Barrie Gilbert on analog ubiquity

This article originally appeared on the ChipCenter domain at

http://www.chipcenter.com/analog/c001.htm

I have added a newer bio Barrie sent me, as well a nice picture of him with one of his lab assistants.

Analog Ubiquity

By Barrie Gilbert,

ADI Fellow; Manager, NW Labs, Beaverton, OR

Compared to the time we hominids have been around, the drama of electronics is still on page one of the first act. The dazzling achievements in this field so far, and its penetration into every conceivable application, guarantee that developments in electronics during the coming millennium will shape the very nature of human existence in unimaginable ways, assuming we can solve our chronic sociological problems and learn to live in harmony. Indeed, we are already totally dependent on electronic products. With further advances in complexity, accompanied by relentless miniaturization and inexorable cost reduction, electronics will slip even deeper into the fabric of our lives: silently, invisibly and unobtrusively. Even “naturally.”

The word “electronics” originally referred to signal generation, processing and display techniques which were largely analog in nature. Most automatic control systems were also analog, except relays, found by the millions in telephone systems, and other electromechanical devices. When I started my first job at a research establishment in the early fifties, the only computer to be found in this sprawling complex was a temperamental analog machine, chock full of vacuum tubes, whose needs were attended by a hushed priesthood. Apart from uncountable slide rules, well-lubricated by daily use, it represented the extent of computing at a daily level, though the digital giant was already stirring.

The bipolar germanium junction transistor was still a novelty and a curiosity. Scarce, enigmatic, unreliable, yet nonetheless tantalizingly promising, these little three-legged black cylinders were regarded as far too expensive and risky to employ in the next generation of speech encoders on which I was working, even though these systems employed a heap of shift registers and other logic. Instead of scores of 12AT7’s, the newer equipment used wire-ended trigger tubes (a kind of miniature thyratron) as logic elements.

However, within the space of a fleeting decade, numerous bipolar transistors, fabricated from the earth’s cheapest and most abundant solid element, silicon, in the form of wafers slightly less than an inch in diameter, were being combined in a single “chip” to make logic gates. MOS devices were being demonstrated in research labs around the world; the simpler PMOS types were more common, with faster NMOS just around the corner; finding ways to combine these in order to provide complementary polarity types - CMOS - would take a little longer.

With these developments, electronics was poised to undergo a transformation of massive and profound importance, destined to alter every human life, in ways, and to an extent, few people at that time - perhaps no one - could comprehend. Part of this impact would come from the possibility of making integrated circuits containing thousands of transistors at very low unit cost. But another trend of signal-processing and digital automatic control systems was threatening to eclipse erstwhile analog methods of performing these functions.

With the advent of the microprocessor in the seventies, the future of analog electronics for a while looked very bleak. I recall being asked, on joining Analog Devices in 1972, whether I’d be interested in designing a monolithic analog-to-digital converter. My reluctance to get into this initiative stemmed from the distasteful prospect of only hastening the demise of analog techniques. After all, I had accumulated a sizable collection. I didn’t welcome a world in which these would be obsoleted by a string of mindless ones and zeroes. It was a job-security thing. So I deflected that invitation, and instead pursued the perfection of analog nonlinear circuits. In retrospect, that stance was about as grandiose and futile as King Canute’s efforts to hold back the raging tides.

As this century of discovery and innovation draws to a close, everybody, even the school-kid - especially the school-kid - knows that the future of electronics is digital. That message oozes from every media outlet and permeates our education system, being taught in kindergarten classes (implicitly, by the toys and tools to which children are nowadays exposed) right through to graduate courses in electronics, which have for some years shamefully neglected to inculcate the very foundations of this discipline, often jumping directly to system-level VLSI design.

“Being Digital,” in all its glorious totality, is seen by people like Nicholas Negroponte of MIT’s Media Lab as something akin to the quest for the Holy Grail, or at the very least, an indispensable component of human life, from here onward. And of course, that much is largely true. Even this die-hard analogger advises his team members that, if a function can be executed in the digital domain, don’t even think about laboring over some neat, classical, elegant analog alternative. This is an inescapable modern truth, and it’s quite redundant to delineate all the reasons why. One is forced to admit that it’s an awesome truth when a company with a name like Analog Devices sinks serious money in a Digital Signal Processing group.

The question arises: Where is all this leading? If we try to imagine the world of 2100 - or even 2050 - we are bound to ask: Will analog design be a forgotten art? What will be at the heart of these new, ultra-sophisticated systems, throbbing with a certainty to which only digital engines can aspire? Even more billions of CMOS transistors? Yes, undoubtedly; we are already at the portals of gigabit memories. Even higher speeds? Unquestionably; clock rates in advanced systems may go unnoticed: from a philosophical viewpoint, they are only incidentally electronic anyway! Digital systems opportunistically employ electronics for three important reasons: electronic gates are exceedingly fast (picosecond delays); they are incredibly tiny (challenging viruses); they are vanishingly cheap (of the order of a penny per thousand). But their electronic heritage is rapidly becoming secondary, and it would make little difference, in principle, if the underlying technology were based on fluidics, optronics or conductive peanut butter.

