Page 3 of 7
Re: big mirrors and fireflies
Posted: Thu Jul 25, 2019 9:57 pm
by marcatore
Dade wrote: Thu Jul 25, 2019 9:25 pm
Otherwise, have you tried to enable hybrid forward/backward path tracing in your original scene, it should transform that fireflies in nice caustics
How to? Enabling "add light tracing"
I've tried to create caustic using caustic caching but it crashes everytime...it seems that it eats a lot of RAM
Re: big mirrors and fireflies
Posted: Thu Jul 25, 2019 10:01 pm
by marcatore
B.Y.O.B. wrote: Thu Jul 25, 2019 7:58 pm
You could use point lights with some radius > 0 to have smooth shadows, maybe with a spotlight IES profile to shape the cone.
With point light in the inner part of the light fixture or at the end of?
If the second is correct... how should I light up the inner part of the light fixture? Emission material?
Re: big mirrors and fireflies
Posted: Thu Jul 25, 2019 10:08 pm
by B.Y.O.B.
marcatore wrote: Thu Jul 25, 2019 9:57 pm
How to? Enabling "add light tracing"
I've tried to create caustic using caustic caching but it crashes everytime...it seems that it eats a lot of RAM
Yes, "add light tracing", but don't enable the caustic photon cache.
marcatore wrote: Thu Jul 25, 2019 10:01 pm
With point light in the inner part of the light fixture or at the end of?
If the second is correct... how should I light up the inner part of the light fixture? Emission material?
Doesn't really matter, the cone should not touch the fixture anyway. And a point light is always invisible.
To have a bright surface inside the fixture, you can refer to this example:
viewtopic.php?f=4&t=1129#p13268
The ceiling spotlights contain a meshlight with disabled DLS (direct light sampling) and very low gain, just so the camera sees a white surface. Their main lighting comes from a spotlight positioned directly below the meshlight.
Re: big mirrors and fireflies
Posted: Fri Jul 26, 2019 6:56 am
by FarbigeWelt
marcatore wrote: Thu Jul 25, 2019 10:01 pm
how should I light up the inner part of the light fixture? Emission material?
I usually use light points. I make lampshade from two parts. The inner a volume object with material matte (white 1,1,1) and the outer part (actually a copy of the inner but a bit larger and overlapping) with a second material, one that makes the lamp‘s appearance. Do not use single layer objects. Build the lampshade with e.g. boolean operation difference with e.g. cones or make a spin object from a mesh converted curve and then fix the normals (recalculate or flip directions, all face normals should point outwards) in edit mode.
Re: big mirrors and fireflies
Posted: Fri Jul 26, 2019 10:56 am
by marcatore
Thank you guys, I'll check your suggestions.
Re: big mirrors and fireflies
Posted: Fri Jul 26, 2019 11:52 am
by marcatore
Below a test with point light (with radius 10cm) + IES file.
It seems it's really working...it remains the "fake" illuminted light fixture.
I'll try what suggested.
It's working.
Next week, at work, I'll try to switch the scene to this setup and check with everything how it works.
Thank you for your support.
Re: big mirrors and fireflies
Posted: Mon Jul 29, 2019 10:31 am
by marcatore
Tried to port that setup to my original scene.
It seems it's working really good.
Below the un-denoised and denosied results.
20minutes rendertime - 957 samples reached.
Re: big mirrors and fireflies
Posted: Mon Jul 29, 2019 10:42 am
by B.Y.O.B.
Glad to hear that, nice results.
Re: big mirrors and fireflies
Posted: Mon Jul 29, 2019 12:30 pm
by marcatore
Now I have to finish converting some materials with the correct textures and assign different material to the walls like the project is designed.
And then there are some graphical lightbox I have to tweak.
But the main issue is solved about lighting and fireflies.
Re: big mirrors and fireflies
Posted: Tue Jul 30, 2019 7:08 am
by marcatore
Some questions (again on mirrors...but not only)
Look at these pictures.
I've launched yesterday some HD images (2300px wide) to test at "full res" how it performs.
Most of the image areas are clean (halt condition to 2000 samples)and it's quite good considering that they were completed in 35minutes but there still some areas with strong noise like highlighted in the previous pictures.
what should I have to do to make them cleaner? Just more samples? Why, except for that wall area (there are some small areas on the wall with noise like that...but more subtle and not so strong), the mirrored objects have so low sampling?
Thanks in advance.
EDIT:
Tested now without any halt conditions and it seems that at 4800 samples (I did a render region, I hope that samples count is correct for comparison) it still noisy.