Particles all drift the same way

[ChrisCousins]

Hi - hope there's a simple answer here. I've got particles on a plane, which I'm planning to add forces, friction attribs to so the particles all tend to drift in certain ways. What I've noticed though is that even with a totally flat grid, the particles initially bounce straight up but once the vertical movement stops they all drift in the same direction. See silly GIF attached. I'm guessing a rounding error, substeps don't help but is there any way to avoid this? I thought using a planar patch not grid would fix it, but they just move in different ways but still the same direction.

Plane is not rotated, no forces other than gravity... help! HIP attached.

Cheers - Chris

sillyParticles.zip

[dleonhardt]

Hey Chris,

In my experience, particles can behave like in your example when you provide an object with zero thickness inside your Static Object, so using Extrude Volume should fix that. But if your collision plane ends up being completly flat, I'd recommend using the Ground Plane DOP as your collider instead.

Edited May 13, 2020 by dleonhardt

[ChrisCousins]

Thanks Dennis - nice idea but no dice, they all still dutifully roll the same way. It needs to be geo they're bouncing off as I'm planning on adding variations like friction, v etc to the plane, to get the particles bouncing in certain ways (think sand on a chladni plate).

Appreciate the suggestion however! Cheers - Chris

[dleonhardt]

So I took a quick look at your scene and it seems that giving thickness to your plane and switching the collision detection to Volume Collisions fixes the issue. But I'll assume that you had a reason for using Surface Collsions in the first place.
 

Quick test with a 10x10 division grid

Surface collision (with a very familiar pattern)

Volume collision 

[ChrisCousins]

Thanks for looking into this. Yup the setup I had in mind needed a dense grid, the idea was that the collider has attributes like bounce, friction that can change over time, allowing the sand to settle in patterns, and shift as the attributes shift - but they always just run off the edge in one direction unless I turn up friction so high they just get stuck.

Hmmm... I guess there must be other ways of getting the motion I'm looking for, maybe using point clouds and getting the data via nearpoint() - will take a look.

Strange that POP is affected by topology in this way, but I guess the optimisations have to come with certain limitations.

Thanks again - Chris

[anim]

6 minutes ago, ChrisCousins said:

Strange that POP is affected by topology in this way, but I guess the optimisations have to come with certain limitations.

it's clearly not expected behavior, so rather than assuming it has to be this way submit a BUG, seems like you have a pretty nice reproduced case which clearly shows it, especially the one with 10x10 div grid

worst case you'll get a reply that it will not be fixed if there is actual reason for this behavior, best case it will be fixed in no time

 

[ChrisCousins]

13 minutes ago, anim said:

worst case you'll get a reply that it will not be fixed if there is actual reason for this behavior, best case it will be fixed in no time

On it...

[ChrisCousins]

Update - heard back from sidefx, helpful and speedy as ever. I'll quote it rather than mangling it with my own interpretation ;-)
From our developers:
The particles do per-triangle collision, so have the chance of colliding at the edge of a triangle rather than a face, thus imparting horziontal motion. This is because they are colliding with a soup of polygons, not with a mesh.
The other strong artifact in this situation is that particles get a default radius of 0.02 if no pscale is set. This is why if you scale this system up 10x (by transforming source, collision, and gravity) it will seem to work. Alternatively, you can set pscale to 0.002 on the particles to get less silly behaviour.