terrain + atmospheric scattering test

Post by **OnoSendai** » Fri May 30, 2008 11:34 am

Since I have implemented view-dependent sub-d + displacement, I decided it's time for another atmospheric scattering test, this time with terrain

So in this scene there's a gigantic Earth-sized sphere (radius 6378135 m).
There's an atmosphere sphere around that, with radius 6428135, i.e. earth radius + 50 km.
Then there's a 200km*200km quad, which is subdivided and displaced using a FBM procedural map.

The lighting is provided solely by the sun, using the spectral power distribution as measured at the edge of the atmosphere.

All sky lighting is due to Rayleigh scattering (simulates scattering off air molecules). Water aerosol scattering was disabled for this test.

I've attached the scene file, it should work with Indigo v1.1.1. (except you will get some water scattering)

If anyone has a spare supercomputer, I'd like to see what it looks like without noise

EDIT: if you want to run this scene, make sure you set bidirectional = false

Post by **fused** » Fri May 30, 2008 11:38 am

wow.... nice....

....

edit: tried to render, "found unexpected element 'center' in 'athmosphere'."
edit2: after removing all 'center' elements, indigo just hangs
edit3: disabled bidir, works. looks different tho

Post by **OnoSendai** » Fri May 30, 2008 12:01 pm

ah damn, that's right, I added some code to set the density center of the atmosphere medium. So it won't work correctly with v1.1.1, unless you move everything up by the Earth's radius

pixie · Post by **pixie** » Fri May 30, 2008 12:09 pm

Gives me the same problem too

zsouthboy · Post by **zsouthboy** » Fri May 30, 2008 12:24 pm

Looking good boss, even if we can't render it for you

Post by **OnoSendai** » Fri May 30, 2008 2:19 pm

Update at ~ 3 hours

Stinkie · Post by **Stinkie** » Fri May 30, 2008 4:37 pm

Damn, that looks cool!

BbB · Post by **BbB** » Fri May 30, 2008 6:48 pm

Very good looking. Now we can render broad vistas without need for postprod: (I tried it here and had to do all the volumetric with Blender Internal and Photoshop)

That's assuming it's not too slow. Judging from your render time, there's a bit of a speed hit, though, isn't there? I assumed you cooked this one on your quad, right?

Post by **OnoSendai** » Fri May 30, 2008 6:50 pm

There's a massive speed hit relative to the usual sky model.
This one is using 3 threads on the quad

BbB · Post by **BbB** » Fri May 30, 2008 7:03 pm

Ouch.

(then again, postprod atmoshperes are a lot of work)

PureSpider · Post by **PureSpider** » Sat May 31, 2008 3:52 am

What about a shot of that scene where you can see the sun?

Post by **OnoSendai** » Sat May 31, 2008 11:57 am

The final image.

CTZn · Post by **CTZn** » Sat May 31, 2008 12:48 pm

Approved

Where are these wonderfull clouds you showed to us (2 layers please

) ? Shadows on the rock, mmh...

Kram1032 · Post by **Kram1032** » Sat May 31, 2008 10:05 pm

Those beautiful clouds with the same kind of noise in 3D ;D xD

Here, a possibility of add/substract in normal direction would be handy... At least I guess, that'd be needed for a sphere prim with disp... - That and a more general primitive model which'd allow general prims

xD (Or an a bit pointless mesh generator where you as well could directly use a sphere...)

Hmmm.... with a correct displacement map, it should be possible to displace a low-poly sphere as if it was "perfect", shouldn't it?

Ah well.. Great test, seems to be alot faster now

(after 9h on a single core, it already looks reasonable

)

What about a marine test with that water on a certain level?

(If the water is perfectly plain^^)

Deus · Post by **Deus** » Wed Jun 18, 2008 12:07 pm

Ill use this