2 Aug
2016
2 Aug
'16
4:54 p.m.
I've also got two additional boards in a bag that I haven't taken out yet
https://goo.gl/photos/D4quXT992Y6huQoAA . They're in there quite tightly,
but I did take a peek at some of the chips that I could see from the edge.
Couldn't read the part number, but they were Intel chips and almost looked
to me like they were early static ram.
If they are I know it might not help with testing the memory but might help
with making sure the CPU works first. Not sure which slot though they'd
install to even if they were, I'd presume the white slots.
For the memory test I don't see any (visible) front panel switches that
would indicate a memory test, the ones that I can see just look like
they're for direct depositing into core. But I'll have to check the manual
for that.
On Tue, Aug 2, 2016 at 5:33 PM, Jay West <jwest at classiccmp.org> wrote:
Glen wrote...
------
Isn't there a looping memory test capability built in through the front
panel? Maybe you have to flip a switch or something to enable that. I'd
have to look at the manual to remind myself.
------
My foggy memory is telling me the same thing as Glen mentions above. The
21MX (m/e/f) definitely has a built in microcoded memory test, but I don't
recall 100% on the 2100.
Hayden wrote....
-----
Well I'm hoping for the memory section that I can find at least one of the
4 core boards that works right away. I've got an 8k driver board I can
swap it so I could run the system on only a single core board if needed.
-----
That's the problem, what I was referring to earlier. You can't really test
just a 4K board. Its been a long while since I had to pop the cover on a
2100 thus my recent brain memory is all 21MX stuff. But as I recall a
minimal 2100 memory section is memory controller, ID, XYD, and core. You
will be unlikely to tell which of the 4 is causing issues just with front
panel tests. A typical program doing read/write/compare can fail with no
indication which of the 4 boards is really the problem. Swapping is
problematic if there are several bad boards. You can also run into a
difficult to test situation if a 16K ID board is bad "at the top", meaning
an 8K board tests ok, but the same board tests bad if it's at "the next
8K". Or if you add another 8K, determining if it's the 8K or the upper part
of the 16K ID that's faulting.
Troubleshooting the memory of a 2100 via boardswapping with no known good
boards is ... irksome. Understatement of the year.
One thing to keep aware of (this is sketchy, as it's very foggy memory)...
I think there is a switch on the panel (under the cover I think) that lets
you clear or set all memory INCLUDING the parity bit (memory is 17 bits on
these). If you don't do this between memory tests - and you have some board
that is not setting memory parity correctly for each word, you are really
gonna be scratching your head. Once you get bad parity bit status, you have
to wipe before retesting.
One other thing I remember from working on the 2100's.... there was some
setting - probably power supply voltages, or something with the memory
section - that was temperature sensitive. You had to adjust some trimpot or
the like based on the room temperature (+/- a window of course).
J