[ODE] redundant constrain problems?

[Martin C. Martin]

Graham Fyffe wrote:
 > Hi, Martin.  My problem actually arises with a kinematic loop of
 > joints in a suspension rig.  I've got this:
 >
 > car chassis - ball joint (horizontal) - suspension arm - ball joint -
 > spindel - hinge joint (vertical) - strut bottom - slider joint - strut
 > top - hinge joint (horizontal) - car chassis

I couldn't quite picture your setup from your description.  Is it like 
one of these diagrams?

http://www.chris-longhurst.com/carbibles/index.html?menu.html&suspension_bible.html

> Now, a real car has a hinge joint instead of a ball joint for the
> first joint, attaching the car chassis to the suspension arm. 
> However, if I do put in the hinge joint, then the car does
> physics-defying acrobatics on its own.  Without the hinge joint,
> things work fine except for extreme conditions.
> 
> Is it possible that somehow these two hinge joints are in fact
> conflicting constraints?  Is it possible that they are actually
> redundant, but that floating point inaccuracies are making them ever
> so slightly different, and therefore conflicting?

That's possible.  Are you sure you haven't accidently mixed up the axes 
somewhere?  It's pretty easy to rotate something the wrong direction or 
otherwise end up having an axis facing the wrong way.  I've always 
thought the demos should be able to draw e.g. the hinge axis or other 
relevant axes.

Anyway, if that's not the problem, it could be ODE trying to overcome 
small errors.  Although the solver is supposed to be good at dealing 
with those.  Have you tried using different solvers?  Does the choice of 
solver make any difference?

- Martin