konradh
Geoff, What do you mean by stacking the dither over and over? I think of dither as the last step when exporting a finished product and going down to 16-bit for the purpose of making a CD.
There aren't any cases I can imagine where audio isn't passed around via a network. Imagine this use case.
1. User 1 plays guitar and sings and uses a
ShoutCast stream to send the stream to his neighbor in another country (User 2).
2. User 2 adds keyboards to the stream and forwards it to someone in another state (User 3).
3. User 3 adds drums to the mix and relays it to another stream with 100 real-time listeners.
Now you have a performance situation. The first musician cannot hear and react to what the second two musicians add to the mix, but he can get a bounce later to see how it came out.
The second musician can hear the first, but can't hear the drums since they haven't been added yet and he too would have to wait for a bounce to hear the final.
The third musician would actually hear the whole mix and stream it up to the wider audience.
In an internet chat room environment, like SecondLife or any other, a group of listeners could enjoy the performance and fairly real-time feedback. They would hear the performance within a second or two of it being played.
In this case, you have three stages at which dither may be required. User 1 would add dither before converting to a 192 bit stream (MP3 quality) to go to User 2. User 2 converts the 192 bit stream to 24 bit audio, adds his part, converts back to 192 bit stream and sends to User 3.
Now you have three dithers stacked. In this case, the first two could be triangular and the last could be Pow-R3 shaped dither.
If each musician uses a shaped dither, the shaped dithers will compound. Since Shaped dithers often have frequency regions whether the dither is 3-4 times more than other regions, it could reach the audible range on a 192 bit MP3 quality stream if shaped dithers were stacked.
So, for this case, Users 1&2 would use triangular and user 3 Pow-R3.