Hello,
I have a serious problem with speckles this time. Mostly, these are generated by HDRIs with values way above 1. In this case however, it is different. I have a car part with a silver metallic paint. The paint is using glossy reflections, reflection depth for the paint is 5, glossy subdiv is 56. Clamp output and subpixel mapping is on.
When I hit render, I get a lot of white speckles in all the areas where a car part is reflecting itself, let say the bumper for example. If I set reflection depth for the carpaint to 1, these areas are speckle-free, but I loose the self-reflection of course.
I tried all sorts of things: Setting the AA to 50 min and 100 max, no change. Setting subdivision for the metallic paint to 256, no change. Is there any workaround? This has been annoying me for months…
High AA will undermine your high material samples - for the laugh try setting your aa max to 12 and keep your really high material samples and see if it’ll do it.
Here is the file. The speckles are visible in the metallic layer of the carpaint (see MtlSelect render element). I narrowed it down to one light in the scene. Increasing the subdivs of the metallic layer helps a bit, but I need that scene completely noise-free.
EDIT: wtf, file size limit for zip is 97kb? Come on… Please send a pm and I will mail the file to you. It is about 600 kilobytes
We are sorry about the delay of this answer.
Could you please send this scene to vray@chaosgroup.com
Please include the link to this post , and mention what is the issue with the scene.
Svetlozar, I have already forwarded this scene to your email a while ago. The issue was the directional parameter which he had used in his vray light, it caused the unresolvable noise. If you’d like I can resend it.
Another way to avoid the problem is to use the VRayOverrideMtl material and use a different and simpler material for the reflections.
There is no bug here, just the probability of a camera ray to go through the two reflections and hit the bright light source is very very small and happens very rarely. We must find another way to solve this issue, perhaps using some kind of multi-bounce adaptive sampling.
I don’t think dome light has a directional parameter. For this issue the rectangular light had a value of 0.75 if I remember correctly. Setting the dir value to 0 and making the light have no decay with value of 0 instead solved that current noise problem.
I feel that this also the case when using hdri’s of sunlit locations in a dome light. The sun in these hdri’s is often very bright, causing the same speckles. Single pixels in the framebuffer sometimes have 3-digit float values. That’s the white speckles. I have a scene file for that problem, if that helps. It is quite large though, because of the hdri.
