I have seen people and me myself using twist joint setup (usually while rigging forearm twist) using the good old rotational weighted constraints or direct rotational connection with offset values to give even twisting. It works great for straight aligned joint rotated or twisted in just single direction for example rotate X, but once you start rotating in all direction which we all tend to, some thing weird happens, something not so good that is known as flipping. This flipping behavior is the result of GIMBAL LOCK! Well then, that is a major issue :):
I have blogged a little overview about how we can avoid this :
Until recently I was also struggling with twist joint setup. On your blog you seem to know Felix Joleanes website: what I found really informative was the part on flipping less rigs. I don’t know if it can help but I started investigating an alternate workflow : I have converted the information found in this tutorial into a quaternion based aim constraint node using the python maya API. From this experience I was able to write a custom twist reader which ultimately drives a custom torsion deformer. .
Coincidently from your thread I also start writing on this subject on my blog: http://circecharacterworks.wordpress.com/2011/07/20/toward-a-quasi-time-independant-twist-node/
Cool! Yeah I know Felix and his website; his works are inspiring! The setup that I posted on my blog is pretty straight forward and simple to use, just the regular World Up Object vectors.
I like what you have achieved with the custom python node. I have linked your post to my blog post as well. You and your blog is going to be on my hit-lists. Cheers :):
(Maya) I used a pointer and connected in the x rotation via the connection editor to the wrist joint. Then used that pointer to control the twist. This way your only getting the rotation data for the axis you want, eliminating the flip.
So you are using a purely manual setup, like http://www.3dfiggins.com/writeups/forearmTwist/ . I see many relevant point on Kiel Figgins article, as long as work is done on time and on budget whatever techniques is used isn’t important. Does your solution scales well with 20 , six-armed character on screen in a 5min shot? Personally i think an animator time is worth more working on timing, silhouette and acting than carefully defining the twisting behavior of a character skin.
.