Using Models to Predict Capacitor Leakage Current

[PA4TIM]

I think testing ceramic caps is not as easy as it looks.
Many ceramics have a big voltco and piezo electric effect. Maybe the voltage pulls the plates together, that increases the leakage, that causes some reaction in the material, voltage drops, the plate distance increases and current drops, this goes on until the voltage becomes high enough to overcome the force and hold the plates and the current increases further.
Besides that, the sudden "release" could cause a voltage by the piezo electric effect. Some ceramics have a rather big tempco.

And referring to the pdf about partial discharge, maybe there are bubbles in the material .

Maybe also some effect as mica's have on very low frequencies, as if they can not follow the change in polarity. 

This could also explain the non lineair behavior in tests.

In electrolytic caps there is the chemical problem as an extra.

I am curious what would happen if we make a big 2 plate cap and measure the movement, force and leakage. And then place several dielectrics between the plates (like a ceramic tile  ) And then look if they behave also "strange" But I do not have the gear to do those measurements.

I made something like that years ago to test several materials behave as dielectric. (not with the sensors) http://www.pa4tim.nl/?p=2929 the last picture on the page.

[joeqsmith]

I think testing ceramic caps is not as easy as it looks.
Many ceramics have a big voltco and piezo electric effect. Maybe the voltage pulls the plates together, that increases the leakage, that causes some reaction in the material, voltage drops, the plate distance increases and current drops, this goes on until the voltage becomes high enough to overcome the force and hold the plates and the current increases further.
Besides that, the sudden "release" could cause a voltage by the piezo electric effect. Some ceramics have a rather big tempco.

And referring to the pdf about partial discharge, maybe there are bubbles in the material .

Maybe also some effect as mica's have on very low frequencies, as if they can not follow the change in polarity. 

This could also explain the non lineair behavior in tests.

In electrolytic caps there is the chemical problem as an extra.

I am curious what would happen if we make a big 2 plate cap and measure the movement, force and leakage. And then place several dielectrics between the plates (like a ceramic tile  ) And then look if they behave also "strange" But I do not have the gear to do those measurements.

I made something like that years ago to test several materials behave as dielectric. (not with the sensors) http://www.pa4tim.nl/?p=2929 the last picture on the page.

I picked them for their low leakage and easy to source if someone wanted to try and replicate anything I had done.  Alum elec may have been easier to measure but I was not sure they would repeat.  Then again, these have done anything but be repeat. 

I wanted to know how running the HV test effected cap #4 and #8,  so I set the system back up and ran a few parts including a 20Meg. 

[joeqsmith]

I repeated 5 sweeps for both #4 and #8 and compared them with the control.   #8 is still a bit unstable but the high voltage has had a pretty dramatic impact on both parts.

[joeqsmith]

Using a wet stone, I wanted to see where the device broke down. 

[PA4TIM]

PA4TIM: And referring to the pdf about partial discharge, maybe there are bubbles in the material .

Cool,now we see the bubbles. Very nice controlled sanding

[joeqsmith]

And here is the problem:

[tautech]

Very very cool Joe 'Sherlock' Smith. 

[joeqsmith]

I think testing ceramic caps is not as easy as it looks.
Many ceramics have a big voltco and piezo electric effect. Maybe the voltage pulls the plates together, that increases the leakage, that causes some reaction in the material, voltage drops, the plate distance increases and current drops, this goes on until the voltage becomes high enough to overcome the force and hold the plates and the current increases further.
Besides that, the sudden "release" could cause a voltage by the piezo electric effect. Some ceramics have a rather big tempco.

And referring to the pdf about partial discharge, maybe there are bubbles in the material .

Maybe also some effect as mica's have on very low frequencies, as if they can not follow the change in polarity. 

This could also explain the non lineair behavior in tests.

In electrolytic caps there is the chemical problem as an extra.

I am curious what would happen if we make a big 2 plate cap and measure the movement, force and leakage. And then place several dielectrics between the plates (like a ceramic tile  ) And then look if they behave also "strange" But I do not have the gear to do those measurements.

