In an industry that has a reputation for high turnover, there's a surprising amount of personal longevity, especially at analog vendors. A case in point is Lew Counts, vice president of analog technology at Analog Devices Inc., who is retiring today (Oct. 31) after 38 years at the company. Counts recently spoke to EE Times about the changes he sees in process technology, products, foundries and people, and their implications for the industry.
In his view, it's all about IC process technology, where you uncover potential rather than discover new laws of physics. You need to use what you have to its utmost, but you also need to do more in order to go forward. "The fundamental point of Moore's law is to create more value and expand horizons," he noted. "View technology as something you can drive forward and change. Electronics has moved from the shrine to the mainstream. Arcane stuff has become commonplace."
At a deeper level, he said, "electronics is the projection of the human psyche; it's about ego. Our field has become a means of expression, with so much value in tiny chips. Content is in product, with the high degree of value now encapsulated for all.
"The real strength of the industry is process, and everyone else gets a free ride. It affects everything, not just density and feature size. If you don't have a proprietary version of a process, it won't be fully effective. A process encapsulates all the learning that precedes it. It is the collective wisdom of the organization." Big chips are not a point event, Counts believes, but are "the accumulation of so much work that has preceded them, and in so many disciplines."
Analog companies have to do a lot of that work themselves, since analog poses a distinct set of trade-offs, noted Counts. For a long time, the analog mind-set was to reject departures from the status quo. "But you can find a way to move industry forward by process, by squeezing a lot out of each generation. Moore's vision was right, but new processes were needed to lower cost and power per function. New processes are a direct result of knowing that process technology must be driven."
The industry, he said, "had a naive belief that cramming more in, and scaling, would lead to something significant. We have a durable model of growth and advancement, but we were not so sure of this 30 years ago. We now provide 128 channel data acquisition for ultrasound in a chip, improve image quality, [enable it to] be faster and be lower power." That example, he added, demonstrates one big change: " 'And' is the new critical word. You have to provide Benefit A and Benefit B."
Stepping back for a broader view, Counts said that the biggest surprise is to see "the whole point of connectivity and networking of individuals as a social phenomenon." Another trend is the "convergence of network science, silicon processes and sophisticated design through an interdisciplinary approach and a climate for entrepreneurs."
The biggest change, he said, has been "the universality of electronics, and sophisticated electronic products. The model has been powerful at generating unexpected things, because you can connect so many people."
What about the talent needed to make this happen? "There have never been lots of analog designers; there never have been a lot of people on the technical side," Counts noted. "Butand this is criticala few people can do the work of many." While higher education in the United States is envy of the world, with a system of private and public colleges that has been very successful, "electronics is worldwide, and the industry has challenges to find enough of the right talent. By pulling in this talent, a small percentage of the population can create so much value."
As for the debate on whether analog vendors will embrace fab lite or even fabless models, as their digital counterparts have done, "the fabless model has opened up opportunities for many innovative startups, and that's a good thing," Counts said. At the same time, "proprietary processes such as CMOS on SOI [silicon on insulator] are still needed for high-end applications. Foundries have to broaden their process offerings, such as to 50 V, since electronics will be everywhere."
Adding to the challenge, capital markets "are short-term oriented, but our industry has a longer time constant. The road map is a plan of the future, and it takes a long time to implement. Investment models can be precarious, in terms of payback."
What about the future of analog in an increasingly digital world? The "need to interface to the real world mandates analog," Counts said. "Analog circuits are still cheaper for some applications. Look at [the work under way to use] CAT5 [Category 5 cable] and analog to distribute video, compared with HDMI."
Looking ahead, Counts noted, "We have moved to having system experts, in addition to circuit and IC experts." Conventional, "ideal" op amp thinking, he said, is of no use in tackling some problems, when the real issue is error management. "You need to get information into digital form with no errors. You need to understand intermodulation distortion, SNR [signal/noise ratio] and similar parameters to do analog IC design and meet channel requirements. In basestations, for example, dynamic range is the key factor."
In the shorter term, Counts believes "we'll have a profusion of inexpensive electro-optic links in the next five years, using an optical link between clamshell halves [of a cell phone], and with chip-to-chip optical communications coming" as well.
Looking further ahead, "the convergence of biotech and electronics will prove to be very fruitful, with interdisciplinary efforts. Engineers are going to have to get closer to the fundamentals of physics and biology."