Re: [osg-users] Matrix Quat precision
Hi Vincet, Even with doubles there are still precisions issue, just much smaller than with floats. If you are using a iterative calculation then if you aren't care the errors could accumulate until their become visable. The best thing to do is to avoid an iterative calculation and calculate the new camera position based on the position of the node in the scene on each frame. Robert. On Thu, Mar 27, 2008 at 8:38 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi all, I'm currently making a new Manipulator inherited from MatrixManipulator. The specificity is : On the call of a function, the manipulator (camera) have to follow a node which is moving. I mean that the camera have to be exactly immobile relative to the moving node. For that i've used Matrix, Vec3 and Quat. All looks great, but after 10 minutes of running, I can see that a little moving difference between the node and the camera... if the node rotate about 2 or 3 axis, the difference is more important... Using Vec3d, Quat and Matrixd, is there any reason for this lack of precision ? Or is it my method which is not good... ? Thanks. Regards, Vincent. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Matrix Quat precision
Hi Robert, For the moment I use a relative calculation (I compute on each frame depending on the last frame's position). I suppose that computing in absolute mode could be better... but I would be more difficult to implement. Thanks Robert, I'll will try this if I have some time... One more question : If there any conversion between Quat and Matrix, or Quat and Vector, or Vector and Matrix which can not be precise ? I mean some conversion function to avoid for example in the goal of beeing always precise ? Regards, Vincent 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincet, Even with doubles there are still precisions issue, just much smaller than with floats. If you are using a iterative calculation then if you aren't care the errors could accumulate until their become visable. The best thing to do is to avoid an iterative calculation and calculate the new camera position based on the position of the node in the scene on each frame. Robert. On Thu, Mar 27, 2008 at 8:38 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi all, I'm currently making a new Manipulator inherited from MatrixManipulator. The specificity is : On the call of a function, the manipulator (camera) have to follow a node which is moving. I mean that the camera have to be exactly immobile relative to the moving node. For that i've used Matrix, Vec3 and Quat. All looks great, but after 10 minutes of running, I can see that a little moving difference between the node and the camera... if the node rotate about 2 or 3 axis, the difference is more important... Using Vec3d, Quat and Matrixd, is there any reason for this lack of precision ? Or is it my method which is not good... ? Thanks. Regards, Vincent. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Matrix Quat precision
Hi Vincent, You approach is iterative so the errors will accumulate, so it's based to avoid this and go fetch the world matrices from the node your tracking each frame. The osgGA::NodeTrackerManipulator does this. Robert. On Thu, Mar 27, 2008 at 8:57 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi Robert, For the moment I use a relative calculation (I compute on each frame depending on the last frame's position). I suppose that computing in absolute mode could be better... but I would be more difficult to implement. Thanks Robert, I'll will try this if I have some time... One more question : If there any conversion between Quat and Matrix, or Quat and Vector, or Vector and Matrix which can not be precise ? I mean some conversion function to avoid for example in the goal of beeing always precise ? Regards, Vincent 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincet, Even with doubles there are still precisions issue, just much smaller than with floats. If you are using a iterative calculation then if you aren't care the errors could accumulate until their become visable. The best thing to do is to avoid an iterative calculation and calculate the new camera position based on the position of the node in the scene on each frame. Robert. On Thu, Mar 27, 2008 at 8:38 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi all, I'm currently making a new Manipulator inherited from MatrixManipulator. The specificity is : On the call of a function, the manipulator (camera) have to follow a node which is moving. I mean that the camera have to be exactly immobile relative to the moving node. For that i've used Matrix, Vec3 and Quat. All looks great, but after 10 minutes of running, I can see that a little moving difference between the node and the camera... if the node rotate about 2 or 3 axis, the difference is more important... Using Vec3d, Quat and Matrixd, is there any reason for this lack of precision ? Or is it my method which is not good... ? Thanks. Regards, Vincent. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Matrix Quat precision
