ECE alumnus Abdallah develops robust, ultra-low-power ECG processor
Kim Gudeman, Coordinated Science Lab
- Alumnus Rami Abdallah won the International Symposium on Low Power Electronics and Design Low Power Design Contest with his robust, ultra-low-power ECG processor.
- His device is five times more energy efficient than current processors and is more resilient to errors.
- His innovation may have benefits in medical technology.
An ECE alumnus has developed a new chip that extends the battery life of portable electronics while maintaining reliability. The work could enhance the practicality of personal medical devices that allow physicians to remotely monitor coronary heart disease, atrial fibrillation and other medical conditions.
ECE alumnus Rami A. Abdallah (PhD ’12) recently won the International Symposium on Low Power Electronics and Design (ISLPED) Low Power Design Contest with his robust, ultra-low-power electrocardiogram (ECG) processor. The contest recognizes “innovations targeting power efficiency” and submissions must be implementable, according to ISLPED’s website. Abdallah's work is the first to address resiliency in low-power devices.
“The main requirement for portable electronics is that they are low power,” said Abdallah, who was affiliated with the Coordinated Science Lab during his time at Illinois and who now works for Intel. “If the device is implanted near your heart, you clearly don’t want to be switching out the battery all the time.”
Abdallah’s silicon prototype, developed as a student under ECE Professor Naresh R. Shanbhag, is nearly five times more energy efficient than current state-of-the-art processors.
It is also more resilient to errors—a problem that plagues low-power devices. At low voltages (<0.5 V), device variations increase by three to four times. Current design philosophy is to tighten these variations at large energy overhead. By employing statistical error compensation, a design technique developed in Shanbhag’s group, device variations are embraced and exploited to match the application-level requirements at low overhead, leading to significant robustness and energy benefits.
Abdallah used two processors in his chip—a main processor and a simpler processor—that run in parallel. If the main processor makes an error, the simple processor compensates for the error by employing techniques from detection and estimation theory. Because of the large design margin and low overhead, the processor allows for 16 times more voltage variation than the current standard while conserving energy.
The research comes at a time when spiraling health care costs and a growing physician shortage make it more difficult for patients to receive care for chronic conditions. They may also enable physicians to better treat patients through telemedicine and offer those living in remote locations better access to care. In addition, wearable electronics could be used for entertainment purposes.
“Rami’s work clearly shows the benefits of taking a communications-inspired approach to the design of energy-efficient information processing systems,” Shanbhag said. “This approach represents a paradigm shift from traditional von Neumann style computing to a Shannon-style information processing.”
Abdallah presented his research at the ISLPED ’12 conference in San Francisco July 30-August 1. The work was conducted as part of the Alternative Computational Models research theme in the Gigascale Systems Research Center and in collaboration with faculty from Illinois (including Douglas L. Jones, Andrew Singer, and Rakesh Kumar) and elsewhere.