Day 1, Tape 1, Side 1
In 1972, Joseph Desch, Robert Mumma and Donald Eckdahl of NCR were approached by the Smithsonian Institution and asked to donate their recollections to its Computer Oral History Collection, now part of the National Museum of American History. Desch and Mumma were interviewed on the 17th and 18th of January, 1973 by Henry Tropp in former NCR Building 10.
HT: This is a discussion with Mr. Joe Desch and Mr. Robert Mumma and we’re sitting in a nice, little office at NCR. The date is the 17th of January, 1973.
And, as I told you, before we started, I — I really wanted to have a – an informal kind of give and take session on what it was like in that early period. And so why don’t we start with the formation of your electrical engineering department and how you got interested in the development of a computational device.
JD: All right, I will be glad to tell you.
I came to this company on April 25th, 1938 at the request of Mr. Williams, who was my superior at Frigidaire. My instructions were to develop an electrical research laboratory. The company, up till that time, had no electrical research laboratory. It was not called an electronic laboratory in those early days. There was no Electrical Research Department in existence, I had to start one from scratch. .. There was no program given
to me, because no one knew exactly what application electrical devices would have in the cash register or accounting machine business that NCR was in at that time.
JD: For a few months, I didn’t hire anybody. I just .. .. studied thoroughly the product line of the company, .. visited all the areas that possibly could offer me .. .. ideas for the use of electricity, and — and then formed a department .. with myself as the only employee.
HT: Before you go on to .. the, the next stage, it might be worthwhile to describe the product line at that time. We all know about cash registers, and know pretty much what they looked like in that period, but what are – what were the other elements in the major product line?
JD: Well, the company manufactured, in Dayton, Ohio, cash registers of a large number of models; .. in the hundreds of different models. .. And they. manufactured accounting machines. .. Rather simple by today’s standards, like the 3000 – Class 3000 accounting machine was made here. We also .. made .. posting machines for banks. We made . . .
RM: The Class 2000.
JD: the Class 2000, which is an accounting machine and also is used as a posting machine. We made it generally for banks and retail stores. Cash registers went to supermarkets, department stores and so forth, that sort of thing. .. I think, at that time, we owned the, the Ithaca plant .. for the manufacture of adding machines, and we still do, only the adding machine has been downgraded by electronic devices now. It’s no longer the same machine.
RM: That was the Allen Wales plant we took over.
HT: Allen Wales?
JD: Allen Wales plant. RM: W-a-l-e-s JD: .. I don’t know, there may have been some other devices. Bob, can you think of any? RM: Well, there was that charge system that we put in these stores. The telephone system that enabled them to check their charge accounts. JD: Credit authorizations. RM: Their credit authorization, yes. [page 4] JD: Yes. HT: Was that a mechanical — ? RM: We had an Electrical Engineering Department, one of whose functions was the development of this communications system for credit authorization. HT: Oh, I — I don’t remember that. Could you describe it in just a few sentences? RM: Well, it was a telephone attached to a sales slip punch. When the sales person wants to check your credit,she lifts the telephone which connects her directly to the Central Office. At this time she also puts the sales slip into the slot in the punch. HT: Mhm. RM: The sales person gives your name, account number, and amount of purchase to the credit authorizer over the telephone. If your credit is satisfactory the punch will operate punching a pattern of holes: in the sales slip. The sound of the operation of the punch indicates that the transaction is complete. HT : Aha RM: But if the punch didn’t operate, you were not a good risk. HT: Aha. RM: It was that simple. [page 5] HT: It was a very simple, direct system then. RM: That’s right. HT: Gave the information from a central source. RM: Where they looked through the files. JD: I’d like to make a point here. .. Bob mentioned, or you mentioned, the electrical engineering department. I have to distinguish between what we called the electrical engineering department and this electrical research department that I was forming. The electrical engineering department was concerned primarily with motors which we manufactured here RM: That’s right. For the cash registers. JD: and with this communication system for credit authorization. That was about the extent of the electrical effort and it was all very simple. HT: Mhm. JD: .. Most of the work, nine-tenths of it — and there were only two or three people in the department anyway — was directed to motors, because we manufactured a wide line of motors. Practically every machine had a different motor and we made the motors. RM: You see, that — that motor was developed by Kettering for the cash register. [page 6] JD: Yes, that’s right. RM: It was a series motor. HT: Oh. JD: And to keep that into production, the Electrical Engineering Department had to provide engineering effort. HT: Mhm. JD: They weren’t inventing anything particularly. HT: I see, aha. JD: Earlier in this interview I mentioned that I studied the product lines then existing at NCR Corporation. I was also looking for applications which could use electrical or electronic technologies applicable to these product lines. My past experiences were largely electronic interspersed with mechanical engineering as it was at the Frigidaire Corporation. To increase my awareness of electronic applications in other fields I subscribed to many technical journals and proceedings and discovered that in cosmic ray research, bistable electronic tube trigger pairs were used to count down the rate of arrival of cosmic rays so that the output of the cosmic ray counter could drive a slow mechanical counter. Trigger pairs were used in series, each [page 7] pair counting by two, that is it was really a series of binary counters. Wynn-Williams in England was using thyratrons instead of vacuum tubes. It occurred to me that this approach could be arranged to provide a scale of ten electronic counter instead of a binary counter.
