Greg,
Can you confirm that Radiance still does not subdivide illum spheres during its source sampling routines? If this is still true, then no amount of "-dj" is going to help the situation that John is having. I have run into this problem where the lack of source subdivisions on spheres cause some nasty side effects. For example, I used a sphere to enclose the illuminating part of a torchiere. The reflection on the polished floor appeared as if there was a ball of fire floating a few inches above the torchiere. Other times, supporting geometry would not be illuminated correctly as it gets close to the illum sphere--similar to the problems John was describing.
So obviously, high on my wish list is adding illum spheres to the source sampling subdivision routines.
-Chas
Hi Chas,
It is true that sphere sources aren't subdivided, but sample rays to them do get jittered using the -dj option, so the main problem is one of noise, not regular artifacts. Subdividing sphere sources would be computationally expensive, and it seemed to me like the circumstances when you are so close to a sphere (illum or otherwise) that subdivision is required are rare.
-G
From: Charles Ehrlich <[email protected]>
Date: January 9, 2004 10:28:51 AM PST
Greg,
�
Can you confirm that Radiance still does not subdivide illum spheres during its source sampling routines?� If this is still true, then no amount of "-dj" is going to help the situation that John is having.� I have run into this�problem where the lack of source subdivisions on spheres cause some nasty side effects.� For example, I used a sphere to enclose the illuminating part of a torchiere.� The reflection on the polished floor appeared as if there was a ball of fire floating a few inches�above the torchiere.� Other times, supporting geometry would not be illuminated correctly as it gets close to the illum sphere--similar to the problems John was describing.
�
So obviously, high on my wish list is adding illum spheres to the source sampling subdivision routines.
�
-Chas
�
