An Interview with Ken Larsen

Ken Larsen


KL = Ken Larsen
GAM = George Michael



GAM: Today is the 16th of March, in the year 2002, and I'm going to be talking to Ken Larsen, former DEC representative who had a lot to do with the Laboratory's early days in computing. Ken, why don't you start by telling us where you came from, how did you get out here, and so forth.

KL: Well, how I got out here is a long story. Around the end of the Korean War, I was drafted into the Marine Corps. After Boot Camp, I was assigned to the "Airman School" at the Naval Air Station in Jacksonville, Florida. Upon completion, I was assigned to the Aviation Electronics Technician school at the Naval Air Station at Memphis, Tennessee. The training was specific to aircraft radio and radar equipment.

My duty station was with a Marine helicopter squadron, HMX-1 at the Marine Air Station, in Quantico, Virginia. This squadron is the experimental and tactical development arm of the Marine Corps "Air Wing". Although my major task was to maintain radio and other electronics equipment, I was also assigned to help out with the installation and operation of some infrared equipment being developed for the early night vision experiments. (HMX-1 currently operates H-1, the helicopter used by the President of the United States.)

I was offered a "Field Engineering" position by IBM and was scheduled to go to SAGE (ANFSQ-7) Computer School. Because I arrived a couple of weeks before the classes started, I was placed in charge of maintaining the classified manuals library. Since there was little activity there, I began reading the books and studying the logic diagrams. The manual for the SAGE main processor manual had foldout logic diagrams. When I opened the foldout called "accumulator", it looked like a radar range counter with whole bunch of other gates and logic elements attached. So, that gave me a head start since I could relate to equipment that I had already been involved with, so that really made it easy for me to get acquainted with that computer quickly. I was originally hired to go to Battlecreek, Michigan to the Sage site there, but one very cold morning in Kingston, New York, they called us all together and said, "We're starting a new engineering group in Santa Monica, California. There's a company out there that's breaking away from the Rand Corporation and the new company name will be Systems Development Corporation and we are forming an engineering support arm to them and would anyone like to go to Santa Monica? I think I had my hand up first, so that's how I arrived in Santa Monica on July 4th of 1957.

Our group installed an ANFSQ-7 in the building behind 2500 Colorado Blvd. in Santa Monica, then the Systems Development Corporation headquarters building. There are a number of interesting things about that installation. For example, Southern California Edison didn't believe a building that size could consume the amount of power specified in the service order. They must have unilaterally decided to provide power based on some formula relating to the square feet of the building. During the power test phase, we were powering dummy loads before connecting to the computer frames. When we started the third motor generator set, all the lights went out and we apparently blew up something significant back at some transformer sub-station. We called them up and they came and had a meeting with us. I remember during the meeting we discussed that they hadn't supplied us with adequate power. We had not been involved with specifications or service orders, we were just trying to bring up the power. Another engineer and I were assigned to bring up the power before the processors arrived using dummy loads for check out of all power supplies and distribution panels.

When the Southern California Edison people came in, they asked, "well how much power do you guys really need?" With air conditioning and the whole machine, displays and support equipment; from the books we calculated the power requirement and came up with a number close to the original service order. Their representative said, "That's not possible, you can't use that much power in this one building." The engineer who was with him says, "You mean to tell me you're going to use one-sixth of the power in Santa Monica?" I don't recall the amount of power required, but I will never forget his astonished comment. That computer, in this one building, would consume one-sixth of the power going into Santa Monica in 1957. In about three weeks, power was restored, we did get both processors up and running with up-time better than 95%. At the time, the reliabily studies determined that the SAGE computer could never operate due to the failure rate of the vacuum tubes and related components.

Then I went to work for a company called Telemeter Magnetics that made memory cores and memory stacks and even built memories. They manufactured memories for the General Electric ERMA computer a decimal machine developed for banking applications. It was quite interesting to see how a decimal machine operates with comparison to the straight binary machine. While at TMI, it was my responsibility to specify and purchase memory test equipment which we purchased from a company called Digital Equipment Corporation (DEC). As a result, I became very well acquainted with several of their people and one day they came by and offered me a position in field sales, working in the Los Angeles office. After several discussions, over a number of weeks, they made an offer including dual assignments, as applications engineer for the memory test equipment and to begin to be a liaison with the Jet Propulsion Laboratory (JPL). My task was to pursuade them to design in DEC modules in their data acquisition and recording facility. They were also considering the purchase of a PDP-1 for a wind tunnel laboratory. It was my task to help them determine that the PDP-1 would be able to perform the on-line data acquisition and data reduction needed in the laboratory. (DEC salesman were accused of running around with a $120,000 solution looking for a problem.) I did become well acquainted with JPL personnel and their requirements.