I made something like that years ago to test several materials behave as dielectric. (not with the sensors) http://www.pa4tim.nl/?p=2929 the last picture on the page.

Interesting read on your fF project.   

[joeqsmith]

I tried to leave a post on the Carlson YT video to see if I could purchase a completed unit but I get this strange message that my comment failed to post.  I looked around his channel and could not find where he has any type of store.  Does anyone know if he offers to sell these as a final product?   To try and review it, I want to start with one that the author has given his seal of approval on rather than something that was hand built by an amature with who knows what parts.   

Also, if you noticed that I had edited a few of my posts over the weekend.  Lots of grammar and spelling made a few difficult to read was all.

[BravoV]

I tried to leave a post on the Carlson YT video to see if I could purchase a completed unit but I get this strange message that my comment failed to post.

Wonder what is that error message mean ?

[oldway]

I tried to leave a post on the Carlson YT video to see if I could purchase a completed unit but I get this strange message that my comment failed to post.  I looked around his channel and could not find where he has any type of store.  Does anyone know if he offers to sell these as a final product?   To try and review it, I want to start with one that the author has given his seal of approval on rather than something that was hand built by an amature with who knows what parts.   

Also, if you noticed that I had edited a few of my posts over the weekend.  Lots of grammar and spelling made a few difficult to read was all.

I thought we were not going to talk about Mr Carlson's tester anymore ..... That's off topic for the new title.

We will start again with the same problems as in the topic on the Mc Carlson tester. 

[oldway]

You are topic starter, if yourself is speaking about capacitor tester, then I think that I can also post about testing capacitors:

It is already obvious that the capacitors tested by PA4TIM and yourself have extremely different and even erratic behaviors and that it is therefore impossible to create a model that can predict the leakage current from a low-voltage measurement.

Whatever the reasons for this abnormal behavior, it simply proves that a low-voltage test does not make it possible to verify whether a capacitor is good or not.

Mr. Carlson's claims that the leakage current behaves like a resistance , that's not what you found .... As Dave would do, it's time to push the "fail" button.

My point of view is not that of a theorist, nor of a researcher, but that of a practitioner who seeks practical results.

I work in electronics repair (currently vintage audio with vacuum tubes and transistors) and:

1) It is already extremely rare that I have to test a capacitor with regard to the leakage current because a simple measurement of the voltages of the circuit or of its functionality makes it generally possible to detect this kind of problem (only test needed: measure of capacity)

2) If I have to measure the leakage current, I have to do it at the rated voltage to be sure that the capacitor is working properly when it is powered on.
If the test does not guarantee it, this test is useless and is a waste of time.
Cheaper and simpler to replace the capacitor.

What is a capacitor good or bad? What are the criteria used to determine it?

There are two possible points of view:

1) to measure all the characteristics of the capacitor (measure and not predict!) And compare them with the technical data of the manufacturer.
At the level of repair, it is inapplicable because it takes too much time.

2) to be satisfied with a less rigorous test that detects the breakdown voltage ....
If the breakdown voltage is lower than the nominal voltage, the capacitor is either defective or must be reformed (electrolytic)

It should be noted that the circuits are not all equally sensitive to the leakage current.
For example, in a leakage biased tube circuit with a 10M G1 resistor, the least leakage current of the capacitor coupling it to the preceding stage will prevent the circuit from operating.
On the other hand, if it is a cathode resistance biased stage, where the resistance of g1 is generally of the order of 100K, a low leakage current is acceptable.

There are hundreds of different types of capacitors and dielectrics.

Each has a different behaviour and need his own model. 

It's obviously impossible to use hundred of models to predict leakage current from a low voltage test....

The leakage current must be measured at rated voltage, this the only solution.

[GEOelectronics]

Oldway said :The leakage current must be measured at rated voltage, this the only solution."

