electrolytic cap polarity questions

[DF96]

If you are testing that way there is a chance that the current in the output capacitor is too high. That would explain why the smaller cap survives. It limits the current because of its higher impedance at 20 Hz.
Paralleling two caps would mean the caps share the current.
Please be aware that there is a maximum permissible value for the current in a cap. The data sheet calls it max ripple current. The data sheet gives you a value for 100 or 120 Hz because that is what the cap will see when used as a smoothing cap in a classic power supply.
The max ripple current at 20 Hz is much lower. Maybe another forum member can explain why that is so.

I think you have solved the puzzle. It turns out that the clues were there:

The signal swings max 7.5vrms @20hz with a 10,000uf cap on the output. Surely enough to fry the cap. With the same signal strength @20hz the output with the 3300uf cap reads only 2.65vrms, which explains why the 3300uf does fry as it is not pulled down pass the -2V limit.

So 10mF passes too much signal (for the cap), 3.3mF passes too little signal (for the application). Nothing to do with polarity or voltage limits, but current limits and high pass behaviour.

 

[Charlie123Maldor]

Yes that's a very good explanation garommeteer.

I test it with the 20hz and below only to adjust the low frequency response, and that was not for a long time.

With the 10,000uf cap, 20hz sine at 7.5vrms into 2ohms was at 0db and not at -3db cutoff point. I was just pushing it to see how low it would go and see how low it would go. I would imagine the 3x 2200uf ones in parallel is plenty to get <20hz at -3db but i haven't tested it yet. There is really no need for so much capacitance there anyway as the PSU can't keep up when playing demanding music at high levels. I was just shocked and confused as to why it happened, as with Df96 also.

Its a good point, I think it could be the current too because I remember one of the caps going off when playing music at a only a watt or two.

I am no expert as I am only a hobbyist but I know that class A amp draws a high continuous constant current even with no signal. The emitter has 17.5v dc on it and with a 2ohm load attached, it's pushing nearly 4amps. Maybe it's too much charge for one big cap there and why ''Paralleling two caps would mean the caps share the current'' is working.

I will plan on getting some 2200uf caps with a higher voltage rating (63V) and very high ripple rating also.

Fixing the problem is great but I think understanding why is just as important and better.

Thanks for everyone's input.

 

[DF96]

10mF and 2 ohms will be -3dB at 8Hz, so about -0.6dB at 20Hz.

If there is a problem with lower signal levels ("only a watt or two") then we may be back to the parasitic oscillation hypothesis.

Class A current draw is irrelevant, except to the extent that it limits the possible output.