Atmosphere Simulation

Post your tests, experiments and unfinished renderings here.
Fox
Posts: 435
Joined: Sat Mar 31, 2018 11:17 am

Re: Atmosphere Simulation

Post by Fox » Thu Mar 12, 2020 3:11 am

With latest numerical precision fixes, the BiDir can render large scenes.
3h on e5 2620 v4
Atmosphere_3h_BiDirVM.jpg
I used the "Glare_Layer_vol" for camera vol.

And the hetero volume type with multi / max steps 80 for all.

The step sizes are insane, from the direction of Sun, between Atmosphere_3 and Atmosphere_2, there is close to 600km.
So i divided it by 30 and get 19503.3m, same with others:
Atmosphere_1_vol step size 4926.7
Atmosphere_2_vol step size 10230.0
Atmosphere_3_vol step size 19503.3
Glare_Layer_vol step size 4243.3

Smooth shading seems to not work with that insane scale for Moon, i use shade flat and upped the geometry.

wasd
Posts: 315
Joined: Tue Apr 24, 2018 7:20 pm

Re: Atmosphere Simulation

Post by wasd » Thu Mar 12, 2020 10:29 am

Moon lighting is not realistic there, because there's (almost) no reflection from earth.
I made the earth to be a complete sphere.
moon2.png
moon2.png (11.06 KiB) Viewed 831 times
Albedo for earth and moon:
emalb.png
emalb.png (3.6 KiB) Viewed 831 times
CPU Bidir (sometimes GPU Path) + Metropolis | Log Power | Core i5-4570

frank_yifei
Posts: 16
Joined: Fri Mar 09, 2018 1:46 pm

Re: Atmosphere Simulation

Post by frank_yifei » Thu Mar 12, 2020 2:23 pm

the blue sky is amazing
Attachments
untitled.jpg

User avatar
Glendaloch
Posts: 23
Joined: Fri Jul 06, 2018 8:35 pm
Contact:

Re: Atmosphere Simulation

Post by Glendaloch » Thu Mar 12, 2020 6:46 pm

Super interesting work here!

Yes that blue one has real depth, but also the sunrises (sets?) are great too.

Fox
Posts: 435
Joined: Sat Mar 31, 2018 11:17 am

Re: Atmosphere Simulation

Post by Fox » Fri May 08, 2020 12:29 pm

Clouds didn't come out that life true, but still something.
Clouds.jpg
CPU rendered 6 hours at step size 500m, singlescattering.
Another 11 hours at step size 100m, multiscattering.
The first one is much faster, but resolution didn't seem to improve with 100m step size.
BiDirVM and filmic tonemapping. Light tracing seems to not work with this scene scale.

Post Reply