Possibly the Class D or swithing supply high frequency noise is corrupting the signal source or confusing the Sound Card Analog to digital converter.
Interesting you should say this; during the development of the front-end signal source, I discovered the Class D amp would send bursts of energy into my phase locked loop LP filter, causing it to wobble in and out of phase lock. I had to put the circuit inside a shielded enclosure, now all is perfect. Also, note in the very first post that I was able to get a good FFT of the input signal, which was taken simultaneously with the loaded output. So worst case it would muck up ch. 2 but not ch. 1.
It might be worth investigating the high frequency noise generated by the Class D and the power supply and trying to reduce it.
It appears the majority of noise is generated by the amp, not the power supply. Although the amp has its own output filter, I also used the leakage of the transformer and a 0.015uF film/foil capacitor on the transformer secondary to filter out that last bit of noise. Didn't want to add too much capacitive load, as this can make the amp run unstable. Other than these small steps, I wouldn't really know what to do without upsetting something.
Is the plot of the clean sine wave measured at the amplifier input with it driving the load?
Yes, definitely. There is a 0.01% sine wave feeding into the amp while driving loads from 50W to 100W. Same exact output distortion characteristic.
Has anyone else independently verified the distortion of a PowerPhysics A-404?
Not to my knowledge, and if you aren't an OEM, the manufacturer has little interest in speaking with you. I had some email correspondence with the president, and it was short-lived.
It is good that the distortions do not change much by the transformer load, but less good that the amp generally has high distortion.
Yes, that is encouraging. I am considering throwing in my AB subwoofer amp just to prove everything else is proper and reasonable.
Do you use a low pass filter to keep the carrier artefacts away from your sound card?
Apparently not. This would imply the output is inadequately filtered, relying on the additional assumed inductance of the loudspeaker. This is typical, but I do have a large leaky power transformer in series which I would think is sufficient. If not, I have two thoughts:
1. The very first post showed a simultaneous capture of the input and output, while driving load. Since the blue input is verified to be clean (same signal I get from the source when Class D is powered down) I think I have validated that my sound card is not suffering from some type of EMI issue generated by the amp or switching supply.
2. If the amplified output is indeed clean, but the sound card is incorrectly capturing carrier artifacts and reporting as harmonic distortion, I need to filter this signal. Can I do this passively? Is this some type of aliasing problem? I will say, by the scope the sine wave looks great, but it's not easy to detect 0.5% THD with the eye. What I have been using is a 'filter' consisting of an auxiliary split bobbin power transformer to isolate and scale down the output suitable for the sound card. Would this not be sufficient to keep the sound card happy? If not, I wonder if I want this carrier-ridden signal driving my sensitive load.
Thanks for your knowledge base; I admit I have limited experience with Class D, but I do have enough equipment to troubleshoot and enough analog understanding to be dangerous and creative.