Interview with Paul Gray
August 21, 1998
Orinda, California


RW: Paul is currently Dean of Engineering at the University of California, Berkeley. He and his researchers did the seminal work in MOS analog circuits, such as filters, analog to digital converters, and operational amplifiers. This work effectively obsoleted bipolar technology for mixed digital analog integrated circuits. Tell us about your early background.

PG: Well I was born in Arkansas in Jonesboro, a town in north east Arkansas, came from a family of, who had lived in the area for many years, merchants mainly. My mother came from that family; my father was a West Point graduate, Army officer, also from Arkansas. I grew up there then mother and father were divorced, and moved out to Tucson, Arizona in 1957. She supported herself as a schoolteacher, and raised my younger brother and I first in Arkansas and then in Tucson. Went to high school in Tucson, went to college at the University of Arizona starting in 1960. Stayed there, went to undergraduate, finished in '63; graduate school, finished a masters in '65; and then stayed and got a PhD degree in 1969.

RW: Were these in electrical engineering?

PG: All in electrical engineering, yeah. Right, and that was really kind of how I got into the integrated circuit world. At that time, in the late '60's there were three laboratories in the United States in university settings that had the ability to fabricate an integrated circuit. One was Berkeley, one was Stanford, and the next one was at the University of Arizona largely because my thesis advisor, who was a professor named Doug Hamilton, had visited Berkeley and went back to Arizona and started that laboratory. So as a graduate student in the late '60's, I had a chance to actually push wafers in and out of a furnace tube and learn about integrated circuit technology kind of from the basic technology up, and that was something that really, really kind of shaped my future career, because it really gave me a feeling of what the technology was like.

RW: So did you teach?

PG: I did some teaching there, I was a lab TA and got introduced to teaching that way, but the course I took was really more of, in terms of the research, was more focused on circuit design, IC design, mixed signal and we, Doug's research area program focused on a number of aspects. My particular one was electro-thermal interactions, that is how in something like a power integrated circuit, like a voltage regulator or an audio amplifier that might drive a stereo speaker, how the heat generated by transistors on that integrated circuit chip would interact with the other transistors on the same chip. And in addition to the electrical interactions that you'd normally expect, you'd have the potential of getting things behaving in strange ways because the heat transferred rapidly from the transistor that's dissipating power into the other transistors and circuits. So, that turns out to be, it was sort of an early example of work on a topic that subsequently became pretty important in the mixed signal world, handling heat in IC packaging. But, so I worked on that as a PhD topic and later on applied some of that and some of the work at Fairchild and other places to designing power amplifiers.

RW: So what moved you to California?

PG: Well, I, when I was finishing up in '69, I knew I didn't want to stay in Tucson so I did start an interview process, interviewed a lot of high tech companies, and actually met a lot of people, that was a lot of, I tell my students now whenever I can that when it comes time to go out and interview make sure you, you know, take advantage of the opportunity to meet, to visit many companies. It's one of the few times in your life when you have free, you know, pretty much unrestricted entrée to a wide variety of different kinds of companies and get to meet some of the key people in those companies. So for example I met Jim Solomon, who had a long career in the CAD industry after that. At that time he was a design manager at Motorola in Phoenix. I met Bob Pepper, who at that time was a design manager at Sprague Electric and went on to have a long career both at RCA and later on several other places. And anyway, so I interviewed a number of places, one of those was Fairchild. I came out to Fairchild R and D, visited, interviewed there. That was quite an exciting place at that time. That was early 1969 and that was a cast of characters out of whom spun many companies and so I've, I was quite taken with Fairchild R and D, went to work there, started there in June of 1969. I remember driving out here and listening to the radio driving up from Tucson, and that was the time of People's Park, hearing people getting shot in Berkeley from roof tops and thinking oh my god, what are we getting ourselves into here? But anyway, that's how I ended up at Fairchild.

RW: Did you take Andy Grove's semiconductor course?

PG: I took, I actually took that course and I, a couple, year or two later was one of the instructors in that, but I didn't cross with Andy, Intel was formed just before I got there in late sixty, actually not, it was just, it was sometime just before I arrived. So the semiconductor course was still going, the people teaching at it by the time I got there were folks like Ron Whittier, Albert Yu, Phil Downing, and some of those other folks that didn't go on to Intel until somewhat later. But by the time I got there both Bob Noyce and Andy had gone onto Intel. And, but that course was really a remarkable thing at the time, you know, each of the new employees who came to Fairchild R and D and many of the people who went into the operating divisions came up and took that course. Of course at that time, training in semiconductor technology in the universities where these people were coming from wasn't terribly strong, and so you really needed a way to sort of bring them up to speed on what was happening in the technology at that time. And at that point of course, the silicon gate tech, silicon gate and MOS had just sort of made its way into the world or into a sort of mass production kind of environment, and that was part of what Intel was based on. And they were, you know, propagating that knowledge in technology to a lot of the people who took that course. And then a year or so after I got there we added to the course kind of a mixed signal or analog component, and so my part of it that I did a couple cycles of was, you know, basic analog design, operational amplifiers, things like that. And, but that was, those were good years. There was a lot of excitement at that laboratory, and of course, it spun off an awful lot of companies, including some important mixed signal and analog companies at that time, PMI, and some others like that.

RW: Hogan is one of the interviewees in this series, and he told me that, I asked him how come things kind of started downhill at Fairchild under your watch, and he said "I was not well the whole time," he said, "I'd had an operation, and I wasn't up to speed and that's when it all started to fall to pieces."

