Fun with Capacitors

[jcandy]

Nothing get an audiophile more riled up than a discussion about "cheap" versus "boutique" capacitors. Using my awesome new Smith and Larson WT2, I set out to test not only capacitance, but also more subtle effects like dissipation factor (limited to the basic series resistance model).

In principle, the WT2 can make super-accurate measurements, but I still had to spend a bunch of time doing tweaks in order to reliably measure ESR (the series resistance). The built-in capacitance measurement tool was giving negative series resistances in some cases, so I needed to construct a very-low-resistance test jig and then make separate frequency-dependent calibration measurements. I can discuss the details if there is interest.

Basically, I bought a bunch of audio-grade metalized polypropylene 8.2uF capacitors (and one crappy 4.7uF NPE cap) and measured capacitance error and loss angle (just the ratio of real to imaginary impedance).

The cool result is that the most expensive cap (Auricap) was the worst of the all audio-grade units (the NPE was far more horrific, but that was expected) in the sense of have the most variable capacitance.

Here are the results:

[jcandy]

Erse, Jantzen, Mundorf

[jcandy]

Bennic, Solen, Axon

[jcandy]

Dayton 1%, Auricap, Sonicap

[jcandy]

Solen PB, Dayton 5%, ClarityCap PX

[jcandy]

ClarityCap ESA, ClarityCap SA, Rat Shack NPE

[[email protected]]

Re: Fun with Capacitors

Can you try to summerise your findings

[AMC]

Re: Fun with Capacitors

Very interesting, but I'm not sure what a good result would be in the second figure. Could you explain what would be ideal.

[billfitzmaurice]

Re: Fun with Capacitors

Obviously if the most expensive cap had the worst result it's only because you didn't properly break it in. Figure ten hours for each dollar of its cost. If that doesn't do the trick then just as obviously it was your failure to have the cap cryo-treated first that ruined the result. And of course you can't get the superior performance of a high end cap unless you use solid silver teflon insulated multi-stranded wire. Cryo-treated, of course. And broken in. :p

[AMC]

Re: Fun with Capacitors

Was your test equipment put together with steel screws or brass screws? And where did you have your rose quartz positioned? (sorry, just had to pile on a bit) This does clearly show, that good caps are better then NPE, at least sometimes.

Here is something you could test, I have always heard that you can combine a good cap with a NPE and it sounds good, could you test this? (i.e. a 7uf NPE in parallel with a 1uf audio cap, sounds just like a 8uf audio cap)

[neildavis]

Re: Fun with Capacitors

Wow -- excellent work!

But I don't have a clear idea yet how to link the results to audible effects. I assume that the delta C vs F characteristics might result in a significant distortion product...maybe someone can figure out how to state the data as distortion. Did you consider measuring delta C versus V? That might be a lot easier to relate to distortion.

Not trying to focus on the mundane, but I'm interested in the DF measurements for that NPE cap. I had seen some manufacturer's data that showed a moderate change in DF vs F, but your results show much more variation, and it shows that DF for an NPE can be quite high at the crossover frequency. I modeled the effect of DF in my passive crossover program, but I assumed DF to be constant vs frequency. Oh well.

If you want to see the effect of DF on a crossover, you can download the "alpha" code at the link. It lets you check a box to add the effect of DF to the crossover model (modeled as a frequency-dependent resistor). For DF less than 5%, you won't see any difference. But once you get above 10% the effect is easy to see. BTW: you will probably want to look at this code anyway, because it has the network equations for the crossover module in a help file--might help with your Linux project.

http://www.audiodevelopers.com/Software/PSD.zip

[carlspeak]

Re: Fun with Capacitors

Wow -- excellent work!

But I don't have a clear idea yet how to link the results to audible effects. I assume that the delta C vs F characteristics might result in a significant distortion product...maybe someone can figure out how to state the data as distortion. Did you consider measuring delta C versus V? That might be a lot easier to relate to distortion.

Not trying to focus on the mundane, but I'm interested in the DF measurements for that NPE cap. I had seen some manufacturer's data that showed a moderate change in DF vs F, but your results show much more variation, and it shows that DF for an NPE can be quite high at the crossover frequency. I modeled the effect of DF in my passive crossover program, but I assumed DF to be constant vs frequency. Oh well.

Let's not rush to judgement on NPE's deficiencies due to elevated DF. Know this: Not all NPE's are the same in this regard. Some do have high DF and other's have very low ones. At least, that's been my experience. A $1 RS NPE had a DF over 10%, wheras, others I've purchased from audio retailers like Madisound and PE are well less than 5%.

[mzisserson]

Re: Fun with Capacitors

SO.... there are measurable differences? Hummmmmm.......

Nice work, however there are other items which need to be measured to come to anything conclusive including impulse response, slew rate, impedance curves, etc...

More importantly is can one decipher what the measurements mean related to sonics.

Assuming "the most expensive did the worst" is foolish when we do not know exactly what it all means in the application of sound reproduction. I mean, there are general assumptions, but that seems to be as far as it gets and we speak in the generalites of electricity vs. audio band reproduction and the reactive devices they are proprigating through. Most times, these measurements are mase with simple tones, or sweep tones, because there has not been a way yet to see what happend when complex reactions such as strange electrical phase angles are presented by the driver/crossover reaction.

This is a good track! With a little ingenuity it may be made into something truly conclusive! Thanks!

[lhwidget]

Re: Fun with Capacitors

So, if I'm reading the DF plot right, I'm seeing a DF of about 0.15 at 2 kHz = 15%, or are we seeing 0.15%?

[andrewkim]

Re: Fun with Capacitors

DF of typical capacitors in question is in the range of 0.1% to 0.2%.

ESR -vs- |Xc| is in range of 1:1000. Considering the typical distortion of the woofer, I doubt that this is audible.

Andrew