quick way to measure C-V curve

No, but the implication is that you know the parts you using. As I've said before, when I design power supplies (or practically anything else) a web page is open to the Murata site and I'm looking up the parts I'm using before they get plopped down on te schematic.

Wouldn't the jazzy way to do this be to digitize the voltage versus time and do the derivative numerically, calculate capacitance and print it out ?>:-} ...Jim Thompson

Well, its kinda that. The value of dI divided by 2pi F dV isn't 'the capacitance' so much as it is A capacitance. Q/V is an entirely different number, because you get Q by integrating I over a range of V values; Q is a different function of V than just proportional.

These capacitor materials are also history-depedent, it's possible that you'd get different results if you twiddled the DC offset too high during setup. It could also be different if you changed the 'small voltage' 0.5Vrms (the peak value might dominate the capacitance value, if polarization is quick-onset and slow-decay).

For the same reason, there is likely a bit of dissipation (which will be temperature and vibration and age dependent).

It could even (like an electret) hold a history for years until popped into an annealing oven.

It's a nonlinear material, and C is a linear property, so it'll never be an accurate characteristic because it characterizes a linear model.

No, this technique needs a voltmeter:

(I had one of those when I was a kid.)

I'm making switching regulators, which will have a small ripple voltage on top of a large (if you consider 1 volt to be large) DC voltage, so the incremental capacitance is what I care about. Above

I do see sinewave distortion if I use a scope for this measurement. 1 volt RMS is maybe a tad big.

As you say full, accurate characterization of these caps would be a lot of work, and they aren't all that repeatable anyhow.

Polymer aluminums still have limited ripple current and sucky reliability. I use them but always with a parallel ceramic to take the ripple current.

Bingo! ...and what you learn on one can't be easily generalized. I met with the Kemet rep last week. She promised Spice models. We'll see. They also have a line of "low-distortion" X7Rs but I think they're in limited production (and have been for a decade).

Oh, right, two input channels plus the trigger.

Interesting...

I used your model to see what the waveforms would look like for a square wave and a sine wave applied to an R-C integrator:

It can be seen that the non-linear model does not work for negative values of DC bias. Probably needs an abs().

Obviously the non-linear voltage coefficient causes distortion of the sine wave. But I wonder if the effect is time-dependent? Perhaps at high frequencies the effect could be less. This might be the case if it is caused by a physical effect like compression of the dielectric, although I would expect that to give a positive voltage coefficient. So there must be some other mechanism. It should be fairly easy to perform a test by applying waveforms of different frequencies. Perhaps I will do that.

Paul