Young designers can nowadays be quite effective in wielding 10,000-gate chunks of logic function without the slightest idea of how the gate-level devices work. That’s not necessarily a bad thing; indeed, it is one of the marvels of our age, and it is reminiscent of the way we have progressed in software from the era of tediously manipulating bits, one by one, then byte by byte, until today we nonchalantly string together reams of code involving many megabytes of memory, like so much DNA. Indeed, one is bound to wonder how soon the fusion between electronics and molecular biology will be upon us.

With the prevailing importance of logic function, there is necessarily a diminished concern for the underlying means on the part of those people seeking to achieve complex effects. I have proposed the term “epitronics” to describe contemporary digital electronics, meaning that the function “hovers above” its earthy silicon substratum, in much the same way as human intellectual activities transcend their physical framework.

In strong contrast, classical electronics may be described as Newtonian. It requires an approach to design in which the focus of attention remains entirely on physical devices and effects. The operational variables have dimension - volts, amps, charge, frequency, power - and these arise and propagate in elements that likewise have dimension - resistance, capacitance, inductance. Notice, too, that all of these bear the names of great discoverers: Volta, Ampere, Coulomb, Hertz, Ohm, Faraday, Henry. Considerations is energy - the most basic of all physical phenomena - abound, a sure sign of the deep roots of analog concepts.

Analog circuits settle into states without algorithms without recourse to the concept of number. These state-spaces are the autonomous solutions to equations describing the response of interconnected elements to their various excitations. Yet, they don’t “need” any equations; only we do, in order to understand analog systems. Right now, in millions of homes, cars, offices, airplanes, hospitals, laboratories and spacecrafts, analog circuits are solving massive simultaneous, nonlinear, differential equations, at rates up to several billion solutions per second, for every node voltage and branch current, without a single calculation. Analog circuits are thus inherently computational; for them, finding a square-root or a vector sum may take but a few nanoseconds.

The Newtonian quality of analog electronics runs like a silver thread all the way from PN junctions to communications networks. Some of the pre-requisites to a mastery of this quick-silvery and quintessentially physical domain include a thorough understanding of the semiconductor materials and manufacturing equipment, device physics and modeling, circuit analysis and relevant mathematical methods, including those of Taylor, Fourier, Laplace, Bessel, Kelvin, Boltzmann and many other pioneers.

There is one further important reason why contemporary digital systems depend on electronics. They touch, and in places overlap, the domain of analog signals, arising in the physical world. Thus, we need those A/D and D/A converters, that somebody else bravely designed, as language translators. And even the boundary between these disciplines remains fuzzy. Thus, a sigma-delta converter involves an inextricable confluence of analog and digital principles, which developed out of switched-capacitor techniques. A band-pass sigma-delta converter is both an analog entity, having frequency-selective properties like a classical filter, and a digital one, utilizing numerous logic gates. However, attention to the analog behavior of the individual devices is of crucial importance in ensuring performance and robustness.

Increasingly, analog ICs use technologies originally developed expressly for purely digital applications, and they rely on finding innovative ways to overcome the inherent weakness of such technologies when viewed in analog terms. A growing trend is the contemporary use of deep sub-micron CMOS in the realization of LNAs, mixers and IF sub-systems for telecommunications applications, operating at frequencies up to 2 GHz. It remains to be seen whether all of the problems related to a pure CMOS solution can be solved, or even whether all of the problems related to finding a pure-CMOS solution can be solved, or even whether there is any real cost advantage compared to using the bipolar content of BiCMOS processes. The question “Why bipolar?” will be left for another time. But there can be no question about the permanence of analog techniques in the future world. And, while transistor count in VLSI continues to climb, there will always be a place for analog cells comprising just a dozen active devices, maybe…even…just…One.

Barrie Gilbert is a Life Fellow of the IEEE , ADI Fellow, and a Member of the National Academy of Engineering.  Born in 1937, in Bournemouth, England, he pursued an early interest in the then-new “transistor” at Mullard Ltd, later working on first-generation planar ICs. Emigrating to the US in 1964, he joined Tektronix, in Beaverton, OR, where he developed the first electronic “knob-readout” system, and other advances in instrumentation. Between 1970-1972, back in England, he was Group Leader at Plessey Research Laboratories. In 1972 he worked as an IC  designer for Analog Devices Inc. and joined that company full-time in 1979, as their first Fellow. He now directs the development of high-performance analog ICs at the NW Labs in Beaverton, which he founded.

For work on merged logic (later called I2L) he received the IEEE “Outstanding Achievement Award” (1970); and in 1986 the IEEE Solid-State Circuits Council “Outstanding Development Award”, citing his earlier invention of the Translinear Technique. He was Oregon Researcher of the Year in 1990, and in 1992 received the Solid-State Circuits Council Award for “Contributions to Nonlinear Signal Processing”. He has received ISSCC “Outstanding Paper” awards five times, the ESSCIRC “Best Paper” award twice, and several industry awards for “Best Product”, etc.  He has written extensively about analog design and is a frequent lecturer. He has been issued over 100 patents worldwide and holds an Honorary Doctorate of Engineering from Oregon State University. He is a Member of the National Academy of Engineering.