Re-Capping A PC Motherboard

Another recent home project, in the category of things that made me unreasonably pleased with myself, was replacing two capacitors that had failed on the motherboard of my Shuttle SN68SG2 PC.  I actually first wrote about this computer in the context of the number of restarts required while installing Windows XP.  The machine is now 2 years old, and is my primary desktop PC at home.  About a week and a half ago, I was working one night when it just clicked off.  No warning, no shutdown, just >blink.<  This is never a good sign.

After a moment of swearing, I realized that virtually all of the apps I use these days save automatically and so the actual risk of lost data was minimal.  I powered back up, and everything worked fine, so I convinced myself that it was just some power glitch on the mains, and nothing to be particuarly worried about.

Two days later, it happened again.  And this time I had to admit that it was probably a real problem, and I would have to at least keep an eye out for it.  I powered back up, and went back to work... for about an hour.  When I powered up following that third failure, I didn't quite get all the way through Windows booting before it dropped dead again.  Once the mean time between failures of a PC is less than the boot time for Windows, it is pretty well shot.

Given the nature of the failure, I was almost certain that it must be the power supply.  Nothing else, it seemed, was likely to cause such an instantaneous power-off failure like that.  So I took the housing apart, measured the power supply, and figured out that there was a replacement available at Fry's.  With the new power supply installed it came up, booted to my desktop... and then died.  More swearing.

Finally I Googled around and came across this forum posting about my model.  Sure enough, upon closer inspection, those same two capacitors in my machine were bulging badly:

When I removed them from the board, one of them also showed signs of electrolyte leakage on the bottom side:

Interestingly, when I measured the capacitance, both the failed caps actually measured *above* spec.  The parameter that has gone bad, I suspect, is the equivalent series resistance.  These are "ultra-low ESR" caps, meaning, essentially, that they store and release charge very efficiently.  Or at least, they used to before their guts started leaking out.  I don't have an ESR meter, but that is the usual failure mode.  This is an example of the capacitor plague which dates back to the late 90's when some of the Chinese manufacturers got a bit sloppy with their electrolyte formulation.  But since the parts only fail after several years of use, there are many millions of them in service.

Getting replacement parts was an interesting challenge.  The original caps are 8 mm in diameter and 23mm long, with a capacitance of 1800uF, and a voltage rating of 6.3V.  They were also nominally rated at about 0.015 ohms at 100KHz.  I could have ordered from Mouser to get some almost exactly the same size and ESR, but it would have taken at least 2 days and cost $38 in shipping, for $2 worth of parts.  The closest part I could find over the counter locally was at HSC Electronics Supply in Santa Clara, and they were 1800uF, 0.025 ohms at 100KHz, but rated at 25 volts.  This meant they were appreciably larger:

Even the leads on the new caps were larger, 0.2mm wider than the originals, so I had to grind them down with a Dremel tool, and bend them inward to make up for their wider spacing.  Once installed, the new caps stood a bit off the board, but it was a situation of "close enough."  The off-board mounting proved fortuitous, because their extra height also caused them to interfere with the heat pipe between the processor heat sink and the fan.  I ended up bending them over slightly to fit the system back together.

So in the end it was not a particularly clean and elegant fix, but it worked, the parts cost $2, and I was able to get them locally.  I put the machine all back together, and it booted right up.  And when I came back about 20 minutes later, it was off again. (#$%@!)  Then I realized that I'd forgotten to reconnect the power to the fan when I put it all back together, and the machine had shut itself off because it had overheated.  I let it cool down, plugged in the fan, and it has now been running for about a week.  I'm hopeful that I'll get a couple more years of use out of it before I need to replace it.