I think you can find some answers in
this post (from 09.U-0a setup with external USB DAC).
And if, what effect should thiss have on sound quality. Less dynamics, bigger noise floor?
Less dynamics is what everybody would say. This is stupid theory though (IMO), because *or* you are not able to hear the full range, *or* -when you think you do- you will hear distortion at the low end instead.
Just try it with XX which very officially presents you the attenuation in dB. Take a 16 bit random file, set XX to -0dB, the preamp to any volume you think is bearable. After this, set XX to -90dB. Officially one bit is left now, and no music can come from it (and no dither is applied). However, when you are able to hear it, you'll hear it as distortion. Now, can you hear anything ?
Keep in mind, the volume of the preamp should be kept at bearable listenin levels !
My suggestion is : no. You can't hear it.
So now tune up XX until you start hearing something. Ok, in the before you obviously were with your head in the speaker, but keep in mind this is not much realistic for obtaining "dynamic range". So better step back to the listening position, and find out when you are starting to hear music at tuning up XX' volume.
Say your answer is -66dB. This would mean that the dynamic range you can obtain is ... well, 66dB. Keep in mind that the 0dB is the "bearable level !".
Sidenote : "bearable" might be 105dB SPL (listening position !). So what you -in my example- actually still can hear is 110 - 66 = 39dB.
With -66dB you'd have lost 11 bits so 7 remaining. Way too few for good music. But *now* you'd have to search for music that allows playback at 105dB SPL *and* has music information at -66dB. Of course this exists but it will be more rare.
A bit depending on the outcome (and your system), you can imagine what happens with a 24 bits DAC. This means 144dB so called dynamic range, and there is NO WAY to unveil it.
The noise floor too, IMO is a phenomenon which does exist in theory, but in practice works out slightly different.
My point would be this (disclaimer : by lack of real signal process knowledge, haha) :
In a DAC there is nothing like analogue noise with will show underneath its capabilities. That is, not derived from the
digital capabilities (from the analogue output stage there is, but this is not about the "noise floor" as we speak about it). You must look at this relatively :
When a 24 bit DAC is fed with 16 bit material derived from a 24 bit file (like you can do now since 0.9u-2) the cut bits will create noiSES. This is nothing else but harmonic distortion occurring at a certain level, and that level will be just around the noise floor of 16 bits.
Now here we go again : the noise floor of 16 bits would be -96dB, and I think in the above we determined that there is no way that you can hear that. The dynamic range you'd need to obtain it, is too large.
The same noiSES will appear when you feed a 16 bit DAC (or any bit DAC for that matter) with, say, 12 bit material derived from a 16 bit file. Again, the relativeness is important, because again the *cut* bits will create noiSES, or better, the remaining 12 bits actually needing the 16 bits will do. So the noiSES floor is 12 bits, and this is -66dB. See the story above.
What actually happens ?
A more fluent sine is cut into more square cutoffs because there's not enough bits anymore to keep it as smooth as it was. It is those squares causing the harmonic distortion, or maybe better : squares have a signature of over tones that don't belong to sines.
A side effect of squares is that they bear more energy than sines. And this is why this "distortion" can be audible above the noise floor. They *will* be when the needed attenuation is so large - and which follows from too much gain - that they're in the audible level.
PS: for fun, look at a 24 bit file opposed to a 16 bit file. When you follow my reasoning, you'd conclude that a 16 bit file is full of (relative) harmonic distortion opposed to the 24 bit file. The sine in the 24 bit file is rather endlessly more smooth opposed to the 16 bit file.
Now, who says - or what determines whether only 24 bits is good enough, 16 bits can do fine, or 14 bits are okay just the same ?
The dynamic range you can perceive does !Now I think you can better understand why I said in the Releasenotes of 0.9u-2 that to my theories you wouldn't hear the difference between a 24 bit file played as such, and the same 24 bit file cut to 16 bits (when played with XX
).
