05-09-2012, 04:53 PM
Capacitor Corona
Its 3:00 A.M. A Saturday morning, Los Angeles time. My cell phone is going off, two, no three times. It’s
got to be a wrong number, so I ignore it. Then I hear a text message come in and that funny ding your cell
phone makes when an email arrives. I decide it’s worth it to get out of bed. So I get to the phone and it
immediately rings again. I don’t recognize the number, but I pick up. “Is this AEi Systems?” the caller
asks. “Uh yea,” I answer in my best professional 3AM voice. “We have a big problem and we need your
help.”
It turns out the engineers on the other end of the phone are as serious as a heart attack. They work for an
important government agency. We can’t and won’t identify the agency or the specific system they were
having a problem with, or exactly what it controls, other than to say it’s extremely costly on several levels
and for more than one constituency when it doesn’t run. Here is what I can say: one of their circuit boards
keeps failing and they don’t know why. There are several hundred boards in the system, and as they fail,
the system gradually becomes less valuable as it loses all output. They need to find out why a capacitor
in an R-C snubber across a thyristor is failing, and they need to know now!
As AEi Systems’ main service, we do a lot of worst case circuit analysis. As part of that process, we see a
lot of designs and do a lot of SPICE modeling, as well as circuit troubleshooting. This particular customer
wanted us to model the circuit in SPICE in order to see the stresses on the snubber, with the idea that we
might be able to determine the reason for failure.
By about 5 A.M. our investigation had begun.
Failed parts were flown by special charter flights to us in California both Saturday and Sunday. Regular
meetings every three hours were used to reinforce the sense of urgency as well as to assess our findings.
Over the weekend, we modeled the circuit in SPICE, assessed the stresses on the elements, and even
managed to contact the capacitor manufacturer (yes, manufacturers do answer the phone sometimes on
weekends).
The capacitor that failed was a radial-leaded polyester film cap rated at 600 VDC and 200 VRMS. Upon
failure, the capacitor was found to be shorted. The manufacturer’s derating for the polyester capacitor
was 70% of the 85°C rating up to 105°C and 50% of the 85°C rating up to 125°C. Military Handbook
1547A is typically used for military and aerospace programs, and is a good guide to follow in hi-rel
applications. The 1547A derating for polypropylene is 50% up to 70°C and linearly down to 25% at 100°C.
We identified that the capacitor did not fail its rating (although it did fail its derating). The measured
voltage stress was +317 volts peak and -430 volts peak. The equivalent RMS voltage is 173 volts. The
voltage stress ratios were 71.7% for DC and 86.5% for AC, both well above the recommended 50% max.
The ratio of the rated AC to the rated DC voltage, however, was only 0.33 – this is quite low, and
indicates the capacitor was not well suited for high-frequency AC applications.