InComEss introduces ultra-low-power wireless sensor node

Energy harvesting background

The ever-increasing efficiency of electronic circuits has recently brought up a qualitative change in the way wireless sensing can be implemented due to power consumption being reduced to the realm of environmental energy harvesting. Self-powered wireless sensors can now rely on capturing waste energy emanating from near physical phenomena such as temperature differences, mechanical vibrations or even residual radiofrequency from telecommunications, instead of requiring single-use batteries. The generated power is not only extremely low (typically from microwatts up to tens of milliwatts, depending on the source), but also tends to be uncontrollable and unpredictable. This variability challenges many assumptions taken for granted in the design of conventional measurement devices and requires a different conceptual approach.

In general, the harvesting strategy consists of accumulating energy in a small repository until a pre-defined operational threshold is reached, then sensors are activated until power is depleted or consumption is voluntarily turned off. Usually, the storage devices are supercapacitors which, even though can reach up to tens of Farads, are still several orders of magnitude behind common batteries.

Focusing on this measurement cycle imposes a paradigm shift in the understanding of these systems, since they work in a discrete mode rather than a stationary state, and so are better described in terms of energy required to perform a single measurement rather than the average power consumption over an arbitrarily long period of time. One consequence of this model is a tradeoff between the measurement frequency and energy consumption. Assuming a given charging rate, sensors that consume more must wait longer for enough energy to become available. Similarly, speeding up measurements requires lowering consumption. Thus, time can become a free parameter used to match an energy source with the consumption requirement. Another trade-off can be found between the system complexity and the energy consumption, as depicted below in Figure 1. If the supply is extremely constrained, sensors are relegated to having no control over the measurement frequency (1º scenario). As more power is harvested, more sophisticated features can be implemented such as synchronized measurements triggered remotely (2º scenario). Finally, an unbounded power source would translate to a more conventional measurement device with exact timing, a real-time clock (RTC), etc. (3º scenario).

With the least amount of harvested energy, the worst-case scenario, sensors could only spare to send the measurements as soon as enough energy becomes available, and this task would probably have to be broken into several subtasks in case the energy storage is very small. In this scenario, and especially considering that the nature of real-world energy harvesting is highly unpredictable, it would not be possible to follow a constant measuring period, it would just work when it can.

With more regular and predictable energy available, sensors could be turn on all the time waiting for an upstream trigger. This would effectively sync all the measurements, regardless of whether the transmission happens at a later stage and would require a harvesting rate higher than the standby consumption. With much more energy available, sensors could, additionally, keep a real-time clock synced with high precision and follow strict measurement periods, among other more sophisticated features.

Wireless sensor node developed in InComEss

With some months to go before the InComEss project can be wrapped up, it is gratifying to see that the team of experts has delivered significant results. In the frame of the project, an ultra-low-power wireless sensor node was developed, demonstrating the capabilities of the novel piezoelectric and thermoelectric generator materials. Achieving this objective requires a great deal of effort not only in consumption measurement but also in identifying where energy is spent to enable optimizations.

Consumption measurements were performed using the amperemeter Power Profiler Kit II, as shown in Figure 2, which features a logic analyzer that is used to record traces of the internal program. This makes it possible not only to calculate the overall energy expenditure but also to sort it out into multiple segments that directly correspond to distinct tasks or portions of the source code. Therefore, an energy profile is constructed, allowing developers to practice energy-aware programming and to focus optimization efforts more effectively.

Measurements are automated using the amperemeter software and Python scripts alongside several electronic equipment such as relays and LabJack data acquisition tools to reduce the testing time, allowing many more tests to be performed thus increasing the statistical significance of the results.

Figure 2: WSN prototype partially attached to the amperemeter.

As an example, one such consumption profile is presented in the next two figures below, with only a few states for simplicity. These states are:

SETUP: Initialization of the base hardware in the WSN.
SENSE: Sensor measurements.
SETUP_BLE: Initialization of the Bluetooth transceiver
BROADCAST: Wireless data transmission.