Film and Ceramic Capacitor Leakage Current

[Echo88]

Very nice, thanks
I set up my K617-testbox last week again also to measure some more FETs, caps and SSRs.
Is it possible to get your list as an excel-file/similar, to which i could add my measurements? Just if thats okay for you.

[3roomlab]

I entered some of the data from various threads here and there (mF=1000uF)
I included some supercaps (datasheet values)
And there is the uA/Farad leak column value for rough comparison
The UKL 2.2mF  is really something

*the ECQ-E1106 in the list is 100v (2uA/F), the 250v E1106KF would have 0.8uA/F

[chuckb]

Very nice, thanks
I set up my K617-testbox last week again also to measure some more FETs, caps and SSRs.
Is it possible to get your list as an excel-file/similar, to which i could add my measurements? Just if thats okay for you.

Here is the Editable file. Add Change the date at the end of the file name so we can keep track of versions.
Add you name on the left column if you add an entry.
I use a Keithley 617, if you use something else make a note in the comment column.
An organized person would probably have a seperate TAB for each dielectric.

Thank you for contributing.
Chuck

[chuckb]

Here's the file.

[chuckb]

When you "form" the dielectric of an Electrolytic capacitor does it stay formed? Can you form a batch of capacitors then put it on the shelf till you need it?

I attempted to answer that question with a Nichicon low leakage 2200ufd 16V cap. This cap took about a month at 16v for the leakage to reduce from the 1 day value of 200na to less than 20na. This was at 32 deg C. Higher voltage rated Nichicons (63v and 100v) seem to form within a week.

For this cap I applied rated voltage for 20 days, applied 0v while traveling for 2 weeks, then reapplied rated voltage for 10 more days. Within a day of reapplying voltage the current was the same as before I shut it down for travel. So the Dielectric Forming was not lost.

Then I tested leakage at 10V. The leakage dropped from 17na to 4na as the voltage changed from 16V to 10V, and a few more days passed.

I will check this same capacitor for leakage next year, if I remember.

[Kleinstein]

The build-up / degradation of the dielectric layer is expected to be a little asymmetric, especially as the idele state is with zero voltage and not a reversed voltage.
AFAIK the formation is from genrating new oxide layer by oxidation. The degradation is more by the formation / groß of holes (e.g. at defects / grain boundereies) in the oxide layer by movement of oxide to a different, slightly more favorable position near by.  So the 2 processes are quite different. Normally the formation should last quite long, like a few years, but things can vary with temperature and type.

The other superimosed effect is the dielectric absorbtion. This part should be similar for charge / discharge - here the 14 days off can reverse most of the effect of 20 days under voltage.

[Echo88]

Attached is the reformatted file from chuckb, with added measurements from my side.
I hope i didnt delete or change any measurement-infos that belonged together from chuckbs-measurements.
The measurements take a lot of time, so i havent added much yet. The list may contain errors.
The leakage current/capacitance column was just done by me to have some figure of merit when intercomparing capacitors.

Especially measuring film-caps in the µF-range have shown to be really difficult as even a roomtemp-stability of +-1°C is apparently enough to mess with the pA-measurement due to K617-TC and filmcap-TC: the current measurement then reads for example -40pA because the cap has a slightly higher voltage.
Usually i use a 33M-resistor in series with the capacitors, as thats necessary to get stable low current noise readings. What resistor and procedure do you use chuckb?
Forming some 400V electrolytic caps meanwhile and measuring some supercaps for additional fun.   

[bsw_m]

What resistor and procedure do you use

Hello! I'm use this setup for these measurements. Best if diode have leakage current as low as possible.
The resistor may vary from 10kOhm to 1MOhm.

[chuckb]

... What resistor and procedure do you use chuckb? ...

My setup includes a large metal cast box, an RC filter for the Applied Voltage and two Kapton heaters to change temperature.

I use a a Hammond 1550H enclosure, it is roughly 100mm tall, 140mm wide and 200mm long. The current output connector is a Triax connector with two small alligator clips on the inside. Smaller caps are suspended in mid air by the connector leads. Larger caps rest on a teflon sheet covering the bottom of the box. The metal cover of the box has to be fastened down or a weight applied to keep it from moving. Any moving metal will add noise to the readings.

The Excitation voltage from the K617 passes through a 200k resistor and a 100ufd PE cap for 20 sec of RC filtering. This reduces the apparent current flow caused by voltage noise on one side of the cap. All charging and discharging of the cap under test happens through the 200k resistor.

