32-bit float to 24-bit: To Use Dithering or Not?

If importing a 24-bit FLAC file into Audacity and doing some minor edits such as trimming and not processing, is it advisable or not to disable dithering? I read that if importing a 16-bit WAV file and doing non-processing type edits, it’s best to select 16-bits for the Default Sample Format selection under the Quality dialogue in Preferences, and to disable dithering. I was under the impression that when importing a 16-bit WAV file and leaving the sample format set to 32-bit float, that even if doing non-processing type edits, you may end up with unwanted sound artifacts with dithering disabled. So, the two best options I read about were: 1. 16-bit to 16-bit with dithering disabled, or 2. 32-bit float to 16-bit with dithering enabled; the second option being the “non-audiophile” way (in the context of not doing any audio processing).

Since there is no 24-bit to 24-bit option in Audacity, what is the “audiophile” way to do the conversion as compared to the “audiophile” way that has antecedently been written about in discussions about 16-bit WAV files?

If the in and the out are the same bit depth and you only did simple cutting in Audacity, then you don’t need the dithering.

Obviously, if you have a mismatch, you need it, but even if your in and out match, the instant you do anything that affects the size or quality of the blue waves, you’re dead. You need the dithering. Dithering keeps the output data errors from lining up in a way that they’re audible.


So, are you saying it’s not a good idea to do simple cutting on a 24-bit audio file in 32-bit float mode without using dithering? And if so, is there a reason why Audacity can’t natively handle 24-bit files without being in 32-bit float mode, just as it can do with 16-bit files (16-bit to 16-bit w/o dithering) – and export those without any changes to the audio?

Audacity does everything internally at 32-floating because its quality is higher than anything else you’re likely to find, and it doesn’t overload. If you Export the show at 32-floating, then, no, you don’t need dither, there, either.

Photoshop does internal color at L-a-b colorspace (not RGB or CMYK) for the same reason. L-a-b is higher quality than anything else.

You only need dither when you go from 32-floating (internal - can’t be changed) to Something Else and that Something Else doesn’t exactly match the input.

24 bit to 24 bit by cutting-only should work without dither.
16-bit to 16-bit is OK.
32 bit not floating to 32 bit not floating is OK.
Doesn’t-Matter to 32 bit floating is OK. The output bit depth and the Audacity internal bit depth match. No dither there.

If the in and the out don’t match or you apply filters or effects, you’re dead.


Really, it doesn’t matter.
When dithering to 24 bit, the level of “dither noise” is incredibly low, as is the harmonic distortion if you don’t apply dither.

If you’re doing non-processing type edits only, and import/export are 24 bit, then for “perfect” quality you can turn dither off.
For best quality when processing, dither should be enabled, but when exporting to 24 bit the difference is too small to be audible.