This is a continuation of an effort started a few years back. I had purchased one of these off eBay, but found that only every other bit would get stored. The troubleshooting steps in the manual indicated that this meant either a bad A3 buffer or A4 controller card. Although the schematics are included, there are no board-level troubleshooting instructions, or detailed theory of operation. Furthermore, HP says to not even return these boards for repair, simply discard them and they will send replacements. I because they’re made up of inexpensive 74LSxx series chips, even back in the late 70’s when the device was built, it would not be economically viable to have a technician spend their time diagnosing and repairing these boards. Yet that’s what I’m going to attempt.

When I first dug into this, I soldered dozens of little jumpers of magnet wire to various pins so I could probe the inner workings while the card was inserted. This got me nowhere, and so I shelved the project.

A year or two later, I saw another one come up on eBay, and one it for a surprisingly reasonable sum. It was listed “as-is, for parts”, but I figured I’d take the risk. Although in worse physical shape, this unit actually was fully functional. So now armed with a set of working cards, I was able to troubleshoot the bad unit.

Swapping cards, I was able to confirm that A3 was the problem. Fortunately, this card is accessible in place if you remove the HPIB (GPIB) card, which will not hinder operation. This finally let me probe the device in action.

The A3 board is essentially the edit buffer. Bits are read in, displayed on the front panel, edited, and then placed back into memory. The read & write is done serially, so there’s a shift register made up of 7474 flip-flops, and 7451 And & Nor gates that allow for the bits to be modified. The circuit is split up into even & odd sections, each with their own clock, serial in & out lines, and a few other control signals. (note the odd symbol for a NOR gate in the schematic below).

I first suspected the the flip-flops, so I unsoldered all of them, and installed sockets. I tested them in a little chinesium chip-tester, and one of them was bad, but the problem remained. I turned my attention to the and / nor gates, which unfortunately my little tester didn’t support. I wired up my 16500 Logic Analyzer to the output of each of those gates, and found that U32 was glitching. After replacing that I had what seemed like two working units.

Maybe this happened during repair, or maybe it was an issue that I didn’t catch at first, but now on the second unit (the one I thought was perfect), The 16th bit now mirrors the 2nd bit, on all words. This problem seems to follow card A4, which is not accessible when installed. It’s also a more complex circuit.

I flipped the unit on it’s side, so I could start probing the edge connectors. On the analyzer, I’ll designate the last pin on the pod as PROBE, so I can move it around looking for interesting signals. When I find one, I’ll add a grabber to the next free line on the pod, and add it on the trace. I have it in continuous capture mode, to trigger on the clock signal (which only runs during a load or fetch). I turned my attention to J1-8 & J1-5, which are odd & even lines out to the RAM card. I expected them to look similar, but they don’t.

The odd line, which is working, seems to be in phase with it’s clock (this is showing no bits active), while the even one looks different.

I expected to see a similar pattern as the odd line on both lines of the good unit. I was wrong.

It looks totally different. Not shown here, but when there are bits selected on the bad unit, I’m seeing short glitch pulses on both the odd & even lines. Those are not evident on the working unit.

So without being able to dig deeper into the circuit while it’s working, I’m at a bit of a loss. For the time being, I’m going to be content with one working unit and one parts unit.

I decided the next step would be to build an extender card so I can freely probe both good & bad units while in operation. The edge connectors are 3.96mm pitch (?!), and I found 4 on eBay at a reasonable sum. They’re right angle connectors, but I’ll make ’em work. The bigger annoyance is there are jumper cables between cards along the top, so I both need to make an extender for those to connect to card A4, and I need to accommodate cables that need to pass across A4, without having to extend a bunch of other jumpers. I think I’ll design this card with a large enough hole in it. It’ll be mildly annoying, and I think a job left to Future Paul (screw that guy…).