Sunnyvale, Calif. -- Taking a cue from the movie Back to the Future, designers at startup GTronix Inc. have developed analog signal-processing circuits that deliver higher performance at lower power levels and much smaller silicon areas than most of today's digital signal-processing solutions. The Fremont, Calif., fabless IC supplier says it is bringing local intelligence to application-specific sensory interfaces.

This description of the company direction reflects GTronix's focus on microsensors, and on adding varying degrees of local intelligence for information extraction, actuation and wireless communications, said CEO Hubert Engelbrechten.

Initially funded by the Georgia Tech incubator program, the company has created a portfolio of 15 patents (granted or applied for) and has series B funding by Menlo Ventures to the tune of $13 million. Microsensors represent of a 2.5 billion-unit market, said Engelbrechten, and many of these sensors have between 70 cents' and $1.50 worth of pure signal-processing content and another $2 to $4 of associated IC subsystem support.

$5 billion pot (and growing)
The microsensor market represents a $5 billion opportunity today, and is forecast to grow at nearly 16 percent per year over the next five years, according to market research firm Frost & Sullivan.

The need to miniaturize adds pressure on sensory processing to improve the human interface. But that increases software complexity and power levels, since the DSP chips have to run faster to keep up with the computational tasks, said chief science officer Paul Hasler. For example, a cell phone or PDA might have a fingerprint sensor for security, a speaker and microphone for conversation, an image sensor to capture pictures, a touchpad to capture data, a position sensor if GPS is built in and an accelerometer to protect micro disk drives from shock.

But all this has to be supported on microwatt power budgets, since features are always on and battery capacities are declining as systems shrink. Countering the trend to cram more and more analog functions into DSPs, Gtronix has developed analog signal-processing functions that can offload the DSP engine and reduce system power consumption while requiring significantly less silicon area than equivalent digital functions.

"By leveraging mainstream CMOS processes using floating-gate analog programming techniques, we are able to craft programmable adaptive structures," said Hasler. The analog signal-processing technology leverages a unique combination of three elements, he said: proprietary analog algorithms; a programmable floating-gate transistor structure; and a unique circuit design technique for creating signal-processing functions.