I have a smoke sim that emits from a characters upper body and flows outward and along the surface of the character, but I'm getting really gnarly collision banding. You can really see it in the alpha of the smoke render. I have been unable to clean up the banding artifacts in the sim, and I'm currently resorting to a post process.

The colliders are airtight polys with a collision sdf. The collision voxel size matches the size of the pyro sim.

The source is points that are peaked outward from the collision surface so they don't sit directly on the character surface.

Things I have tried to fix the artifcats.

• Tried both static object collider (with deforming on) and volume source to source the collsion field and collsion velocity.
  
• Tested increased max substeps between 1-3.
  
• Increased collision field resolution and subdivided the collsion poly surface.
  
• Tested with "fill object" on the collsion field and/or increased bandwidth.
  
• Applied vdb_smooth_sdf to the collsion field.
  

I'm out of ideas at the moment. Willing to try anything.

The collision is going to do pretty much exactly what you've shown. Your SDF collider is not smooth, it
has ridges.

Put down a VDB smooth SDF, this will smooth it out.

lewis_T
Put down a VDB smooth SDF, this will smooth it out.

I already tested with a vdb_smooth_sdf node as well as subdividing my geo prior to the collision_source_sop as noted above. But just to be doubley sure, I tried again and I'm getting the same results. I've included screenshots of my smooth sdf and the resulting smoke sim.

check the collision when is solving you might have a pretty low res sdf in there.
How are you feeding this into the solver? Can you post the .hip

sepu
check the collision when is solving you might have a pretty low res sdf in there.
How are you feeding this into the solver? Can you post the .hip

I'll see about sanitizing the hip to post, but that'll be a bit of time. I have checked the collision field in the solver by turning on the visualizer for the pyro sparse object and it looks as expected, which is to say it looks like my collision sdf field at the sop leve.

I've tried sourcing the collider both with a static object and with a volume source dop (using collision and collisionvel fields).

I'll work on posting something maybe tomorrow-ish.

That looks correct - still tough to say without looking at your .hip
You might want to look at this (https://www.sidefx.com/forum/topic/103550/); you might be running into a similar situation.
I don't need the whole character, just where you are having issues or some repro.

sepu
That looks correct - still tough to say without looking at your .hip
You might want to look at this (https://www.sidefx.com/forum/topic/103550/); you might be running into a similar situation.
I don't need the whole character, just where you are having issues or some repro.

Thanks for taking a look. Here's a sanitized hip. I have freeze-framed and stashed the geo.

So either disabling Correct Collisions or removing density from the Field to Correct parm "fixes" the issue. I know this is not what I should do. But based on this it seems that collision field isn't doing a great job reseting the density field and creating these artifacts.

Edit: Interestingly if you use a Velocity Voxel Size multiplier, it will also affect the collision field.

In case anyone runs into this issue in the future, I'd like to help them out. So I reached out the jsmack to try and get some resolution (pun intended) to this issue, and he said the following:

I think might be on to something with the accurate collisions toggle. sometimes setting the density to 0 inside the collision field is too harsh and a custom solution is needed. I haven't done any pyro like this for VFX in quite some time, but I remember using custom smoother volume fields to mask the density field that overlaps the collider with more falloff and buffer than the actual collider. this is needed especially if the roto animation doesn't match the plate super well.

I hope that helps someone else out.