Question about Bit-depth and dithering

Sorry if I seem a bit clueless. Im new here.

Does mix and render count as processing in 32-bit, which would require a dither to prevent rounding errors, or can I just downscale it back down to 16-bit without any issues? The tracks im using were all originally 16-bit and are only being spliced together (no overlay mixing is happening), so i dunno if its even worth using a dither, and if all it will do is slightly decrease the track’s quality, when there weren’t going to be any rounding artifacts in the first place.

Im aware that small edits, such as cutting and pasting dont require any dithering, but im just not sure about this particularly.

No, you don’t need to dither.

If you are actually mixing, mixing is done by summation and mixing two 16-bit tracks requires 17 bits, but one extra bit probably doesn’t need dithering. And there are other effects/processing that effectively increase bit depth.

…It’s really not a big deal one way or the other because at 16-bits or better under normal conditions you can’t hear dither or the lack of dither.

“Mix and Render” does count as “processing”, but in this case Audacity will automatically know that it does not need to dither and will not apply dither.

About 2 years ago (Jan 2023), this change was made to Audacity: Avoid unnecessary dithering (achieving bit-perfect import and re-export of 16 bit data, and other things) by Paul-Licameli · Pull Request #698 · audacity/audacity · GitHub
In short, what that does is it allows Audacity to know when dither should be used and when it should not. Assuming that dither is enabled in Preferences, Audacity will only apply dither when needed, and will not apply dither when it is not needed.

I was originally a bit suspicious of this change as I thought there may be edge cases where Audacity might make the wrong decision, so I tested this change very thoroughly and was pleased to see that Audacity correctly handled everything I could throw at it.

In short - just leave dither enabled in Preferences and let Audacity figure it out (because it can).

Thanks, I appreciate the help.