Miller Cap: C0G, Mica or Polystyren?

[thorstenlarsen]

Which dielectricum do you prefer for millercap? Please give objektive reasons and/or subjektive impressions.
👂 Thorsten

 

[Vovk Z]

I can't give subjectives, but objectively C0G I think is the best (C0G is better then any film).

 

[DPH]

There's also a potential benefit in parasitics from a smd c0g cap and a very tight layout for feedback.

 

[Elvee]

I can't give subjectives, but objectively C0G I think is the best (C0G is better then any film).

I don't think so: high quality films, like PS, PTFE, PE are probably the best after vacuum caps, but it is difficult to be definitive about ceramics, because so much depends on the exact process, impurities, etc. when you look into really fine details

 

[Mark Johnson]

Why not build the amp with one type and listen to it? Then switch to another type and listen again. Whichever one sounds the best, IS the best. At least to you.

 

[simon7000]

My experience is that the inexpensive polystyrene capacitors available today are nowhere near as good as the old ones that are only available in limited old stock values.

The C0G types may actually have a voltage coefficient even thought they do not have a temperature one, so they can change value slightly if the applied DC voltage varies.

I do have a nice assortment of Mica capacitors, but when using them for some reason that I have not yet found they just don't seem to be a favorite.

I tend towards Rel-Cap which makes polystyrene and foil capacitors and Teflon/foil ones. Naturally they are ridiculously expensive.

 

[Bigun]

Subjectively, I've only compared ng0/cog with silver-mica in a Class AB amplifier; it was a sighted comparison although at the time I felt it was a comparison made without expectation bias - except perhaps that I should hear no difference. I heard a very small subjective difference, the ng0 sounded cleaner, without any artifacts I could detect whereas the mica had some slight haze / micro texture (forgive the b-s adjectives) to the mid-treble. Of course, there are tolerances with these caps and although they were both marked at the same value I have also found, subjectively, that the value of Cdom can have a clear impact on the sound.

I always choose ng0/cog for my builds unless wanting to use something else because I am trying to clone a particular design as faithfully as possible. I use both SMD and through-hole ng0/cog.

I've never tried Teflon, PIO etc. - generally I prefer small physical size for tight pcb layout around the VAS device.

 

[Elvee]

The C0G types may actually have a voltage coefficient even thought they do not have a temperature one, so they can change value slightly if the applied DC voltage varies.

If the value varies with DC excitation, it also means that it changes with <large>* AC levels.
In filters, Cdom or similar roles, it will cause odd-order distortions if the effect is polarity-independent, or even-order ones if the effect is asymetrical.

That is exactly the reason why I said these caps were objectively inferior compared to high-quality plastics: the dielectric absorption of, say mica might look horrendous, but in fact it remains a (mostly) linear phenomon, meaning it might slightly impact the frequency response; by contrast, anything that is amplitude-related will ultimately result in THD increase.


*the definition of large depends on the actual signal level compared to the voltage rating of the cap, and also on the level of tolerated non-linearity: for ppm levels, a COG should be safe.
If your target is in the ppb range, that will no longer be the case, and Cdom will typically see the full output voltage, making things worse.
At this level, the PCB material and many other factors will also come into play, so the capacitor could only play a minor role in the final THD result

 

[Elvee]

My experience is that the inexpensive polystyrene capacitors available today are nowhere near as good as the old ones that are only available in limited old stock values.

BtW, poor standards PS caps are not a new thing: Suflex for example managed to do it wrong in the seventies.
I do not know how they achieved it: after all, putting together a good plastic film, foil and termination doesn't seem insurmountable, and manufacturers (mostly German) did it right 20 years earlier, but somehow they missed it.

No wonder they disappeared from the manufacturing landscape...

 

[Bigun]

https://www.tscgroup.com/acrobat/sm01002b.pdf

http://www.cde.com/resources/catalogs/ceramperf.pdf


no voltage dependence noted here

 

[jcx]

use Two Pole Compensation?

TPC greatly reduces the feedback thru the compensation C at audio frequency

and high V rating could help distortion too - kV availabe in typical 10s of pF Cdom sizes

 

[DPH]

Samuel Groner's treatise on low distortion oscillators talks about materials for caps. C0g have the advantage in stability but some of the film/foils are about 12-18 dB better. That's the difference between <-130 dB and -150 dB however.

 

[Bigun]

remember, the VAS transistor will have an intrinsic capacitance, made of a semiconductor region that exhibits a temperature and voltage dependency out of all proportion to the external capacitor

 

[jcx]

yes, beyond some distortion chasing point you can't simply use just one 'VAS transistor'

use a cascode that cancels the Ccb (needs an emitter degen R)

slope dependent distortion hawksford - Google Search

 

[phase]

Polystyrene, not the ones that have the leads laid onto the foil.
After that silver mica, followed by cog.

 

[Vovk Z]

I don't think so: high quality films, like PS, PTFE, PE are probably the best after vacuum caps, but it is difficult to be definitive about ceramics, because so much depends on the exact process, impurities, etc. when you look into really fine details

I am almost sure that "in average" present C0G is better or much better than any film.
For example - C0G gives the very best performance in low-distortion oscillators. It means a lot to me (as I am now mostly measurement equipment DIY-er).

 

[chris719]

Yes, and even if the absolute best film cap is better, the manufacturing of C0G MLCC are probably more consistent. They are also much better in terms of taking abuse during soldering.

 

[Vovk Z]

The C0G types may actually have a voltage coefficient even thought they do not have a temperature one, so they can change value slightly if the applied DC voltage varies.

It may have (and definitely have) but it is at ppm level.

 

[chris719]

If the value varies with DC excitation, it also means that it changes with <large>* AC levels.
In filters, Cdom or similar roles, it will cause odd-order distortions if the effect is polarity-independent, or even-order ones if the effect is asymetrical.

That is exactly the reason why I said these caps were objectively inferior compared to high-quality plastics: the dielectric absorption of, say mica might look horrendous, but in fact it remains a (mostly) linear phenomon, meaning it might slightly impact the frequency response; by contrast, anything that is amplitude-related will ultimately result in THD increase.


*the definition of large depends on the actual signal level compared to the voltage rating of the cap, and also on the level of tolerated non-linearity: for ppm levels, a COG should be safe.
If your target is in the ppb range, that will no longer be the case, and Cdom will typically see the full output voltage, making things worse.
At this level, the PCB material and many other factors will also come into play, so the capacitor could only play a minor role in the final THD result

A quick search indicates that calcium zirconate C0G have almost zero voltage coefficient of capacitance.

http://www.kemet.com/Lists/Technica... Class-I C0G MLCC and SMD Film Capacitors.pdf

They are used in equipment that measures better than an AP can measure. I highly doubt this is the issue that Ed makes it out to be.

 

[Mark Johnson]

He's just pointing out that C0G may not perform quite as well in other categories, as it performs in temperature sensitivity. Merely a heads-up: don't forget to read the application notes.