[Bf-committers] some notes about new armatures

Ton Roosendaal ton at blender.org
Wed Oct 12 11:22:04 CEST 2005


Hi,

> in MHO the refactoring of the animation system is really good,but  
> there are some minor issues.
> The new dag seems not to work correctly with constraint.I try a  
> strechy arm setup for a rig and there is some lag (the same thing  
> happens if I mix ipodrivers(great feature,thank you Ton) and python to  
> control objects transformations.

Python has not been 'dagged' yet fully, so that I cannot speak of. For  
the rest, the Constraints and Drivers should all work without lag.

For setting up correct and solvable dependencies, you have to keep  
these rules in mind;

1- Cyclic dependencies are not solved at the moment. One reason is  
because I know it can be often bypassed quite simple by artists, the  
other is to really get a full understanding of what kind of (useful)  
cyclic issues we need to solve.
The current system is based on a very strict "calculate everything only  
once" rule, which ensures a predictable cpu overhead (instead of  
exponentially). I don't intend to sacrificy that, although on a  
'render' command, we could calculate everything twice to get rid of  
lagging caused by cycles.

2- There are two DAG systems now, one for the Scene (object relations),  
and one for each Pose (bone relations). The Scene DAG also has all  
relations from Objects to Bones and viceversa.
However, you cannot expect dependencies to work mixed then, meaning: a  
Bone changes an Object, which changes a Bone in the same Pose again.  
All relations within the entire Pose should fit in the Scene level DAG.

Further I am extremely interested in seeing good demo files (even when  
very complex) of character rigs that have issues with lagging, that way  
I can get a good reference suite to test further improvements with.
If you have lagging Poses, and they're not violating the above rules,  
please post it in the bug tracker!

> Another problem(in MHO a big problem)is the difference between  
> rest pose and animation pose.The 2 things should be related in some  
> way,I was able to replicate an ikhandle setup(speaking in maya  
> therminology) with the follow path constraint,but the skinning is  
> done in rest pose and there is no way to syncronize it( except  
> manually)  with the animation constrainted pose,a function that can  
> transform the rest pose into the pose created by the constraints could  
> be very important for setups(I'm not a coder,I can't resolve the  
> problem myself).

Uuuh... I really don't understand this. It would help if you refer to  
the article I wrote about Armatures/Pose though;
http://www.blender.org/cms/How_Armatures_work.634.0.html

The 'rest pose' doesn't really exist, it is the armature itself (what  
you edit in editmode). All deformations - of course - are based on  
changes with respect to this initial setup (that's what we call Pose).  
There's no difference between "poses" or "animation poses", it's really  
the same stuff.

> Sorry if this is not the right place to speak about this,but it seems  
> to me that really few people tries medium-advanced setup in blender  
> and, as a consequence, some problems arising only in extreme  
> situations cannot be found and analyzed.

Reviewing current development is fine on this list. :) What we rather  
not do here is endless discussions on feature requests, especially when  
no developer is working on that.

I also work almost daily with the animators in the Orange Studio,  
they're very picky too, so you can bet they will kick me when I talk  
nonsense. :P

Thanks,

-Ton-

------------------------------------------------------------------------ 
--
Ton Roosendaal  Blender Foundation ton at blender.org  
http://www.blender.org



More information about the Bf-committers mailing list