Transistor Museum Transcript of Interview with Wilf Corrigan, Sept 2006

Transcript of Interview with Wilf Corrigan, Sept 24, 2006

Interviewer: Jack Ward, Curator – TransistorMuseum.com

1) Hello Wilf. Thanks for agreeing to participate in this interview. It is notable that you have been involved with semiconductors for over 45 years, and that you were a very young man when you started to work with silicon transistors in 1960.

Yes, it was unusual. Most of the guys were a generation or half a generation older than I was at the time. So, I just happened to come in at the particular point in the industry where it was very easy for younger people to get senior jobs. I knew, for example, Bob Noyce, and he was running Fairchild Semiconductor well before he was 30.

2) It must have been quite a change to come from England and go to Transitron in Boston.

Well, it was my first job. In fact, George Wells just brought a book out. In the book, he goes into that period of time at Transitron. He was about a year ahead of me - I was there for only six months. (This was pretty crude technology?) Oh yeah. The crystals, for example, looked like an onion, not all at what we think of as a crystal. The actual crystal, as it came out of the crystal grower, was almost spherical. When you think of a (modern) silicon crystal, whatever the diameter, you think of a (cylinder) salami. And what we used for the junctions was like a small tennis ball. And really, the device was made in the materials dept. By the time you had grown the crystal, you have really made the device, and all the slicing into the bars was simply mechanical. You had no control of the device characteristics at that point – that was all determined.

The (Transitron) revenues at that time were around $100 million. They made a range of devices, power transistors as well. John Royan (John and I started another company) was involved in the power transistors. He started in 1959. Another source (for you) would be Pierre LeMond. He was the top engineer. I doubt whether they used the term “VP” at that time, but we was the chief technologist and all of engineering reported to him. I doubt if Pierre was more than 27 at the time.

3) What types of issues did you address there as a manufacturing engineer?

Usually, it would be that the yield went to zero - “What happened?” No one really understood how these things worked. It was very much an “end of the line” yield. On some transistor lines in those days, a respectable yield would be 20%. I remember one transistor line at Motorola, where I was simply supplying the germanium material, but they used to have me sit down at the meeting. The yield was down around 3%, and the guy who was running the transistor line pulled all the engineering guys in and said, “Hey, we are never going to get there from here by just modifying what we’ve got! So, what I’d like you to do this weekend is divide up the process and each of you have 10% of the process and write down what you think should be the process and Monday morning we’ll put these together and that will be the process we are running next week.” And at that time, it took about two weeks to get through all the steps. Magically, the yield came out at 30%, so that was the new process.

In those days, the customers didn’t really know what was happening in the production line. We did a lot of reliability testing, so you could have some assurance that the stuff was ok – it really was fairly loose. That’s one of the things I think the military contributed, was insisting on specifications, written specifications. A little bit later than this, by the time I was at Motorola in the 1960s, a lot of these devices were going into the Minuteman missile and other missiles, and so the military was all over it: “Written specifications, any process change had to be certified and approved by Quality Control). It put a lot of methodology into it. Years and years later, when I was dealing with a Japanese partner (Toshiba in the 1980s), and we went into total immersion in one of the Japanese production lines, we realized that they did not have any written specifications, as had become the norm in the American industry. This was because it was so stable there (in Japan). You taught somebody the process, and the people were there a long time. I think that is one thing the military involvement did, was to put a lot of formality into the manufacturing process.

4) As early as the 1950s, silicon transistors were used by the military. In fact, the Explorer 1 satellite, launched in 1958, used the 2N33X silicon bar transistors of the type you worked with at Transitron. One can only imagine the level of “End Test for Reliability” that must have been used for these early devices.

You must have heard of the famous solder–ball problem. The way in which you sealed the devices, on certain packages for silicon devices, was in nitrogen. It would be in a “dry box” that was fully nitrogen inside the box. There would be a pin hole in the top of the (transistor) can and around the pin hole would be a solder “donut”, and the operator, using rubber gloves for inside the 100% nitrogen dry box, the operator would wipe a soldering iron across the top of the cap, and that would fill the pin hole. So now, the unit would be sealed – this was (the process) done in the early 1960s and late 1950s. But what they found was that quite often, inside the can, there would be a solder “dew drop” down into the can, and when there was any vibration at all, this would break off, and you’d have these solder balls rattling around inside the device. When a rocket takes off, there is quite a bit of vibration. So, you’d have these solder balls, and if these hit some sensitive part of the device or bridged a junction, all Hell happened. Testing for this was a problem, because it took a long time to understand what was actually happening.

Another example (of early Transistor Reliability testing) was in the early stages of plastic transistors, around 1963/64 at (Motorola). We tested the plastic transistors virtually to the same military specification that we tested the sealed metal can transistors, and they came through wonderfully. They just did great, and we were all very pleased about that. And then, they went into production and we tested them the same way, which we thought was very rigorous, and we life tested them, and then I remember getting a phone call from Bob Galvin at Motorola, who was the CEO at the time. And he said, “I want to know why our television sets south of the Mason-Dixon line are starting to die in August. We think it is your transistors.” We then went and evaluated them and that’s when we found that about 85C and 85% humidity was death to the then plastic transistors. What happened was that we moved on to a new technology (plastic) but the testing methods don’t apply to the next generation of technology. It was “touch-and-go” there for while. We had people at Motorola who wanted to withdraw the (use of plastic transistors) from the field, because of this temperature/humidity problem. But we said that we would fix the problem fast – we’ll quarantine the lots and fix the problem in a couple of weeks, which we did by using a different formulation of plastic.

