Which Capacitors for Audio?

[DaveP]

I saw an article in Electronics Weekly about a year ago, it was a proper investigation into various audio grade types over ordinary types and the main difference was down to the Piezo-Electric effect.  It seems that some caps have a more rigid structure so less energy is lost in vibration.  Now that this has moved from the realm of mystery and rumour into mainstream production techniques, we might see better caps at lower prices.

 

[living sounds]

Polystyrene, russian PIO/Mica, Wima FKP2 (non-metallized polypropylene) are the most transparent from my experience. MKC (polycarbonate), MKP are good where higher values are needed. Tantals in the audio path are great IF there is a significant bias (= DC voltage) present. For electrolytics use Panasonic FM/FC, Nichicon PW. Better caps of smaller values can be used in parallel with bigger ones (like a 22uf electrolytic with a 100nf polypropylene) to significantly improve performance (a practise called bypassing).

Make sure the caps that look like polystyrene actually are made of that material. I've seen a lot of eastern german polyester caps around sold as polystyrene. Used them in a lot of applications and wondered why they sounded pretty underwhelming until I found that out.

 

[johnR]

I saw an article in Electronics Weekly about a year ago, it was a proper investigation into various audio grade types over ordinary types and the main difference was down to the Piezo-Electric effect.  It seems that some caps have a more rigid structure so less energy is lost in vibration.  Now that this has moved from the realm of mystery and rumour into mainstream production techniques, we might see better caps at lower prices.

I wouldn't count on it. I still have a copy of a test report on "low distortion" ceramic caps from 15 years ago, from one of the major manufacturers (I think it was Murata, but I've lost the pages with the company details on). It explained about piezoelectric effect in capacitors, which was new to me at the time, but it has been known for quite a while now.

The high dielectric constant materials like X7R and Y5V are OK for dc supply decoupling. I try and avoid the SMT ones unless they're "fail open" rather than the more common "fail short" type. Thermal stresses and PCB flexing can cause mechanical failure (as can high ac voltage due to piezoelectric stress) without wire leads to take some of the strain.

For high quality audio signal paths, especially where they would have a significant ac voltage across them, they are best avoided.

 

[boji]

Old thread, but I have a cap question, please.

In a 2520 summing amp audio path, would anyone take issue with using a KZ 10uf that is rated at 100v instead of a KT that is rated more in line with the suggested 25v? 
I remember someone saying that overrating the voltage by more than x2 what you expect from the signal can start to reduce the cap value...

I can imagine some of you thinking the choice is bunk anyway as 'Muse' series is probably as good as KT...  ;D  but can't hurt to ask?

 

[Monte McGuire]

Higher voltage electrolytics have lower tan-theta losses, and usually lower impedance. Their larger can size is also usually accompanied by longer life. So, yes, use the 100V part. I typically select the highest voltage that will physically fit into the application since they perform better than lower voltage parts of the same series.

Edit: The high voltage / low value thing doesn't seem kosher. You might be confusing it with the problem of high-K ceramic capacitors losing capacitance when run at a high DC bias, but I assume you're talking about electrolytics here, and not high-K ceramics for audio coupling, which would be a no-no. The tolerance of an electrolytic capacitor does not depend on its voltage rating, and they generally don't "squish up" like ceramics under DC bias either.

 

[boji]

Thanks for helping me complete my mouser order!
cheers

Edit: Just read your edit.  ;D  Ah ok, only generally applies to ceramics then, thanks again!

 

[GreyboxAudio]

While we're at it, I found this article quite worth a read: http://sound.whsites.net/articles/capacitors.htm

 

[ruffrecords]

While we're at it, I found this article quite worth a read: http://sound.whsites.net/articles/capacitors.htm

I have a lot of time for Rod but he does occasionally make silly blunders in his explanations. In his first paragraph what he means to say is:

"if the cap is large enough (compared to frequency and circuit resistance), it will never have any appreciable AC voltage across it. "

I added the AC. He then goes on to say "With no voltage, there is no stored energy." which, if the capacitor is blocking dc is incorrect.  Other than that it is a very good article.

Cheers

Ian

 

[rackmonkey]

Higher voltage electrolytics have lower tan-theta losses, and usually lower impedance. Their larger can size is also usually accompanied by longer life. So, yes, use the 100V part. I typically select the highest voltage that will physically fit into the application since they perform better than lower voltage parts of the same series.

Edit: The high voltage / low value thing doesn't seem kosher. You might be confusing it with the problem of high-K ceramic capacitors losing capacitance when run at a high DC bias, but I assume you're talking about electrolytics here, and not high-K ceramics for audio coupling, which would be a no-no. The tolerance of an electrolytic capacitor does not depend on its voltage rating, and they generally don't "squish up" like ceramics under DC bias either.

Monte’s right on here, but you’re talking about 10uF in the signal path, not 150uF,  and if you’re trying to optimize for  audio quality, why wouldn’t you look at stacked polyester film/foil caps? They’re small, and even the fanciest, most expensive electrolytic is going to have tan theta losses and Q/Dissipation figures (or any other possible figure of merit) ten times or more worse than stacked poly.  And they’re much cheaper than high quality electrolytics for a given size.

