An Engineer’s Memoir of Teletype Corporation

It’s a little odd that I had not given any thought to working in the computer field at the time I graduated from college. At that time I had thought that computers involved some interesting technology; but I wasn’t interested in working on things that sat in glass-enclosed rooms and received all their input from punched cards. It was only when computers began to be used in online applications that they attracted my interest. I also was well aware that stored-program logic and magnetic storage were going to displace wired logic and punched paper tape in all kinds of systems applications.

I remember seeing at Teletype a Model 33 printer with a cash drawer underneath. It might have had a 10-key keyboard. There was talk of getting into the point-of-sale terminal market. In view of the consent decree restrictions I suppose the concept was for the telephone company to offer point-of-sale terminals and communication with the customer’s computer somewhere else.

I hit the road for Phoenix two weeks later, in the early summer of 1966, leaving behind one unfinished project. Teletype was getting interested in numericallycontrolled machining. There was a desire to connect a DRPE punch to one of the 1401 computers so that we could generate our own tapes to control the machines. I was asked to do the connection. IBM offered a customer-hardware interface to their machine as an option. They charged something like $150 a month for this feature. Perhaps in retrospect that would have been the most reasonable way to go; but I decided the price was too high and that I would instead make the DRPE connection emulate a magnetic tape drive. The tape drive interface was available at no extra cost, as tape drives were simply daisy-chained off the machine. This was not a totally harebrained scheme; some other company produced a successful time-of-day clock for IBM products that connected to the tape drive chain. Most of the tape drive logic was done inside the computer. The drives themselves were pretty simple, having only read and write amplifiers and the tape motion signals. One complexity of writing to a tape was that the drives had read-after-write heads and sent read signals back to the computer during writing. I don’t think these had to be the same characters that were written; I think they had only to have correct vertical parity. I was going to use the logic circuits I had experimented with earlier: high speed transistors on standard Teletype circuit cards. I don’t know whether the project ever got completed. About the time I was leaving Teletype had taken delivery of a Honeywell 200 computer, a machine that was directly targeted at IBMs installed base of 1401s. Perhaps by the time the DRPE project

74

had been completed there would not have been a 1401 to connect it to. From this point on I will leave out a lot of personal history, confining the account mostly to things that involve Teletype.

75

V. Go West, young man

entire technical staff. A new idea or interesting observation, mentioned casually during the morning coffee break, would be known throughout the R&D organization by noon. At G.E. the cafeterias operated only at meal times. There were coffee and soda vending machines throughout the plant and office areas. The engineers got their refreshments from the machines and returned to their desks to consume them silently. There was practically no inter-cubicle conversation. No doubt the vending machines were a great dollars-and-cents saving over keeping the cafeteria open; but I am convinced that any savings were far outweighed by the loss in informal communication. Another problem with the cubicles was that a talkative co-worker had a captive audience to distract from their work.

Soon after I went to work for G.E. I ran into Dennis Bobka, who had also left Teletype. G.E. at the time had an aging but highly regarded data communications processor, the Datanet-30. Dennis worked in a data communications group, where his expertise with Teletype’s customer applications was quite valuable. Later I think he went across town to work for Motorola.

G.E. at the time seemed to have the best knowledge in the industry of data communication and data flow in the enterprise. Perhaps this was one area where the high status of management was a help rather than a hindrance; professional managers knew what they wanted from a data system. IBM, it seemed to me, had a much too computer-centric view of the world. They regarded a terminal keyboard as a particularly unruly species of card reader. Labor relations at G.E. were strange. The computer department in Phoenix was a non-union plant. Phoenix natives tended to be anti-union. Arizona is one of the sparsely-populated not-very-industrial states. The natives tended to think of unions as those things they have back East in the steel mills, and in New York City where graft and corruption are endemic, and in California where the union workers are all communists. They preferred for themselves the Western lone cowboy ethos, and voted for Barry Goldwater. Other G.E. workers had come to Phoenix from G.E. plants elsewhere in the country, and were generally fed up with all of the petty nonsense that the unions there habitually engaged in. One of my fellow workers, a technician from Canada who was something of a conservative provocateur anyway, told of a trip he had made to see about something at a G.E. plant in Syracuse. Within fifteen minutes of his arrival he had been the cause of a union grievance. Whatever he was going to do required an oscilloscope. As there was not one in the room he walked across the hall and found one not in use and rolled it to where he needed it. The union contract required that only union technicians move test equipment between rooms.

