4.1. GAS Sensors
Before providing a detailed explanation of each component of APOLLO, we first introduce the principles and characteristics of the chemical gas sensors mounted on our sensor board. Several types of COTS chemical gas sensors exist, but each sensor has different operation principles. The operational characteristics of a gas sensor are generally classified into three types: heating semiconductor, non-dispersive infrared (NDIR), and light emitting diode (LED). The size, accuracy, and power consumption of a compact gas sensor all vary with sensor type.
The heating semiconductor sensor evaluates a target gas concentration by measuring the electrical conductivity of a sensing layer that is composed of a metal-oxide material such as tin dioxide (SnO2
) or zinc oxide (ZnO). When toxic gases reach the sensor's surface and are absorbed, its electrical conductivity changes. For semiconductor sensors, a warm-up time is needed because the semiconducting oxides react sensitively to vapor and other chemicals. For example, the output of the CO sensor, MiCS-5521, is stabilized when the temperature reaches about 340 °C. Pre-heating eliminates both vapor and impurities on the sensing layer so that the chemical equilibrium is achieved. With the use of MEMS technology, this type of sensor is small in size and has a fast response time [6
]. The CO, NO2
, and VOC sensors belong to this category.
The NDIR sensor consists of an infrared lamp, a sample chamber or light tube, a wavelength filter, and the infrared detector. Gas is pumped into the sample chamber, and the gas concentration is measured electro-optically by absorbing a specific wavelength in infrared. The infrared light is directed through the sample chamber towards the detector, which has an optical filter in front of it to eliminate all light except the wavelength absorbable by the selected gas molecules. Ideally, other gas molecules do not absorb light at this wavelength and do not affect the amount of light reaching the detector. NDIR sensors usually consume more energy than the semiconductor sensors; however, they provide accurate measurements. The CO2 sensor belongs to this category.
An LED sensor such as the PPDNS4 counts the number of particles based on the amount of LED light a particle blocks when passing through the detection area of the sensor. Since this type of sensor has a heater for air circulation inside the sensor, its power consumption is significantly higher than that of the other sensors.
4.2. Sensor Board
Since the primary goal of APOLLO is to provide information about the EPA-specified criteria pollutants using inexpensive compact sensors, we used COTS gas sensors that would satisfy the requirements of low cost and applicability. The chosen sensors, which detect CO, CO2, NO2, particulate matter (PM), and volatile organic compounds (VOCs), were integrated into a single board for analysis of the characteristics of each sensor in a consistent environment. The VOC sensor was selected because it detects SO2. A temperature/humidity sensor was also mounted on the sensor board, since the sensing results of the gas sensors are sensitive to ambient temperature and humidity.
Mounting several gas sensors on a single integrated sensor board has advantages in terms of energy use, cost, installation time, and pollutant detection compared to multiple and separate sensor board designs. As described in Table 1
, many noxious gases are generated from the same contamination source. Therefore, placing the sensors close together enables the valid detection of air pollutants. The accuracy of the semiconductor gas sensors is normally influenced by temperature and humidity; therefore, a considerable baseline drift may be caused. To overcome this problem, the humidity and temperature sensors were attached to the sensor board and correction algorithms applied.
However, integrating several sensors into a sensor board presented a series of unexpected problems in terms of power supply. Table 2
shows that the rated voltages of the six sensors differ significantly.
In an early design, we used a voltage boosting circuit that enables 12-V supply from the 3-V power source. This method caused a critical voltage drop, and, consequently, the radio transceiver did not function properly. In addition, the sensor node consumed two AA batteries after about one hour of operation. With the hardware revision, the power supply was separated from the sensor node and changed to a 12-V rechargeable battery. To support the three kinds of rated voltage, two DC-DC converters were employed: one for step-down 12 V to 3 V and the other for step-up 3 V to 5 V.
After changing the power supply designs, the sensors such as the PPD4NS and D-120 still did not function properly after a certain time of operation, due to their excessive power requirements. We concluded that a continuous power supply was not feasible, even with larger batteries; hence, sensor operations should be conducted as power-manageable components. In our final design, all the gas sensors had electrical switches to enable software-based power management. The final hardware is shown in Figure 1
. The sensor board was basically powered by a lithium-ion rechargeable battery, but, for high spatial flexibility and convenience, the sensor board was designed such that power supply from power outlets is also possible.
One interesting observation was that the heat from chemical sensors such as the CO, VOC, and NO2 sensors could greatly influence the temperature measured by the SHT11 unit. During the hardware revision, we recognized that the measured temperature from the sensor was significantly higher than the actual temperature, and the diffusion of thermal energy affected the sensing accuracy of the nearby SHT11. Therefore, in the final version the SHT11 sensor was located further away from the chemical sensors.
The energy consumption of sensor board could be further reduced by eliminating the DC-DC converters. In our prototype sensor board, two DC-DC converters were employed to provide three kinds of voltages: 3 V, 5 V, and 12 V. However, since the DC-DC converters typically have low efficiency (around 70%) [12
], a large amount of energy was wasted during the voltage conversion. Therefore, the unification of rated voltages among several sensors is recommended because the DC-DC converters can be removed from the sensor board.
For the base hardware to host the integrated sensor board, an IEEE802.15.4-based sensor node [13
] was used. The board basically consisted of an MSP430 MCU and TI CC2420 [14
] transceiver offering a data rate of 250 kbps at 2.4 GHz. The base board provides a 51-pin connector for the add-on sensor board and to forward sensing data to the base station.