So lets say you are playing back a 20 Khz signal that most you probably can't hear anyway. Since the number of samples on a Redbook CD is around 2 for one cycle of the signal, the resulting waveform is misshapen. So a person that does not understand will think that this doesn't work well for sound reproduction and that person is incorrect. The 20 KHz signal is properly reproduced, however, the mishappen signal produces harmonics. Well the next (2nd) harmonic will be at 40 KHz which is definitely way above a human's hearing capabilities so it has no meaning in context of audio reproduction for human beings. To quote Harry Callahan (Dirty Harry), "A man's got to know his limitations." Sorry to the ladies out there, Harry wasn't very PC.
The fact that you have asked this means you understand the principle of digital-to-analogue, and vice versa. It's what my thread is based around.
It's what I was explaining in post 4. I included a picture of a signal there which was for a much lower frequency sound than 20KHz. I explained though that as you go up the frequency scale you get less samples per cycle. Until almost two per cycle as you said. (No need to worry about harmonics at this point, but bear with because I will get to that.)
I think at higher frequencies the low number of samples fails to determine the amplitude of the signal. Or as you say misshappen. E.g. if the two samples happen to fall at the very start and mid point of the sample. Then sampler will miss the amplitude of the signal. It will still determine the frequency by knowing how many sample points were active where over time. If there were lots above and below zero amplitude it knows the frequency is low etc (like in picture in post 4).
However the answer to what happens is this. There is little fundamental frequency stuff going on at high frequencies like 20KHz. Most of it is harmonics of lower frequency sounds. Harmonics being what determines the difference between one sound at a given frequency and another sound at that same frequency. Harmonics being overtones of the base frequency at intervals of octaves about the base frequency. It's the differing amplitude of these overtones which tells us what the sound is. E.g. a piano playing 440Hz sounds different to a trombone playing 440Hz. It's because the instruments have different levels of amplitude at different harmonics.
Whether data is lost by missing these upper harmonics another debate.
The issue with 44.1KHz sampling is, how much amplitude data is lost and where. You example of frequency half of sample rate, can be extended to quarter frequency of sample rate. A 11.025KHz sound only has all amplitude data taken if the samples fall at the right places. Meaning the start, quarter point, half point, and three quarter point, of the cycle. Outside of this and digital to analogue conversion makes the peaks and troughs, i.e. amplitude, misshapen.
This is
crucial. Once you start loosing small amounts from the original signal, you start to loose the original sound. It's not a massive amount of loss, some could argue, while others argue the opposite. However once you sample in higher frequencies, you rule out the issue.
