1 SEACvideo.mov History of SEAC Presented by Russell A. Kirsch Russell Kirsch:

1 SEACvideo.mov History of SEAC Presented by Russell A. Kirsch  Russell Kirsch:
1 SEACvideo.mov
History of SEAC Presented by Russell A. Kirsch
Russell Kirsch:
The National Bureau of Standards built the first computer with an
internally stored program in the United States. The Standards Electronic
Automatic Computer, which we call SEAC, did its first productive
computation in April of 1950 here at the National Bureau of Standards,
which is now called NIST, the National Institute of Standards and
Technology. We are here today to show you how everything was started
inside the federal government, although of course there are some
commercial interests, which would have you believe otherwise.
2 SEACvideo.mov
Members of the Team
We have here today three of the engineers among the group of eleven
who received a medal from the Department of Commerce for the design of
SEAC, which in turn is part of a larger group of about 33 engineers and
technicians and mathematicians who were involved in the design of this
first computer. We have first Ruth Hauter Cahn, who was one of the first
early design engineers. And, Robert Elbourn, Bob, perhaps came earlier
than any of the other of us. And, Sidney Greenwald,. My name is Russell
Kirsch and I came to the Bureau of Standards shortly after SEAC began its
first productive computation and I was involved in the design of additions
and modifications for the SEAC over a period of its whole lifetime and
have remained at the Bureau of Standards for a continuous period, which
is now in its 52nd year.
3 SEACvideo.mov
What makes it the first computer
I mentioned in my introduction that SEAC was the first computer but I
think it’s worthwhile for us to explain what we mean by the “first
computer.” Certainly, the first computer was Babbage’s computer in the
1800’s but in no sense was it the computer that we think of today. would
one of you like to discuss the question of what was the first computer and
in what sense was SEAC the first computer? Do you want to try that Bob?
Elbourn:
Well, ENIAC is usually been pushed in the newspapers as the first
computer but it was not a stored program computer it normally was a
wired program that had to be re-wired for any other problem for the next
problem Actually, it was designed to compute range tables for Aberdeen
Proving Ground. That’s what it was set up for the single program and it
was usually not changed from that. John von Neumann suggested that the
table of drag coefficients could be used for storing instructions for the
machine and make it into a stored program computer. but not one in
which you could change the instructions stored there through the machine
so it really wasn’t the kind of machine we’re talking about but what John
von Neumann suggested did lead to the type of machine we’re talking
about.
Of course, all of this was after April of 1950 when the SEAC did its first
productive computation with the exception of the ENIAC which was 4
years earlier. But it was not, as you point out, a stored program computer.
Elbourn
April of ’51, that’s right.
Cahn:
It was much slower, also, wasn’t it?
Kirsch:
The ENIAC?
Cahn:
Um hmm
Kirsch:
slower, yes, but in all these cases so much faster than
electro-mechanical machines that the age of electronic
machines was really introduced by the ENIAC but the big
jump was to the stored program concept which SEAC led
with.
Elbourn:
Yea, I would say that is correct.
4 SEACvideo.mov
Why was it built
Kirsch:
I think it may be worthwhile to mention something about
why the SEAC was built. , , the government was interested in purchasing
a commercial computer, and let a contract to the UNIVAC company for
the construction of what became the first commercial computer. And,
there was the question of who would be the contract monitor, and a bunch
of you smart guys were likely candidates to be the paper pushers who
would monitor the, commercial contract. The, UNIVAC computer was, in
a certain sense, more ambitious than SEAC because Eckert and Mauchly
were, had their mind, had their eyes on commercial business applications
which would involve numerical as well as alpha-numerical computations.
And, that made the delivery time of the UNIVAC quite a bit more delayed
than had originally been anticipated. SO, the Bureau of Standards was
asked, volunteered really, to go ahead and sort of put together what was
then called, “an interim computer,” and in those days I guess we knew
that “interim” meant “permanently on into the future.” So the SEAC was
sort of thrown together, as it were, and in fact went into operation, was it
about a year or so before the UNIVAC was delivered and became the first
machine.
