Dynamic voltage and frequency scaling for on-demand performance and availability of biomedical embedded systems

  • Authors:
  • Dejan Raskovic;David Giessel

  • Affiliations:
  • Department of Electrical and Computer Engineering, University of Alaska Fairbanks, Fairbanks, AK;Poker Flat Research Range, Fairbanks, AK

  • Venue:
  • IEEE Transactions on Information Technology in Biomedicine - Special section on body sensor networks
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The goal of the study presented in this paper is to develop an embedded biomedical system capable of delivering maximum performance on demand, while maintaining the optimal energy efficiency whenever possible. Several hardware and software solutions are presented allowing the system to intelligently change the power supply voltage and frequency in runtime. The resulting system allows use of more energy-efficient components, operates most of the time in its most battery-efficient mode, and provides means to quickly change the operation mode while maintaining reliable performance. While all of these techniques extend battery life, the main benefit is on-demand availability of computational performance using a system that is not excessive. Biomedical applications, perhaps more than any other application, require battery operation, favor infrequent battery replacements, and can benefit from increased performance under certain conditions (e.g., when anomaly is detected) that makes them ideal candidates for this approach. In addition, if the system is a part of a body area network, it needs to be light, inexpensive, and adaptable enough to satisfy changing requirements of the other nodes in the network.