During this time, I received a call from DEC's chief engineer, Ben Gurley who said, "I'm going to Livermore to install a PDP-1 can you come and help me?" So, I met him at the Lawrence Radiation Laboratory in Livermore and did my best to help him complete the installation. That's the first time I met you, George, and you took us out to Concannon Winery and introduced us to Captain Joseph Concannon and his sons. As I recall, after the computer was installed, you were doing some experiments with the precision CRT.

GAM: Yes.

KL: So, I think I was up here a couple of times making some adjustments on it based on instruction from Ben Gurley.

GAM: That was a type 34?

KL: It seems like that was the right number, yes. Ben Gurley was the only one that really knew anything about that equipment. So he would instruct me to do this or do that and I would come to the lab and do as instructed.

Some time later I received a call from, I believe it was Glen Strahl, stating, "The computer's been down for three or four days and we've talked to the people back at Maynard and we need a memory stack. They are unable to send one to us; can you help?" My response was only, "OK, I'll see what I can do." Well, I called back to Maynard and found out that Foroxcube, their supplier for memory stacks, was behind in delivery. They had four or five computers on the floor that they were trying to check out and they had only one stack and were moving it from computer to computer in a twenty-four hour operation. There were different groups that were testing out these other computers. They would be using the stack sometimes at night and early mornings and all kinds of schedules, trying to get the computers ready so when the stacks arrived they could ship them.

When I called Glen back, and told him what the deal was and he said something like, "Boy that's terrible, we've got this program going and we're really in tough shape." I responded with a statement to the effect that, "Well, Glen if you want me too, I'll come up and I'll try to fix that stack for you." And he asked, "Well, what kind of experience do you have on fixing stacks?" I explained that I had worked for Telemeter Magnetics and I knew a little bit about how they are put together and I won't promise anything, but if you want me to I'll take a chance on trying to get it going." So he said, "OK, what do you need?" I think I asked him for one of these magnifying lamps with the circular florescent lamp, a tiny soldering iron, some rosin core solder, an exacto knife, some tweezers and some very small bitty pliers, and things like that, and I said I'd catch the next airplane up and meet him at the gate. So he made some comment about, "Well, I'll have the guard ready" and, of course, in those days you came up to the gate and signed in and they assigned a guard to you and that guard stayed with you even when you went to the restroom.

So, anyway, when I arrived, they had a work table and everything ready for me. I came in and I pulled the stack out of the computer and laid it on the table with several people carefully watching me. The best I can remember, it was Glen Strahl, Ed Lafranchi, Bob Abbott, I can't remember all of them, but there were at least four or five people around there watching me and the pressure was really on. I just began to roll the stack over and kept looking at it until I spotted a broken digit wire. (Probably no one who is going to listen to this tape understands what a digit wire is but it's also called an inhibit wire.) With the exacto knife I scratched the varnish off both ends and built a solder bridge across, trying to make it as clean and secure as possible.

"Well, let�s give it a try." As I installed the stack, Bob Abbott was ready with his diagnostic tape ready to push the load button. I got the stack in and brought the power up and he pushed the load button and the diagnostic tape loaded and ran. Bob expressed his delight at this success. Glen Strahl looked at me and said, "I'm getting you a security badge". I had no idea what a security badge was, but he put through the paperwork and I got a badge. My rescue effort at that time, getting the stack repaired, was appreciated enough that he put me through a rush request. I think I got the badge within six or seven weeks. I know that there were people that had been hired by the Lab who were still sitting outside the fence in that other building for three and four months, sometimes longer, doing whatever work they could do out there. So, I was impressed with Glen's ability to put in that rush order. It is my recollection that Bob Abbott was writing a program for the Precision Scope, for some sort of measurement purpose. He was elated that the machine came up and...

