Researchers at MIT in coordination with Texas Instruments (TI) have developed a new energy-efficient microchip for low-power portable devices. The microchip, which is still a proof of concept, was demonstrated on the TI MSP430 and operates at 0.3V, which according to an MIT release is up to 10 times more energy-efficient than current technology.
MIT researchers Anantha Chandrakasan, professor of electrical engineering and graduate students Joyce Kwong, Yogesh Ramadass and Naveen Verma of the Department of Electrical Engineering and Computer Science (EECS) faced two challenges in getting the chip to operate at a low voltage.
“One challenge is that some forms of logic just don’t work at low voltage,” says Dr. Dennis Buss, chief scientist at TI. “You either have to eliminate those kinds of logic or modify them so they do work.” When the logic is changed to allow for lower power consumption, the circuit doesn’t operate at ultra-high speed anymore. “I should point out at low voltage these parts don’t operate at 1 GHz; they operate at 10 MHz or even 1 MHz at ultra, ultra-low voltage,” he says.
The second challenge in designing the energy-efficient microchip, according to Buss, was controlling the range of variance present in these chips, because imperfections in the silicon chip present more problems at lower voltage levels. “When you're operating in some thresholds, the drive current is exponential in the voltage, so then very small change ratios in the threshold voltage can give you very large fluctuations in drive current and in delay,” says Buss.
According to Buss, in order to optimize the design of these microprocessors new techniques using statistical timing analysis (SSTA) need to be developed. “People are starting to use SSTA in normal designs but since the variance in delay is so much larger at low voltage I’m not sure and the students aren’t sure whether the tools being developed today will be adequate for this more demanding challenge,” he says.
The microchip will most likely only have applications in portable devices — specifically portable network devices, military devices, communication devices and especially implantable or battery-operated medical devices. “In the same way electronics revolutionized computing 30 years ago and the same way it revolutionized communication 10 years ago, my guess is in the next 10 years electronics will revolutionize medical electronics,” says Buss.