Back to audio, since humans can hear only up to 20kHz at best there is no reason to go above that as Nyquist frequency.
This is not an inherent property of what the nyquist limit does. Your statement is only true if you've used a lowpass filtering in your mastering process so that frequencies and harmonics above the Nyquist limit are filtered out before they are recorded.
For example: A Nyquist limit of 22.05kHz does not mean that it will not sample 33kHz. ANY sampling frequency will intersect with that 33kHz wave in a way that produces a series of values that result in a different wave at a lower, and often audible frequency.
There are only two ways around this phenomenon:
1. Use a lowpass filter at the nyquist limit to completely prevent any frequencies above that limit from being captured at all. The nyquist limit is a conceptual limit, not an actual barrier to sampling frequencies above it.
2. Choose a Nyquist frequency so far above the A-weighted range that frequency roll off and frequency aliasing become completely immaterial, as they will not produce other artifacts inside the range of human hearing.
The other general rule is to work in multiples of 24. So if you're sampling at 192kHz during the recording stage, don't mix down to 44.1, but rather 48 or 96. This reduces downsampling error... though computer processing is so powerful these days that this is a far smaller concern than it used to be.
But again, I want to be clear that the Nyquist limit does not prohibit frequencies above it from being sampled. Even the sampling frequency doesn't prohibit that. If you sampled a 33kHz wave thirty-thousand times a second, it would produce an audible distortion... not the original wave, mind you, but the resulting ALIAS of the wave--a lower frequency wave--which ends up inside the A-weighted spectrum.
More detail here.
Excerpt:
"Although sampling at twice the Nyquist frequency will ensure that you measure the correct frequency of your signal, it will not be sufficient to capture the shape of the waveform. If the shape of the waveform is desired, you should sample at a rate approximately 10 times the Nyquist theory."
192 is closer to 10x nyquist than 22.
Last edited: