On Thu, Jan 5, 2012 at 2:35 PM, Tony Duell <ard at p850ug1.demon.co.uk> wrote:
Don't know
if it's similar to what you've seen or not, but from my
experience testing M-series modules for the -8/L and -8/i, I've found
more dead 7474s and 7440s than any other type of chip.
When I was repairing m HP98x0 machines I noticed that 74Hxx parts were
significantly more likely to fail than plain 74xx parts. And I think the
74H40 and 74H74 were high up the list of those that had failed.
There are some H-series logic parts used in the PDP-8/L, but not
enough, IIRC, to be statistically significant compared to straight TTL
parts. That said, it doesn't surprise me that you saw a similar
pattern with a different logic family in a different machine.
For my
testing, I rigged up a 3M test clip...
into the ZIF socket of a hand-held IC tester.
...letting the tester
power all the ICs on the board but only fiddle the lines on the chip
under test.
That's OK if the flipchip sinmply brigns out each IC to different pins on
the connector (a lot of the simpler ones do),
A lot do. In my original post, I gave several examples for which this
works. It turns out to be the vast majority of the single-height
M-series modules in an -8/L or -8/I (the M452 clock generator and M310
delay line being two exceptions I can come up with off the top of my
head).
but surely on more complex boards where the ICs are
interconnected,
you're going to have contentions between the outputs of the IC tester
andf the outputs of other ICs on the
board (which are being powered from the tester).
Quite so. I mentioned the M220 Major Registers module as being one of
those cases. The M706 and M707 TTY interface boards are two others.
There is, of course, HP's logic comparator, which
is a device that takes
a good IC of the type you are looking at, feeds it the same inputs and
the drive on the board you're tersting and compeaes the outputs of the
good IC and the one on the board. It then indicates which pins are different.
Yes. I don't happen to have one of those, nor have I noticed one for
sale. I already had the ZIF-socket PIC-processor-driven IC tester - I
did not purchase it specifically to test DEC boards. I made a cable
to adapt an existing tool for this specific job.
Personally,giiven all the restrictions, I am not sure
either method is ant
faster than the old-fashuioned way of seeing what doesn't work and tracing
signals to find why it doesn';t work.
In the specific case of the PDP-8/L, my experience is that if you are
seeing single and double-bit errors with address/data loads and
registers on the lights just from a few tweaks of the front panel
switches, you have a high probability that the problem is on one of
your M220 boards. Outside of that, you can have the printset open and
*quickly* pull each single-height, Chip-Tester-friendly board from the
path you wish to investigate, move the clip and press the "test"
button 3-4 times and replace the board faster than you can install a
board extender, replace the board, and re-run the test you just ran to
look for the failure.
In no way did I ever say that many or most or every IC failure can be
detected with a hand-held tester and a custom clip cable. I _did_ say
that I built one and repaired my machine with it and I'm happy with
how quickly I could confirm for myself that hundreds of ICs were
likely to be functional.
Yes, there are failure modes for simple TTL gates that won't be picked
up by a hand-held tester. In my experience, those are rare compared
to a chip with a blown input or output that just plain produces the
wrong output when the inputs are tickled. *That* will be found by the
tester and clip.
-ethan