Greenwald:
Well, , as far as I know, in 1947 we were just working on
various components, coders, decoders, , I remember
working on, , looking and seeing how many, how much we
could put on a tape per inch and all that kind of thing, you
know this wire recorder. And, there were just many
miscellaneous little projects that didn’t seem to have any
coherence. And, then all of a sudden in 1948, I wasn’t sure
exactly what happened but we got the go ahead to build
something and that set the spark. I mean all of a sudden the
place really came to life and, for the next twenty months in
building that machine I have never felt anything such, so,
so much exhilaration. It was just a tremendous part—I
would say the best part of my career, anywhere at any time,
it was just fantastic. You were discovering new things all
the time. You didn’t know what was finally gonna come
out of this because digital computers were a very unknown
quantity. They called them giant brains and there was all
kind of hoopla in the newspapers. But, you really didn’t
know what was going to come out of this. It was just a
fantastic time and I don’t expect I’ll have an experience
like that again.
Kirsch:
That was all caused by $150,000 that George Dantzig
managed to liberate from the Air Force.
5 SEACvideo.mov
Kirsch: There were a number of applications of the SEAC computer
that we introduced. Of course it’s well know that we started the field of
image processing in 1957 with a scanner based on a rotating drum lathe
type device that Ruth’s husband, Leonard Cahn built and we demonstrated
the first image processing on the SEAC computer. A more unusual
example was a problem that was posed by the Patent Office, which is part
of the Department of Commerce. They wanted to be able to search for
chemical compounds in the patent files and until that day the only way of
searching was based on the nomenclature, on the names of the chemical
compounds. It became clear to us that one could actually feed in the graph
theoretic description of an organic molecule to the SEAC and one could
search for fragments based on the topology of the molecule rather than the
name of the molecule, you see. And, so we wrote this code for structure
searching on steroid molecules and fragments thereof and it was an
extremely ponderous code, but it ran, you see. And, we did convince the
patent office that this was the way to go, which subsequently they did do.
But because the code was so ponderous it was something that we hoped
you would never have to use. Well in those days the idea of something
you that you hope you would never have to use was the H-bomb. So of
course we called the code the “H-bomb” code. Well, some how or other
the code sheets got misplaced and when it became known that the H-bomb
code sheets had been misplaced, the FBI came in. And, of course it took a
lot of explaining to explain that this had nothing to do with the H-bomb, it
was just a name that we chose to use for the code sheets. But the Patent
Office did benefit by this introduction of the use of computers for
searching in chemical literature, which of course now is a routine thing
that not only the Patent Office but all the chemical companies use every
day.
Greenwald: Well, for example, they studied the flow of traffic. I mean,
you know, today we have we talk about gridlock and so on and it seems
to me, I remember they ran programs on SEAC that would try to talk
about what the timing should be with lights on this avenue and that avenue
and so on to make the maximum progress. There are very many
interesting programs
Cahn:
I am very much aware of that one because Leonard worked on
that. And, I remember very well when they were giving the presentation
at some hotel or something or other. What they did was have a movie
camera and they would shoot every so often and they had this, when they
ran that thing that showed the traffic moving and so forth. It was just dots
on it and so forth. They made this movie tape and sent it to be developed
and so forth. The post office lost it somewhere- somewhere in the mail.
So, they ran the whole thing again and Leonard brought it home and
developed it in our dark room in case they didn’t find it.
Kirsch: Another example of a very interesting simulation is the one that
Bob and I worked on. It was an attempt to simulate the behavior of naval
vessels on a radar display. And, Bob got the differential equations for the
motion of a ship given a specific rudder and velocity command. And, I
remember a demonstration that we did for some navy admirals in which I
was running this program using actual numerical data that we got from
David Taylor Model Basin about the dynamics of ship motions. And, I
was in the process of explaining to an admiral that here is a certain type of
ship and if you give it a certain velocity and rudder you can see that it will
collide with another ship, which is of course much more desirable than
doing it on the ocean. And, this admiral said no, no no that won’t collide.
I said, oh no, you’re mistaken because you can see that with this rudder
and velocity command the two ships will collide. He said, run it. And, I
ran it and they did not collide. I said, oh, how do you know that. He said,
I used to be the commander of one of those ships, which is of course is a
great form of compliment. And, we had many experiences like that in
which we were convinced that the kind of simulation programs that we ran
were in fact good simulations because they held to the mirror to nature, as
it were.
