Energy Efficient Wireless Sink Node for Monitoring of Snow Environment

Conventionally, sink node is considered to have large hardware and energy resources; however, many times sink node is working in same conditions as source nodes, especially when deployed for monitoring of the snow environment. In this paper, an effort has been made to practically realize a sink node which is energy efficient and cost effective for monitoring applications. To save energy, the Main Power Module is designed to provide controlled powers to sensors and sub-modules. The paper discusses design aspects of the sink node and its long-term field evaluation with environmental sensors, especially the Snow Depth Sensor of MaxBotix. Field performance of Snow Depth Sensor has been enhanced by Euclidean Minimum Distance filter which improved the correlation of data to 0.997. The proposed design helps to achieve energy consumption of 42.72mWh which is significantly lower than the previous work. The reliable working of the sink node in the long-term field evaluation indicates that snow environment can be monitored at less expense of energy by employing proposed sensors and the specially designed sink node.


Introduction
Wireless Sensor Networks (WSNs) have been conceptualized as the collection of large number of inexpensive nodes which have huge potential to monitor the environmental parameters at low cost. According to Akyildiz (2002), the wireless sensor nodes have limited resources in terms of memory, processing power, communication range and energy. The communication module onboard WSN node has small range, so by utilizing various energy efficient routing protocols ( Erlacher (2016) has embedded wireless sensor nodes in the snowpack before release of avalanches for better understanding of inner movement of snow-mass. The comparative evaluation of state-of-the-art ultrasonic snow depth sensors has been carried out by installing the sensors of M/s Campbell Scientific (SR-50) and M/s Judd Communications at nine sites across the United States (Ryan, Doesken, & Fassnacht, 2008). These sensors are field proven, but they are expensive and consume more energy. Energy consumption has been an important concern in field based WSN applications.
The researchers (Alippi et al., 2009;Raghunathan, Ganeriwal, & Srivastava, 2006) have focused on maximizing the lifetime of nodes, which derive energy primarily from the small capacity batteries. Simjee and Chou (2008) used super-capacitors and efficient rechargeable batteries in sensor nodes as they can be replenished from natural resources in the deployed areas by tapping solar power, wind energy and vibration energy ( Kolios et al., 2016) envisaged to send data only on occurrence of events. So, energy consumption has to be minimized in the deployment of Wireless Sensor Nodes. The source nodes of WSNs are energy efficient but they have lower hardware resources. The recorded data of source nodes are processed and disseminated to central station through sink node; thereby requiring more resources. Standard Arduino Mega Board meets this requirement of practical sink node, but consumes ≈ 336mWh of energy in power down mode. Usually, it is presumed that source nodes of WSNs have energy constraint but sink node has large resources in terms of energy, computational power and hardware features. However, energy consumption remains a challenge when deployed for environmental monitoring of snow bound regions of Indian Himalayas where all nodes are subjected to same environmental conditions.
In the past, conventional nodes have been designed with other popular microcontrollers for various types of monitoring applications. Gutierrez et al. (2014) developed automated irrigation system by monitoring the soil moisture and temperature of root zone of plants. The system was built around PIC24FJ64GB004 microcontroller, which had deployed Soil Moisture (VH400), Temperature Sensor (DS1822) and water pumps. The system was tested in a crop field for 136 days in the temperature range of 15ºC to 38ºC. The microcontroller-based monitoring system consumed 80mAh (360mWh) in operational mode excluding the power of water pumps. Yan, Sun, & Qian (2013) presented energy aware sensor node design using MSP430F microcontroller. This paper presented "node-level energy saving" by adaptive transmission, power setting, periodic sleep/wake-up scheme, and "network-level energy saving" by adaptive network configuration. The power saving has been confirmed by experimental tests conducted in ambient conditions of greenhouse environment. Bengherbia et al. (2016) proposed FPGA based WSN sink node. The proposed architecture for the Sink node was built on MicroBlaze soft-core processor, and XBee communication module for communication. The proposed node was tested and implemented on the Digilent Basys3 Artix-7 FPGA board. Test application for temperature measurement (20ºC) in ambient conditions had been developed to ensure the proper functioning of the sink node. The proposed FPGA board was operating with power supply of 5000mWh, which is very large for the standalone applications of snow monitoring. The novelty of this research work lies in the design of S-Node by which the energy consumption has been brought down to 42.72mWh and field performance of MaxBotix Snow Depth Sensor has been improved by incorporating Euclidean Minimum Distance (EMD) Filter. Design methodology helps to monitor the snow environment with less energy consumption. Energy efficient sink node has been evaluated experimentally for long term at remote site in Indian Himalayas at low temperature up to -9.8ºC.
The remaining article is framed as below.
• Section 2 describes the design approach and various hardware components used along with different modules/ components of S-node.
• Section 3 presents the results and discussion in detail regarding the present work.
• Section 4 summarizes the major findings as conclusion.

