Return to Components/Devices

Rick Dill's comments on Circuit Sets

Comments on Ken Shirriff's
Germanium transistors: logic circuits in the IBM 1401 computer

March 3, 2021

Well, you have managed to pick out the core of Joe Logue's group ... i.e. the best of them. I was a summer student in 1954 and these were the key to building IBM's circuit families. Jim Walsh and Bob Henle were more visible to me, although I came later to almost revere Yourke was less visible to me, but had my admiration for the current switch (i.e. ECL) circuit which was the core for IBM's fastest machines.

Jim Walsh became the leader of the circuits group in later years and was pretty much responsible (at a review level at the least) for the circuits that IBM used. Henle was involved in that, but he had a core role in moving IBM into semiconductor memory. Both are on the now long list of IBM Fellows .. something I have come to take pride in not being one in spite of my early belief that it was the way to a great career. That turned out to be right for me, but certainly not for them.

The circuit families were staged by cost and performance. My memory might not be great, but I think the sequence was:

• Resistor Transistor logic (RTL) at the bottom, low in speed and cost
• Diode Transistor (DTL) next where the bulk of the logic was done with diodes and transistors kept the levels OK and the logic simple
• Above that was Transistor Transistor logic (TTL), more costly and faster
  (These all were slowed by charge storage in the transistors.)
• Current Switch (ECL) at the top, by far the fastest. It also needed the most power but the transistors never went into saturation.

That line-up went until the last bipolar based computers, which was a lot of orders of magnitude in speed and performance.

I was starting my second year in grad school when Bob Henle showed up to try use support to lure some students they knew about. It was the result of Dale Critchlow (a brand new grad and faculty member) going to IBM the year after that. His thesis was in magnetics. His student (Bob Dennard) also was doing magnetics for a thesis. They wrote contracts for all of us and it doubled my income as a graduate student to $200 a month. We all ended up at IBM in 1958 or so and ended up working together when I thought I wanted to look at something beyond just semiconductors. We were in a group using multi-aperture cores a transistors to do logic. It was fun and a place to have friendly competition about how many transistors and cores were needed for a function. The 1800 was just coming out and we asked management to let us get us permission to build a 1401 with our technology. That didn't work and I turned back to semiconductors ... first tunnel diodes, then watching Von Neuman's parametric logic using diodes for the voltage variable capacitors. I was skeptical ob both that and tunnel diode logic, but involved within both thrusts. There were some really neat concepts thought of in them. This was also about the time of IBM's attempt to make circuits and hopefully computers with cryotron circuits. In the background was a group of physicists who thought that using electro-luminescent/photoconductor circuits which they could simply print or silk screen. It would never have happened if it worked and it was basically unworkable at anything short of multi-second switching speeds.

[An idea that comes up is that maybe once we got semiconductor light emitters and lasers, we might have revisited that technology]. That would not result in anything useful at the physical size needed for competition today.

Yes IBM patented lots of things.

It's late I will sign off.

Rick