JW wrote:
On Mon, 20 Sep 2010 22:48:04 -0400 PC Guy wrote in Message
id: :

I'm aware of a massive problem with capacitors that really affected Dell
around the same time that these boards would have been made, and we are
experimentally taking a few boards and changing the electrolytic
capacitors (1000 and 1500 uf) with new ones to see if that solves their
problems with operational stability.

I'll bet that fixes your problem. Not just Dell was affected, and
electrolytic caps only have a lifetime of 2000 hours or so anyway. Make
sure you get low ESR types for replacement. If the problem persists, try a
new power supply as well.

You could try this diagnostic, there's a 30 day evaluation:
http://www.passmark.com/products/bit.htm
If it does what you need it's pretty reasonable price-wise.

Look for information on "modified Arrhenius equation". It's the
equation you use, to convert the "2000 hours", into the real
life span.

Arrhenius is a chemistry equation, related to reaction rate.
It's something like "reaction rate doubles for each 7C rise
in temperature". It is used to model how reaction kinetics
change with temperature.

Electrolytic capacitor reliability, is "curve fitted" to the
Arrhenius equation. But the exponent in the equation doesn't
use 7C. It is a different value, specific to each product.
Curve fitting reveals the characteristic value.

Say the capacitor is rated 2000 hours at 110C, and the inside
of the computer is 35C air temperature. The temperature difference
is 75C between spec point and application temperature. Lets say
the curve fitted Arrhenius factor is a doubling per 15C. The
temperature difference leaves room for five doublings or a
factor of 2**5 or 32. The predicted lifespan of the capacitor
would be 64000 hours at 35C.

If the equation pops out an answer higher than 15 years, then
the assumption is, the rubber bung on the bottom of the cap
fails. Early in the life of the cap, the "chemistry equation"
applies. Later in life, the packaging falls apart, and that
is the proposed failure mechanism above 15 years. (Once
the rubber plug fails, it allows the capacitor to dry out.
Ozone will attack the rubber.)

When replacing caps, I would replace "like with like", in the
sense that I wouldn't change the class of capacitor used.
If a "switching power supply grade" electrolytic cap is used
in the Vcore circuit, I would select a similar type as a
substitute. If I were to place an OSCON or a polymer cap
in there, then re-evaluating the equations in the Vcore
design datasheet would be required. The following is an example
of a Vcore datasheet, with maths. Each datasheet will have different
kinds of advice, particular to the circuit and its
requirements. If you don't wish to use one of these,
then select caps from a "like" family, to what you
find on the motherboard.

(See, for example, page 14 "Cout Selection")

http://web.archive.org/web/200403310...5ADP3180_0.pdf

Paul