Fine detail and scale issues / explanation of scale and time values?

[ben_hamburg]

Hi there,

I´m simulating something on a macro scale project, that needs to look massive, so I´m working in meters as scene units.

I have a tube and the liquid moving down it, with the camera moving away from it inside the tube.
Right now I´m simulating 170 Mill cells with flip solver, but I still get "chunky" geometry in the preview and test renders.
When I switched to "implicit" surface in the rendering tab and changed the sampler to spherical, I got the liquid to look smoother.

I just recently started playing around with the flip solver, so I´m not sure, if it just needs more cells to get the fine detail I want, or if its simply a scale issue.
I´ll have to admit, I never quite got the connection between scene scale and detail.
I guess If I have a 10m by 10m grid with 50 mill cells and then move the camera up close to lets say 10 cm of the liquid, it will look chunky, while it will look fine from 5 meters away, so maybe I´m just expecting too much?

Maybe I´d need to set up a cascading grid array for this scene, since the liquid covers a long, curved area, which is a total waist of grid space if done with just huge grid.
Would that even be possible?

And while I´m on it.

Could you elaborate again on how the dynamic options exactly work?

There is the gravity, the scale and the time spinners, which all default to 1.

If I want to make the sim look more massive, which basically just means "slow it down" visually, what should I do?

Increase scene scale?
Decrease gravity?
Decrease time?

If my scene is in centimeters: Does increasing the scene scale from 1 to 100 mean, it treats everything as if 1 unit is 1 meter now?
Does increasing teh scene scale also increase the default values for foam and splash settings?
What exactly does the time spinner do?

[ben_hamburg]

Here is a screenshot:

[Ivaylo Katev]

170M cells is a huge res, it's not the reason, but as you mentioned if you have 50M cells and the camera is closeup , this is not enough. actually the camera setup is very important for the simulation, with proper setup you can spare a lot of the resolution.
the cascade technique used in flip does not have the same quality as in the grid core, because it relies on the brush source technique, that is not natural for particles, the grid is a 3d bitmap and you can paint overt it, but with particles this is a bit harder, you have to produce particles between the existing ones, that leads to undesired changes in the behavior

[ben_hamburg]

Ah, now you lost me, that explanation was a bit too technical for my simple mind...

I figured its a huge res, but since I only have to sim about 100 frames, I just let it run through.
This sort of setup (liquid inside a tube) is unfortunately something I have to do quite often, its a pain in the a** to setup anyways, because you can´t properly place lights to make it look good, you always have to have the camera target pretty close to the camera, so It doesn´t penetrate the tube when it goes around some corners, etc.

So by camera setup you mean how close the camera is to the liquid?
Because like I mentioned above: Since the tube is curved, the camera can´t be too far away from the liquid, or it would lose it from view around the curves.

Would it maybe work to check "inertia" and move the grid along with the camera, to avoid those huge cell counts?

[Ivaylo Katev]

if the camera is following the liquid and can see only small part of the pipe, the moving simulator is not a bad idea, you can give it a try