Radiosity & Reflection

I’m trying to render a scene with two spheres reflecting a big light-source above them (like neon lamps for example). For the spheres to reflect the environment I activated the Ray Mirror option in Mirror Transp Panel. Since the lamps provided in blender are invisible and therefore not reflected, I decided to try radiosity-modeling. I added two rectangular-shaped planes to the scene and made them emit light (Emit in Shaders Panel).
Now I had blender do it’s radiosity-calculations (select the spheres and the rectangles, press Collect Meshes, press GO, Replace Meshes and Free Radio Data in Radio Tool Panel / Calculation Panel).

The result is the following:

Illumination looks good, but the reflections aren’t smooth. Why is that so? Has this something to do with how blender calculates a radiosity-solution (creation of Patches and Elements, adaptive subdivision)? Is there a way to get smooth reflections?
I’ve tried to render the same scene with yafray and global illumination. I get smooth reflections there, but the spheres are all grainy. Is there another way to receive such an effect?

Thanks for any help.

Do the radiosity calculation at render time. Don’t solve the radiosity first.

Use the RAD option in the render settings.

Try that.


Actually, that is the problem. I just tested it myself.

When you run the solution and the replace the meshes, Blender reverts the smoothed mesh back to the standard poly mesh, so you get a faceted sphere which causes the bad reflections.

Run the calc at render time and all should be fine, as it was for me.


That’s the explanation, thanks. Blender really creates an unsmoothed mesh during radiosity calculation. A simple click on Set Smooth for the spheres solves my problem.

I also tried doing the radiosity calculation at render time. That also worked, but didn’t give me the exact same result.

Pre-calculated radiosity:

Radiosity calculated at render time:

I guess my actual question is: What’s the difference between solving the radiosity before rendering and solving it at render time?

Thanks a lot!


Solving before rendering allows you to perform adaptive subdivision of surfaces - Blender adds more vertices and faces where they are required to improve the quality of the radiosity. That doesn’t happen if it’s solved at rendering, so if you don’t have a dense enough mesh your radiosity can look blocky.

Solving before rendering also allows you to perform the expensive radiosity calculations once but to use the results in an animation - handy for lighting environments in a soft, realistic way.

Thanks for your explanation rawpigeon. I just found another post that says the same thing:

leon wrote: (about solving radiosity before rendering)

Can produce much better results than rendertime radiosity, since it modifies the mesh to put vertices where needed for accurate lighting…
I checked that, but the changes blender did to my scene when calculating radiosity in advance are minimal and affect only one sphere. Is there maybe another difference?