[inforoots] Burroughs & Von Neumann, ElectroData, ERMA, drums, design automation

[Stan Sieler]

Hi,

While trying to figure out what a "B-Box" is ('cuz of Ed Thelen's post),
I stumbled over an interesting PDF of an oral history at the Smithsonian:

   http://invention.smithsonian.org/downloads/fa_cohc_tr_brad730201.pdf

...with information about ElectroData and Burroughs (two innovators in 
computers).

The file is a transcript of an oral interview of James Bradburn,
a founder of ElectroData and later a VP at Burroughs.

In it, we learn that the first ElectroData computer (later acquired by
Burroughs) was designed by in part by Dr. Ernest Selmer:

   And so I asked Lindvall one day, you know, who have you got around 
   Cal Tech who can really help us on digital computers, because other 
   universities are vying.  "Well," he says, "there happens to be a 
   Dr. Ernest Selmer, [...] here, who is a Norwegian, and he has been 
   with Princeton with Von Neumann for a few months, and then he has 
   been up with Paul Morton [.]
   [...]
   So we got a hold of Selmer, and he came over and consulted for us 
   for a period of time, and played a major part in the initial logical 
   design of the ElectroData computer, as we called the thing.

Bradburn goes on to mention that they built their own drum 
for the computer.  (Unclear if it was before/after ERA's drum.)

Indeed, echoing what I learned about Burroughs self-sufficiency when I worked there:

   Yeah, we built our own drum.  Yeah, right from scratch. 
   We built everything from scratch.  There wasn't anything purchased 
   but the typewriter.

ElectroData was also involved with ERMA:

   This was ERMA. So we built a special drum. We built the drum for ERMA. 
   And we built a tape unit for ERMA. So we got into tape units in the 
   late forties, too, making the first tape unit. So, we built our own mag tape units.

   ...

   Then we went from there to putting mag tape on the machine [the ElectroData
   computer], because we developed these machines for ERMA. 
   As you'll recall, we built the tape unit, we built the drum for ERMA; 
   Bendix built the electronics through Stanford Research Design, 
   and I think Friden or Marchant built the keyboard devices for it. 
   Stanford Research put the whole thing together, and then subsequently, 
   GE came in and won the contract to make the whole thing and carry it on.

An interesting claim about using transistors:

   Well, yeah. We got in very early in the transistorized. 
   We were probably the first machine to make a transistorized device.
   And this was a--this was just tape to printer alone, 
   with complete format control.  And this must have been, oh, 
   someplace in 1957 or 8, I think it was. 
   And this was a pretty sophisticated thing, in that we'd take the standard tape 
   off of the, the computer and put in separate format information into the, 
   into its own computer, if you'd call it that. I don't know what 
   else you'd call it. So we could end up making printing in the type 
   of format that you wanted.

Since the Bell Labs TRADIC was in 1954, and the MIT TX-0 was 1956,
he may have misremembered his dates ... the interview was done in 1973.
...or he may have been unaware of the other items, or he may have been
referring to *commercial* transistorized devices.

An interesting reference to patents for revolving drum storage devices:

   Oh, Cliff Berry [from Iowa] worked for ElectroData. 
   He was the mass spectrometer guy, see.  [...]  And when they got into the 
   revolvers, and IBM questioned the validity of the patent, 
   we says, "No, no way." Because Berry says, you know, 
   "I've seen the guys working on a capacitor revolver-type thing 
   and that way preceded that."

Finally, a reference to a B-Box, presumably on early ElectroData computers:

   But, there was another advantage we had; and that was, 
   we had additional registers known as a B-box that entered into 
   our machine, so you could store added factors or additions in here 
   to modify instructions or to update instructions or modify the 
   numbers, particularly on repetitive programs.

(Somewhere, I read that the IEEE mentioned that the ElectroData 205 was the
first commercial computer with an index register.)

On early memory:

   When the very first of the old memories were just under development, 
   we didn't want to get into the situation of those cathode ray tube 
   electrostatic memories because they were on the way out just 
   about when... shortly thereafter as people shifted over to cores. 
   But cores hadn't gotten going.  So the best we had to work with 
   was drums and trying to speed them up.  And disks, to a degree. 
   But the drums seemed to be a better answer than the disks, 
   although subsequently, of course, disks proved out pretty well.

On the development of the head-per-track disk drive:

   [Bradburn was asked "So you actually didn't do any work in the disk field?"]

   Well, we did at ElectroData, at Pasadena, at ElectroData along about 
   '58 or '59 [as part of Burroughs], because it became pretty evident 
   that any of the moving 
   arm disks of that period were very, very unreliable. 
   And the maintenance costs were eating them up.   
   This was before the advent of the 2311. And even the IBM ones were 
   highly unreliable.  So we built them totally enclosed, 
   a head per track, big disks, at Burroughs.

On design automation:

   MAPSTONE [the interviewer]: I suppose this was predominant in the 
   aircraft business where they were building guidance machines. 
   There are two schools of logical design. You know the East Coast 
   block diagram approach and the West Coast Boolean algebra logical 
   design approach. 

   BRADBURN: Oh, I see what you mean. Yeah, that's true. We were proponents 
   of getting all the logic, particularly on the B-5000 was written 
   in Boolean algebra.

   MAPSTONE: Oh, it was.

   BRADBURN: Yes, it was all simulated, and checked out in time on some of the best, 
   fastest machines we could get a hold of at the time. 
   I think we used one of the fast Philco machines, 
[SS: perhaps like the one we have in Visible Storage? ... this would be circa 1959/1960]
   they built an earlier one there, or one of the big UNIVAC 
   So we could prove out every program step before we actually built 
   the thing in hardware. It was quite a step in those days. 
   But the whole logic of the machine was worked out in 
   Boolean algebra. And then we had some of the earliest 
   design automation. I don't mean design automation in 
   the simple sense of just blowing up a parts list and 
   splitting them up in orders like a lot of people call 
   design automation. But the machine itself was designed 
   by virtue of the logical guys writing Boolean algebra 
   equations, and then the design automations pulled that 
   together and actually produced block diagrams.
   [...]
   Rather than going to block diagrams and then working out there what 
   the circuitry is going to be. [...]  So this was some of the first 
   work done, I think, at ElectroData, was involved in using that 
   kind of an approach to the problem.

        
More interesting oral histories at:
   http://invention.smithsonian.org/Resources/Fa_Comporalhist_Index.Aspx

Stan