My 'simulation' / thought experiment didn't assume that any speed related limitations like stray capacitance were significant, or additional complexity like thermal effects, or direction of flow. I only focused on a non-linear gain function, such as that of a MOSFET. Afaict, if the system is "linear time invariant" as an approximation, instead of being forced to go sample-by-sample, a sine wave could be processed in parallel with a large number of samples spread across an array of identical amplifiers and then recombined as a composite signal. Ie: for the purposes of the example it doesn't matter in which order they are processed, but with each iteration we get a waveform with new frequencies added while others are subtracted.
I'm well aware that in practice things could be far more complex, and I'm certainly not opposed to having NFB available as a tool in the toolkit.
I'm well aware that in practice things could be far more complex, and I'm certainly not opposed to having NFB available as a tool in the toolkit.
The main advantage of a THD measurement is that it will show up any failings in construction or stability, particularly important if you are producing amps for resale. Even if you like the sound of 0.5% THD, you would want all your amps sold for money to be the same and working to spec. Sadly many suppliers of modular/kit amps do not test for THD as they do not believe in the measurement, and hence quality can be 'variable'!
Not true in a literal sense. Might serve as a rough check though, which might be better than no check at all.
Perhaps should say 'most' failings. I think you will find most major manufacturers would do a THD check before shipping anything, some might even provide the chart to the customer!