How to determine the voltage rating of an unknown capacitor?

[Jay_Diddy_B]

Hi,

For anybody interested in modelling the voltage coefficient of a ceramic capacitor in LTspice I developed this model.



Instead of specify the capacitor in uF it is specified in terms of charge. in LTspice X is the voltage across a component. The model shows a capacitor that has a value of 1.08 times in nominal value, falling by 40% at the rated voltage.

In the test circuit I compare a linear capacitor with a nonlinear capacitor.

Here is a scope shot showing a real nonlinear capacitor:


Top trace is the voltage across the capacitor. The input step is 6V.

The lower trace is the input current.


Jay_Diddy_B

[(In)Sanity]

So in the interests of actually being able to use the capacitors I have on hand I've employed the Fred Flintstone method of testing.   It's called a current limited 1000v DC supply.   So what's an acceptable leakage current for an average ceramic SMD cap,  or should I just blow one up and see where it fails ?

I figure if I can leave it run for days at voltages it's going to be seeing in whatever circuit I'll be using it in..then it has a fighting chance.   The VC however might be a big issue.    How about curve tracing with a variable high voltage AC supply ?

Thanks,

Jeff

[(In)Sanity]

Ok,  so I just tested a 3.3nf cap from the batch of several thousand that I have and it failed @ 930 volts.   Just made a little "tick" sound and the current dropped off.  So I can assume the batch of caps is probably gong to work at a few hundred volts. That pesky VC is what will haunt me.  I had a few others not related to this batch fail @ ~300 volts.   I think I'll try to see if the leakage current is non linear with voltage.

Thanks,

Jeff

[AndyC_772]

The rated voltage of ceramic capacitor is nothing to do with the voltage at which it actually breaks down and fails, it's to do with the voltage at which it retains its capacitance. If you could test the effect on capacitance as the dc bias increases, you'd find that it falls to a fraction of its rated value long before it's actually destroyed.

[ftransform]

I had this theory regarding a 555 timer which uses a capacitor to determine frequency or duty cycle. There is a formula for output waveform of the 555 timer. Can the input voltage of a 555 timer be varied (while holding capacitance constant) in order to perform quick and easy capacitance measurements with regards to voltage (at least from 5-15 volts or whatever it is). Or will it exhibit nonlinear behavior because of other circuit properties?

Or of course if any one can suggest a IC that will work for the scenario that I have described it would be appreciated. I think a single chip solution without micro controller would be in everyone's best interest.

[(In)Sanity]

The rated voltage of ceramic capacitor is nothing to do with the voltage at which it actually breaks down and fails, it's to do with the voltage at which it retains its capacitance. If you could test the effect on capacitance as the dc bias increases, you'd find that it falls to a fraction of its rated value long before it's actually destroyed.

This part I completely understand.  It's why I mentioned in my post about the VC being the one pesky thing.   My goal was to simply determine a worse case scenario.    If the capacitor physically fails at 300 volts then clearly it's not going to have a working voltage of 500.   Not at all scientific I know,  but it gives me some idea what I'm dealing with.   

[(In)Sanity]

I had this theory regarding a 555 timer which uses a capacitor to determine frequency or duty cycle. There is a formula for output waveform of the 555 timer. Can the input voltage of a 555 timer be varied (while holding capacitance constant) in order to perform quick and easy capacitance measurements with regards to voltage (at least from 5-15 volts or whatever it is). Or will it exhibit nonlinear behavior because of other circuit properties?

Or of course if any one can suggest a IC that will work for the scenario that I have described it would be appreciated. I think a single chip solution without micro controller would be in everyone's best interest.

In one of my earlier posts I had mentioned a way to use a micro with a DAC, op-amp and ADC to determine the VC at least.   It might need to be a boot-strapped op-amp with the higher voltages.   A PWM controlled switch mode supply might be easier.   My 1000v supply is just a PWM controlled high voltage transformer from some LCD back lighting with a high voltage bridge and high voltage caps.   It does the job.   

The theory was to charge to a fixed voltage and measure the amount lost at a constant current for an exact duraciton,   then charge to the next higher voltage and repeat the process.   As the capacitance falls off do to the VC the percentage of loss will increase.   My other post might clarify this more.

https://www.eevblog.com/forum/chat/how-to-determine-the-voltage-rating-of-an-unknown-capacitor/msg184583/#msg184583

[TerraHertz]

I have the same problem. I got 3 component drawers for free.

