I am recording my audio cassettes as 44.1 kHz 16 bits files.
The ‘Dither’ page on your Wiki says no dither is necessary when exporting 16 bits files that have not been processed.
If using ‘Amplify’ is really dither useful ?
I suppose it add only the same integer number of bits to all the samples.
Or does it make an interpolation ?
So what is the best way :
record at 44.1/16 bits, apply Amplify, export 16 bits
record at 44.1/32 bits, apply Amplify, export 16 bits
It’s not likely to make a huge difference either way.
Ensure that you have the current 2.0.5 version of Audacity: http://audacityteam.org/download/
(this version fixes a problem with applying too much dither when exporting stereo tracks).
Amplification of 16 bit audio would produce harmonic distortion due to the limited precision of 16 bit audio. Audacity avoids this problem by processing as “32 bit float” which is massively more accurate than 16 bit integer. Audacity uses 32 bit float format for all processing (for the same reason - it is a lot more accurate and so better quality).
As processing is done in 32 bit float format, it is best to work entirely in 32 bit float format and only convert to 16 bit at the end. Ideally the conversion to 16 bit should be done with shaped dither.