View Full Version : Changing rotation axis
David
03-08-2008, 05:05 AM
Hello all,
In a rig, to avoid gimbal lock, it is a good trick to change the order of the rotation axis.
What I usually do is take the axis that is less used and put it in the middle.
For example, a heel will turn sometimes in X, sometimes in Y, but rarely in Z. So the rotation axis will be XZY. It sometime helps avoiding gimbal lock.
Do you have similar tricks about this?
Thanks
David
McNistor
03-16-2008, 08:05 AM
Hello all,
In a rig, to avoid gimbal lock, it is a good trick to change rotation axis.
What I usually do is take the axis that is less used and put it in the middle.
For example, a heel will turn sometimes in X, sometimes in Y, but rarely in Z. So the rotation axis will be XZY. It sometime help avoiding gimbal lock.
Do you have similar tricks about this?
Thanks
David
after you're done animating, you can convert the rotation curves from euler to quaternion
lloyd
03-16-2008, 10:48 PM
i am having a problem. why cant i animate local!!!!!
make a cube. rotate it in Y abit. then in local try and animate it spinning round facing its new direction, it wont work!!!!
and in add move the axis is not right to rotate it how i want. whats the point in local if you cant animate with local!
David
03-17-2008, 07:44 AM
Hello Lloyd
First, read this : http://www.anticz.com/eularqua.htm (http://www.anticz.com/eularqua.htm)
Local rotation axis are just a visual help.
The true visualisation of an euler roation is in add mode.
You could go for quaternion rotation but you'd run in different problems.
McNistor
03-17-2008, 09:08 AM
in real life objects can rotate freely on any combination of axis.
3d apps can't evaluate these rotations simultaniously in euler (see rotation order). here comes the quaternion algebra which uses 4 (insted of 3) values for describing rotations.
the link posted by David should enlight you on this matter.
good luck!
pretty much the same thing, but i tend to think of it in different terms.
the first angle will move freely, the second will move the first with it, and the third will move all the angles.
so the last one would be the main rotation, and in fact the middle one is the "less animated" one, which will be the roll/twist.
4602
acorn
03-18-2008, 01:30 PM
-David
Local rotation axis are just a visual help.
The true visualisation of an euler roation is in add mode.
You could go for quaternion rotation but you'd run in different problems.
yeah, the thing is though when you get complex rotations on top of rotations the add mode axis get all screwed up. Was doing something with this and when I showed my teacher he had a kinipshin-"you can't do that" is essentially what he said.
-acorn
franky
03-19-2008, 10:08 AM
you always rotate in hierarchies. use freeze nulls as parents for your controls.
the first control rotates a freeze null (parented) and under that freeze null you have you second control that gets rotated. this way the parent/child relationship for the second control is always the same. means, you animate the first control, and the second moves to its new position, but is still at 0/0/0 in its rotation axes.
freeze1
....|
.....--- Control1
.............|
.............--- freeze2
.......................|
.......................--- Control2
it even has the great advantage that you can adjust the position of a control without chaning the animation by touching the freeze null....
kim aldis
03-20-2008, 02:21 PM
pretty much the same thing, but i tend to think of it in different terms.
the first angle will move freely, the second will move the first with it, and the third will move all the angles.
so the last one would be the main rotation, and in fact the middle one is the "less animated" one, which will be the roll/twist.
4602
you can visualise it in your head; rotation is performed sequentially, one axis after the next. So consider what will happen if you rotate about x, y, then z, each time about the global axis and you'll have a pretty good idea of what's going on.
Generally it's correct that you shouldn't animate rotations in local mode. Apart from the gimbal thing you'll tend to find that although your endpoints look correct the path that the rotation takes to get there will be unpredictable.
Also, you can get to any orientation using only 2 axes of rotation and you should avoid using 3 axes in your animation. You don't need to and it will confuse things and make it difficult to analyse what's going on when the unexpected happens. KISS :)
vBulletin® v3.7.0, Copyright ©2000-2008, Jelsoft Enterprises Ltd.