[Becco_UK]

I'm running a simple scene test - emitter light through glass onto wall - with another mirror wall - as expected the caustic light is not shown in the mirror wall.

Perhaps in time it may start to show but I'm not waiting a lifetime!

The preview is more revealing - a mass of coloured noise where the reflected caustic should be. This is probably why scenes that should feature reflected caustics, jewellry for instance, always look incorrect.

Only a major rewrite of Maxwells' engine is going to fix this problem.

[sandykoufax]

here is a SL 25 image, but I can't see anything.

[Becco_UK]

sandykoufax: Is it really necessary to project through glass, which isn't going to give desired results, for reflected caustics, no matter how high the sample level goes?

The above image is my simple (sample level 17) test. Light through glass (which is a caustic) does not reflect correctly.

[sandykoufax]

If we can't get it with maxwell, we have to do it in post-process.
But I hope that maxwell can do this with it's power.

[Becco_UK]

sandykoufax: I agree Maxwell is very powerful but in some areas it performs poor. This problem of caustics not reflecting correctly is significant - it was known about some years ago so you may have to live in hope for quite some time yet!

[KurtS]

It works if you make a "camera obscura" and place an mxi emitter inside the camera (projector), but this requires a very high SL

[Becco_UK]

KurtS: What sample level do you think would have to be reached for a noise free caustic reflection to be reached?

[KurtS]

I don't know. i just made a very quick test, and I have to admit it did not look good at all. I doubt that the camera obscura-method I suggested will produce descent results at all, because there is so little light coming out of the hole in the box.

[lebbeus]

sandy--

none of the maps come with your file, mind sharing? If they're proprietary, I totally understand…

[sandykoufax]

lebbeus, I didn't archived that image for the big file size.

You can use any images you have to do this.

[wimver]

it is just a matter of time (measured in centuries) before this image clears out.
I let it run till SL20, and disabled the ambient light, so it shows a bit of the reflection.
I guess with SL250 all the noise will have gone ))

cheers
wim

[KurtS]

about the caustics problem in general: no doubt this is one of the most critical weaknesses of the Maxwell render-engine. I really hope that the NL teams comes up with a solution to this problem soon.

[Becco_UK]

KurtS: It would be nice if someone from Next Limit would comment on progress. However I suspect this isn't going to be fixed in Maxwell 1.# Seems as though it's the code being used which for any hope of noticable would require a major rewrite and even then I have doubts. Unbiased render engines are still far ahead of hardware, similar to how advanced ray trace engines seemed years ago.

The solution might be for programmers to write code that calculates where the reflected caustic is to positioned, calculate it seperate and superimpose it in the correct position - just a thought.

[tizxx]

What do you think about this image ?



3 planes

1 plane hdri emitter
1 plane the Terra Cotta floor
1 plane the mirror - ( Reflectance 0 and Reflectance 90 = 255,255,255 - Custom IOR Nd = 10000 - Roughness = 0)
1 sphere (Green Glass)

The caustics works very fine with Maxwell Render 1.7.1

probably I not understand the problem...

Sorry...

[lebbeus]

tizxx--
this issue is about caustics of caustics (projected image) showing up in reflections; you're right, direct caustics do show up in 1.7.1