Feb 2, 2014

Guru #3, George Erdi

George Erdi is one of the five “gurus” in the early Linear Technology advertisement. His colleagues poked fun at him for being called an “elegant” precision amplifier designer. But he certainly was the pioneer in precision (“Mr. Precision” according to the Wall Street Journal). He was able to make a precision amplifier on a process that wasn’t built for it at Fairchild. Then he built an entire company around the concept, Precision Monolithics Inc. And then he joined Linear Technology as a founder and got his face on this advertisement.




George Erdi was born in Budapest, Hungary in 1939. In 1956, in a brief period before the Soviets crushed the revolution in Hungary, he and his mother escaped to Canada (his father had died in WWII). In much the same way as his childhood friend, Andy Grove, they walked across the border at night in the snow. In those few weeks when the revolution was succeeding, the border control was loose and 200,000 Hungarians fled the country. His aunt lived in Toronto, so that was the destination.

George had completed high school in Hungary and did not need further secondary school in Canada. In his own words, he was not a “born engineer.” Coming from Soviet-controlled Hungary, he did not speak English. “After a couple of years, I enrolled in the Engineering Department of McGill University, assuming that command of the language was not a necessity in a scientific field. In my freshmen year, however, the first course I had to take was English Literature. Fortunately, the first topic we covered was Geoffrey Chaucer’s Canterbury Tales, written in middle English. None of my classmates understood a word of it either. Thus, I became an electronics engineer almost by default.” McGill University, in Montreal, is one of only three English-speaking universities in Quebec. (Not to be a name-dropper, but Les Vadasz was also Hungarian and also attended McGill – although George didn’t know him. Vadasz worked at Transitron, then the digital side of Fairchild before joining Intel.)

He left Montreal and went on to pursue a Master’s Degree at the University of California in Berkeley in 1965. He had been accepted at both Stanford and Berkeley, but had a scholarship to Berkeley. In those early years of the semiconductor industry, Berkeley had a small fab. He took some IC classes, but wrote his thesis in statistical communications because the IC route would have taken much longer. That is essentially where he started in analog. Andy Grove was the key process engineer at Fairchild R&D and made the recommendation for Fairchild to hire George. He says when he joined Fairchild he was basically a novice, “To quote an old joke, I did not know an integrated circuit from a segregated one. But neither did anybody else, the situation being strikingly similar to my first English class at McGill.”

Dave Fullagar joined Fairchild a week after Bob Widlar and Dave Talbert left; George joined a month later. Fullagar was George’s office mate and was intended to work on a precision amplifier, the µA725, but it didn’t work quite right. Then Widlar’s LM101 came out from National. The two office mates asked themselves why the compensation capacitor wasn’t integrated. Perhaps National’s process couldn’t accommodate a cap, but theirs could. Fullagar then did the µA741 (integrating the comp cap) and George picked up the µA725. He essentially started over and created the first precision monolithic op amp. He wrote a paper on the part, but left Fairchild before it was published – so Fairchild removed his name from it. Paul Gray, a professor at Berkeley, did give George credit, however (you probably studied from his Gray and Meyer textbook). George is also credited in various sources with the first monolithic D/A converter. The µA722 was started by Tom O’Day, who left in the middle of the project to join H-P. George made some significant changes and accepts credit as the co-designer. Not bad for a kid not even 30 years old on his first two circuits. By the way, Mike Markkula was responsible for the marketing – legendary investor and second CEO of Apple Computer.
Fairchild µA722
10-bit current source
for D to A converters
(Credit: Fairchild
Camera & Instrument
Corporation)

“In those early years of ICs in the mid-1960s, experienced design engineers had great difficulty abandoning the well-established design concepts of discrete or even vacuum tube circuits. The idea that a transistor was cheaper than a resistor was revolutionary. Everything depended on matching between resistors and transistors, not on absolute tolerances. Inductors and capacitors were unavailable.”
“I, as a beginner, did not have any preconceived notions. IC design was just as easy, or as difficult, as discrete design. My timing was right,” recalled George. 
Bob Pease was one of those designers who were raised in vacuum tube and discrete amplifier design. He devoted one of his columns to his reaction to Erdi’s first op amp.


“When monolithic op amps came along, there were some influences to ‘keep it simple, stupid’. I designed a T52AH – also labelled as Amelco's 809BE – with just 10 transistors, which worked pretty well. But op amps with 20 or 30 transistors soon had just as good a yield. And they offered more features. So, we kept learning how to add more transistors for better performance.”
“But when the Fairchild μA725 came along, we designers were really puzzled. Why would anybody use FOUR input transistors? What the heck was George Erdi smoking? If you set up a diff amp with two transistors in parallel at the plus input and two transistors paralleled on the minus input, why would that give an advantage? But the specs showed real superiority – low offset voltages, good bias currents, and low offset current. Hey, this was about 1971. Not many engineers were climbing inside their suppliers' ICs and studying the layouts. If you didn't, though, you could be stuck with a lousy layout. I know.”
“The basic feature of the μA725 was the common-centroid layout of the input transistor "pair."
“If you took those four input transistors and laid them out in an X pattern, it would be denoted by: 
AB
BA
Connecting them properly in parallel, you can get the linear gradients of Vos, to cancel. And the gradients in beta to cancel. Gradients caused by heating from the output stage – and even from other asymmetrical sources of heat -- tend to cancel. Any linear gradients caused by imperfect die attach tend to cancel. (Nonlinear gradients do NOT get cancelled, of course, but these are usually small.) And these cancellations all happen thanks to a common-centroid layout, which is just another way to say that the ‘Center of Gravity’ (CG) of one input "transistor" is at the same place as the CG of the other transistor.”

Dave Fullagar responded to the Pease column:


“…I had designed the prototype µA725 in 1966 and presented a paper on it at the International Telemetry Conference in Washington in 1967. This early version was quite crude and did not contain a cross-coupled input pair."
"When I proposed, and was given the go-ahead to work on, the µA741, the µA725 project was assigned to my officemate, George Erdi. I remember several brainstorming sessions in which we discussed ways to improve on its performance, including the use of a cross-coupled input stage —we didn't have a fancy name like "common-centroid" for it in those days! The sketch we ended up with on the blackboard was very similar to Fig. 2 of your article. George vastly improved on my earlier design, and the product was announced by Fairchild sometime around 1971 …”
George said it seemed like every Friday was a going-away party at Fairchild. In 1969, Marv Rudin (manager of linear circuit R&D in Palo Alto) and Garth Wilson (circuit design manager under him) left Fairchild to found Precision Monolithics. PMI was founded with financing from Bourns, Inc. Immediately after financing and incorporation, they hired George Erdi who had worked for Wilson at Fairchild, and then Dan Dooley from TRW Microelectronics; both were offered founder stock. They hired a brilliant process engineer, Jerry Bresee, from Techtronix, who developed a semiconductor process far superior to Fairchild. Semiconductor and materials engineer Wadie Khadder was hired with founder stock from Fairchild to support Bresee in both semiconductor process engineering and also the critical precision thin film technology.

Dooley left after about 4-5 years and joined National. With Bresee's and Khadder’s superior process, Erdi was able to design and achieve breakthrough advances in precision and low power for amplifiers, voltage references and data conversion. In 1975, Erdi reported on an offset trim technique that used 300mA over-current pulses, to progressively short zener diodes in a string. This so called “zener zapping” could be used to trim the offset of an op amp on the wafer. The first op amp to utilize this new trim technique was the OP07. (I was always a big fan of their part number prefixes: OP, REF, DAC, etc.)

The OP07 and DAC100 were huge successes. Even though Bourns had repurchased all the shares to make PMI a wholly-owned subsidiary, George felt well compensated. But eventually, George looked around and thought that if 60% of the sales were coming from his designs, then they weren’t going to grow very fast. He needed more contributors. George and Wadie Khadder left PMI in 1981 and eventually, PMI was finally bought by Analog Devices in August 1990.

George knew Bob Dobkin from conferences and called him when he read about him leaving National to start a new company. Linear Technology had not yet been formed, so George joined as a founder. He was employee number five. The other four founders had all quit National at the same time. It was hoped that hiring George from PMI might reduce the legal action from National. But no. George said the legal action with National went on until the same day as the big earthquake in 1989. The settlement came in at 3pm and the earthquake hit at 5pm. As with Fairchild and PMI, he was immediately successful at Linear Technology developing many families of op amps. He retired in 1993 at age 54 but continued to come in once a week or so to support his products.
2012 Analog Aficionados Party
Photo credit: Fran Hoffart

I asked George about the Wall Street Journal article citing Linear’s first new product as the LT1013. He said it wasn’t the first product. Earlier parts were improved versions of existing parts from competitors, like his LT1001 as an improved version of his own OP07. The LT1013 was the first proprietary new design released. I also asked about Bob Widlar. As mentioned, George joined Fairchild a month after he left. George’s first task was to look over Widlar’s projects and see there was anything to salvage. But he said there was nothing left behind. George actually met Widlar a few times at the ISSCC when it was held in Philadelphia – always with a drink in his hand. They worked together at Linear Technology, of course, while Widlar was living in Mexico. Widlar would come in to town for a few weeks and Dobkin would drive him since he didn’t have a license. He said Widlar would never admit that there were things he didn’t know about. Yet he always managed to find his way into George’s office to ask question.

For a man who wasn’t a “born engineer”, having designed some legendary op amps, having Bob Dobkin hire you as his company’s first engineer and also having Bob Widlar ask you about op amp design – that’s pretty impressive.

NOTES

“In a Tech Backwater, A Profit Fortress Rises; Maker of Arcane Chips Earns Better Margins Than Google, Microsoft”, Wall Street Journal, George Anders, Updated July 10, 2007

Computer History Museum - The Silicon Engine | People www.computerhistory.org, 30 June 2011

“Analog Circuit Design – Art, Science, and Personalities”, edited by Jim Williams, 1991 (Chapter 18. Starting to Like Electronics in Your Twenties)

http://www.computerhistory.org/semiconductor/timeline/1968-Data.html
Bob Pease: “What’s all this common-centroid stuff, anyhow?”, Electronic Design, Sep. 30, 1996

Oral History – Dave Fullagar, Jack Gifford, Garth Wilson, Computer History Museum, http://archive.computerhistory.org/resources/text/Oral_History/Fairchild_at_50/102658281.05.01.acc.pdf

IC Op Amps Through the Ages, Thomas H. Lee, rev. October 31, 2003

Precision Monolithics, Inc., Wikipedia

Walt Jung’s Op Amp History, http://www.analog.com/library/analogDialogue/archives/39-05/Web_ChH_final.pdf

No comments:

Post a Comment