Hi,

I'm currently working on a flip fluid viscous project but there seems to be an error with the meshing. As you can see in the attached images, there are some strange “blocks” or “bars” occurring beneath the model. The first block appears on frame 127 and then randomly changes size, position and quantity. I've no idea where this is coming from, in my terms it all seems normal.

The compressed Cache seems to work normally though, so this is Happening on the way to the Surface Cache.

Thanks

I don't have Houdini in front of me but definitely play around with the settings value in Particle Fluid Surface. I guess the generated VDB volume that influence the meshing pickup something that it is not supposed to.

Try Spherical if you're using Average Position and also the Erosion/Dilate.

Hey thanks,

the erosion scale was the culprit indeed. I've already fiddled around with this setting, but always between 0 and 1 and there was no difference. While heeding your hint i set to 3 and it worked. The blocks are gone.

But I've come across another problem. I try to render it out now but struggling with unbelievable high render times.
The first image i attached looks okay but it took me 15 minutes on my 16 cores machine, and that with only 3x3 pixel samples. The particle separation is on 0.02.


The other thing I noticed is, that i don't use the surface cache correctly, i think. When the blue flag is on the particlefluidsurface it looks like the first image, but with the blue flag on the surface_cache it looks like crap. I canceled the second image on minute 22; as you can see, it looks pretty bad.

Rendering a relatively noise free reflective/refractive material with Mantra are indeed time consuming.

You might want to reduce the reflection and refraction limit to 3 or 4 if you use the default 10 which are overkill personally unless photorealism accuracy is your goal.

If you're rendering as image sequence, upping the noise threshold to around 0.03 for 720p output or 0.05 for 1080p should still be acceptable (provided your min/max samples is 1/9). Leave the pixel samples at 3x3 unless you need to render heavy motion blur/DOF that benefit from higher pixel samples.

I'm unsure what is causing the issue when you set the display flag to the surface_cache node but do remember that you can override the render flag (the purple colour) on any nodes which Mantra will use for the actual render so you can have a separate display flag (like a dummy geo for faster viewport display etc).

Ctrl+LMB on where you usually set the display flag to set it as render flag. There's a reason why there is the occasional RENDER null that the Shelf tool creates!

To really answer your issue… I got no idea at all. Maybe try bake out the problematic frame and read it in to compare the mesh with the one generated by the Particle Fluid Surface.

ok thank you,

i was able to reduce rendertime from 15 to 5 mins per Frame with your tipps. Set refract and reflect Limits to 5 and noise to 0.05; not to much visual loss, still looking good.

Didn't solve the flag issue though, but i will ignore it for now and come back to it, if it gets a real Problem.