Chocolate on cookie ripples artifacts.

[Ivaylo Katev]

ok,
the floor is nothing special, just add it
about the scale - the scene have non realistic scale. we always say to the users "use realistic scale", and now we violate our own rule
why? the simulation of small scaled and big scaled scenes is not equal by performance. the small scaled scenes simulate slower, because the higher dynamics that they usually have. this scene have no high dynamics, but nevertheless it will simulate very slow if it is set in realistic scale, because it will require higher viscosity to achieve the same look. in fact, small scale is equal to higher gravity (measured in simulator units) and in real scale you will have higher gravity and higher viscosity to oppose it. this is a waste of performance, so the better solution is just to make unrealistic scale (or use gravity multiplier) in order to save calculations. now, how to recognize the cases when we can simulate in non realistic scale?
the rule is - you shouldn't have free falling liquid or good visible waves. the free falling liquid and the waves betray the real scale and the observer feels that something is wrong.
and now about your increase of the viscosity and non newtonian parameter.
actually for thicker chocolate you need to decrease the non newtonian parameter. the viscosity must be increased but not as high. value 1 of the viscosity means infinite viscosity. it would take infinite time to calculate, so we have a limitation. viscosity 0.9 is about three times higher than viscosity 0.7. why? the first one needs 0.1 to reach 1, the second needs 0.03 to reach 1. 0.1 is about three times bigger than 0.03
what i observe in your simulation - despite the three times higher viscosity (that takes three times more time to calculate), you actually have lack of viscosity, you lose the coiling , that is present in the original sample. this is because the higher non newtonian cancels the viscosity efforts.

[jstrob]

Thanks for those tips!

By increasing the meshsmooth level by one iteration on the colliding cookie I could finally get rid of the big ripples. But I can still see those grid artifacts that were also in the original sample scene: here you can see my render and the almost original sample scene both using implicit surface (I only activated the floor collision on the sample scene):


To get rid of those grid artifact, I use the mesh instead of the implicit surface so I can add modifier on top. I use a vol. select and a relax to relax only the top part. The problem now is that I the perfect highlights that I had in tthe implicit surface became jaggy. I can't remove that jaggyness with a meshsmooth or a relax... it's not so bad on a frame, but in the animation it's even worse.


By changing the isosurface level from 0.5 to 0.25, the jaggynes is reduced in some parts but appears in other parts...


the options are:

-makes 2 render (one with implicit and one with mesh) and try to fix that in comp

-find a way to add modifier on implicit surface (is that possible?)

-generate particles on the geo, exctract prt from that and send it to frost. I still have to figure out how to do that...

[Ivaylo Katev]

well, you are now at point which i have never reached. what about to increase a bit the surface level and to use push or turbosmooth modifier? the increased surface level will make the chocolate thinner, but the push modifier will return its original size

[jstrob]

I tried relax push turbosmooth and different isosurface level, everything on the mesh and when animated it looks really ugly. For now I will do 2 renders just to get a perfect result fast (or see if i really can that way). But I will then try to use stoke and frost to generate particles from the volume of the mesh and try to mesh it with frost.

From that I see that phoenixFD really needs a way to tweak the isosurface with modifiers or a way to be meshed rapidly with frost.

[Ivaylo Katev]

there is a technique that i forgot to mention, it was used in the previous chocolate sample (of 2.1) but in the last sample is not used because it looks smooth enough. i speak about the smoothing feature placed in the input panel. this feature modifies the input data before it is passed for rendering
see the attached image, there is a big difference

[jstrob]

Hi Ivaylo,
It looks to work very well in your example. It was already activated in my test. When I remove it I can see it is even worse. I played with the settings but I can't improve it.

[Ivaylo Katev]

ok then, you can try to switch strong surface mode off and to use low sharpness, for example 0.3
this is not a good solution, some problems should be expected, but it may solve the grid artifacts

[jstrob]

ok good to know I will that. Do you think increasing resolution could help too?

In the meantime. I have an better result by combining the good parts of the mesh and the isosurface in comp. I have to make 2 renders and some roto though... and it may not work in all situation.

[jstrob]

And what about the Geometry volumetric mode? I haven't been able to render anything from it.

[Ivaylo Katev]

geometry volumetric is for fire and smoke, not for liquids.
well, increasing resolution will make the grid artifacts smaller, i was considering to suggest it but didn't do it.

[jstrob]

Maybe if the artifact is smaller I can get rid of it more easily or maybe the mesh mode will behave better in higher resolution. If they become very small I could even maybe make them disappear in comp with a filter like "remove grain" in after effect.

Thank you so much for all your advices!

here is the result with strong surface OFF and sharpness at 0.0: the sim is pretty different. the artifacts are decrease but still apparent.

[jstrob]

I just discovered that if i generate particles with the phoenix source, I can then pick the grid with frost and it works! I thought it was not supported directly. I am doing a sim with 20M cells and I will try to mesh it with frost.

[Ivaylo Katev]

good to know! btw without strong surface mode the sharpness must not be zero, this is not a liquid.
anyway, i still wonder how this problem arose. in the sample scene there are no significant grid artifacts, and it is even not smoothed. i wonder what we are missing. well, your images have bigger resolution but this is not enough.

[jstrob]

If you look at this:


This is from the sample scene when we let it settle a bit more. The grid artifacts are present.

I just look at my 20 M cells sim this morning and I had a problem with the particles generation. It seems that because I stop the sim in the middle at some point and then restarted it, the particle were resetted... In my result, the liquid is ok but the particles completely disappear in the middle and then start being emitted in the middle... I think it's a bug. i will resimulate witrhout stopping.

[Ivaylo Katev]

well, then i know what is it. your words "to settle" made it clear. this is actually the wet cookie geometry, there is actually almost no chocolate over it.
so, the solution must be sought in direction lower non newtonian and perhaps higher resolution. there must be chocolate over to cookie if we want to avoid the ugly grid.
one weird suggestion - try with as low as possible non newtonian, and some attraction to the cookie (use the right button dialog).