[Fernando Tella]

I think Martin talks about the shelves at the bottom part of the windows. I think bounced light comes from those surfaces too.

[Hugh]

I'm not clear on any of this to be honest - as far as I was aware 'caustics' is a word that cg people use to describe all direct reflected/refracted light in nature - regardless of specularity or nd. Perhaps my understanding is too limited but it doesnt strike me as an issue to provide a black and white function for turning this off.

I agree, the definition of caustics within an unbiased engine is a little murky.

The simple answer is if a material layer has a roughness of 0 to 1; light reflected from it to other surfaces will be defined as caustics. If the roughness value is 2+, bounced light will be classed as indirect. So in the second instance, turning caustics off will not turn off the reflected light.

I don't know if this will follow for refraction also, I haven't tested it. My guess is that it's easier to define refracted light and so easier to eliminate.

[mrcz]

I think Martin talks about the shelves at the bottom part of the windows. I think bounced light comes from those surfaces too.

Yes, exactly. Top of the shelves which looks like a cover of the heating units.

[Mihai]

Well I asked around It looks like it's a lot more complex than simply what/when it's considered caustic light or just indirect reflected light. Deflix could you please send me that scene for checking? It would help a lot to find out if there is a problem here with caustics/roughness. I tend to agree with the other posters here, that noisy light on the ceiling doesn't look like it's coming from the floor by the shape of it.

[Hugh]

Well I asked around It looks like it's a lot more complex than simply what/when it's considered caustic light or just indirect reflected light.

Could you be a little more specific please, Mihai.

My test was admittedly very simple. It consisted of an emitter inside a black tube to create a collimated beam; A flat plane in the path of the beam but set at a 45 degree angle to it; and a diffuse object to illuminate, placed in the path of the reflected beam. With caustics turned off, the illuminated object only receives light when the roughness value of the 45 degree flat plane material is at least 1.013. Anything below that value, requires that caustics be turned on, otherwise the scene remains perfectly black.

In this scenario, it doesn't seem at all complex. So it would be most helpful if you could expand on your statement in order that I can make my reflector materials in such a way that I can turn off caustics reliably, if needed.

[Hugh]

I think Martin talks about the shelves at the bottom part of the windows. I think bounced light comes from those surfaces too.

Yes, exactly. Top of the shelves which looks like a cover of the heating units.

I think that the caustic patterns would tend to run parallel with the units if this was the source, instead of at an angle. Is there any geometry outside the room that could be responsible. It would have to be below the line of the windows, reflecting the sunlight back up.

[Mihai]

Hugh, I'm sorry but I don't know how it is decided exactly either. But perhaps there is a problem here and the 'cutoff' shouldn't be that sudden. If you still have that scene please send it to me, it would help to check what's going on.

[deflix]

The patterns on the ceiling are definately from the floor - 100%
This is from direct sunlight coming thru at an angle. You can not see the highlights on the floor due to nd/specularity.

Mihai - I have sent the file to you via email.

[deflix]

Mihai,

Did you have a chance to look at my scene file?
This project is overdue now....

thanks in advance

felix