphone are in the same local area network (LAN),
they can communicate with each other, while the platform as
the client and the smartphone as the server. After successfully
building up the Wi-Fi link, App can either retrieves sensor
data by BLE or Wi-Fi. For Retrieve Sensor Data command,
user encapsulates the ID of the gas sensor of interest in the
command’s Sensor ID field. Moreover, if user wants to know
all sensors’ value, user can set the sensor ID to 0x00 so that all
sensors’ data will be sent to the App at once to improve the
communication efficiency.

B. Power Consumption Reduction

The BLE module is a low power device that the radio peak Fig. 7 Screenshot of the App and the form factor of our platform
current is less than 10 mA when the VDD is 3V. Our
experiment shows that the Wi-Fi module consumes about 70 Conclusions
mA. Therefore, when user chooses to retrieve sensor data In this paper, a wireless multi-gas sensing system is
under BLE mode, we put the energy-hungry Wi-Fi module presented to detect the concentration of three harmful
into deep sleep mode to decrease the overall system power compounds, i.e. CO, CO2 and CH 2O as well as PM2.5. With
consumption. Power consumption under different operating the wireless communication modules, sensor data can be
condition is as Table II shows. It can be seen that since BLE shown on the smartphone in a distance. Since the user
consumes really low power, the first and last items’ power interfaces is implemented on the phone and the hardware
consumption are almost the same. The power reduction resource sharing scheme, hence the hardware is small and
scheme just mentioned saves about 45 mA for the system. compact. The calibration schemes for all the gas sensors are
also presented to ensure the correctness of the sensor data. In
TABLE II the near future, we will provide a cloud server that facilitates
POWER CONSUMPTION ANALYSIS long-term sensor data monitoring by using the built-in Wi-Fi
Communication method Average current (mA) module in the platform.
BLE+Wi-Fi 560
BLE only 514
Wi-Fi only 559 References
[1] "An introduction to indoor air quality", United States
Environmental Protection Agency, [online] Available:
App Development https://www.epa.gov/indoor-air-quality-iaq/introduction-indoor-
Upon opening the App, users can choose to retrieve sensor air-quality.
data via BLE or Wi-Fi. The initialization of Multi-Gas is done [2] iCeeO2-700 CO2 monitor [online] Available:
through BLE, so first the App would check whether BLE is http://www.iceeo2.com/
enabled or not on the phone. If not, App would ask the user to [3] Bosswell PM2.5 detector [online] Available:
enable BLE. Afterward, App starts to scan the surrounded https://mall.pchome.com.tw/prod/QBAV3J-A9008R9JY
multi-gas devices, and displays the discovered devices’ MAC [4] A. Priyadarshini et al., “A real time portable embedded system
addresses and device name (i.e. Multi-Gas) on the screen. design for particulate matter monitoring,” Bombay Section
After user selects the desired device, the application will Symposium, Sep. 2015
[5] A. Kashid, et al., “Real Time Indoor Air Quality Monitoring
connect and initialize the multi-gas system through BLE
System,” IJIREEICE, pp. 36-38, vol. 4, issue 11, Nov. 2016.
protocol. Initialization includes retrieving firmware version, [6] C. Borrego et al., “Assessment of air quality microsensors versus
setting data transfer mode, etc. If the user chooses to transfer reference methods: The EuNetAir joint exercise,” Atmospheric
mainly through BLE, the App starts sending commands and Environment 147, pp. 246-263, Sep. 2016
receiving sensor data to/from the multi-gas platform. If user [7] Infrared CO2 Sensor: Gascard NG CO2 sensor, Edinburgh
chooses to operate under Wi-Fi protocol, user informs the Sensors
platform the target Wi-Fi AP’s SSID and password through [8] C.-C. Chen, et al., “A Wireless and Batteryless Intelligent Carbon
BLE. Multi-gas will start to connect to designate Wi-Fi AP. Monoxide Sensor,” Sensors, vol. 16, issue 10, Sep. 2016
After connecting to Wi-Fi AP successfully, the platform is [9] Electrochemical CO Sensor: LPT-A-COB, Critical Environment
Technologies Canada Inc.
ready to receive commands from App. App will send the
[10] High-Accuracy Laser PM2.5 monitor: DustTrak II Aerosol
retrieve sensor data command. Fig. 7 shows the screenshot of Monitor 8530, TSI
the App that continuously retrieves the sensor data and the [11] ME2-CH2O-Ф16 Electrochemical Formaldehyde Sensor
form factor of our proposed multi-gas platform. Manual, V1.3, Winsen Electronics Technology

ISBN 978-1-5386-4342-6
837