<swoon>

[ftransform]

I had this theory regarding a 555 timer which uses a capacitor to determine frequency or duty cycle. There is a formula for output waveform of the 555 timer. Can the input voltage of a 555 timer be varied (while holding capacitance constant) in order to perform quick and easy capacitance measurements with regards to voltage (at least from 5-15 volts or whatever it is). Or will it exhibit nonlinear behavior because of other circuit properties?

Or of course if any one can suggest a IC that will work for the scenario that I have described it would be appreciated. I think a single chip solution without micro controller would be in everyone's best interest.

In one of my earlier posts I had mentioned a way to use a micro with a DAC, op-amp and ADC to determine the VC at least.   It might need to be a boot-strapped op-amp with the higher voltages.   A PWM controlled switch mode supply might be easier.   My 1000v supply is just a PWM controlled high voltage transformer from some LCD back lighting with a high voltage bridge and high voltage caps.   It does the job.   

The theory was to charge to a fixed voltage and measure the amount lost at a constant current for an exact duraciton,   then charge to the next higher voltage and repeat the process.   As the capacitance falls off do to the VC the percentage of loss will increase.   My other post might clarify this more.

https://www.eevblog.com/forum/chat/how-to-determine-the-voltage-rating-of-an-unknown-capacitor/msg184583/#msg184583

That method looks good but a single chip solution would be nicer. Especially if it could be done with a 555 timer. I think a duty cycle change would be the best because then you would not need a oscilloscope and it could be measured with the lowliest of multimeter, perhaps if some math is done? I think some multimeters even measure frequency?

I'm just trying to think of the simplest solution with lowest part count and greatest availability for everyone.

[(In)Sanity]

I'm just trying to think of the simplest solution with lowest part count and greatest availability for everyone.

I had mentioned even perhaps just old fashioned curve tracing?   I've not tried this concept yet,  but might be useful.   It should show the drop in capacitance with increased AC voltage.   

Jeff

[AndyC_772]

I think it's worth revisiting what the voltage rating of a capacitor actually means.

With a ceramic, as we've discussed, it's primarily to do with the fall-off of capacitance with voltage. It's a limit imposed by the useful functional performance of the device. Exceed it and the part just doesn't do its job properly - and that's easily defined and measured.

With other types it's not so clear cut, though. Tantalums catch fire and go bang, so I'd be very wary indeed of trying to measure the rated voltage of one at all.

Other types may simply suffer a reduced lifetime at higher voltage, so you might find that a 10V cap works perfectly well at 20V for a while, but dies prematurely. A test instrument won't tell you that, unless you specifically set out to perform long-term reliability testing.

Personally I still think the cheapest, easiest way to get a capacitor of known value and rating for a project is to buy a new one from a reputable supplier - but I applaud the academic exercise nonethelsss.

[ftransform]

I think it's worth revisiting what the voltage rating of a capacitor actually means.

With a ceramic, as we've discussed, it's primarily to do with the fall-off of capacitance with voltage. It's a limit imposed by the useful functional performance of the device. Exceed it and the part just doesn't do its job properly - and that's easily defined and measured.

With other types it's not so clear cut, though. Tantalums catch fire and go bang, so I'd be very wary indeed of trying to measure the rated voltage of one at all.

Other types may simply suffer a reduced lifetime at higher voltage, so you might find that a 10V cap works perfectly well at 20V for a while, but dies prematurely. A test instrument won't tell you that, unless you specifically set out to perform long-term reliability testing.

Personally I still think the cheapest, easiest way to get a capacitor of known value and rating for a project is to buy a new one from a reputable supplier - but I applaud the academic exercise nonethelsss.

me and insanity are chinese quality control.

I think I am going to do some experiments in order to measure these guys tonight.

[robrenz]

Don't forget the dielectric is an insulator and has a breakdown voltage where it will puncture. A higher voltage capacitor is going to have a thicker dielectric.

[Fraser]

I bought a Duoyi DY294 Semiconductor analyser a couple of years ago that also claims to provide the breakdown voltage of capacitors. I have never used it for such but suspect it works on the leakage across the dielectric as the voltage is ramped up to 200V  I don't like causing a breakdown in a dielectric....it just seems the wrong thing to do somehow. I suppose if the capacitor is sacrificial and not for use it is of little consequence.

