Am Sun, 16 Aug 2009 15:00:21 -0300
schrieb Arthur Pires <arthurprs@...>:
> I'm testing this new feature before upgrading,
> but i can't hear any diference using dither, my ears are bad or the
> difference is very small?
The difference is small. Compared to full scale, we are talking about -96dB level here. It depends on your hardware (both biological and electrical) if you notice. Note that the generic mpg123 decoder used (and by default still uses) truncation instead of rounding for conversion to integers... it took quite some time until someone noticed -- and then, only after looking at a spectral plot.
So, the dithered generic decoder combines proper rounding with the dithering (wouldn't make sense otherwise) -- and it's all about the lowest bit in the sample value.
An extreme case to make it obvious is offered on
There's a file, hotel.mp3, that is encoded at very low volume, using only lowest bits of samples for all the sound. Decode that to wav with and without dithering, then amplify (normalize) the WAV, listen.
Both will sound awful, but the noise pattern is different.