Still it seemed that the G.E. management in Phoenix did not appreciate the

78

anti-union sentiment in the plant, nor attempt to foster good labor-management relations. Perhaps the management training was geared toward dealing with the militant unions militantly, and didn’t allow for the possibility of a labor force willing to cooperate with management. Or perhaps it was that in the absence of a union management couldn’t figure out a mechanism for two-way communication with the labor force, there being no persons formally designated to represent the workers in dealing with management. Whatever the reason, time after time management made arbitrary, unilateral decisions with no apparent concern for the feelings of the workers. Each time I was surprised that the workers did not flock to the representatives of the big electrical unions, who were always hovering around, and demand union representation.

Another personnel phenomenon at G.E. was that the first-level managers of engineers were extremely retentive of their employees. If some other manager wanted you to work for his unit, and if you wanted to do so, it was practically impossible to get your own manager to let you go. It was said, and perhaps not jokingly, that it was easier to quit and get rehired than to get a transfer from one unit to another. I didn’t try it.

My first engineering assignment at G.E. sounds like a half-day Teletype project: design a transmitting distributor to send several characters of ASCII to aModel 35. It took several months to complete. First I had to learn the G.E. logic circuit set. This was a bit strange. There were logic-level circuits like NANDs and flipflops. Then there were pulse circuits using blocking oscillators and pulse transformers. To trigger flipflops one had to use these pulse circuits; it was not permitted to use the edge of a level pulse to trigger flipflops. So my initial design was turned down by the more senior engineer in the cubicle, although it probably would have worked just fine. After getting a satisfactory logic design I had to go through design automation. Everything possible was wired by automatic Gardner-Denver machines; so it was necessary to put logic diagram information into a form that could be processed through several stages and eventually produce a deck of cards for the wiring machine. I was somewhat familiar with IBMs design automation, which allowed for logic signal names long enough to be self-explanatory. The G.E. system then in use allowed only ten characters for the logic name, and then required a four-character suffix composed according to some arcane rules that I never fully understood. I screamed that this was intolerable. It turned out that there was a new system then under development that would allow considerably longer logic names. Our cubicle had the opportunity to be the first to use the new system, so we decided to do it.

My circuit was to be a part of a larger item; and G.E. didn’t have the ability to

79

around on some rollers inside the front door of the machine. The machine had a watchdog timer that would expire unless the software frequently executed an instruction to restart it. If the watchdog timer expired the machine would automatically reload its software from the paper tape.

I’ll include a story about G.E. magnetic tape drives, just because it is funny. IBM and everybody else in the industry made tape drives with vacuum column buffers so the tape could be accelerated very rapidly at the heads in spite of the inertia of the reels. G.E. had bought the rights to a patent for a different buffer technology, using scramble bins rather similar to the bins Teletype used on RT stands to store paper tape. The amount of tape in a bin could be measured approximately by shining lights at the edges and measuring the light striking solar cells at the back of the bins. I’m not sure why G.E. did this; perhaps to get around IBMs patent on the vacuum columns. The scramble bins were much smaller than vacuum columns, but the tape drives as a whole were as large as vacuum column drives. This worked just fine mechanically; the tape could be accelerated rapidly and the servo system usually kept the right amount of tape in the bins. What nobody had thought of was that the tape in the bins rubs against itself, surface to surface. This continual rubbing tends to demagnetize the tape; so we had the phenomenon that a freshly written tape could be read easily, but as the tape was read repeatedly the recording would deteriorate until it could no longer be read successfully. G.E. eventually had to replace all of those tape drives in the field, at great expense.