I have attached the DY294 manual

A review is here:

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-1/

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-2/

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-3/

For GBP30 delivered, it wasn't a bad buy.

Regards

Fraser

[(In)Sanity]

Well one thing I can say about breaking down the dielectric on a capacitor.   If you thought it was rated at say 200 volts and it goes up in smoke at 150 volts then clearly you were mistaken.   If on the other hand it survives to 930 volts like one of mine did,  then it might be anything from 1 volt to 500 volt rated    Yep I know..so helpful isn't it. 

Jeff

[saturation]

That's a really interesting device.  Rather than just a transistor or any other semiconductor tester, its really a 1kV or 200V current limited supply that you could use for testing anything for its reverse breakdown or voltage limits and a handheld at that.  I don't think anyone else makes such a thing in such a form or cost.

Its very helpful for salvaging parts or overstock and characterizing them.

I bought a Duoyi DY294 Semiconductor analyser a couple of years ago that also claims to provide the breakdown voltage of capacitors. I have never used it for such but suspect it works on the leakage across the dielectric as the voltage is ramped up to 200V  I don't like causing a breakdown in a dielectric....it just seems the wrong thing to do somehow. I suppose if the capacitor is sacrificial and not for use it is of little consequence.

I have attached the DY294 manual

A review is here:

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-1/

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-2/

http://www.jestineyong.com/review-of-the-dy294-digital-transistor-tester-part-3/

For GBP30 delivered, it wasn't a bad buy.

Regards

Fraser

[JackOfVA]

The high K ceramic dielectric capacitors have complex C versus V behavior, to say the least.

The first plot below demonstrates two behavior characteristics.

First, hysteresis - the capacitance at a particular voltage depends upon the history of the voltage. This is the electric field analog of B-H hysteresis seen in magnetic core inductors. Note that the starting point, just like a B-H curve, does not fall on the hysteresis loop.

Second, the C versus V relationship depends upon the interval between voltage steps. See Bob Pease on "dielectric soakage" for a discussion of this phenomena. A short version of it is that the dipole charge elements in the dielectric do not all orient and reorient at the same speed with respect to voltage changes.  This is modeled by Pease as multiple RC elements in parallel.

The data is taken with an HP4192A impedance meter using the internal bias voltage, with the HP 4192A controlled over the IEEE-488 interface by a program I wrote. Data analyzed and plotted with Origin software.  I intentionally used a Z5U part for these measurements to illustrate how far their behavior diverges from the perfect capacitor we often think of.

To illustrate the dielectric soakage further, I ran the same 0u1 Z5U capacitor through a series of alternating +35V/-35V cycles. The first plot shows the full data run and the second shows an expanded view of a couple of reversal periods.

To show how badly the non-linear capacitor can be in real life, I looked at current through a Y5U at 1000 Hz and compared it with a high quality film cap. Voltage is the upper trace, current through the capacitor is the lower trace. The distortion is obvious just looking at the current waveform.

Finally, it's not just capacitance that can change with applied voltage; so can the dissipation factor. The D versus V plot is older data, taken with a General Radio bridge.

[robrenz]

Nice work JackOfVA

[Fraser]

Saturation,

I bought the DY294 because it looked useful for all manner of tests and appeared well made for the money. I must be honest and say that I have not used it in anger. It does seem to be a lot of test capability for not a lot of money. I know of no other kit in my lab that will voltage test a capacitor or VDR so easily.

The user manual is pretty appalling and no new versions beyond V3 seem to be available. The ergonomics of the unit are not optimum either....but for GBP30 delivered....its not bad.

Worth having in the test kit inventory ? I think so.

Fraser

[(In)Sanity]

Well the graphs provided by JackOfVA certainly add more complexity to this puzzle.   It's starting to sound like their is no simple way to perform these tests.   That's not to say I've given up.   

Thanks,

Jeff

[JackOfVA]

To return this to the original question, if you have sufficient number of capacitors of a particular type to test to destruction, apply a current limited DC voltage to a representative sample of the lot, increasing the voltage until the dielectric ruptures, calculating the max, min, mean and standard deviation. 

