How do engineers conserve energy in IoT devices? Let me count the ways…

Find energy savings in process technology, node scale, core architecture, hardware and software optimization, energy harvesting, and engineering grit.   

IoT devices collect data through physical sensors, process sensor input on a low level (e.g., filter noise, discard unwanted data, etc.), a control scheme, or execute an action (e.g., local indication, alarm, fail-safe actuation, etc.), store data, and transmit data either via wire or wirelessly (i.e., telemetry). These functions consume energy.

Figure 1: The layers of IoT computing and the communication between them. (Image: F. Samie, L. Bauer, and J. Henkel, “IoT Technologies for Embedded Computing: A Survey”, in CODES+ISSS, 2016.)

Methods to reduce consumption can include using a low duty cycle, sampling data as infrequently as possible without affecting the validity of data interpretation. Keep circuits in a sleep state or off until needed. Circuits for wireless communication can be woken up to transmit a compressed data file and immediately put back to sleep. Power consumption can be unpredictable, as a core can consume different amounts of power at different times, depending upon the application. Using sleep power states can decrease power consumption but take progressively longer time for the CPU to act. A lower voltage supply will reduce energy requirements (P=VxI). However, supply voltage can be so low that external noise competes with true signals. Operating at a higher frequency can also save power, because more cycles can execute in the same period of real time.  A drawback is that operating at higher frequencies can cause parasitics to increase, and more heat is generated. Another trade-off with operating voltages is that a higher frequency core necessitates a higher minimum voltage level. The challenge of lowest power consumption for computing power means more careful engineering with meticulous attention to detail.

Materials science also plays a role in balancing power consumption with higher performance. Ever-smaller technology scaling nodes have consistently lowered energy use but have become more complicated to implement. The core architecture and process technology make a difference in power consumption. Choosing the most power efficient architecture for a given IoT application is a challenge in trade-offs, as architectures offer different features. A newer process technology, Fully-Depleted Silicon-On-Insulator (FD-SOI), can optimize leakage in both active and standby modes, allowing one to adjust optimization for either power or performance dynamically, as needed.

Instruction Set Architecture (ISA) can affect power consumption, since some instructions can take more cycles to execute than may actually be needed. Software can be employed to optimize active modes by carefully managing or scheduling tasks. Compressing files for transmission so that the wireless circuit is active for the least amount of time also helps. “Race-to-idle” is a term that implies that execution should be sped up as much as possible so that execution (active) time is reduced (assuming that operating voltage is not adjustable.) One can also trickle-charge an IoT device via solar panels, harvesting byproduct energy from RF signals, or by converting rejected heat from a nearby device to energy.

Doubtless there are other ways to conserve energy, but the above illustrates several methods engineers use to lower power consumption. The real challenge is to select the best methods for your application and judge worthy design trade-offs, remembering that tactics in one area can affect your strategy in another.




Share and Enjoy:
  • Digg
  • Sphinn
  • Facebook
  • Mixx
  • Google
  • TwitThis