Elbourn:
That brings up the fighter simulator. The combined digital
and analogue computing operation that we had. We had
built a model of the cockpit of a jet aircraft and put all the
instruments in there. Our analogue computer systems
operated the instruments but the SEAC did the trajectory
and the altitude of the airplane in space although it was just
sitting there in the lab. In there the jet pilots came over
from Andrew’s and tested our simulator, got out and said
yes that’s a pretty realistic simulation. So the guys came up
and invited me down to get in it. I got in and the first thing
that bothered me was the artificial horizon because it
looked like a little airplane that you would control with a
joystick. As a model builder, I was familiar with radiocontrolled airplanes. It turns out that the artificial horizon,
to try to correct that requires exactly the opposite motion
and I was just beginning to get squared away with that
when my hand reached in and pulled the throttle back and
pushed this thing up to simulating 1500 knots. I’d get
things straightened out and then my mind would flip over
and use the other model and everything would start
spinning crazily. When I finally gave up and crawled out,
they said well Mr. Elbourn you’re 900 feet under ground
and inverted. The boys got a big kick out of that.
Elbourn:
we mentioned Joe Wegstein. On TV nowadays you see a
lot about computer matching of fingerprints and Joe was
really a pioneer in that field. He developed computer
matching under FBI sponsorship for a number of years.
6 SEACvideo.mov
Kirsch
Another one of the successful experiments with the
SEAC computer was to start the field of image processing,
digital imaging as we know it today. In 1957 we built the
first scanner which could scan a photograph and read it into
the limited memory that we had on the SEAC computer.
This is the first picture that was ever scanned in a
computer. It's a picture of my son when he was born in
1957 and when we read it into the SEAC computer it was
possible to print out the scanned image . And this is the
first picture that was ever fed into a computer. It consisted
of only 176 by 176 pixels, each pixel being only one binary
digit, that is black and white and from this grew the whole
field that we now know of which includes everything from
CT scanners to earth satellite moniotoring to bar code
readers to all the amazing things that are done with digital
imaging. And so we see that there have been changes that
took place. Some of the changes are very big., the changes
in the technology. Some of the changes are less big.
Here's my son today and you can see that people have
changed less than computers have in these 40 or 50 years.
7 SEACvideo.mov
Users of SEAC
Kirsch: We are all engineers having had responsibility for the design, the
construction, the maintenance and operation of the SEAC
computer. But, there was an entirely separate group of
people, mathematicians, whose primary goal was to use the
computer for productive computation. Bob would you like
to describe the role of the mathematicians in starting the
SEAC computer?
Elbourn: Well they worked at , designing computers in a broad, general
way. They had a machine development laboratory, which
produced descriptions of the list of instructions that the
machine if constructed should perform. And, with that list
they wrote programs. The programmers got tired after
several years of writing programs that they could never
test- they had no machine to run them on to see if they
would run. They complained they wanted to build a
machine. That I think was primarily what motivated Sam
Alexander to go after that $150,000 for the Air Force to get
our machine started. Before that time we had been directed
by the committees that , oversaw the mathematical
development by the government to work on computer
components on our funding as components for the next
generation of computers. That’s why we were after that
time had not built, undertaken to build a machine.
Kirsch: The, the community of mathematicians working on SEAC was a
truly illustrious bunch of people. Their field of numerical
analysis was sort of in the doldrums until a truly
spectacular group of mathematicians arrived in the
Washington, D.C. area after the war and were all assigned
by various agencies to work at the Bureau of Standards
using the SEAC computer. And, this created a sudden
great leap forward in the whole area of numerical methods
for doing mathematical computation. And so that analysis
using formal mathematical methods began to be replaced
by analysis using numerical approximations. As we know
today the field of numerical analysis got a very strong head
start through the use of the computer, the SEAC computer,
by the illustrious group of mathematicians, many of whom
were at the National Bureau of Standards as employees,
others of whom were delegated to the National Bureau of
Standards by other agencies.
Elbourn: We had many alumni at the mathematical- WPA Mathematical
Tables Project under Arnold Long during the depression.
They did their calculations with desk calculators, and
learned a lot about numerical analysis.
Kirsch: And actually the government’s, one of the government’s most
popular technical publications, the all time sort of “best
seller,” was the Handbook of Mathematical Functions,
produced by, edited by Irene Stegun and Milton
Abramowitz, based entirely on the contributions from
mathematicians who were using the SEAC for numerical
computation. And, to this day, that becomes an important
publication, which I understand, I’ve heard recently is now
being updated. Instead of being just in printed form, will in
fact be available on the Web so that people can continue
the process that was so illustriously begun by
mathematicians on the SEAC but now one can do this at
home, on one’s own personal computer, using the best
that’s being thought and said by contemporary
mathematicians.
