[Q] How can i use 3x7x1.5m with real water measurements fluid sim

Ive been trying to make a fluid sim in a container of about 3x7x1.5m. After checking tons of tuts about they all go about 200 in resolution but with such dimension it looks like crap. The water flow and mass is way to big, bumping it up takes much more time, tried subdivisions but i this increases bake time a lot.

Is it better to bake this in a smaller scale and then scale it up? I had been testing a other sim which is a model in cm. This looked so much better.

Ive got a feeling fluid sim is a not complete feature at this moment.

The larger the fluid sim the larger you’ll need to have the resolution to show any fine detail of the water.

As a basic example
10m / 200 = 5cm resolution
10cm /200 = 0.5mm resolution

okay thanks, what about scaling it to 1/10 of the scale doing the sim and then upscaling?

PS do you know why the jagged edges appear?

Well, doing the simulation at a tenth of the final size and then upscaling will probably work alright. In fact, it might allow for finer control over the “behavior” of the fluid. However, I don’t think it will solve your problem regarding the big drops of fluid. After all, all details of the fluid will be enlarged, as soon as they are upscaled again. Personally, I tend to use resolutions higher than 250 for the final bake most of the time and I find this is sufficient for most renders to look decent.

As to the jaggy edges: Frankly, I’m not quite sure, why they appear. But I do know that a lot of people have this problem. I believe the developers are actively working on a way to solve this, but don’t pin me down to that. Meanwhile, one way to improve these jaggy edges would be to increase the “Smoothing” parameter in the fluid domain to a number higher than 1.0. Another possibility (and the one that I am using) is to use the “smooth” modifier after the bake. This gives me more flexibility, but tends to eliminate some of the finer details. But I can live with that. :slight_smile:

By the way: You mentioned your fluid looks “like crap”… You didn’t forget to set the shading of your fluid domain (in the toolbar) to “smooth” instead of “flat”, did you?

Thanks Marcus, i did all what you said already. I also found something about adding a smooth mod to it afterwards. The details of the water are sort of okay, i most go higher but it will take much more time.

I also tried adding more details to the ‘obstacle’, but this doesnt help either.

I would like to know why the jagged edges appear on sloped surface or organic shapes. The mesh is detailed enough but it seems the ‘obstacle’ gets a low res cage or something. Ive seen this question a lot and noticed it has been around for years??? Quite awkward they never really looked at this yet.

I just came across a demo for a Blender Mantaflow (“a unified smoke and liquid simulation that might replace the ones in Blender at some point in the future”) integration that’s currently been worked on: http://www.blendernation.com/2016/10/06/blender-mantaflow-integration-demo-2

Just thought that might be interesting… So, those jagged edged might “at some point in the future” be a problem of the past.

Well i learn a trick by a different user. If i make the liquid domain longer at the bottom, at the end of the bake i can apply a boolean. This makes the jaggered/stairs a lot better and almost is gone then.

Thanks for the link! Ill check it out

I’ve used the manta flow branch. It’s good :slight_smile:

But bare in mind, the workflow for getting the liquid sim to work is very different to the smoke sim in this build. You first need to export a python script (Theres a button to do it automatically in the mantaflow physics tab) then, you must change the cache option to bake to .obj and choose light compression (Heavy does not work as of yet) then you can bake and it will work :slight_smile: it took me ages to figure out how to get it to work on my own! lol

It seems to produce good results at low resolutions (70 - 128). and also bakes faster too

heres the manta flow branch - http://graphicall.org/1199

i looked at it, only windows ;(

heres a linux build with cycles disabled - http://graphicall.org/1197

Sorry bro, but im on OSX.

Hi rombout

I was curious if you had any issues using the boolean modifier with the fluid simulation
to remove the jagged sloped surfaces?

I’ve found that while it works quite well for a still image, for animations, it can be
problematic, since every once in a while it’ll mess up on a few frames; where the fluid
will just disappear or you get an incorrect solution. Which, of course, ruins the animation.

For instance, using a boolean modifier with your file on frame 249:

http://pasteall.org/pic/index.php?id=108694

and then for frame 250:

http://pasteall.org/pic/index.php?id=108695

The other downside of using a boolean modifier with a fluid sim is that it really slows down the UI, although you can disable it using the eye icon in the modifier options.

I see you’ve been searching for an answer on the internet

As you said there, smoothing excessively, can delete the finer details of the fluid simulation.

There’s also this (in case you haven’t seen it):

Perhaps the best solution at the moment is to use a moderate amount of smoothing in combination with the boolean modifier?

Shame you cant use manta :frowning:

Hopefully someone will port a version for OSX soon :slight_smile:

strangely, here it says it is supported for mac and linux, but not for windows ?? i’m confused, guessing it’s an old post… but if it was integrated with mac before windows, why is there not a build available?

https://wiki.blender.org/index.php/User:Sebbas/GSoC_2016/Documentation

Jamie B

I guess just no one has compiled and posted a version for windows yet.
I got tired of waiting and just built it on windows myself.

You can see that sebbas has fixed the windows issue (Last item in the ‘Done’ column at https://trello.com/b/EmyScKlZ/blendermantaflow )

It seems like Manta Flow still uses a domain and is missing quite a few features from the elbeem fluid simulator.

There is builds for windows now :slight_smile:

Just not for mac… i got my Windows version from GraphicAll :slight_smile:

And yes it is limited, you do need domains, but you can have open domain borders :). Also, you cannot export to .obj yet. But it is definatley useable, and produces very nice results with low resolutions :smiley:

The last link i find weird. I though Boolean always work better with quads?

PS i used the Boolean just to check it. Someone else pointed me on using that as a tip. If its freaking out at the end of the animation thats okay, just make it couple frames longer

I’m not 100% sure if booleans have a problem if the mesh is quads or tris.
I converted the boolean object’s mesh to tris, and it had the same issue.

I’ve searched around a bit more today, and it appears that the boolean modifier does have an issue
with open/non-manifold mesh:

https://developer.blender.org/T38193

I converted frame 250 of the fluid sim to a mesh and look at all the non-manifold parts:

http://pasteall.org/pic/index.php?id=108711

The following link seems to state that the problem occurs when using the ‘Carve’ Library/Solver:

https://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.62/Boolean_Modifier

However, in the instance that failed, the solver was set to ‘Bmesh’; so maybe this solver has the same issue?
I tried switching the solver to ‘Carve’, but it just produced an error: /!\ Cannot execute Boolean operation.

Looking through the documentation, it seems that you can also run into issues when the faces, of the
two objects that the boolean modifier is being applied to, overlap or intersect.