[ki_cz]

I understand what you are saying, but the fact that with an HDRI emitter placed on a plane (not in the image based environment slot), makes me wonder if this is specifically a problem with only the way IBE works. Putting an HDR on a plane functions normally ...

Would be nice to hear from NL on this one for sure.

[Forum Moderator]

Hi,

Some interesting facts:
FIRE uses a path tracer engine; calculates paths coming from the camera looking for the emitters (Luis, correct me if I'm wrong as I'm a layman in this; maybe it was the other way around).
The Production engine is a bi-directional path tracer, which means it calculates paths coming both from the camera and looking for the emitters and paths coming from the emitters and looking for the camera.

In general terms, this makes the Production engine more capable than the Fire engine.

As you have noticed, the bigger the emitters, the quicker the calculation of the caustics. That's a matter of probability, it's easier for a ray that bounces, refracts, etc, to find a large emitter than a small one.
Please, check this thread that shows an experiment about caustics seen through a transparent material while changing the sun radius: viewtopic.php?f=135&t=41876&hilit=sun+radius

There are some (dirty) tricks when dealing with refractive glasses. Not very realistic, but much quicker to render and sometimes better than getting dark areas behind transparent objects; please, check this sequence:

01 Refractive glass normal.png

1.- A refractive glass material.

02 Refractive glass hidden to GI.png

2.- A refractive glass hidden to GI (Global Illumination) > No shadows at all.

03 AGS glass.png

3.- An AGS material > No refractions but acceptable shadows and reflected caustics.

04 Refractive glass + AGS glass coincident - weird results.png

4.- Refractive glass hidden to GI + AGS glass hidden to camera and reflections/refractions. There are two coincident geometries and for this reason, we get weird and unexpected results.

05 Refractive glass (H to GI) + AGS glass NOT coincident (H to C and RR).png

5.- Refractive glass hidden to GI + AGS glass hidden to camera and reflections/refractions. This time they are NOT coincident (I shifted its position just a millimeter in each axis) and we get an acceptable result: a good-looking glass + decent shadows (although not perfect).

This can solve many situations and the render is very fast.


Regarding the lack of caustics when using IBL (and I'm guessing right now) it might have to do with the way it is treated as it's a really big emitter, usually with soft light and just a few intense areas. Most probably, the general "soft light" has a higher priority than the smaller intense spots that are supposed to generate the caustics. I bet that assigning an HDRI to a sphere with the normals pointing to the inside so it creates a virtual environment, would have a different result. I'll try to make some tests in that regard.

All the best,
Fernando