I remember a visit by Ernie Kettnich and somebody else fromTeletype. Maybe they were trying to sell G.E. on theModel 37. G.E. used IBM Selectric typewriters as console terminals on the computers and had reliability problems with them. I guess with the fairly low volume of computer sales, and only needing one console terminal per computer, it was considered too expensive to redesign the console to accomodate a different kind of terminal. I’m pretty sure I remember seeing a G.E. video terminal with a Model 33 keyboard; but I don’t know if it was a production item or a laboratory prototype.

Although I had seen integrated circuits before leaving Teletype, the G.E. equipment I worked on was still discrete component technology. We did build a couple of integrated circuit modification kits for one of the computer models. It was fortunate that the machines in question used NPN silicon transistors, so we could connect directly to integrated circuits and intermix them with the rest of the logic. G.E. went through a tortuous process of evaluating competing integrated circuits and managed to standardize on the wrong family: the SUHL line made by Sylvania and Raytheon. Soon those companies dropped out of the I.C. business while

81

Texas Instruments dominated.

At the time there were some fairly cheap plastic cased Motorola RTL integrated circuits on the market. I could afford enough of these to undertake a project I had been wanting to do for some time. Back in the ’50s I realized that synchronous transmission offered advantages over start-stop for radioteletype operation. The worst thing about start-stop on the radio is that with a mechanical selector a mutilated stop pulse allows the receiving shaft to keep going around, without latching up the clutch. Typically several characters are printed in error before the shaft gets back to where it is ready to stop for the stop pulse. I have seen a figure that this problem costs something like 15db of signal/noise ratio in character error rate versus bit error rate. With synchronous transmission there is the problem of remaining in synchronism, but not this problem that one bad bit causes several character errors. The second advantage of synchronous operation is that the detector knows where the bit boundaries are. Hence it can accumulate all the energy in a signal pulse, rather than point selecting in the middle of the pulse as a Teletype selector does. Frank Biggam and I had talked about this in connection with the multiplex equipment, it being synchronous. We couldn’t do anything about it because we were being handed a signal that had already gone through a detector; to take advantage of it we would have had to build the FSK or PSK demodulator into the multiplex equipment. Collins Radio had a frequency division multiplex product, Kineplex, which did take advantage of synchronous detection. Teletype had made, in AN/FGC-5 days, a thing called a monoplex. It used multiplex parts to achieve a single-channel synchronous transmission. I never found any detailed information about how well it had worked in comparison with startstop. At the time I didn’t realize it had become a product; but Bob Reek tells me it was used for DEW Line communication over scatter circuits. I believe he said it was sufficiently promising that a synchronous two-channel multiplex was also operable over those circuits.

Thus I had long wanted to try synchronous radioteletype transmission in ham radio, but could not before then because I would have had to build a lot of equipment and the FCC rules required start-stop transmission anyway. After a while (OK, I admit I can be pretty slow sometimes) I realized that it would be possible to transmit a synchronous signal compatible with a start-stop receiver. It was just a matter of sending 7.0 unit or 7.5 unit code at an absolutely regular rate, putting in an idle character if there was no message character available for transmission. 7.0 unit code at 45 baud can be received without difficulty on a mechanical teleprinter and is easy to send synchronously and is fast enough to stay ahead of the keyboard or tape reader. Now with cheap integrated circuits I could easily build a start-stop

82

receiving distributor feeding into a synchronous transmitting distributor. I could use blank as an idle character, or LTRS or FIGS as appropriate. At the receiving end a similar circuit could delete the idle characters, just to keep the machine a little quieter. I built the transmitting circuit and it worked fine, except that in those days we didn’t have much in the way of I.C. breadboarding equipment so it was hard to keep wires from falling off. Then when I tried to transmit on the air with it the RF ran all through the circuit and paralyzed the logic. These problems could have been solved, but I put the project aside as I was doing other things at the time.

