Cage opened this issue on Nov 15, 2002 ยท 9 posts
Cage posted Fri, 15 November 2002 at 1:37 PM
Ooh, it drives me nutso. Why do the axes become confused? How can the problem be avoided? I did an animation recently in which I had a prop spinning around its z-axis while it also zoomed along the z-axis toward the camera. Everything was fine until I saved the PZ3 and re-opened it. After that, the prop was no longer rotating solely around the z, but also along the x and y, flailing about wildly. Was this Gimball Lock? Why did it happen? I rebuilt the whole scene and it saved correctly the second time, but I have no idea why it was inconsistent. Should I have paid better attention in Trigonometry class? Dang, they warned me that I really would need to know all of that one day....
===========================sigline======================================================
Cage can be an opinionated jerk who posts without thinking. He apologizes for this. He's honestly not trying to be a turkeyhead.
Cage had some freebies, compatible with Poser 11 and below. His Python scripts were saved at archive.org, along with the rest of the Morphography site, where they were hosted.
VK posted Fri, 15 November 2002 at 8:29 PM
Hello Cage, I don't know what was going on with your scene, but I think I know what "Gimbal Lock" means: Your model has 3 rotation channels. If you use only one of the channels, then the rotation axis is aligned with one of the main axes of the Poser scene. But if you set two or three rotation channel at the same time, then the rotation axes are modified, and strange things can happen. Whenever you set the middle (second) rotation channel to 90 or -90, the first and third rotation channels rotate the model around the same axis of the Poser scene. So, your model loses a rotation axis, whenever the second rotation channel is 90 or -90. This is "Gimbal Lock". For example: yRot = 54 xRot = 90 zRot = 0 looks exactly like: yRot = 0 xRot = 90 zRot = -54 The middle rotation (xRot in this example) at 90 locks the gimbal completely, so that the first (yRot) and third (zRot) rotations share the same rotation axis. This happens, because Poser computes the rotation channels one after the other, and the rotation axis of later channels is modified by the rotation of prior channels. So you can't avoid this effect. You can sometimes arrange the rotation channels (the rotation order) properly, to minimize gimbal lock effects: You find out, which of the rotations won't be close to 90 or -90, then you place this rotation between the remaining two rotation channels. The arms and legs of Poser humans have for example the rotation order "Twist", "Side-Side", "Bend", because the middle rotation ("Side-Side") will never be close to 90. When you try to set "Side-Side" to 90, you lose a rotation axis, and "Twist" looks exactly like "Bend".
Cage posted Sat, 16 November 2002 at 4:00 AM
I see. They spent a lot of time talking about gimballs in that astronaut movie with Tom Hanks. Is this sort of problem a universal trouble with 3D computer coordinates? Does NASA have this problem? Hmm.... Hmm. I will have to think about this. Makes me glad I live outside of Poser and look in, instead of vice versa. Thank you!
===========================sigline======================================================
Cage can be an opinionated jerk who posts without thinking. He apologizes for this. He's honestly not trying to be a turkeyhead.
Cage had some freebies, compatible with Poser 11 and below. His Python scripts were saved at archive.org, along with the rest of the Morphography site, where they were hosted.
bloodsong posted Sat, 16 November 2002 at 9:08 AM
heyas; where's that post with the illustrated posette gimbal lock torture device? you know the one i mean...! where... who was it?? figured out to avoid gimbal lock by parenting the hip to a prop for the one rotation....? come on guys, help a senile old poser techie out, here!
bloodsong posted Sat, 16 November 2002 at 9:12 AM
Attached Link: http://www.renderosity.com/messages.ez?Form.ShowMessage=878305
here it is. (you thought i was nuts, huh!?) and animated too! thank you, ockham!ScottA posted Sat, 16 November 2002 at 11:37 AM
Gimbals are angles (which we write as numbers of degrees). So it relates to a lot of things. If you were physically able to bend your knees the same amount in both directions. Then your brain would most likely suffer from gimbal lock if you straightened your legs out in a straight line when you were walking. As long as you kept them slightly bent one way or the other at all times. You'd be fine. ScottA
wadams9 posted Mon, 18 November 2002 at 3:00 AM
I think VK deserves a round of applause for one of the simplest, shortest, yet clearest explanations of gimbal lock I've ever seen. Whenever I've tried to explain it to someone else, I get brain lock.
Anthony Appleyard posted Mon, 18 November 2002 at 5:12 AM
That is why I would like to be able to switch to a mode where directioon is described as direction cosines, which do not have locked positions.
VK posted Mon, 18 November 2002 at 4:06 PM
wadams9, thanks for the compliment :)