Talk:Bit Depth

32 bit depth
Do you know of any effect that will cause the noise level generated by the 32nd bit of resolution to appear in the output file at 16th bit depth or higher? I cant think of any.

Nor can I even for 24 bit recording. To get noise of the 24th bit to appear at the 16th bit one would have to increase amplitude by a factor of 256 (= shifting by 8 bits). NT 22:17, 9 October 2007 (PDT) Amplify will do that! An AVC (automatic volume control) effect that attempts to boost quiet passages could have that effect too, if the parameters are sufficiently extreme. So could an equalization effect. It's not so much the effect that matters as the parameters passed to the effect. You'd generally only get away with such extreme parameters without clipping if the original source was recorded at too low a volume - the real source of the problem.

I think this question hinges on magnitude. To shift a signal 8 bits means amplifying it 256 times, or 48dB. Are there any real world apps where the level of the recorded signal is 48dB or more too low? If not, choosing 32 bit over 24 gives no advantage.

Audio AGC doesnt normally cover anywhere near this sort of dynamic range, least not that I've seen. NT 20:23, 10 October 2007 (PDT)

When I record conserts at my church, between songs there often someone talking at a much lower level compared to the choir singing at max level. Often they need to be raised as much as 32 dB. That would mean that rendering to 16 bits, at least 16 + 6 bit make a difference. So at least 24 bit recording will make a difference compared to 16 bit.

Moreover; whether 32 bit makes any sense is if you have a 24 bit soundcard and want maximum quality even after a lot of editing. But it's good that software is a bit ahead of hardware - no soundcard today (as far as I know) can record and play back 24 bit of accuracy (see here for some tests). --Diwic 13:04, 7 January 2008 (PST)

Space used by 24 bit depth
Q: ''When I record 24 bit, the text indicating how many hours/minutes there is left to record is the same as for 32 bit float. Is this a bug (in Audacity 1.3.4) or is Audacity throwing away one byte of four (possibly to help alignment) when I do 24 bit recordings? ''

A: It looks like an error in calculation (incidentally going back to 1.2.4 or before). The disk space used to make the recording is correct. We'll take a look at it. By the way, asking questions on Wiki pages isn't usually the best method of getting an answer. The Forum is usually best. Please see Reporting Bugs - Gale