neuronaut wrote:This looks like you're not doing proper overlap-add or overlap-save. Are you windowing the output signal and adding the windows?
Your observation is a good observation. It was the first thing I controlled. Here it's simply impossible because you are looking at the grain of sound without OLA.
This is the result of Phase Unwrap of a single windowed (hann) grain of a 400Hz sinus. It's so strange that i can't believe it, but my phase unwrap destroy the grain.
In the C++ code that I have uploaded there's no OLA, but you can look at single grains after Phase Unwrap. Obviously even OLA is out, but only because grains are really deformed in Phase Unwrapping. I'm making a mistake that I'm not finding.
I have no idea....
Sob sob sob.