Measuring capacitor leakage

[vindoline]

I'm testing 12 Nichicon KL series series caps right now. This is Nichicon's "low leakage" type. I'll report the results when I have them - probably this weekend.

[vindoline]

OK. As promised here are my initial capacitor leakage results. All of the caps in this test were brand new and purchased from Mouser. I Used Nichicon UKL series 25V (2 @ 3300 uF, 5 @ 2200 uF, 5 @ 1000 uF).

To start all of the caps were connected in parallel on a breadboard and charged to 24.5 V via a 1 K resistor and a PD TP340 power supply. After the caps had charged to near 24.5 V the resistor was removed. The caps were allowed to form at 24.5 v for about 2 days. The caps were then discharged for a day. At this point the circuit was modified to give each cap its own 1 Meg resistor to ground to monitor the voltage drop (and thereby leakage). The caps were then charged to 10.00V. The total charging/leakage current was monitored by a Keithley 196 DMM. Once the total charging current was in the microamp range, the 1 Meg resistors were shunted by jumpers to speed up final charging. After about 3 days the total leakage current was 100 nA. At this point, the jumpers were removed from the 1 Meg resistors and the individual cap leakage current calculated by measuring the voltage drop across the resistors with an HP3456A DMM (1 mV = 1 nA). The results are:

1.  3300 uF    3.60 nA
2.  3300 uF    2.78 nA
3.  2200 uF    0.63 nA
4.  2200 uF    0.19 nA
5.  2200 uF    0.04 nA
6.  2200 uF    0.25 nA
7.  2200 uF    0.17 nA
8.  1000 uF   17.14 nA
9.  1000 uF   10.65 nA
10. 1000 uF   37.22 nA
11. 1000 uF   14.10 nA
12. 1000 uF   12.54 nA

The real surprise for me is how poorly the 1000 uF caps performed. For some reason, I expected the lower value caps to have lower leakage. The 2200 uF caps appear amazing.

**edit: corrected pA to nA due to my inability to do math!**

[Andreas]

. At this point, the jumpers were removed from the 1 Meg resistors and the individual cap leakage current calculated by measuring the voltage drop across the resistors with an HP3456A DMM (1 mV = 1 pA).

Hmm,

shouldn´t it be 1 mV/1 Meg = 1 nA instead of 1 pA? if you measure the voltage across the 1 Meg resistor?

with best regards

Andreas

[vindoline]

. At this point, the jumpers were removed from the 1 Meg resistors and the individual cap leakage current calculated by measuring the voltage drop across the resistors with an HP3456A DMM (1 mV = 1 pA).

Hmm,

shouldn´t it be 1 mV/1 Meg = 1 nA instead of 1 pA? if you measure the voltage across the 1 Meg resistor?

with best regards

Andreas

Yup, sorry. Corrected. Lack of coffee!

[DC1MC]

You've got a real winner on 5.   

Could you please use a cheap ESR meter to see if there is a correlation between these measured leakage and ESR ?

 Cheers,
 DC1MC

[splin]

1.  3300 uF    3.60 nA
2.  3300 uF    2.78 nA
3.  2200 uF    0.63 nA
4.  2200 uF    0.19 nA
5.  2200 uF    0.04 nA
6.  2200 uF    0.25 nA
7.  2200 uF    0.17 nA
8.  1000 uF   17.14 nA
9.  1000 uF   10.65 nA
10. 1000 uF   37.22 nA
11. 1000 uF   14.10 nA
12. 1000 uF   12.54 nA

The TP340 24h stability spec @10V is 3mV and the temp coefficient is 2.3mV/C. Given the 3300uF/1Mohm time constant is nearly 1 hour, it is very difficult, if not impossible to make accurate sub nA leakage current measurments with this setup - 1mV of change in the power supply voltage translates to 1nA of apparent leakage current if the P/S voltage change occurs much quicker than the time constant.

DA makes it worse as the time constants are much longer. Every measurement will reflect the P/S ripple, drift and LF noise over the last hour or so. (I guess your meter will reject the 100/120Hz ripple and higher frequency noise when measuring the DC voltage across the 1Meg resistor). Your measurements are probably OK down to the nA level but not the 2 decimal points you show. A bit like measuring sea level rise by making spot measurements in the face of tides and waves.

