[grenert]

I want to replace the DC blocking capacitors in some portable audio devices of mine with better quality ones (I did test DC offset to see if they were necessary, and they definitely are). It is difficult to find 330uF or larger caps that will fit, but I can probably cram in some 220uF ones. Would I be better off with non-audio 330uF low ESR caps, or "audiophile" 220uF caps such as Silmic or Cerafine? Thanks!

 

[n_maher]

Are the stock caps 330uF? If not, what are they?

 

[grenert]

Stock are 220uF. I'm going on the typical modder's (maybe incorrect) assumption that the stock caps are both too small and poor quality.

 

[Pars]

220uf is pretty large for a coupling cap. There is usually a resistor to ground after these. The formula f = 1/2 * Pi * R * C (where C is in farads) determines the -3dB frequency. This is a high-pass filter. If for example the resistance is 100K ohms, the cutoff (-3dB point) would be 1/2 * 3.14159 * 100000 * 0.000220 = 1/ 138.22996 = 0.0072 Hz, which is really close to DC. Even a resistance of 1000 ohms is going to give a pretty low value (0.7Hz).

I would go better cap, but maybe smaller value. Out of curiosity, are 220uf caps used elsewhere in the device? Sometimes manufacturers will just use or reuse parts that they are already using elsewhere because it is easier and cheaper for the to do so.

 

[jcx]

depends on your cans too, some are as low as 16 Ohms while others are sensitive enough to be OK with DAPs despite 70 Ohm impedance

thats > 4x range in C for the same low frequency cutoff

 

[holland]

There may be an output resistor to bring the load up, such as 120ohms so that a smaller C can be used.

For the resistor to ground, wouldn't that parallel with the headphone Z in the RC circuit for a single ended amp?

I know some amps, based on chip amps, like the uAmp107 has a 1000uf coupling cap.

 

[AudioCats]

I vote for better quality. The last thing you want is muddy boomy sound from the headphone out....

 

[ericj]

Quote:

Originally Posted by Pars /img/forum/go_quote.gif 220uf is pretty large for a coupling cap. There is usually a resistor to ground after these. The formula f = 1/2 * Pi * R * C (where C is in farads) determines the -3dB frequency. This is a high-pass filter. If for example the resistance is 100K ohms, the cutoff (-3dB point) would be 1/2 * 3.14159 * 100000 * 0.000220 = 1/ 138.22996 = 0.0072 Hz, which is really close to DC. Even a resistance of 1000 ohms is going to give a pretty low value (0.7Hz).

470uf is quite common for output coupling in DIY headphone amps.

And yes, there is often a resistor to ground after the coupling cap, but it's there to allow the cap to charge and discharge normally whether or not headphones are plugged in.

And you've forgotten that the impedance of the headphone driver is in parallel with this resistor.

So lets say that you have a pair of Grados at 32-ohms and a charging resistor of 10k, which is a totally believable charging resistor value. The effective impedance 'seen' by the capacitor is 31.8 ohms. We'll just say 32 ohms because .2 ohms is irrelevant here.

A 470uf capacitor thus gives us a corner frequency of about 10hz for grados, though bass starts rolling off at about 40hz.

That's only half the issue, though. You also have phase distortion to consider. If the output cap is too small relative to the impedance of your headphones, there will be significant smearing of bass frequencies. With 32-ohm grados and a 470uf output cap, there will be some phase distortion as high as, oh, 80hz or so.

The downside: a larger output capacitor takes longer to charge, so the turn-on 'thump' that occurs while the cap is charging takes longer the bigger the cap is. if this is bad enough, it could potentially harm some headphones if they are plugged in when the device is switched on.

 

[regal]

Quote:

Originally Posted by ericj /img/forum/go_quote.gif That's only half the issue, though. You also have phase distortion to consider. If the output cap is too small relative to the impedance of your headphones, there will be significant smearing of bass frequencies. With 32-ohm grados and a 470uf output cap, there will be some phase distortion as high as, oh, 80hz or so. .

Phase "distortion" is a misnormer, it just means there is a shift in phase relative to the other frequencies up to 10x the cutoff freq. On headphones I believe it is a non issue. I have used VST plugins to purposely change the phase below 80 hz and it is completely inaudible. I could be wrong but I think this is a bit of a head-fi myth.

