Linear/Mixed-Signal Chips: Jim Williams
Understanding analog is an art

At a recent luncheon talk, Jim Williams offered an anecdote to describe a significant shift in the world of analog design.

The staff scientist at Linear Technology Corp. (LTC) couldn't seem to keep his Jaguar XKE-with 167,000 miles on it-in tune, although he had the original repair manual and followed its instructions exactly. In desperation, he took it to an old-salt mechanic, who got it humming and purring again. "The manual was written for a new engine," said the mechanic. "Not one with the wear and tear yours has had. You need to change the viscosity of your oil."

Similarly, Williams concluded, there are a lot of textbooks on analog. There are a lot of digital cells that claim to replace analog circuit functions. There are a number of EDA tools that claim to encapsulate analog expertise. None of this stops Jim Williams' phone from ringing at Linear Technology Corp. "A lot of the time, they don't even know what you make," Williams complains. "All they know is that you do analog."

Fundamental changes in the electronics industry-viciously brutal time-to-market pressures combined with a distressing deficit in analog expertise-have changed the nature of the applications work he does. Instead of building a generalized circuit for one of LTC's part types and publishing an app note for everyone to marvel at and use, at least 50 percent of Williams' applications circuits are now built directly for customers who will simply add it-without modification-to the product they need to ship.

"You ship a piece of copper-clad board from your lab (sometimes with wires dangling from it) and if it works in your customer's system he 'postage stamps' it into his design," said Williams. "You're designing circuits for customers who are going right into production."

"That's a big change from 20 years ago," he said. "Twenty years ago, linear companies would not have wanted that kind of responsibility. Now, the service issue is dominating the business. The circuits you ship have to work in your customer's product." That makes analog design very targeted, very dedicated, he said. And it has a marked effect on how linear companies are run. "It means the value of your products [linear ICs] gets mixed up with the services you perform.

"Ten years ago, people generally knew what they were doing in analog; knew how to do it. If you showed them an interesting part type, they could probably figure out what to do with it. If they encountered you at a conference or got you on the telephone, it was to talk 'subtleties,' " said Williams.

"Lately, there's been a proliferation of people who are not particularly interested in analog but who are forced to do analog just to get their product out. It doesn't mean they lack the capability," Williams observes. "It's just that analog has become an item on their checklist. It's not a technology trend; it's a change in the user base."

Such a shift is undoubtedly distressing to someone who loves electronics-old electronics, analog electronics-as much as Williams. Though he seldom gets to spend much time relaxing there, his home laboratory is populated with working Tektronix 500-series oscilloscopes. He collects antique instruments, in fact, and has an electrometer that Lord Kelvin was said to have built.

Another prize is a 200-year-old Marine chronometer that's accurate to 1 ppm per month. Williams is fascinated with the way people solved measurement problems with the technology available to them at that time. "How did you hold 1 ppm 200 years ago?" he asks. "Questions like that make you truly interdisciplinary. You can't get too invested in the technology that's in front of you; you have to seek out all-encompassing solutions."

Like many engineers, Williams finds a certain beauty to electronics. The wall of his living room is graced by what, from a distance, looks to be a highly textured, vividly colorful tapestry. Moving closer, it becomes a series of populated electronic circuit boards, laid out flat in a frame. "The nose cone of a Minuteman missile," Williams said. "It was always armed, but the electronics ran so hot, that you'd have to cool it with liquid Freon to keep it from burning itself up. If the missile was ever fired, you'd have to disconnect the cooling system. The missile would have about 10 minutes to reach its target before the electronics overheated." It turns out, the liquid Freon was an excellent preservative for the color-coded resistors, the copper printed-circuit board traces, the gold and silver-cased transistors, and their green phenolyic substrate.

The hand-made digital clock in his dining room is a 3-D sculpture, using 14-gauge solid copper wire, copper tubing and plumbing pipes, which is in sharp contrast with some of the "rats' nest" try-out circuits that periodically pepper his test bench at Linear Technology.

Williams puts in a 60- to-80-hour work week. His time is equally divided between solving specific customer problems and completing long-range design projects. Some of it is done at home; some of it in his laboratory at work. "They have a tendency to transmute," he said. A "total disconnect" from the engineering world is skiing with his wife, he said, or playing with his son. Most of the time, with customers begging for assistance, he can't disconnect.

Such pressures can make Williams leery about answering his telephone. The callers have immediate analog-applications problems to solve. Their tone is often desperate. "The time between the customer's first call and your shipping him a copper-clad board to try out is sometimes no more than a week," said Williams. But because the caller's analog knowledge is increasingly limited, it is impossible to tell sometimes whether he is specifying the right things, and-as with all custom design-there is always the danger of delivering the wrong thing.

Board-level products offer a very flexible platform for analog circuits. "At the board level, you're working in the customer's domain," said Williams. You can get a sample circuit out the door very quickly, and can make easy changes if it doesn't seem right the first time. If you understand the customer's problem, you can win a slot for your company's parts.

Measurement and control, and low-end signal amplification always remain a challenge, Williams confirms. "Often, you have a 'bag of tricks' you can rely on to solve a customer's problem, but sometimes it's something original that will take some time and thought.

"The end user doesn't care if the solution is a single chip or not, especially if he's on a time line," said Williams. The goal of the LTC customer usually isn't to come up with one integrated chip, he believes. The goal of the customer is to ship on time-with performance and cost parameters intact.

LTC targets the ICs it decides to build very carefully. "We ask, 'Can we do this? Should we do this?' " Williams said. "You want to be careful of doing too many circuits that others can do, because then you're selling a commodity. People will obviously get better at putting big hunks on one chip," he acknowledges, "but can you get it out on time? Is it flexible? Can you test it? And can you make any money at it?"

Despite the pressures Williams feels from the scarcity of analog designers, he is not convinced the engineering colleges and universities can alleviate the problem. The best analog designers showed their interest in this long before college, he believes. Ironically, the situation is characterized by a kind of either off-or-on digital metaphor: chances are you are either an analog engineer or you're not. "If you're a dog," he quips, "you're gonna sniff trees."

Return to: TimesPeople 98