Simulating a gas giant results

[japetus]

The last week or so I've been trying to simulate something like a gas giant planet after becoming curious if I could. Been a lot of mistakes along the way but I think i'm getting results I like finally...as always I'm wondering if there's a better way to do it. It mainly consists of using Max's Motion Field helper to spin the smoke around in opposite directions around the planet, and then using a sphere to cut out a perfect sphere shape since the smoke does tend to flatten out at the poles. Still messing with the best sort of render settings to use (ie: phase or not, absorption color, opacity, etc) But I finally8 started to get promising results last night, so I thought I'd share. The grid is about 300 million cells and takes about 5 min/frame to calculate, this is frame 95. I'm open on comments on how to make this better, it's been pretty interesting to figure out how to do.

[fixeighted]

We all of course want to see it animate!
Good share though...will look out for it developing.

[Joelaff]

What about a VolumeGrid with a texturemap (possibly VRayDistanceTex is involved to get that sphere)? Are you intent upon simulating the actual swirls of gases? Or a sphere with a texturemap used to emit for a few frames to build up the atmosphere. Then perhaps volume displacement (which is slow to calculate)? Just thinking out loud here.

[japetus]

What about a VolumeGrid with a texturemap (possibly VRayDistanceTex is involved to get that sphere)? Are you intent upon simulating the actual swirls of gases? Or a sphere with a texturemap used to emit for a few frames to build up the atmosphere. Then perhaps volume displacement (which is slow to calculate)? Just thinking out loud here.

Yeah, the idea was to simulate the swirly gasses to get something more interesting than usual (a Jupiter texture, or swirly bitmap in Substance Designer). I also just realized a Body Force might be a good option to keep things together better than the spherical gravity I was using. Test 500 here we go!

[Joelaff]

Could you use an outer shell of some sort to keep the gases inside of an enclosing sphere?

[japetus]

Could you use an outer shell of some sort to keep the gases inside of an enclosing sphere?

I'm trying that now too..before when i started this, was getting really bad dark voxels, but maybe now It'll work better that I have everything else dialed i!n!

[Joelaff]

Possibly emit from a textured sphere, let that build up to get your base of colored gases. Then use that as your initial state for some turbulence or other gas animating techniques you have discovered. ??

Interesting experiments. Keep us posted.

[japetus]

So here's the simulation hitting against the outer sphere...you can see the shadowy voxels, which might get better at a higher voxel count, but i can also shell the inside just a bit more and cut off that bad stuff and reveal swirly clouds. This mighit be my favorite test yet!

[fixeighted]

Yes that one is a considerable upgrade

[Joelaff]

That's cool. Maybe you could play with the scatter settings to get some interesting effects on the terminator line.

[japetus]

Did an overnight sim and got some good results, of course now putting it in the scene I realize I need to go even higher! Tried out some methods including Resimulation, which ate up all of my 96 GB of ram, but then I discovred I could get a lot of the detail I needed by using some grid-based motion blur to smear the clouds more (set to about 3.0), and adding a little vector displacement baesd on speed. The results are pretty stunning I think!
Edit: Advection Displacement works almost as good as Vector and is much faster to load and calculate, so I'm going with that.

[fixeighted]

That's the ticket! That first one looks like the direction to go....lots of lovely detail with real perceivable depth and bigger mushed stuff too, largely like the references I've looked at.
What a fun experiment

[japetus]

Been messing with this way too much, learning more ins and outs as I read the documents (apparently the information never sticks in my brain). Right now I'm trying to avoid a cutter object and am using a Confinment object to keep things in a spherical shape. But I do run into nasty grid artitfacts still, especially when the smoke is pretty dense. Using a cutter object seems to freeze or slow way down on my render, so I was wondering if using a different Voxel type like Inscribed or Centered would help with that? I'm reading some conflicting information on when to use Inscribed or the default Circumbscribed. I'm guessing I'll be forced to use a cutter object in the end...at least to clean up the top layer of voxels.

[japetus]

Finally realized I could use a VRayDistanceTexture to cut out my perfect sphere instead of the cutter object, now my render isn't freezing up (taking forever). I've been maxing out my machine with 96GB of RAM with a 345 million voxel grid, which, I wish I could go to 500 million, i just barely have enough detail to get a medium close shot. I tried a Grid Resim at just 0.25 resolution and that crashed the sim, but at least I am able to do a time slowdown resim to get a little cloud movement going on in my animation so you can get the sense that it's not just a texture, which is pretty cool.

[Joelaff]

Looks awesome! That was what I was getting at in post number 3 above. It seems fairly controllable, and makes for that nice falloff.

I wouldn't mind a little more "detail" in there, which might just be some grain added in post. Though possibly if you doubled the size of your texturemap and added some noise to it before simulating... Then you would get more fine streaks.. maybe?