More capacitor testing

[Conrad Hoffman]

I've put up some results of this weekends cap testing. I'm a measurement junkie and tend to be verbose (ok, verbal diarrhea at times) but the results are quite interesting whether you read the numbers (recommended) or just look at the pictures.

Article and write-up page

The new write-up is at the bottom of the page.

Thoughts and comments?

Best,
Conrad

[Face]

Nice read, thanks.

[Elvee]

Interesting work. It broadly agrees with my own tests, in particular the indicative value of the residue of the bridge.
Normally, the bridge should be able to null completely the linear parts of the impedance; what's left is thus a measure of the intrinsic non-linearity of the capacitor.
Dielectric absorption, series or parallel resistance all belong to the linear category, and should in principle not impact the linearity.
But I noted that non-linearities are in fact associated with the dissipative part of the impedance, especially the "excess" wrt the dielectric in ideal conditions: a high voltage plain foil polyester cap can be highly linear, despite having a relatively high DF of say 0.0035, but a metallized one having a 0.005 figure can be highly non linear.
I also noted that non-linearity in synthetic dielectrics is always associated with construction issues: this is especially true for metallized types.

On the other hand, mineral dielectrics tend to have a small, intrinsic non-linearity: I noticed the effect on mica, glass and ceramic capacitors.
The effect is barely noticeable for high quality ceramics like COG, but can become very large for high k materials, such as Z5U.

You probably made a small error in your table: you categorize the Comp Research cap as a polyester, but the DF is 0.000791, which is lower than the PETP material itself.
It is more likely polycarbonate.

Here are some more examples of DFs for various vintage types (there are one or two typos in the decimal places):
"Best" film cap you've never heard of?
"Best" film cap you've never heard of?
"Best" film cap you've never heard of?
"Best" film cap you've never heard of?
"Best" film cap you've never heard of?

[Conrad Hoffman]

Wow Elvee, that's quite a neat collection of caps. I guessed at the Component Research cap based on the residual, but you're quite right that it can't really be polyester. Data on CR caps can be hard to find and most of what I have is ex aerospace surplus. I have some huge polystyrene (again, a guess) CR caps that are quite remarkable.

I don't find that DF or DA have any good correlation to listening tests, thus my interest in the bridge residuals, which seem like a more rewarding path to pursue.

I later added an addendum to the write-up expressing my concern that internal bridge components can alter the results and that "zero" may not be "zero". Do you have any thoughts on whether my results of the better caps really rank them correctly? Also, any explanation on why the residual is always strong 3rd order harmonic? I should probably put a spectrum analyzer on the bridge output, but there's a limit to the number of hours in a day.

Thanks for having a look!

Best,
Conrad

[diy_audio_fo]

Originally Posted by Conrad Hoffman

I've put up some results of this weekends cap testing. I'm a measurement junkie and tend to be verbose (ok, verbal diarrhea at times) but the results are quite interesting whether you read the numbers (recommended) or just look at the pictures.

Article and write-up page

The new write-up is at the bottom of the page.

Thoughts and comments?

Best,
Conrad

Dear Conrad,
thamk you for your valuable work!
It refreshed my (scarce) knowledge about capacitor.

Digging in my memory I recalled an old paper by John Curl and Walt Jung

A real-time signal test for capacitor quality
http://waltjung.org/PDFs/A_RealTime_...or_Quality.pdf

and I wonder what is (if any) the correlation between the bridge residual of your work and that of Curl / Jung's paper.

[Conrad Hoffman]

Yes, I'm familiar with the Jung/Curl paper. IMO, it's a step in the right direction, but it has a flaw. I don't know if it's a fatal flaw or a minor flaw. The problem is that esr is frequency dependent. The standard model of a capacitor where you have a reactance in series with a resistor only holds at a single frequency. You can't match their two capacitors for loss as they show, except for a single frequency. It isn't valid over the audio range, and it isn't valid for a square wave, so I don't know how good the comparison is. I'm speculating- I haven't sat down and tried their circuit to see if it's really an issue.

CH

[john curl]

We did a first order correction for ESR in the tests. I used a 1 ohm 20 turn pot. It is not much of an issue.

[Conrad Hoffman]

Hi John! Thanks for that- I should wire up the opamp circuit, as it's probably easier to get what I want with that, than with the big bridge. I'm still concerned about the reference cap. I wonder if using a 1000 pF air variable and ranging it up would work, or if that's just too far to go? Though that's what my bridge does, there are some other parts in there that might contribute errors. What I really want to do is resolve the difference between Teflon, polypropylene and polystyrene with no question as to the contribution of the test setup.

Best,
Conrad

[Elvee]

Originally Posted by Conrad Hoffman

I later added an addendum to the write-up expressing my concern that internal bridge components can alter the results and that "zero" may not be "zero". Do you have any thoughts on whether my results of the better caps really rank them correctly? Also, any explanation on why the residual is always strong 3rd order harmonic? I should probably put a spectrum analyzer on the bridge output, but there's a limit to the number of hours in a day.

Thanks for having a look!

Best,
Conrad

Hi Conrad,

I think your ranking is correct, but there might be a small offset, of 20 to 30ppm: a very good PS cap is normally at ~50ppm @ 1KHz.

It doesn't shock me the residual is mainly 3rd harmonic: it has to be odd-order, because there no way an asymetry or anisotropy can appear in the capacitor's construction, and low order components have to dominate, because the change in dielectric constant with the electric stress (what causes the non-linearity) is soft and progressive, a bit like a spring stretched too far.
I have seen exceptions however: capacitors having operated for years at a high ambient temperature with a strong DC voltage applied tend to generate also even-order harmonics.

I suspect some of the distorsion generating mechanisms are mechanical: either electrostriction or Coulomb's force.
Sometimes, when you test a capacitor with a relatively high voltage (10V or more), it generates a very faint sound, at twice the test frequency: these caps tend to display a higher residual than "quiet" types".
The "prebiased" ones, when they generate sound, do it at the fundamental.

Something in your test set-up might also contribute to the 3rd harmonic: one is never too careful with those sensitive measurements: you could try alternative configurations, signal sources, ....

[SemperFi]

Very interesting. I love you guys who nerdily sit around your test bench and give us the goods afterwards I do too, (the nerd part), but am too lazy to write down my findings and make something out of it.

How'bout the difference in film vs foil?

How'bout biasing the cap at several steps up till max rating? Will a bias create less symetry and add some even order harmonies?

Keep at it.