PG: Yeah, I've seen, Les is a big supporter of ours at CAL and we see quite a bit of him and I've had similar conversations with him and he was of course an idol at the time, at least for me, he was quite a figure having coming from Motorola and I know there was, he was, I'm aware of that fact and I think it really did affect him. But you know, there were sort of another, there was another problem with the basic model that we had this, you know, Fairchild invested a tremendous amount in R and D without much ability to protect that investment, and so they had a marvelous collection of technical talent out there doing sort of university setting kinds of research projects, very, very successful, but they didn't succeed very well in translating a lot of those things into the manufacturing domain. There were, they hadn't developed a good way to rapidly propagate the things from the R and D environment into the manufacturing and product areas. And there was also not a very strong tradition of intellectual property protection. Not that that necessarily would have been the right solution to the problem, but the combination led to this situation where a lot of the money that was invested in things at Fairchild R and D at that time ended up being the basis of founding other companies. And, and I think that was, you know, there's no question that Fairchild would, was affected in negative ways by having all these competitors springing up all over the place, particularly in the silicon gate MOS, which really took a vastly longer time to propagate into the Fairchild product line than it did to propagate into the Intel and the other competing companies product lines even though it was basically big, large chunks of that technology were developed at Fairchild at the time. And that same pattern happened other places within Fairchild also, so it was a fantastic collection of people and ideas and activities, but it didn't, the model didn't work well for translating that to the bottom line of Fairchild's business at the time, and I'm not sure what, you know, that may have been something that Les could have figured out how to deal with had he been more able at the time.

RW: Yes, Andy Grove gave a talk about the 1101 or the 1103, I don't remember which, when it had come out, and Fairchild was, Intel was about a year and a half old and he said of this product, this proves that we can transfer technology from Palo Alto to Mountain View in a year's time. Meaning Fairchild R and D in Palo Alto to Intel production in Mountain View, so it was really blatant at the time and people would leave with processes and technologies.

PG: You know the Intel culture, as, you know, which the leadership of Intel came out of that Fairchild experience, and they, you know, they've adopted some interesting differences in the way they've approached things. They really don't have centralized R and D per se. They do a much better job of, I think, of translating things rapidly from conceptual stage into products by moving people, a lot of people movement, you know, typically people will carry things from conceptual stage to product stage instead of having sort of R and D researchers over here and product guys over here and they sort of talk to each other now and then. If they can keep that flow going, it helps translate things much more rapidly. But, you know, solving that's a real issue I think for any industry that has a basic R and D organization. How do you get that flow going, get the flow, break down the barriers for getting products in the markets?

RW: Well, jumping ahead, another outfit that's learned how to do that is the University of California, where research projects turn up in products relatively quickly. How is that done?

PG: I think that's an interesting point because it's a little bit of, that has been true historically but is especially true now. We have faculty members taking leaves to go start companies often. It's becoming a pattern, it's happening at Stanford also. And I think it's people, it's, you have create an ability for a faculty member or graduate students to conceive, work on, develop an idea in a university setting and then through a mechanism of either taking a leave of absence if it's a faculty member or if it's a graduate student they might graduate and go work for a company, but that person, the same people who kind of have the fundamental concept in their mind of what this is, this idea or this technology, migrate into the domain where they've got to take this thing and translate it to something that's actually manufacturable and gets into the real world and is a product. That's a powerful tool because it, you don't have this translation and if you look at the instances and some of them are in the mixed signal domain which we'll probably talk about, but other ones, in the software internet search domain, the key is you've got to personally, you've got the person migrating and I think part of the reason you're seeing a lot of that now is that we don't, it's a little bit of an issue for us just how strongly to encourage this because it could get to be, you know, a real issue if you have too many at any given time, faculty members off starting companies, but it seems clear that if you can create a way for university research to get into the, you know, you're whole idea is to have impact, you'd like faculty to do research that creates ideas, technologies that get into the commercial domain that have impact and create jobs and improve society and all that. And to the extent you can get faculty to go and take time off and go help do that by starting enterprises, that's good, but you just don't want to have it become sort of too pervasive so that it distorts what the university is really all about, which is creating new knowledge. And, so anyway, it's a, but it's, there's a, that's a, particularly in today's environment with, you know, with the entrepreneurship going on, it's really becoming a common pattern.

RW: Well we kind of jumped ahead there, going back to Fairchild.

