Film and Ceramic Capacitor Leakage Current

[chuckb]

I have been using my Keithley 617 Electrometer to measure the leakage current of large Electrolytic capacitors for several years. It has an internal +-102VDC source for biasing the Capacitor Under Test.
I was working on an error budget and I needed to know the actual leakage current of some small filter capacitors. I ran into some surprises with C0G ceramic caps and PE film caps.

The AVX and TDK C0G caps are very low leakage. Other brands of C0G have 20x more leakage and a large Dielectric Absorption. To be fair the leakage of the other caps is within their specifications. However, if you want some very good caps, see the attached pdf.

I do not have the time or budget to test all available brands and configurations. This is just what I have measured.
NOTE: This article is just to caution you to check all your components before you use them in critical applications. I just did a sample from one batch of parts. I may have a bad batch or a great batch of parts.

I looked into some WIMA film capacitors and was surprised at the rated leakage currents. I simply had not researched them before.
The WIMA MKS2 data sheet states >1000 Megohm-uF for the 6.8uF 50V part. I take that to mean over 147 Megohm or less than 68na of leakage at 20 deg C at 10V. Well, that’s a lot more than I assumed.

When I checked the WIMA MKS2B046801M00JS, 50V 6.8ufd small red film (Polyester film, PE) cap they 30 minutes it has 1500pa of leakage with just 10V applied at 27 deg C. After 12 hours the current had settled to 650pa of leakage. With 50V it had 22,000pa after I let it settle for 30 minutes.

I spent the next week checking a wide selection of new small capacitors. Then of course, I had to expand the testing to two temperatures.
When I warmed the MKS2 cap to 60 deg C (the temp in circuit) the leakage was 5.5na at 10V. With 50V it was 259na, wow.

A physically larger Panasonic 100V 10uF ECQ cap (PE) had 50pa a 10V. This settled to 10pa overnight at room temp. At 60 deg C the leakage current was 150pa at 10V.

What are your experiences with small (<100uF) capacitor leakage current?

[bsw_m]

Time ago I'm measured leakage of elcap Epcos B41859 100uF 35V. After 1 hour this cap has leakage 180nA at rated voltage and 11nA at 10V.
I can add measurement results as they are carried out.
To measure I'm using V7-49 electrometer.

[Kleinstein]

It is really difficult to tell apart true leakage and dielectric absorption. The settling over time is a characteristic of dielectric absorption current.  When testing with an applied external voltage source, one also has to care about drift of the voltage source - this would look like additional current.
A way to separate leakage from DA would look like first recording the apparent leakage with voltage applied to the capacitor, keep the capacitor charged for a longer time (e.g. 10 x the initial test, e.g. over night) and than discharge the capacitor and measure the apparent current from charge recovery after the same time. The 2 nd current would be a approximately (lower limit) to the DA contribution in the first test.

I have tested few small (2.2 nF)  C0G and PP , PS caps and found about comparable DA. The TDK brand C0G was considerable (e.g. 5-10x) lower DA.  Over the time scale tested (100 µs to 40 ms) the DA is dominating over leakage.

Polyester film is way higher DA than PP or C0G. So it can take considerably long time (e.g. more like days) to really see leakage.

[bsw_m]

Murata RPER71H104K2S1A03A after 5 hours settling time: 29pA at rated voltage. But this cap have big DA.

[bsdphk]

When I did something similar, I found that a lot of the actual leakage was crepage across the exterior plastic surface of the capacitors.

Wash the capacitors in isopropanol and distilled water and measure again, without touching them with your fingers.

The lowest leakage I found were high voltage (>=500V) wide lead-spacing (37.5mm) film capacitors.

Art of Electronics has useful wisdom on this topic.

[T3sl4co1l]

A few other data from John Larkin: https://www.dropbox.com/sh/m0m2zjljysms4yc/AACA2-BIGYDgHNRQN0PuaQOAa?dl=0

Tim

[chuckb]

Here are the results from 6 more capacitors I tested.

I checked a TDK X7R, lots of leakage and DA.
The leakage current for the KEMET X8L capacitor did not change much with temperature. It was high all the time. The leakage current for other caps typically changed 5-10x when the temperature changed from 31C to 61C.  The X8L leakage current only changed 3x. The current seemed to settle quickly like it did not have much DA but it may have just been the high leakage hiding the DA.

The WIMA Polypropylene cap (MKP4) had much better leakage (100x less) than the WIMA PE (MKS2) units.

The leakage of the 1.8u Panasonic PE caps compared to the 10u version was reduced by a factor of 5 as you would expect.

I checked a new Nichicon 63V 100u 1000u Low Leakage Electrolytic cap, UKL1J102KHD. After 1 day at 7.15V it had an steady state leakage of 10na. Then I did a day at 63V to form the dielectric. After I discharged and reapplied voltage the 7.15V leakage was 1.4na.

[chuckb]

One addition this time. It is a Wurth 0.033u, 50V, 1812 package. With it's low capacitance it achieved a very low leakage current.

I corrected the capacitance of the Nichicon low leakage cap in the last post.

[TiN]

I can also share my old results with wet slug tantalum XTV caps. It's the same cap that Jim Williams used in AN124 preamp.