Apply a reasonable safety factor to the result, perhaps 3:1 or 4:1 from the mean.

That strategy won't be useful if you have 5 or 10 parts of each particular type, but if you have 100 or 200, sacrificing 5 or 10 for a destructive breakdown test isn't too bad a price to pay.

With a knowledge of a reasonable maximum recommended operating voltage, you can then look at C versus V over the rated voltage to get an idea of the dielectric class. NP0/C0G ceramic will have next to no change in capacitance with applied voltage. X7R will have a lot and Z5U even more. (I assume these are ceramic capacitors.)

If you want, I'll can run a destructive test on a few parts for you - at least up to 1KV, don't think I can easily go beyond that point without kludging up a HV transformer and a Variac and I don't have time for that. But, I can run up to 1 KV with a standard lab HV supply. Contact me directly if you are interested. I can't run hundreds of different parts, but be glad to look at a dozen or so samples of four or five different values and/or capacitor types.

With respect to surface mount ceramic caps, most every C0G/NP0 part I purchase is  light gray color whilst the X7R types are brown or tan. That's not universal and I don't suggest it as a definitive test for dielectric class, but I can say that I've never seen a light gray ceramic surface mount cap that was not a C0G/NP0 type and that I've never seen an X7R of light gray color. I have seen a few C0G/NP0 parts that are brownish, particularly in values of a 100 pF or less. 

 

[(In)Sanity]

I have no problems with 1000V AC or DC.   I've only done DC destruction tests so far.  I can try AC of various frequencies.   I'll see what I get.  My focus has shifted to trying to fix a 60 inch TV with back-lighting that acts up after 10 hours.   But that's another subject.

Jeff

[Babak7]

I like both idea of VCO & LCR and isolating the device under the test ( i.e. the capacitor) with high voltage capacitor.

However, I have an idea that might be closer to actual physics of cap.  As the voltage increases the leakage of the cap increases.  So let us use a current source, connected to very high voltage say 500V, or more if you want to use Vacuum tubes  to create the current source -- I know we have transistors that can handle up to 1000V--,  Let's adjust the current source to have current of about 10nA or whatever leakage value is acceptable to you.  Then start charging the capacitor from 0V.  At the beginning the capacitance voltage increases linearly with time, but at some point leakage current becomes equal to current source current and the voltage does not increases anymore, and that voltage is you Capacitor's maximum voltage.  Here are some physical derivation:

Q=CV ,

take derivative of both sides.

dQ/dt=d(CV)/dt

dQ/dt is definition of current  or I  (I=dQ/dt)

so

I= d(CV)/dt 

so

V(t) =(1/C) I t

However as voltage increases the leakage current increases and at some point leakage current = to the current source and voltage across capacitor does not increases with time anymore and stays constant.  you can choose this value as voltage rating of your cap.  So, you could say voltage rating of my cap is defined, when the capacitor have say 10nA of leakage current.

To generate high voltage you could use Tesla coil, boost convertor, or voltage multiplier ( voltage multiplier ladder) using diodes and capacitors.  For current source you could use high voltage transistor, or vacuum tube connected to this voltage multiplier circuit.

[james_s]

For many ceramic capacitors, the capacitance starts to fall off quite sharply as the dc bias is increased. Perhaps you could devise a circuit which measures capacitance with a gradually increasing dc bias on the cap under test until the capacitance falls by, say, 20%, or you chicken out - whichever happens first!

I don't think this will work. I am not an expert on this topic but I remember reading that the capacitance drop is more closely related to the physical size of the capacitor than the rated voltage, so for example using a 50V 0603 cap instead of a 25V 0603 part won't solve the problem, but using a 1206 cap will reduce the effect.

[shapirus]

...yes I know this is a very old topic :).

I am here to leave a link to an article that describes a clever technique and proposes a very simple device that allows to measure capacitance under dc bias relative to zero-bias capacitance: https://www.analog.com/en/technical-articles/how-to-measure-capacity-versus-bias-voltage-on-mlccs.html

Having that one, and a device such as DY294 or other that is able to ramp up the voltage until a leakage current is detected, one could characterize unmarked caps without any (or without trusted) specifications, which can be parts salvaged from discarded boards or bought in kits from aliexpress etc.

I'm going to build it some day soon.