PG: Right, so I went there in '69; I was working in a group there called the Advanced Linear Technology Development Group, something like that, led by a fellow named John McDougal who was a supervisor at the time. I got involved in a couple of projects, one of them was development of a power amplifier stemming on this thermal work I'd done in college; another one was doing some technology development on a particular type of mixed signal process called the Super Beta Technology. I worked on that, on those things for a couple years, learned a lot about the, you know how a semiconductor company works and then I got a chance to go to Berkeley as a visitor. I ran, met Don Peterson at a conference. Don at that time was leading the CAD effort at Berkeley and his students were busy developing SPICE. Don asked me to come to Berkeley as a visitor for a year, just to do some teaching, try something different. I thought that was a great idea. So I did that, I went there in September of '91, taught some courses and got involved with some graduate students and started doing some research on various more advanced things than I'd been working on at Fairchild and ended up staying. The long and short of it was I ended up staying there. But one of the things we got started on there was all of the mixed signal integrated circuits there. At that time the word 'mixed signal' hadn't really come along yet, it was analog. If you had, you had analog and digital. And of course digital had been implemented up to maybe 1966 or '67 and all bipolar, and then when the MOS technology began to come along, first as, you know, a metal gate PMOS and then various flavors of MOS technology, the digital world was beginning to move rather rapidly, especially memory into the digital and the MOS technology. But at that time the mixed signal and analog functions were, these were the functions that allowed digital systems to talk to the real world. So you have a, you know, the real world is a world of analog quantities like pressure and sound and light and quantities that vary continuously and digital processors and memories are digital. They store information in the form of ones and zeros, so you need this translational circuitry that converts the, conditions these analog signals, takes them off the sensors that detect them and convert them to digital, and back again. So a good example's a CD player, where you have an analog sound that you're trying to generate but it's stored in the form of digital bits so you have all this translational digital-analog and analog to digital conversion to do. And lots of other examples: communications systems, data storage systems. All most all electronic systems have this and so it's an important part of the IC world. Maybe somewhere between 15 and 20 percent in terms of actual manufacturing of revenues today are sort of in mixed signal and analog chips if you take the microelectronics industry as a whole. So anyway, at that time, all of the mixed signal analog functions were done in bipolar technology, which was the original IC technology. And the work I did at Fairchild was in that same domain, more of that same sort of thing. At Berkeley I worked with my colleagues, Dave Hodges at that time, and it struck us that it was obvious that MOS was going to really take over the world of digital because it had self isolating technology, much better density. Lots of reasons why it was a better technology and it sort of hit us, you know, someday you're going to need to build a, you know, one of these digital chips where this analog interface stuff right on it and, you know, integrate everything. And so we ought to be looking at how you build these mixed signal analog, you know, filtering and analog to digital conversion and the conversion from digital to analog and those kinds of things in the MOS domain. So it turns out it's really hard to do that because MOS is really different, it's a different kind of device, it has a lot of less desirable characteristics as a transistor than a bipolar does and so we started working on that. I had a couple of graduate students who worked on analog. How would you do an analog to digital converter in this MOS technology? So, you know, it turned out that you needed a completely different approach, because an MOS transistor is so different, it has a, one of the things that makes it really useful for memory is, you know, an MOS transistor has an ability to detect a charge on a capacitor non-destructively. You can sort of sense a charge on a capacitor, so you, without destroying that charge, which you can't do with a bipolar transistor and so that allows you to build these dynamic memories. The D-RAM business is based on the fact that you can store information on a capacitor as a charge and detect that information with a gate of an MOS transistor and that can't be done in bipolar. So if you translate that to the world of mixed signal, it means that maybe you should be using capacitors as the sort of elements that you build a circuit out of rather then resistors which is what you do in bipolar transistors. So, and to make a long story short, we figured out ways to build A to D converters and then later on various, something called a switch capacitor filter which is a filter that filters analog signals using capacitors and MOS transistors instead of the traditional approaches in bipolar technology which use resistors and bipolar transistors. And, so that was percolating along, going very well, but still at that time, that was a research project in a university that nobody in the real world of Silicon Valley was paying any attention to. This was, when I talked to, I remember I was on a panel discussion at ISSCC about this one time at about that time frame and about A to D converters and I remember I gave my little five minute spiel at the beginning of that and I piped up and I said, yeah we think you could probably build these A to D converters out of capacitors and the audience just burst out laughing, you know, "Oh what do you, what is that capacitor" you know, that was a nutty idea at the time. Nobody could really see why you would want to do that, you know, it just didn't seem to make much sense. So anyways, along about that same time, I was doing a lot of consulting with, first with Signetics, but then later with Intel on, working with Ted Hoff, and so we had this stuff going on at Berkeley and parallel with that Ted, who had just at that time finished his work on the micro-processor, he was kind of, kind of got that finished and he was kind of ready for his next thing and I met him in about 1974-'75 and I went ultimately from that went and worked with him as a consultant one day a week for, while I was on the faculty at Berkeley and he was interested in that time, he was looking around for new markets, for chips, I mean, what's the next thing we're going to do here? And being the bright guy he is he sort of figured out well, you know, one way to do that if you're a semiconductor guy is to think about what are the things in the world for which there is one for every human being, you know, and those are good markets because there are a lot of human beings, and well there's TV sets, and there's, but there's telephones. You know, and it sort of hit him, Gee, Intel really ought to be looking at the electronics that goes in association with the telephone and in particular the entire of transition of the whole telephone system to electronics from what it was before which was relays. And the up-shoot of that was we got involved in a project down there to do a, the coding, the filtering in analog to digital conversion that goes at the other end of the telephone line. You pick up your telephone and down at the other end of the wires there used to be relay now there's an analog to digital converter and a filter and then behind that is a big computer. So everything switched digitally and a telephone central office that your phone wires go down to really looks like a giant digital computer with a bunch of A to D converters and D to A converters connecting to telephone wires. So Ted's idea was Gee, we ought to be in that business because there's a lot of telephones. So we worked together on this and it sort of, it's one of those kind of serendipitous things. We're doing this thing at Berkeley on these filtering techniques and A to D conversion techniques together with my colleagues, both Bob Broderson and Dave Hodges contributed a lot to that. A lot of graduate students along the way were working on that, and Jim Creary, and Ricardo Suarez Gardner, and then it sort of matched up. I remember driving home in my car one day and saying Gee, you know, this codec stuff, this telephone thing is just an A to D converter and we've got this scheme up here for doing these A to D converters and not only that you need this filter thing and we've got this way of doing filters and then there's a lot of back and forth. There was another filter technology coming along at the same time, CCDs, CCD filters and we thought that was going to be a way to do a good approach, turned out it wasn't. Long story short, it sort of developed that once we published the work at Berkeley on these filters and A to D's, that I was able to then go down to Intel for a year, rather quick, shortly after that, about 6 months after that and help Ted and that team translate these university nutty ideas into some real products that ended up sort of being, you know, today if you pick up your telephone in North America there's about a seventy five percent chance you're talking through one of those chips of that type, obviously not that very one. But, so it had a pretty big impact, and serendipitous, we just, I guess I attribute a couple things to that: I know we're close to out of time on this tape but, picking the right thing to work on. If I hadn't been going back and forth and talking to Ted and sort of understanding a nice application, we might not have worked on that particular thing. Second thing is getting the thing, making, getting the idea from a university setting where it's just a nutty idea into a real product as fast as you can. And the people, having the person, either a faculty member or a graduate student go do it is really the best way because it, you know, you don't lose anything in that transition.

RW: Well then your work and your people at UC that was really the fundamental work done in MOS and mixed signal applications, is that true?