8 SEACvideo.mov
Influence of SEAC
Kirsch: The SEAC had an immense influence on computing government
in academia and industry. At first the SEAC, of course,
was the only computer available to the government. But
we did have a missionary in the form of Sam Alexander,
the head of the computer laboratory. And, Sam would go
around to other government agencies explaining what
computers were, giving them the opportunity to actually do
computation on the SEAC and in the process learning not
only about computers but learning about the business they
were in ordinarily. And, of course there are tales perhaps
appocraphil about people who, when they were exposed by
people like Sam, to the notion of computing, went back and
did a study of how to use computers and proceeded to
improve their own operating procedures before they
actually used the computer and of course achieved an even
greater benefit after implementing some of the operative
procedures on the computer. We had a very distinguished
group of alumni from academia and from industry who
were working on the SEAC computer who went out to
become heads of large corporations, heads of computer
science departments and carried the experience that they
started gathering on SEAC onto later computers and into
other institutions. So, in a certain sense, SEAC not only
did valuable computations but it did very valuable
education. Are there comments that you want to make
about some of the influence of the SEAC on , computing
everywhere else?
Greenwald: Well, we know for example, that our whole logical design
group as it turned out one day up and went to IBM and
Kirsch: This was Al Leiner and his group.
Greenwald: Yes, yes. So, that happened to a lot of people- going to
various corporations and taking what they had learned from
SEAC to other fields.
Elbourn: Well we worked closely with the Atomic Energy Commission.
Sam Alexander had clearance and he was a consultant to
the committees that recommended computing facilities for
all the atomic energy laboratories. I went with him on
those trips a time or two but they used a a very large
amount of time on the SEAC when there were no other
computers available that early day. And of course we
couldn’t go near the machine when they were using it.
Greenwald: But one of the - I think we outta point out was that one of the
difficulties with SEAC was that it was serving two
purposes. On the one hand, it was an experimental
machine, on the other hand , it was doing useful work.
And, it would work 24 hours a day and the engineering
staff had something like eight hours of prime time. We
would come on sort of during the afternoon and so on. But,
that was very difficult because on the one hand if you’re
building something and experimenting with something,
you’re making changes and so on and then someone comes
on at five or six in the evening and they want to use the
machine. And, who knows what you’ve done to upset
things. So, it was a very difficult task to keep those things
going but somehow we managed
Kirsch: In the area of education, the National Bureau of Standards actually
paid for my honeymoon. When we were married in 1955, I
was given the assignment to go around to Air Force bases
all over the country teaching them about computers. So,
we had this wonderful honeymoon, thanks to the federal
government, in which we to these God forsaken places on
Air Force bases and I would teach the general staff about
computers which was quite a startling experience because
here you would have majors, officers, generals and stuff
deferring to this young kid who was telling them about this
whole new world of the future. And, of course, the world
of the future came to pass and is now, of course, very
routinely part of the operations of everybody, including the
Air Force.
9 SEACvideo.mov
Components Tubes Wires How Did It all Work?
Kirsch: Sidney mentioned working on components on the SEAC and we
are fortunate to have today from the storehouse at the
Bureau some of the components of the original SEAC
computer and we’d like to discuss those with you today.
I’ll tell you what we have here and then we can discuss the
individual devices. This cabinet contains two tubes that
were part of the mercury memory, which in its vast
capacity consisted of 10 K bites of storages. That’s ten
kilobytes, some hundreds of thousands times less of what
we have nowadays in a small laptop computer. This is the
amplifier that was used to re-circulate the acoustic pulses in
the memory. On your left you see some of the typical
wiring of the kind that was used in the SEAC, with vacuum
tubes showing and all the hand wired soldered joints. And
finally we have here a magnetic wire cartridge which was
the main device used for putting information into the
computer and taking it out of the computer. Would any of
you like to discuss any of these devices in particular?
Greenwald: Well, I’d like to make one comment about… I said that I had
worked on radar equipment before this and in working with
other engineers on the equipment, you built it and you
would design the circuit and you put it in this, rusty
equipment, and if something failed you would adjust the
value of a resistor, a capacitor and so on. That wouldn’t do
for computers and I found that out very early and it was
really an eye-opener. We, we when we, computers are
have to have extreme reliability. Remember that when you
have a television set, you have in those days, you might
have, say twelve, fifteen tubes or twenty tubes or whatever
it was and they were always burning out and you had to go
to the drug store and replace them and so on. Well, can
you imagine a computer then, SEAC had something like
1300 tubes, but you’d be forever replacing burning out
tubes and resistors would go bad and everything else. But
how do you get a machine with that many components, and
I think that had, I don’t know how many resistors, I think it
had maybe 15,000 diodes on that order. It was just a
fantastic amount of equipment and it all had to work.
