I'm curious - it seems a little around-the-houses to have to use an approximation to erf when it's not obviously relevant. Why not just use an alternative sigmoid (e.g. the logistic function) or whatever the trigonometric solution falls out to be if you model it as a 'real' slewing or rotation motion around the scene?
Interesting! I honestly only used erf because that shape felt good to me. It looks like that logistic function is much simpler, and should give me pretty much the same result.
But modelling it as a real scene sounds awesome, I'll experiment.
