November 2013
This has to do with electrolytic aluminum filter caps for switching power supplies.
No matter what type filter cap I try, they blow out (become pregnant) after months or a few years. I repaired cable boxes for many years that had the exact same problem.
This is only a three volt supply at about two amps. Ten volt 1,000 mfd caps are used in the stock supply. Also, a three amp Schottky diode (burns up) supplies the DC to a 15 amp logic N-channel MOSFET with a heatsink. It gets hot. Then, the output of it gets a cap, a choke, and a cap. Nicely filtered three volts.
This is my final change-out and it is lasting the longest. So far, no blow outs, but it has only been seven months.
Now, the two five amp Schottky diodes in parallel. Using only one still gets super hot. Caps 25 volt at 1,000 mfd. I’m only using general type filter caps at 20% 105° C. Why has this been such a big problem?
The cap that usually blows is the first one after the MOSFET. I see no spikes on the output of the MOSFET either. I could use a TO-220 pack with dual diodes in it, but no room. The two 5 amp in parallel work just great and only get warm.
Please log in to post an answer.
The likely cause is the high-frequency ripple; aluminum electrolytic capacitors don't deal well with high frequencies (high loss factor) or large AC component of a waveform (capacitor may be depolarized).
He could substitute a tantalum capacitor, and/or parallel some ceramic capacitors across the aluminum one, e.g. 0.01 µF and 0.5 µF (or 10 nF and 500 nF, if you prefer), keeping the leads of the ceramic capacitors short. He could also use a small ferrite choke in series with the capacitor to decrease the AC component of the waveform.
A schematic of the power supply would help to pinpoint the problem.
The quick answer to your inquiry is that you're using the wrong type of filter capacitors.
From the description given of the power suppy — three volts, two amperes (6 watts), and the heating problems encountered, I surmise that you have a small flyback switching power supply that is not operating very efficiently. If the supply is rated for a mains input range of 90-130 volts AC and you're operating near the top end of that range — 120 volts or so — the ON time of the switch transistor will be quite short relative to the OFF time. All of the input power to the flyback transformer must be delivered during that short ON time, so the switch current will be high. Similarly, the flyback (secondary) current pulse through the output diode will be high because it must deliver all of its energy to the output capacitor in a short time. Both of these conditions serve to elevate the operating temperature of the switch transistor and output diode.
Finally, that output current pulse is dumped into the output capacitor. A real capacitor can be visualized as an ideal capacitor in series with a small resistance; the latter is known as the Equivalent Series Resistance (ESR). You need to use capacitors having a very low ESR value and a high ripple-current rating. I would expect that your output capacitors are seeing very high instantaneous ripple currents. Capacitor heating is a function of the ESR value and of the square of the rms value of the ripple current. A suitable capacitor might be a Panasonic EEU-FR1E102, available from DigiKey, their part number P14424-ND, $0.91 each. The ESR of this device is 0.020 ohms and it will tolerate over 2 amperes rms.
As far as paralleling diodes goes, I've had bad experience with that. You cannot guarantee exactly when the diode will switch from non-conduction to conduction, so for a very short instant, one diode may be exposed to the full current pulse. I can't visualize why you have room to fit two 5-ampere Schottky diodes but not room for one 10-ampere TO-220 package.
The reason that the capacitor nearest the MOSFET switch is the first to be destroyed probably relates to the board layout, and the fact that the other capacitors have additional lead inductance (including the etched conductors) in series with them. If possible, try to equalize distribution of current from the switch to each of the capacitors, and do the same for their returns to the Common bus.
I hope that these suggestions help.
