[Radiance-general] Redirecting+scattering

[Greg Ward]

Hi Federico,

It sounds to me like you are "diving into the deep end" as they say  
about new swimmers who are very ambitious.

> - As to the prismatic structure, i will use the PRISM2 material as  
> virtual source. No redirection of diffuse incoming light and no  
> dispersion are the only limitations of this material. And it would  
> make no sense to use the PhotoMap port, because these materials are  
> not implemented in this tool.  Is it right?

I don't know about support in the photon map add-on, but the prism  
material types do redirect diffuse rays as well.  They just don't  
have a diffuse component.

>  - As to the diffuser, i would intuitively treat it as an ILLUM  
> with a BRTDF function as alternate material (I´m measuring,  
> modelling and comparing different kind of diffusers) and run   
> MKILLUM.  If i understood correctly, the BRTDF materials are only  
> considered in the direct calculation and indirect specular, but not  
> in the indirect diffuse, where they are approssimate to lambertian  
> diffusers. Should i try another material?

Your information is correct as far as I understand what you are  
saying.  If you can fit your measurements to a Gaussian lobe  
distribution, you are better off using one of the native models  
(trans or trans2), as all ray paths will be followed appropriately.   
The general BRDF types, transfunc, transdata, and BRTDfunc, only use  
the directional-diffuse lobe specified for interactions with light  
sources.  General ray sampling from an arbitrary distribution is  
quite expensive, and not done in Radiance for that reason.

>  - How are PRISM materials computed in the MKILLUM calculation of  
> the window light output? Should I better try to run the simulation  
> without MKILLUM and high ambient parameters (-ab, -ad, -as)?

The diffuse portion of the window distribution through a prismatic  
glazing will be computed by mkillum, leaving the  
"specular" (directional) portion to be done during the final  
rendering.  This is the correct behavior, and you don't have to treat  
it specially with mkillum, which is helpful for considering  
contributions from the sky, ground, and neighboring structures.

>  - What if i measure the BRTDF of the complete system and i model  
> it as a normal light source? It is really difficult (or impossible)  
> to get sharp peaks in light output with  high subsampling and  
> jittering of the source?

I'm not sure what you mean, here.  If you have some way to simulate  
your exterior and measure the actual output distribution of the  
window system as a light source, then you wish to represent that as a  
brightdata modifier on a window polygon, this would work.  It's a  
very challenging measurement problem, though, which is why no one has  
done it to my knowledge.  (Anyone out there have information on this?)

The purpose of mkillum of course is to create a light source from  
your surface(s) by simulating an outdoor situation.  It doesn't  
always work, though.  Specifically, mkillum has trouble resolving the  
contribution from the sun through curved, specular reflectors and  
devices.  It can handle most everything else.

>  Any other suggestion to simulate this system correctly  
> (quantitative)?

Measure and validate.  Repeat.

-Greg