Reading a little I understood that when re-samplig (in my case from 44,1khz to 48khz), sample are obviously re-adjusted (since 3900 samples has been added for each 44100 ones) with a 32-bit precision. So when exporting (to 16-bit) dithering should be applied.
However, this “theory” is either wrong or i didn’t correctly understood. Because in practice it is right what you said. I carefully listened with a pair of good headphones, the only think that dither does in my case is adding a rustle (quite annoying at a track beginning).
With no dither you can hear perfect silence at the track beginning and as for the rest it is exactly the same.
So I think it is better to disable it when just resampling. Even if you don’t mind the rustle, I can’t hear any actual benefit in adding dithering.
It adds noise everywhere, not just during silence. Most of the time it’s masked (drowned-out) by the music so it only affects the very-quite and silent parts. Without dither you get quantization noise which “rides on top” of the signal. Like regular analog noise, quantization noise is most noticeable during quiet parts but unlike regular noise it goes-away with digital-dead-silence.
Dither noise is supposed to sound better (on the quiet parts) than quantization noise.
Somebody probably makes a “smart dither” plug-in that turns it off during pure-silence.
Awhile back, someone was working with 8-bit files and they also preferred it without dither. At 8-bits you can easily hear dither or quantization noise.
Most recordings made with a microphone have analog noise so they are “self dithered” and adding dither noise when you reduce the bit depth only makes it worse.
I think there is something similar to dither happening when you change the sample rate because with interpolation there are random (uncorrelated) rounding errors. (Of course this doesn’t affect silence.)
If you leave it cranked-up loud enough to hear the dither you’ll probably get bad clipping (distortion) when the music is playing. Or you might get a temporary threshold shift (temporary hearing loss) so you probably won’t hear the dither again until your ears recover!
I moved from the good old notebook to a PC a friend gave me. It’s a professional workstation mainly for 3D rendering purposes (it has Xenon, Quadro) but I guess it has a quite serious sound card too (I couldn’t hear any rustle on the notebook).
Yes, that’s what I read around. Someone said that dithering not only add the swish but it actually prevents quantization errors…
I don’t know, I was a good 10 minutes with the headphones, closely listening to the two tracks like an idiot. I couldn’t hear any “quantization distortion” on the not-dithered one. The other (dithered) is nothing better, exactly the same but with added rustle at the beginning. It is not that much of an issue, however why having the initial rustling if I can get rid of it simply disabling dithering?
The dithering noise is very subtle on the just exported wav file. But when converting to adx it increases a lot. If there’s no dithering applied, the wav to adx conversion doesn’t add any noise (it just increases if there is already).
In fact, I think the tracks I’m using already have a conspicuous floor noise. I can actually hear the floor noise even at the beginning of some of the game’s original musics (it is a Ps2 game I’m modding) on this PC (…it is like you can hear almost every bit of sound. Even the difference between wav and adx compression wasn’t that obvious with the other PC).
Interesting, it makes sense. I’ll just leave dithering disabled then. I’m making a mod for everyone to use, so it’s better this way for people with a good sound card.
