Mar 27, 2014

Guru #6, Nello Sevastopoulos

The original “guru” advertisement from Linear Technology only promoted the designers since they had not released any products by that time. The ad ended with “…watch this space in the near future for the first in a series of new product announcements.” Several months later, the second “gurus” ad appeared. And this time it introduced a dozen new ICs. And it also added a new “guru”, Nello Sevastopoulos.

Sevastopoulos joined Linear Technology in August of 1983 to bolster the CMOS design effort with Tom Redfern. He was friends with Redfern and Jim Williams back at National Semiconductor and they were the ones who convinced him to join LTC. Employee number 138 overall, he was the fifth IC designer. But how did he become a linear IC design “guru”? His footsteps didn't go through Stanford or MIT or Fairchild and he didn't take apart tube radios as a child prodigy. Surprisingly, it almost didn't happen at all; at several points in his life fate conscripted to point him in that direction.

Born in Athens, Greece, in 1945, the Sevastopoulos family was relatively well off in the context of post-war Europe. Other countries in Europe were worse off, of course. But he recalls the Sixth Fleet of the U.S. Navy being anchored off-shore and he looked at them like gods. They were wearing blue jeans, smoking Pall Malls and had Zippo lighters – luxuries for his family. In his youth, Nello was good at sports. Unlike the United States where most organized sports fell under the control of the schools, in Europe sports were organized by neighborhood clubs. There were stadiums within walking distance for soccer and gymnasiums for basketball. The coaches were community volunteers and there were club leagues. Sevastopoulos excelled, particularly in track and field. By the age of 17 he was relatively famous, traveling throughout Europe competing in the high jump at world-class tournaments and being interviewed for newspapers, TV and radio. Nello was part of the Greek National team in 1964 – the year of the Tokyo Olympics. To this day he’s quite athletic and in exceptional shape.

Times were different back then and there were no prospects for a professional career in track and field. His father was a textile engineer, but Nello had no interest in engineering. He was good with his hands and considered being an architect – but with Europe in turmoil after World War II, those didn't seem like good career prospects. As fate would have it, the King of Greece had also been a high jumper and knew of Nello Sevastopoulos. The King sent telegrams and offered him a scholarship to college. Sevastopoulos chose to attend a technical school associated with the University of Brussels in French-speaking Belgium. He was good at math and physics, but struggled with electrical engineering. Although he says he took a plurality of electrical courses, he majored in nuclear engineering. The electrical courses were based on vacuum tube circuits in those days.

Sevastopoulos did some post-graduate work in audio, working on speaker systems with electrical feedback for about six months. In 1969, our future linear circuit design guru, Olympian, who spoke Greek and French, returned to Greece for his military service. He finished his service as a 2nd Lieutenant in 1971.

Even in those days, a bachelor’s degree was not always sufficient and Sevastopoulos felt he needed to go to graduate school. Since he spoke French, he wanted to attend McGill University in Montreal, Canada (McGill in those days had a French-speaking section). But without much money it was going to be difficult, so he decided to go to California where is brother worked at H-P, in Palo Alto. Prior to leaving Greece, Sevastopoulos took the GRE and an English proficiency test – which he failed due to his poor knowledge of English. So he came to the United States on a tourist visa and had two months to improve his English, retake the tests and get accepted to a University. The day he arrived was September 4, 1971. He spent 15 hours a day studying English and watching “Dialing for Dollars” on television with a dictionary. This time, he easily passed the tests.

So now you are probably expecting our nuclear engineer and future linear IC design guru to go to Stanford and study under Fred Terman. No, be patient my friend.

Sevastopoulos liked math and had a little bit of feedback experience and wanted to study control systems. He had applied and been accepted to several big schools in the area but he went to Santa Clara University.

