25 Apr
2005
25 Apr
'05
6:09 p.m.
Tony wrote:
With the emphasiss on 'likely'...
I have actually seen a case like that (somewhat simplified) :
2 modules, A and B. Module A outputs a signal to B. A was marginal, but
withing spec on timing, B was just out of spec. System failed. Replacing
_A_ with one that was less marginal got the system to work. So you spend
a long time working on the old module A, which in fact is working within
spec, missing the fact that B is the problem.
I am never in favour of shotgun replacements.
It's the brother of
board-swapping in that you don't know what the fault was, you don't know
you've found it, so you can't know it's fixed.
I'm not a fan of shotgun replacements. Board-swapping can be a useful
diagnostic measure, in that if you swap a board out and the machine
starts working, you have most likely isolated the fault to that board, making it easier to focus your efforts to track down
the actual failed
component.
I've found this technique to be especially useful when the symptom is
sporadic flakyness.
And that's the one time I'd never use it. Intermittant faults are a pain
to trace because you don't know you've fixed them (Murphey's law will
ensure tha the machine works fro a bit after you've done something to it
:-)). With such faults you _must_ find the real cause and put it right.
Swap modules, you might think you've traced the fault to a particular
module, only for it to come back and bite you later on because it was
somewhere totally different.
-tony