Odd memory error in PDP-11/04

Noel Chiappa jnc at mercury.lcs.mit.edu
Wed Sep 7 10:44:27 CDT 2016


    > From: Ethan Dicks

Let's look at this one first, this is probably the easier to solve.

    >>> 2) Setting D10 in location 000000 results in D10 set in all the
    >>> locations

    >> Sorry, didn't follow that? Did you mean that if you store 02000 in
    >> location 0, all other locations now report the 02000 bit set?

    > Only 04000, but, yes.

Ah. That's D11. :-)

    > If I set that bit in location 0, or other locations, it gets set in all
    > locations. If I clear that bit, it clears.

So that's likely in the memory (although I suppose it could be the CPU,
_somehow_). It sounds like there's a latch somewhere in the output path
(because it affects all locations right away, not just once you've written to
them) that's getting set one way or the other, and and then, won't change. I
suspect the problem is with the flop for that bit, not in the circuitry
that's clearing/clocking that flop, since it only affects that one bit.

Looking at the MS11-J prints, there is indeed a '174 latch in the output
path; the one for D11 is in E30 (input pin 14, output pin 15). You might want
to throw a scope on it, and see if it is indeed acting consistently with the
symptoms (to make sure this is actually the cause).

Although why it can be either set or reset with a write, but freezes in one
state for reads, is puzzling. I'd suspect control circuitry, but it's only
that one bit. I don't think it can be something on the input side, because
the memory chips have input and output on separate pins, so if something was
hanging on the intput side, it shouldn't make it through the chip.

    > Something appears to have died while I was powered-on and testing last
    > night and now, the run light goes off right away after hitting boot,
    > and I don't see the address lines or the data lines flickering.

Yeah, sounds like the CPU is halting. It's probably going to take a logic
analyzer to figure out what's going wrong. Too bad that machine doesn't have
a real front console, that would probably let you figure out what the problem
is.

	Noel


More information about the cctech mailing list