An Interview with Roger Anderson
RA = Roger Anderson
GAM = George Michael
GAM: Today is March 8, 2001. We are talking with Roger Anderson.
The usual thing, Roger, is to start by telling us how and when you got to the Lab. We'll take it from there.
RA: OK. I came to Livermore in the fall of 1959. I was hired into the Explosives Research Group of the Chemistry Department,
primarily to develop a k-band interferometer to determine the kinetic parameters of exploding chemicals.
I had a very interesting time there, learned a lot more about microwaves, and developed a few unusual pieces of microwave hardware.
The interferometric techniques worked fine, but the project hit a standstill in obtaining appropriate impact materials.
This is at the explosive and inert medium interface. Air will give you one point on the Chapman-Jugiet equation, but
you need other densities of materials for the shock wave to impact, to get further points on the curve, and to
determine the thermodynamic parameters of the explosive. That is, of course, the feedback into the codes that are
trying to model the tests. Experimental measurements are great, but they don't extrapolate very well unless you
really know the true equation of state. The problem was that I needed materials that were transparent at
k-band microwave frequencies, of known and relatively uniform densities and, in fact, I needed the spectrum of them.
It's very possible, at that frequency, to use foams, but no company was willing to attempt making them at controlled
densities. I started doing more work in kinetics, and using the microwave equipment for microwave spectroscopy.
Sometime in the early 1960s, maybe 1963, I went to a Wescon trade show. I liked to go to engineering trade shows
to see what instrumentation opportunities were coming along. By this time transistors had really come into their
own, although I had my first experience with transistors while at the Aberdeen Proving Ground in 1955. At that conference,
one of the exhibits was by a relatively unknown company called Digital Equipment Corporation. On duty at that booth
in San Francisco was Ken Larsen. He showed me a small computer called the PDP-5, actually drop-shipped there, on
its way to delivery in Berkeley. It was one of the first small computers to arrive on the West Coast. At that point,
it was a general purpose machine that someone put together that should be really useful, so I started pushing that
concept for instrumentation use at the Lab. Also around that time, the PDP-1 had been delivered to the Lab and I
was introduced to it. I guess it was primarily through Ray DeSaussure that I really got into it. I was trying
various and sundry things, but also saying that such computers would be useful for the Laboratory in general.
In my case, the interest was primarily for instrumentation. Around that time, I transferred from the Explosives
Research Group to the Physical Chemistry Group in the Chemistry Department. I was pushing what could be done with
additional computers to Jack Frazier, who was then Department Head. I was really excited by what could be done in
the analytical area, because the computers were good for taking and processing data. I don't remember the exact
sequence of what went on, but adapting to computer use became a primary effort while I was still in explosives
chemistry. I spent a lot of time investigating the use of computers for the chemistry department in early 1963.
We were looking into all the various and sundry machines that were being developed and what was being done in real-time
data acquisition and processing. I think the proper description is in spite of the Computation Department as opposed
to with its cooperation. I can still remember talking to Sid Fernbach and Bill Masson about various and sundry things
that were going on and I remember Sid said, "Well, what did you look at?" and "What is this?" and things like that,
and he said, "Well did you look at the IBM 1800?" I said, "No, I haven't heard of the 1800, when was it announced"?
He said, "Last week". That was when they were making the initial investigations on the Photostore and, of course, that
was the controller for the Photostore and he had pre-announcement information.
GAM: Well, the 1800 inside the Photostore was highly modified. It wasn't the kind delivered to most people.
RA: Right. But the primary thing is it was an unannounced product at the time I was looking at computers. We looked at the
unmodified 1800 and decided it was too awkward to interface to a variety of lab instruments. Anyway, we continued
looking at various small machines. Oh, I should also point out that while looking at various machines, the PDP-4 came out,
It was an 18-bit machine like the PDP-1.
DEC was starting to push the fact that the PDP-7 was going to come out using their new circuit cards called Flip Chips.
I can't remember the company now, another Massachusetts startup, but we also looked at their machine called the, DDP-24,
which was a good machine. It looked between these two that the PDP-7 was easier to interface and was the most reasonable
approach for us to follow.
Then the politics really started being visible. I can remember one day when I was officed in one of the trailers outside
of 107C that I got a visit from Ed Lafranchi and Glenn Strahl. They told me that authorization for all computer procurement
had been given to the Computation Department; to Sid actually, and they were acting as the Engineering support for him.