GAM: Well, I don't blame him.

KL: That was my first point where I really got acquainted with Glen and others. I think they accepted the fact that, maybe, I was more than just a peddler that came by once in a while. It was a great start for getting acquainted with key people at the lab. I was secretly just tickled that I'd been able to get that stack working.

GAM: Well, the relationship with DEC at the Lab was always a very positive thing. We liked everybody that came from DEC. The guys were very good and they were interested in helping and they could help. They were really quite good. And, well, DEC swept all over the electrical engineering department�we had DEC equipment everywhere.

KL: Yes, the lab bought a lot of modules in the early days and many computers later on.

Well, we just talked a little bit about how I became involved with the physics community. I think it started with Victor Perez Mandez.

GAM: In Berkeley.

KL: Yes. He's the first person I talked to that spoke in the "Physics Language". He had an experiment running there that had a three-month shot on the Bevatron, and when he finished the experiment, the data was useless because various pieces of equipment had died during the project and they didn't know it. Since the data was useless, he didn't want that to happen again, and wanted to know if I could solve his problem. He said several other people had been there, talking to him, and concluded that it couldn't be done. He explained, in great detail, what his project was designed to do, in his elegant post-doctoral physics terminology. I had no idea what he was talking about, and I was just trying not to look more stupid than I was. When I left, promising to try and come up with some ideas to solve the problem, I went out to my car and I sat down and wrote down phonetically all of the terminology that was foreign to me. I then went to the Stanford technical library and started looking up terminology and began to get an understanding of what it was he was describing. The next time I called on him, I was more able to communicate a couple of ideas to him.

We installed a PDP-5 with a simple programmable tape formatter, designed to control a Datamec tape drive. To show a concept of using the computer to monitor the instruments and record on IBM compatible tape, I had set up a prototype test on a PDP-5 that was shipped to our office by mistake. I had some modules and worked over a weekend to make a tape controller, and Tom Tracy at Datamec loaned a tape drive for a demonstration to Victor and a couple of his staff. He brought a programmer, who reviewed my concept and, as a result of his positive comments, they gave me the order for a PDP-5. From my logic drawings, his technicians built the controller and, with the tape deck from Datamec, they got it going successfully. My suggestion was to have the data pass through the computer, record it onto the tape and then take every 10th data point as a sample to be displayed on a Tektronix scope so they could monitor each instrument. I think the programmer was just immensely sharp, he was able to accomplish far more than I estimated could be done, so they really knew the results of the experiment before they took the tape over to the 7090 in the computer building for final analysis. That was a great experience.

Now I could understand some physics terminology, so discussions with my contacts at Livermore made a lot more sense from then on. At Berkeley, we put in five duplicates of that original system. I never sold a PDP-5 at Livermore, but when DEC announced the PDP-8, I think it was Richards Thomas who bought the first PDP-8 for the Chemsitry Department.

Then Jack Fraser and Roger Anderson got a PDP-7. I can't remember now who got the second PDP-8, but after Richards Thomas, several others got PDP-8s and, as I recall, there were many PDP-8s ordered and listed in the Lab's stores catalog, so they could be checked out like a scope or other instrument.

GAM: Yes. We had one guy who's what I call a real doctor, Don Watson, whom I think got an LINC-8. And he was doing microelectrode work studying brains mapping where the neurons were, and he used that LINC-8 for everything. You know writing, reports and keeping track of the experiments, and he swears by the 8. In fact, he said he would write a little article for me.

Do you remember the "uptime" card reader?

KL: Oh yes. We just had more down time than up time. That was what? 2000 cards a minute or something?

GAM: Two thousand cards a minute, but if one didn't use fresh cards, there would be a card jam and you got an eight foot fountain of cards. We learned quickly about the need to sequence number the cards. But, we got rid of that and put in a 1402 card reader/punch. We had a 600 line a minute Analex printer on the PDP-1, but it didn't work well. We took it off and interfaced a 1403 N-1 printer, the IBM printer that was on the 1401.

KL: Yes, the chain printer.

