[So, errr, that ‘6dB on pain of death’ waffle was a bit opaque. Someone actually told me they had no idea WTF I was talking about. Sorry about that.
Fortunately, this question comes up a lot. And it came up again today. Hopefully this explanation is a bit more straightforward…
Cheers all!]
Can I ask you a question about decibels? Does it make a difference if you are given a mixed down wav which peaks below -3db and another one which peaks around -8 to -10db, even if both are mixed down well and there is no compression or amping?
It absolutely doesn’t matter at all, as long as you export at 24-bit. In fact you can be anywhere over -48dB and still be better than 16-bit resolution.
More bits = lower noise floor. Rule of thumb is 6dB per bit, so in your example the track peaking at -3dB is using 24-bit resolution, the other is effectively at 23-bit.
24×6 = 144dB of dynamic range, peaking at -3dB means you’re 141dB above the noise floor. Peaking at -8dB means you’re 136dB above the noise floor.
So, you know, it’s probably ok 😉
Bob Macc's Blog 
#1 by Alex / Stillhead on February 13, 2017 - 1:35 pm
Well damn, here I’ve been exporting at 16bit assuming it’s always fine – I’ll need to sort my exports out now!
#2 by bobmaccsblog on February 13, 2017 - 2:21 pm
Yep, just 24-bit with no dither and you’re good to go. With the 6dB per bit rule-of-thumb, if you’re exporting under -6dBFS then you’re at 15-bit (or less), as opposed to 23-bit.
Do this quick experiment:
Get a track, whatever. Turn down master fader to say, -86dBFS, export it to 24-bit file with no dither, nothing. Then export it to a 16-bit file, no dither. Then export it to a 16-bit file with dither (whatever flavour, doesn’t matter here).
Reimport, turn back up by 86dB, listen.