I would think... possibly problems to do with data sampling / truncation in the realm of interpolation? Truncating a data set finitely (quantizing) to a number of samples generates the "ringing" known as Gibb's Phenomenon, right? So, said ringing would (on paper) distort spectral bandwidth in the frequencies of the spectrum, causing false high-end aural artifacts. Sort of like giving the treated sound a bad case of tinnitus... I suppose.gol wrote:the paper wasn't telling why that gibbs effects was actually bad, audio-wise.
It's all math to me:
http://www.sosmath.com/fourier/fourier3/gibbs.html
Either way, I'm VERY glad to see continued work on Sytrus, the new features look awesome!
Didier