I've been working a lot with linear and angular springs, and during this work a simple yet hard-to-solve problem has occured, and I hope that you might be able to help me...
When simulating springs it is very helpful to ad a velocity dependant damper. The one I use looks like this (in 2D):
linear damping force = -Kd * ((V.x*L.x+V.y*L.y) /|L|)
where Kd is the damping coefficient, V is velocity vector, L is spring vector, and |L| is spring length. Now, it is very tempting to add a similar angular or normal damping in the form of:
angular damping force = -Ka * ((V.x*L.y-V.y*L.x) /|L|)
Where Ka is the angular damping coefficient. This works fine in the sense that it makes wobbly soft bodies act much more like bags of oil or even sand, and it makes rows of connected point masses and springs behave much more like a real rope or chain.
The trouble is that the angular damping (as opposed to the linear damping) is not a strictly internal force. When applied to a free falling soft body it also dampens its movement as if under influence of air drag, and angular damped ropes stop swinging much faster than undamped ones.
I apply the angular damper pairwise for each set of endpoint masses at a time, and in opposite directions, so it should act as an internal force.
So, the big question is: How do you implement a strictly internal angular damping force?
Cheers Michael
as it does introduce world torque.
