Reverse-engineering old aerospace/military computer peripheral boards

[D Straney]

Here's some interesting mystery boards I found being sold in a batch as gold scrap (I've got a bunch of extras too I'm selling at-cost, if anyone wants one).



First clues are the late-70's date codes: 1977 and 1978.  The "IBM FSD" label on the back stands for the IBM Federal Systems Division, which was responsible for the System/4 Pi aerospace computer series which was used in the Space Shuttle (the famous AP-101), the A-7 aircraft, and others.  (I also have an interesting-looking power supply from the A-7's System/4 Pi tactical computer, but that's for another post once I have time to take good photos and trace the circuitry)

We can safely say that these cards aren't directly from one of the System/4 Pi computers themselves: the form factor is all wrong.  See some great photos here, here, and here.  Rich Katz (klabs.org) has a couple other Space Shuttle computer photos though where you can see the resemblance in the design, especially in the first one: photo 1 and photo 2.

So what are these from, exactly?  I couldn't find a definitive list of IBM Federal Systems Division products or projects, but nothing besides the System/4 Pi series showed up as any kind of brag-worthy milestone from IBM.  There's a decent chance, then, that these are from some kind of peripheral or related electronics unit that's meant to work with one of the System/4 Pi computers.  There's supposedly a separate I/O unit, which I couldn't find any more information about (disappointing as the "TRANSLATE RECEIVER" label implies digital I/O), and I'm sure there must've been some application-specific external peripheral boxes too.  Let me know if you have any more info on that front.

Besides the "TRANSLATE RECEIVER" label, the edges are labeled "MDQ", which doesn't help much either:


Now for the much more fun part though, which is to look at the actual circuitry.  Each card has 2 gold hybrid modules with identical part numbers, and two tantalum caps, which are connected across the power rails (which you can see as copper pours on the back side of the board).  Searching for the hybrid's or the card's part number turned up no matches or info on parts supplier websites, so I cut open one of the hybrids on the one card where the metal was bent up anyways:



With a mediocre-quality microscope image from the friendly local makerspace (got to try and tune the settings next time I visit), there's enough info for tracing the circuit inside each hybrid:


I can't tell what the ICs are without getting much higher zoom levels, and then matching the images against die shots posted by Zeptobars / Silicon Insider / Evilmonkeyz / etc. etc. but at least both dies are identical and connected in identical circuits.  Based on the vague functional description and the connections, my best guess is that they're comparators, which would make sense for a receiver/level-shifter for digital inputs.

The section that looks like a pull-up resistor on an open-collector output (just like the popular LM311/LMx39/LM290x comparators) is what makes me fairly confident that these are actually comparators and that the "TRANSLATE RECEIVER" label doesn't refer to analog scaling & level-shifting instead.  It's possible that what I've labeled "VCCO" is actually a negative power supply, the ICs are actually op-amps, and the pull-up resistor is actually a pull-down resistor to let the output reach closer to the negative rail, but then the design seems weirdly asymmetric for no reason and some other things don't make sense (like why the op-amp would have a separate ground connection).

One rare very straightforward section here is the internal voltage regulator for one of the power supplies, in the top-left corner of the schematic.

Each hybrid might represent a receiver for a differential digital input.  There's some weirdness around common-mode vs. differential input scaling: the resistors to ground (R6 & R10) scale the absolute input voltage (both differential and common-mode) while the resistors between inputs (R7 & R9) only scale the differential input voltage while leaving the common-mode input voltage intact.  The common-mode-midpoint voltage (at pin 6) gets restricted to positive voltages by D1, along with a fixed resistance (R8) which pulls the common-mode towards an unknown external voltage (pin 7) that's common to both hybrids but is not externally connected to any of the other power rails on the board.

The strangest part about the common-mode circuitry is the section outlined on the schematic, with Q1, R17, R13, C1, and C2.  I'm not 100% confident about Q1's type & orientation but it makes the most sense based on diode tests and physical layout.  What seems to happen here is that as long as the input signals at pins 16 and 14 are purely differential, their AC sum (at Q1's base) stays at zero and Q1 is off, but as soon as there's a common-mode transient, Q1 turns on and clamps the common-mode-midpoint in the resistor network to the unknown external voltage.  This might be a way to maintaining a known output state in the presence of interference? (large common-mode voltage spikes induced by noisy external equipment, lightning, etc.)  It likely wouldn't be for protection, as there's still R7 & R9 in series, unless those resistances are very small compared to R5 & R11 (a large scaling ratio for the voltage dividers).  Would love to hear any ideas about the circuit details.