In many applications, “pushing the design envelope” means faster, higher-resolution performance. So it is easy to forget that for a large and growing class of applications, such as personal medical instrumentation, slow and low-resolution is not only the meets the requirements, but also lets you establish different system architectures and approaches.
Of course, medical applications such as MRI and CAT scanners need lots of channels, bits, and MSPS. But for physiological monitors—blood pressure, blood oxygen, electrical and acoustic cardiac signals, and even brain waves—the situation is quite different. At the recent Texas Instruments Developer Conference (TIDC), Professor Rahul Sarpeshkar of MIT's Analog VLSI and Biological Systems Group lab discussed the bio-related ICs that his group has architected, built, and tested (www.rle.mit.edu/avbs). He repeatedly emphasized that bandwidths are on the order of 10 Hz, and resolutions of 8 or at most 10 bits (commensurate with 1% accuracy).
Designing for performance beyond these boundaries causes dramatic increases in power dissipation and die size, and ha a ripple effect into system architecture. Of course, it's not just biomedical applications that have these characteristics: many geophysical applications (seismic systems, weather monitoring, and acoustic monitoring) have them as well.
Before you “look down” at this slow, low-resolution world, remember that it's the world of all living organisms, including you. By constraining your design requirements to match, you achieve more than obvious, immediate benefits. It's an opportunity to step back and say “if my design only had to perform to this modest level of performance, how could I radically change the overall approach to the problem, rather than using the obvious choices? What other dimensions could I pursue, such as extreme low power, if I could do additional analog pre-processing and minimize digital and software-based processing?”
Take advantage of the unique target specs to rethink you entire design topology, instead of the standard “amplify, digitize, and process” signal-chain approach.
-x x x-