Jeff Liebermann Inscribed thus:

On Thu, 06 Jan 2011 09:38:53 -0700, D Yuniskis
wrote:

I'm wondering what the prevailing approach is when
it comes to dealing with kit that has been found to
have "bad cap-itis". Particularly, PC's.

That's rapidly becoming a major part of my business, second only to
malware removal.

Agreed ! Malware is most definitely No:1 Bad caps No:2 !

Do folks exploe the current operating conditions
of the machine *before* re-cap-ping (i.e., see
what supplies are affected, what sorts of ripple
is present, etc.)?

No. It takes too much time. If it's obvious that there are bad caps
in the box, the only decision is whether to replace ALL the caps, not
probing around looking for sections to avoid. Where I get into
trouble is when I get a working machine, that has obvious signs of
leaking or bulging caps, that will eventually die. The customer often
does not want me to do a major motherboard capacitor replacement if
it's still working. Oddly, some of these machines have lasted years
with obvious bulging caps.

I tend to point out to the customer that there is a liability in leaving
bad caps alone. At some point there will be a catastrophic failure and
that they will have to pay again at some time in the future. Since
they are paying a fixed service charge its in their interests to have
the caps replaced before that happens. Yes it bumps up the bill, but
not as much as it would if they have to replace the machine before they
need to !

Or, do you just replace and
*functionally* retest (i.e., if you haven't
probed the circuit to get an idea for the exact
nature of the problem, probing afterwards at
that level of detail seems to be a contradiction)?

No. I use the shotgun approach and just replace every cap of the same
type. The time I spend doing the "extra" caps is about the same as
what it would take to probe around looking for the worst offenders.
The idea is to do the job once, and not have it come back with
additional blown caps.

I would do the same. I have also started paying attention to the caps
in the RAM PSU circuits and replacing them as well. Having had some
really weird machine behavior because they have become bad.

In particular, how do you know if the system has
been operating within "absolute maximum ratings"
while this fault has been present? How do you
know the system's reliability/integrity hasn't
been compromised (so your "fix" is just a
band-aid)?

Ouch. I've bought bulk caps on eBay suspecting that I may be asking
for trouble by installing garbage caps. When possible, I'll buy
Panasonic caps because I've had no problems with them. If it will
fit, I try to increase the voltage rating, even if it means using a
smaller capacitance value. I have attempted some accelerated lifetime
testing using an ESR meter, with inconclusive results. At this time,
I can't tell the difference between a new bad cap, and new good cap. I
consider uprating the caps near the CPU as mandatory as I know (and
have measured and calculated) that they are running too hot for their
voltage rating. Also, I try to use polymer caps, but they're only
available in smaller values and kinda expensive.

I've had some very odd behavior that has disappeared when replacing caps
not directly involved with the CPU PSU circuits. I recall one machine
that had a very odd patterning on the screen as the resolution was
increased. Replacing the RAM PSU caps cured that problem. The clue
was only the on board graphics showed the problem. Putting in an
external graphics card was just fine.

E.g., I've been recommending disposing of machines
with this problem. The labor to re-cap just doesn't
make sense given the book value of machines that
exhibit these problems -- especially in light of
the fact that you can't be sure that the "repaired"
machine is really "100%" (maybe OK for a machine you
use informally at home, but would you want to *rely*
on it in a commercial environment?).

Not totally true. The problem with getting a new machine is that the
customer will have to pay me to transplant their software to the new
machine. Often, that's more expensive than the machine. If it
involves an OS upgrade, it gets more complicated. Of course if the
machine is ancient, it's not an issue.

There's little difference between home and "commercial" machines these
days unless you're talking about servers. Well, maybe more chrome and
glitz on the home machines.

Very true nowadays. :-)

I suspect most machines are just recapped and re-used
but wanted confirmation or repudiation of that opinion.

It's about 50% here. If the machine is more than about 6 years old,
I'll usually suggest a replacement. If fairly recent (Core2Duo or
better), I fix it. My recent batting average is about 75% success. It
was much worse in the past when I didn't replace all the caps.

However, a major problem has been BGA array soldering. If I dive into
the machine, replace all the caps, there are still plenty of other
things that can cause a subsequent warranty problem. I try to warn
the customer that replacing the caps is a "best effort" and that if it
fails, they're committed to buying a new machine, and that I owe them
a new machine. It's awkward, messy, and full of problems. They have
to recognize that recapping involved some risk. It's often easier to
just suggest they buy a new machine than to deal with the ethical
issues required to convince them that simply touching their machine
does not create a lifetime warranty.

Yes that can be a problem with some clients. They expect you to be fair
with them, but they aren't prepared to be fair with you.

I've actually had an instance where a client sued me and the Judge found
in my favor. They then refused to abide by the judgement, so I counter
sued and won. They still refused to accept that they had to pay and
ended up with the court bailiff paying them a visit. I'm currently
getting £2.00p per week and expect to wait about twelve years before I
get paid in full.

http://802.11junk.com/jeffl/pics/drivel/slides/diploma.html
This is from about 1987.

Hey Jeff, can I use that ! Big Grin.

--
Best Regards:
Baron.