strange bake results

[percydaman]

trying to bake the lighting into an interior wall and am getting these strange white rings around the edges. The interior is just a box with inverted normals and a window cut out of it. When mapped back onto the geometry, the corners and edges are obviously crappy looking. Anyone have an idea? It was originally rendered 2048x2048 with pretty good irradiance map settings.

[percydaman]

ok it looks like its caused by the edge padding settings.

[n6]

Re: strange bake results

trying to bake the lighting into an interior wall and am getting these strange white rings around the edges. The interior is just a box with inverted normals and a window cut out of it. When mapped back onto the geometry, the corners and edges are obviously crappy looking. Anyone have an idea? It was originally rendered 2048x2048 with pretty good irradiance map settings.

I have sadly discovered the same thing... the problem is that the geometry need some thickness otherwise it is like rays could "cross" your geometry and appear inside the room.

I have tried 100 differents parameters (check visibility samples etc...) but it didnt help

The only solution I have found is to modelize the room's exterior geometry... which is realy anoying when you work on real time scene.

Dont know if I m clear enough

Also values>0 for the "secondary rays bias" tend to increase those visual bug while it is absolutly necesary otherwise VRay crash....

[percydaman]

Im actually trying direct comp for the first bounce and lightcache for the second bounce. Settings the edge padding to 0 and rendering is looking pretty decent.

Why cant the edge padding go outside the mesh instead of inside?

[n6]

I give it a try on my test scene.

[percydaman]

well it looks alright. It certainly pays to intelligently lay out your UVs whenever possible instead of using max's automatic flattening.

[n6]

Im actually trying direct comp for the first bounce and lightcache for the second bounce. Settings the edge padding to 0 and rendering is looking pretty decent.

Why cant the edge padding go outside the mesh instead of inside?

What did you mean by "direct comp"?
irad map, photon, lightcache or Montecarlo ?

[percydaman]

sorry monte carlo

[vlado]

The problem is that with the edge padding, you can get samples exactly on triangle edges and some GI rays may escape towards the outside of the room. It would be best to add thickness to the walls.

Best regards,
Vlado