If you happen to record at a very low level and you raise the volume in postproduction that extra resolution should help.
^
What extra resolution? This is the recurring claim which no one has been able to back up. Please explain or point us to a technical explanation for this.. other than a gut-feeling "that's how it seems like it should work", which seems to be the only basis for the claim thus far.
I think part of the issue here is the failure to grasp how the omnipresent noise impinges on the desired signal.
There seems to be a common misconception that the noise floor in a system is something "down there" which the signal needs to climb above if it's to be heard and, once that's happened, it is magically noise-free with the potential for infinite resolution (limited only by the convertors). This is simply not the case. The noise, whether we like it or not, floats on our signal, effectively adding a random value to each sample and inherently limiting its resolution.
Here's a simple comparison between two digital systems. It assumes that the convertors and electronics are "perfect" & is very much a first-order approximation. It doesn't take into account the dithering & resolution enhancing effect of noise, nor all manner of psycho-acoustic phenomena, etc. It's just a relative comparison. (Disclaimer!)
Say we're recording an orchestral concert and set our mic gain so the recorder clips at 120dB SPL.
We're in a *very* quiet hall, everyone holds their breath, and the background noise is only 24dB SPL. (if only...)
Then the music begins with a single quiet note, which peaks at 54dB SPL.
That note has a dynamic range of 30dB. As the reverb dies away, those of us present hear the sound smoothly decay into the background ambiant noise.
24 bit recordingThe 24 bit convertor has a dynamic range of 144dB. We've set 0dB (clipping) to occur at a sound pressure level of 120dB SPL, so the background noise at 24dB SPL is 96dB down from clipping, & occupies the bottom 8 bits of our digitized signal .
Our Convertor samples the note, capturing the peak at the 13 bit level . The "gut feeling" is that, the peak of the captured waveform will smoothly follow the instrumental sound in small increments of 1/8192 (13 bit resolution) of the peak value.
But we've forgotten about the noise! The bottom 8 bits of the convertor are jumping about randomly, with the effect that they add a random value to each sampled point. We can't be sure exactly where the wanted signal is, as the bottom 8 bits are meaningless. The peak of the musical note is only 30dB above the noise, so we have only a 30dB dynamic range. (Resolution approx 5 bits). But it won't sound "granular" or look like a staircase on the screen. If we zoom out our display we'll just see noise superimposed on our waveform; which is exactly what we hear when sitting in the hall.
32 bit recordingSo what if we're also recording with a (hypothetical) 32 bit convertor? The sound levels are obviously the same, and again we set our clipping level to be 120dB SPL.
Now our convertor has a dynamic range of 192dB. The extended low level resolution means that the noise now occupies the bottom 16 bits of our digitized signal, with the first musical note peaking at the 21 bit level. Wow! we think, "the resolution is now 1 part in 2.097 million!"
But it isn't. Again, we've forgotten the noise! Now the bottom 16 bits are meaningless in terms of resolution as they're simply noise. The signal peaks at 54dB SPL, the noise is at 24dB SPL, the dynamic range is 30dB.
Nothing we can do in the digital domain will increase the resolution of our wanted signal. In order to achieve this, we have to reduce the noise level at the recorder inputs. All we can do is to place the mic closer to the instruments, switch off the a/c, record at night,etc.
It's a pity that the conversation has gone down this particular rabbit hole, as the upgraded MixPres potentially have a lot more to offer. As a few people have already mentioned, the dual convertor topology should allow us to record low level signals using maximum gain from the low-noise preamps, whilst also allowing capture of high level transient peaks up to the clipping level of the input. Effectively that allows us to utilize the full dynamic range of the analog front end, and not worry about clipping. Ever. And that can only be a "good thing".