Well, this also works, but it takes way more work to setup.Urs wrote:I do a lot of tests that look like unit tests but plot out number of iterations and measured time for, say, 100000 samples. I do those for different settings (cutoff, resonance, input material) and different accuracy (use sqrte() for tanh or 1/sqrt(), use 3rd order polynoms or 7th order etc.). On top of that I do different initial estimates (last sample's Vs, linear prediction, Vs from linearly computed filter). For this you better use C++ templates.
