I have a capacitor bank consisting of five 3.3 mF electrolytic capacitors. (Itelcond 400 V 3300µF datasheet here.) The five capacitors are connected in parallel and there's a charging circuit with a 24VDC power supply, a 12 ohm resistor and a switch. (I'm testing the setup with 24 V, but plan to go up to live voltage via a rectifier.)


simulate this circuit – Schematic created using CircuitLab

I calculated the leakage current with the formula given in the datasheet, for an individual capacitor:

\$I_{f0} = 0.004CV = 0.004 \times 3300\ \mathrm{µF}\times 400\ \mathrm V = 5280\ \mathrm{µA} = 5.28\ \mathrm{mA}\$

Since the capacitors are connected in parallel, the total leakage current is 5 times that:

\$I_f=26.4\ \mathrm{mA}\$

However, when I close the switch to charge up the bank, there's a constant voltage drop of 8 V across R_v. That means there's a constant leakage current of 666 mA flowing.

When opening the switch, the bank also discharges much more quickly than it should. (With a theoretical constant leakage current of 26.4 mA, it should be mostly discharged only after ~10 seconds, however it's definitely quicker than that)

There's a couple of possible explanations I can think of:

  • One or more of the capacitors is faulty and leaks significantly more current than it should (although I did verify that the capacitance is in-spec, measured it with my multimeter).
  • The formula for the leakage current is only valid for the rated voltage of 400 V, whereas I'm testing it with only 24 V. However, it's counterintuitive to me that the leakage current would decrease if I go up to a higher voltage.

If anyone of you knows what the problem might be, let me know.

  • 1
    \$\begingroup\$ Interesting! It would be very useful to test it at slightly higher voltage to see if it’s linear, but I suspect a bad cap. Or perhaps if they are very old, you need to reform the dielectric layer inside them. For how long did you apply voltage? \$\endgroup\$
    – winny
    Mar 10, 2022 at 18:25
  • \$\begingroup\$ Just to confirm, the switch is indeed a mechanical switch right? That is in fact disconnecting the capacitor bank from the power supply? Not something like a switch that is powering off the power supply, right? \$\endgroup\$
    – DKNguyen
    Mar 10, 2022 at 18:28
  • 7
    \$\begingroup\$ Just to cover the obvious- are you absolutely sure you got the polarity correct on the capacitors? That's a very large amount of current. Something is seriously wrong. \$\endgroup\$ Mar 10, 2022 at 18:33
  • \$\begingroup\$ Also- that's 10W going into the capacitors. Whatever is absorbing that is going to be getting warm or hot, at least, depending on the physical size etc. At worse it may vent some electrolyte suddenly. \$\endgroup\$ Mar 10, 2022 at 19:02

2 Answers 2


Charge those capacitors to at least 200V for a couple of minutes, then discharge, and repeat the measurement. The leakage current you're measuring should not occur for properly formed capacitors. Measure the current individually for each capacitor - one of them may be "dead", or perhaps one or more are inserted in reverse.


Apparently I'm an idiot who can't tell plus an minus apart. Thanks to everyone for the helpful answers.

I somehow managed to mount the capacitors in the wrong polarity. This is especially bad since they are electrolytic caps. However, they seem to have survived my stupidity and still work fine.

The voltage drop across R_v is now at a much more reasonable 16 millivolts, i.e. there's a leakage current of only 1.3 mA (Although it will probably increase when I go up to a higher voltage).

  • 2
    \$\begingroup\$ I'm glad you didn't apply 200V to the caps. That would have been .. dramatic. The current should drop over time at bias. \$\endgroup\$ Mar 10, 2022 at 19:56
  • \$\begingroup\$ @SpehroPefhany Yep, that's one of the reasons I tested with a lower voltage at first, to verify basic stuff like the leakage current. And I'm very happy that I did. \$\endgroup\$
    – Zciurus
    Mar 10, 2022 at 21:17

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.