The drive to integrate analog circuits onto systems-on-chip in increasingly impossible processes has revolutionized analog design.
The drive to integrate analog circuits onto systems-on-chip in increasingly impossible processes has revolutionized analog design. Gone are most of the familiar theories of operation, topologies and tools that generations of analog designers have learned. The approach is entirely new, built upon the stuff almost everyone tried to avoid when they were studying circuit theory.
Teresa Meng, a professor of electrical engineering at Stanford University, put it succinctly in a panel at this month's Design Automation Conference. The new concept in analog design, she explained, is to ask how few analog transistors are absolutely necessary to the design, and how much can be done in digital circuitry.
The concept is already familiar. It gained its foothold in the design of partial-response, most-likelihood (PRML) read channels and sigma-delta converters, where the front-end circuitry is analog, but almost immediately the signal becomes digital, and nearly all the signal conditioning and processing are done to the right of the analog-to-digital conversion stage.
We have also seen the idea increasingly applied in the periphery of more-conventional analog designs. Digital circuits have been used more often over the last few years to trim, calibrate or self-adjust analog signal paths. There's even been some though not nearly enough progress toward digital self-test of analog circuits.
But Meng is thinking of an entirely different order of digital intervention. She offered the DAC panel two op amp approaches:
One, a traditional op amp, had perhaps two dozen analog transistors, lots of signal paths and everything based on feedback to linearize the device.
The alternative, entirely different, had two analog transistors arranged as a gain stage. All of the amp's linearization, gain adjustment and other jobs were handled digitally most of them in a complex DSP feedback loop.
Meng said the digital circuitry, despite its complexity, was smaller and lower in power than the analog transistors it replaced. And the device could be made to yield in a deep-submicron process.
This is the future that some analog experts say is already upon us. As another case in point, expect to see a huge increase in the use of Class D amplifiers to replace linear amps in all kinds of applications not just in audio output stages.
Ron Wilson covers microprocessors, programmable/reconfigurable logic and the chip design process. He can be reached at firstname.lastname@example.org.