5) When you joined Motorola in 1960, mesa transistors were the best available?

Yes, and planar. What got Fairchild going in 1958 was basically a mesa transistor – the PNP was the 2N1132 and the NPN was the 2N697. Around that time, Fairchild implemented the planar transistor, while everyone else was working on the mesa. We were late at Motorola with the silicon devices. So we were right in the process of bringing up the mesa product, and suddenly we find that the others guys have got planar. So we made the decision to abandon mesa and go straight to planar. Of course, all the germanium (transistors) was mesa, because you couldn’t apply the planar process to germanium. So we went immediately to planar. Fairchild was still making the mesa for PNP - we fixed that too, because our R&D department, under Arnie Lesk and Jack Haenichen, came up with the annular ring, which was a real break through at the time. We were probably the only people who could make planar PNPs for awhile.

6) Was it the development of the annular ring technology at Motorola that allowed the manufacture of PNP planar transistors? How was epitaxy used?

Well, what you were getting was surface inversion on the PNPs (a common failure mode at that time). And, basically, the annular ring solved this problem. This is what enabled us to come out with the planar version of the PNP, which rapidly replaced the 2N2800, which replaced the 2N1132. We also decided that epitaxy was the way to go, if you could do it (10 micron thick layer, with controlled resistivity). Once you could do that, it meant that the series resistance from the junction to the back of the chip was dramatically less, so that the saturation voltage, VcSAT, was dramatically lower and this positively changed the characteristics (of the transistor). The other benefit was that you could do it with a much smaller chip, and in the chip business, these things are important - the smaller the transistor and the higher the speed of the transistor, the more valuable the transistor is, and that’s true even today. This was the real key for epitaxy. I remember that when I got to Fairchild (in 1968), the difference in chip size was horrendous. It was around 50 mils sq (for Fairchild chips) compared to the equivalent Motorola device of 25 mils on a side. This was like 4 to 1. So when you have a chip that’s four times cheaper, and the electrical characteristics are quite a bit better, its almost no contest.

7) There were other companies at the time that were manufacturing planar epitaxial silicon transistors. Was it the combination of this technology with the annular ring process that allowed Motorola to become the leader?

Yes, and the other (advantage) was being the first mover. For example, Motorola was able to establish the Motorola “2N” numbers and the specifications were written around our devices. The other companies either came up with a different “2N” number several years later or they attempted to make our “2N” and meet every single specification that we had, and these were very tightly specified devices. So it was really very difficult to hit every single parameter, and it they couldn’t, then it wasn’t a 2N2218/19 or 2N2904, unless you hit the (specs) exactly, you didn’t have the device. This first mover is like with Intel, with the 8080 architecture in microprocessors. Once they had that out, and all the software is written around that architecture, somebody else has to attempt to replicate exactly, and then they are in violation of Intel’s patents and copyrights. So, (at Motorola), it was combination of things. First, you must have the process technology, then you have to make the devices. For example, one of the things we specified was beta linearity, or the gain of the transistor over a whole range of current levels, and everybody else had trouble doing that, particularly at the low current beta – we had some “tweaks” in the process that enabled us to get at 100 microamps virtually the same gain as at 100 milliamps. Whoever tried to compete with us had to meet this same specification.

8) I am interested in your comments regarding the 2N2222 transistor. As you know, this device is a silicon epitaxial annular ring device introduced by Motorola in 1962, and it has since become the most widely used transistor in the industry, with billions of units sold.

Once I moved on to the integrated circuit business, I really didn’t keep track (of the transistor industry). (Originally, the 2N2222) was the high beta version of the STAR chip, in the TO18 package. I was surprised to see (in the data you sent me) that the same number has been applied to the TO92 (plastic) version. We deliberately at the time came up with totally different “2N” numbers for the plastic, because we didn’t want our existing customers for 2N2222s, who were paying us $4 a piece, saying that they could take the 25 cent plastic device and plug in into the socket and it works exactly the same. The (original 2N2222) was a general purpose amplifier, with a 40V specification. And it was interesting that people learned to use it in a lot of different ways, not just as an amplifier.

9) What was the transistor organization structure in place during the development of the 2N2222 product line.

The R&D portion of the development was under Arnie Lesk and Jack Haenichen (who was about my same age). And Harry Knowles, he was brilliant guy. I had interfaced with Harry, when he reported up to Hogan and was in charge of all R&D. Lesk worked for him. My interface with Knowles would have been only sitting in a few meetings, as a peripheral and fascinated person. Originally, I was given the job of Director of Pilot lines, so that was my initial job, which didn’t last very long, because not long after that I got the whole transistor group. Initially, my job was the PNP transistor product line, to take it to a real volume product. I got involved in the 2N2219 (the brother of the 2N2222) about six months later.

