solderdude
Grand Contributor
Do you realize this is not a graph of a 20kHz signal, but instead a graph of a complex waveform consisting of more than one frequency?
yes.
Why is in bold a false statement? Because you don't see a 20kHz waveform on the graph. And you can't actually tell the amplitude on the 20kHz part of the waveform on graph either. Because it's mixed with the rest of frequencies in single graph from which you can't visually separate a 20kHz from the rest.
Actually I can .. IF it were a 'mix' of frequencies the peak of the 20kHz would be higher than the few peaks you see there. That peak actually is the -1dBdB FS and is what the original signal SHOULD be and is when a proper reconstruction filter would be used behind the R2R conversion.
But what we do know? Physics. Well, at least I know it and physics tells the continuous tone of 20kHz, which is a fundamental tone in this graph, well...won't have its amplitude changed because of presence of the rest of frequencies (which are higher than a 20kHz tone). Because amplitudes of various frequencies don't add.
If it were an addition (superposition) of 2 frequencies then the average level would be -3dB. Which is what we see on a frequency plot.
So If the extra frequency would be removed by a sharp filter the average would be -3dB because you remove the extra signal.
THERE is the roll-off which you insist does not exist.
You keep insisting that your ear 'removes' the ultrasonic content and thus the 20kHz which is 'hidden' in the complex waveform re-emerges and the amplitude is correct again. This would mean that amplitudes do not add but only subtract.
And we do know that superposition (mixing) of signals is an addition of signals not a merging.
Leaving out the (obligated) reconstruction filter and assuming the ear takes care of this by 'ignoring' everything above 22.05kHz is a novel idea.
Let's just say that you haven't come with compelling evidence you are correct in this.