HT: That’s right. They were getting up to about 100,000 pulses per second, weren’t they? JD: I don’t know. I don’t remember exactly how high they went. Well, anyway, .. he showed a circuit diagram which .. which became common knowledge and also an article appeared by some other writers in the Institute of Electrical Engineering journals, which gave other diagrams. Then it occurred to me that all of this business of using .. a pair of tubes as a binary counter where you had only , one — one bit at a time could trigger back and forth – that .. the thing to do is arrange a decade of them, ten of them instead of only three or four or five that they were using. HT: Mhm. JD: And, by doing that, I could count up to ten — in one counter using all tubes. JD: And a thyratron [see Wikipedia for more information about thyratrons] only requires one. But the thyratrons were big – just as big as the vacuum tubes, except I didn’t need as many. So the first machine that this notebook of mine here shows used thyratrons.
HT: This is the one dated – let’s see. This is the one. JD: That’s right. The question immediately arose as to how I could make a compact counter, an electronic counter that would be of higher speed than a mechanical machine. Mechanical machines are rather slow in this operation.. HT: Mhm. JD:.. and require a lot of time. HT: Mhm. JD: At that time, it was not necessarily economic to put a thing like that into production, but a — a model, a prototype model could be built from the [page 9] information that was available. That information was transmitted to Mr. Mumma who came to work for me– about a year later, Bob? RM: Yes, it was April, ’39. JD: It was just a year later about. And I gave him the project of creating a three stage — a three bank electronic counter.
JD: Was it? RM: And got a working model; as you see in the record. [View the model and the patent drawings here] JD: Now that model exists and we’ve used it several times in trial. Well, we used it when we had our interference with IBM and we’ve used it again now in this — in this new UNIVAC-Honeywell contest. HT: You say that’s a two bank model? JD: Well, the accumulator – I must explain an accumulator. An accumulator takes the place of an adder and a memory. HT: Mhm. JD: That is, what you put in it, stays in it. HT: Mhm. JD: And if you put some more in, it just adds to it, until you reach an overflow point in which case [page 11] you can add another bank, and then add another bank. The same thing is true in an adding machine. It’s really an accumulator and then you impulse it mechanically. Well, so that we didn’t have to have a tape, a magnetic tape or punch paper tape, although the concept appeared to us later when we actually did work on paper tape punches and magnetic tapes and so forth, just before we started on … OSRD work. HT: Mhm. JD: So, this machine – wonder if we have photographs of it around. There’s plenty of ’em around. HT: I think there are some in this stack. JD: Well, there’s one – excuse me – there’s one in this – one that has no cover that I wrote. RM: Do you want a detailed technical explanation of how it operates or what is your — ? HT: Well, .. as much detail as – say, looking at this photograph which — RM: I — I see — I see a mistake we made already; look at the photograph. It has a four bank accumulator and a three bank impulse generator. HT: A four bank accumulator so you can handle then [page 12] RM: We had overflow. HT: four digit, .. three digit input and a fourth digit for the overflow. RM: Overflow. Yes. JD: Was that a full scale of ten overflow or just a single tube? RM: I am not sure. No, it had at least five thyratrons in the 4th denominational order. I was restricted in space. At the top of the picture you will see the four stepping switch indicators. JD: Yes, that is right. RM: Telephone stepping switches were used for the indicators. They sensed for the conductor in each counting ring with the highest electrical potential. They were wired so that they would store the information from the last conducting thyratron and then turn off the power to all of the thyratrons. At the start of the next operation the switches would restore