Volume sourcing with reoriented simulation container

[ChristopherC]

Hi everyone!

Let's say that I have some smoke shooting in one direction. I would like to have the orientation of the simulation container aligned to that direction to try reducing the number of unused voxels.

To do that, I'm using a Point Position DOP and all is fine until I try to source my volumes. In the SOP context, the velocity definitely goes into the direction that I want but, when imported into DOP, it looks like it's being interpretend as being in a space local to the one defined by the Point Position DOP. In other words, it doesn't go anymore in the direction that I'd like to.

I fixed that by doing an inverse rotation on the source volumes before they're being imported in DOP but it looks a bit hacky. Is there a better way?

See attached scene.

Cheers!

reorient-simulation-container.hiplc

[StepbyStepVFX]

If you want to reduce the footprint of your sim, try to split your domain of computation by instancing containers (and also re-orient them), and use Gas Resize Fluid node (to adjust the size of containers dynamically, but don't forget to limit the max size so that they don't all grow at the same time and cover the same area... would be useless). Here is a nice tutorial (you'll need to pay), but I think that instancing is not to hard understand with the docs :

https://gumroad.com/l/TqUNR

By the way, if your problem is also, after sim, to save space on hard drive, think about using the Volume Compress node, who allows you to "quantize" the values of the fields you want to save on disk, and therefore encode them with less memory footprint :-)

Hope this helps

Edited April 11, 2018 by StepbyStepVFX

[ChristopherC]

Thanks for the tips!

I already know about clustering but my sim really has no need for clustering (it's nothing like a smoke trail with sparse areas), the only unused voxels come from the container not being aligned properly, and so I was imagining that reorienting a container was a standard practice?

A proper alignment would even help the Gas Resize Fluid Dynamic DOP to possibly grow in 1 dimension instead of most likely 2 or 3 dimensions otherwise.

... aaand, as a bonus, the volume slices become much more helpful when aligned to the natural flow of the smoke!

[vtrvtr]

Sourcing assumes world space, you need to change that to get the correct orientation

Here's one way to do it

 

 

orient_pyro_container_odf.hiplc

[ChristopherC]

Cool! It's still hackier than what I hoped for but it's a good alternative! I think I'm going to wrap your approach into a custom Source Volume DOP digital asset, and that'll be it.

Thanks!

 

[ChristopherC]

I'm having another issue with this scene.

I'd like to set-up some vorticles but can't figure out how to make them to play nicely with the position data.

In the attached scene, the vorticles are being sourced and are advected over time, as expected, but they have no influence on the smoke simulation. I believe this is because the vorticles are defined in world space while the simulation space has been modified through the Point Position DOP.

Alas, when setting the 'Position Data Path' from the 'gasvorticlegeometry1' node to '../Position', things are breaking in the SOP Solver DOP named 'source_vorticles'—the vorticles are sourced properly but the vorticles merged from the previous frames through the DOP Import SOP go all over the place. I tried playing around with parameters like `SOP Output is in Simulation Space', to manually build the transformation matrix from the position data to then apply it in the SOP Solver DOP, and whatnot, but to no avail.

Any idea of what needs to be done to get this setup working?

reorient-simulation-container.hiplc

[ChristopherC]

I gave another go at manually building the transformation matrix from the position data and this it worked out, not too sure why it didn't in the first place.

See the node '/obj/geo1/dopnet1/source_vorticles' for the fix.

reorient-simulation-container.hiplc