I have a number of HV (100V-600V) film caps that I've removed from old equipment. I set up a crude leakage tester by putting a microammeter in series with an adjustable DC power supply. For each cap, I ran it up to the rated voltage. They fell into roughly three groups: First, some showed leakage right at the resolution limit of my meter, showing 0.00-0.02uA. Second, some showed leakage around 0.2-0.3uA. And third, some showed leakage around 1uA.
I understand that ideally, a capacitor should block DC perfectly, with no leakage. But these leakage amounts seem so small that these should work fine in a lot of positions in most circuits (say, a tone stack with no DC, or a cathode bypass cap), assuming that their capacitance value is still in spec. So what would you consider the upper limit for leakage current in a film cap?
I understand that ideally, a capacitor should block DC perfectly, with no leakage. But these leakage amounts seem so small that these should work fine in a lot of positions in most circuits (say, a tone stack with no DC, or a cathode bypass cap), assuming that their capacitance value is still in spec. So what would you consider the upper limit for leakage current in a film cap?
A typical worst case would be a HV coupling capacitor into a 1M grid resistor.
If the following grid DC voltage should be < 0.1VDC maximum,
then 0.1V/1M = 0.1uA capacitor leakage maximum.
If the capacitor had 1uA leakage, then the following grid DC voltage across 1M would be 1V.
That would clearly be too much leakage.
If the following grid DC voltage should be < 0.1VDC maximum,
then 0.1V/1M = 0.1uA capacitor leakage maximum.
If the capacitor had 1uA leakage, then the following grid DC voltage across 1M would be 1V.
That would clearly be too much leakage.
Last edited:
Jim Williams described a method of leakage measurement : see appendix B
https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf
https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf
Last edited:
I'd say your measurement floor: datasheets have limits for understandable reasons, and the limitations of ATE's, but a good plastic film cap should have a minuscule, almost undetectable leakage.
If it shows some, it means it is damaged in some way, and should not be used in HV situations.
Probably OK for zero-bias cases though
The most probable cause of leakage in an old cap is the dirt and grime accumulated on the outside - I hope you gave them a good wash and dry before testing.
The film capacitors with epoxy cases used to crack round the leads if badly handled and let moisture in. This lead to leakage and eventual failure
I bought some 4 uf film caps from a surplus house that were obviously rejects. No markings. They measured 300000 ohm at 2 v on 20 meg scale of dvm. That is .66 ua.
Probable cause, overlap of metal film when spraying before winding up.
As the case was not the problem, leading to moisture leakage, I was able to use them in my application. That kind of leakage should be linear over voltage I was exposing them to, up to 180 v. My circuit had a source impedance of 47 kohm, much less than 300 k.
Probable cause, overlap of metal film when spraying before winding up.
As the case was not the problem, leading to moisture leakage, I was able to use them in my application. That kind of leakage should be linear over voltage I was exposing them to, up to 180 v. My circuit had a source impedance of 47 kohm, much less than 300 k.
Last edited:
I reckon for high voltage caps once leakage starts to increase its only going to get worse and worse and could ultimately lead to catastrophic failure. Of your caps I'd suggest only retaining the best measuring category for HV use and derate the others.
However leakage does depend on total capacitance, for the same type/voltage of capacitor, so its possible the variation is due to this.
There are probably some more sophisticated ways to characterize leakage by measuring the noise spectrum of the leakage to determine whether its white or 1/f or popcorn/sporadic.
However leakage does depend on total capacitance, for the same type/voltage of capacitor, so its possible the variation is due to this.
There are probably some more sophisticated ways to characterize leakage by measuring the noise spectrum of the leakage to determine whether its white or 1/f or popcorn/sporadic.