Dithering audio is mixing lo-res (16-bit) audio data with special noise (dither) for supressing of quantization noise. It make sound more clear.
Quantization noise is nonlinear distortions appear by reason of errors of conversion of hi-res (24 bit and more) signal from low-res and vice versa.
Discordant of quantization noise as assumed is consequence of correlation the noise by musical signal. Other words (not exact definition), musical signal generates quantization noise similar to the signal.
Thus dithering masks quantization noise suppressing it.
Of course, dither is additional noise of resulting sound. But it is very quiet.
Exists several algorithms of generation this noise. Its spectrum may be adapted to spectrum of musical signal (masked by music, as example).
alphaD is original dithering algorithm of AuI ConverteR 48x44 mask quantisation noise with minimal intervention into music.
Download DEMO AuI ConverteR 48x44 and try alphaD dithering algorihm.
Recommendation of alphaD applying for sample rate conversion:
1) Input audio file 16 bit integer --to--> Output file 16/24/32/64 bit integer and 32/64 bit float;
2) Input audio file 16/24/32/64 bit integer and 32/64 bit float --to--> Output file16 bit integer.