serengetiplains, what value, voltage and film type are the caps that you're are going to use.
Generally the bigger the cap the lower it's resonant frequency will be. If the resonant frequency of the cap falls below the frequency range you are trying to affect the bypassing effectiveness will be greatly diminished. As an example I called Hovland a few years ago when they were still in business and got Bob Hovland on the phone to ask him what the resonant frequency was of an 8mfd 100 v MusiCap. Bob replied that it was 140kHz. I went ahead and bypassed the 4 pole Jensen caps in the power supply of my Tripath amp anyway even though the switching frequency was over 700 kHz. It made a small worth while improvement but I don't believe it was due to noise reduction. I think it might have had more to do with having a small amount of very low impedance energy storage available for reproducing the leading edge of the waveform. There certainly wasn't an adequate amount of power stored at 54volts to run the amp for any significant period of time.
If you are actually trying have an effective bypass at the amps switching frequency the smaller caps will have to have leads that measure in length in the Millimeters or the lead inductance will defeat your purpose.
By the way have you thought about replacing the wiring that came stock with something better,that would surely be a quick and simple upgrade path.
Scotty
Scotty, I understand that any bypass capacitors will operate in the lower KHz region depending on lead inductance, meaning they won't affect switching-related artifacts. Per your experiment, and experiments I have performed with class D amps (Tact, Tripath), bypassing to my ear creates obvious audible effects, the overall sum of which I find quite pleasing.
I suspect bypass capacitors help shunt DA recovery voltages to ground. DA is generally modelled as follows:
You can see from the model that capacitors release spurious voltages related to the original input (charging) voltage at frequencies determined by the parallel RC combinations, the R indicating, I think, a lower frequency given the lagged release. Here is a graph that purports to show recovery voltage/time after release of the original charging voltage. Nasty stuff.
How this data reflects a working circumstance is beyond my guessing abilities, except perhaps to say this: DA recovery voltages will carry a frequency related to the overall peak charge/time of a capacitor. I suspect that the pulse-current on the output of the Ncore (or any class D amplifier) can be decomposed into lower (including audio) frequency and higher (including switching) frequency components. This decomposition is performed quite handily by the LCR filter on the output. I don't see why a similar decomposition cannot be performed prior to the output, with L provided by the capacitor leads. This would suggest the capacitor can operate at audio frequencies despite the high switching frequency, thus removing audio-frequency noise from the output rails. This is actually what I hear.
Fwiw, here are a few charts I created to test DA recovery voltages. I used a good DMM with a 10Mohm input, which is too low for consistent data for capacitors of differing values. But the charts at least give some correlative evidence of DA effects, particularly for same-size capacitors.
I doubt I've answered your question to anyone's (including my) satisfaction, but what we need is some explanation, however tentative or cursory, to give some insight into what's happening.
I will use 1000V 2.2uF Solen metallized teflons for the large bypasses (smallish!), and 100V 0.1uF and 0.01uF film/foil teflons for the on-board bypasses.
