[Radiance-general] Cutting a cylinder, or modeling smooth geometry...

[Lars O. Grobe]

Hi,

I am running into problems due to the lack of csg in radiance. I need to 
accurately model a tube with a very specular material, which is bent. 
The real tube is produced by cutting a cylinder under 22.5 degree and 
assembling it, leading to a nice 45degree angle with one ellyptical 
edge. As I am not able to cut a cylinder geometry in radiance, I have 
modeled it as a mesh, but the specularity leads to visible edges.

I have been sonsidering several solutions so far:

1) Using antimatter to cut away parts of the cylinder would probably not 
work, as the second cylinder, which also has to be cut, would lead to 
two overlapping antimatter volumes.

2) Refining the mesh would result in huge geometry, as the segments 
would become thinner while keeping their length, this would lead to a 
real worst case for oconv. Also to make the edges disappear, I would 
have to get into dimensions where one facet should not take much more 
then one pixel in the final rendering.

3) Using "smooth" surface normals would lead to an visual improvement. 
But I have my doubts if these would be considered for photon mapping and 
especially caustics? In my setup, the pipe ends with a diffusor 
(currently modeled as a box modified by a rather generic trans 
material). I would expect caustic effects to become visible caused by a 
round tube geometry with very high specularity.

4) The last option I see is using a cylinder object with a mixfunc 
modifier and a simple cal file containing just the line

condition=if(Pz-Px,1,0);

to cut my geometry in 45 degrees, using void to get parts of the 
geometry invisible. Would that work with overlapping geometries (which 
would still exist, even if invisible), and would it be relieable with 
the pmap extension? The main disadvantage here is that I would have to 
maintain an extra set of geometry for radiance, while I could use 
converters to import from other programs so far.

CU Lars.