Ah, no. Rise time of 5 microseconds translate into a bandwidth of 70 kHz.
Couple of papers taking about the differences between the sinusoids and transients:
https://pdfs.semanticscholar.org/d2...3.1619984010.1561013925-1040519108.1558324547
https://pdfs.semanticscholar.org/b6...3.1619984010.1561013925-1040519108.1558324547
If you look at the graphs with extracted transients, you'll see that the transients are aperiodic and not symmetrical relative to the zero amplitude axis. As the papers say, trying to Fourier-transform them is not productive: it is much easier to treat them as signal components of a different kind.
We don't have to go far from your site to see examples of real-life signals recorded with higher than 44/16 resolution. For instance:
https://www.audiosciencereview.com/...kris-dam1021-r2r-dac-measurements.2324/page-8.
Do you think the signals depicted in post
#143 are naturally band-limited to 20 KHz?
If they were so band-limited, everything would be peachy. In my experience, cymbal signals usually aren't. Look at the signal #4 for instance. It starts with what looks like nine periods of a sinusoid. Try to imagine removing every three out of four samples, simulating the sampling at 44.1 KHz.
On this particular graph, the samples just happen to be captured in such a way that if you start removal from a certain point, then the signal sampled at 44.1 KHz, for these "nine periods", will almost disappear. The three samples taken out represent what I was talking about - rise and fall of a pulse. In this case, they are 5.7 microseconds apart.
You may argue: but the underlying signal corresponding to these "nine periods" is still there, look at the graph - it kinda reappears later. I would say: maybe, but the mechanical momentum corresponding to the "nine periods" wasn't transmitted to cochlea with the 44.1 KHz sampling
on time, so the timing of the perception of the overall transient, even if it is eventually heard, is going to be off.
The sampling rate of 176,400 Hz appears to be adequate for this type of music signal. I can imagine how the reconstructed analog signal will approximate well the mechanical momentum transferred, and thus the transient will be heard at a time close to the time when it would be heard live, assuming same SPL for the live performance and reproduction from the captured PCM.