15 Nov
2004
15 Nov
'04
2:27 a.m.
Testing power supplies before powering up a system is just plain paranoid! (And I'm
trying to be polite here!) There are a thousand parts in modern computers that can cause
just as much damage!!!!!!!! Just last week I THOROUGHLY tested a HP 1000 including
running test on ALL of the firmware ROMs (About three dozen of them!) plus CPU tests,
memory tests, PSU test etc etc etc. I'd had the machine running for several days
testing it. THEN I connected the Floating Point Unit (this was a F series HP 1000). As
soon I turned it on the internal memory tests started but it acted goofy. After removing
the FPU and further testing I found that the FPU had toasted the CPU card :-( It has one
bit in the Microcode section that's stuck high. I'm guessing it's a bus
transciever or something similar since the stuck bit appears in every memory location and
every register. (Anybody have schematics for the HP F-series CPU?)
Further example: YESTERDAY I was testing a "new" HP 1000 (an E-series this
time). A quick check of the PSU outputs showed it to be normal and we powered up the
computer, cleaned and reseated some cards, reconfigured some cards, installed other cards,
etc etc and powered it up and down numerous times over the course of several hours. THEN
it happened! The PSU started making a loud sizzling sound and poured out copious smoke!
All that testing and prior running gave no hint of the impending failure! What's
more, this is the SECOND time that I've had a HP 1000 PSU fail after running for over
an hour. FWIW I pulled the PSU out of another untested machine, put it in this one
(without testing it!) and everythings working again.
IF I had an absolutely unrepairable, irreplacceable computer and IF I had all the PSU
specs and pinouts it MIGHT justify the trouble of testing a PSU before using it to power
up a computer but otherwise forget it!
Joe
At 11:13 PM 11/14/04 +0000, you wrote:
I check the PSUs of machines that I own and have run before if I've not
powered them up for, say, a year oe more.... It doesn't take long to do
(especially not if you're familiar with the machine and/or have
schematics), putting right the damage can take a lot of time and money.
Most of the time you'll have no problems, sure. Most of the time when I
get a machine on the bench (wheter a 'new toy' or something from my
existing collection), the PSUs behave perfectly on dummy load. But if
something does got wrong, you really don't want to kill all the chips....
There is anotehr issue, actually. PSU problems -- marginally low
votlages, ripple, and so on, can cause all sorts of odd behaviour that
will take hours to find if you don't initially suspect the PSU (like the
time my PDP11/45 would run for 1botu an hour and a half before falling
over in all sorts of odd ways -- I traced that to a '+5V' supply to the
memory that was actually about +4.6V). Checking out the PSUs can
eliminate such problems.
Which is 11 years ago....
At the very most 11
years ago, which still isn't bad. I think it suffices to say that the guy who cut
his teeth on the machine
and has supported them since 1986, and *owned the company since 1992*
didn't have any problems with it... He said that the PSUs were sturdy
That would not be enough for me. Heck, the assurance of the designer
wouldn't be enough for me (and nor, for that matter, would anybody else's
assurance). The risk is just too great.
Sure the PSU will have built-in protection -- all decent machines do
(although HP omitted it from their desktop calculators for some unknown
reason -- the HP9815, 9825, 9832, etc desparately need a crowbar!).But
I'd rather not rely on it. Better to be safe than sorry.
I may have been mistaken, but at the time, as
presently, I didn't see
the need to delve deeper into the PSU.
All you need to do is disconnect the PSU from the machine, connect it to
dummy loads (car bulbs are what I noramlly use -- 6V ones on the +5V
line, for example) and check the output voltages. You don't need to
investigate the internals of the PSU unless there's something wrong.
-tony