There was no redundancy in this machine, remember.
There was no checking. And, it still had to work hour after
hour to be useful. Well, what we did when we designed
this circuit, the main circuit for the machine, we employed
what was called to me, a very novel thing, “worst case,”
and you assume that every resistor can have its worst value
or its best value, same thing with the diodes, same the tubes
and so on, and you designed such that no matter what
happens within those limits of parameters, that thing is just
gonna chug away and work and work and work. And, you
ended up with an extremely reliable main stage that was
used throughout the machine. That to me, I said, was a real
eye-opener and is, I think, what I think made the machine
work, and work as well as it did.
Kirsch: Well did the fact that there was one main stage consisting of a
vacuum tube and a post- transformer feeding the diode
logic network, made it possible to do a kind of marginal
checking that we did in which we could lower the return
voltage from all of the vacuum tubes uniformly through the
whole computer, thereby decreasing the tolerances that you
had designed into the components and we could cause the
machine to marginally fail during testing time. That was a
very important notion of having, create a worst case which
of course you had designed for.
Greenwald:
Worse than worse. [laughter]
Cahn: Well I think that one of the main things was the type of way we
designed the thing there. The tubes were used only for amplification. All
the switching and so on was done with the diodes.
Kirsch: Ruth refers to the germanium diodes. And, at the risk of
[laughter] doing something very dangerous, I’m going to unplug one.
This is an old vacuum tube base in which there are one, two, three, three
diodes wired in. And, this is the way in which all the logic was done on
the SEAC computer with germanium diodes, which were a notoriously
unreliable component.
Cahn: They were pretty much unknown.
Kirsch: Unknown, yes. Remember, this is before there were transistors so
this is one of the early solid state devices and it was not
possible to do marginal checking so what we would do is
what I just did: unplug them, test them, put them back in
thereby possibly creating more problems than had actually
existed ahead of time. We also compiled vast records on
the tests which were utterly useless. Another one of the
components that you see here is this magnetic wire device.
Ruth…
Cahn: It’s not in the original computer.
Kirsch: Right. Would you like to describe the out scriber that you and I
built for using the wire cartridges?
Cahn: Yes. The original, first, the original input and output of the
computer is a tape or typewriter…
Elbourn: Perforated paper tape
Cahn: …which is very slow getting into the computer and very slow
getting out. So, one of the first things they wanted to do was speed that up
and they went to this wire cartridge here which you could somebody built
what we call an inscriber which put information from the tape onto the
wire there and this one into the computer---seconds instead of hours that
perhaps a long tape would have taken and so forth. But that means that
outside of the computer you had to put the information on the wire and
take it from the wire back on to tape and that one that did the second thing
was what we call the out scriber, which was a little more difficult than
putting it onto the wire.
Kirsch: One of the difficulties with the out scriber was that the magnetic
wire on which information was recorded quite densely and
could be read very rapidly by the computer, had to be read
very slowly for purposes of driving paper tape punch and
since the wire was moving very slowly, the magnitude of
the signals that we read were very, very weak. One of the
problems we encountered was some noise that was very
hard to detect and it would occur in bursts at various times
and I remember trying to determine where it came from.
And, so I did some radio detection triangulation and we
discovered that there seemed to be bursts that were coming
from the National Security Agency at which point we
decided there’s no point in trying to find out what was
causing it but we did manage to circumvent the difficulties.
We had other kinds of problems on the out scriber that I
remember.
Cahn: Yes, that was our baby.
Kirsch: Right. One of the problems, perhaps one the earliest problems in
human factors design was that there was a reset button on the out scriber
and this is one of those buttons that you push that doesn’t give a click. So,
it was what I called a “non-satisfaction” type button. And, we would find
people pushing the thing, almost pushing over the cabinet of the out
scriber because they didn’t hear anything click, you see. Very elementary
thing that we learned later on—is necessary.
Cahn: I think they had the same problem when they controlled the
computer…mathematicians push those buttons to death.