We talked about it a little bit and they said, "Well, what do you think if we said this procurement was not a good
idea for the Chemistry Department"? I just looked at them and I said, "Well, all I can say is I know several persons
who are going to be awfully mad." They agreed and indicated that the question would have to be settled by upper
management. All of a sudden the problem got escalated to the director's office. Eventually, an order got placed
for a PDP-7. I did some early programming on it, but I was still more interested in instrumentation than data processing per se.
I can't remember exactly, but sometime shortly after the PDP-7 and the flip chips came out, DEC introduced the LogicLab,
which was a bread-board type of hardware design tool. Also someplace in there, the PDP-8 was introduced.
It was Sol Golsman, who was with DEC, who was trying to figure out all the things that the Logic Lab could be used for.
Well, he liked the flip chips and he really liked the Logic Lab because it allowed him to try things out.
He was also trying to figure out what to do with some of the memories that didn't meet the PDP-8 spec.
So, out of individual parts on the Logic Lab, he built a computer that ran the PDP-8S instruction set,
but it was serial, so the out-of-spec memory was plenty fast enough. He had a little bit of a vision when
he laid out the computer, and he put the core memory right in the middle of the Logic motherboard.
He left a hole because he was going to build another machine that used larger memory, to service a 32K-word disk.
The motor would fit where the core memory was in the 8S. That was the 8D and a prototype ran.
But, for people who thought the 8S was slow, the 8D had all of the rotational latency problems of the
IBM 650. The time it took to get a word out of memory depended on where in the address space it was.
With all the subsequent problems, the 8D was built in prototype form, but was never sold.
Well, I should say it was never manufactured. But the 8S was manufactured and, much to my surprise,
I was able to get one purchased by the Chemistry Department. In my view that was the world's
first personal computer. It was small; it was hardly bigger than your tabletop Mac there and it didn't
have enough computing power to think about sharing it. Timesharing was not an alternative.
But it was inexpensive since that machine was only 10K. Thus, you could think about it in terms of
interfacing directly to experiments. I built up the machine in a portable rack and constructed some I/O interfaces.
Eventually, they added one of those 32K disks to it. It had a 4K memory and processor,
and a full 32K to hold the operating system, the applications, and the data. It worked and it was
on wheels so we could plug it into various experiments and try various things. By that time, even
the rest of the world was beginning to think computers were useful. So, that was the start of it.
We had bulk storage in that machine�it was called paper tape. We could punch and read paper tape.
I had one full box of paper tape which contained all of the then currently-known digitized spectroscopic data.
Recently, in fact, it ended up in the Computer Museum. However, at the time, people were starting to look at other things.
In the Spring of 1968, I went on a short sabbatical to the University of Missouri. I worked in their Computing Lab
on Tom Brubaker's hybrid computer. I also studied a lot about signal processing and data processing.
I taught a FORTRAN class; my first real introduction into higher-level programming.
Before that, everything was assembly language. We had an IBM 360-44 in the Engineering Department,
so I really got my first good flavor of what IBM and IBM systems were like. It was rather tedious,
but it was also very interesting. I got an understanding of hexadecimal arithmetic and how hex works,
and found out what was really going on inside the machine as opposed to what you could see from a FORTRAN listing.
It was very interesting, I learned all about WATFOR and WATFIV; they were good educational systems.
In the Summer of 1968, I returned to the Lab. I took over and implemented the prototype experiment for the CAIN Operation.
That had been proposed in Engineering as a service for the Security Department to make a different electronic badge
access method for the Laboratory. CAIN is an acronym for "Control Access by Individual Number."
The idea was to make badges that had coded into them a number and make a reader that would read them
and then computer control access into various places within the Laboratory. For the demonstration system,
we were going to use the PDP-8S computer and disk located in the Plutonium Building that was not used too much.
We were going to control the access gates into the Plutonium Building restricted area. That was the concept.
They already had security booths of some kind. I got volunteered to put that particular system together.
It involved everything from designing the logic for the access to writing the software that would run on the 8S.
I even wrote the code for the digital drafting machine that generated the code strips for the badges themselves
and put them together. Anyway, it was put together and we did control access to that system on a prototype basis
to prove it would run. Then, not too long after that, I transferred to Computation.
The CAIN Program continued, but it was not making very great headway. I don't remember exactly what it was but,
someplace in there, Ed Lafranchi asked me to come in and take a look at what was going on and evaluate it.
He had all of the "flow diagrams" or "logic diagrams" which, for me, might as well have been written in Chinese.
I thought it was a waste of my time. There were some discussions as to who was going to manage what.
The long and the short of it was that management, what there was of it, was ineffective.
