[Radiance-general] Higher surface reflectance simulation in buildings

[Greg Ward]

The higher the surface reflectances, the more bounces you need to  
compute.  In pathological cases where you have all white surfaces  
facing each other in a closed environment, you may need to disable  
the ambient cache (-aa 0) and rely on Russian roulette sampling (-lr  
-20 or so).  To get a large -ab setting to work (-ab 20 or whatever),  
you will also need to decrease the -ad setting to make this tractable  
(-ad 64 or so with -as 0).

-Greg

> From: Jiangtao Du <jiangtao.du at yahoo.co.uk>
> Date: November 8, 2008 9:45:48 AM PST
>
> Dear All,
>
> I am doing Radiance simulations in atrium well with higher surface  
> reflectances ( 0.6 ~ 0.8 ~ 1, diffuse surface).
> But according to   "Aizlewood M, Butt J, Isaac K & Littlefair P.  
> Daylight in atria: a comparison of measurement, theory and  
> simulation:  Proceedings Lux Europa. Amsterdam, 1997."   and     
> "Littlefair P. Daylight prediction in atrium buildings. Solar  
> Energy, 2002; 73: 105-109.",    Radiance underestimated the  
> reflected light (Internal reflected component) in atria with high  
> reflectance (some measurements in the paper used a reflectance 0.85  
> and got the biggest discrepancy in IRC was 4.87). My some  
> simulations  got the similar results: simulations underestimated  
> the measurement with reflectance 0.6 and 0.8 and  the higher of  
> surface reflectance the bigger is the discrepancy. All the studies  
> used overcast sky model and set a higher ambient parameter setting.
>
> Could you help me explain the divergence from high reflectance  
> simulation using Radiance. Also, if these demonstrate Radiance  
> simulation could bring some bigger errors to the dayligthing  
> simulation in high reflectance environment?
>
> BTW, I will appreciate it very much if you could share some  
> experience or knowledge  and give some suggestion about higher  
> reflectance simulations.
>
> Cheers
>
> Jiangtao
>
> PhD student in University of Sheffield