Greenwald: Well, you talk about problems, I was involved in in, in
designing the circuitry for the wiring, for the wire input, we
used to call it the wire dumper, and also the tape units and
one of the problems we had was on the floor above us there
was what they called the tube lab and they fabricated
original tubes of various kinds and in the fabrication they
have to out gas these tubes. You know first they get a very
good vacuum and then they have to out gas it. Well, in
order to do that they had a RF transmitter there of I don’t
know some unknown huge power, and it would pulse
repeatedly while we were trying to work the machinery.
Kirsh: Not only was the pulse repeated but it was aimed at us. [laughter]
Greenwald: And trying to get that interference out we had to do an awful,
awful lot of grounding and shielding and so on to try to avoid that. But
they made life difficult, but we, we overcame it.
10 SEACvideo.mov
How Did It All Work Part 2
Kirsch:
On the SEAC computer we had a device which was a
modification of an old office dictating machine which was
used to let the programmers store their programs and their
data in such a way that they could remove it from the
computer . This wire cartridge is one of the devices that
was used to store information external to the computer. It
consisted of two reels of steel wire with a thin magnetic
plating on the wire and the information stored in binary
form in such a way that the programmers could record their
programs on the wire, bring it to the computer and then
record the results on the wire from the computer and
remove it where a separate device called the outscriber
could convert the information from the wire to a punched
paper tape and then from the punched paper tape to a
printed form which people could read. Here you see the
operator plugging one of those cartridges into the device on
the computer which would read the information from the
wire cartridge. And that was the ponderous method that we
used to put information into the computer and to remove it
from the computer. I have a prize find here i"d like to
show you . Ruth was mentioning the input/output problem
and after the wire cartridge was transcribed to paper tape, it
became necessary to run the paper tape on a Teletype
machine to actually make a printed hard copy. And, those
of us who were involved in working with computers and
using limited resources, at times had to use all kinds of
ingenuity to be able to make do with our limited resources.
Here’s one such example: [laughter] One night, we ran out
of paper for the printer and some clever person managed to
print out the results notwithstanding our limited resources,
which demonstrates what you can do with ingenuity and
limited resources. It also demonstrates that they don’t
make toilet paper the way they used to. [laughter]
11 SEACvideo.mov
Memory of SEAC
Kirsch We would like to talk a bit about the memory of the SEAC
computer. My own laptop here contains 500 megabytes of
information which to most people nowadays is not at all
startling. What might seem more startling is to understand
that on the SEAC we had 10 kilobytes of memory, that’s
25,000 times less memory on the SEAC than we have in
my little laptop computer here. Can we discuss the question
of how the memory on the SEAC worked, two different
kinds of memory?
Elbourn: Presper Eckert Jr. of the Moore School project had worked
during the war on the problem of illuminating the ground
return from radars and the trick they adopted to do it was to
store one return from one thing in an acoustic pipe of
mercury and then subtract that from the return of the next
thing in the radar and then a different stationary part
cancelled out. Only moving targets produced an output that
was detected.
Kirsch: This is one of the pipes in which the mercury was stored on the
SEAC memory.
Elbourn: So, in that machine they adopted this, , mercury acoustic line and
we copied their plan for the SEAC memory and that same
type of memory was adopted in England for the ACE and
the EDSAC. Except in their version, their lines went down,
they used a 45 degree reflector at the bottom of another that
came back up. The amplifier circuits were on top. But this
was a fairly widely used form of computer memory at that
time.
Kirsch: The acoustic memory for SEAC consisted of glass tubes of which
these are two examples here with mercury and quartz
crystal at opposite ends of the glass tubes. Then, the signal
that would be fed into the quartz crystal at one end would
propagate down in exactly 384 microseconds to the other
end where the signal would be read by another quartz
crystal transducer fed into this amplifier and re-circulated
back so that one could put information into the memory or
take it out by using the correct timing and pulling out the
signal at just the point when it came out of the mercury
delay line. The total size of the memory consisted of eight
foot high racks of amplifiers and mercury tubes like this,
two big cabinets, which would have taken at least the space
from me to Ruth, eight feet high in two large cabinets, and
that was a total of 512 words of 48 bits each of memory.
There was also an electrostatic memory.
Greenwald: Well before we go to that I want to point out that there were, I
think 64 of these to make the memory, right?
Greenwald: And it’s kind of interesting, you know, we, we, we play with
our machines today and we don’t care what the temperature
is in the house and so on. This memory was so the
temperature had to be keep, kept so precise, they had a
thick aluminum plate, the whole size of the cabinet inside
and that had to be kept plus or minus a quarter of degree
Fahrenheit, I believe, in order for this thing to work.