PG: Well, there were, I think we made some real contributions there. I should say there were a lot of other people working on this at the time, and I should maybe try to detail that out a little bit. Of course, Ted himself you have to give a lot of credit to for recognizing that this was going to be important. He at that time, you know, he was really in the digital world. And in fact in parallel with this project we're talking about, he had another pet project which was digital signal processing. He was one of the first people to recognize that this was going to be really important and that integrative digital signal processors were going to be a big deal. And he had another parallel project at that time to do one of those at Intel, which ultimately was not successful for a variety of funny reasons not relating to the fundamental concept. But then TI went on and took that in to make today, it's a major business for TI and some others, but getting back to the point, at that time, I think we were, you know we, our graduate students, Jim Creary, and Harry Lee and some of those folks were I think the earliest to utilize the capacitive A to D conversion, chargeer distribution A to D conversion technique for A to D conversion in MOS technology and that technique was patented and is widely used in PCM Codecs even up to the present time. Another element of this is the switch capacitor filter idea, which is another sort of key element of analog digital interfaces. When you convert something to analog, from the analog domain to digital, you have to first filter it. You have to take out some of the high frequency parts of the signal and then you sample it and analog to digital convert it. So you have a filter part and an analog to digital conversion part, and the filter part, we had another sequence of graduate students, David Allstot, Ian Young who's now with Intel, and a number of others who I'm sure I'm going to leave out so I won't try to name them all, but developed that switched capacitor filter technology which was another sort of pretty new idea using switches and capacitors and op amps to do filters. That technology was worked on in parallel by quite a few other people. There was a fellow named Miles Copeland at Carleton University who was pretty close to parallel in developing that with us and his sort of visibility came quite a bit later, but I think in reality he did quite a bit on that and I think deserves a lot of the credit for that. But that technology had a lot of impact and, you know, if you look at the world of analog and mixed signal today, there aren't too many switch capacitor filters around. Now that technology has migrated into the A to D converter world. There's sort of a, there's a technique called sigma-delta modulation which uses switched capacitor integrators in a way that accomplishes A to D conversions and some ancillary filtering together. But that technology had a lot of impact and, you know, if you look at the world of analog and mixed signal today, there aren't too many switch capacitor filters around. Now that technology has migrated into the A to D converter world. There's sort of a, there's a technique called sigma delta modulation which uses switch capacitor integrators in a way that accomplishes A to D conversions and some ancillary filtering together. But that technology has had quite a wide spread use in mixed signal analog digital interfaces and I think you know, it's right to say that a number of people in university settings were working on it at the time. I think we made a good contribution to that technology but a lot of people were working on that. And I should also say that in terms of the commercial development of that, the Codec filter for telephone applications was the very, that was kind of the prototype. So if you look at the world today you have all kinds of these things, you have ISDN transceivers for telephony, you have disk drive interfaces, you have A to D and D to A converters for instrumentation, sensory interfaces, all sorts of things with, that involve the same kinds of functions, ADSL, XDSL, all kinds of modems, all sorts of things that involve the same basic sets of functions of filtering and analog digital conversion, digital analog conversion, filtering, some sort of media driver, line driver, what have you. But the Codec filter was really, for voice telephone, was the first example of that. And so it sort of was, you know, a really important first demonstration that you could actually integrate on a single chip all the elements of this analog digital interface function. Today, here in the 1998, that entire function, all those things together is, you know, you get a number of manufacturers sell multiple, chips that contain multiple instances, cost per line is fifty cents, something like that. Back in 1975, when we're talking, in 1978, it was ten dollars, twelve dollars a line, so there's been a huge reduction in price.

RW: But Ted Hoff and Intel were the first.

PG: Ted, the first commercial, let's see, the first commercial codec A to D converter was the Intel 2912. And that was the first, commercially available A to D converter for, and let me tell you what else was going on at the time. Another graduate student of ours who had worked on the same project went to Siliconics and Siliconics developed an A to D converter for this same application. Another graduate student, Dave Allstot, and two graduate students, Bill Black and Dave Allstot, went to National. National under that time under Jim Solomon's leadership, developed a codec and filter chip set. This was all going on in parallel, and, oh and there was another graduate student, by the name of Gordy Jacobs who went to work for at that time SSI, Silicon Systems, and they developed another switch capacitor circuit, I think this, if I recall right, that was a modem front end filter, and all of these were happening. So I'm not exactly sure I can recollect the exact sequence of, you know, when something's really a product and when there are samples and when the data sheets out and when people are shipping a million month, it's sometimes hard to identify who's really there first. But, to a first approximation, I think it's, if you really look at the chronology, Ted was the first guy to really, that project was really leading it. And whether or not the actual product to market happened first, I'm not actually completely sure, but I think he deserves the credit for having that vision ahead of the other folks. I think it's true that, from a product standpoint, the other people were following. They saw what was going on and they really jumped in.

RW: Yeah, well he's famous for the microprocessor, but the codec is essentially unknown. I have a 1998 Ted Hoff story. Ted's a friend of mine, and my wife and I had just written a draft of what we hoped to be an article called 'Please Don't Call Intel A Semiconductor Company.' And the gist of the draft is that Intel is now a computer company and the difference between them and IBM is miniscule, and they started off as a semiconductor company, but they're now a computer company. So I was talking to Ted about that and he said Bob Noyce would turn over in his grave if he heard that because Bob Noyce once told Ted we know the computer business and we don't want to be in the computer business. We don't know the telephone business, so let's give that a try. So a little aside. Well we kind of jumped ahead, before MOS there was bipolar, the first integrated circuits were bipolar circuits. Were you involved in that at all?

