Issue: Dark wall in reflection

[Zafar Iqbal]

I'm doing a number of indirect lighting related tests and I've stumbled upon something that is puzzling me with this latest test.

Following are frame 1 and frame 34 along with 1:1 crops on the sides. I manually rendered the last one, frame 100, because the ceiling would be directly visible in it:



I'm using a single photometric light, FG (3 bounces) and GI (both FG and GI regenerate maps on each frame). View Dependency and Importons are off. Arcitechtural material on everything with AO Color Bleed enabled. I'm also using the Photographic Exposure filter. 2 refl ray bounces in the render options btw.

The "room" consists of 3 materials - orange, green and the white one. The room is entirely enclosed with faces inverted and visibility changed to front faces only in Visibility Options for that mesh.

Everything being too dark in the reflection would perhaps make more sense - what I'm seeing here doesn't and I'm hoping someone can shed some light on this.

Ps. the lighting in the animation from frame 1-34 (which is all that's been rendered so far) is rock steady. No issue except for what you see in the reflection.

[Zafar Iqbal]

Just noticed (on an accident) that the reflection looks fine (regarding brightness) with GI only, but is darkened a lot once FG is enabled (with or without GI).

That gave me some keywords to feed to Google and this is one of the better threads I found about it : http://forum.mentalimages.com/showthrea ... eflections - but it's not helping unless I increase rendertime with an insane amount. Thats just unrealistic in a production scenario.

I'm actually very surprised by this - but luckily we can append to the FG map. I'll try a test test with a multicam setup (for FG calculations only) and see how that goes.. should work.

[Zafar Iqbal]

Adding multiple cameras for FG only calculations did the trick, and even though I have multiple cameras, four in fact, the combined average is 1½min per frame for FG calculations, and the solution is rock steady. The entire frame takes 7½mins to render on my i7 920 with motion blur enabled and some 35mins with Color Bleed on top of that. I'm starting to love the FG+Color Bleed combination but it's a shame it's that slow.

[Kzin]

Adding multiple cameras for FG only calculations did the trick, and even though I have multiple cameras, four in fact, the combined average is 1½min per frame for FG calculations, and the solution is rock steady. The entire frame takes 7½mins to render on my i7 920 with motion blur enabled and some 35mins with Color Bleed on top of that. I'm starting to love the FG+Color Bleed combination but it's a shame it's that slow.

have you used enough reflection depth's for fg computing?
if you use raytracing, then disable mb for fg compution. render fg computation only and after that, use "use fg map from file" with motion blur at rendering.
for motion blur, try to use the unified sampling, renders faster with mb enabled.

[Zafar Iqbal]

The FG ray depth isn't the issue here but It seems like FG does only look at the complexity of the mesh with the reflective property - or so I read.

I used the extra cameras to do FG only, as mention previously, and then read the appended solution in my actual render pass. I also used Unified Sampling. It looks very promising although I can't make sense of what the Quality parameter is doing. Whats quality "1"? quality 100? When is it "best quality", and if there aren't any best, since you can just up the Quality parameter, then whats the Max Samples for anyways? I like Unified Sampling and can see the benefits in quicker rendertimes, but I'm not fully understanding it yet.

I stumbled upon MX_fgshooter. this pretty much simplifies and make the setup with multiple cameras (passes) and appending FG solutions much much simpler and intuitive. Nice!