Because embedded sensor systems dissipate part of their power in moving data between sensors and processors, it is important and interesting to develop techniques to reduce the energy costs of such communication.
Value-deviation-bounded serial (VDBS) encoding is a new technique to reduce power dissipation in bit-serial communication interfaces such as I2C and SPI. It reduces I/O power dissipation by exploiting the tolerance of many embedded signal processing applications to small errors in their input sensor data.
In an end-to-end evaluation of an OCR application, VDBS encoding reduces transitions (and hence dynamic I/O power dissipation) by 55% on average, while maintaining OCR accuracy above 90% for
previously-correctly-recognized text. And in a pedometer application, VDBS encoding reduces transitions on average by 54% in exchange for step count errors of less than 5%.
Status: This is an active project. If you are interested in getting access to the implementation of either the optimal VDBS encoders, the Rake VDBS encoder, or encoder VHDL / Verilog generators, please get in touch!
Collaborations: This is joint work with Martin Rinard (MIT) and Pier Andrea Francese (IBM Research).
- P. Stanley-Marbell, P. A. Francese, and M. Rinard. "Encoder Logic for Reducing Serial I/O Power in Sensors and Sensor Hubs", In 28th Annual IEEE Symposium on High-Performance Chips (Hot Chips'16), August 2016. [Download PDF]
- P. Stanley-Marbell and M. Rinard. "Reducing Serial I/O Power in Error-Tolerant Applications by Efficient Lossy Encoding", In 53rd Annual ACM/IEEE Design Automation Conference (DAC'16), June 2016. [Download PDF]
- P. Stanley-Marbell and M. Rinard. “Efficiency Limits for Value-Deviation-Bounded Approximate Communication”, In IEEE Embedded Systems Letters Journal, 7(4), 109-112, 2015. [Download PDF]
- P. Stanley-Marbell and M. Rinard. “Value-Deviation-Bounded Serial Data Encoding for Energy-Efficient Approximate Communication”, MIT-CSAIL-TR-2015-022, 2015. [Download PDF]
- P. Stanley-Marbell. “Encoding Efficiency of Digital Number Representations under Deviation Constraints”. In Proceedings of the IEEE Information Theory Workshop, 2009. [Download PDF]