My mistake, isn't that just what JA has been using for several decades as his clipping point, or was it 1%? I'm not sure now.
JA uses 1%, as that’s the same as -40dBFS, which has been shown by some as the audibility thresholds for music. Take note that audibility changes based on frequency, which is why a THD plot across the frequency range is very helpful, and we can more easily hear distortion as the frequencies go higher, and it’s also important to note that audibility also changes with volume, distortion is slightly more audible at louder levels.
However, THD also stacks across components, from your DAC to your preamp to your amp to say a MiniDSP between your preamp and amp, etc. So while -40dBFS (1% THD) is mostly inaudible, if all your components are 1% in the treble, they will add together and now be audible.
I’m not positive on the formula, but if it’s the same for how I’m been told to calculate stacking for channel separation (crosstalk), if we err in the side of caution and assume 6 components, in order to maintain 1%, each component would need about -56dBFS (0.1584893%), so 0.1% THD for each component would be inaudible; for 3 components, 0.3% is enough to be inaudible.
Looking at the 0.1% THD point of this Onkyo, we get about 160W into 4ohm.
If using DSP and really wanting perfect amp performance in regards to impedance handling (2x wattage for ever halving of impedance), this would mean 80W into 8ohm.