2
\$\begingroup\$

I have a commercial replacement PSU that I am using for a vintage XT class computer which has 4 DC lines (5v, 12v, -12v and -5v). On fitting the PSU however, the PC operated but after a while under load one of the capacitors E7 within the 5V DC section got very hot and vented, followed by a second E13 within a different line. The seller has given me some advice and kindly sent me a schematic. From this I have replaced the caps with rubycon 35V versions and E7 gets to about 50c tops from my testing, E13 barely get's warm. I will fit a fan. I wondered if anyone here from the schematic could please give me there thoughts on if there is a reason one failure led to the other? Many thanks

The specs for the replacement are approximately correct. The PSU can supply +5v @6A, +12v @ 3A, -12v @ .8A, -5v @ .8A

My PC is needing (tested via shunt resistor) +5v @ 2.1A +12v @ .34A -12v @ .02A -5v @ 0A

Picture of schematic

\$\endgroup\$

2 Answers 2

4
\$\begingroup\$

The power supply schematic indicates this is a flyback converter, total output power rating = 80W (sum of output power rating of all rails), which is quite modest by today's standards. Your load, however, is only about 12.6W.

Failure of E7 (on the +5V rail) is quite understandable, as it was subjected to the most stress (ripple current). Even more understandable if this capacitor was several decades old - it would likely to have suffered "drying out" and its capacitance and ripple current withstand would have suffered as a result.

Failure of E13 (on the -5V rail) is less understandable, since there was no load on it, however, once E7 failed then E13 may have been subjected to higher ripple current and/or ripple voltage (and possibly DC voltage) than normal since E7 would have been providing the majority of the voltage clamping - details would depend on the magnetic coupling between each of the secondary winding pairs.

You mentioned that the replacement for E7 got to 50C - that is of some concern since it is not necessarily the temperature per se that does the damage, it is the temperature rise above local ambient. Check the datasheet of the capacitor used, which will provide guidance on the effect of (a) ambient temperature, and (b) temperature rise.

I would be tempted to replace E7 with two or possibly three capacitors in parallel, to get the ripple current stress down - ripple current is what casues most of the internal heating. Reducing ripple current by factor of 2 reduces the power loss by a factor of 4 (power = I^2 * R). Look up electrolytic capacitor datasheets to ensure the ripple current stays within manufacturer specifications. For these applications, the value of the capacitance is not as important as the ripple current withstand. There is a second filter after E7, E7 is intended to filter the flyback ripple current and get +5V supply voltage ripple to an acceptable level that is then cleaned up by L2 & C8.

\$\endgroup\$
6
  • \$\begingroup\$ What do you mean, exactly, by "not the temperature per se that does the damage"? Most thermal failures are related to absolute temperature. \$\endgroup\$
    – Hearth
    Commented Nov 29, 2023 at 4:53
  • \$\begingroup\$ @Hearth A capacitor running at 80C when local ambient is 20C is going to have a much shorter life than the same capacitor running at 80C when local ambient is also 80C. \$\endgroup\$ Commented Nov 29, 2023 at 7:04
  • 1
    \$\begingroup\$ @FabioBarone Interesting. I never thought about it like that, and I'm a little sceptical. As I understand it, electrolytical capacitor aging is largely due to the electrolyte drying out, which happens faster at an elevated temperature. To my understanding, it does not matter much if this temperature is ambient or caused internally. Do you have a source with more information? \$\endgroup\$
    – marcelm
    Commented Nov 29, 2023 at 11:05
  • \$\begingroup\$ By adding an additional capacitor in parallel to E7 (in order share the ripple current) I believe this would also double the capacitance value, so should I get new caps ~500uF to keep an overall 1000uF? Thanks \$\endgroup\$
    – Willinliv
    Commented Nov 29, 2023 at 12:16
  • 1
    \$\begingroup\$ @Willinliv Yes, for capacitors in parallel the total capacitance is the sum of the individual C values. When selecting caps for filtering the output of a flyback, the trick is it know the ripple current (RMS), and select capacitors that will provide the lifetime required at that current stress. Reputable manufacturers will provide that data, if the data does not exist then assume the worst. You can try to keep overall C the same as per the original design but that is secondary to achieving lifetime at the current stress. \$\endgroup\$ Commented Nov 29, 2023 at 23:31
2
\$\begingroup\$

The circuits are unrelated in a sense, but also coupled through the flyback transformer.

Both of these caps were 1000uF 16V, maybe there is a design error or manufacturing error, and cap model not handling enough ripple current was manufactured in for both circuits.

Also the fact that new ones you bought are of some specific brand means very little. Sure they could be quality capacitors, but if they have inadequate specs for the job they need to do, then being a good quality capacitor is not enough, as caps have various other parameters too, other than capacitance and voltage.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.