power and cause the proper digit representing thyratron to conduct. The accumulator was cleared to zero by pressing a button to restore the stepping switches to zero. HT: Mhm. RM: I failed to mention that the conduction of any [page 13] thyratron in any denominational order of the accumulator will cause any other conducting thyratron in that order of the accumulator to be extinguished, and will also cause the next higher digit representing thyratron in that order of the accumulator to be prepared to accept the next pulse which is applied to the common input pulse line connected to all of the thyratrons in that denominational order. There is a separate input pulse line for each denominational order of the accumulator. HT: Mhm. RM: This process will continue for as many pulses as are sent to the selected order of the accumulator. Five pulses will cause the first five thyratrons to fire or conduct in succession with the five representing thyratron conducting and all other thyratrons in that selected order non-conducting. The number 115811 would be transmitted by sending five pulses to the thyratrons in the tens order of the accumulator and eight pulses to the thyratrons in the units order of the accumulator. When the zero representing thyratron in a denominational order of the accumulator conducts [page 14] at any time other than during the zero clearing operation, it will cause a pulse to be sent to the input line of the next higher order causing the thyratron conduction in that order to advance one digit representing position. The counted pulses to the accumulator are sent to the highest denominational order first and to the lower orders in sequence, so that the transfer pulses when generated will always appear on an input line not receiving counted groups of pulses. This is the reason that five pulses were sent to the tens order before the eight pulses were sent to the units order of the accumulator in the above example. We could have put all of the pulses in the units order and let them transfer out to the higher denominational orders. In the above example this would have required the sending of fifty-eight pulses instead of the thirteen pulses actually used. The savings is more evident with larger numbers. For example, if we wanted to enter 500 we would press the five key in the hundreds order of the pulse generator and send five pulses to the hundreds order of the accumulator. [page 15] HT: Instead of five hundred pulses. RM: Instead of five hundred pulses. HT: Right. RM: In this system the pulse generator is controlled by a keyboard and will send groups of counted pulses sequentially by denominational order to the corresponding denominational orders of the accumulator. The multi-digit number to be added to the number in the accumulator is set up on a full keyboard, one row of nine keys for each adenoid- national order. No zero keys are required since zeros are automatic with no keys pressed in that particular denominational order. The adding operation is started by closing an electrical contact with a push button or motor bar. The adding operation is terminated when the depressed keys on the keyboard are released. HT: Mhm. RM: The groups of pulses from the pulse generator are fed into the accumulator starting with the highest denominational order. As mentioned before, this is to prevent any interference from the transfer pulse generated in the accumulator, as thyratron conduction changes from nine to zero. This pulse is sent to [page 16] the next higher order of the accumulator. You will notice that the transfer progression is in the opposite direction to the input progression. HT: Aha. RM: If you think about it you can see why that would happen. HT: Yes, yes. Mhm. RM: This is basically the way it is done. [See Patent 2,404,697 filed 21 Mar 1942; issued 23 Jul 1946, inventors Joseph R. Desch and Robert E. Mumma, assignee the National Cash Register Company, and Patent 2,595,045 filed 20 Mar 1940 and issued 29 Apr 1952, inventors Joseph R. Desch and Robert E. Mumma, assignee The National Cash Register Company. Both patents can easily be found online at Free Patents Online or at Google Patent Search] HT: Aha. JD: Very fast.
