The betrayal of constant power × time: finding the missing Joules of transiently-powered computers

Abstract

Transiently-powered computers (TPCs) lay the basis for a battery-less Internet of Things, using energy harvesting and small capacitors to power their operation. This power supply is characterized by extreme variations in supply voltage, as capacitors charge when harvesting energy and discharge when computing. We experimentally find that these variations cause marked fluctuations in clock speed and power consumption, which determine energy efficiency. We demonstrate that it is possible to accurately model and concretely capitalize on these fluctuations. We derive an energy model as a function of supply voltage and develop EPIC, a compiletime energy analysis tool. We use EPIC to substitute for the constant power assumption in existing analysis techniques, giving programmers accurate information on worst-case energy consumption of programs. When using EPIC with existing TPC system support, run-time energy efficiency drastically improves, eventually leading up to a 350% speedup in the time to complete a fixed workload. Further, when using EPIC with existing debugging tools, programmers avoid unnecessary program changes that hurt energy efficiency. CCS Concepts • Computer systems organization → Embedded and cyber-physical systems.

Battery-less zero-maintenance embedded sensing at the mithræum of circus maximus

“…In Republic, we use a static checkpoint approach [15,72], which inlines calls to a checkpoint library to copy the complete system state onto FRAM. To place checkpoints, we profile the energy consumption of different parts of the code [3] and accordingly inline checkpoint calls. At each call, a checkpoint takes place if the capacitor voltage drops below a threshold that barely guarantees the energy to dump the state on FRAM.…”

“…A further voltage detector turns the "power good" signal up to make the buck converter activate the device whenever the selected input comparator switches its output. As device activation is now separately controlled, we configure the output of the buck converter exactly to 2.1V, which represents an energy-efficient regime for both the MCU [3] and the radio [25] As in Republic, the converter operates in pass-through mode whenever the capacitor voltage falls below this value. The TPS62736 also features an independent "enable" signal that can be used by the MCU to disconnect from the power sub-system, effectively implementing a software-controlled shutdown.…”

“…Energy management in Republic was totally unaware of these aspects and only operated based on the net energy inputs from the harvesters. Existing literature, besides a few exceptions [3,35,39], offers little support for gaining or exploiting this information, and almost never includes peripherals in the picture.…”

“…One may argue, for example, that our design in Empire is enabled by a priori knowledge of energy demands, which is generally not available and may change at run-time. Tools and techniques to accurately gain this information are, however, emerging [3]. Moreover, as we learn from Lesson 3 in Sec.…”

Accurate Power Consumption Evaluation for Peripherals in Ultra Low-Power embedded systems

“…For ultra-low power systems, allowing intermittent execution, a variety of models has been proposed [22]- [24]. The EPIC modeling tool [25] supports temperature variation and clock drifts in power consumption estimation. These models do not address power consumption of the peripherals.…”