Comment 20 for bug 1086999

So, if I understand correctly you meant to normalize dt so that the time step is 1. Now I understand but I do not think that this works. If the normalization is just a notation change then it should have no effect on the calculated values. But you can see that it does have an effect because the integral and derivative terms are calculated differently. The integral is discretely evaluated by summing error * dt whereas the derivative term is error/dt. If you divide the inputs to observation() by dt then I agree the derivative term is equivalent to having removed the /dt factor but the integral term is not the same as if you had not applied the transformation.

This is all a little moot since we do not use I or D control but we should stick to the standard equation:
http://en.wikipedia.org/wiki/PID_controller#PID_controller_theory