JD: .. Now, also before Mr. Mumma arrived – and this is a part that I couldn’t work in very well until just now – .. it became apparent that I had to develop a thyratron tube of small dimensions if I was going to make a compact unit,
HT: Mhm. JD: and so I set up a tube laboratory. It isn’t the one that you saw. HT: Mhm. JD: It went through a lot of phases of development over the years. .. And finally it reached that point where I could, I could make practically any tube that I wanted to make. However, did they show you a box of these little thyratrons? HT: Yeah, yeah. In fact, I saw some on a little board that you had labeled. JD: Yeah, well, that was … But, but there is a box with a whole – with a couple of hundred or so in there, and those were the best looking ones. HT: Yea, they’re about three inches high, and — JD: No, a little smaller than that. HT: A little smaller than that [page 23] RM: Of course — HT: and a little over an inch in diameter. JD: Now, later, after this, you’ll find that Mumma built another electronic counter but that was for NDRC. HT: Mhm. RM: But this — JD: But he used the small tubes then. RM: But these two original ones, these are the original ones, .. they had wires coming out of the base. In other words, they were soldered in. There were no sockets or no plugs on the base of the tubes. HT: Aha. RM: I don’t – which are the ones you saw? The — HT: I saw some — JD: He saw ones with faces. RM: with the — HT: Yeah, I also saw some with wires dangling. RM: Just a flat press on the bottom and all the wires came out of that flat press. JD: You mean small ones? RM: Little — little miniature, about this high. JD: There’s a lot of tubes over there that have wires just coming out with no bases but they are of a [page 24] different vintage. They — they were built in the forties. HT: Oh, they are later. JD: Later. HT: Yea, I — I don’t know that I saw any of the large thyratrons that were common, as you mentioned, commercially in this period.
JD: I’m surprised they haven’t shown you this first machine — RM: But that model is full of – that model still has those thyratrons in it. That we just saw. HT: And they are the size of the conventional vacuum tubes of that period. RM: They are original. HT: And I remember the size of those. [Laughter] RM: They are the original ones. HT: Well, how did you — how did you? JD: You see, I realized that that was going to be a very serious drawback in selling this if we were going to have big tubes. So the problem was to, to start up a tube laboratory and make small ones, reduce the size of the thyratron. So I hired a man named Cone. That was the first man I hired on that. [page 25] RM: Coe or Cone? JD: Cone. RM: Cone was the first man. HT: How do you spell that? JD: There was a Coe also. RM: Coe was in the electrical engineering department. JD: Yes, he was in another department. RM: But Cone’s – C-o-n-e. JD: Cone, and I don’t know where he came from anymore – but he had had experience in — in doing this kind of work. RM: Glass blowing. JD: And, so .. he — he built the first thyratrons and, of course, we had .. the usual long series of experimentation and redesigns to get the exact characteristics that Bob required in this next model that he was now building. HT: Mhm.
HT: Mhm. Mhm. JD: More so than size and cost; because they had military applications and they, they.. they, they were going to try to get the speed up as fast as they could. And, we did finally, .. after, I don’t know, a year or a year and a half, we reached over a million counts per second. And we .. RM: Well, yes. JD: I think we ought to report it as high as two million. RM: Well, the one we took to Aberdeen for this fire control was a million counts a second. We had a one .. megacycle crystal oscillator. JD: Yes, I know that one. But- HT: One megacycle crystal- ? RM: Oscillator. HT: Oscillator. That was getting up to a million pulses per second. JD: And we had a gate that time. RM: See this was done to, to measure these flight times of a projectile between two loops and we could start and stop that impulse generator at one million [page 29] times a second so every count was a microsecond. JD: picture of that – RM: You read out the accumulator, you directly had your count in microseconds between this known distance between the two loops. JD: But, finish with MIT here temporarily because I’m trying to keep this in chronological order HT: Right. JD: if I can. We worked very close with MIT during the period of 1939 and ’40. HT: Outside of Sam Caldwell, did you have any other contacts, or contacts with other people like Harold Hazen or some of the other people? JD: Not Hazen, but.. but Dr. Molnar .. and indirectly then with Dr. Harrison, because Dr. Molnar reported to Dr. Harrison. And then later on, .. when we went into OSRD work with Dr. Coleman, who — who was really the monitor of all our work, for the