Hehe, thanks.Galleoneer wrote:Umm, with these ones you can make extra strong coffee!
To be honest they rock![]()
Are they really modified Antti ones?
Yeah, they are basically Antti ones, but not much like the source code examples that float around. I.e. all examples use 2 tanh()s per stage while only 1 is needed (nobody seems to notice this). What I did was, I simply removed a lot of stuff, kinda like nomalizing the whole equation. The basic problem is the oversampling. Normal double sampling does not work properly, because tanh() on the input sample *does* introduce aliasing, no matter what. So I came up with some kind of "drive progression" over the stages which reduced it a bit. Hope you can follow. I can't. I don't even understand the resulting code anymore.
Cheers,

