Hi,
I'm getting some very weird behaviour with some of my particles that collide with my VDB.
They should come to rest but a small percentage don't ever stop moving.
The collider is a VDB created using the Collision Source node and represents the original RBD pieces converted to a volume.
I've tried using Auto Sleep and various other methods but these same particles just never stop. In one of the videos the purple colour is the active particles and the red are the sleeping particles.
I've upped the substeps in the pop solver to 20 and they still move.
Any advice would be much appreciated.
POP Particles keep moving when they should stop
2434 9 0
-
- stevegreentechy
- Member
- 10 posts
- Joined: Feb. 2019
- Offline
-
- sekow
- Member
- 238 posts
- Joined: Nov. 2013
- Offline
-
- stevegreentechy
- Member
- 10 posts
- Joined: Feb. 2019
- Offline
sekow
hard to tell without seeing an hipfile. but I reckon something is up with your collision source.
Thanks. I had made a stupid error with the collider. I was deleting the packed fragments which were outside the camera frustum.
I did this in an attempt to reduce the size of the VDB generated by the collision source.
This had the side effect of changing the geometry point count which was throwing the velocity point calculations out in the Collision Source node.
When I sorted this it stopped the particles drifting but I was then getting some crazy bigger particles vibrating up and down at rest. See Vibrating.mp4.
I reduced the max pscale of the source particles from 0.05 to 0.03 and the bigger ones stopped vibrating. I can only assume the debris source points are so dense that they cause overlapping particles in the POP sim and they get stuck inside one another. I did try remove overlapping etc but that didn't cure the issue so for now I'll live with slightly smaller overall pscale as the sim is now stable.
Edited by stevegreentechy - Nov. 12, 2021 06:10:00
-
- sekow
- Member
- 238 posts
- Joined: Nov. 2013
- Offline
-
- stevegreentechy
- Member
- 10 posts
- Joined: Feb. 2019
- Offline
sekow
any reason why you are using vdbs for collision and not straight up geometry?
I was told POP sims need volumes for collisions.
This was from a Jeff Wagner live stream.
In the first 20 mins Jeff covers POP colliders and the volume created internally by the shelf tools.
https://vimeo.com/252645795?embedded=true&source=video_title&owner=1723479 [vimeo.com]
Also by doing it with a VDB Proxy volume means I can cache it to disk before the DOP network and not rely on the solver creating collision geometry for complex RBD data every frame.
Edited by stevegreentechy - Nov. 12, 2021 07:58:12
-
- sekow
- Member
- 238 posts
- Joined: Nov. 2013
- Offline
To be honest I never used volume collisions with pure pops. Flip and smoke something different.
Even with some very heavy destruction.
I would also argue that you would have the disadvantage to be stuck on the timescale you cached those volumes out. As those are not very handy to interpolate for substeps.
Back to your issues. Usually pop particles are pretty "dumb" and do not have a collision relationship between individual points. SO it does not matter when they overlap due their pscale.
What is always helping is a bit of drag via pop drag or pop wind.
Like I said without having a look at your setup its hard to say whats up
Even with some very heavy destruction.
I would also argue that you would have the disadvantage to be stuck on the timescale you cached those volumes out. As those are not very handy to interpolate for substeps.
Back to your issues. Usually pop particles are pretty "dumb" and do not have a collision relationship between individual points. SO it does not matter when they overlap due their pscale.
What is always helping is a bit of drag via pop drag or pop wind.
Like I said without having a look at your setup its hard to say whats up
-
- sekow
- Member
- 238 posts
- Joined: Nov. 2013
- Offline
-
- stevegreentechy
- Member
- 10 posts
- Joined: Feb. 2019
- Offline
sekow
To be honest I never used volume collisions with pure pops. Flip and smoke something different.
Even with some very heavy destruction.
I would also argue that you would have the disadvantage to be stuck on the timescale you cached those volumes out. As those are not very handy to interpolate for substeps.
Back to your issues. Usually pop particles are pretty "dumb" and do not have a collision relationship between individual points. SO it does not matter when they overlap due their pscale.
What is always helping is a bit of drag via pop drag or pop wind.
Like I said without having a look at your setup its hard to say whats up
I only base it on the fact that the shelf tool for deforming object creates a node network with a collision source node and a VDB file cache.
My collider is set to volume sample in the DOP network.
Try putting the pscale attribute randomise before the pop network back to 0.05 and you'll see the crazy vibrating points. When it's set smaller like 0.03 the crazy points disappear.
Edited by stevegreentechy - Nov. 12, 2021 11:09:37
-
- sekow
- Member
- 238 posts
- Joined: Nov. 2013
- Offline
I had a look, and yeah I can reproduce that jumping behavior with the increased pscale. No clue why though.
I've switched over to surface based collision and changed the collision sourcing by skipping the collision source sop, see attached screenshot and file.
Mind thats a possible way to fix your issues, and by no means the only right way. But it might help
I've switched over to surface based collision and changed the collision sourcing by skipping the collision source sop, see attached screenshot and file.
Mind thats a possible way to fix your issues, and by no means the only right way. But it might help
-
- stevegreentechy
- Member
- 10 posts
- Joined: Feb. 2019
- Offline
sekow
I had a look, and yeah I can reproduce that jumping behavior with the increased pscale. No clue why though.
I've switched over to surface based collision and changed the collision sourcing by skipping the collision source sop, see attached screenshot and file.
Mind thats a possible way to fix your issues, and by no means the only right way. But it might help
Brilliant. I appreciate you taking your personal time to look at my issue.
Are there any downsides to using surface based colliders?
Do points ever pass through the surface or are they pretty robust?
I can see advantages in them easily supporting sub-steps with the retime node to interpolate.
Edited by stevegreentechy - Nov. 12, 2021 18:34:52
-
- Quick Links