[Radiance-general] Modeling Paradox

[Carsten Bauer]

Hi,

I remember it, I was at this point, too, some years ago in my life. Its 
one of the oddities of Radiance, having a powerful rendering engine but 
- at the first glance - no 'convenient' way to set up / put in a model. 
This is no complaint, consider that the stuff is for free, and you 
already get a rendering engine more or less surpassing many commercial 
packages in terms of quality and being currently accepted as the de 
facto standard for accuracy in terms of lighting analysis.

I once thought about writing some import module for Radiance geometry, 
but soon made up my mind. It probably is not very difficult, but it 
needs a lot of time and tedious work, and after all you'll end up 
reinventing another CAD program, although there are lots of them already 
on the market.. Additionally, with modern PC equipment and their high 
amount of RAM, treating geometry as polygon-meshes is not so much a 
limitation as it has been before (I remember that POV ray uses lots of 
different primitive types, for the more complex ones you need 
sophisticated root solvers to determine ray intersections, meaning 
longer calculation time)

Nevertheless, so far I worked exclusively with an editor and the 
Radiance native formats (Hi Rob, you're absolutely right, with one 
exception of course, I'm insane, but I'm not masochistic, so I don't use 
vi.... :-) . But even when working with CAD programs, its good to rely 
on exact coordinate specification rather than on mouse-drawing, to avoid 
inaccuracies like accidental small gaps between edges resulting in light 
leaks or other artifacts which need a lot of time to get tracked down 
later.

Another Tip: decompose your model in separate items as much as possible, 
and let the Radiance input engine put them together when calling oconv, 
the simplest means are often the most effective ones.

-cb