Design Approach
The energy efficient design of the proposed S-Node is explained in the following sub-sections of Major Hardware Components & Design Methodology, Working Sequence and Modules/ Components.

Major Hardware Components & Design Methodology
The S-Node is designed around

Working Sequence
On triggering of hourly Alarm, RTC interrupts microcontroller by pulling down INT0 pin and ISR is executed. The warm-up time of 15s is enabled to settle the signal of sensors to steady state output. The values of sensors are sampled and stored in SD card with date and time stamp. The data are transmitted through GSM link to central station at interval of one hour. It takes maximum 02 minutes for warm up, acquiring data, storing sample values and sending transmission through GSM module. At end of next hour RTC wakes up microcontroller. So, low duty cycle of 3.33% is achieved in reference to periodic transmission of one hour.

Modules/ Components
The S-Node relays the data of remote sensors nodes to the central station. For economical and quick turn out design time of S-Node, COTS Arduino shields are used. ATmega2560 microcontroller is the heart of S-Node which is used in Arduino Mega board, therefore, programmable by ArduinoIDE. A large number of C/C++ libraries are available at GitHub repositories for all the sub-modules of S-Node, thereby, reducing the effort in embedded programming. The current consumption of ATmega2560 in power down mode is small 7.0µA @ 5V, 8MHz at operating temperature of 25ºC and it works at lower temperatures upto -40ºC. Though, ATmega328 has smaller current consumption 4.2µA @ 5V, 8MHz in power down mode, however, ATmega2560 at clock frequency of 8MHz has been preferred for design of S-Node due to higher number of resources. The block diagram of the S-Node is given in Figure 1 and hardware layout of the prototype is shown in Figure 2.  The details of customized module and standard modules are given in following sub-sections:

Main Power Module
Main Power Module is specially customized to implement the provision of switched powers to all modules and sensors. ATmega2560 controls the Main Power Module for activating/ snapping the power lines of the shields and sensors/ peripherals. Whole switching is achieved with combination of N-Channel Enhancement Mode MOSFET (VN2222LL) and P-Channel HEXFET Power MOSFET (IRF9530). Digital high signal at Gate Terminal of VN2222LL drives IRF9530 to pass the voltage at source terminal to the active load. Heat losses at higher currents are reduced as the source-drain resistance (0.2Ω) of IRF9530 is very small. Linear regulators and IRF9530 are chosen in TO-220 package which helps in easy mounting of heat sinks. The Main Power Module derives switched voltages from 12V external supply for operation of all the peripherals and sub-modules. Low-Dropout Voltage Regulator MIC2937A provides +5V round the clock to the microcontroller and RTC. The regulator has very low quiescent current of 160µA, which ensures low power consumption of the S-Node. Fixed linear regulators LM1117I-3.3, LM7805CT and LM7809CT generate switched voltages of +3.3V, +5V and +9V to power Zigbee Module (CC2530), MaxBotix Sensor and GSM Module, respectively. Switched voltages of +5V and +9V are designed with external pass elements TIP125 to enhance the surge current capacity of the power system. Additionally, external voltage (+12V) is bypassed to power up Multi-parameter Sensor (FWS500).

Communication Module
Communication segment has two sub-modules GSM SIM900A Shield and CC2530 Shield. GSM Module is built on low-cost dual band GSM/GPRS engine-SIM900A that works on frequencies 900/ 1800MHz. The module is directly operated with TTL voltage levels compatible to ATmega2560. The module typically consumes average current of ≈80.0mA from 9V and has intermittent high surge power needs for communicating with GSM network. AT-commands have been implemented on ATmega2560 for automatic network identification, registration and sending of SMS in ASCII mode. The module is connected to Serial1 port of ATmega2560. S-Node stores and sends the data to central station at one-hour interval as per frame structure given in Table 1. By using positional encoding and ignoring special characters, frame size has been brought down from 88 to 72 ASCII characters. As single SMS contains maximum 160 ASCII characters, so, the data of at least two hours are transmitted, thereby, providing redundancy of one hour in data transmission. Even if SMS is not received for one hour, the previous data can be replenished. The position of each field ASCII character helps to decode the data at central station.
The node to sink node communication is with Zigbee compliant CC2530 radio module of Texas Instruments. The module is compliant to IEEE 802.15.4 standard. Serial2 port of ATmega2560 drives the Tx and Rx lines of CC2530 through the logic level translator which has been realized using 2N7000 N-Channel Enhancement Mode MOSFET. The power on switch of CC2530 Shield configures the radio module for Point-to-Point or Broadcasting Mode of transmission.

