A coupleof years ago I made a shaper with used the first 10 chebby polynomials
It worked great and correctly , since a 0db sine wave as input give all the overtones for each polynomial .
BUt , I am having someserious brain fart issues
Take this chebby shev one
4x^3 - 3x
I implemented this is reaktor like this
Which I think is correct ..
The input signal is cubed , then mulitplied by 4 , and eventualy the signal mutiplied by 3 is subtracted from it .
Let's focus on the first portion , should the cubed input sig be mutiplied by 4 , or should the input sig be multied by 4 and then cubed ?
Two different things , as seen here in pure data
edit .the input signal in the pure data patch is a ramp going from 0 to 1 over 512 index points
It's not real time audio , but a pure recursive calculation writing the formula in table .
There is an obvious difference whether the input signal is first Cubed then multie by 4 (bluebox ) compared to (4*sig)^3
It's impossible that my reaktor implementation is wrong since it bahaves as chebby poly should behave.