Energy rationing for harvesting: power consumption limits for wireless sensor nodes

Figure 1 shows an example network and the subsystems of each node. For ease of deployment and lower installation cost, each node is required to be able to communicate wirelessly. To reduce communication overhead and response time, we want the nodes to be able to process sensor data locally and control actuators. The cost of routine maintenance (e.g., battery replacement, etc.) for a large number of nodes can be extremely high. Ideally, the sensors would work continuously for several years relying only on stored/harvested energy.

The choice of sensors, radios, and microcontrollers (MCUs) depends on the nature of the application. This article focuses on sensor networks in office environments, targeting applications such as energy management, security, or resource planning.

Energy and Storage

Light energy is usually the most abundant form of ambient energy in indoor environments. Some modern solar cells (made of amorphous silicon) can generate about 5uW/cm2 under a 200 lux fluorescent light source. Table 1 lists estimates of energy capture rates, which show that a 10cm2 solar cell can generate 70–120 uW.

Table 1 Approximate energy acquisition rates under typical indoor fluorescent lighting conditions

Microthermal generators use a certain temperature gradient to generate electricity. However, to produce a power density of 15uW/cm3, the thermal harvester needs a thermal gradient of about 10oC. Many application environments, especially indoor environments, do not have large temperature fluctuations. Therefore, the applicability of thermal harvesters is limited to these environments.

Some of today’s vibration energy harvesters require accelerations of about 1.75–2.00 g (which are generally not that high in indoor environments) to generate 60 microwatts of power.

The on-board capacity for energy storage is very limited, and the opportunities for harvesting ambient energy are also limited, so the sensor needs to use energy very sparingly. For example, a solar cell with a battery capacity of 100mAh can get 70uW to provide power for half of the 10-year node life. The node must keep its subsystems working and the average power consumption must not exceed 39uW.

Node Subsystem

MCUs, radios, sensors, and actuators have very different power/performance characteristics. Meeting the system power budget requires the sensor node to manage its subsystems in an optimal way. Figure 1 shows some of the subsystems used to implement a node.

Some modern low-power MCUs have a peak power consumption of about 345uW when operating at a clock frequency of about 1MHz. Assuming that the sensor data processing requirements are generally medium, the MCU's duty cycle can be extremely small (for example: less than 1%) to reduce the average power consumption.

Sensor nodes usually transmit information such as physical phenomena and related control messages at a relatively low rate. Table 2 summarizes the salient features of some important low-power wireless communication technologies.

Table 2 Comparison of some low-power communication architectures

The power consumption numbers listed in Table 2 are intended only as general guidelines for system design. As transceiver designs evolve, they consume less power. It is important to consider all aspects of the design when choosing a transceiver architecture. Wireless local area network (LAN) transceivers consume less energy per bit than Zigbee® transceivers, but they are optimized for higher data rates and have higher peak power consumption.

Some examples of sensors relevant to indoor applications include: thermometers, temperature sensors, microphones, and passive infrared sensors. Some current temperature and humidity sensors and microphones have a peak power consumption of about 70–80uW. Some passive infrared sensors that can detect human activity generally have a peak power consumption of 100–500 uW. Temperature and humidity sensors monitor slowly changing phenomena and operate at a low duty cycle, while other sensors used to detect motion can be turned off to reduce detection performance. In many applications, sensors require more energy than data processing or wireless communication. Therefore, meeting the system power budget requires innovative methods to manage sensors.

in conclusion

Despite tremendous advances in computing, communications, and sensing, the lack of adequate power and energy remains a daunting challenge in implementing wireless sensor networks. Technological advances in energy harvesting and storage are easing the power bottleneck, but the demands of end applications are pushing it higher. Closing this persistent power-demand gap requires a system-level design approach that optimally compromises performance to achieve energy savings while maintaining a minimum quality of service. Future wireless sensor nodes will autonomously adapt to changing application demands and energy availability over time.