Santa Clara had been a key research institution for NASA during the Apollo moon missions, supporting the control systems. Professor Dragoslav Šiljak’s methods were used at NASA Marshall Space Center in control design of Saturn V rocket which powered Apollo astronauts to the Moon. Coincidentally, Šiljak made the 1952 Yugoslavian national team for water polo and traveled to the Olympics in Helsinki. The final match was against Hungary for the gold. The match ended in a draw; with the winner determined by cumulative goal ratio, the Yugoslav team was awarded the silver. The following year the team won the World Cup in the Netherlands, beating Hungary in the finals, avenging the loss of the gold medal. He missed the 1956 Olympics due to injury, but he was back for the 1960 Olympics in Rome – but finished fourth; no medal.

At the University of Belgrade, Šiljak sought out books by Russian mathematicians like Lyapunov, Pontryagin, and Krasovskii. And, in a country where the supply of basic goods and services sometimes made it difficult just to make multiple copies of an article, Šiljak managed to get papers published in the top U.S. journals in control engineering. He published a paper in 1962 with co-author R. Petrović, which is credited for the beginning of stochastic computing. He joined the Department of Electrical Engineering at Santa Clara in 1964, where he taught courses in control theory and applications. After hearing Šiljak lecture on new control methods for larger booster rockets, an attending NASA scientist invited him to the Marshall Space Center in Alabama, where he soon began work on control design for the Saturn rockets. In the early 1970’s, Dr. Šiljak initiated research in mathematical theory of large complex dynamic systems and applied his analysis to a wide variety of models in areas as diverse as control of electric power systems, population biology, large space structures, competitive equilibrium in mathematical economics, and the arms race. Although he had some electronics courses, Sevastopoulos wrote his thesis on the stability of larger power machines.

This time around, Sevastopoulos excelled academically – graduating in 1973 with about a 3.9 grade point average. But some of his friends did even better. He recalls Slobodan Ćuk, another Serbian from the University of Belgrade. “He was an animal in how hard he worked and studied,” said Sevastopoulos. He finished with perfect grades, and went on for his PhD at Santa Clara University. If you recall the infamous Apollo training accident where a fire occurred during a test and the door would not open – we believe Ćuk predicted the failure from his own analysis. Ćuk then went on to teach at Caltech. Perhaps you are more familiar with the Ćuk Converter – the two-inductor, switched-mode power supply topology where the output can be lower than or higher than the input. Yes, that’s him.

Other classmates were Volkmar Shaldach who eventually ran Analog Devices in Europe, and Evan Moustakas who went on to get his doctorate at Santa Clara University and he became the EE department chairman of San Jose State University. At this time Marvin Vanderkooi was the Linear Applications Manager at National Semiconductor. Moustakas convinced Vanderkooi to get Sevastopoulos an interview at National – even though he had only taken a few electronics courses and had no other relevant experience. Sevastopoulos’ roommate happened to work at Ampex and knew a lot about semiconductor ICs. He pulled Sevastopoulos aside for a couple hours before the interview and said, “This is what you need to know.” He taught him about op amps, linear voltage regulators, current sources, comparators and voltage references. And that was exactly what the interviewers at National asked him – and amazingly, he got the job. Sevastopoulos was assigned to the applications department in the standard linear IC (SLIC) group. At first, this was a huge struggle for Sevastopoulos since, in his own words, he didn't know much.

At the time, there were three linear design groups within National: standard linear (SLIC, run by Jim Solomon – founder of what became Cadence), advanced linear (ALIC, run by Bob Dobkin) and consumer linear (CLIC, run by Tom Isbell). Sevastopoulos worked in standard linear applications and occasionally needed to ask questions of Dobkin and Carl Nelson. The ALIC guys certainly looked down on the SLIC guys and, even more so, on a new applications engineer who had to learn everything on the fly. He also saw Bob Widlar a couple of times (since Widlar had retired to Mexico by then). Widlar left an impression because of how well he promoted his products: great application notes, presenting at seminars around the world, very thorough.

