Dielectric Over Light Source


I’m pretty new to Radiance and I stumbled on a problem I probably should have predicted, given more experience…

So I have a scene where the light source is behind a curved dielectric medium, and both are in front of a plastic wall. The source has a very narrow spatial extent, almost like a point source. Now what happens is that a ray will start tracing from the camera, hit the wall, but will never reach the light source. This makes sense since it’s not possible to do any direct calculation this way - there’s only a very specific direction in which the ray will actually reach the source.

This seems like something that probably was encountered by others and should have a solutions, so I’d be very grateful for any assistance here.


You can use the photon-mapping method, which includes the ability to throw caustics in an environment. There’s rather a lot there to get into for that, and I’m not the expert. I know you can achieve some very nice (and accurate) results with it, given some time and effort.

Thanks I hope I’ll get a chance to look into this.

In the mean time, I’m having some more problems with this glass slab:

I’m running rtrace with -lr 1, and expecting to see a single reflection/refraction ray, but I’m actually seeing from the report that the ray travels back and forth inside the glass about 6/5 times before being refracted out.

Any ideas why that might happen?


That doesn’t really seem possible. The -lr option with a positive number is a hard limit on reflections. Are you sure you aren’t over-counting refracted rays, or shadow rays, somehow? If you send me the results of one ray query through the material using rtrace -lr 1 -otdnM, I might be more help.

After investigating further, indeed what I’m seeing is multiple refractions.

Turns out that the object I’m using is actually composed of three groups, and I define all of them as dielectric.

It looks like radiance decides on wether it goes from inside/outside of the material according to how the normal is oriented. Now, one of these groups, has an outer/inner surface with the same normal direction, which means Radiance thinks it goes from air to dielectric when it enters the material, and then again from air to dielectric when it leaves the material, which means that it’s like a dielectric shell with an air bubble inside of it. Does that make sense?

You should really use the “interface” type for a surface separating two dielectric media if neither of them has a 1.0 index of refraction (like air).

Would it make sense to use mkillum here? Perhaps it’s possible to replace the curved glass with a secondary source?

The docs say it’s not appropriate for directional sources, and my source is basically a point source (although with a wide angle), so I’m not sure…

The difficulty with mkillum is the same as the difficulty in the interior, namely finding the light source through your curved refractor. If your light source is large enough, and you turn it into a “glow” type, and send a LOT of samples with mkillum, then it may work, but not better than photon-mapping with caustics.