Why does this segment of the industry attract such odd-ball personalities? In this inaugural installment of "The Scarlet A," Stephan Ohr maps out some of the ways that analog is different
That was the title of a column I ran in the analog pages of a now defunct magazine. In it, old salts like Analog Devices' Dan Sheingold talked about trying to keep vacuum tube-based instrumentation from drifting. Bill Pascoe of Mentor Graphics told about his work designing op amps into flow meters for sewage treatment plants. The picture we ran had him posed in front of a Sun workstation, a slide rule in his lap.
The late Art Fury did a column for me about his design work on Tommy Turtle, one of the very first electronic toys. Like a lot of analog engineers, Fury had a colorful career. He was an evangelist on behalf of power management, writing and interviewing well on the energy savings that would accrue from lights, appliances and entire buildings that are smart enough to turn themselves on when they're needed and turn themselves off when they are not. Fury did more to promote the notion of smart power (power and logic on the same chip) than any six marketing guys. Remember, too, that Art Fury was the one who launched the 555 timer at Signetics and got Ford to design sequential rear directional signals into the Mercury Cougar by ripping out the taillights of an old car, refitting them with SCRs (the same device that triggered Tommy Turtle) and driving the jury-rigged car to Detroit.
I identify with another guest columnist, an engineer with Analogy, who wrote about working for the DOD--and driving a Harley. In engineering school and on my first engineering job, I read the Village Voice and grew my hair long (I had hair then). I wound up designing video projection and sound systems for theaters in New York--not because it paid well, but because it was more fun than building a nuclear weapon. Now I wear the Scarlet A: I'm the Analog Guy in a world gone digital. Why is it that I always feel as if I'm on the outer edges of something?
We are NOT at the dawn of a new era, but in fact well into it: The Digital Revolution. And guess what, folks, it is NOT the death of analog. It is a whole new business opportunity. As Robbie McAdams, senior vice president at Analog Devices, pointed out in a keynote address at an Analog and Mixed-Signal Conference (and as Texas Instruments vice president Del Whitaker wrote again in an op-ed piece in EE Times) digital technology drives analog business. The digital revolution--personal and networked communications, multimedia and PCs--creates entirely new opportunities. These guys are not old salts. Nor are they hippy longhairs. McAdams and Whitaker are seasoned business types, every bit as 90s as the rest of America. Hold onto your stock certificates, they tell us; we are in for a ride!
Analog, it turns out, is where digital meets the real world. The currencies of analog design, the control of voltages and currents over time, are what are used to launch digitally manipulated signals onto a telephone line or Ethernet cable. They are what tune radios and cell phones, drive motors and hi-fi speaker cones. The basic building block is not a logic gate, but a signal-conditioning circuit, typically an integrating amplifier (an op amp), and guys who know how to use them--to juggle the tradeoffs among linearity, bandwidth and power consumption--are worth their weight in gold. (Maybe we ought to say that, instead of analog designers have more fun, analog designers make more money.)
The behavior of digital circuits can be represented abstractly, without reference to the devices or systems that implement them. Logic states--ones and zeros--are rather abstract entities that in principle can be implemented with flashing lights, CMOS switches or (as with IBM's early computers) clicking relays, whose coils are driven by vacuum tubes. Analog circuits, in contrast, are generally tied to specific implementation schemes. Voltages and currents can be represented by ones and zeros, but they are very physical things. Digital logic states can be represented by TTL (5-V) voltage swings, 1-V voltage swings or considerably smaller transitions. But ensuring that those voltage swings follow a linear ramp generally requires designers to focus on the capabilities of a specific fabrication process.
To be sure, the technology is changing. A 16-bit D/A converter based on a laser-trimmer R/2R resistor ladder and bipolar current sources would be too expensive to manufacture for consumer audio and would hardly maintain its trim. Sigma-delta technology--effectively a digital signal processor--implements the audio data converter in CMOS logic. (It uses a summing capacitor at the end, to be sure, but it is a digital logic circuit nonetheless.) The economies of scale enjoyed by big CMOS foundries means they can build an IC with 20,000 logic gates cheaper than a bipolar device with 32 transistors and a thin film resistor ladder. The best part is that sigma-delta audio D/A converters can routinely provide a 16- or 18-bit dynamic range--24 bits if you want it. In real-world audio terms, that kind of dynamic range is equivalent to recording the pit-pat of caterpillar footsteps on an airport runway while an SST takes off.
Similarly, the disk drive read channel is no longer a bipolar circuit that detects analog peaks coming off the read head. It is a digital CMOS circuit that takes multiple samples of the phase-shifted waveforms coming off a disk and, using a polynomial expansion code, makes an educated guess as to what the serial data pattern should be. This year, we're taking about circuits that can correctly guess all the bits in a 750-Mbits/second stream.
Thus, a lot of the cleverness we now see in mixed-signal design is in the ability to render analog functions in digital CMOS. It is not the same sort of cleverness that builds a pretty good filter with a fistful of resistors, capacitors and op amps, though the people who can do this intuitively--like National Semiconductor's Bob Pease or Linear Technology's Jim Williams--in the general scheme of things are golden. But even here there are caveats: While the digital world races headlong toward 0.15- and 0.13-micron semiconductor line widths, a lot of really interesting control work is done with cheaper 0.6- and 0.8-micron manufacturing geometries. In fact, the shrinking geometries may not help analog design because this reduces drive capability and signal-to-noise ratios. As Broadcom's Klaus Bult pointed out at to standing-room-only crowds at last year's ISSCC, for data converter design shrinking geometries can be a royal pain in the butt.
Those of you who wear the Scarlet A know what I'm talking about. No matter who's doing it, this column has only one theme: Analog is different. Whether the focus is on standalone building blocks (op amps or data converters), ASSPs (application-specific standard products like modems or disk drive read channels), market forces and strategies, venture funding or the incredibly interesting personalities this part of the industry seems to attract, it is still one theme: Analog is different. We'll recount the many ways in this space. For those of you coming to analog and mixed-signal design for the first time, we invite you to, at the very least, have fun.