Finally, they just decided the support for CAIN was going to be out of Engineering and by a totally different group.
We started to do it. It was my opinion that it would be faster to start from scratch and redo the whole thing
than it was to try and figure out what was usable. So, by that time, the hardware for the system shifted to the PDP-11.
We were designing interfaces and we were designing a new type of booth, with different access controls and with
the idea of using it lab wide. I whould say the bulk of the work for developing that system was actually done
on the PDP-10 in Computations using Fletcher's PDP-11 cross-assembler and then transferring all the code via magnetic tape.
GAM: Are you sure it was Fletcher? I thought it was Gail Marshall.
RA: It might have been, but I think Fletcher had something to do with it. I don't remember exactly.
At some time, Dave Nielson joined the project. He and another Engineer, whose project had disappeared
after he had been hired but before he appeared. All he knew was FORTRAN. It was a little difficult for that
particular system, essentially since I wrote the entire system in assembly language. It included badge entry
as well as access control. We got a disk that was word addressable so, in essence, I had a virtual memory
for the access files out on the disk. We put it together to control them and slowly built up the whole system.
RA: The control program itself was about 16K words and the access-entry thing was about the same size. So, we put
it together, and after testing, we saw that it really worked.
GAM: Another triumph of science over witchcraft.
RA: The interesting thing to me was when that system was redone. By that time, the VAX had come out and they had to
upgrade to the VAX. A former student from DAS, who'd come along and had gone to work at the Lab, decided that,
for the best document control and so on, he should write it in ADA. And so the rewrite of the CAIN system, when
it was moved over to the VAX, became an ADA project. Personally, that was the only place I know where ADA was
directly used at the Lab. Someplace in there, the DAS program came along, and I started teaching the computer
architecture courses. Of course, the bureaucracy got bigger as the program matured.
GAM: Do you remember what year you finished the CAIN system?
RA: 1972 or 1973, something like that, because I'd been transferred into EE (Electrical Engineering) by that time.
Someplace in the middle of that, I went down and did support for Bio-Med for a couple of years. There, we used
ModComp computers. They were starting to do communication work on those systems. We had a two-computer network
that we were using to do real-time instrumentation (data acquisition, actually) from some of the cell-flow analysis systems.
GAM: Was Tom Slezak there then?
RA: Yes, Tom was a DAS student at first, and then he was hired into Biomed. I was down there a couple of years
trying run their computer system, also trying to get into a little bit of pattern recognition and maintaining
the system and keeping it running.
GAM: Well, I think they had some very nice things down there.
RA: Yes, they got a new staff member, who'd come in from MIT, and he wanted to start over with UNIX machines.
There was trouble enough with the ModComp systems. It was about then that I went over and spent a little time supporting MFE.
A good share of the reason for that was due to my collaboration with the DAS program. Part of the time that I was at DAS,
I had helped Charlie Wetherell on the Ironman Evaluation Project. That was the DoD contract compiler, the predecessor of ADA.
It was a very interesting experiment. We had the red, blue, green, and yellow languages, from unknown authors.
We had six weeks to learn the four languages, compare them, and evaluate them. Great experiment.
We had Jim Greenwood to help us too, with all the online language work that was going on down in the Sherwood Project.
The DAS system was progressing, and I was much more interested in the small, individual computers.
I'd gotten my introduction to UCSD PASCAL and we brought that in on the system. I was also interested
in PASCAL from a personal point of view, using it on my own. That was my greatest introduction into
high-level languages. With PASCAL, if I could get it through the compiler, it ran. It was a good language.
I don't use it that much anymore, but I do still have it.
That's when I did some work for MFE. Actually, I was working for Bob Wyman at that particular time.
I was evaluating the use of PASCAL for the MFE computer control system. Then I switched and supported the Test Division in EE for a while.
I worked with Bill Lennon, whom I'd known for years in some of the DECUS work. I influenced Bill to come to the Lab.
He later talked me into taking over a project on the communication link between the test site and the Laboratory.
He said, "We have this program here and we've been working on it, and it almost works. It's still got a bug or two to clear up."
And this program was actually taking some specific data from the test shot, bringing it up to the surface where an
LSI-stored it in core memory, and then powered down the computers. After waiting for the ground shock to go by,
the computers were powered up again and the data transferred to floppy disks for transfer to the CP.
The amazing thing is the computers survived the quite large shock wave. We physically carried the floppy disks
back to the CP and then read them back into another LSI-11. This, then, got interlinked into the Nevada System,
was transmitted to Las Vegas, and put on a microwave link to Los Angeles, then put on the phone lines to San Francisco,
and then onto phone lines to the Lab. Needless to say, it got encrypted on each end during these transmissions.
