Stuck bits on 11/73 Clearpoint 4MB memory - how to repair?

[Noel Chiappa]

    > From: Jacob Ritorto

    > I struggled for hours with inadequate eyesight, tools and materials,
    > but I think I got this mod done!

That was quick! I had an 11/23 on my workbench that had had that backplane
mod, so I took it apart to take a picture, to help you, but I guess that's
OBE... :-)

    > the machine's stuck in self-test at error 47, Memory CSR error.

It's not clear that that's really a hardware problem. DEC makes assumptions
about how many memory CSRs there will be, etc, and some third-party board
'fail' when they don't meet those expectations. I have a Clearpoint QED1 (I
think that's the one, didn't check) that does that.

    > What I assume is the parity light on the third party "Clearpoint QRAM-2
    > SPB-1 88B" lights during self test. I've found no documentation for this
    > sucker as yet..

I have a PDF of the "User Information Manual" for the Clearpoint Q-RAM 88B.
(Oh, I see, later message says you found it.) No prints, alas. Anyone have any
we can scan and put online?

    > I'll have to find my rl02 controller and build the system up some more
    > so I can run xxdp and find out what exactly died.

I would tend to use diagnostics as 'help', not my main tool in diagnosing
faults. This is particular true as really bad issues can prevent the
diagnostics from loading/running, so if you get skilled at fault analysis
without depending on them, you're better situated to deal with the failures
that prevent using them.

    > Can anyone offer hints as to how to identify which component is broken
    > and how to go about repairing this?

As other have said, lacking the prints, pulling chips (if they are socketed)
ought to enable you to work out which chips goes where. (I'v worked out the
chip layout for a number of un-documented memory cards this way.)

I don't know about your 88B, but the one I have photos of, and my other
Clearpoint cards, the chips are in sockets. With luck yours will be too... If
not, ask if someone has one with sockets (alas, I don't have an 88B), who can
do the mapping for you.

Important: once you have worked it out, pass it along! I'm trying to upload
all the data I've collected about cards for which no documentation is
available to the Computer History wiki.

    > It's the only memory board in this machine, so I guess the problem
    > might actually be a bus or processor board, right?

Could be. I have an KDJ11-B which has stuck bits, and that's the CPU board,
so yours could be to.

    > I have no other q-bus memory to test with, so can't do swapping /
    > process of elimination to be sure.

It's definitely worth having another small QBUS memory card to use in fault
isolation when debugging. M8044's are readily available on eBay for about
$20. They are only Q18, so not usable in the same machine as Q22 memory, but
they are useful for debugging. I would definitely invest in one.

(If you luck out and get a bad one, send it to me, and I'll swap it for a
known good, tested, one. I've fixed a whole bunch of them, got to the point
where the last one I did for someone I didn't even have to pull out my
'scope! I could tell from the symptoms exactly which chip to replace! :-)


    > From: Jay Jaeger

    > Well clearly it is only affecting certain address bits - or the
    > diagnostic would not run at all

Well, like I said, I do have a KDJ11-B which will run the on-board startup
diagnostic, but which has a bit stuck hard on the QBUS interface! So those
machines seem to be pretty resilient.

Although as you point out, if he can load and run the diagnostic, it's
probably not the CPU.

    > note that it is starting at 010000000, so that points to the memory

Yup. If some locations work, and others don't, it's almost certainly the
memory. And since he's only dealing with a single memory card, it's probably
not the bus drivers/receivers on that card, either.

One can use ODT on the KDJ11-B to poke around, and find the envelope of the
problem. (See comments above about not relying on diagnostics! They are
better for saying something's busted, than for accurately telling what _is_
broken.)

With that, and a chip->memory map, it should be fairly easy to replace
the offending chips.

	Noel