GAM: Very nice. We had an enormous number of different things on there and they all worked and they all taught us something. I think the proper name for it, as far as we are concerned, is "Romper Room". I mean, you get an idea and you go in there and try it and there wasn't any screaming and yelling about "get off the machine because it's my turn", stuff like that, it was very nice.

KL: You know one of the things I remember about the Lab was you got a couple of displays made by DDI?

GAM: Data Display Incorporated, yes. They were superb.

KL: I was very intrigued with them and I was trying to get DEC to build something like that, but I couldn't convince key people to invest in the product. On the SAGE machine, they had that Charactron Tube.

GAM: Made by Stromberg-Carlson.

KL: Stromberg-Carlson, oh yes, it was in Hazeltine console, but it was a charactron and then the Memotron made by Hughes, an early storage type display. For character displays, they used a defocused beam and sent it through a matrix and steered it back to its position on the display. I was very intrigued with the beautiful characters that DDI had.

GAM: The DD-80 was electrostatic deflection, so it was much faster than the magnetic deflection on our precision display. In fact, the type-30 display also used magnetic deflection. With the dd-80, you could really move the beam around very fast, so the characters were being made between four and nine microseconds. Much faster than the Charactron Tubes. And we made heavy use of the DD-80. It was a good machine. We had one on the 6600 too, and we tried one on the LARC, and we also put a film-processing unit on, but that didn't work out. What I learned in all the fooling around that we tried was, especially with the help of the PDP-1, you must keep it simple. If it isn't simple, it isn't worth it. The more elaborate things just aren't any good for the users. They like it simple and they like it fast. And if it requires that they be involved with it slightly, to push a button, realign a piece of film, or something like that, they'll do it. But, if you try and make all that stuff automatic and you pay for it in time and complexity, nothing works. It really was a major lesson.

KL: I can't remember all the different people that I met out there, but I always had a good time. Really stimulating people to work with and talk to and I learned a lot from just meeting with them and trying to understand out what they were trying to do. I can't remember where I met Bob Wyman, but I had a lot of contact with him during the building of OCTOPUS and all the add-on memories to the PDP-6's you bought. I hope I wasn't too detrimental in his project there, but I know Gordon Bell was concerned about whether the system was going to work or not. His comments were that I was to help wherever I could and, if they needed something, he would smuggle it out to me one way or another. Anyway, I got pretty well acquainted with Bob in those days.

I was trying to remember when it was, but you had some sort of a supercomputer symposium and you had people from Westinghouse, and IBM and CDC and others. Gordon Bell was supposed to come out to that meeting. I was in the office in Palo Alto and was getting ready to pick up Gordon Bell at the air port to make the meeting that was supposed to start around three o'clock or so. The phone rang and it was Gordon Bell, it had snowed in Massachusetts and his car slid off the road and he didn't make it to the Boston airport. After explaining to him what I understood about this high level meeting, he said, "Well, here's what I want you to tell them." I sat for about half an hour writing down things that Gordon Bell wanted me to say. I came in and was on the panel with all the other people who were "real chief engineers" and "real experts" on this subject.

GAM: I don't think you should call them real. I mean those things were just targets of opportunity. That meeting was almost like a bidder's conference. It was a perception that Dan Slotnik's Soloman computer would be a good basis for a new SIMD kind of organization. He was supposed to represent Westinghouse and we had CDC and IBM. IBM decided they weren't going to bid on it, but CDC bid and they bid the STAR. It would have been better, in the long run, had we gone with Westinghouse with a more subtle design, but we ended up with the STAR. The STAR was a very interesting machine; it helped to formalize the whole idea of vector computing to the programming community. I am sure others have done the same thing, but that was our introduction. I remember that someone told me the STAR was ninety per cent of the first two chapters of Ken Iverson's book, "A Programming Language," and if you understood APL and the first two chapters of his book, you understood the STAR. The STAR was probably the largest machine we ever had delivered to us. Even bigger than the STRETCH, and I would say that, in general, it was a success from a computer science point of view, but from a physics point of view, it might not have been so much of a success as people had hoped.

KL: Well, when you mentioned STRETCH I was thinking of Norm Hardy. I'm not sure, did he come with the STRETCH?

