Schematics of Atanasoff-Berry Computer logic circuits?

[William Blair]

> > A multi-stage direct coupled vacuum tube circuit.  Fun.

> Yes, be prepared for some fun, in both senses of the word: it is fun to play
> around with, but the implemention of tube logic can be problematic or
> unreliable, at least in the way the ABC tried to implement both NAND and NOR
> gates with resistive input circuitry.
>
> Atanasoff makes it sounds easy in his paper, but if one reads it closely
> it's not quite so, at least as measured by modern standards where one comes
> up with a gate design and then simply repeats it ad infinitum.

After reviewing the ASM circuit on your web page, it became readily apparent to me that it would likely be a major challenge to fill in the blanks for those missing resistor values and end up with something that actually worked reliably.

> The ABC reconstruction and the original required (at least some) hand-picked
> resistors in the gate circuits.

I bought the excellent book “The First Computers – History and Architectures” mainly because of its content on Konrad Zuse’s relay-based machines.  However, in its included paper on the reproduction of the ABC, the authors state:

“In building add-subtract modules, we found the circuits very demanding of precise resistor values.  We have evidence that Berry hand-selected resistors from bins until he found ones that worked, and we attempted the same tactic.  In measuring the characteristics of 10% resistors, we discovered the distribution about the nominal value shown in Fig. 3. (It’s a graph of a bell curve between -10% and 10% with a big notch in the center of the curve). Apparently, the manufacturer had already segregated the resistors close to nominal value.  Hence, we found it necessary to use 1% tolerance resistors.”

So, here’s a 14 gate, multistage, direct coupled vacuum tube circuit that apparently requires various hand-selected values of 1% resistors to work properly.  Whew!!!  I think I’m going to start with something a great deal simpler, like a single flip flop or astable multivibrator using an inexpensive 9-pin miniature dual triode at low voltages to see if I can even get that to work.

> Some of the issues:
>
> 1. Tubes can be insufficiently non-linear. Driving between saturation and cutoff
>     can take a bit of swing and saturation is soft (curved).
>
> 2. Plate circuit impedance is high. If the difference between the plate circuit
>     impedance (lower better) and the grid circuit impedance (higher better) is
>     insufficient, then different fan-outs in the logic circuit (loading), and
>     varying tube/component characteristics, can pull the logic levels away from
>     the design targets and upset everything.
>
> Going to the low plate voltage is a nice idea from a practical view but I
> wonder if it may compound issues of point 2 above. It will reduce the voltage
> shift between the plate and grid circuits that needs to be accomplished, but it
> may also reduce the voltage swing between the 0/1 logic levels. How it works
> out in the balance will be interesting.

Thanks for that info and it will be interesting indeed to see if this will work.  I was surprised to find nothing on low voltage tube logic when there’s so much on the web about low-voltage tube audio circuits of all kinds.  True, tube logic is a fairly esoteric subject with no usefulness beyond that of a novelty, but with all of the audio low voltage stuff, I’d expected some tube enthusiast somewhere to have already done something on the logic side just for the heck of it.

> Straying completely away from technical issues, and I'm somewhat loath to
> mention this, but the animosity engendered by the early-70s court battle
> continues decades later. The somewhat nasty inter-personal battle
> found it's way into the Amazon book reviews as recently as 2004.
> Another bizarre twist in the ENIAC patent saga on the legal/social side.

Yes, I saw that when I went to look at the Amazon entry for Burks' "The First Electronic Computer." Wow, it’s a long-time, knock-down feud for certain.

Bill