Ive been using clamp output around 1.5 to stop fireflies, and it works fine. Just recently the same scene has started to render areas visibly much darker than it was and only render correct when I turn off clamp. Before, the clamp wouldn’t visually change the image that much, it would just fix fireflies. No idea whats going on here. I reset my render settings to default and that didnt fix it. Andy reason why clamp output would suddenly darken an image by quite alot?
Do you have max ray intensity set to something low?
Also the reason for your fireflies could be a hot single pixel light source weather in hdr, or a distant light which causes the fireflies so best to isolate that and fix it there. Clamping might change the way the image looks *(though shouldn’t too much as you describe)
Max ray intensity I think is usually a value from 5.0-20.0.
The fireflies are from small studio spot area lights. Ive got a result overall for my car Im very happy with so I don’t want to change the lighting, and the clamp output was helping calm down specular dings that were picking up on glossy and matte black plastics. I’ve noticed that when I rolled back to an earlier light rig the problem went away so today will be methodically working through to find the problematic light source / tick box / user error.
I’ve found so far its related to animating visibility of the light rigs.
I’ve got 12 shots and each shot group has a studio group and a light rig group. The shot group animates on so at frame 01 shot 101 comes on and 102 at frame 2 etc. The master camera is keyed to move to each hero shot on the correct frame. When I don’t keyframe the group visibility, the shot renders correctly with clamp output turned on. When I keyframe the group visibility and re-render or use RT, the shot visibly changes and gets darker. I dont know if maybe certain lights are being left out of the calculation and arent being added to the scene when the group vis is turning on. I am only using area lights.
It makes me think its something Ive done because its never happened before. Ive had animated vis on lights rigs before with no problem. Ill simplify the scene and see if the problem persists, then ping it over.
Not really sure whats causing it… It seems the issue is controlled by a combination of colour mapping settings. Ive been using clamp on 1.0 and exponential mapping to get the look I want in the framebuffer, then we take that render and, match the grade in nuke and export a LUT back into Vray so I can render without exponential. That was works fine, but with some small spot lights and glossy / shiny materials I need to control fireflies and clamping seems to be the only way to fix that, and it does it well.
My issue though is somewhere along the line using clamp was drastically changing the lighting. The attached scene as Ive left it should be RT’d at frame 1 with camera 101. If clamp is on, Im getting a dark grey background, if clamp is off, its light grey (as it is meant to be).
Also, this now all has to be done in Linear multiply as Exponential + colour napping no gamma also massively darkens the image (more than usual clipping of highlights).
Im not sure if its to do with multiple light rigs and animating groups on / off but the problem seemed to appear when I tried to setup the scene to render each light rig per frame and have one camera animate to each studio shot position.
Thanks for the scene. I had some time to play around with it and that’s one strange thing, indeed.
I can’t really say what causes this, as switching from adaptive lights to full light evaluation seems to fix the problem, but that’s really strange, as you’re using BF+BF for GI and the adaptive lights rely on the light cache.
Another thing is the subpixel mapping - disabling it without changing the ‘light evaluation’ parameter also seems to fix it.
I know subpixel and adaptive lights don’t play well together, but we’ll have to see why this is happening with BF+BF (and it will likely turn out that it’s caused by some other thing, but the devs should be able to debug it).
Actually, I think when light cache is not used and the ‘light evaluation’ is ‘adaptive’, we use probabilistic lights (uniform probabilistic) and that’s where the problem comes from. Not sure how those play together with sub-pixel mapping.
Still, what I suggest is either switch to full light evaluation, or disable sub-pixel mapping.
I heard back from the devs, it seems that this is about how it works - sub-pixel just doesn’t play well with probabilistic and/or adaptive lights and V-Ray also prints a warning about it in the log.