Dither was originally introduced as a way of linearising the analogue-to-digital conversion process. By adding a small amount of noise at very low level, around that of the least significant bit, the quantising distortion that would otherwise occur could be converted into benign noise. Magic to many, possibly counter-intuitive, but very useful. In an all-digital signal chain, conversions from one resolution to another in the digital domain can be necessary, such as when reducing the resolution from 24 bits to 16, for delivery on CD or other consumer media. If you simply throw away the bits you don’t need the old menace of quantising distortion reappears, rather as if you had made the recording at 16 bits without using dither. That’s why you need to reintroduce dither at the correct level for the new resolution. The simplest flavour of dither has a flat spectrum, but various attempts have been made to shape its spectrum so as to create something less ‘noisy’. By reducing the level of the noise in the most sensitive region of human hearing the apparent dynamic range is improved, leading to claims such as ‘20-bit resolution on a 16-bit CD’. The trade-off is increased noise in other parts of the spectrum, but you can’t hear this as well so it may not matter. The choice of these dither flavours has become something of an artistic decision, providing mastering engineers with a selectable tool for optimising the sound of a recording. The differences are very subtle but people have their favourite plug-ins and curves for different types of music, for example. You’ll find a lot of disagreement about which is the best or whether you need to use it at all. Trust me, though, you should use it unless you particularly like the effect of low level distortion when you reduce the resolution of a recording. More esoteric approaches have also been used successfully, such as Apogee’s UV22, which adds modulated dither on a high frequency inaudible carrier at roughly half the sampling frequency (sometimes called Nyquist dither). Apogee’s more recent UV22HR algorithm is now available as a plug-in for workstation software such as Logic, Cubase and Nuendo. It claims to offer automatic selection of the most appropriate algorithm for the destination resolution and format. It also optimises the dithering process for data-reduced destination formats such as AC3 and MPEG. POW-R dither is another option (standing for Psychoacoustically Optimised Wordlength Reduction), which includes different options such as two flavours of noise-shaped dither and a form of Nyquist dither. Waves also has a number of tools in this market, including its IDR360° plug-in that claims to be optimised for requantising surround recordings, and has various degrees of noise shaping from none to lots.