On Mon, Jun 07, 2010 at 10:41:07PM +0200, Philipp wrote: > This is probably a stupid question.
Not stupid, but maybe worded in a way that makes answering it quite impossible. > My guess is that quantisation noise is only something present between > the input signal and its digital representation, and hence no change of > the digital representations can do anything about it. Noise shaping and dithering make sense only for 16 bit or lower. For input (A/D conversion), if your converter is only 16 bit then very probably the analog part isn't really high quality, so even in that case analog noise will probably dominate any quantisation noise, making the latter irrelevant. If your A/D converter is 24 bit, then analog noise will always dominate, so again quantisation noise is irrelevant. The only case that remains is when a digital signal is converted to analog using a 16 bit D/A, or converted to 16 bit digital, e.g. for CD. The purpose of dithering in that case is to convert systematic quantisation errors (i.e. errors that would be correlated with the signal, and therefore appear as distortion and not as noise) to noise. In its simplest form this is done by adding noise, resulting in a S/N ration that would be 3 dB worse than without dithering. This completely removes any correlation between signal and error. Noise shaping and error feedback are used to avoid that S/N ratio degradation. It works by moving most of the noise energy to frequency regions where it matters less. You can see some examples of this here: <http://www.kokkinizita.net/linuxaudio/dithering.html> The last one (using noise shaping) has the worst S/N ratio if you measure it without any psychoacoustic weighting. But it will sound the best. -- FA O tu, che porte, correndo si ? E guerra e morte ! _______________________________________________ Linux-audio-dev mailing list [email protected] http://lists.linuxaudio.org/listinfo/linux-audio-dev