I learned a lot about computers at G.E. but I was never very happy working there. I just didn’t have any confidence that the management was leading us in a direction that would be successful. A good telling of the misadventures of G.E. in the computer business is now available in a book, “King of the Seven Dwarfs” by Homer Oldfield, published by IEEE Computer Society. I decided that if I worked there for two years I would feel that I had paid my dues to the company for hiring me and moving me there and could leave in good conscience. I was also starting to think about a non-industry job where I could perhaps pursue some of my own ideas about computer design without the constraints of an existing product line. I made an interview trip to the San Francisco Bay Area, getting in a visit with Bob Weitbrecht in the process. I decided to call someone at Stanford University to see if there was any work there. Whomever I wanted to talk to wasn’t in, so I decided to try Harry Huskey at Berkeley. I knew Huskey only by reputation; he had been in the computer industry since ENIAC, had worked in England for a time, had built the SWAC computer at UCLA, and then headed Project Genie at Berkeley, which was sort of aWest Coast version of Project Mac at MIT. He was also editor of the IEEE Computer Group publications.

My call to Berkeley revealed that Dr. Huskey was no longer there; he had moved to a new U.C. campus in Santa Cruz. I didn’t know such a campus existed; but it sounded like a nicer environment than Berkeley for all the latter’s eminence. I reached him by phone at Santa Cruz; he told me that they were just getting started there and didn’t have any work for an engineer right then, but to call him back in a few months to see how things were coming along. I returned to Phoenix and waited a few months. One day I had a particularly disgusting morning at G.E. and went home at lunch time and called Dr. Huskey. He said he was ready to hire someone, and that he had to make a trip to San Diego and could go by way of Phoenix and meet me at the airport for an interview. This must have gone well, as he offered me a job a few days later and I accepted. This was in the fall of 1968.

During the last week I worked at G.E. I was shown a prototype of a new prod-

83

uct, the Termi-Net 300 terminal. This was not a Phoenix production; it came out of a communication products department in Virginia. It was a beautiful machine, a little larger and lower than a typewriter. The printing principle was a rubber belt carrying metal type fingers, each with a piece of type on the free end. There was a hammer for each print position. Thus it was the same in principle as the IBM 1403 line printer. No doubt there was a lot of electronics to make it work. They told me the production models would be made with CMOS and that it would all go onto a surprisingly small number of chips. The keyboard used some magnetic principle and had a very light touch. The speed was 30 characters per second. Mentally comparing this with the Model 37 I had thoughts similar to the infamous proclamation of Lincoln Steffens: “I have seen the future, and it works.” A few years later there would be a 120 character-per-second version. It was pretty amazing to me that a company like G.E. could and would produce such an innovative product. There was nothing else in the entire company’s product line quite like it, so far as I know. The company had no presence at all in office machines. I’m not aware that in Phoenix there had been any discussion of CMOS and the ability to make large custom integrated circuits. After the Termi-Net 1200 G.E. apparently lost interest in making terminals. By then there were several companies making daisywheel printers for formed-character printing and others making dot-matrix printers.

84

VI. Back to the halls of ivy (a little) and concrete (a lot)

I arrived in Santa Cruz near the end of September 1968. Dr. Huskey had a couple of barns on his property, so I had a place to store all my stuff while looking for a place to live. I moved in temporarily with Bob Weitbrecht; his home was about an hour’s drive from the Santa Cruz campus.

The faculty of Information and Computer Science (I & CS) was headed by David Huffman, of Huffman coding and sequential circuit synthesis fame, who had come from MIT. The other faculty members were Dr. Huskey, who worked half time for the computer center and half time as a faculty member; and Bill McKeeman, who had just come from Stanford. Bill had given a talk at a recent joint computer conference which impressed us mightily at G.E. when we read it in the proceedings. I&CS was to have been the first board of studies in a new school of engineering at UCSC. Then a high-level committee declared there were already enough engineering schools in that part of the state and that it was not a good time to start a new one. This was something of a disappointment for me, as I had wanted to pursue graduate study in engineering. I&CS remained as a viable program under the Division of Natural Sciences. I met the man who had been hired to head the engineering school shortly before he left.

end of excerpt of book