The fade calculation cubed, as opposed to cube rooted, gain->amp and also did it before subtraction defeating the purpose. So change the fade calculation to have a standalone fade gain to amp calculation (which here we use an approx to cbrt which works pretty well) and then apply it after the f, 1-f not before.
This makes fade loops sound way less uneven. Still some focus and debugging I'm sure but generally better.
The fade calculation cubed, as opposed to cube rooted, gain->amp and also did it before subtraction defeating the purpose. So change the fade calculation to have a standalone fade gain to amp calculation (which here we use an approx to cbrt which works pretty well) and then apply it after the f, 1-f not before.
This makes fade loops sound way less uneven. Still some focus and debugging I'm sure but generally better.