We got the data from the test fairly quickly. I spent time at the CP getting the data onto the floppy disk,
but the real problem came in transmitting the data from the floppy disc through all that linkage to the Lab.
I think I figured out once that it had to go through seven or eight computers.
In the process, it involved Laboratory people, NSA people, EG&G people, Southern Bell,
and Pacific Bell people as it made its way to the Lab. And, if you want an interesting debugging problem,
here you've got a real place for finger pointing. I was trying to communicate in that system and get it to run.
There were actually only four bugs in the program, but it took me over a year to find them.
It would have been faster to start from scratch than to go through it. One bug was a system error.
Another was an error in translating some of the PDP-11 code that had been run on the PDP-10 cross-assembler
over into the PDP-11 code that was assembled on the PDP-11. One letter was transposed.
Some of the code was in FORTRAN, and it had to do with how it did sign conversion going from integer to fixed point.
One of those real obscure errors that the programmers were supposed to be aware of it.
We did eventually get that whole thing put together and it ran.
I used to go up to Las Vegas once a week. Then we found out there were actual hardware problems.
The actual hardware wasn't what they said it was, so I ended up with an electronic tech that flew out with me each time.
We carried part of the interface hardware for each run or test, to debug things.
It was after I left that they did a real shot and we had pictures displayed in the CP twenty minutes after the experiment.
So it worked.
Somewhere in there, this crazy guy called Michael was trying to convince Computation that a computer research group should exist.
GAM: That moves us up to 1980 or a little earlier.
RA: Yes, I think it was 1979. That's when I joined the CRG (Computing Research Group) and started trying to do some multi-processor work.
During that time, I worked with Brian Lavler at DAS. He worked down in the group that did the weather monitoring for fallout.
RA: Yes. He worked in ARAC. As part of his doctoral thesis in Computer Research
he put together several LSI-11s to do multi-processing on that particular problem.
That is where I really got into multi-processing. He was proposing various things and trying to build a system.
I was trying to propose a more modest system using LSI-11s. Gordon Bell happened to be visiting on one of his typical tours,
and I was telling him about it and he said, "Oh no, you can't do it that way. But we've got an answer to your problem."
He started telling about the 784, which was an unannounced product. It had a shared memory communication device.
The thing that I was trying to do was actually communicating via a serial communication port between a central
machine and a ring of machines; trying to do some data-flow type processing. But Gordon convinced us we had to
think on a grand scale, as we had done earlier using the VAX. We had gotten the VAX shortly after I'd come to the CRG.
Steve Skedzielewski had joined the group by then. We implemented that system and got it onto the ARPAnet,
and then started on the multi-processor work. Somehow, which still amazes me, we were able to purchase a 784.
The 784 was approximately four 780s with individual memory and multi-ported shared memory.
GAM: We negotiated with DEC and got it as part of the DoE/OBES contribution to the research they were sponsoring here.
DEC brought all that stuff in, we didn't buy it all. We had a couple of joint research projects that we were going to be doing on the thing.
RA: Well, I don't remember the financing of the thing. I remember the configuration, trying to get the thing together.
GAM: Our Comp Department handled the financing, We expected to develop our Data Flow research on it.
RA: Oh, yes, I do remember vaguely, because Berkeley got one also. They had a 784 and there were two others that went to
undisclosed recipients. That was the extent of that type of activity. And that was the last of the shared memory
multi-processors that I think DEC ever built.
GAM: Well, under certain conditions, that would have been a very good set of research projects.
But the project was a little flaky at that point, and that tended to distract everyone a little too much.
RA: And the support for the effort in that area was fading. Personally, I believe the Lab was a bit too impatient for results.
GAM: Yes, it was flaky and they were not interested in the needed research. When did you start at DAS?
RA: It was early to mid 1970s. Fletcher had been doing some teaching there.
It must have been about the time Wetherell came back. (ed: He had gone to graduate school at Cornell) Charlie was there,
McGraw hadn't been hired yet into DAS.
GAM: Charlie and I hired McGraw. And then he, in turn, helped me hire Steve Skedzielewski. These were all very good computists.
RA: And we went through several rounds out there, producing some very good students.
GAM: Yes, I surely do agree. Well, thanks Roger, your career certainly helped start the use of computers as general experiment controllers.
I also want to thank you for taking the time now to recall some of your early adventures at the Lab.
For information about this page, contact us at: email@example.com