I’m sorry, I’m not a programmer. However it should be possible to use if statements to detect the precision of the current sample, and multiply the dither by the appropriate amount depending on precision.
And isn’t the statement “to the developers” redundant, as you are the main developer?
Exactly the “weird effect” I am talking about. Is it evolutionary or something? (the answer to this question does not belong in this forum; it belongs in some biology forum)
I’m not the main developer. I’m the forum administrator, and I do some development, some bug fixing, and some documentation.
It doesn’t have to be code written by you. If you can show me an implementation of u-Law dither by anyone, that is implemented in the way that you describe, and works, then I’ll bring it to the attention of the (other) developers. As far as I’m aware, no such implementation exists, because dithering as you propose does not give useful results, so unless you can show me otherwise, I’m done with this topic.
This is your last straw. I’m sorry, I can’t find any dithering implementation for U-law online. By the way; Rectangular dithering has also been tested with u-law, it did not work either. Rectangular dithering is supposed to add a random value between 0 and 0.9999999998 times the step size, and then round DOWN to a possible sample value, right? (similarly to no dithering, which adds 0.5 times the step size before rounding down)