Dropping a glass in water

[syberand]

I'm dropping a wineglass in water and it is giving me some issues when i turn up the resolution.

It is all in slowmotion 0.03 on the timescale, so only using one step per frame, and when running the sim at 2mm size with a 3mm push on the glass, everything seems fine and working. When removing the push (the glass is like 1.95mm at it thinnest) and turning up my res to 0.8mm (50m voxels), it goes wrong. Air seems to push through the glass, causing a ton of unwanted bubbles and the geocutter in the isosurface seem to be struggling too.

[Ivaylo Katev]

by default the geometry mode is inscribed, that "eats" one cell in both the sides, the geometry must be at least 3 cells wide to be handled properly. your option is to switch to circumscribed or center mode, that can handle thinner geometry, but can have rendering artifacts. or, you can try with a nightly build, the geometry representation of the new liquid solver can handle objects thinner than single cell and there is no requirement for minimal width. if you decide to try the nightly build, switch the advection method to flip, otherwise you will use the same core as now

[syberand]

https://www.dropbox.com/s/mbszbfbl9h...HX_04.mov?dl=1 - 2 or 3mm voxelspacing with a 3mm push on the glass
https://www.dropbox.com/s/4ajsad8oc7...HX_05.mov?dl=1 - 2mm thick glass with 0.8mm voxelspacing
https://www.dropbox.com/s/2uor6vtpos...HX_07.mov?dl=1 - 2mm glass, 0,8mm voxel spacing and 2 steps per frame. I have no idea what is happening here...

We are using the nightly builds. 1511096441

[syberand]

by default the geometry mode is inscribed, that "eats" one cell in both the sides, the geometry must be at least 3 cells wide to be handled properly. your option is to switch to circumscribed or center mode, that can handle thinner geometry, but can have rendering artifacts. or, you can try with a nightly build, the geometry representation of the new liquid solver can handle objects thinner than single cell and there is no requirement for minimal width. if you decide to try the nightly build, switch the advection method to flip, otherwise you will use the same core as now

I am currently using the nightly build, but cant see where I can change the solver?

[Ivaylo Katev]

i suppose if you are with nightly build and there is no advection in the UI, you are using the flip solver already, so forget my suggestion.
not sure what is happening in the second video, it may be kind of temporary instability, however, to get similar result as video 1, you have to use spf about 2 or 3, because the cell size is decreased 2.5 times.
i think there is some kind of instability in the last video too, but there are a lot of foam bubbles, better disable the foam until the liquid is adjusted, then you can adjust the foam using resimulation (don't forget to export the grid velocity).
just few words about the foam rendering - use deeper bouncing for the rendering, it seems to be 0 or 1 now, use 2 or 3. there is some kind of bubble size flickering, are you using particle size variation in the foam shader setup?
i think you have to try to simulate with tomorrows build, because we had to fix some issues in the geometry-liquid interaction, and it's possible to improve your results

[syberand]

I hope you bear with me, I'm just trying to understand what to turn on, in which situations.

So, when using the nightly builds, I no longer have to have 3 voxels between eg. each side of the glass? How far can you push this?

Also, what is the connection between spf and voxelsize? This animation is run in 0,03 timescale with 1spf. Then when reducing voxelsize, you say we should also increase spf. How is this connected?

[syberand]

https://www.dropbox.com/s/loyyoe38z8...ly_01.jpg?dl=1 - glassthickness is now 2mm, gridsize 1,4mm and spf is 3. Can you spot why we get a jagged mesh inside the glass? I'm using the newest nightly

[syberand]

https://www.dropbox.com/s/ctcibjq3rw...HX_09.mov?dl=1 - same settings as before, just the animation. The only issue now appears to be the "jagged" water against the glass.

[syberand]

Okay, so update time.

The part where the air/water passes through the glass is fixed, which was done by upping the spf.
What is left is a new problem, where we can see gridsized slices through the glass, as you can see in the attached picture.

We have improved on this, but it is not "clean" enough yet. What is the correct way to handle details like this? We are using implicit surface. What we last changed was higher res, more steps and enabled "clear inside". Basically I threw the book at it.
https://www.dropbox.com/s/doip6ksuxv...d_PHX.mov?dl=1 - mesh surface here. Better but still some lines. Suggestions?

[Ivaylo Katev]

yes, i forgot to mention a very important information: the penetration through the walls may be caused not by "holes" like in old simulator, but it may be related to the movement. the grid core should handle moving objects with no porblems, however i think we may have an issue in the situations when the object is moving fast enough, have to investigate this.
about the grid, it's very probably that it's related to the thin glass walls. the flip technique uses grid to resolve the pressure management, and when the walls are too thin they are considered one cell wide, usually this is not a problem because the liquid particles are allowed to reach the real glass geometry, but in case of moving objects who knows... did you update the build? there are some improvements in the geometry-liquid interaction added two days ago.

[syberand]

Currently on the two day old build (1511176469)
https://www.dropbox.com/s/2954ka96of...HX_13.mov?dl=1
It helped with the water getting through by turned up the spf. The jagged edges at the glass was resolution and maybe higher spf? Turned everything up and went to bed. So honestly not sure which helped there.

Now the final issue appears to be the "lines" you can see in the water as it gets sucked down. If I was using mesh, I would just start smoothing, but then I would loose the close and clean contact with the glass, which is more important, so currently on implicit. Is there any way to smooth that?

[syberand]

I have marked the lines I'm talking about. Any idea of what to do?

edit: no idea why there are two pictures.

[Ivaylo Katev]

the mesher now has ability for particle based smoothing, with it you can make the mesh thinner or thicker using the particle radius.
not sure what you mean with "lines", but i see grid artifacts on the surface. my suggestion is to try the mesh, make the experiments with some backup frame, it will contain the particles, and of the result is satisfying you can simulate with particle export enabled.

[syberand]

I think in this case, grid artifacts = lines. I have tried with mesh and smoothing and it looks better in the area i have marked. But it looks worse at the glass, where I need the geometry-cutter to have a clean contact-area. Or do you have another solution?

[Ivaylo Katev]

i think you have to check your isosurface level, i didn't observe any grid artifacts on the free liquid surface ever, all the cases with grid artifacts on the liquid were caused by changed isosurface level.