High Quality Fluid Recipe?

[Atom]

Hi All,

I have a mid-resolution fluid sim that is behaving as I want. So lower the particle separation to go to a higher quality but the fluid breaks up so much that the result is useless.

Why does fluid behave this way, it is a bug?

In Jeff Wagner's FLIP tutorial he sets up a sim then lowers the particle separation and cooks it in the background. I kind of got the impression from watching the video that was all there is to it...

Does anyone have a recipe for generating high quality fluid?

MID RESOLUTION: Particle Separation 0.03

 

HIGH RESOLUTION: Particle Separation 0.02

 

Edited February 6, 2017 by Atom

[Hans Peter]

Did you lower the particle separation only on the meshing node or on the flip object too? 

[Atom]

I lowered the particle separation then re-exported the compressed cache files. For surfacing I matched the particle separation to the new value in the re-exported compressed cache.

[chrisdunham95]

I tend to always have just enough power to sim at high res and then mesh aswell - I often play around with droplet and influence scale in the meshing or mesh using vdb from particles and other vdb nodes, - dont know if this is much help but just thought id input

- looking at the higher res image to me looks asif the meshing droplet/influence wise isnt working correctly or maybe even something to do with the particle radius size? - are you using the default off the shelf 'meshing' or your own technique? 

Edited February 6, 2017 by chrisdunham95

[rbowden]

Do you have enough particles in those areas that look sparse? How many particles are you dealing with in a particle separation of .02? Did you mess with any of the other parameters in your particle fluid surface SOP? Have you tried putting a bounding box around a specific problem and deleting all the particles except for in that box and tried meshing that? Could just be GL that is messing with you.

Seriously a ton of factors to look into and I don't have an updated file from you to help debug so hopefully those questions give you a start. 

 

[Atom]

Thanks everyone, I have been playing around with this all day and have better results already. I think you were right about the droplet scale being a bit too close to the influence scale.

Here is my surfacing recipe, values for the fluid surface node.

Voxel Scale = 0.92

Influence Scale = 0.92

Droplet Scale = 0.72

Edited February 6, 2017 by Atom

[lukeiamyourfather]

Needs more cowbell particles. If you're at the limits of your machine consider narrowing the scope of the simulation. For example only simulate what is directly around the character versus everything everywhere. Then combine meshes or simulations later in the pipeline so they seem like one.

[Atom]

I think I am doing better with fitting my simulation into my current ram limit. Ryan's previous tip on exporting the compressed cache first and controlling particle count helps out a lot. I have actually been able to run my current simulation on a 14GB AMD 4 core. I just let it slowly churn out compressed cache frames then grab a set of completed frames and transfer those to another machine for previewing and generating the surface cache. I have even been able to leverage my i7 16Gb MacMini again. Even though I can't really do any modeling or animation work on it it does not seem to have any problem cranking out a surface cache. I find the eight core i7@2.6Ghz is actually about the same speed as my eight core AMD @4.0Ghz.

Now that I have the camera work done I was able to crop out a lot of the lower falls. Knowing what needs to be in the shot and out of the shot does help in narrowing scope of the sim.

So what approach do you take for the combining step? Is that just compositing trickery or do you create a Houdini scene and somehow blend multiple surface caches together?

Edited February 7, 2017 by Atom

[chrisdunham95]

What are you combining? - i know when you combine a ocean gridwith the simulation that you have separately (submaring emerging/boat etc) you use the same ocean spectrum etc. node settings to deform your simmed mesh to match your ocean grid i believe  

and then a lil comp work too sometimes  

Edited February 8, 2017 by chrisdunham95

[Atom]

And 72 hours later I have a new surface cache at 0.02 particle separation using the surfacing recipe I mentioned above.

@Chris: I am not combining anything, I took the make it all in one big sim route. I don't really know how to combine two or more fluid simulations in a Houdini-centric fashion, but I was hoping someone might have an example file they could post? The speed boat example would be great.

[chrisdunham95]

Awww i see, looks real nice man! 

Ive just put together a quick file, using literally off the shelf tools - dive in and take a look its pretty self explanatory i think - but its how i approach it anyway, - you could also add a wave layer tank from the off the shelf tools for nicer whitewater on the ocean rather than using the foam within the ocean shader. 

(So my process would be render the ocean surface at whatever size it needs to be and then simulate the small area using the splash tank for interaction of that - merge that into the mesh off the ocean surface using VDB's - Or another technique ive seen is deleting parts of the water mesh and whitewater based on position and velocity leaving you with just the outer or 'surface' whitewater and splashes which you can then combine with your ocean surface and other elements) 

or just render them both and combine in comp carefully. 

 

Hope this helps  

Ocean_seamless_file.hip