Measuring capacitor leakage

[argintviu]

I have a few recovered polypropylene capacitors (1uF ... 3.3uF) that I'd like to use but don't fully trust them. They are all in spec. capacitance and ESR wise, but to be safe, it would be nice to also know their leakage. Reading from various sources, I settled for the following procedure:

- Charge the cap to 7V for 1 hour
- Measure the exact voltage across it then disconnect cap and start a stopwatch
- Let it rest for 1h ... 24h
- Read stopwatch time, briefly connect the cap to the attached voltage follower circuit and take a reading with multimeter
- Calculate cap leakage: ILeak = C × (InitialVoltage - FinalVoltage) / RestTime

The lowest input bias current opamp I have is TLC271 with 0.6 pA typ. and 60 pA max. @ 25 deg. C.

Would this be a good way to get an idea if the caps are still good?

[Edit]

Modded the circuit (forgot to add decoupling cap).

[kripton2035]

here are some examples of capacitor leakage testers, to compare with your design.
http://kripton2035.free.fr/LCR%20meters/lcr-capaleakmete.html
http://kripton2035.free.fr/digital%20esr/esr-sprut-de.html

[Vovk_Z]

Would this be a good way to get an idea if the caps are still good?

It depends on what you expect from them and what is their destination. For example, as for me, they are probably good as is, or they are good if my multi-meter shows OL (overload) in resistance measurement mode. Or if they are measured fine with RLC-meter. 

[RoGeorge]

If you have a Gohm impedance voltmeter, you can put the discharged capacitor in series with a power supply and a high impedance DMM on DC volts.  Not all DMM have a setting for high impedance.  Mine (mtx3283) can be set set to either 10Mohm or 1Gohm (for ranges less than 1V only).

With all 3 in series, the DMM on high impedance setting is like a 1Gohm resistor, so the indicated voltage is proportional with the nA of leakage current through the capacitor.  If you read 1V on a 1Gohm DMM, it means the leakage current is 1nA, and 1mV corresponds to 1pA.

[argintviu]

It depends on what you expect from them and what is their destination.

They've been inside power supply units under pulse stress and I want to use them inside low voltage RC filters. I think leakage measurement is a good way to know if the insulation is still intact.

If you have a Gohm impedance voltmeter, you can put the discharged capacitor in series with a power supply and a high impedance DMM on DC volts.

Unfortunately I don't have a high impedance voltmeter, my best one is 10 MOhm. I have an UT61E which theoretically can go into Gohm territory on the millivolt scale, but the actual impedance is unknown so it's no use. Thanks for the idea though.

[RoGeorge]

You can try anyway, the user manual says 3G\$\Omega\$ for the mV range
https://www.batronix.com/downloads/UNI-T/UT61_Manual.pdf (page 17)

[argintviu]

Interesting. I don't remember seeing that when browsing the manual . I'll give it a try.

[Vovk_Z]

They've been inside power supply units under pulse stress and I want to use them inside low voltage RC filters. I think leakage measurement is a good way to know if the insulation is still intact.

In such case I would rather be concerned with their capacitance but not insulation. If they still didn't blown up then theirs insulation most likely is good enough. Film insulation don't deteriorate usually, unless it is some special type of bad known RIFA cap.
But capacitance of film caps may deteriorate under mains voltage being self-healed.

[BrokenYugo]

I would think the usual 10 megaohm DMM would plenty adequate for a current shunt in a full voltage capacitor leakage test, 1mV across 10M is 100pA, or 10mV/nA.

[helius]

Leakage is the usual metric for electrolytic capacitors; for film capacitors, a reciprocal metric is used, called insulation resistance.
Testing R_I is a function of some Hi-Pot testers, from suppliers such as Haefly or Biddle. If you only care about R_I up to 1 KV, most "Megger" brand testers will measure it.

[Kleinstein]

In principle the circuit can work. The input current from the OP-amp can still be a bit of a problem. It is still much better than most DMM inputs.
Besides capacitor leakage there is still the OP-amp input current and dielectric absorption (DA). The DA part gets less the longer the capacitor was charged before the test. PP capacitors are quite good in this respect and should not show much DA effect.
For comparison one could also test a known good capacitor.

[argintviu]

In such case I would rather be concerned with their capacitance but not insulation.

Indeed, some of them had their capacitance degraded severely, e.g. from 3.3uF down to 600nF, that's why I wanted to test the seemingly good ones more thoroughly.

Leakage is the usual metric for electrolytic capacitors; for film capacitors, a reciprocal metric is used, called insulation resistance.
Testing R_I is a function of some Hi-Pot testers, from suppliers such as Haefly or Biddle. If you only care about R_I up to 1 KV, most "Megger" brand testers will measure it.

Cool testers, although a bit too much for the task at hand .

I would think the usual 10 megaohm DMM would plenty adequate for a current shunt in a full voltage capacitor leakage test, 1mV across 10M is 100pA, or 10mV/nA.

