February 17, 2010 –
Eltawil’s proposal earned an NSF CAREER award.
Cell phones and laptop computers never seem to hold a charge for as long as we would like.
Now, CALIT2-affiliate Ahmed Eltawil is developing technology for mobile devices that can analyze wireless conditions, recognize when lower power consumption is adequate and automatically reduce power accordingly. This allows the batteries to hold their charge longer without sacrificing reliability.
Eltawil, assistant professor of electrical engineering and computer science, recently won a prestigious $400,000 NSF CAREER award that will allow him to further develop his “Cognitive Power Management for Memory Dominated Mobile Devices” proposal.
Current mobile devices deplete their batteries so quickly because the enormous number of memory cells packed densely onto their integrated circuits require a high-supply voltage to function reliably.
The devices consume both dynamic power, which actually allows them to store and retrieve data, and leakage power, a wasted byproduct. “A lot of power is used and wasted to make sure the memory is dependable,” Eltawil explains.
He proposes a feedback loop for analyzing data to determine when the system can operate effectively using less power. In the buffering phase, before the data from the wireless channel is processed, they are analyzed to see if they meet certain quality standards. If they do, aggressive power management techniques automatically reduce power to the memory; constant monitoring allows the power to be increased if quality begins to diminish.
“Essentially, we’re using the knowledge of the channel to modulate both the power consumption of the device and the reliability of the hardware,” he states. “As long as the sum of the errors from the hardware and the wireless channel remain under a certain threshold, the data will still display perfectly.”
He is concentrating his efforts on device memory because it is becoming such a major component of the integrated circuit; up to 70 percent of a chip can be devoted to memory. “You can reduce power in the logic section but you’re not going to get as much bang for your buck,” he says.
“By lowering the power in the memory, we’re allowing the battery to run for a much longer time.”
— Anna Lynn Spitzer