Yes I know that I currently accumulate error, but when I did it I was thinking about a zero-error so it wasn't a problem... The NodeTrackerManipulator does it, yes, but I'm working on a manipulator still existing in the project, and make a whole new manipulator looks a lot of work and I don't have enough time to do it... Maybe in the future... Thanks for the help :-) Regards, Vincent. 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincent, You approach is iterative so the errors will accumulate, so it's based to avoid this and go fetch the world matrices from the node your tracking each frame. The osgGA::NodeTrackerManipulator does this. Robert. On Thu, Mar 27, 2008 at 8:57 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi Robert, For the moment I use a relative calculation (I compute on each frame depending on the last frame's position). I suppose that computing in absolute mode could be better... but I would be more difficult to implement. Thanks Robert, I'll will try this if I have some time... One more question : If there any conversion between Quat and Matrix, or Quat and Vector, or Vector and Matrix which can not be precise ? I mean some conversion function to avoid for example in the goal of beeing always precise ? Regards, Vincent 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincet, Even with doubles there are still precisions issue, just much smaller than with floats. If you are using a iterative calculation then if you aren't care the errors could accumulate until their become visable. The best thing to do is to avoid an iterative calculation and calculate the new camera position based on the position of the node in the scene on each frame. Robert. On Thu, Mar 27, 2008 at 8:38 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi all, I'm currently making a new Manipulator inherited from MatrixManipulator. The specificity is : On the call of a function, the manipulator (camera) have to follow a node which is moving. I mean that the camera have to be exactly immobile relative to the moving node. For that i've used Matrix, Vec3 and Quat. All looks great, but after 10 minutes of running, I can see that a little moving difference between the node and the camera... if the node rotate about 2 or 3 axis, the difference is more important... Using Vec3d, Quat and Matrixd, is there any reason for this lack of precision ? Or is it my method which is not good... ? Thanks. Regards, Vincent. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Matrix Quat precision
Hi Vincent, Tweaking how you compute the position shouldn't take more than a couple of hours, and really doesn't require one to write an new manipulator. Robert. On Thu, Mar 27, 2008 at 9:30 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Yes I know that I currently accumulate error, but when I did it I was thinking about a zero-error so it wasn't a problem... The NodeTrackerManipulator does it, yes, but I'm working on a manipulator still existing in the project, and make a whole new manipulator looks a lot of work and I don't have enough time to do it... Maybe in the future... Thanks for the help :-) Regards, Vincent. 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincent, You approach is iterative so the errors will accumulate, so it's based to avoid this and go fetch the world matrices from the node your tracking each frame. The osgGA::NodeTrackerManipulator does this. Robert. On Thu, Mar 27, 2008 at 8:57 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi Robert, For the moment I use a relative calculation (I compute on each frame depending on the last frame's position). I suppose that computing in absolute mode could be better... but I would be more difficult to implement. Thanks Robert, I'll will try this if I have some time... One more question : If there any conversion between Quat and Matrix, or Quat and Vector, or Vector and Matrix which can not be precise ? I mean some conversion function to avoid for example in the goal of beeing always precise ? Regards, Vincent 2008/3/27, Robert Osfield [EMAIL PROTECTED]: Hi Vincet, Even with doubles there are still precisions issue, just much smaller than with floats. If you are using a iterative calculation then if you aren't care the errors could accumulate until their become visable. The best thing to do is to avoid an iterative calculation and calculate the new camera position based on the position of the node in the scene on each frame. Robert. On Thu, Mar 27, 2008 at 8:38 AM, Vincent Bourdier [EMAIL PROTECTED] wrote: Hi all, I'm currently making a new Manipulator inherited from MatrixManipulator. The specificity is : On the call of a function, the manipulator (camera) have to follow a node which is moving. I mean that the camera have to be exactly immobile relative to the moving node. For that i've used Matrix, Vec3 and Quat. All looks great, but after 10 minutes of running, I can see that a little moving difference between the node and the camera... if the node rotate about 2 or 3 axis, the difference is more important... Using Vec3d, Quat and Matrixd, is there any reason for this lack of precision ? Or is it my method which is not good... ? Thanks. Regards, Vincent. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org