Sevastopoulos survived the layoffs and rose to be the applications manager for SLIC. They had just invented the BiFET process and brought out the LF156 and, later on, the quad “741”: LM148/149. Quad op amps lent themselves to sophisticated active RC filter topologies, then in fashion. Sevastopoulos got interested in filter topologies. He still didn't know how to design an IC or know much about semiconductor processes; after all he was an applications engineer. He, however, traveled the world presenting on how to use these new amplifiers. The uncompensated variant, LM149, was particularly well suited to making higher frequency filters. And for a control systems guy with a penchant for transforms, this was fun. The timing was perfect for the rise of PCM in telephony applications.

About the same time, needing extra cash, Sevastopoulos decided to design a graduate course in op amps and teach at Santa Clara University. This was an applications class – how to use op amps – not a design class. That went well and he did that for a few years. Tom Odell, then a product engineer and later president of National Semiconductor (standard products group, including linear), was one of his students. The course was successful and continued to be taught by Professor Sergio Franco of San Francisco State who also wrote a textbook.

When National decided to build CODECs for the PCM telephony market they hired two consultants, David Allstot and Bill Black who graduated from UC Berkeley. The consultants needed help with noise analysis of CMOS transistors and Sevastopoulos volunteered for the task. The monolithic CODECs would use the new switched-capacitor topology for their filters, and this excited him. To his knowledge, it was the first of its kind and started studying all the newly written papers mainly published by UC Berkeley.

At the same time, he got a call from San Jose State University, asking him to teach a graduate IC design course in place of a professor who was leaving for his sabbatical. They knew of his previous teaching work at Santa Clara University so they chose him for the task. Of course, he was a bit uncertain about it since he had never actually designed an IC before. His friend and mentor, Tom Frederiksen (author of "Intuitive IC Op Amps") encouraged him to do it. So he agreed. This forced him to learn how it was done. It was a lot of work, studying evenings and weekends to prepare the lectures, but again the teaching experience was a good one and he was even voted best instructor. Circuit design then seemed easy enough while he was teaching, that he decided to do it for real. He managed to convince his boss to let him quit his Linear Applications Manager job and move over to design engineering. He also hired a young graduate from Berkeley, Sammy Lum, presently at Linear, to form a small design group.

This is not your typical “linear IC design guru” path to IC design. For what he was interested in, the key was the topology more than the analog aspects of the amplifiers. So he didn't need to push the envelope on op amp design. To make a successful product, it needed to be a building block so that one part could be sold to many customers. What he came up with for his first design was the MF10, “MF” standing for Monolithic Filter and the number 10 from the film “10”. This type of proprietary dual 2nd order switched-capacitor filter device had never been done before, so Sevastopoulos had to convince management to let him do it. That meant proposing it in front of Charlie Sporck and all the senior managers and IC designers in their big National Semiconductor amphitheater. First silicon worked, Sevastopoulos hit the road to start doing seminars and promote the part and it turned into a big success. His training represented his view of the part – here’s how it works, here’s why and here are all the types of filter responses you can create with it. But he clearly recalls one seminar in England, where an engineer said “Great, now tell us which mode we should use for our product.” Sevastopoulos thought to himself, “Don’t you want to figure that out?” That’s when he started to realize that many engineers want less theoretical information and more cookbook style guidance.

The MF10 and all of the subsequent promotion made him a famous name in the analog arena. There were other chips, but the MF10 vaulted Sevastopoulos to “guru” status. The product was used widely and eventually second sourced by many major semiconductor companies.

And now we can rejoin our story in August of 1983 when Redfern and Williams asked him to join Linear Technology. As a footnote, he was also lured by Maxim which was founded two years after Linear Technology. He could have been employee #18 at Maxim instead of #138 at LTC. Since Maxim was just starting up, they placed a priority on doing quick second-source products to generate revenue. “We want you to do this part and then this part – copy these industry standards” they said. Linear was looking more long-term and that’s where his friends were.

At LTC, he and Redfern were working in CMOS while most others were working in bipolar. The bipolar designers bread-boarded everything with kit parts – transistors wired up in three-dimensional fashion to prototype each new chip. That wasn't really practical for CMOS since there were so many variations possible for transistor length and width. In those days, simulation was still rather basic – level 1, maybe level 2. There was one computer terminal (no actual computer, but a service that was contracted for computing). They didn't get accurate models until the late 1980s. They really busted their butts in those first few years to make the company grow. It was hard. There was a lot of pressure. They did their own layout if the scarce resources were being used elsewhere. The first chip he did was the switched-capacitor building block, LTC1043, which Williams used extensively and creatively in many application notes. To avoid getting sued by National Semiconductor, Redfern didn't make data converters and Sevastopoulos didn't make filters until later.
From the Linear Technology Annual Report in 1986; L-R, Nelson, Dobkin, Erdi, Redfern and Sevastopoulos

The MF10 and the following family of monolithic switched capacitor filters qualified Sevastopoulos as a guru, and it served its marketing purpose by helping the name-recognition of the company in the early years. He said, “We were it, man, we were the state of the art!” Sevastopoulos said it actually surprised all of them that customers didn't immediately buy their parts. Even with all these big names, customers would ask about a second source. He credits George Erdi because his parts immediately generated revenue. In most cases, they were standard pinout op amps with greatly improved specifications. His parts were beautifully crafted. By “crafted” Sevastopoulos meant that Erdi would thoroughly measure everything and study the parameters’ distributions and make fine tweaks, over and over until the part was as good as it could be. “No one else had the patience or the passion.”

The MF10 wasn't his favorite chip, nor the LTC1060 which improved upon it. “The problem with the switched-cap topology was that it wasn't DC-accurate.” That was okay for “AC” telephony, but you couldn't really use it in data acquisition systems. With the LTC1062, he created a DC-accurate low-pass filter (LPF) – the first of its kind. He said this was an expansion of an idea that Jim Williams showed him from an H-P DVM. His favorite chip was the LTC1063, the next DC-accurate LPF that was very easy to use. It happened to be used by a company called PKI (German division of Philips) which he believed made an early portable phone. But it was the LTC1064 that made a lot of money for the company – hitting the market at the right time for rapid growth in base stations and in data acquisition systems. The LTC1064, quad filter building block, incorporated a very space efficient proprietary thin filter resistor network. The product with a simple metal mask change could be configured as a complete monolithic 8th order filter or dual 4th order filter, giving it tremendous flexibility and speed to market. Give him time and he’ll talk about all his chips. He’s clearly passionate about them.

Sevastopoulos was never really in love with transistor-level circuit design. Instead, he used circuit design to create complex products. An Olympic-caliber athlete, a nuclear engineer, an appreciator of an elegant filter topology, he created some masterful circuits nonetheless.

The world “guru” in Sanskrit means “teacher”. A real guru not only has students but he divulges his knowledge and wisdom. Sevastopoulos’ real legacy, beside all the products he designed, is the creative IC design group he formed, led and taught. Together, they expanded the product line beyond filters. They designed monolithic chopper stabilized op amps, monolithic chopper stabilized instrumentation amplifiers, programmable gain amplifiers, continuous time (not switched capacitor) high frequency filters, high precision sense amplifiers, precision CMOS amplifiers (not choppers), differential amplifiers, a new RC oscillator family, even a couple of “system on a chip” products commissioned for a prestigious instrument company. Many IC analog designers stuck to their “knitting”, albeit crafting beautiful designs. Instead, Sevastopoulos loved creating new functions, and he loved adding, often to the dismay of management, new process steps to benefit product topologies. He left behind talented and creative engineers, the future gurus.

1 comment:

  1. This comment has been removed by a blog administrator.

    ReplyDelete