My ears are middle-aged now, but when the Nyquist prompt is applied, I hear a slightly richer “tintinnabulation” (for want of a better word) but more noticeably I sense an slight increase in pitch, especially for the three-samples-repeated script. With three samples repeated, the spectrum itself begins to look “stair-stepped”.
So are we saying that the sound of the recording you found that had the desired sound could be genuine, that is it could be captured and saved as a digital audio file if the DAC played it at 20000 Hz and providing the recording was made at a minimum of 40000 Hz?
Or must that recording by definition be itself some kind of “fake” or “emulation”?
Yes, that appears to be the case, though we have no information about how the 40,000 Hz version was made.
The original SpecDrum hardware was very inexpensive (£29.95 GBP) add-on for a ZX Spectrum in the mid 1980’s. It was quite common in those days for computer D/A converters to produce unfiltered stepped output. The “steps” would generate additional harmonics, which were only limited by the bandwidth of the analogue circuitry. It was generally expected that the loudspeaker would filter out these harmonics, but in the case of “noise-like” sounds (such as a snare drum) the additional harmonics would help to make the low sample rate audio sound a bit brighter. I remember being impressed by the “quality” of 8 bit “real sampled audio” (compared with the more common 1 bit PWM sounds).