Vellum jitters non-stop without loss of energy!

[miccetro]

Hello, I'm just getting into vellum and the solver seems fast and cool but I'm trying a sim here where the geometry tries to attach to its collision geo, but that ends up being very jittery at the end with what seems to be no loss of energy. I've uploaded a scene file. I've tried to set @v to zero at the end and that still hasn't worked. What're the steps to making this solver behave? 

vellum_basics__jitter_02.hip

[Atom]

Locate the vellum solver and set the Constraint Iterations to something higher than the default of 100. Try 400-1000.

[miccetro]

thanks, @Atom, unfortunately though, that doesn't resolve the issue. It's jittering out of control.

[kleer001]

You're going to have a very very hard time with this technical exercise of wrapping a rotating cloth around and around a small tube. This is a non trivial problem. You may need to crank up certain values like the density of points on the geometry or crank up the overall substeps to such a point that it either blows up or just barely works. In other words Vellum, IMHO, isn't designed for wrapping and sticking cloth. Instead I think it's designed more for draping cloth style with one or two wrinkles, not eight layers.

But I'd be grateful if you proved me wrong and were able to find the right settings. To that end you'll want to change your polygon group selections to a more procedural nature and use a bounding box instead of an explicit manual selection of certain numbered polys. That way you can change your geometry  more easily.

Is this for a specific effect in a shot that you're working on? Is it a personal challenge? Is it for a class?

There might be another way to solve this as a visual problem instead of a naive simulation. I would suggest a more manual animation style or blend shape style solution if you need a ring of cloth wrapped around a pole.

[Andrea]

Ok, a bit off topic I should have tried to do it with vellum but I had this idea. What about using chop as post-process for filtering the noise/jittering?

 

Usually the jittering starts or at least is visible when the object is almost static. In this case you can use the length of velocity to blend the cloth/vellum object between the filtered/non-jittering simulation and the non filtered object. This way you get the best from the two. This trick saved me a lot of times (if you consider three a lot) and the supervisor was happy about the "clean sim"  

 

vellum_basics__jitter_02_chop.hip

 

Edited May 15, 2019 by Andrea

[miccetro]

15 hours ago, kleer001 said:

You're going to have a very very hard time with this technical exercise of wrapping a rotating cloth around and around a small tube. This is a non trivial problem. You may need to crank up certain values like the density of points on the geometry or crank up the overall substeps to such a point that it either blows up or just barely works. In other words Vellum, IMHO, isn't designed for wrapping and sticking cloth. Instead I think it's designed more for draping cloth style with one or two wrinkles, not eight layers.

But I'd be grateful if you proved me wrong and were able to find the right settings. To that end you'll want to change your polygon group selections to a more procedural nature and use a bounding box instead of an explicit manual selection of certain numbered polys. That way you can change your geometry  more easily.

Is this for a specific effect in a shot that you're working on? Is it a personal challenge? Is it for a class?

There might be another way to solve this as a visual problem instead of a naive simulation. I would suggest a more manual animation style or blend shape style solution if you need a ring of cloth wrapped around a pole.

@kleer001 Thanks for the insight, I was attempting to solve a problem for an upcoming project that might involve a similar effect. But I will definitely look into solving this using a different method. But it's nice to know the limitations of what is achievable with PBD.

[miccetro]

5 hours ago, Andrea said:

Ok, a bit off topic I should have tried to do it with vellum but I had this idea. What about using chop as post-process for filtering the noise/jittering?

 

Usually the jittering starts or at least is visible when the object is almost static. In this case you can use the length of velocity to blend the cloth/vellum object between the filtered/non-jittering simulation and the non filtered object. This way you get the best from the two. This trick saved me a lot of times (if you consider three a lot) and the supervisor was happy about the "clean sim"  

 

vellum_basics__jitter_02_chop.hip

 

the CHOPS post-processing is not bad at all, it seems like it can produce acceptable results to me. Thanks for the tip, was trying something similar using a SOP solver.

[Andrea]

I am sure is possible also with the SOP solver to filter the data using several timeshifts but with CHOPS is a lot faster to setup. For this kind of things often it can save a lot of time being just one node and one line of code

Edited May 16, 2019 by Andrea

[Tex]

I had a look at your scene - lowering the cloth constraint thickness to 0.001 and raising some of the dampening settings got most of the jittering out.

[cudarsjanis]

On 5/15/2019 at 5:30 AM, Andrea said:

Ok, a bit off topic I should have tried to do it with vellum but I had this idea. What about using chop as post-process for filtering the noise/jittering?

 

Usually the jittering starts or at least is visible when the object is almost static. In this case you can use the length of velocity to blend the cloth/vellum object between the filtered/non-jittering simulation and the non filtered object. This way you get the best from the two. This trick saved me a lot of times (if you consider three a lot) and the supervisor was happy about the "clean sim"  

 

vellum_basics__jitter_02_chop.hip

 

Thanks for the hip! Works like a charm.

[dyei nightmare]

On 19/5/2019 at 7:08 PM, Tex said:

I had a look at your scene - lowering the cloth constraint thickness to 0.001 and raising some of the dampening settings got most of the jittering out.

dude! you saved me hours of this problems! yess this is

[Benyee]

On 2019/5/15 at 10:48 PM, miccetro said:

the CHOPS post-processing is not bad at all, it seems like it can produce acceptable results to me. Thanks for the tip, was trying something similar using a SOP solver.

have you figured it out with some other solution? thank you

Edited July 2, 2022 by Benyee