10) How would you describe the Motorola silicon transistor program when you left in 1968?

Certainly, we felt that we were number one in the business at that time. Then, almost immediately, I had a lot of insight into Fairchild. It was interesting to see that they had gone in a totally different track. Our whole thrust at Motorola had been that we were going to kill the whole industry with mechanization – automate everything, which we felt we had. When I got to Fairchild, they had a very low level of automation, and they had been very aggressive in what we called in those days, “Jet Age Automation.” They had figured out that with jets, you were only 12 hours away from Hong Kong. At that time we were paying about $3/hour for labor in Phoenix, and if you could get the same job done at 10 cents/hour in Hong Kong, there is a lot of cost to be saved. So they had pretty labor intensive ways of putting things together, but, actually very low cost. I was astounded when I got to Hong Kong to see how low the costs were. Where we (Motorola) were beating them was not on cost, but on superior device characteristics and better logistics and better customer interaction. Actually, we were pricing all of our stuff higher. Their costs were actually pretty good, but they had done it a different way. They were using what we called “Glop Tops” at the time, which was a ceramic base with metal pins pushed through it. Basically, you bonded the chip to one of the pins, and wired to the other two. But again, you’ve got a pretty low cost transistor. Where cost was the only issue, they (Fairchild) were pretty competitive.

11) You’ve indicated that by the early 1970s, the primary focus for semiconductor companies was integrated circuits, and not discrete transistors. Please elaborate.

Well really, for Fairchild, this in the 1960s. Motorola had made a bet on integrated circuits to go with Emitter Coupled Logic, ECL, and that’s where they put all of their efforts when, in reality, DTL, which was Fairchild, and TTL, which was a combination of Fairchild and TI, rapidly became the dominant approach. So they were somewhat behind technically in integrated circuits, and Fairchild had done much more in linear integrated circuits. The whole thrust of Bob Noyce and Gordon Moore was to move on to the next generation of technology, and just leave the existing technology behind. So, in transistors, Motorola paid much more attention to improving what they had in transistors, and ran the transistor business successfully – but to some extent, this was” robbing Peter to pay Paul”. For example, in about 1965, our transistor business was doing very well and the question was when we built a new wafer fab, what should we do. I remember at the time that the decision was to build a new silicon transistor wafer fab, and then hand down your fab to the integrated circuit guys, because they couldn’t afford a new one. Now, McKinsey or Bain or one of the consultants would have said, “Hey guys, this doesn’t make sense.” You put your money on the next generation. From my standpoint, it was wonderful. I’d love to build a new wafer fab. By the time I got to Fairchild, I found, essentially, that the transistor group was running on old tools, and they were relying on the low cost foreign labor. On the other hand, we found that Fairchild had big computers, very expensive $5 million computers, applied to design automation and Motorola wasn’t anywhere near doing that.

12) Since you have had experience with both Motorola and Fairchild, would you comment on what has happened to these companies?

Motorola was very successful in the discrete area, whether it was diodes, zener diodes, germanium transistors, silicon transistors, power transistors, and so on. And that became the focus of what they did at Motorola in Phoenix. They did not put adequate weight behind the integrated circuit push, which put them a critical couple of years behind Fairchild and perhaps TI. That was, I think, the big difference, but they certainly did extremely well with discrete devices over the years. Now, you’ve get the discrete device portion of Motorola is ON Semiconductor, which is a couple of billion dollar a year company, and Freescale is the other piece of it, which they spun out last year which was just bought for $15 billion.

Fairchild, in 1979, we had an attempted hostile takeover by a company called Gould. And so eventually we sold the company to Schlumberger, the oil field services company, and at a very good price - $66 per share, which in those days was a pretty good market cap. Later, it was acquired from Schlumberger by National and then even later, National spun out a big portion of National, and they spun it out with the name Fairchild on it - it wasn’t complete, the whole Fairchild product line. And that has done pretty well. They are now back to a couple of billion a year.

One other comment about what was a big difference for Motorola – there was total dedication for making all of your own equipment. The reason that Motorola focused on the automation was that we were very confident that we could automate anything. There were 16,000 people in one building that did everything. And, in the basement there were 1000 mechanical people, engineers, machinists and so on, building the equipment. So out of 16,000 people, 1000 were building equipment. To some extent, the separation from Chicago was probably good. Hogan was a very independent guy. He had been a professor of Physics at Harvard, and this was his first real job, running the Semiconductor Division. Motorola Chicago had a wavering commitment to the semiconductor business, and the semiconductor business used to go through a four year cycle that was fairly predictable, and every time we were in a down cycle, Motorola Chicago would be ready to sell it, but just at the point where they had done the deal with somebody, the business would take off again, and they said, “Why should be sell the business, it is doing so well?”

13) Thank you, Wilf, for agreeing to this interview. I’ll follow up with an update to your Oral History at http://www.transistormuseum.com

and will be in touch when this is completed. You’ve been very gracious with your time.

#####################################################################################