Sensors & Standard Modules
Multi-parameter Sensor FWS500 has been selected viz-a-viz individual sensors for saving the energy and making the system compact.

Site Selection & Sensor Parameters
The design so realized had been subjected to long term testing at remote field site from 9th February to 8th May, 2017 at Dhundhi (Figure 3), which is located in Pir Panjal Mountain Range at Latitude (32º21'19.5"N), Longitude (77º07'42"E) and Altitude of 3050m above mean sea level. Compact visualization of all environmental parameters has been presented in Figure 4 in normalized scale for the period of evaluation i.e. from 09th Feb, 2017 to 08th May, 2017. S-Node processed raw value of MaxBotix Sensor to derive filtered value using EMD filter and transmitted both values to central station.

Energy Consumption in S-Node
By employing above said design methodology, working sequence and modules/ shields, the overall energy consumption of S-Node has been reduced. The S-Node achieves energy consumption of 42.72mWh in standalone mode which is much lesser than 80mAh of Gutierrez et al.  Table 2.

Euclidean Minimum Distance (EMD) Filter
The frequent transitions in output signal (>4.5V) of the MaxBotix Sensor are expected due to the reflections of ultrasonic signals from the falling snow-flakes. The probable reason of inconsistency is attributable to the most likely filter implemented in the firmware of sensor which prefers the signal of closest target when it receives multiple reflections of similar amplitude. In order to improve the error so associated, the readings are staggered throughout the hour by taking 5 samples at an interval of 5 minutes. At end of hour, sample size [S] of 60 values is available for processing. As snowfall is non-uniform, there are some instants in [S] when reflecting surface distance is measured accurately. Sensor value having minimum Euclidean distance to previous value is selected as final value and higher values are discarded as spurious values. Methodology of recording is given below:- Figure 6. Scatter diagram of raw and filtered data with manual data Let Xold be the initial value before snowfall starts and Xnew be the subsequent hourly reading to be transmitted. On hourly trigger, take sample size of 5 readings [S] for measurement consistency. If the median value in sample [S] is not high (<4.5V) then finalize the median as new sample to be transmitted, otherwise, increase sample rate to 5 minutes till end of hour and select the value having minimum Euclidean distance to previous value of sensor. The variation of filtered snow depth after incorporation of above process sequence is also shown in Figure 5. The algorithm for implementation of the process sequence in embedded program of S-Node, is given in Algorithm 1.

Comparison of MaxBotix Data
The MaxBotix Sensor was mounted at height of 3.50m in Dhundhi field site. So, output readings of MaxBotix Sensor were offset adjusted and compared with manually recorded readings at 3-GMT and 12-GMT. Figure 5 plots the variation in the snow depth of MaxBotix Sensor and manual snow depth. Scatter diagrams have been plotted for the better insight of measurements. Figure 6 shows the variation of raw snow depth of MaxBotix sensor w.r.t. manual snow depth. Due to spurious values of raw data, low correlation coefficient (0.075) is observed. Figure 6 also shows the variation of filtered data w.r.t. manual snow depth after incorporation of EMD filter. We observe high correlation of 0.997(RMSE=0.054) due to the removal of spurious values.
Design of S-Node is able to achieve low power consumption of 42.72mWh (Figure 7 Figure 6) of correlation coefficient (0.075). However, the minimum Euclidean distance filtered data showed strong, positive and linear relationship with high value (Figure 6) of correlation coefficient (0.997) when compared with the actual snow depth. Therefore, by incorporating the proposed design methodology and combination of sensors, the snow environment can be monitored with less expense of energy.

Conclusions
This paper reported the development of S-Node which is designed around COTS shields and open-source Arduino-IDE. S-Node is able to integrate many types of environmental sensors. It can supply switched power to all types of sensors as per the operating voltages with the customized Power Module. S-Node is able to achieve the power consumption of 42.72mWh which is significantly lower than the previous work. It consumes 62.05mWh energy to operate for one hour while recording environmental parameters from Multi-parameter Sensor (FWS500) and MaxBotix Snow Depth Sensor. S-Node worked reliably at remote observatory at low temperature of -9.8ºC. All events of snowfall and day-to-day variations were recorded in the period of evaluation. By using customized frame structure with positional encoding and energy efficient design, consistent data were recorded at remote and central stations. The performance of low cost MaxBotix sensor is further improved by incorporating Euclidean Minimum Distance filter which improves the correlation to 0.997 between the output of MaxBotix Sensor and actual data. Finally, the proposed hardware configuration of sink node (S-Node) and sensors can be deployed for practical and economical monitoring of snow environment at energy constrained remote locations like Pir Panjal Mountain Range of Indian Himalayas.