For interpolation, I've used Linear, Cosine, Catmull ROM, and higher methods (using even more points) and frankly, there's very little difference to my ears, or to my spectrum analyzer. I think you really need to go up to 16-point or better before things get noticeable, but 4-point methods should be fully adequate without murdering the CPU.
Statistics: Posted by syntonica — Wed Apr 10, 2024 3:02 pm