PG: Well, as a late comer, but I was more of an observer of that era, and, you know, one of the things that got me interested in this, in integrated circuits as a graduate student was reading some papers and journals and also magazine stories about the work that Bob Widlar and Talbot at that time at National had been doing along with a guy named Tom Fredrickson who was at that time at Motorola. This was, these were the, 1966-1967 when we were just coming out of an era of discretes, right, the digital integrated circuit had been invented, now you could go out and buy TTL and, well maybe the predecessors to TTL, so there was quite a body of digital ICs out there and the technology was kind of getting to the point where you could make 50 transistor circuits reliably, and manufacture them, and Bob Widlar, I think deserves probably more credit maybe then any other individual for really launching the analog integrated circuit world. He, you know, there were discrete analog, mixed analog discrete circuits like op amps and A to Ds used at that time, used a lot of precision elements, you know, a lot of NPN, high performance, the different kinds of transistors. So if you're building a discrete circuit on a printed circuit board, you were able to buy different flavors of transistor, you know, to suit the particular role that was playing in the IC. You could use precision elements; you could go buy resistors and capacitors that were one percent accurate, or 0.1 percent accurate. So designers at that time were used to having all these things to work with. And you know, the IC technology's very different. You know, in an integrated circuit at that time, a bipolar integrated circuit, you've got one kind of transistor, an NPN bipolar transistor, you've got some resistors that were, you know, 20 percent accurate if you were lucky, you maybe got another kind of transistor, a PNP which was really not demonstrated to work at all at that point in time, a lateral PNP. So Widlar came, from he was with an aerospace firm before that and came to National and, you know, was able to, he and his co-worker Talbot who was a process guy were able to take that technology and say now, let's throw away all the rules we ever used to design these, how would you do it taking advantage of what you can do in an IC. And one of the things you can do really well is that if you're making all these transistors at the same time, so they all match, even though they maybe have absolute characteristics will be different, but those two will be the same. And he, that single sort of observation or characteristic is sort of the core of a big chunk of mixed signal IC design utilizing matching of components as opposed to their absolute characteristics. And he recognized that and, you know, developed a series of, at that time, just totally ground-breaking products. The first one was a series of operational amplifiers which are, you know, a type of amplifier used in various, lots of different applications that had characteristics that were not quite as good as what, well, fell short of what you could do with a, you know, a full blown circuit made on a PC board with all these precision components, but were pretty darn close. And of course form factors, much smaller cost was vastly lower and really, you know, also pioneered the use of a particular kind of transistor called a lateral P and P which wasn't intended to be provided in the technology but which you could get if you sort of did things a certain way and made the diffusions a certain shape. And it was an example of a person with a deep understanding of sort of physics, how the technology works, you know, how a transistor works, the different shapes and the diffusions and the, how the minority carriers, not to use jargon, but how the transistors worked and then you'd look at how am I going to take this technology and do things in a way that people haven't even thought about to make this function I need which is this operational amplifier, using techniques that nobody's ever used before. And that had a huge impact and really launched the 702, the 709 op amp, and then he went on to do another project called, well, a voltage regulator, which is just what it says. You got a computer board and you got a power supply voltage that's varying and you want to generate a closely regulated constant voltage on the board, and you want to do that with a chip, and never been done before with a single chip, and develop some more new ideas. Band gap...something called the ban gap regulator, which is just a way of generating an absolute voltage reference on a chip using bipolar transistors, another new idea, and launched a whole, what now is a huge business of continuous linear voltage regulators and just had a tremendous impact. So if you look at the bipolar world, which is still important, you know, I don't know the breakdown but bipolar is probably still some reasonable fraction, maybe twenty or thirty percent of the world of mixed signal analog ICs, uses his techniques all over the place. And they're pervasive. So it had a huge impact. And of course he sort of flourished in that era and then later in life faded a little bit from the IC scene in the '80's and then died subsequently, but he was really a pioneer of this business. There were several other people that came out of that era, Dave Fullagar, who's, went on to be, work with Intersil and Maxim was the developer of the 741 op amp, which was the first internally compensated operational amplifier, tremendous market success. I don't think it's true anymore, but for many years that was the, by far the largest, manufactured in larger volume than any other operational amplifier. So there were a number of people from that era, Barry Gilbert. Barry was a worker, now with Analog devices, who pioneered a number of other basic circuits that are used to this day in many, many places, has a circuit named after him, the Gilbert Multiplier, widely used in communications. So another very creative, very original person who pioneered a lot of the basic bipolar, analog, circuit design and so those guys were real pioneers. There were a lot of other folks who made contributions, but if you sort of think of who really drove the field, they've had a lasting impact. And so that, the bipolar world really flourished and came along and then the MOS kind of the MOS kind of development came in the mid '70's and you began to see a transition where, you know, the bipolar is really best suited to, and it's still true to this day. You sort of have, here's a way to look a the mixed signal world, this sort of egg shell part is earlier, you know, the standard implementation of some mixed signal function of let's say a modem or a PCM codec or something like that, was to buy off the shelf components, op amps, things like that, and put them together with resistors and things like that and make a board level implementation out of off the shelf components. In the digital world, that era is gone, right, the era of sort of TTL low level integration is pretty much gone, its given way to things like ASICs and micro-programmable micro-processors, programmable DSPs. So you have a lot of different technologies that allow you to achieve high integration even if the manufacturing volume isn't very high. So you have the ASIC technology that can allow you to get a high integration solution to your problem even if you're only going to make, you know, 10,000 of them or 5,000 of them, or a few hundred of them. But the analog world is not like that. The mixed signal world, you know, we've got the discrete components, sort of TTL off the shelf solution, which, and then we've got the high integration solution where things like PCM codex for voice band data modems, single chip implementation of that, ISDM transceivers, these high integration sort of application specific standard components that tend to evolve where you have a very high volume application with standards driven and you can define a standard, high integration component to do that, that's flourishing. But we don't have the thing that the ASIC world has, the digital ASIC world, where you have an ability to define, to customize a function in an analog digital interface, and manufacture it economically for a low volume application. Let's say you need ten thousand of these things and you don't have an ability to do that even today, we don't have anything analogous to the ASIC technology in the mixed signal world. So the mixed signal world tends to divide into these sort of high integration, customized, application specific kinds of solutions that are sort of manufactured by people like Lucent for telecommunications, a wide variety of network driven people for Ethernet and kinds of data communications and other kinds of sort of custom standards driven kinds of applications. And then you have the world of off the shelf standard building blocks, op amps, voltage regulators, A to D converters, D to A converters, sample in holes, stand alone filters that you typically build a solution out of those on in some sort of PC board level solution. That is a, you know, a very large fraction of the mixed signal world is still being done using that off the shelf building block type of approach because we don't have this ASIC world.