Tried this but couldn't get a stable result (possibly because I used low voltage).

In principle the circuit can work. The input current from the OP-amp can still be a bit of a problem. It is still much better than most DMM inputs.
Besides capacitor leakage there is still the OP-amp input current and dielectric absorption (DA). The DA part gets less the longer the capacitor was charged before the test. PP capacitors are quite good in this respect and should not show much DA effect.
For comparison one could also test a known good capacitor.

Thanks for the input. Next time I'll buy a LMC662 like Bob Pease used in "What's All This Capacitor Leakage Stuff, Anyhow?". To account for the DA, I'll keep them powered a bit longer, 3 hours for 3.3uF 630V should be enough.

[MathWizard]

I did to make some DC leakage tester too. What about testing the health of ceramic SMD caps ? I have a few on a PCB from a SMPS, and IDK for sure their spec's, but I know I exposed them to a lot of heat desoldering stuff on the PCB.

I have an LCR meter, but I can never find out what's good or bad for small ceramics, no one knows or ever talks about it I guess. So maybe I should try leakage testing them, and maybe at different temperatures. But IDK what kind of caps they really are, like NPO, XR7, or their real voltage rating or any other spec's. I'll have to learn what's required, to choose a replacement.

[tatel]

I have seeen somewhere a guy using aneng MH12 to check capacitor leakage up to 1000V?

[argintviu]

For ceramic SMD caps, thermal cycling could make them develop small cracks.

Differentiating between them is not difficult. One could wire / attach the capacitor to a tester and get the soldering iron close to the cap. X7R and the like will drop their capacity as they heat up. For example doing this to a 100nF X7R made it drop to some 83 nF when heated while a C0G remained perfectly stable.

Leakage wise, I'll let others comment more, but C0G / NP0 should behave more like film caps, so maybe you could use the method presented in post 1. Just make sure to keep them powered for more time as they exhibit more dielectric absorption than film. One big problem, as you said is that their voltage rating is unknown. Maybe try to analyze the PCB and guess under what voltage they were used.

[Kleinstein]

The dielectric absorbtion if C0G capacitors is usually quite good, about on par with the better film capacitors (PS or PP).  C0G class is about a low TC and the exact composition can vary between series and manufacturers.  For a TDK brand C0G capacitor I even measured dienectric absorbtion about a factor 5 lower than PP or PS film capacitors and thus about on par with the exotic PTFE film capacitors.

With the usually small values for C0G capacitors testing them can however be challanging as amplifier leakage and possibly the parasitic capacitance on a PCB can become relevant.

I remember a NASA publication on checking ceramic film capacitors via leakage.

[argintviu]

Finally managed to complete the tests for some of the capacitors and it seems to have been a good idea as one of the bigger caps has big leakage (picoamperes while the others have femtoamperes). For the test, I used the circuit in post no. 1 to measure the voltage before and after the ~10 days rest period.

Here's the list:

Code: [Select]

1. 1034 nF 250V new, ESR 0.18 ohm, start voltage 7.000V, end voltage 6.586V after 968100 seconds => leakage is 0.000001034 x ((7 - 6.586) / 968100)     = 442.55 fA
2. 1048 nF 250V new, ESR 0.19 ohm, start voltage 7.002V, end voltage 6.612V after 962580 seconds => leakage is 0.000001048 x ((7.002 - 6.612) / 962580) = 424.44 fA
3. 1065 nF 250V new, ESR 0.19 ohm, start voltage 7.005V, end voltage 6.884V after 958050 seconds => leakage is 0.000001065 x ((7.005 - 6.884) / 958050) = 134.19 fA
4.  981 nF 100V old, ESR 0.19 ohm, start voltage 7.012V, end voltage 6.879V after 904355 seconds => leakage is 0.000000981 x ((7.012 - 6.879) / 904355) = 144.2  fA
5. 2162 nF 250V old, ESR 0.1  ohm, start voltage 7.000V, end voltage 6.847V after 974100 seconds => leakage is 0.000002162 x ((7 - 6.847) / 974100)     = 339.43 fA
6. 3621 nF 630V old, ESR 0.08 ohm, start voltage 7.000V, end voltage 0.9487 after 979430 seconds => leakage is 0.000003621 x (7 - 0.9487) / 979430)     = 22.37  pA


[wizard69]

Unless $$$$$ are a serious issue I wouldn't bother with used capacitors of any mainstream type.    Decent capacitors, at least the main stream types, are so cheap that it really isn't worth the trouble.   This especially if caps have come out of hard service or are extremely old.   

[argintviu]

You are right, it's not worth the trouble from an economic standpoint. For me, this was more of a learning exercise, but I do feel warm and fuzzy when recycling parts .

[David Hess]

I would think the usual 10 megaohm DMM would plenty adequate for a current shunt in a full voltage capacitor leakage test, 1mV across 10M is 100pA, or 10mV/nA.

That is how I do it, and a voltmeter input has considerable over-voltage protection so there is no worry there.