A few pictures of my yesterday's test of a new-to-me 50(?) year old GR 1419-B Polystyrene  Capacitance Decade. These are still made today by IET Labs and cost $3581 USD.

Everything tested perfect at 50V and 2-Terminal mode (one side not grounded to case). Second set of tetsts with the common grounded to the case was a different story. One component failed at 200V. They are all rated for 500V continuous. Good thing I tested it thoroughly before using it in my shop.

Pictures reduced in size to save bandwidth but full resolution originals will be linked in my upcoming topic "GR 1419-B Polystyrene Capacitance Decade Teardown"

The last picture shows leakage test with the defective component removed and all the others in circuit at same time
500V/ 0.00 uA />>500M Ohm leakage

Geo>K0FF

[oldway]

At $ 3581 USD the capacitor decade, I hope you did the tests with sufficient current limiting resistance to not exceed a few mA.
With such current limitation, the test should not further damage the capacitor.

As you know that the breakdown voltage is 200V, you can still use this capacitor at a voltage below 200V.
However, the leakage current of this defective capacitor may be larger than the normal leakage current.

[David Hess]

1) to measure all the characteristics of the capacitor (measure and not predict!) And compare them with the technical data of the manufacturer.

Sometimes even that is not enough.  I have a Sprague series 30D aluminum electrolytic capacitor which was part of a decoupling network in a Tektronix 7904 oscilloscope and the capacitor meets all of its specifications when tested at 1000 Hz but fails for the high frequency decoupling that it was intended for.  This leads to your next point:

At the level of repair, it is inapplicable because it takes too much time.

The only test short of using a network analyser was to replace the capacitor which in combination with the typical low frequency tests revealed the issue.

[GEOelectronics]

At $ 3581 USD the capacitor decade, I hope you did the tests with sufficient current limiting resistance to not exceed a few mA.
With such current limitation, the test should not further damage the capacitor.

Good advice, a few microAmps limit even better, that's all it takes.

Geo>K0FF

[oldway]

1) to measure all the characteristics of the capacitor (measure and not predict!) And compare them with the technical data of the manufacturer.

Sometimes even that is not enough.  I have a Sprague series 30D aluminum electrolytic capacitor which was part of a decoupling network in a Tektronix 7904 oscilloscope and the capacitor meets all of its specifications when tested at 1000 Hz but fails for the high frequency decoupling that it was intended for.  This leads to your next point:

At the level of repair, it is inapplicable because it takes too much time.

The only test short of using a network analyser was to replace the capacitor which in combination with the typical low frequency tests revealed the issue.

An electrolytic capacitor test should include an ESR measurement at 100Khz ...
If you had done this test, you would probably have noticed that this capacitor had an abnormally high ESR.

Of course, you will answer that in the 70s and 80s, capacitor manufacturers did not give ESR values at 100Khz, that's right.

[oldway]

At $ 3581 USD the capacitor decade, I hope you did the tests with sufficient current limiting resistance to not exceed a few mA.
With such current limitation, the test should not further damage the capacitor.

Good advice, a few microAmps limit even better, that's all it takes.

Geo>K0FF

Some current is usefull to reform the electrolytic capacitors you test...You see the breakdown voltage increasing and the leakage current decreasing.

Current limited at 5 mA seems to be the best choice both for efficiency and safety.

[PA4TIM]

oldway: An electrolytic capacitor test should include an ESR measurement at 100Khz ...
If you had done this test, you would probably have noticed that this capacitor had an abnormally high ESR.

The ESR is 99% of the time not given for 100 kHz. They state the impedance for 100 kHz. You measure that with an network analyser. (that is for normal electrolytics, for solid smd versions they often state the ESR) But for high enough capacitance the Xc part is so small that you can use it for ESR. (most ESR meters use a squarewave with a root frequency of 100 kHz. It contains odd harmonics probably upto 1MHz so it is a very crude test. The best way is often a scope

I often have to test caps, there is in those cases no other way. I then must be 100% sure I found the fault.
I measure C, D, leakage at WV and almost always look up the datasheets.

