LONDON Researchers at IMEC are looking at the use of silicon transistors in the sub-threshold region of their operation as a way of pursuing ultra-low power goals. A future SoC for biomedical applications could have blocks designed to operate at 0.2 or 0.3 volts, researchers said.
The threshold voltage (Vt) is nominally the point below which a transistor turns off and this has typically represented a limit to voltage scaling as a source of improved power efficiency. Vt is typically 0.7-volt going down to 0.5-V in modern CMOS processes. However, in reality there is some sub-threshold conduction. Indeed, in normal transistor operation that sub-threshold conduction is viewed as leakage and wasted energy. If a transistor could be operated in this sub-threshold region significant power savings could be achieved.
But operating transistors at voltages below Vt is likely to difficult and its use will not suitable for all circuits or application cases. A key factor is that it comes at the expense of frequency performance, which can be a problem for digital applications. It is easier to apply to low-frequency analog but transistor operation it is also strongly dependent on manufacturing variations in such things as oxide thickness, junction depth, and body doping, which could make analog performance hard to calibrate.
Nonetheless power-saving is becoming the primary driver of much of IMEC's research. For example, ultra-low power operation is important to the operation of body-area network (BAN) radios for use in medical appliances, one of the selected areas of research at IMEC's outpost at the Holst Center (Eindhoven, The Netherlands).
The goal is to power wireless electronic systems from harvested energy, which currently offers about 100-microwatts of power. Unfortunately, it is hard to get even the simplest of electronic systems, including a sensor, microcontroller and wireless transceiver, below about 10-milliwatts, according to Bert Gyselinckx, general manager of the Human++ research program at IMEC. So batteries are likely to remain a reality for some time to come.
Gyselinckx's group plans to produce a short range RF transceiver ten times more power efficient than today's Bluetooth and Zigbee chips (see Video: RF uses one tenth the power of Bluetooth, Zigbee. This progress has been made by developments in radio architecture and circuit design applied to conventional CMOS process technology at 90-nm minimum geometry.
Giselinckx said IMEC is beginning to look at sub-threshold operation: "We are doing some work to characterize silicon. The issues around transistor operation are completely different at 0.3-volt. So we are starting with modeling the transistor behavior. It is not a full program."