The TP340 isn't bad but it would be much better to use a precision reference, even one of those dirt cheap AD584 ebay ones. Of course you could monitor the P/S voltage if you have a good enough meter to verify the P/S drift is not significant. Even then, ideally you would want to average the leakage current measurements over long periods - several hours at least, preferrably a day or more.

Alternatively, just open circuit the capacitors and measure the voltage every day for a few days, weeks or longer, but make each measurement as short as possible if your meter has 10M ohm input impedance, especially for the smaller capacitors - or compensate by estimating the charge lost during each measurement.

Don't forget to note the temperature as the electrolytic's leakage increases significantly with temperature. The change in dimensions of the capacitor with temperature will likely also change its voltage given that the plate spacing changes but the charge doesn't.

[EDIT] Added comment about P/S ripple.

[Kleinstein]

....

Alternatively, just open circuit the capacitors and measure the voltage every day for a few days, weeks or longer, but make each measurement as short as possible if your meter has 10M ohm input impedance, especially for the smaller capacitors - or compensate by estimating the charge lost during each measurement.

Don't forget to note the temperature as the electrolytic's leakage increases significantly with temperature. The change in dimensions of the capacitor with temperature will likely also change its voltage given that the plate spacing changes but the charge doesn't.

[EDIT] Added comment about P/S ripple.

A agree that open circuit measurement is likely the better way.

With electrolytic caps, there is not much change in plate spacing as it is the surface oxide layer, but there can still be some effects from temperature and air pressure, moving the cap. This can make the voltage go up or down.

[vindoline]

The TP340 24h stability spec @10V is 3mV and the temp coefficient is 2.3mV/C. Given the 3300uF/1Mohm time constant is nearly 1 hour, it is very difficult, if not impossible to make accurate sub nA leakage current measurments with this setup - 1mV of change in the power supply voltage translates to 1nA of apparent leakage current if the P/S voltage change occurs much quicker than the time constant.

I was concerned about the PS stability as well. My TP340 performs significantly better than the specs you quote above. I'm seeing about a 200uV drift over 18 h, and if all of that is due to Tc, the Tc is about 240 uV/C

Every measurement will reflect the P/S ripple, drift and LF noise over the last hour or so. (I guess your meter will reject the 100/120Hz ripple and higher frequency noise when measuring the DC voltage across the 1Meg resistor). Your measurements are probably OK down to the nA level but not the 2 decimal points you show. A bit like measuring sea level rise by making spot measurements in the face of tides and waves.

I completely agree that the measurements are not accurate to two decimal points, however I do think that the test is valid to compare the relative leakage of the group of caps I'm testing. All of the voltage measurements were taken within several minutes, after the system was allowed to stabilize for about 3 days. The purpose of the exorcise is to choose the best input caps for Andreas's .1-10 Hz amp, not to obtain a theoretically best accurate measure of leakage.

Alternatively, just open circuit the capacitors and measure the voltage every day for a few days, weeks or longer, but make each measurement as short as possible if your meter has 10M ohm input impedance, especially for the smaller capacitors - or compensate by estimating the charge lost during each measurement.

I have a Keithley 616 electrometer with about 10^14 input impedance. While it might be fun to measure the caps open circuit, it would take an unreasonable amount of time and I want to actually build the amp someday!

Thank you though for your careful thinking on the subject. You've brought up some excellent points!

[jfphp]

A very useful -- but old -- instrument for capacitor leakage mesurement  is the HP 4350A.

[BFX]

And what about using SMU for example Keithley 2400 is able measure current in pA.

[vindoline]

You've got a real winner on 5.   

Could you please use a cheap ESR meter to see if there is a correlation between these measured leakage and ESR ?

 Cheers,
 DC1MC

Sure! Here it is:

Cap   Nominal   Measured   Leakage   ESR   D
1   3300   uF   3.28   mF   3.60   nA   0.0   0.084
2   3300   uF   3.26   mF   2.78   nA   0.0   0.080
3   2200   uF   2.28   mF   0.63   nA   0.0   0.089
4   2200   uF   2.26   mF   0.19   nA   0.0   0.086
5   2200   uF   2.28   mF   0.04   nA   0.0   0.084
6   2200   uF   2.27   mF   0.25   nA   0.0   0.087
7   2200   uF   2.27   mF   0.17   nA   0.0   0.085
8   1000   uF   978.9 uF   17.14 nA   0.1   0.085
9   1000   uF   974.3 uF   10.65 nA   0.1   0.082
10   1000   uF   974.8 uF   37.22 nA   0.1   0.083
11   1000   uF   976.0 uF   14.10 nA   0.1   0.089
12   1000   uF   979.5 uF   12.54 nA   0.1   0.085

All the measurements were made with a DER DE-5000 LCR meter. I'm afraid that there doesn't seem to be enough resolution to find a correlation 

[Vtile]

These current measuring setups in nA/pA range are also a good demonstration of the pressure sensitivity of the capacitors.