government, NDRC and OSRD. And – but, in any event, they were working along parallel lines but with high vacuum tubes; but there was no use for us both to be doing the same thing. So, we supported them in that work but – they never did build a complete [page 30] machine that I can remember. I don’t think they ever finished a complete machine because they were switched to war work before they had a chance to finish; whereas, .. we got switched after Bob had built that first model. HT: Mhm. JD: He built the second model but he did it on NDRC money. And, there’s a picture of that in there also. RM: Are you sure that I did the second machine, the second electronic calculator – the big – that great big electronic calculator on NDRC money? I thought that we did that on NCR money. JD: I am talking about the calculator we sent to NDRC. RM: Apparently we are not talking about the same machine. JD: I am sure that you built the small calculator we sent to NDRC before you built that big calculator. RM: Yes, I remember, it was an improved model of the first electronic calculator, using the small NCR made unbased thyratrons. It had a three denominational order keyboard, which controlled the pulse generator, which fed counted pulses into a three denominational order adding accumulator, complete [page 31] with an electro-mechanical read-out indicator. JD: Yes, that was the machine. This was a straddling period between our commercial work and NDRC work. It was hard to keep the two separate. HT: Mhm. RM: The machine that I referred to above as the second machine was actually the third machine or electronic calculator. JD: We were trying desperately to finish this machine so we left a man or two on it, RM: Yes, I was taken off of it at the time. JD: even though we were going on to NDRC work. RM: See, Larry Killheffer finished it up. I was taken off of it. HT: Larry? RM: Killheffer. JD: Killheffer also worked on these things. Now here is the impulse generator of the second model he built and you’ll see the small tubes in this one. HT: Mhm. RM: That’s the one went to Aberdeen. JD: No. No. This is not the one. RM: No? JD: This is the one that went to NDRC. [page 32] RM: Oh, I see. OK. JD: But – and here was the receiver with an indicator on it. We had a motor RM: Yeah. JD: driven indicator as you can see. And this was — this was demonstrated down at MIT.
JD: Yeah. That’s the one — that’s the one that went to University of Chicago. HT: Yeah, Fermi’s project. JD: Yeah. .. Here, we come to the one for the University of Chicago, Metallurgical Laboratory – there was the inside RM: Inside of that. JD: of that particular one, see. And look it, you see all the small tubes. HT: Mhm. RM: And then they’re all thyratrons that are in there. These thyratrons were NCR made with plastic bases.
JD: This was a – RM: That was a test. [page 33] JD: That was a test device. Killheffer was also working on, at that time – this was done by Killheffer. This is a binary decimal converter. .. We kept this one man working for us instead of OSRD at this cross-over point and he — he developed this binary-decimal converter. And, these are the three pictures of it. RM: There’s the next one. JD: Now, here’s the one that went to Aberdeen Proving Ground. And here is the receiver and you’re gonna see the insides of that in another picture. There now there is the inside of that receiver with its indicator. HT: Mhm. JD: And it would count up to six digits HT: Mhm. JD: and here was the receiver. Now, at this point – here’s a good point to bring it in – this back in here is a .. counter from the University of Chicago type. HT: Mhm. RM:. Well, there was a recycling binary counter and it used 6AG7 tubes that counted a million pulses per second. We used this counter to divide by ten to [page 34] drop the pulse rate down to 100,000 per second for these gas tube counters. JD: There’s where we scaled down by a factor of ten to one. HT: I see, aha. RM: We couldn’t go any faster than a hundred fifty thousand pulses per second with thyratrons. JD: Now, here’s what — here’s what happened at the University of Chicago. .. Coleman, who came out here frequently to — to monitor our NDRC work, which was now starting to build up in the area. He was on it and other people were on different things. .. He came out and said he wanted me to go to Chicago to take a look at their work. It was a Dr. Wilson that I had to see. And we found – we saw the work they did. They had built a -a counter that was reputed to be a million cycles per second or something like that; but Coleman, for some reason or other, was a little worried about it. And, the net result was, and I didn’t promote this, but the net result was that he — he tranferred that project here, and we did a lot of work — you’ll find in the notebooks of all the work that Bob did on perfecting the University of Chicago [page 35] counter until it did reach