RW: It's not for lack of trying. A variety of people have attempted that including
LSI Logic and it's failed, and one of the reasons its failed is that there are not testers available that have been developed that are general purpose and can be simply programmed in some automated fashion to test all these parameters. So that tends to be, the tests tend to be hand crafted which means you could have a hundred thousand dollars just involved in the test alone and the tester might be a five million dollar item.

PG: Yeah. Yeah, that's a big factor. Another big factor is these things tend to interact, you know, you put a bunch of mixed signal stuff on a chip and if you're not careful about how you route the power supplies, they're talking to each other through the power supplies, talking to each other through the substrate. There's a lot of process technology dependencies in these mixed signal circuits, so you have a lot of difficulty in migrating from foundry to foundry. And you put all that together and what it ends up meaning is that to actually execute this complex, integrated, what you hope to be an ASIC, ends up taking a lot of engineering effort on the part of the vendor. And engineering effort spells money, and you know this is a low volume, I mean the whole motivation is to provide this high integration solution for something that's not particularly high in user volume, and it's hard to do that if you have to invest a lot of engineering in it. So nobody has really cracked that problem of coming up with a CAD tool and a testing methodology that's smart enough and sort of bullet proof enough and robust enough that it can actually deliver a solution that works most of the time. So the companies, like Maxim and LTC and the other off the shelf component suppliers have been very, very successful over the past, you know, twenty years or so and growing because that approach to implementation is still a really main stream part of the mixed signal world, where as, you know, in the digital world, it's really gone. Now it's all ASICs pretty much. So it's a different. But anyway, so I say that for the, to illustrate the point that the contributions of, you know, why is bipolar still important, why hasn't the mixed signal world made a transition to MOS completely like the digital world pretty much has, and the real reason is that this off the shelf building block sort of element of the mixed signal world is still very important, it's still a big, big part, and bipolar's fine for that, you know. If all you're building is a, you know, a quad op amp or a voltage regulator or something that's really a building block, well it's fine it can do a fine job of many of those functions and so the bipolar design sort of expertise and methodology and implementation is still really important. And so a lot of the innovations that Barry Gilbert and Bob Widlar and those folks developed for bipolar is still used a lot in those kinds of products. And then the MOS, you know, the MOS and CMOS now, it's almost all CMOS now of course, you know, part of that world is kind of off with a different set of design methodologies. You've got the, you know, the crystal semiconductor, a very innovative company that's had a big impact in that and particularly with CD, you know, voice quality, sigma delta modulators for high quality voice for CD players and PC voice systems, you know, telecommunications kinds of applications, highly integrated, high integration solutions for communications, can be either wire line, or no, RF, a lot of RF activity in high, in MOS. So sort of the stand, the built for standard application, high integration mixed signal A to D interfaces are mostly going to the scaled CMOS kinds of implementations. And there it's pretty different, there you tend to see design techniques and design approaches and building blocks that are really drawn from the MOS world that began to build up in about 1975 with sort of a different cast of characters. And of course a lot of people thought that BiCMOS we were talking earlier that, you know, somehow this would all merge into a BiCMOS world where the bipolar and the CMOS would go together and then you could bring in the best of both worlds and the bipolar designers could bring their expertise to bear, and the MOS designers and so on. And its certainly been a factor in some areas and for some manufacturers Analog Devices would be one good example. It's been really important because it gets you time to market. You know if you have this BiCMOS technology, then you don't have to reinvent your bipolar circuits into MOS or vice versa, so it's been important in some areas but most, its also been true that in most areas where cost is important, it's developed that a really good design team can usually find a way to do the bipolar type functions in the MOS technology and if you can do that you always win because the pure, you know, pure old scaled CMOS is a lot cheaper than the BiCMOS is and so, you know, BiCMOS hasn't ended up being the sort of the end all technology that a lot of people thought it was going to be for mixed signal because of the cost factor and the fact that today, you know, you can go out and find, you know, if you're on a sort of a standard digital CMOS track, you can go to a lot of foundries and find very, very cost effective fab in sort of a main stream digital CMOS technology or maybe slightly modified. Where as finding BiCMOS is a lot more difficult. So you know, the world is sort of divided there into the sort of the pure CMOS domain and the bipolar with some BiCMOS in between.

RW: Well, switching gears a little bit here, the University of California does so much of the fundamental work in electrical engineering in my view far more than MIT for example, how come? What makes that happen?

