Film capacitor as power-supply filter

Great info! Thanks, John.

That fig 4b is an obvious made-up cartoon. It's certainly not exponential. And it has no numbers.

Quite welcome. It's good to have some real numbers on (one) real cap.

I was impressed by how noisy the current is. There's stuff going on inside.

I think it was Werner Von Braun who said "One experiment is worth a thousand expert opinions."

Ditto on the thanks. The Tardin article mentions that the caps behave like they have a zener diode in parallel... So having the current look like noisy zener current above the 'threshold' makes some sense.

Years ago I charge an Al eletro to ~1/2 the max voltage, left it on my bench for the weekend and came back to see the voltage had only dropped by ~10% or so. It would be fun to hook one up to an electrometer and let it sit there for a week or whatever and record the voltage.

George H.

Or just check it with a DVM now and then. I did that with a supercap for a few months.

The curve is exponential over the portion that is increasing. Every capacitor will be different, so if you want the curve for your particular capacitor, you have to measure it.

That's why you need a bleeder resistor for higher voltages.

Yor data is valid for that capacitor only. Every cap will be different. There are many different electrolytes with different characteristics, such as non-solid borax or organic, non-solid water-based, solid manganese dioxide, solid conducting polymer, non-solid hybrid electrolyte, etc., and the leakage characteristics will be different for each type. See Wikipedia, "Aluminum electrolytic capacitor", at

Right... I think I used the trick of putting a Gig ohm in series with my DMM and dividing by 100. I'd bet it can't be exponential* 'all the way down' at some point there'll be some other constant (resistive) leakage path.

Hey are tant's any better than Al-electros. At my ppoe I made this triangle wave generator, current source into a cap, milli second to kilo second periods. I used a 100uF tant for the longest times... seemed to work fine.. but I only 'really' measured the longest times once. (otherwise just made sure it went up and down.)

George H.

A pure RC has a exponential decay, e^(-kt). But the leakage in an electrolytic isn't ohmic, so the decay slows down as the voltage drops.

Pretty complex.

Excellent whining.

It's more data than data on no capacitors. I suspect that all wet aluminum electros will have similar shaped curves, so will be self-stabilizing in series strings.

Superb, well-researched whining.

I used a standard Panasonic aluminum electrolytic.

Measure something else and post it here.

Right, if you monitor the voltage you'd have to integrate it to get the current.. It's not at all clear to me what that double exponential would look like.

George H. something like this?

Why suspect? Why not read some of the links I provided? For example, see Page 3 of

Quote:

The operating leakage current [5] as a measure of the forming condition of anode foil depends on the time, applied voltage, temperature, and history of the capacitor (fig. 4 (a) ~ (c)). Typical values of the operational leakage current range between approx. 5 ~ 15% of the data sheet value of leakage current amount and are usually reached after several ten minutes of continuous operation.

The leakage current specified in the data sheet shall be valid even after a long, voltage-free storage period and has therefore a much higher numerical value than the operating leakage current. The oxide layer dissolves to a certain extent as a function of temperature and electrolyte composition, because without any voltage applied, the oxide layer cannot regenerate ("self-healing") [5].

While low-voltage capacitors (up to 100 V rated voltage) with solvent-based electrolyte systems are usually very stable, high-voltage capacitors (from

15 ~ 30 minutes of operation of the electrolytic capacitors via a resistor (low-voltage: 100 Ohm, high-voltage: 1 k, see [7], section 4.1 "pretreatment") at a voltage increased gradually to rated voltage may heal the weak spots in the dielectric and lower the leakage current below the data sheet value.

You are trying to pretend the measurement you made on one capacitor is the first time such a measurement has been made. In fact, you said so yourself.

I posted numerous links that show your assumption is not valid. There is a tremendous amout of knowlege in those links. You would do well to study them.

Wilson is mentally ill or defective or both.

.... Phil

Well, except for the shape of the curve. And the brick wall on the right.

Every capacitor

I did measure one. I suspect that the general trend is common to most wet aluminum caps.

I know that some polymer alum caps are different.

I said "possibly" the only one available online, after several minutes of intense web searching. If you can find an electrolytic cap leakage measurement, past rated voltage, with real numbers, not fuzzy cartoons, please post the link for us.

I didn't assume anything. I measured something.

Try it; it's more fun than cutting and pasting a lot of text.

There is a

Lotta words, no numbers. Engineers need numbers.

If, I(t) = Io*exp(V(t)/Vx) = C dV/dt where Io, Vx and C are constants

Then I get that V(t) = Vx * ln{(C*Vx)/(Io*t)}.. but I might have made a mistake, and getting the initial conditions right is a bit.. uncertain.

I need a good graphing program. Time to learn python?

George H.

Excel isn't a good graphing program, but you can get it to produce graphs.

IIRR Mathcad and Scicad can do it too.

Why measure past the rated voltage? You are never going to operate there. You simply risk damaging the capacitor for no good reason.

The curve shown in fig. 4 is a generic curve. The actual curves are different for each capacitor, and depend on variables such as those described above. So it is impossible to represent all capacitors at once.

Your data is meaningless. As mentioned elsewhere, capacitor manufacturers can give actual curves for their capacitors. You can compare your data against the spec to see if your cap is out of spec.

Your data is meaningless.

Nah. Real engineers need equations. Do an exponential fit to your data. The Stdev will give an indication of the quality of the measurement. The equation will allow you to compare other capacitors. The curve can be compared to actual specs from manufacturers.

You need to also try measuring the capacitor voltage with a constant current source. That could give valuable new information.