[Tfug] Memtest question

[Bexley Hall]

Hi Zack,

--- On Tue, 12/8/09, Zack Williams <zdwzdw at gmail.com> wrote:

> >> If bad caps are found, unless you are terribly fond of the
> >> machine, off to WT with it (recapping is just too much work
> >> and far too easy to lift foils).
> >
> > I've replaced the through-hole caps on two flaky motherboards and
> > found the job quick and easy.  Even if the recap failed, the effort
> > and cost invested was so small (45 minutes / $10) I would shrug it off
> > without too much trouble.
> >
> > If someone was considering replacing the caps on a their motherboard,
> > I wouldn't discourage them.
> 
> I find that recapping depends greatly on the quantity of bad caps, the
> board layout, and how good of soldering gear you have.

Folks seem to concentrate on replacing "bulging caps" as if they
are *The* "bad caps".  Bulging caps have obviously vented and failed.
There are often similar caps (came manufacturer) on the board
that won't *look* like they have failed but they have, in fact.
The only way to test the caps is to remove them from the board
and put them on a bridge to test their ESR.  (but, if you have
removed them all, why not just *replace* them all?  :> )

Some (Ecap) manufacturers are notorious for making crappy caps.
I look for those brands and, if the board is populated with
them, the board is -- or will be -- toast.

Replacement parts must be hi-temp devices.  This adds a cost
premium -- and sometimes subtly changes the form factor which
might make it impossible to get the new device into the old spot.
Additionally, you can buy caps in various grades.  The cheapest
being 2,000 Hr.  Cost goes up from there.  (if you are going to
invest the time to replace the bad parts, wouldn't you want
to use the *best* components to justify your time?)

> I don't mind redoing a board that has 4-5 caps on a wide trace power
> handling section with loose layout, but if you get close to tight
> logic traces or the components are packed really tight,  the
> likelihood of success goes down dramatically, in my experience.

Agreed.  The inverters in LCD monitors fit your first criteria.
The bulk caps in most (newer) motherboards are all soldered
into large power planes -- often NOT thermally relieved.  So,
you use a soldering iron that is too small and all the heat is
drawn away from the solder joint and into the *internal*
power plane (most MB's are 4 - 6 layer boards with the power
and ground planes internal -- the very places to which the caps
connect!).  By the time the solder connection is hot enough,
you've damaged the pad/foil from prolonged contact with the iron.

Catch 22: you want a small iron tip for the fine geometries used
on these boards but that fine tip doesn't have much thermal mass
to compete with the large heat sink presented by the foil layers.

Then you still have to deal with removing the solder from the
holes (more heat).

And, the ultimate deal breaker is that you now have a board of
unknown quality!  How long has the board been running in this 
*failed* condition?  Don't you think there is a reason for the 
caps being there in the first place?  :>  Do you think their
absence comes without lasting consequences?  (all rhetorical
questions).

The caps all work to keep the voltage at key points in the circuit
WITHIN ACCEPTABLE LEVELS.  But, there are two different sets
of "acceptable levels".  One set governs "proper device operation"
outside of which the parts in question are not guaranteed to
perform as designed.  E.g., CPU doesn't "add" correctly.

But, there is another set of levels that are device "absolute maximum
ratings" -- beyond which, you have *damaged* the device.

If you recapped the machine because it was misbehaving noticeably,
then you obviously have been operating outside of the "proper
device operation" limits.  But, do you know FOR A FACT that you
have *never* crossed beyond -- past the "absolute maximum ratings"
of the device?  (Ans:  "no")  And, how will you determine that
the device still *works* properly -- because it passed the
POST?  Or the manufacturer's diagnostic test??  (you don't
really think that this is how the chip manufacturer's certify
their products??  :> )

Unless you are used to doing board assembly and repair work,
chances are you'll just make a mess of the board in the process.
And, end up with something dubious in the process.  Maybe OK
for rescuing a gaming console but I sure wouldn't want to
rely on anything like this!  :(  PC's are just too damn cheap
and transitory in nature to invest much in them...

YMMV.