Norm X wrote:


Thanks for you professional opinion Paul. On the other hand, if my rig
becomes unstable, maybe I should have the shop put a scope on it. If the
electrolytic capacitors are bad. I could pay them to replace them. Their
labor is cheap but I should look into the price of new capacitors. This
image:

https://www.ascendtech.us/mmASC/Images/SAK09-01.jpg

shows capacitors all lined up in rows and could be desoldered and
resoldered.

In your picture, the caps next to the CPU are polymer caps and they
filter VCore (output).

The row of caps near to the ATX12V are input side filtering
on VCore. Those ones happen to be electrolytic.

The reason there are numerically so many caps, is ripple current.
Regular electrolytics have a relatively low ripple current
rating. By putting six caps in parallel, one sixth of the
current flows in each one.

When the cap has an AC voltage across it, and a reactance Xc,
an AC current flows. This current can be measured in amps. It's
quite a significant amount. It can make the capacitors get
warm. The hotter a cap gets, the shorter its life.

Now, Sanyo Oscons have a much higher ripple current rating.
You could, in principle, select a single Oscon and replace
the six other caps. But, we don't do that. If you have the
app note, it gives all the equations. You could plug in
the Oscon characteristics, and the equations would tell you
how to tune some of the other discretes. In other words, if
you put "too good" a capacitor in the circuit, it actually
throws the parametrics off and the circuit needs a slight
redesign.

So when changing caps, you continue to use as many caps
as were there before. And you use the same "class" of caps.
If it didn't have Oscons when you got it, you don't put
Oscons in it. You use the same flavor of caps as you found,
and then you don't have to redo all the math. And the
capacitance isn't critical. If the circuit had 1500uF caps,
you could use a 1200uF cap, or an 1800uF cap for all six.
It makes a slight difference to ripple amplitude and
closed loop stability.

Both the motherboard designer and the Vcore application note
writer, think alike. Neither is interested in Oscons. Both
know that the price "sweet spot", is with certain kinds of
"decent but ordinary" caps.

The caps have such predictable characteristics, you can tell
if someone is cheating purely by "volumetric considerations".
For example, when I see some caps on a motherboard, I head
over to a catalog and look for a replacement, and an E size
case has to be replaced with an F sized cap, it tells me the
original cap was inferior and the undersized nature of it
means it was probably under more stress than would be desired.
The larger cap I'd be called on to substitute, won't fit, because
the cap outlines are shoulder to shoulder, and if I buy that
F-sized cap, and put six next to one another, the end two
won't be insert-able into the board. (Some of the caps
will be leaning like fallen trees.)

You try not to leave a lot of lead length on them when installing
them. But I have seen people do recapping jobs where they're
forced to do less than optimal installations of them.

*******

It's hard to get those caps out. There are two hole styles.
At work, we had a "relaxed fit" hole for electrolytics, and
used a lead bender to "form" the legs so they wouldn't fall
out. The result of doing this, is when you need to do
maintenance on them later, you heat up the oversized holes and
the cap just "falls out". Great. Then you clean up with
a solder sucker and solder wick and head off and look for
the replacement to insert.

The other kind of situation is more common. The other place
I worked, used "straight" capacitor legs and "interference fit"
holes. These are a bitch. I used a vacuum desoldering station,
and spent around two hours one day, trying to nudge those out.
The thing is, the solder is good and hot and is ready to behave.
But because the cap leg is "jammed" into the hole, even pulling
with a pair of pliers isn't working. And if you pull too hard ?
The entire fillet comes out, and you're ****ed.

So who ever is doing your re-cap job, they have my respect.
It's hard work to rework boards like that. And it so happens
the computer motherboards are inferior quality.

We had a board shop apply a series of optional processes,
that made our board blanks "bulletproof". You could rest a
soldering iron on the laminate, and the foil wouldn't lift.
Whereas if you put an iron on an Asus motherboard for 30 seconds,
a pad could end up stuck to the tip of your iron. (Typically
happens to pads with no copper track connected to them.) Not
only is it hard work, but you're also working with motherboards
that don't take kindly to abuse of any sort. You could
easily ruin the board.

So if the caps are interference fit, bad things are going
to happen, and the dude at the soldering station is going
to be putting coin in his swear jar. I had all the right
equipment for the job, but the stuff just wasn't being cooperative,
and I would not look forward to doing that a second time.

Even if you apply ChipQuik to the joints before attempting
removal, there is still the problem that the legs were
jammed in there. A "relaxed fit hole", on the other hand,
just about any one could do one of those.

Sometimes the fit is so tight, the auto-insertion machine
"crumples" the cap when fitting it the first time. And they're
using this interference fit idea, so the caps can't move
as they go through the soldering process.

Paul