Folks:

At the company I work for we primarily make PCI-bus PC cards and software to
control them (most are designed for Videoconferencing).

We have recently received a significant number of field failures with a
strange (to us anyway) failure mode.

Analyzing the failures, we found that on each board there was one or more IC
that had a significantly lower DC resistance at the power pins than is
normal. Some boards only have one such device while some have four or five
parts doing this.

Note that frequently the damaged parts still work if the PCs power supply
can source enough current!

Most of these are VLSI parts, but occasionally thay are simple 8-bit bus
buffers.

Most of the parts run from the same 3.3 volt regulator on our boards (which
is fed from the PCI bus +5 volt supply), BUT NOT ALL! Some damaged parts are
powered by +5 volts from the PCI bus and some are powered from an on-board
+5 volt regulator that is powered by the PCI bus +12 volt supply.

The only obvious similarity is that all of these failures come from the same
customer and is used in a custom PC (which uses an off the shelf switching
power supply to generate the power for the PC/PCI-bus & our board).

This customer is very important to us, but they are based in a foreign
country across the Atlantic. We do not yet have access to their system that
our board fails in. Even if we did, it appears that most of the time
everything works fine, but every so often they have a system that fails when
first powered up at their factory and becomes functional when they replace
our board with another. I have been told that they are not seeing any such
failures from the field. So it either goes immediately in their system or it
stays "good." And we've NEVER seen any failures like this from any other
customer; this customer receives less than 25% of all the boards of that
type shipped, which makes a design flaw of our board seem less likely.

Worse, we are running at about 15% strength right now (over 80% of the
people that were with us at this time last year have been laid off). Those
of us left are wearing so many hats that we have very little time to spare -
and no budget for "unnecessary travel."

From the bench, we might see this; A field return is tested on one PC and
appears to be operating normally. When placed into another PC (which has a
PCI-bus extender card with over current protection & indication) it
immediately fails and one of the over current LEDs activate. a good board
measures several K-Ohms on the 3.3 volt bus, but one of the failing units
measure only a few Ohms to the digital meter. Lifting IC legs one by one
eventually identifies the faulty chip(s), and when replaced everything
returns to normal. In another case it's the +12 that is over current and it
is traced through a +5 volt regulator to one ICs' +5 supply pins.

The majority of failures are 3.3 volt parts, then 5 volt, and finally those
whose power is derived from +12 volts.

The onboard 3.3 volt unit is a Linear Technology LT1506CR-3.3 switching
regulator (the dash 3.3 is fixed at 3.3 volts), being employed as per their
own specs/schematic. And NONE of these has failed hard; while it is possible
that this circuit is failing, it always appears to be functioning properly
when we analyze the faulty boards. Additionally, if this was the cause, how
would that explain the failures of the parts powered by the PCI bus +5 and
+12 volt supplies? And why only this customer?

It seems reasonable that the power supply that this customer is using is
occasionally doing something that our board finds disagreeable, but what?
And why only when first assembled and never in the field? It would be
extremely helpful if we had some idea what to ask this customer to
investigate...

I'm hoping some of you PIC-listers might present some opinions that help
solve the problem - or at least help us start asking questions that lead to
a solution.

Any thoughts?
Thanks in advance.
Bruce.

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics