Biosensor networks are hypercritical of low-power design techniques

One of the continuing projects at IMEC, the European microelectronics research consortium, is the construction of a body-area network of biosensors wirelessly linked to a central data reduction and wide-area network interface device. Since the sensor amplifiers are not wired to anything, they must operate autonomously at extremely low energy levels. This is proving to be a benchmark design problem for the new discipline of low-energy chip design, engaging every level of design expertise from architectural planning to cell design.

The current sensor module contains two analog sensor receiver chips that perform amplification and initial filtering of biopotential signals. The analog signals are then passed on to a Texas Instruments microcontroller for A/D conversion and coding into a more compact form. The MCU sends the data on to an off-the-shelf Nordic Semiconductor NRF 24xx Zigbee radio chip.

This is a functional demonstration device, but it is just a step on IMEC’s power reduction path. Near-term plans call for a new analog front-end chip that will include a 12-bit, 100 Ksample/second successive-approximation analog-to-digital converter that will consume only 20 microWatts. This will be teamed with a new microcontroller that IMEC is developing from scratch.

Developing an MCU in this day of high-quality commodity MCUs might seem a bit strange, but IMEC program director Bert Gyselinckx explains that in this case it is an appropriate effort. “DSP for biomedical signal processing is a rather new area,” Gyselinckx said. “It requires about 100 MIPS processing power—intermediate between the entry-level DSP chips and the high-end DSPs—to perform the level of not only filtering but signal analysis that is required to avoid transmitting large volumes of data back to the central unit. But to reach the level of energy efficiency required for these applications, you must exploit a great deal of data parallelism.” The IMEC part thus will employ a novel architecture, and will use custom layout rather than conventional cell-based design techniques, based on the architectural decision to trade low-energy processing power for radio bit rate.

Similarly the radio is coming in for change. IMEC is looking at a new architecture for implementing the IEEE 802.15.4a specification for pulsed ultra-wide-band radio in the 3-10 GHz band. The design team believes that the transmitter can be switched off between pulses, as described in a paper at ISSCC this year, cutting transmitter energy consumption to about 1 nanojoule per bit.

All of these efficiencies will allow the sensor module to operate on scavenged energy. It’s not quite as simple as that, Gyselinckx explains, because scavenged energy usually comes in bursts—a few moments of sunlight, the occasional motion of an arm, and so forth—so simply scavenging is not enough. The power supply for the sensor module will probably include a microbattery with a capacity in the 100 microAmp-hour range, a small supercapacitor to handle the bursts of current drain from the pulsing transmitter, and the scavenging device.

Taken together, the sensor module will not only be a viable device for use in medical monitoring applications—it is being developed in conjunction with at least two major medical institutions in the US—but it will be a learning platform for extremely-low-energy design practices. In both roles it will be important.