GAM: No, he was here before that. And the STRETCH fascinated him, and he did go back to Poughkeepsie to work on it, and a thing called the HARVEST, which was like four STRETCHs. And then he came back here with the STRETCH. It was delivered in time for us to "exploit" it for the 1962 test series at Christmas Island and it worked fine. It paid for itself, as they say. We were able to take ideas, test them, analyze the results, restructure the idea, and test it again. It was very good.

KL: Oh, I remember a lot of contact with Sid Fernbach, really a great gentleman. He was kind enough to go to lunch with me, once in a while, over to the bowling alley. Did everyone who went to the bowling alley eat the rib eye steak sandwich?

GAM: I don't know.

KL: It seems to me that the rib eye steak sandwich was a very popular luncheon sandwich. Anyway, Gordon Bell was always impressed that I was able to talk to Fernbach. He was such a busy guy, you just couldn't make an appointment with him in his office. But, for the cost of a rib eye steak sandwich, you'd get a chance to talk to him for an hour. That was a bargain, I thought. But I thought he was just a super kind of guy.

GAM: Well, everybody spoke very highly of him you know. They say two things; first he wasn't much of a manager, but he knew how to handle people very well.

KL: Oh really? That's kind of an odd combination.

GAM: Yes, well, with the idea that an executive attended to things like getting the briefing charts correct and calling meetings, and so forth, Sid didn't do that. He didn't think it was important, and I think he was right. You came to our memorial seminar for him?

KL: Yes, you sent an invitation to me. I appreciated it very much. Yes, that was a good meeting.

KL: One guy we should never neglect is Ray deSaussure. He kept the greatest file of supplier's equipment that I've ever seen. If you didn't get your literature into his file you just never got considered for a potential sale. Ray was one of my stops with whatever the latest literature was available, to make sure he had a copy.

GAM: Well, you know, he and I worked together on those files, but they certainly were his files and they were huge. He kept them in good condition too. When they started moving us around, it was a big inconvenience to have to move all that stuff. So we boxed it up, there were about forty boxes of old manuals and specifications of machines, including some that never got built. Did you ever hear of the Phoenix? The MITRE Corporation designed it. That's in there, there's the RCA stuff that never got built. Anyway, all those boxes are somewhere inside of Tom Bucholtz' house in a locked room and, every now and then, I pulse him and say, "You've should get that stuff over to the museum�the Computer History Museum at Moffett Field." It's a very good collection of literature. The Lab has also donated many old manuals, in addition to some of the original equipment we had; LARC and STRETCH manuals and stuff like that. At the museum, they can become a resource for others to understand what happened.

KL: I donated a couple of boxes of old stuff that I had. They were notes that had almost been thrown away; it was Ken Olsen's and Dick Best's discussion of how to do the logic presentation for the prototype SAGE computer. It had all that stuff in there about how they decided to make the first DEC boards, and gate notations and all these things and how they decided to do that for the SAGE documentation. And they'd been thrown out and I salvaged them because I thought they were kind of interesting because I'd worked on SAGE, and I was very interested to know where all these ideas came from. Someone was copying this kind of stuff, so I gave them that paperwork, and I also had every issue of DECUS Magazine up to about 1972 and some other documents, and I always wondered whether they got into a place where they could ever be viewed again.

GAM: I think that, thanks to the existence of the Computer History Museum and Gordon and Gwen Bell, they may now be part of the document collection at the Museum. They have about all the DEC equipment and documentation they can handle. Yeah, they'll take more, because you can use it to give people as gift's and for trading and things like that.

KL: Yes, I had the central processor SAGE manual and I gave it to the museum. I also had a three-volume set on "Basics of Digital Computing" that was published by IBM and it had very detailed description about how core memories worked.

GAM: Apropos to core memory, there's a book called "Memories That Shaped an Industry" by Emerson Pugh, an IBM guy, and Ken is mentioned very heavily in there. If it hadn't been for him and his insistence on certain aspects of the design, the SAGE memory never would have worked right.

KL: Well, Jay Forrester and Ken worked together on that project so I think Jay Forrester was eventually regarded as the inventor of the core memory. He and another guy. The other guy claimed that he invented it, but there was some sort of a law suit and they decided that, from the records, Jay Forrester was the inventor.

GAM: This has been very pleasant Ken, and I want to thank you for taking the time to reminisce about your role during the early days of computation at the Lab.