Dither is one of the most fundamental yet most misunderstood concepts in digital audio. There’s a lot of theory as to how it works and why it’s needed, but knowing how and when to use dither is actually very simple:
Turn it on when exporting to 16 or 24 bit files.
When you’re working within a DAW, all processing will be done in 32 or 64 bit floating point. Whenever you want to export something to 24 or 16 bits, make sure that dither is enabled. Dither is a fundamental part of making digital audio work as intended, and it’s actually kind of strange that we are even given a choice to disable it when exporting.
There are very few occasions when it’s appropriate to disable the dither, and on those occasions, there’s also virtually no harm in leaving it enabled anyway. We’ll just get a tiny bit more noise in the exported file. On the other hand, if we disable dither when it’s actually needed, which is most of the time, then we might introduce something called truncation distortion, and that sounds quite bad.
With dither enabled, we replace the ugly truncation distortion with noise instead. All of this is happening at a very low level, way below the actual music. But at least for 16 bit files, there are certainly situations where it can be audible.
Ok, but what kind of dither should you use then? Use whatever dither that’s built into your DAW and presented as the standard or default dither.
The standard dither is usually TPDF, which is a technically correct type of dither that will always work. That’s what we use. If you want to try other kinds, like POW-r, UV22, IDR or some other types of dither and noiseshaping, just go ahead and do so.
Just remember that as long as you enable dither, the kind of dither you use is arguably among the least important decisions you’ll make when mastering.