The issue of interpolating across the 0/2pi and -pi/+pi boundaries is a recurring issue, currently "handled" by the fact that FOA transforms run at audio rate only. The current implementation in the UGens is "UNWRAPANGLE" and can be optimized.
The problem still persists when the control value sent in runs at K-rate at some point in the chain (and so is interpolated before getting to the transform UGen). This potentially calls for a new type of interpolation from CALCSLOPE in sc UGens which can deal with interpolating non-contiguous values (essentially unwrap the phase, then mod it back into range).
The issue of interpolating across the 0/2pi and -pi/+pi boundaries is a recurring issue, currently "handled" by the fact that FOA transforms run at audio rate only. The current implementation in the UGens is "UNWRAPANGLE" and can be optimized.
Here's a lead to do that: https://gist.github.com/mtmccrea/8f891a8f15f24b11f5374eeb6f4f26f3
The problem still persists when the control value sent in runs at K-rate at some point in the chain (and so is interpolated before getting to the transform UGen). This potentially calls for a new type of interpolation from CALCSLOPE in sc UGens which can deal with interpolating non-contiguous values (essentially unwrap the phase, then mod it back into range).