 

[grenert]

Thanks for all of the information and the equation. The players do use the same 220uF caps elsewhere. Most of my headphones for portable use will be 60 ohms or more, so that helps somewhat. The problem with larger value caps is the diameter. However, I could probably parallel two good-quality 220uF caps (which have smaller diameter) to get 440uF. Any problem with this solution?

 

[tomb]

Quote:

Originally Posted by regal /img/forum/go_quote.gif Phase "distortion" is a misnormer, it just means there is a shift in phase relative to the other frequencies up to 10x the cutoff freq. On headphones I believe it is a non issue. I have used VST plugins to purposely change the phase below 80 hz and it is completely inaudible. I could be wrong but I think this is a bit of a head-fi myth.

While I generally agree with eric's last post, I think you have a point, there. I've used the term, too (phase distortion), but the actual start of the corner frequency fall off may be more important.

I published this graph way back in July of last year in the Millett MAX thread (Post #479):

It's easy to see in this case that if someone is pursuing flatline amplifier response (why wouldn't we?), then you have to take into account where that -3dB cutoff slope starts deviating from the flatline. To do that, you have to push the cutoff frequency lower so that the start of the fall off occurs at 20Hz or below. Whether the start of the response deviation can be called "phase distortion" seems a bit iffy, I guess, but the desired effect is similar: you want the cutoff frequency almost less than half of 20Hz.

Note -3dB cutoff frequencies:
470uf w/32ohms ~10Hz,
680uf w/32ohms ~ 7Hz,
1000uf w/32ohms ~ 5Hz.

This and the graph shows that 680uf might be a minimum to ensure a flat response with Grados down to 20Hz. That said, no one has ever reported issues with bass rolloff using Grados on a Millett or MAX with 470uf output caps. I think that has most to do with the fact that Grado's own response curve peaks at 50Hz and rolls off under that.

Disclaimer: I am not saying that Grados' bass response is limited. When you get below 50Hz, you're talking about feeling rather than hearing, anyway.

 

[regal]

Good post tomb, Also remember the reason we use -3dB, because no one can hear the less than 3dB difference in sound levels.
So in effect the -3dB point is where the bass starts to roll-off ( to human ears anyway.)


Edit: I guess an argument could be made that if the headphones themselves start to roll-off before the corner freq the combo w/ the cap could be percieveable before the -3dB point!

 

[luvdunhill]

Quote:

Originally Posted by regal /img/forum/go_quote.gif Good post tomb, Also remember the reason we use -3dB, because no one can hear the less than 3dB difference in sound levels. So in effect the -3dB point is where the bass starts to roll-off ( to human ears anyway.)

I don't know if I agree with this. When dealing with speaker crossovers, differences in fractional decibels are definitely perceivable, especially in the upper-midrange. Anyone who has dialed in a number of crossovers in their day will be able to do this blind-folded.

Do you have a source for this claim? Here's mine:

Decibels

 

[ericj]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif While I generally agree with eric's last post, I think you have a point, there. I've used the term, too (phase distortion), but the actual start of the corner frequency fall off may be more important.

I think that, due to the swampy nature of bass waves, as long as phase distortion stays in the bass range it is probably not easily perceived. It's when it reaches up into the midrange that things get ugly.

 

[nikongod]

Quote:

Originally Posted by luvdunhill /img/forum/go_quote.gif I don't know if I agree with this. When dealing with speaker crossovers, differences in fractional decibels are definitely perceivable, especially in the upper-midrange. Anyone who has dialed in a number of crossovers in their day will be able to do this blind-folded.

in multi-driver crossovers, the phase shifts &turnover points of a crossover are essential to get right. in reality though, it does not matter if BOTH are at exactly 3khz, or both are at EXACTLY 2.9khz. As long as they mesh properly they will work. the problems come when things DONT mesh up properly. the sounds from the 2 drivers can cancel or partially cancel (I agree) which is all sorts of screwy.

In a single driver system this whole issue is less important. there is nothing else to reference the mis-phased sound to, and you just go with the flow.