PG: Well I think a lot about that because, you know being the Dean of the school over there we think a lot about what do we do to make sure that, you know, the future is as good and hopefully better than the past in terms of the impact the research can make and the impact the students can make. We have a lot of things going for us: there's a history of excellence and good people beget good people, that helps a lot; but if I thought about the elements that have really allowed the school to have some of the impact its had, you know, risk micro-processor work, raid the mixed signal stuff, development of SPICE, involvement in the UNIX operating system development, some of those impacts; proximity to Silicon Valley's a big factor. If I think about the history of some of those developments, a lot of them have the characteristic we were talking about earlier where you have an environment on the campus at Berkeley where a free flow of new ideas, everybody talks to everybody about everything. The challenge is to invent, you know, develop something that's original, creative and has impact and no limits. You can talk about anything, look at anything, doesn't matter how far in the future. And marrying that to Silicon Valley where, you know, people are trying to figure out how to three month timelines and how to get the next product out, many of the developments I mentioned happen when faculty members, and I'll just take my own case as an example with things like the codec work where there was a contact. We have a faculty member going down to Silicon Valley, he's exposed to a lot of real time development that's going on, limits, problems, and he can think about what's the next thing these guys really need, what development would really enable the next generation of products in this particular area to happen and then go back to Berkeley and have a chance to seed some ideas, talk to grad students, talk to colleagues, talk to people in other areas, maybe system oriented, maybe software oriented, maybe CAD oriented, and get some synergy with other areas of techniques or approaches that might have been used elsewhere. I think that marriage is important. So I guess what I'm saying is the physical proximity. If you look at strong universities in the U.S, top three universities are Berkeley, Stanford and MIT in engineering. Well those are three universities that are located in the two centers of high technology in the United States, third being Austin, I suppose, and they maybe haven't reached that same domain, but that's important, that proximity. So proximity and ability to be closely coupled with what's happening in the industry that you're concerned with, that's really important. You know, obviously getting the best people. If I think of what makes a university great, I, people come and ask me often from sort of Midwestern universities who are trying to improve, they'll come and say, Gee, what do we do to get to be a Top Five school? We'd like, our goal is in 25 years to be a Top Five university in engineering. And you know, I think if you're located in a small Midwestern state a thousand miles from the industry you're trying to serve, it maybe impossible. I'm not sure you can ever do that. But if you were going to try, you know, the key is to hire the most creative young people you can find. If you can go out and in any given year do a recruitment and identify and attract the, one of the top two or three best young people coming out of an engineering program in the particular field you're interested in, maybe it's microelectronics, maybe it's computing, maybe it's networking, maybe it's biotech, whatever that is, if you can identify and recruit the top two or three people and then you, you know, guys with jobs like mine as Deans of colleges, we're a lot like farmers. I mean, we find the best seeds we can find, we plant those, we go out and make sure that we get the best fertilizer and the best environment; we make the sun shine as often as we can and just let them do their thing. And I think if you do that and if you have the proximity the rest happens. You know, we have a tremendous advantage in the Bay Area. When we go recruit faculty members, everybody wants to live in Northern California, so we usually win when we're out there trying to recruit, we go sort of fifty-fifty with Stanford, but when we compete against anybody else for the best people, we get them because with MIT we usually win because Boston is just not as attractive anymore. This is where the action is and so that, it's people and connections to industry and all the rest is sort of, you know, well there's a thousand things that are secondary, you know, making sure you have the best facilities, making sure you foster these relationships, making sure you're coupled to industry, etc., etc., making sure you have the best graduate students. A very important part of any great university is the fact that graduate students do all this stuff. When I outlined that scenario to you and talked about these things, you know, a lot of what happens with, you know, risk, Bill Joy, the UNIX guy, goes to Sun and he's a key guy and founding Sun and he's a graduate student at Berkeley. Our grad students have gone on to seed all of this in industry were some of the best students, so, you know, at Berkeley and Stanford, MIT you get the cream of the crop of the graduate students who come out of the undergraduate programs around the U.S. and, you know, it all kind of goes together. It's a, excellence seeds excellence and, you know, because you have a great university you get, you get the best students coming in, you're able to recruit the best faculty in it. So when a Dean from a Midwestern small school comes and says how do we get to be in the Top Five, boy, I, you know, I don't know, I think it's really hard. You've got this sort of momentum thing going and it's really tough to move up and I think sustaining it is really a, I guess I'd say one other thing, you know Berkeley in the, Berkeley's always been, even dating back into the pre-war era, a top university. Its always been ranked in engineering in the top few. So this isn't sort of a new thing, but what's a new thing is the Silicon Valley, high tech, networking, microelectronics involvement and I think there if you just look back and sort of say well how did that actually happen? Well, you know, you had people like Don Peterson, David Hodges who's another close colleague of mine that worked on a lot of this stuff, also was the Dean of the college before me, you know, Dave Patterson, Randy Katz. You know those guys really took the initiative. You had the individual faculty there who by their very nature looked around and said, Wow, this, you know, there's really something happening here, we've got to be a part of this, I want to go start consulting with, you know, this company or I want to go down there and get a research relationship with that company or we want to work with these guys. And it propagated, now I think that we...You know Stanford had a closer proximity. They had a visionary dean, Fred Terman, who really, you know, you've got to credit with having incredible vision and founding the industrial park and starting that program. In the early days, as I mentioned, you know, you had the two universities and then also Arizona and MIT a little later having on campus microelectronics laboratories, which really went a long way toward creating this synergy because there was a lot of back and forth with that. That created a lot of ties between the campuses and industry. But I think that, you know, you wonder if in the 1960's if the person who had been Dean at the time at Berkeley had decided let's create an industrial park here in the area somewhere, perhaps out in Richmond or, and see this a little more. Whether a part of Silicon Valley might have ended up being up here, there would have been some problems. There wasn't as much land available here, other reasons why that might have been difficult, but you have the feeling we might have been able to do more than that. Of course, what has subsequently happened is biotech has really seeded out here in the North Bay and that has come mainly out of the biology, biotech parts of the Berkeley campus and UCSF. An important ingredient of that is that UCSF is, you know, it's one of the leading medical schools in the United States and the biology part of the Berkeley campus have worked together to really seed that. So that's really grown up around that sort of complex and has happened really more in the North Bay than down in Silicon Valley. So, you know, probably past history, water under the bridge, but it might have been interesting to try to seed more of a high tech high industry presence in Berkeley. There is a substantial software and networking, we have a lot of things going on in the area in terms of small startups. Mostly focused more on software and Internet based kinds of things as opposed to sort of silicon technology per say. But, you know, we certainly have had our major role in Silicon Valley so I'm not too worried about that. And that's really continuing, you know, if you look at things going on now, wow, you have a lot of internet networking kinds of startups, the Ethernet sort of thing, a lot of things related to applications of silicon technology, micro-machine kinds of things, you know, the micro-sensors, micro-actuators, where you're taking silicon technology and using it to make things like micro-accelerometers that might make their way into all sorts of applications like human IO, virtual keyboards where you have an accelerometer on each fingertip and you type on a bare surface and you see on your lap top screen a virtual keyboard and it allows you to eliminate the keyboard because you have these sensors that you can use to do IO with. Stuff like that, there's a lot of that going on and so, you know, it's going to keep on being a center of activity.

RW: One thing I've noticed about the present day EE department, so many of the faculty and the students are foreign born, in particular, ethnic Chinese from the mainland, from Taiwan, from Hong Kong. What's behind that?

