And of cause by observing the behaviour(and error
leds/codes) of a machine
you get a reasonable clou where to search for a fault.
Self-diagnostics always worry me in that you're using a faulty device to
tell you what's wrong with itself. How do you know the bits needed for
the diagnostics are working correctly?
To be fair well-designed self-diagnostics start by assuming that only a
tiny part of the machine is working (and testing that), and then going to
using more and more features, but never depending on a feature that's not
been previously tested. But an awful lot of self-diagnositcs are not
well-desiogned. I've repaired at least one HP instrument where, if one of
the RAMs had failed in a particular way, it did not give the expected 'RAM
U<n> failed' pattern on the LEDs, but something totally different.
Fortunately I only got a list of the erorr codes _after_ I'd found the
fault by the old-fashioned method.
Some of the HP9000/200 power-on self test error messages are little
weird. I got confused by erorrs relating to the disk controller in 9836,
for example. If I was a board-swapper, I proably would have replaced the
rgith FRU first (namely the FDC board), but the message pointed in
totally the wrong direction for component-level repair.
-tony