over a million. It went over a million. RM: I think that represents the facts. HT: I should mention here for the record that we are going through, Mr. Desch is going through, a document of his that has no cover on it that was originally labled SECRET but is now unclassified and dated October 24, 1945. And in this document are a series of photographs of these various machines that he’s describing in sequence as he goes through. JD: So that, in this device, which was the receiver that we – Bob and I – took down to Aberdeen Proving Ground, .. which counted accurately to .. .. one megacycle, but it would go higher, and which upon pressing a button you could read out just how many microseconds had passed, because you were in microseconds. And, I — I think this one is the RM: That’s just opened up, on .. the underside. JD: impulse counter. No, this is the impulse counter. Oh, yeah, that’s right. You’re right. Here — here — RM: I think –
JD: Here is the resetting binary counter. [page 36] HT: Oh yes, that’s the one that’s on the back bank … JD: That’s right. And, here is the underside of the one megacycle resetting binary counter. Now, this thing I’m talking from was written by me to Mr. Williams after the war to acquaint him where we stood HT: Mhm. JD: in the electronic field because a lot of this he didn’t get to see during the war. HT: I see. RM: That’s a gate worth mentioning. JD: Yes, it is. “Auxiliary one megacycle gate for interval measurements with the one megacycle counter delivered to Aberdeen Proving Ground.” RM: See, we had to start and stop the count accurately plus or minus a pulse. HT: Mhm. JD: And that — this is quite a job when you have a constant running oscillator, the magacycle to break it off in one cycle and not let it run over the one cycle. Here was the impulse generator for the – for this combination. RM: That was simply the test accumulator really. Well, yeah – okay and then it .. [page 37] JD: And, here’s the inside of it. RM: You see, that also had a million counter in there, too. JD: Yeah. RM: It had to. JD: Mhm. RM: That’s a [inaudible] one. JD: Now, this one was the signal identification. RM: Oh yea. JD: system that we built, which really was an IF system that they wanted us to hack out real quick and get over to England; so we built that one. Then we built a radio transmitter.
JD: It will hold one to ten. We also had one to five. HT: These were built in the lab that I saw? [page 40] JD: Mhm. RM: Well, those were – HT: Here, what people now call the “Secret” or “Lost” Lab. JD: Well, the lab wasn’t there when these were built. The lab was in this building. RM: Third floor of this building when those were built. Third floor of this building. Just moved over there. JD: See, when I first came here, the first lab I had was on the tenth floor and we moved down to the third floor. And then we moved out — HT: [Chuckle] JD: to Building 26, HT: Aha. JD: because of the NDRC work, and so on. HT: Aha. And this is Building 10. RM: This here is Building 10, yeah. JD: Here is another radio transmitter for the remote control project for the Army Signal Corps. [Contract #W2279sc-203] Now, this one was built, and you can set up combinations here, to trip land mines
HT: Well, let’s see. It says “Experimental three bank counter submitted to Navy for tests and approval before constructing counter printer equipment.” JD: Right. So, you’ll see a picture here of a counter — no, you won’t. I think that they wouldn’t allow me to use it, because it – no, they wouldn’t. HT: Still classified at this period in time. JD: Yes, it was. That was the one we gold-plated, remember? RM: Yes. JD: We had five of these in a rack and I had it gold-plated. A “Test impulser was built for the Navy to test the accuracy of their counter printers.” [page 43] HT: Mhm. JD: Because we have come to the end of OSRD work where we’re going to skip four or five years. HT: Right. Right. I realize that. JD: And this was the thing that did it. HT: That test counter. JD: Well, that and the counter itself. At this point it might be well to explain the gap of four or five years referred to above. The counters built for OSRD were acquired by the Navy and the Navy entered directly into a contract with NRC for a few more. More and more contracts between NCR and the Navy resulted and finally all of our technical facilities were in use. The Navy even supplied additional technical personnel. All of the work was highly classified and still is and the end result was that the only electronic work during the rest of the war period done by NCR was for the Navy. The work …. terminated in 1946. Here is a problem, it occurred to me that this business of binary counters or decimal counters didn’t use the optimal amount of equipment, so I developed a theory here .. notice that it was pretty late. It was in 1945. I developed a theory that Go to Second session of interview