The reformation current behavior of the oxide / dielectric in some metal sheet based (small) capacitors that have been sitting at the drawer for a decade or longer are wild. I'm using my Manhattan mk1 10.0000 V reference and 100 Meg needle FET-VOM in series configuration and the needle doesn't know where to settle with some caps. 

[Echo88]

@Vtile: Do you mean atmospheric pressure dependence?
@e61_phil: Thanks for the tip with the necessary serial resistor for the capacity measurement with the Electrometer. With 1M in series to the measured 40µF foil cap i get about 40nApp noise (way too much for qualitative measurements of different caps) and with a 1G-resistor in series i get about 50fApp.
Just a little bit more than the electrometer-noise itself including triax-cable which is about 30fApp. 

[Vtile]

@Vtile: Do you mean atmospheric pressure dependence?

Pressure source doesn't matter (I would assume), if it translates to mechanical force and then again internal pressure. In my case I just poked the capacitor with finger (which resulted a change of the electron flow) and after that repeated it with a few tools to verify it were pressure.

Long time ago I did read interesting article (from someones homelab) about the ideas how the DA have also mechanical aspects in electrolytics (or let say caps with soft dielectric materials as I can't recal), which makes sense considering the fluid is disturbed and also the distance (etc.) of the plates changes.

[e61_phil]

Is it really a constant current flow or is it a current flow because you changed the capacitance while pressing with your fingers?

What happend if you release the pressure? If the effect comes from changes in capacitance, than the amount of current increase should become a current decrease.

[Kleinstein]

Using the fingers to apply "pressure" would also add a thermal disturbance.

[Vtile]

Is it really a constant current flow or is it a current flow because you changed the capacitance while pressing with your fingers?

What happend if you release the pressure? If the effect comes from changes in capacitance, than the amount of current increase should become a current decrease.

What do you mean about constant? Yes, it is change of capacitance I would assume. Not induced spike / ESD, while my throw together "indicative measuring setup" is far from optimized nor my measuring device even close to "metrology grade".

Using the fingers to apply "pressure" would also add a thermal disturbance.

No this is not temperature (at least not conducted), but good point to remember for hobbyist like me.

Here is a short video clip, as you can see this far from professional setup, but merely indicative. The range is +-2.5mV over aprox. 100MOhms, center zero. For professionals this is nothing new, but us newb on/off hobbyist these at times comes around the corner. Some types are more sensitive than others, like the first polymer film cap I were testing (poking around after a workday) originally could be used probably as a microphone so sensitive it is.
https://youtu.be/3HFX_nMm6D0

 

[Kleinstein]

I had not expected that much effect of mechanical pressure on a film cap. So an interesting video - at least the first half.

What type of capacitor is it, and what voltage ?

[e61_phil]

Great video! Thanks a lot.

With constant I meant if the current decreases again under constant pressure to determine if it is a leakage or capacitance effect.

[Vtile]

I had not expected that much effect of mechanical pressure on a film cap. So an interesting video - at least the first half.

What type of capacitor is it, and what voltage ?

Hehe.  I just left that there that someone might have been interested where the needle moves after.

DC Film capacitor, PP, 0.33uF, 250V - Farnell: 2529137
The pressure applied is much more than what someone would apply to component like this in any normal use. The old film capacitor 'mic' without a body on the other hand doesn't like pressure at all edit: (...nor temperature). The actual leakage might be something else than seen on the video clip because stray fields and the cap had been electrified for an hour or so total.

@e61_phil Thanks.  I'm running a new experiment, let see.
Edit. ....and managed to delete the video, while editing. 

[Vtile]

I did re-run the experiment I managed to delete from the recorder yesterday, a much quicker this time. 
It seems that the current jumps about 2.5 pA under 260 grams of weight and returns back to starting point. When the pressure is removed it go down the same amount and returns.

Funny meter to play with in the bench and at the field. Interesting enough the wooden board (HV test box) behaves much better considering ESD than my "ESD" mat.   
https://youtu.be/K4KDn3gZjao