If you have the luxury of space, you don’t even have to settle for lowly polyester. You can still find old 10uF Polycarbonate caps on ebay that aren’t nearly as large as a comparable Polypropylene caps, and have 10x or more better tan theta/Q/D numbers than polyester.

Just a thought if the purpose of this exercise is maximizing audio quality.

BT

 

[abbey road d enfer]

even the fanciest, most expensive electrolytic is going to have tan theta losses and Q/Dissipation figures (or any other possible figure of merit) ten times or more worse than stacked poly.

Does it really matter? In a well-designed circuit, coupling caps are calculated for very small voltage across them at audio frequencies. All the usual figures of merit count when there is significant voltage, which would be the case in filters, but we're all supposed to know that electrolytics should not be used in filters.
When used as coupling caps, electrolytics may see some AC voltage across only at VLF (<<20Hz); do we care about distortion of VLF parasitics?
A low-voltage 100uF stacked film costs about $10 a piece; a similar electroytic costs about 20 times less and is also 20 times smaller.
I would agree that for vacuum-state circuits, film caps are better suited particularly because of their negligible leakage.

 

[JohnRoberts]

Does it really matter? In a well-designed circuit, coupling caps are calculated for very small voltage across them at audio frequencies. All the usual figures of merit count when there is significant voltage, which would be the case in filters, but we're all supposed to know that electrolytics should not be used in filters.
When used as coupling caps, electrolytics may see some AC voltage across only at VLF (<<20Hz); do we care about distortion of VLF parasitics?
A low-voltage 100uF stacked film costs about $10 a piece; a similar electroytic costs about 20 times less and is also 20 times smaller.
I would agree that for vacuum-state circuits, film caps are better suited particularly because of their negligible leakage.

+1 what he said...

If the pole frequency is 0.2Hz there will not be much terminal voltage change over audio frequencies. Most non ideal characteristics are related to changing terminal voltage.

Save the big bucks for caps used in passive loudspeaker crossovers where they can actually make an audible difference.

JR

 

[rackmonkey]

okay, fair enough.  But as to the cost factor, i was talking about the 10uF requirement here, not 100uF (i think he said “10uF, 100v”). Higher values would be ludicrously expensive. i have bags full of 10uF poly caps that i bought for less per unit than good electrolytics of the same value. price goes up fast from there.

 

[abbey road d enfer]

The value of a coupling capacitor must be considered in view of the impedance it sees.
In order to ensure flat response at 20Hz (-0.1dB), one needs to set the -3dB point at 3Hz.
But in order to minimize the effects of non-linearities (particularly electrolytics), the -3dB point must be lowered, to ensure no significant AC voltage exists across the cap.
With a 10k load, the minimum value for a film cap would be about 5uF, and at least 10-22uF for an electrolytic.
This is assuming the LF response is governed by a single pole (one dominant C-R HPF).

 

[magicchord]

Maybe this is going overboard but I don't think so.

I like to minimize distortion at low frequencies, and to reduce cumulative LF phase shift through successive stages. When choosing a value for an electrolytic coupling cap, I try to follow the Unity Rule (a rule I made up, hehe).
In essence, the value of the capacitor multiplied by the value of the load impedance should be close to 1.

Example: for a 1k-ohm load, you take 1/1000 which gives you 0.001Farad or 1000 uF.
So, 1000uF to couple to 1k;
100uF to couple to 10k;
10uF to couple to 100k, etc. etc.

Say I need to couple to 47k. So, 1/47000 = 21.28uF so I choose a 22uF cap and call it good.

 

[abbey road d enfer]

Maybe this is going overboard but I don't think so.

I like to minimize distortion at low frequencies, and to reduce cumulative LF phase shift through successive stages. When choosing a value for an electrolytic coupling cap, I try to follow the Unity Rule (a rule I made up, hehe).
In essence, the value of the capacitor multiplied by the value of the load impedance should be close to 1.

Example: for a 1k-ohm load, you take 1/1000 which gives you 0.001Farad or 1000 uF.

Tau (time-constant)=1second, -3dB point at 0.16 Hz and -0.00027dB at 20Hz. A little on the overkill side, but nothing wrong with it.

 

[boji]

Thanks guys for helping me get what that cap is doing too! 

 

[jensenmann]

...why wouldn’t you look at stacked polyester film/foil caps? They’re small, and even the fanciest, most expensive electrolytic is going to have tan theta losses and Q/Dissipation figures (or any other possible figure of merit) ten times or more worse than stacked poly.  And they’re much cheaper than high quality electrolytics for a given size.

The ESR of stacked film/foil caps is a lot worse than with electrolytics. 17 Ohm vs <1 Ohm
If that matters for your purpose is up to you.

 

[abbey road d enfer]

The ESR of stacked film/foil caps is a lot worse than with electrolytics. 17 Ohm vs <1 Ohm

That seems very high; where did you get these data?

 

[rackmonkey]

Certainly doesn't reflect the results of my measurements.

 

[jensenmann]

That seems very high; where did you get these data?

I measured it.