I have done some simple tests with reflective caustics in Cycles (Blender 2.90) using the following scene consisting of one perfect mirror and one with roughness = 1. Both are fully metallic and the light source is a sun lamp shining trough two holes in a wall:
Rendering this in Cycles with reflective caustics on and Filter Glossy = 1 gives the following result:
Shouldn’t the reflected light from the perfect mirror give a single spot of light on the floor? I understand that there are limitations when it comes to caustics in Cycles and that LuxCore is a better alternative for this kind of scene, just wondering why the reflected light is so diffused?
Bled file attached:
relective_caustics_test.blend (875.7 KB)
You must decrease “Direct Light”, “Indirect Light” and “Filter Glossy” values and you must use strong light and you must use high render samples like 2048, 4096 or more for Caustics.
Do not put Cycles and Caustics in one sentence.
Thanks for the info, but will it not negatively affect the other stuff in the scene…? Maybe I should give this up and look into LuxCore instead
Test using LuxCore (default settings + “Add Light Tracing”)
Cycles and Caustics can never be placed in one sentence.
Now off to “tests”:
For all tests minimum bumps are set to 2.
All tests are identical, unless something changed in title.
Adaptive sampling, filter glossy 10, clamp indirect 10 (the light on the wall is 0.7 so 10 is not clamping anyway)
filter glossy 5
Filter glossy 0.1
Filter glossy 0.01
Filter glossy 0
No adaptive sampling, 3000 samples.
Filter glossy 1
A crop for the desired arena with 10000 samples
So, no, you will never receive anything similar to caustics in Cycles.
Now, in luxcore…
Halt time = 30s.
Path + light tracing (sobol)
Path + light tracing (metropolis)
Same as above + photongi cache indirect/caustic
7. No caustic cache
8. Caustic cache
If you like caustics:
I’m messing around with caustics and thinking about a scene. Anyone have a nice idea for them?
Those are only 10 minutes sampling renders, with some nice scenery I’m going to leave it overnight
I’m thinking about putting this one for overnight:
Many thanks for your time
@Acrivec testing and providing examples. Your caustics renders are really impressive, I will for sure learn more about LuxCore and try it out.
Maybe a good first project could be Chris Isaac’s “Mirror Suit”
the key is to use branched path tracing and multiple importance sampling on the lamp
NOw caustics in cycles are being developed
Yep - currently the MNEE method only works with refractive caustics, however other methods such as path guiding are also being worked on which should help with reflective caustics.