It was probably inevitable that the mid-June passing of two giants in analog chip design -- Jim Williams and Bob Pease -- would spark self-examination in the semiconductor industry as to whether analog design as a unique talent is hitting its sunset years. Sure, it may be some time before precision mixed-signal products from the likes of Linear Technology, Corp. go away, but with larger chunks of analog interfaces being subsumed within microprocessors and FPGAs, is there much sense in being an analog expert? Isn’t the world going digital anyway?
First of all, the whimsical rise of the steampunk movement should prove to us that anything antique can become retro-trendy before our eyes. If the revitalization of vinyl LP and cassette tape markets has displaced digital file sales in some market sectors, why can’t analog studies become de rigueur again in engineering schools? After all, the same audio snobs who insist a vinyl record sounds warmer (whatever that means), by reproducing the analog slope of the musical scales, are the ones who claim analog synthesizers are somehow closer to the earth than their digital successors.
But there is a more critical misconception shared by those who dismiss the continued evolution of analog interfaces in control applications. The dissemination of more powerful processing roles to disparate locations in a physical network means embedded computing interacts more with the physical world than traditional IT once did. Thus, data acquisition and processing in the analog domain will become more important in the years ahead. Those who say most mixed-signal tasks are behind us sound a bit like the pre-Einstein physicists who claimed all physics research was virtually complete in the years preceding the Michelson-Morley experiments.
Sure, we’ve hit slight speed bumps in integrating analog blocks in standard digital CMOS processes. Placing A/D or D/A converters directly on a digital chip doesn’t make a lot of sense beyond 12-bit resolution. Intellectual property cores for analog functions rarely go beyond a filter, op amp, or voltage-controlled oscillator. But that’s why semiconductor researchers are experimenting with special processes and unique substrates at feature geometries below 45 nanometers. Some special chip layers are oriented to extremely low power consumption, some toward high-voltage interfaces. In all cases, analog blocks can take advantage of the CMOS process gains.
Remember, a few short years ago, skeptics thought DSP blocks within an ASIC or FPGA had stopped at the level of the barrel shifter or multiply accumulator. Suddenly, new process and design efforts made it possible to embed a full DSP processor inside an FPGA or microprocessor. Today, the standalone DSP processor is dying as a result. It is likely that additional precision analog functions will be subsumed inside a standard digital chip soon.
But there’s more to be done in analog design on other fronts besides mere chip integration. Embedded intelligence everywhere requires a lot more than mere digitization of sound, voltage, light, and other electromagnetic aspects of the real world. New automotive and factory floor requirements demand accurate pressure-sensing and accelerometer functions. This has driven the market for MEMS components, which can serve as analog aggregators far better than traditional silicon. New work on 3D or hybrid substrates may allow better integration of MEMS and silicon chip components in one package, or may allow future precision accelerometers to be implemented in silicon.
It’s an inescapable fact that the cloistered digital domain of IT and consumer devices is getting more complex and more accurate all the time. But at the same time, embedded computers must interface with the analog physical world to a greater and greater extent. That means digital design does not advance in a way that leaves the analog world in the dust. Digital and analog design advance in tandem, perhaps not in lockstep as advances are made in one domain that eclipses the other, but certainly in ways that allow digital and analog to feed off each other. The best legacy Jim Williams and Bob Pease give us is the knowledge that analog engineering is on the verge of entering a new and exciting generation. It is hardly on its last legs.