This is another one of the Global Illumination Test Scenes -FROM THE YEAR 2000- presented here.
This scene was designed to test shadows created by an area light floating above 8 rectangular beams. The light is aligned with the bottom set of beams.
This is not the most exciting test scene, so I’m going to spice things up by talking about SAMPLING STRATEGIES!!! WOO!
The closer the beams are to the floor, the sharper their shadows become. This is a 10 second render to make sure my camera is in the right spot. Each pixel has one sample. Each sample has one light sample. Because I’m using the minimum number of samples, there is noise.
One sample per pixel and one light sample per pixel. Lots of noise, but the solution is apparent.
One interesting question posed by this scene is:
Am I better off increasing light samples or pixel samples? What are the trade-offs? I tried one render favoring light samples, and another favoring pixel samples.
1×1 pixel samples, 4×4 light samples. The shadows are cleaner, but geometry edges are more aliased. See the pixelated bright beam edges.
4×4 pixel samples and 1×1 light samples. The geometry edges are antialiased, but the shadows seem more noisy.
A strange problem. The only compromise is not to compromise. Anyone up for chess?
4×4 pixel samples and 4×4 light samples. Nice and clean in 20 minutes. If you don’t count quantizing…
The quantizing of the grey plane is irritating, but should be expected since I’m rendering to 8 bit per channel RGB.
This test doesn’t require indirect bounce, but it’s still fun to test. It’s surprising how much the bottoms of the beams fill in.
Adding indirect bounces increases the render time to 2 hours. Note the filled in undersides of the beams that were black before.
See the stripes on the floor? That’s quantizing. I’m going to work on that in the next post.