Something has been bugging me lately:
If I import a 16-bit file, process it in 32-bit float (amplification, reverberation, delay…), export it in 32-bit float, re-import that output, set dither to “none” in the quality settings, and then export it in 16-bit…Would that avoid applying dither twice while creating no additional quantization errors, or on the contrary create harsh harmonics because no dither was applied when down sampling from 32 bit float to 16-bit?
if the latter point is correct, could someone please tell me how to avoid repeated dithering in addition to not creating more rounding errors?
My sincerest thanks for reading my request! May everyone have a magnificent day !
When exporting as 32-bit float, dither is not required and is not applied. This has always been the case in Audacity.
Also, recent versions of Audacity are a bit more clever about when to apply dither. If you import a 16-bit file, edit it (cut / copy / paste / delete) but do NOT apply anything that modify sample values (no processing), and then export as 16-bit, Audacity will not apply dither, because it knows that all samples can be represented exactly as 16-bit values.
I know audacity does not dither when exporting to 32-bit float and that’s the reason why I included these extra “convert to 32-bit float” steps in my original post. I just wanted to make sure that exporting a processed 16-bit audio to 32-bit float, re-opening that 32-bit float audio file in audacity (without re-processing it) to then export it to 16-bit would not result in additional dither/rounding errors.
but I guess it’s all good if you’re telling me that audacity does not dither if you don’t process the input file
thank you for your time! I wish you a marvellous day!