Hi Bernard. "Link with" operations are entirely stable and
reliable, and I
have used them many times in many productions without incident.
The thing that is potentially unstable are the local Euler
results you get
from a pose or orientation constraint. Constraints are applied
at the global
level and use quaternion math for the calculations. As the
results are
calculated and applied back to the local rotations, they may
shift between
quadrants. When you have a local rotation that relies on a
consistent range
of values, say between -180 and 180, this can destroy the
relationship you
thought you were getting, since local 360 is as good as 0 to the
globalvalue. This applies to normal expressions as well as "link
with" operations.
The solution is, any nodes you are using for expressions or
"link with" ops,
make sure to set animation keyframes on the local rotation at
0,0,0. This
will force the global constraints to always resolve in the same local
quadrant, and thus you can rely on your relationships always
performingproperly.
FYI, I have notified Softimage about this issue many times. I
have logged it
as a bug and also via escalation. I don't know if it has a UDEV
number or is
being tracked. It may be one of those issues that's too deep to easily
solve, and there is a known workaround, so it's low priority.
-Brad
On 9/25/07, Bernard Lebel <3dbernard(at)gmail.com> wrote:
>
> Hello,
>
> Not really a support question, but rather getting a feel people's
> experience in general compared to my own.
>
>
> We have a bunch of characters, that I have not rigged.
> The modeler created some shapes.
> The rigger created a cpset to control the shape weights, with nice
> sliders.
> The thing is that the relationship between the shape weights
in the
> mixer and the cpset parameters are not (=) expressions, but "link
> with" expressions.
>
> Now, the problem we're continually having is that somehow the
> relationship between the shape weights and the sliders break. It
> doesn't look like it's broken (I've checked 100 times, every element
> of the expressions is there and valid), but moving the sliders
has no
> effect anymore on the shape weights.
>
> In time I have written lengthy scripts to workaround this problem.
> It's ugly but it works. However, even though it's no longer a
> show-stopper, animators are keep reporting that they have to
use my
> scripts to repair the shapes. The typical case happens when the
> animator exports the mixer and imports it back onto the model.
>
> We're not using reference models.
>
> Perhaps there is something wrong we did in the first place, I don't
> know. I don't know all the details of the rig (it's extremely
> elaborate). However nothing seems non-legit to me.
>
> Obviously, my faith in "link with" expression has taken quite
a toll
> as a result. I could have fired this to support but we're not on
> maintenance.
>
>
>
> So my question is this: how much use of "link with"
expressions do you
> make in rigging? Is my experience unique? What are the things
to do
> and not do when using those?
>
>
>
> Thanks!
> Bernard
> ---
> Unsubscribe? Mail Majordomo(at)Softimage.COM with the following
text in body:
> unsubscribe xsi
>
