Jambo
100+ Head-Fier
- Joined
- Jul 4, 2007
- Posts
- 206
- Likes
- 11
It's an interesting one, I've honestly never come across it before in coupling caps. As you point out though it is common practice in decoupling caps as in your picture.
The non linearities I was referring to.. Well, say for your load 4.7uF gives you your desired frequency response, if an ideal capacitor is assumed then that is all you need.
However, in your decoupling example, smaller caps are paralleled up with the bigger caps as they tend to have lower ESR and ESL and will perform better at higher frequencies. In other words, you're compensating for parasitics. What you're trying to do is compensate for the non linearities of one cap by adding another one. The range of frequencies you want to cope with in decoupling is in theory infinite, and how linearly you do this isn't all that crucial. Unless loading an LDO or something else that can go unstable, pretty much every cap you can add is a bonus.
To apply this to AC coupling, I would say that in order to make that work you would want a very accurate model of the caps that you are using that you could simulate and see what happens. My point is that if you weren't careful about it, you could easily introduce more non-linearity than you already had by adding two different capacitances, ESLs and ESRs, and that this is a problem this time since it is in the signal path.
By all means though, experiment and see what you can hear, and if anyone can experiment with this on an audio analyser I'd be interested in seeing the results. I just know for a fact that this is not common practice and I figure there must be a reason for it.
As you point out, the audio bandwidth is not very big so I really doubt whether any difference would be audible so long as your main cap was of decent quality.
The non linearities I was referring to.. Well, say for your load 4.7uF gives you your desired frequency response, if an ideal capacitor is assumed then that is all you need.
However, in your decoupling example, smaller caps are paralleled up with the bigger caps as they tend to have lower ESR and ESL and will perform better at higher frequencies. In other words, you're compensating for parasitics. What you're trying to do is compensate for the non linearities of one cap by adding another one. The range of frequencies you want to cope with in decoupling is in theory infinite, and how linearly you do this isn't all that crucial. Unless loading an LDO or something else that can go unstable, pretty much every cap you can add is a bonus.
To apply this to AC coupling, I would say that in order to make that work you would want a very accurate model of the caps that you are using that you could simulate and see what happens. My point is that if you weren't careful about it, you could easily introduce more non-linearity than you already had by adding two different capacitances, ESLs and ESRs, and that this is a problem this time since it is in the signal path.
By all means though, experiment and see what you can hear, and if anyone can experiment with this on an audio analyser I'd be interested in seeing the results. I just know for a fact that this is not common practice and I figure there must be a reason for it.
As you point out, the audio bandwidth is not very big so I really doubt whether any difference would be audible so long as your main cap was of decent quality.