PG: Well that's a good question, a lot going on there. I'll just make some observations about that, it is a fact. At Berkeley we have, let me give the facts because it's important, this is a complicated issue. At the undergraduate level at Berkeley we have about forty percent undergraduate students in engineering that are Asian American students. These are American citizens who apply with everybody else and get admitted, a very small number of foreign students at the undergraduate level, very small, less then five percent. At the graduate level we have about twenty five percent non-U.S. students admitted and actually in the program we have about thirty five percent non-U.S. students at this point in time, 1998. Those are mostly Asian students. Those are non-citizens, they are foreign, from foreign countries, they apply, they come here as a graduate student and I would guess that about twenty five percent of our graduate students are Asian, non-U.S. foreign students. And in addition we have about another twenty percent of those graduate students who are American citizens who are of Asian decent. So it's kind of important to distinguish there. But, mainly, we have a very large number of people of Asian decent of some sort on the campus and in the semiconductor industry and computer industry and so why is that? Well, you know if you, the phenomena we see is that in Asia especially, but foreign countries in general, graduate education, PhD career in the academic world, but certainly a professional career at a graduate level is much more highly sought after. We don't have as large a percentage of American students, particularly students coming out of high school and students coming out of undergraduate degrees who are motivated to pursue graduate degrees. So when we look at the admit pool for graduate school, we see very highly qualified international students from around the world and especially from Hong Kong, Taiwan, Korea, the other Asian countries, not so much Japan, who want to come to Berkeley. And it's true, UCLA, and Stanford and other places. And we had, we actually have a differential admissions that, the threshold of admission to graduate school for those students is much higher, because we limit the numbers to twenty five percent non-U.S. and so they're much more competitive in terms of admits than the U.S. students. Why is that? Why is that? I don't have a good answer to that. It seems that due to a number of factors we just have a smaller percentage of students in the United States who are motivated to study engineering in the first place and then once you get to the graduate level to go on and get graduate degrees and have careers that might be academic or might be industrial at an advanced level. I'm not a sociologist, I'd have trouble speculating on why that is, but it's certainly true. And it's also true that if you, you know, we have this controversy right now with the H1 visas and how many foreign students, how many foreign workers should be admitted to the United States. The fact is if you go down to Silicon Valley, and you know, you walk down the hall at Intel and you realize that we'd of never built this industry without those workers. I mean we have to have these people. They're an incredibly talented talented pool. They add incredible value and they work very, very hard and as a group we couldn't do without them. So you know, I think we have to, we just have to continue to, I think it's a net plus for the country, we import great manpower. And so, but as to why it is, that's an interesting question. I think you do see, I see in the second and third generation Asian American students who've come from those backgrounds, I think I sense among the students I see a tendency to become more like the typical U.S, I think it's a cultural thing. The work ethic and the dedication to the idea of having an academic career or having an advanced degree is something that seems to go with cultures who've been through very difficult post-war periods. Typically the parents may have come from very humble beginnings and they have been wiped out or maybe it's the grandparents now, wiped out in World War II. In the case of Taiwan, you have a very tortured history for that country and I think maybe this work ethic and this desire and willingness to really seek a higher goal comes partly from that cultural background. Kind of speculative, but I think it probably does. But the fact is we have a lot of very highly qualified foreign students and they're a great resource for us.

RW: Yeah, there's no question about it. I think that is one of the reasons we're now doing well vis-a-vis the Japanese who utilized Japanese talent for their engineering while we utilized the best in the world here in the U.S. And of late, we've, as superb as the Japanese are we have to credit them in so many areas, especially automobiles, but in the last ten years or so the U.S. has pulled ahead in a lot of high tech areas where we were, we were thought that we were going to be slip to number second. And I think it, I think one of the big reasons is that we recruit the best in the world and they don't. And we recruit the best in Japan for engineering. Well talking about success, what do you perceive as the elements of success in high tech of the, you've known a lot of these folks?

PG: Yeah, well I, you know, I get a chance to talk to students quite a bit and I try to, you know, I, there's a few things I try to tell them not that I, you know, any particular great shakes as a philosopher, but you know it's, the people, one common thread that I see is that successful people tend to be driven by love of what they do and excitement about what they do and not about getting rich and I try to tell the students to, you know, make sure they don't seek a job because it has a big stock option package or something like that. Go and do the job that excites them even if it means they, you know, they have to eat beans for a while. If they get involved in something that they really can get excited about and put all their energy in and be the best at, that'll, all the other rewards will come later on. And that's the most important thing. And I think, you know, as you pointed out earlier, I think that's a pattern you see among this group of people who end up succeeding at these enterprises. They really love what they do and get a big charge out of the exercise, the enterprise, the thrill of doing it. And sometimes you see cases where the, you know, success is almost not so much a blessing because you've sort of done it and now what do you do next, but that's one thing. Another thing is I try to tell them to, you know, be aggressive, really. It's easy if you succeed at all the things that you try you've almost certainly set your sights too low, you can do, it's amazing what you can do when you take a big problem and you break it down into a bunch of small problems and do each little problem one at a time and you work for a couple years and you look back and you say my god, how did we ever do this? But, if you just work on each individual problem one at a time and then you'll get the big problem solved and you know, don't be too timid, don't, be aggressive and don't set your sights to low. I try to get them to think like that and I think there's a lot of evidence that you can do just amazing things in this, you know, if the Intel guys had said, you know, said in 1968 that they were going to be what they are today, no one would have ever believed it. And never underestimate the power of the technology we're involved in, it's really amazing. And then never believe your own press releases. We were talking about that. I think a lot of people have sort of fallen on hard times by not being honest with themselves about what was really going on and believing the hype and not the reality and of course it's easy to say that, but I think that's, it's nice if the students can kind of get a sense of that. It's being honest with themselves as the first judge of what's really happening and not believing the press releases. So those are some things that I try to get them to latch onto.

RW: Well that's great Paul. Thanks so much for taking the time out to give us these words.

PG: Well Rob I want to compliment you on this series. I think it's, capturing this is really a valuable thing. I really am going to find the time sometime to go down and watch some of these in the library because I suspect they'll, certainly not in my case, but some of the other cases you've captured I think will add a lot of historical perspective for people who look at these some time down the line.

RW: Well thank you Paul.