Canarin node is the air-quality sensing node which is developed under the Canarin project in collaboration with intERLab (AIT, Thailand), LIP6 (Sorbonne Universite, France) and University of Bologna (Italy).
In SEA-HAZEMON project, the Canarin3 (Canarin version 3) is mainly used. The other versions, Canarin1 and Canarin2 are used by the other research groups in Paris, Macau and Bologna.
Left: First batch “lunchbox” version of Canarin3 node
Right: Refined version of Canarin3 node
Each Canarin3 node consists of sensor modules and processing-communication part.
The included sensors with specification in Canarin3 are listed as follow:
- PM sensor: Plantower PMS7003
- Measurements: PM1.0, PM2.5, PM10
- Detection range: 0-500 μg/m3
- Resolution: 1 μg/m3
- Accuracy: 10% @100-500 μg/m3, 10 μg/m3 @0-100μg/m3
- Humidity and temperature
- Detection range: 0-99 %RH, -40-80 C
- Resolution: 0.1%RH, 0.1 C
- Accuracy: 2%RH, 0.5 C
- Barometric pressure sensor:
- Detection range: 950-1050 hPa
- Resolution: 0.12 hPa
- Accuracy: 0.01 hPa
- Carbon dioxide sensor: MH-Z16
- Detection range: 0-2000 ppm
- Resolution: 1 ppm
- Accuracy: 50 ppm
- Carbon monoxide sensor: MQ7
- Detection range: 20-2000 ppm
- Formaldehyde sensor: Dart WZ-S
- Detection range: 0-2 ppm
- Resolution: 0.001 ppm
- GPS: Ublox M8N
All sensors are used with built-in factory calibrated value, except the CO sensor. The CO sensor (MQ7) output raw voltage which cannot be used directly. We use a reliable instrument as a reference to calibrate and convert the raw reading into meaningful reading.
On the processing-connectivity side, Canarin3 uses UDOO Neo Full board (https://www.udoo.org/udoo-neo/). The UDOO board has built-in an Arduino-compatible Cortex-M4 microcontroller, an ARM Cortex-A9 processor, an SD card slot and WiFi. The components are utilised to collect data from sensors, store data to the SD card and send them to the server whenever the WiFi connectivity is available.
Yun Weather station
The weather station is based on Arduino Yun and Weather Sensor Assembly. It can measure wind speed, wind direction and rain.
Argent Data Systems – Weather Sensor Assembly p/n 80422 (https://www.argentdata.com/catalog/product_info.php?products_id=145)
- Wind speed resolution: 2.4 km/h
- Wind direction resolution: 22.5 degree
- Rain gauge resolution: 0.2794 mm
Arduino Yun (https://store.arduino.cc/arduino-yun, https://store.arduino.cc/arduino-yun-mini) is used for processing, storage and network connectivity. It has an Arduino microcontroller, a MIPS processor running OpenWRT based operating system (https://openwrt.org/), an SD card slot and built-in WiFi.
Photos: Weather Sensor before and after deployment.
Sensor testing and calibration
PM sensor, CO sensor and CO2 sensor were tested in both closed and open environment using the certified Federal Reference Method (FRM) and/or Federal Equivalent Method (FEM) instrumentation to compare with the sensor datasets (Sathita Fakprapai, AIT, 2017), (Loi Huynh Tan, AIT, 2016). Sensors were placed in a closed controlled environment and compare the reading with a certified instrument. The open environment testing is done by attaching sensors to PCD stations and comparing the value. The result shows that the sensors have reliable performance as resembling its specification.
The dataset is also used in the linear regression to calibrate some sensors which itself do not output calibrated meaningful data from the factory, such as the CO sensor.
Photo: PM, CO and CO2 sensors testing and calibration with calibrated equipment and PCD stations. (Sathita Fakprapai, 2017)
We also tested the reading consistent between different Canarin3 nodes by putting all nodes together in the same environment. Under the same condition, all nodes give similar reading according to the sensor’s datasheet. From this process, we can identify and fix the node that apparently reports the strange reading from the group.
Photo: Canarin3 nodes in testing.
Figure: Sample PM2.5 data from Canarin3 on a testing period.
Data Collection and Communication
The HAZEMON sensor node read data from the attached sensors periodically. Each reading is uniquely identified by the type of sensor with a timestamp. It will be individually saved to a file on the non-volatile storage. In the case of Canarin3 node, the storage is an SD card.
Canarin node can connect to WiFi network for internet connection. Whenever the connection is available, each stored data files is read and uploaded to the Canarin server (canarin.net) one by one. If an upload is successful, the data file is deleted from the node, freeing up the storage space.
Figure: dataflow from sensors to server.
Blue arrow – sensor data reading and store process.
Red arrow – uploading process.
From typical deployment, the sensor node is installed in a place that always has network connectivity. Immediately after the data file is saved to the storage, it gets read, uploaded and deleted at almost no time.
The sensor node can collect data without a network connection. It has sufficient space to store data, depends on the capacity of storage, from several months up to years of offline data collection. The stored data will be uploaded automatically, whenever the connection becomes available.
Since components in the sensors will not work without electric power, the node cannot measure data during the power lost. However, it will restart itself when the power comes back. The data files that are already stored on the non-volatile storage do still remain to be uploaded whenever it is able to.
Uploaded data from all SEA-HAZEMON sensors can be viewed from the project website on Canarin server (https://canarin.net/seahazemon).
Screenshot of HAZEMON website on Canarin server, showing data in map view.
Sensor node installation
The air-quality monitoring nodes are installed with these requirements:
- About 2.5-3 meters height above ground level.
- Outdoor, open air.
- Far from the kitchen, exhaust air duct and place with an open-burning related activity.
- Protected from the rain, water and harmful environment.
- Has electricity to power the sensor.
- Has WiFi connectivity for data upload.
Maesot 2018 deployment
On March 2018, ten of Canarin nodes were installed in the area around Maesot district in Tak province of Thailand.