12

What kind of capacitors should I use to couple audio? I'm dealing with line level 1Vp-p audio, with a 1V dc offset (typ.) I want to convert this into 1Vp-p AC, at up to 20 kHz. My first version used ceramic 10u 10V caps - a simulation showed attenuation by about 15% at 20 kHz.

Update: I eventually went for 10u 6.3V. I did not need Hi-Fi quality and the fact that 10u 6.3V caps were available in 0603 packages was useful for me.

Community
  • 1
Thomas O
  • 31,546
  • 57
  • 182
  • 320
  • 4
    What load impedance? Is the audio meant to be "hi-fi"? (hi-fi people are very picky about capacitor types.) – markrages Nov 19 '10 at 14:25
  • 1
    @markrages He is at 1Vp-p so I assumed this is a signal running through a filter or maybe into a A/D and he just wants to remove the DC offset and then add a fixed known offset back. If this is the case, asking for the load impedance isn't tons of help as he is probably inputting into a very high impedance device. @Thomas O please correct me if I am wrong. – Kellenjb Nov 19 '10 at 14:36
  • 4
    Hi-fi people are picky about *everything*. Doesn't mean their concern is actually legitimate. http://sound.westhost.com/articles/coupling-caps.htm – endolith Nov 19 '10 at 15:44
  • The load is a maximum of 1kohm (short circuit), typically it will be driving an audio/video transmitter which will amplify the audio anyway. – Thomas O Nov 19 '10 at 17:05
  • 1
    C=1/(2 * pi * F * R), so an 8 uF cap to pass 20 Hz and above. Standard practice would be electrolytic with "+" terminal toward the 1V bias. – markrages Nov 19 '10 at 22:50
  • 3
    @markrages - .... *Electrolytic?* WTF? Use a good quality film cap. – Connor Wolf Nov 20 '10 at 06:17
  • 1
    'lytics are fine. It's not hi-fi and film caps are 10dB more expensive. – markrages Nov 20 '10 at 07:11
  • @Thomas O have you looked at a Bode Plot of your system? You might want to post a schematic if you want any more help. – Kellenjb Nov 21 '10 at 16:15
  • @Thomas O can you let us know what it was that fixed your issue? – Kellenjb Nov 22 '10 at 02:55
  • @Thomas O This question is now being linked to by other questions. It would really be nice if you could come back to it and let us know what fixed your issue. – Kellenjb Feb 08 '11 at 21:24
  • @kellenjb, Done. – Thomas O Feb 08 '11 at 22:36

3 Answers3

18

Ceramic capacitors can be microphonic and might introduce distortion, mylar or polycarbonate would be better.

Leon Heller
  • 38,774
  • 2
  • 60
  • 96
  • Interesting, did not know. – Kellenjb Nov 19 '10 at 15:06
  • 1
    I do want to make a point that although this answers the specific question that he asked, he is seeing 15% attenuation at 20KHz which seems rather large to me. I suspect hes got something more going on then just the wrong type of cap. – Kellenjb Nov 19 '10 at 16:42
  • @Kellenjb, that is a simulation using ideal capacitors. – Thomas O Nov 19 '10 at 17:04
  • 2
    Post your schematic then. There's soomething wrong with your circuit or sim. – markrages Nov 20 '10 at 07:21
  • @ThomasO If you use ceramic it should be C0G for audio purposes. I'd love to know about this 15% attenuation. No ceramic would do that. – user207421 Sep 24 '14 at 06:24
  • Mylar is alas not that good of a choice http://conradhoffman.com/cap_measurements_100606.html If you're going the expensive route practically any of the other ones studied there is better. – Fizz Sep 04 '15 at 05:20
11

When you are using capacitors to couple a circuit you need to be worrying about the low frequencies. A coupling capacitor is by definition a high pass circuit.

The larger the capacitor value you chose, the lower your cut off frequency will be on the high pass circuit. Wikipedia shows an example circuit as well as how to pick your capacitor value in respect to your desired cut off frequency and resistance.

As for the type of capacitor, since audio is AC you will need to have a non-polarized capacitor. In my experience, I have never found anything better then a ceramic capacitor for this application.

Kellenjb
  • 17,509
  • 5
  • 51
  • 87
  • 8
    Never done any audio stuff, but you don't have to use a non-polarized capacitor to pass AC so long as you give it a DC bias. – Nick T Nov 19 '10 at 22:01
  • 1
    @Nick T a coupling capacitor removes the DC bias. – Kellenjb Nov 19 '10 at 22:26
  • You could bias the amp's output before the cap, though I guess your output could have some unknown bias. – Nick T Nov 19 '10 at 22:32
  • 4
    @Nick T I was assuming he was putting this cap on his input (so the offset would be unknown), now I realize it is on his output so he should have a known DC offset. – Kellenjb Nov 21 '10 at 15:45
  • Put bleeder resistors from your inputs and outputs to ground so you have a known bias. – endolith Feb 09 '11 at 19:26
  • 1
    He says he already has a 1V DC offset on the signal, that's what he wants to get rid of, so a polar electro would work fine. – user207421 Sep 24 '14 at 06:22
  • 2
    "As for the type of capacitor, since audio is AC you will need to have a non-polarized capacitor." ST knows better: http://electronics.stackexchange.com/questions/188722/reason-why-polarized-electrolyic-on-audio-line-doesnt-blow-up ;-) – Fizz Sep 04 '15 at 05:09
1

Considering the specific application (1V p-p, 1V "offset" - I'm assuming offset in this case means the output is 1V above the ground reference), there's no reason to worry about any audible distortion or microphonics; these only come into play with high voltage swings and is, allegedly, part of the "mojo" in Trainwreck guitar amps, which have a ceramic in parallel with a poly cap coupling the first two gain stages with about 300VDC at the plate and a signal that can exceed 200V p-p passing through the cap. At line levels, your MLCC should be sonically transparent.

In most line-level coupling applications, anything over 1uF is usually overkill; if your application is anything close to typical, 10uF will pass frequencies down into the decihertz range.

I too am curious about the 20KHz rolloff in your simulation. Perhaps there's some decoupling between the signal and one of the rails that you haven't accounted for, like the output impedance of the amplifier itself? A 1 ohm resistance at the output would give a corner frequency of about 16KHz with a 10uF capacitor...which really isn't a bad thing, since most humans' hearing is also attenuated above 16KHz. If you really needed to pass everything from subsonic to ultrasonic, you probably want something closer to 4.7uF for the coupling capacitor and a higher impedance input for the next amplifier.

Peter Sage
  • 11
  • 2