Not the dump load pwm issue again.
i've repeatedly said for well over 2 years now the issue is supply side inductance, not dump load side inductance.
http://johansense.com/bulk/resistor_pwm.JPGhttp://johansense.com/bulk/pwm_resistor_1.JPGhttp://johansense.com/bulk/pwm_resistor_2.pngprovided the load is inductive or resistive, it doesn't matter. you could use a 500 foot spool of 12 awg as a dump load on *any* pwm dump load, and provided you've got a diode across the coil,* it won't know the difference.
*The dump load should already have one, but it may not be heatsinked sufficiently for continuous currents created by a mostly inductive load, for example, a 500 foot spool of wire.
A capacitor across the dump load will cause the switch to fail due to peak current, that's it, nothing more or less.
An additional failure mode exists if an IGBT is used as a switch without a free wheeling diode across the dump load, or the igbt... in which case, whoever designed it needs to find a new job.
solving the audio problem is more a matter of finding sufficient inductance to generate continuous currents in the dump load resistor+capacitor.
and that inductance is expensive, compared to just increasing the frequency to 16+khz.