Benchmarking-Based Investigation on Energy Efficiency of Low-Power Microcontrollers

As a real boost to competitiveness and productivity for sectors, such as health, agriculture, or industry 4.0, the Internet-of-Things (IoT) market is transforming business models and revolutionizing practices. A major challenge is to develop applications with high computing potential with limited resources. To satisfy both high-level performance and low-power requirements, a deep characterization method for microcontrollers is proposed since selecting the appropriate device is a central pillar of smart energy policy. The high-level analysis investigates the potential of various low-power microcontrollers with a benchmark utilizing a periodic duty cycle model of typical WSN applications. Afterward, deep characterization was carried out in two steps. The preselection phase is based on estimation whose results will serve as input for the second measurement phase. A highly accurate four-wire measurement involving guarding and shielding of all interfaces and the measurement cave is used for all measurements. To make the results comparable, a normalized power, defined as a measure of power per throughput, is introduced. The measurement characterizes the active power consumption depending on the supply voltage and frequency as well as sleep power, data logging power, and peripherals power.