Cap #1, XTV138T030P4L = 1.2 nA, 10V bias, measured by Agilent 41420A SMU
Cap #2, XTV138T030P4L = >50 nA, 10V bias, measured by Agilent 41420A SMU
Cap #3, XTV138T030R4L = 2.8 nA, 10V bias, measured by Agilent 41420A SMU
Cap #4, XTV138T030R4L = 3.5 nA, 10V bias, measured by Agilent 41420A SMU

[dietert1]

Connected a Wima MKS2 6u8 50 V, fresh from the bag. Exponential decay stopped at about 1,8 nA. After about 3 hours it sits at 480 pA, which is a confirmation of the original post. Ambient is 25 °C and i used a 1 MOhm resistor to measure the current with a FLUKE 8600 in its 200 mV range (resolution is 10 pA).

Regards, Dieter

[bsw_m]

Some new measurement for Wima MKP4 capacitors and Epcos B41858C5477M000 100uf electrolytic capacitor.
All info in attached graphs.
Also if You need more data, I'm attached raw data (first column is time in unixtime format, second column - current).
All measurement done by my V7-49 electrometer.

[mzzj]

Tested at 10 volts:
22nF OLD stock Suflex polystyrene: 0.3 pA and the measurement is starting to get quite  sketchy. (would need some shielding or I have to sit quietly without moving)
1.5uF 400v Wima MKC 4: 2pA after couple of minutes
Another 1.5uF 400v Wima MKC 4: 3pA (lazy to wait)
Philips MKP 1.8uF 250V <4pA
Evox MMK 4.7uF 100V 40pA

MMK(polyester) is behaving just like expected with 10x leakage current(or DA, didn't wait for ages) 

[bsw_m]

Wima MKP4 4.7uF test at 50V

[David Hess]

I have tested low voltage polypropylene capacitors for leakage at high temperatures and found considerable variation between manufacturers.  For a 0.47 microfarad capacitor, I needed less than 10 picoamps of leakage at 150F or so.  If you are looking for characteristics which are not guaranteed in the specifications, then this is just what you have to do.

[bsw_m]

David Hess Thanks for Your reply.
I am not looking for any special capacitor at the moment.
I make these measurements out of curiosity (and perhaps, as a reserve for the future, if I need a component with any characteristics that are not indicated in the datasheets, I will know which components should be paid more attention to). And also in order to share with the public. For the same reason, I provide not only the leakage current data at the end of the component measurement, but also the complete measurement graph. Perhaps this information will be useful to someone.
Unfortunately, I am not able to measure a large number of samples of the same type for two simple reasons - they need to be purchased, it takes a lot of time to measure them.

[JohnG]

I have tested low voltage polypropylene capacitors for leakage at high temperatures and found considerable variation between manufacturers.  For a 0.47 microfarad capacitor, I needed less than 10 picoamps of leakage at 150F or so.  If you are looking for characteristics which are not guaranteed in the specifications, then this is just what you have to do.

150F!?! That musta been a hella big polypropylene cap Oh, wait, I should read more carefully...

[David Hess]

I have tested low voltage polypropylene capacitors for leakage at high temperatures and found considerable variation between manufacturers.  For a 0.47 microfarad capacitor, I needed less than 10 picoamps of leakage at 150F or so.  If you are looking for characteristics which are not guaranteed in the specifications, then this is just what you have to do.

150F!?! That musta been a hella big polypropylene cap Oh, wait, I should read more carefully...

0.47 microfarads was quite big enough in a polypropylene part.

[bsw_m]

Hitano 0805N101J500NT. 100pF 50V NP0 capacitor. (8fA leak at 10V after 3 minutes).
In this test, I did not shoot the graph. 3 minutes after applying the test voltage of 10V, the current through the capacitor was 8fA.
With a test voltage of 50V, the leakage was 55fA

[serg-el]

After molding 12.2 V for 16 hours.
Self-discharge.
Not ceramic and film. Electrolytic and solid state.
For comparison and preliminary assessment.

[bsdphk]

Have you measured the leakage of the protoboard you are using ?

I would expect that to account for a relevant part of your result.

[Kleinstein]

The prototype boads are often not that bad with leakage. I have not measured it directly, but the leakage currents can be quite low ( pA range). So with the electrolytic capacitors this is likely still to a large part leakage of the capacitors and maybe the input current to the meter to measure the voltage.  For the initial phase (e.g. less than 20% drop) the dielectric absorbtion in the capacitors can also be an important factor.

[ZhuraYuk]

Here is the leakage of two 10uF polypropylenes from WIMA and JB capacitors. The WIMA is 4 times more expensive than JB.

[Kleinstein]

Much of the current seen with the PP caps could still be from dielectric absorbtion. It is at least hard to tell from the curves shown.
At least the intial part (e.g. the first 15 minutes) it quites clearly something transient and not really leakage.

[MiDi]

Bit OT:
1000µF 50V 85°C Panasonic series M-A @10V is down to 1nA after 3 days.
Leakage current can be approximated by a * t^-b after ~2s (and to at least 3 days), which corresponds to a straight line in a loglog plot.

[chuckb]

Continuing this thread. Attached is some more testing I did while locked in for COVID.

The leakage of film caps (maybe all caps) is nonlinear with voltage.
For example my favorite Panasonic ECQ series 10uF, 100V Polyester cap.
Changing the bias voltage from 10v to 50V results in 5-6 times more leakage.
Changing the bias from 50V to 100V causes 8 times the leakage.

Some spark gaps were tested with fa leakage.