I am having a rather unusual problem with about six hundred embedded motherboards in the field, and was hoping someone might have an idea (or two) on what might be happening, and possibly what the fix might be.
These are embedded low power Via C3 based motherboards that are deployed in the factory automation industry. Power is distributed from the power supply to the motherboard through a 12" power cable which consists of (2)18ga. ground wires and (2) 18ga. +5VDC wires to an ATX style connector on the motherboard; the board draws about 3 amps of +5VDC under maximum CPU load. After several months in the field, contact resistance on the ATX connector increases for some reason, and is causing a voltage drop as seen on the motherboard (sometimes as much as .5VDC!) There is an onboard voltage monitor on the CPU board that triggers a reset at about 4.7VDC, the end result being that the processor gets stuck in a reset loop - it will run for a minute or so and then the CPU load momentarily increases and resets again. The voltage drop is *usually* greater across the ground lines than the +5 lines for some reason. Maybe there's a clue there?
On a system that is failing, unplugging the ATX connector and reseating it will "fix" the problem and the system will continue to work for several months until once again the resistance increases in the contacts and... well, you get the picture.
If the connector is reseated here at the factory and "repaired", we cannot get it to fail again under any conditions: vibration, humidity, temperature cycling in a environmental chamber, etc.
We've tried using contact lube - no luck.
Thinking that there might be a problem with dissimilar metals, we checked that the contacts on the cable and ATX connector on the board are tin. The contact specification claims that they are gas tight.
At first (a number of months ago) we were seeing only a .2 to .3VDC drop, so we increased the power supplies output to 5.15VDC hoping to compensate for the drop (a Band-aide, I know), but as you can see above, the contact resistance kept increasing only to have the same problem occur.
We cannot increase the gauge of the wire, nor can we add more +5 and ground wires due to there being only 2 pins available for each on the power supply. The crimps will not accept double crimping.
I know the problem could be solved by soldering the wires between the power supply and MB, but this would be a huge task due to the number of units in the field, and the difficulty of disassembling the systems, pulling the boards, and reworking them.
Does anyone have any thoughts on what is happening here, and possibly an easier solution?
Thanks very much for any help.