Tantalum Capacitors in Audio

You might want to review what's said at

- They don't tolerate reverse polarity at all well (distortion and/or failure)

- They don't tolerate over-voltage

- They don't tolerate sudden charge/discharge cycles

- When they fail, they tend to fail shorted (sometimes solidly so, sometimes intermittently so)

They're used, a lot, as power-supply bypass caps in older test equipment (Tek, HP, etc.), and I'd say that they're the #1 cause of failure I've seen in such equipment. When a piece of test gear just stops working (and sometimes starts emitting a horrendous bad smell) my first reaction is to look for a tantalum bead cap which has shorted out and fried itself... and I usually find one. I usually replace 'em with high-frequency-rated aluminum 'electros of about 5x or 10x the capacitance, and that has always worked fine.

Depends what you mean by "bandwidth", and how you're comparing. According to Elliot, most modern 'lytics have a bandwidth up into the MHz range... the foils are bonded togther in a way which minimizes parasitic inductance (unlike in some older 'lytic types). So, they should have a high-frequency bandwidth limit far in excess of what you'll need for audio coupling.

Bipolar 'lytics tend to be physically larger than polar ones, for the same voltage rating and capacitance. If you're limited to caps of a particular physical size, then you can get more uF into that space with a polar 'lytic, and this will extend the low-frequency end of the frequency response range (all else being equal). Whether that's relevant or not in your circuit, is up to you to decide.

My general understanding is that polar 'lytics can work as well as non-polar, for signal-level audio coupling, if you keep two things in mind:

- Use a large enough amount of capacitance that the voltage across the cap does not vary significantly during normal use... that is, the low-frequency cutoff is well below audio. Elliot recommends sizing them at 10x what you'd need to get your desired low-end rolloff frequency.

- Bias them properly. Lore usually holds that distortion can begin to appear if the voltage across them dips down below zero (reverse polarizing the dielectric). I've read some opinions that this isn't a significant problem for _small_ reverse voltages (say, a volt or so) but haven't personally tested this.