Problems with weighted constraints

[Sergio Mucino]

I'm running into an interesting problem with weighted constraints. I have a feeling of what's causing it, but I wanted to see if anyone has ran into it before. I've got a fairly straightforward 3-arm animation setup. One chain is the IK arm, another chain is the FK arm, and the third chain is the deformation arm. Each joint on the deformation arm has two orientation constraints, targeting its corresponding joint in both the IK arm and FK arm. The weights on the constraints are controlled by a single Custom Parameter, so I can blend with a single control. There are some Offset controls parented to each joint of the Deformation arm. And there are a bunch of little nulls sitting on the deformation arm. These nulls are position constrained to two of these offsets each. Say, one of the nulls sits midway between offset1 and offset2, then its constrained to both, with a 0.5 weight on each constraint. The problem I'm seeing is that each time I move the arm(s) (either), be it by manually moving the animation controls I have for them, or by playing with my blending slider, these small nulls seem to not return precisely to their original locations. They land somewhere in the vicinity, but they have a hard time returning to their original place. This is more notorious when I perform fast movements (for example, by quickly dragging side by side on the blending slider I have for my Custom Parameter). If I drag said slider very slowly, they stand a much better chance of coming back to their original position. My feeling is that it's somewhat related to the fact that all the objects constrained in all cases, were constrained with Constraint Compensation on. Looks like all these offsets added together are causing rounding errors during the solve. Of course, this is just my gut feeling, but I wanted to see if anyone has ever seen a problem such as this. I'm kinda surprised SI would have problems with such a simple setup... there's a bit of stuff going on (I actually over-simplified it for the purpose of this explanation, but it's nothing crazy really), but nothing that should break like this. We're on SI 2012. Any pointers are welcome. Thanks! --