Otherwise, the delays would change. So, if for any reason
the power went off in the building, let’s say, and that
temperature was disturbed, you might have to wait 24 hours
for the thing to stabilize again. That’s how precise things
had to be.
Kirsch: Actually the mercury acoustic memory was one of the more
notoriously unreliable parts of the computer and so we had
very sophisticated diagnostic programs which could put
information into the memory and read it back out and
compare what had gone in with what had gone out to see
whether the memory was failing. And, quite typically any
failure that was detected would be a failure in the memory,
in the acoustic memory much more so than in any of the
other components.
Greenwald: Yea. There was another trick that we pulled, I don’t know I
guess whether other people copied this from us. But, on
the main bus where all the information went out, we had a
little radio and you could hear, you could hear the
information across change somewhat. But, you could hear
the machine actually working and you could tell whether or
not something had gone wrong. Because usually when a
machine gets into one of these endless loops and spinning
its wheels, you’ll hear the same thing over and over and
over again and you know the program has been lost
somewhere. But this was a very rough diagnostic.
Kirsch: I’d like to perform a little experiment since Bob is interested in
music, I’d like to whistle something to you Bob, and you
tell me the name of what I’m whistling. Okay?
[Whistling] Do you recognize that?
Elbourn: No.
Kirsch: Ruthie, do you?
Cahn: Fiba fiba foo.
Kirsch: Fiba fiba foo. That was what Sid is talking about, the mercury
delay line memory test program, which would run at one
megahertz. So, you tune the radio in to the one megahertz
signal and the envelope, which was an acoustic signal, is
what was the sound that Ruth says was called “Fiba Fibba
Foo.”
Cahn: It also printed it out if it went correctly.
Kirsch: Yes. So, we had many different ways of testing the computer.
Including the audio tests. Can we talk about the other
memory, the electrostatic memory? As I recall, Bob, you
did the great design of the last of the great electrostatic
memories at a time when they were unfortunately obsoleted
by transistors.
Greenwald: …By ferrite cores.
Kirsch: And ferrite cores, yes.
Elbourn: No.
I don’t think I designed it.
Cahn: Bill Davis did
Elbourn: Bill Davis…
Kirsch: Bill Davis built the original electrostatic memory. I was referring
to the last version that you did, Bob.
Greenwald: Oh oh. That’s what Ruth and I worked on. What happened
was that the memory, this cathode ray memory in which
one digit is on one tube and there are 45 tubes actually.
What happened was that the, the machine on this memory
was notoriously prone to error. I mean, it would work one
day and the next day it wouldn’t work and then there’d be a
lot of scrambling and so on. It was finally I think a fellow,
Davis left and I think Holt left or he went on to something
else and Ruth and I were given the job of trying to make
this thing really, really work. And, we went through
everything we could think of. We went through the
amplifiers, we went through the grounding, we re-wired
things. Do you remember the various things
Cahn: Not really.
Greenwald: We looked at them. We looked at the power supplies looking
for transients. We had a whole long list of things. But
when we finished the thing really worked and then became
and that was, I think our pride joy. I think Ruth doesn’t
remember that. But, I remember very well.
Cahn: I remember working…
Kirsch: What you may remember was that in maintaining the electrostatic
memory, since the information was stored on 45 cathode
ray tubes, each binary digit being one spot electrostatic
storage on the face of the cathode ray tube. Occasionally
there would be blemishes on the phosphor, which would
cause one binary digit to fail. John Rafferty, had a
mysterious ability to change the deflection amplifiers on all
of the 45 cathode ray tubes simultaneously so as to move
the raster off of any of the defects, any of the blemishes on
the face of the tube.
Greenwald: And he kept a map of the known defects.
Kirsch: And he kept a map of where the known defects were on the face of
these tubes. Then later on, you managed to use some new
tube designs that RCA came out with which seem to have
fewer defects on the phosphors, but again it was a good
design too late, unfortunately for the technology.
Greenwald: Yes, in 1950 I think a fellow at MIT came out with using a
magnetic core memory and Jay Forester changed the whole
complexion of memory and from then on the mercury
memory and the CRT memory became obsolete.
Kirsch: So, a lot of the technology of the SEAC became obsoleted by
progress. And, that may possibly account for why so few
people know that the SEAC started everything.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement