Voices: Dave Fullagar, analog-IC designer and entrepreneur
Dave Fullagar made his mark at Fairchild Semiconductor in the 1960s. He designed the ubiquitous μA741 op amp, perhaps the most successful op amp ever. Fairchild, Texas Instruments, and National Semiconductor still sell versions of the popular device (References 1-4). After leaving Fairchild, he joined Intersil as the company’s first analog-IC designer. He later co-founded and became a vital technical contributor to Maxim Integrated Products, from which he retired in 1999. EDN recently had the chance to interview Fullagar.
Were you an electronics genius as a boy? Were you building crystal radios and winding your own slot-car motors?
I’ve met very few geniuses in my life, and I’m certainly not one of them! But yes, I was building crystal sets at about age 9. Until I was 12 years old, we lived on the moors in the north of England in a house with no electricity, so a crystal set was my only option. In 1954, I read an article entitled “How to build a radio in a flashlight.” It used something called a transistor—a Mullard OC71. I went down to the local radio store to buy one. “Never ’eard of a transistor, boy. Don’t know naught about that,” said the proprietor in a broad Yorkshire accent. I finally did acquire some OC71s, and still have them. They are glass-encapsulated junction-alloy devices. To make a phototransistor, you just scrape the black paint off the glass.
While at the Fairchild 50th anniversary celebration, I was struck by how few electrical engineers there were in the semiconductor industry in the 1960s. Many were materials people, physicists, and mathematicians.
I think that’s an accurate observation. Understanding how bandgap references work or the limitations of lateral PNPs requires knowledge of the underlying device physics. In my case, the electrical-engineering program at Cambridge University had a prerequisite of two years of either physics and math or mechanical engineering. I had no interest in concrete bridges and sewer systems, so the choice was simple.
Did you consider any other fields of study? What brought you to choose your specialty?
I did consider a career in physics, but I like a work product that you can touch and feel. As soon as the physics curriculum moved into quantum-mechanics and Shrödinger’s equation, I started to lose interest. Physics up to that point had been so precise, and now they’re telling me they can’t even figure out where the damned electron is?
What was your first job out of school? How did you land it?
My first job was with Ferranti [a leading UK defense contractor] in Edinburgh, Scotland. I worked on a terrain-following radar for a bomber that was supposed to fly to the Soviet Union at treetop height. They used to fly the engineers over the highlands of Scotland—at 300 feet—to validate the radar’s terrain-following ability. It was a good incentive to get the design right. Ferranti’s offer was the best paying—at about $2200 a year; no, I didn’t omit a zero—of the four or five jobs offered.
Was working in the United States what you expected? What was your path to Fairchild? Why did you take the job there?
In the 1960s, there were two principal routes to the United States for Brits. One was via Canada; the other, through Transitron in Massachusetts. Tom Longo at Transitron hired me. My colleague and fellow Cambridge graduate Wadie Khadder had joined Transitron a year earlier, so the transition was an easy one. It soon became apparent that Transitron was in decline. (It had been the No. 2 semiconductor company in the United States in the 1950s.) So both Wadie and I moved to Fairchild in early 1966.
I moved to the United States because I was young, single, and looking for adventure. If someone had offered me a job in Hong Kong, I would have taken it.
My assigned task when I joined Fairchild R&D in 1966 was to design the successor to the μA709. The target specification I was given by marketing was of the “let’s-improve-all-the-key-specs-by-50%” variety. However the biggest problems with the 709 were its idiosyncrasies, not its specifications: It was tricky to stabilize, there was no short-circuit protection, and it would latch-up and self-destruct in nanoseconds. National’s LM101, which [the late Robert] Widlar designed, addressed many of the user-friendliness issues but still required external compensation and had a kludgy front-end bias scheme. Widlar must have come to the same conclusion, because he later redesigned it as the LM101A with a much-improved front end.
I proposed the internally compensated μA741 in mid 1967 to Garth Wilson and Marv Rudin, who ran the Linear R&D Group. Next thing I knew, I was sitting in Gordon Moore’s office. He asked me if I’d mind moving to Mountain View because that would expedite the introduction of the part, which occurred in May 1968.
IC designer George Erdi was at Fairchild at this time. Did he help you, or was he more of an insular, loner type?
I wouldn’t describe George as a loner type. He was my office mate and an excellent engineer at Fairchild R&D. In those days, each project was a one-man affair, so, although we might discuss aspects of design or layout over coffee, each person was solely responsible for his own design.
How intertwined was the design of the 741 with the process?
One of the reasons I moved from R&D in Palo Alto, CA, to the Mountain View facility to complete the 741 was that the linear-circuit fab was in Mountain View. The process had been running for a couple of years there and was quite stable. The only process issue for the 741 was the need for a pinhole-free capacitor oxide, but this [requirement] turned out not to be a problem. The design was very tolerant of process variations. In particular, lateral PNP’s betas were all over the map in those days because the lateral PNP is a surface device susceptible to contamination, but the design could tolerate beta variations from about three to infinity.
Bo Lojek’s History of Semiconductor Engineering states that you invented the three-terminal regulator while at Fairchild but that marketing never developed it. Many IC designers seem to exhibit hostility toward marketing types. Do you think analog-IC marketing is all it can be?
It’s true that I proposed a three-terminal regulator well before any existed in the marketplace. Marketing told me there wasn’t a need for such a device because power routed to individual PCBs [printed-circuit boards] was decoupled with just a low-value resistor and a capacitor, and this solution was cheaper than my proposed three-terminal regulator.
To say design engineers have “hostility toward marketing types” is a little strong, but healthy skepticism is in order. Marketing typically reflects what they think customers are telling them. The customer has no idea what can be done at the silicon level, so he says: “I like your existing widget, but what I really need is one with these four key parameters improved by 50%.” So, that becomes the new marketing plan. This approach never spawns novel products.
The really novel products come about in one of two ways. One is when an engineer figures out a totally radical approach to meet a functional need. For example, Widlar’s bandgap reference was sheer genius. We all knew that VBE [base-emitter voltage] had a negative temperature coefficient and delta VBE had a positive temperature coefficient, but Widlar figured out how to put the two together to produce a temperature-stable reference.
The other way is when a good applications engineer works hand in hand with a talented designer. Maxim’s RS-232 products came about in this manner. Charlie Allen, the best applications engineer I have ever encountered, recognized that lots of digital equipment had ±12V supplies whose sole purpose was to power RS-232 drivers. RS-232 drivers didn’t fit into any of Maxim’s analog-product lines, but he persisted. He corralled Dave Bingham, one of Maxim’s most creative design engineers, into taking a look at it. Bingham felt sure he could put charge pumps on the same IC as the RS-232 drivers—no mean feat—thereby eliminating the need for the ±12V supply. Thus was born one of the company’s most successful product lines.
OK, so you design one of the world’s best known op amps at Fairchild and then go to Intersil. Did Fairchild drive you out or did Intersil lure you in?
I think Intersil lured me, but it wasn’t financial. I didn’t really know what a stock option was and couldn’t have distinguished between an IPO [initial public offering] and a UFO [unidentified flying object]. Intersil offered a chance to create my own analog-design group with a clean slate and to work with Jean Hoerni, PhD, one of the true giants of the semiconductor industry.
What parts did you work on at Intersil? Were there any accomplishments you are particularly proud of?
It’s interesting that, as a circuit designer, you get tagged by your best-known design. Hans Carmenzind is known as the 555-timer guy, I’m known by the 741 op amp. But Carmenzind created many great designs in addition to the 555. For me, the Intersil years were the most creative. I developed the first IC logarithmic and antilogarithmic amps and the first monolithic FET-input op amp, the ICL 8007, which dominated the market until the bipolar FETs came along.
I also got to spend a month in Japan designing the first electronic-shutter IC for a single-lens-reflex camera for Canon. This circuit took the logarithm of three inputs—film speed, aperture, and light intensity—summed the result, stored it while the mirror went up, and then took the antilogarithm to generate the shutter speed. It used about 20 transistors in total and was probably my most elegant design. Nowadays, this is done with about half a million transistors in a microcontroller—how prosaic.
From Intersil, you went on to co-found Maxim in 1983. Did you catch this entrepreneur’s spirit while you were at Fairchild, or did you have it from the beginning? Did Jack Gifford seem like a good partner?
I suppose anyone who leaves his native country to start a new life in a foreign land has, by definition, an entrepreneurial spirit. Certainly, Silicon Valley has a huge number of successful contributors who started their lives in places as diverse as the United Kingdom, Bangladesh, and China. I first teamed up with Jack Gifford at Fairchild, where he had marketing responsibility for linear products. When I went to Intersil, it was frustrating to be designing products that I felt weren’t selling because of inept marketing. I encouraged Jack to join me there, so we were both business partners and personal friends for many years before Maxim.
I heard 15 years ago that Maxim’s strategy was to hire inexperienced people fresh from school and put them under the wing of an experienced IC designer. Is that how things worked?
Yes, and it was a very successful program. Most of the people I hired out of college are still with the company—some in vice-president slots.
Both you and Linear Technology’s Bob Dobkin showed me logarithmic-amplifier circuits during interviews. What is it about that circuit that you like?
Interesting—I didn’t know that Dobkin also used that circuit. I’ve been using it since I designed a logarithmic amp for Intersil in 1970. I like it as an interview topic because it isn’t something you find in a textbook. The first—relatively simple—challenge is doing the dc analysis. The more interesting part is figuring out the impact on ac stability of having an active element in the feedback and what it takes to achieve unconditional stability.
Maxim copied a lot of Intersil parts as a second source at the beginning. The famous ICL7660 charge pump, the 7106 and 7107 DVM (digital-voltmeter) chips were a few of the parts. Did they keep you going while you designed proprietary Maxim parts?
Absolutely. The original business plan called for the speedy introduction of 14 second-source parts to generate quick cash flow, followed by proprietary parts. Generating positive cash flow in a start-up is key: If you have to go back to the venture capitalists for an unscheduled round of financing, you get taken to the cleaners.
Considering your accomplishments at Maxim, was it a part, a process, or a way of doing business that you feel proud of?
I feel proudest of the design team I recruited. They were—and still are—some of the finest people in the industry, having designed an incredible number of innovative products, as well as serving in senior-management positions, including current President and Chief Executive Officer Tunc Doluca and several of the vice presidents.
Maxim was one of the first analog-semiconductor companies to organize itself by market as opposed to process, design difficulty, or part type. What gave Maxim this insight more than a decade ago?
There comes a point in any company’s growth when having one person run all of design, product planning, or sales stretches that person too thin. At that point, it makes sense to organize by market. Apart from being more efficient, this structure creates a number of individuals with profit-and-loss responsibility, who acquire the skills necessary to run all aspects of the business: chief executive officers in training. This approach facilitates a smooth succession as the original management team retires.
When you think about all the great things you accomplished at all the glamorous companies you worked at, which was the most fun?
Tough question because it was all fun. I thrive in a small-company atmosphere, so I would have to say the early days at both Intersil and Maxim. The lab work was always especially rewarding. Checking out a new breadboard with a cup of coffee warming on top of a big Tektronix 545 oscilloscope—that was real engineering! Sitting in front of a screen running simulations just isn’t the same.
Any chance that you’ll get bored in retirement and design a few parts for some lucky company?
That’s very flattering, but, after being retired for almost nine years and in management for 20 years before that, no one would want my designs today! But if anyone would like me to put together an electronics package for their sailboat, with a voyage to Tahiti included, well that might be different.
- Fairchild Semiconductor LM741
- Fairchild KA741
- Texas Instruments UA741
- National Semiconductor LM741
- The ua741 operational amplifier, University of Wisconsin-Madison College of Engineering.
- Thomas H. Lee, Tales of the Continuum: A Subsampled History of Analog Circuits, IEEE/SSC, Oct 2007.