18 replies to this topic

[hopbin9]

Another thing to try. Connect the last locked SOP to the input of a Box SOP. That will create a box that surrounds the bounding volume of it's input. Then increase it's polygon divisions. Let us know if that geometry also jitters.

[tomwyn]

Macha, on 14 February 2012 - 02:43 PM, said:

Well, I have a feeling you run into precision errors. There are ways around that, none of them easy. One would be to temporarily move your stuff to origin, and then back once you're done. On huge scene scales or differences in scales this does happen all over the place, and the further you get from the origin the more you loose precision. I think it has to do with the decimal point shifting position but I'm not really sure. Or some strange mantissa thing -who knows how these things really work other than people who peruse IEEE. Also, try H12 if you can and see if it is better.

It's just a suspicion of course. I'm not sure if it is the real problem in your case.

When you first mentioned distance from origin, it kind of clicked that precision might drop the further away the geometry gets. I've been finding that I'm using smaller & smaller values at the far end of the animation. (For example, the difference between position 0.00 and 0.01 along the curve is much smaller than the difference between 0.99 and 1.00.)

Hopefully one or some of the above suggestions will reveal some kind of solution!

As I mentioned, if I can I'll upload a similar .hip to the one I'm using. I understand it might by a little frustrating try to decipher my descriptions!

Thanks again.

[tomwyn]

hopbin9, on 14 February 2012 - 03:11 PM, said:

Another thing to try. Connect the last locked SOP to the input of a Box SOP. That will create a box that surrounds the bounding volume of it's input. Then increase it's polygon divisions. Let us know if that geometry also jitters.

Will do!

Thanks for the suggestions. I'll work my way through the scene in the morning & update when I can. I'm sure there's a solution In there somewhere!

tom.

[tomwyn]

Hi all,

I've spent some hours working through all your suggestions, but unfortunately I have not found a solution to the problem. Here's a brief overview of what I've tried (after all of these, the jittering remained):

'Locking' final SOPs within assets to bake geometry.
Attaching box to locked geo to create bounding box.
Scaled down scene so that object does not travel so far away from origin.
Checked & iterated 'transform object' within my merge SOPs.
Non-integer frames seem fine.


The only slight improvement i've managed is going in to my assets at geometry level & increasing all the scales x10, which meant not needing to scale down the overall asset in my new scene as much, but the jittering is still there & ultimately, I am still using the same scale ration between my asset & the obstacles, so I end up with the same problem.

I have a serious hunch that the scale difference between my curve/asset is the issue. But it could also be the parameterization along the curve (reaches many decimal places).

Anywho, I've constructed a super basic setup which displays the same problems as my scene, so if any of you guys could take a peek to either confirm my suspicions about the scale ratios (or expose me as an utter fool for missing something blindingly obvious) that would be amazing.

in the scene, the asset is a sub group (rather that just sharing an otl) & contains a simple parent/child relationship, some geo with promoted paramaters to control from the assets at scene level, and some baked geometry.

The scales in the scene are as large as I would like to take them, so feel free to scale down anything if needed. However, the scale ratio between the asset & the obstacle need to remain the same - or as close as possible.

Thanks again for all the helpful suggestions.

Attached Files

•  jittering geometry example.hip   173.53K   21 downloads

[hopbin9]

Macha was right. It is a precision problem.

I fixed it by giving everything a common parent and scaling the scene down by 0.001

Seems to have removed the jittering.

Attached Files

•  jittering geometry example fixed.hip   164.77K   19 downloads

[tomwyn]

hopbin9, on 15 February 2012 - 07:17 AM, said:

Macha was right. It is a precision problem.

I fixed it by giving everything a common parent and scaling the scene down by 0.001

Seems to have removed the jittering.

Hey, yes this does indeed solve the jittering issue in this scene! Thanks a bunch!

I didn't think to use the uniform scale slider as I haven't used it before... I had a feeling I would have overlooked something fairly obvious, sorry about that.

How would you suggest I apply this technique to assets which contain locked geometry? (i.e. the rescale input breaks the locked nodes) Or should I just save my .otl's again but make them editable? I'd rather not do this as I'd have to pretty much reconstruct my scene, but if it's the only viable option then I can give it a go.

Here's another example file which includes the .otl of the subnet from the previous example so you can see what I mean by the nodes 'breaking' when I try to rescale them. If you could have just one more quick look I'd really appreciate it! (If you need to look, that is.)

Thanks again, really appreciate the help.

Attached Files

•  asset example 002.zip   22.83K   21 downloads

[tomwyn]

hopbin9, on 15 February 2012 - 07:17 AM, said:

Macha was right. It is a precision problem.

I fixed it by giving everything a common parent and scaling the scene down by 0.001

Seems to have removed the jittering.

Hey,

Thanks for all the help, but having spent a few more hours on this problem, i have unfortunately still been unable to fully solve it.

Although the file you previously posted was indeed free from the jittering geometry, I think that was because the subnet was actually only travelling a very small distance & it's scale in comparison to the path curve was very large.

After experimenting with uniform scale, I've managed to reach a point where the geometry/path curve/sphere obstacle are scaled to the correct proportions, but the jittering remains.

I've attached the .hip here for anyone who wight be interested to have a look.

I fear I may be missing something obvious again, but I get the feeling that something just isn't right.

Thanks!!

Attached Files

•  jittering geometry example with correct scales.hip   211.25K   21 downloads