🚀 go-pugleaf

"Neil Wilkes" <neilwilkes@opusproductions.com> wrote in
message news:3bbb80b6.1@webx.la2eafNXanI

> The thinking behind this is to hide the quantization
> distortion incurred when reducing the bit depth of an
> audio file, as simple truncation will result in rounding
> errors - and very probably noise.
> Dither adds a more pleasing noise instead.

Dither does not hide quantization distortion, it changes the
basic nature of quantization error so that it is not
distortion at all.

Distortion has a formal definition, it is something that is
correlated with the signal.

Without dither, quantization error is correlated with the
signal (and also the sampling frequency) so it is properly
called a distortion.

Noise has a formal definition, it is something that is not
correlated with the signal.

With proper dither, quantization error is not correlated
with the signal (or the sampling frequency) so it is
properly called a noise.

Dither changes the basic nature of quantization error from
being a distortion to being a noise.

The justification for dither is that noise is generally more
tolerable than distortion.