[David Hess]

1) to measure all the characteristics of the capacitor (measure and not predict!) And compare them with the technical data of the manufacturer.

Sometimes even that is not enough.  I have a Sprague series 30D aluminum electrolytic capacitor which was part of a decoupling network in a Tektronix 7904 oscilloscope and the capacitor meets all of its specifications when tested at 1000 Hz but fails for the high frequency decoupling that it was intended for.  This leads to your next point:

At the level of repair, it is inapplicable because it takes too much time.

The only test short of using a network analyser was to replace the capacitor which in combination with the typical low frequency tests revealed the issue.

An electrolytic capacitor test should include an ESR measurement at 100Khz ...
If you had done this test, you would probably have noticed that this capacitor had an abnormally high ESR.

If that had been an option I would have used it but my impedance bridge is limited to 1000 Hz.  Many of the modern multi-frequency LCR meters are essentially network analyzers and either would likely have revealed something.

Of course, you will answer that in the 70s and 80s, capacitor manufacturers did not give ESR values at 100Khz, that's right.

Even the modern datasheet for that part does not list specifications at 100kHz but I could have compared it to similar parts which do have such a specification.  I was just using it as an example where the most common diagnostic tests are not sufficient and it is more economical to test it by replacing it which is likely whether the test reveals that it is bad or not.

[GEOelectronics]

Good advice, a few microAmps limit even better, that's all it takes.

Geo>K0FF

Some current is usefull to reform the electrolytic capacitors you test...You see the breakdown voltage increasing and the leakage current decreasing.

Current limited at 5 mA seems to be the best choice both for efficiency and safety.

I've heard about reforming electrolytics (one of high $ value like those can type "computer grade" or high Voltage) but can't remember actually doing so before, except the old trick of bringing the AC supply up slowly on tube gear that has been in storage for a long time. Every service bench should have at least a small isolated Variac transformer with meters.

K0FF's Proverb concerning 50 year old electrolytics: "Fixation by Replacation".

Here's a trick for doing just that easily on old style leaded caps or other components wired to terminal strips:

https://www.eham.net/articles/23125

Geo>K0FF

[PA4TIM]

I have done a lot of reforming and most times with succes. (only in my own gear but also just as measurement experiments)

[oldway]

I just finished to repair a SANSUI AU-9900 of 1976....

ALL the electrolytic capacitores are from 1976, none have been replace, and they are all good...

The problems ? Bad contacts, four bad transistors, tests and adjusts....

Whe should stop to have allergy of capacitores....they are not the culprit of all the failures....I have had far more failures with transistors than with capacitores.

[joeqsmith]

I am not too surprised that there were cases that the low voltage tester could detect faults that the old high voltage ones could not just based on the few parts I have looked at.

Some searching, this person had a few pages dedicated to vintage capacitor test equipment.  They to the time to measure the thresholds for the various capacitor types.  Not too surprising that these old relics are not very sensitive and have no way to do a peak detect.   I suspect none of these would catch #4 and #8.  Possibly #2 and #5 as well.   #6 is a lost cause.  The variance is just too small.   

https://people.ohio.edu/postr/bapix/

In the Carlson's Lab video, he mentions the threshold he used for the alum elec.   I plan to run a few NOS alum elec 150V parts next to see how they behave at 20 and 150 volts and then again attempt to very carefully and methodically damage them.     

[joeqsmith]

This plot shows three different tests.  All caps were swept to 20 volts.    Current is in mA. 

Cap #1 (green) is tested without forming, while caps 2-4 have had a half hour form.   

I then formed cap #1 for an hour and repeated the test (white).     

I then formed all four caps an additional hour.  As I mentioned, there was a time we were doing this in a fairly large scale.   

All capacitors are new old stock.  The file names follow the same format, DUT#, capacitance, ESR in ohms and dissipation factor.   At 20 volts, the only part that exceeded 5uA was the unformed part.