I’m trying to render what I think is a relatively simple scene in Cycles but I’m getting horrible noise, even when rendering at 1000 samples. I’m trying to light the inside of a dome with a few spotlights. Simple, mostly diffuse materials. For this test I have a light emitting ring around the base of the dome and 8 spotlights pointing up.
Caustics are off. Shadows are off. I’ve tried clamping direct and indirect paths. Filter glossy at a high number. I’ve tried everything I’ve been able to find about reducing noise. I know some scenes are difficult for Cycles (like indoor scenes lit only by bounced lighting, hence light portals). But this seems like a straightforward one.
Can you think of something I’m doing completely wrong? Should I just ditch Cycles for this and use Blender internal?
Here’s a version of the scene rendered with Blender Internal in about a minute. It looks nice but not as realistic as I can get with Cycles. I still think this kind of thing should be pretty simple for Cycles. Any help figuring out what is causing all the noise would be appreciated.
Because of using spot lights use branched path tracing. After you can finish it with compositing. I rendered it, but I can´t upload it here and it looks much more better. Try it.
One thing I found helped, almost the majority of your lighting is direct lighting, I turned the clamp indirect to 0.1, see if you like/dislike how that looks (seems similar but less noisy to me)
In this case, the major other source of your noise is the spot lamp size, if you set it to 0 instead of 0.5 (but of course becomes sharp) it renders the majority clean in 20 samples, with the edges of the ring taking 250 samples to clear. IMO (and not necessarily fact) I find the edges of a non standard camera are often more noisy, like in your project.
I found turning the lamp down to 0.1, the blend factor up to one and the degree size of the cone to 55* for example produced something similar that rendered faster.
@Kraward and @zeealpal, excellent advice. It really helped me fine tune the scene and the settings to achieve a much more acceptable result.
The change that had the biggest impact was setting the “Clamp Indirect” to 0.1, followed by changing the size of the spot lamp size to something very small.
I also realized that the noise on the edges was not really caused by the non-standard camera but because most of the light was hitting the surface from a very shallow angle. I was able to get much better results by changing the flat light ring to a torus and moving it inwards so that the light fell more evenly on the dome edges. I also moved the spotlights inward to widen that angle as well.
With all of those changes I was able to get this result with only 100 samples.