The metal box sits directly above the K617 to keep the leads short. The triax is fixed in place so it does not move around and cause noise.

I attached two round 75mm dia. kapton heaters to the outside of the metal box. At 10V they draw 2 amps total. This 20W of heat raises the box temperaure up to 63 deg C. If the box is 63C the capacitor inside will eventually reach 63C. I usually wait an hour for the cap to catch up to box temperature before I start a test. The box is covered with a heavy cloth for insulation. I measure the box temperature with a handheld Fluke Termocouple thermometer. I do not try to measure the capacitor temperature directly as this would upset the current leasurements.
 
With smaller caps the meter input current temperarily changes when the K617 changes ranges. With some caps, if I leave it in Auto Range, the meter can go into RANGE oscillation. For these caps I watch the display and manually change the ranges. The meter could use a little more hysterysis on the Auto Range change.

I have a Agilent GPIB to USB adapter and software to collect and store the data from the test.

[ArgyllGargoyle]

Have you tested any tantalum capacitors for leakage?
I recall having good luck with the orange Vishay Tantamount variety (e.g. 597D family).

[chuckb]

In general, Tantalum caps have a lower leakage current than normal aluminum electrolytics. I have checked several Tantalum caps and I'm finishing up a week of looking at one right now. I will update the list in a few days.

I'm testing a 35V 6.8uF, series 150 from Vishay (Sprague). The leakage current is very nonlinear with applied voltage.
At 32 deg C,

Bias voltage
3V 4pa
5V 20pa
10V 100pa
35V 1.5na
 

[T3sl4co1l]

I once did an accelerated aging process, which seemed to help; I took a few 1-4.7uF 100V metal case (M39003 style) dry tantalums, biased them to near 100V (I forget if it was over or under the rating), and sat them inside a power resistor cooking at, something elevated, 100, 130°C, something like that.  Let that sit for a couple days while monitoring leakage.  Then cooled back down and tested at lower voltage.

I forget what the exact figures were though, or other relevant information (brand, production date, etc.).  It was good enough for my application (a little sample and hold circuit), an improvement over polypropylene.

Tim

[TimFox]

I once did an accelerated aging process, which seemed to help; I took a few 1-4.7uF 100V metal case (M39003 style) dry tantalums, biased them to near 100V (I forget if it was over or under the rating), and sat them inside a power resistor cooking at, something elevated, 100, 130°C, something like that.  Let that sit for a couple days while monitoring leakage.  Then cooled back down and tested at lower voltage.

I forget what the exact figures were though, or other relevant information (brand, production date, etc.).  It was good enough for my application (a little sample and hold circuit), an improvement over polypropylene.

Tim

Was that an improvement in leakage current over polypropylene, or just an improvement in size/footprint?
According to Vishay, a 5 uF polypropylene capacitor should have a leakage resistance of about 20 G (100,000 M -uF), falling by a factor of 100 from 20^o C to 100^o C.

[T3sl4co1l]

Improvement in time constant; the others I had used (including polystyrene) just didn't perform well.  I'm not entirely sure where the leakage currents were in there, but it didn't seem to improve with capacitance, across film types.  For a while, I had a 10nF PS in there, but almost 1nF (C0G) to 100nF (PE, PP) didn't seem to make much difference, strangely.  I also don't have any PP in ~uF handy, that would fit in the space -- so the tant was a reasonable choice.

One-off proto thing BTW, not a production item.

Tim

[Kleinstein]

For a S&H circuit the other problem is dielectric absorbtion. Already mica and polyester capacitors are thus not a good choice. Electrolytic capacitors usually have very high dielectric absorbtion.
The case with a very long hold could be a reason to look for rather large capacitance. Other leakage paths (e.g. FET or OP-amp) could be a reason that a relatively large capacitance is needed and very low leakage of the capacitor not really needed.

Build up of the oxide layer at a higher temperaure makes sense, as a thicker and stable oxide can help to keep leakage low, though possibly at the price of a reduced capacity.

[T3sl4co1l]

Not actually much of a problem in that one, as updates were small changes. But a big deal for general applications of course!

I don't think I had tried an electrolytic, assuming they'll be too leaky (and even with the small changes, the absorption probably sucks anyway).  With some of the numbers here, maybe I should've!

Tim

[MiDi]

Attached some leakage results for different film caps @100V.

[MiDi]

Attached some DA results for different film caps discharged from 100V to 0V.

[MiDi]

Attached some leakage results for different film caps @10V.

[MiDi]

Attached some DA results for different film caps discharged from 10V to 0V.