What is Dither?

In an audio context, dither is random low-level noise added to the signal during conversion to a shorter word length (bit depth - for example from 24 bit to 16 bit).

It has the effect of recovering much of the quietest part of the audio that would otherwise have been lost in the conversion process.

There is a tradeoff – a slightly increased level of background noise. However there are two factors that mitigate this: Firstly, humans are able to hear sounds that are quieter than the background noise, and secondly, a technique known as noise shaping can be employed to move this noise as much as possible into frequency areas to which the human ear is less sensitive.

As a result, it is universally accepted that dither is a good thing, and we always apply it (with a suitable noise shaping curve) when carrying out conversions that involve reducing the word length.