start porting pd-mkmr filters for signal rate modulation of filter coefficients

[dromer]

Still need to do allpass (missing from pd-mkmr) and need a better solution for [expr~ sine($v1)] since it seems that [-~ 0.25] - [cos~] blows up with certain frequencies.

[MikeMorenoDSP]

I found the issue, we need to delete the *~ 2pi multiplication from the filter coefficients since cos~ is a cosine wavetable reader which takes inputs from -1 to 1, while expr~ cos($v1) takes inputs in radians.

[dromer]

Good catch! :D

[dromer]

So we see some minute differences in filter response going into low frequencies, which was already observed years ago: https://patchstorage.com/question/pitch-difference-in-biquad-filters-without-expr/

The solution proposed there uses a wavetable to construct the (co)sine, however this would make any patches using these filters incredibly large (and would likely not even fit on a target like Daisy).

For now I'm very hesitant to merge this, as this feature would also prevent the user from initializing the filters using static arguments. So we see breaking changes on different fronts, which is not good.

Maybe we should first get expr~ working in hvcc: Wasted-Audio/hvcc#21
And then have a more proper implementation for the (co)sine functions in there.