evo ti objasnjenje za dither by Paul Frindle
Dither is not a workaround in the least - it is a requirement in order to turn the quantised mathematical representation into a continuous transfer function - i.e. without the statistical discontinuities. This is what is needed to make it match the properties of the physical world - where it will inevitably emerge - hopefully
Random and uncorrelated dither adds at 3dB for each doubling of the number of sources.
The simple formula for this is
Noise in dB = 20 * log(Sources ^ 0.5)
So 48 of your dithered 24 bit sources added up in silence would give:
20 * log(48 ^ 0.5) = 16.81dB
The dynamic range of a dithered 24 bit signal is around -143dB
So the total dynamic range of your 48 dithered sources all added up would be
-143 + 16.81 = -126.187dB
Clearly plenty good enough for rock and roll - and of course 16dB better than your DAC. BTW watch out for DAC noise figures because they are inevitably quoted as A-weighted - which takes out the top end freqs where most of the noise actually is. This has the effect of producing figures that can be up to 10dB better than your actual experience of listening to it.
So in concept whenever we do anything to a digital signal (even change the gain) we have made a whole new signal that requires dithering (adding noise - just like in the real world - just like in analogue). However because digital representation (being entirely mathematical) does not have to be constrained by the real world and the laws of physics, we can throw whatever precision we want at it - so the results can be better than anything possible in the real physical world - if we so chose. But what we actually do when designing systems (within reason, based on common sense and science) is to dither those signals that would give rise to significant quantisation distortion if left undithered. So for example we are prepared to ignore stuff at 48bits down (or 64bt float down) on the basis that distortion at -288dB is not going to be significant - and of course we take into account other factors like the error increase due to processing algorithms etc.
But we do dither signals that end up being represented at 24bits (i.e. PT TDM, PowerCore etc.) because at the very limit it is possible that distortion at -144dB might just get to be audible. And of course at 16bits for the final media when the distortion would be at -96dB, it is obviously quite important - even in a domestic listening environment.