[Bf-funboard] Eular rotation vs Quarternion rotation

[Martin Poirier]

--- Bassam Kurdali <bkurdali at freefactory.org> wrote:

> and eulers would be good for wheels/geers/spinning
> things.

For those, axis - angle would probably be even better.
(axis - angle => x, y, z, angle. like quats but x, y,
z does not vary when angle changes and vice versa)

One of the problem with quat for those type of joints
is that it's rather hard to set a continuous
extrapolation due to the way the three curves interact
together.

Axis - angle doesn't have that problem. Like eulers,
you'd set two keyframes and then set the angle curve
extend to extrapolation.

Another advantage is that converting between quats and
axis - angle is bijective and quats interpolation (I
would wager) translates directly to axis - angle
interpolation.

> We could also ask that quats could be editable as
> eulers- the IPO editor can pretend/lie to us :) and
> show eulers. Again , this should be optional.
> Animation master has a scheme like this.
> The one thing that could be a problem is the
> conversion from quat to euler. they are not one to
> one, but I think it is doable with some limitations.

That would probably work ok with single values, but
dealing with handles would be hell due to the
difference between quats and eulers interpolation.

> It's a bit arrogant to say quaternions are there for
> the benefit of the programer. I was animating in the
> past in euler only packages, and *everybody* was
> screaming about gimbal lock. That's why programmers
> gave us quaternions, and gimbal lock is virtually
> gone.

Not only that but interpolation between euler angles
tend to do wild swings around polar axis, a behavior
quaternions don't present at all.

Martin


      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs