I received a comment regarding three-phase method, which stated that for 3-PM, the number of sky patches should be 145. Errors occur when you increase MF from m1 to m4. I cannot understand it, according to my undertanding, the more the sky patches, the closer the Reinhart sky is to the continuous sky. I don’t konow if someone publish the latest results indicated that the increase of the sky patches will caused the inaccuracy results when using three phase method. However, I can’t find it after searching many literaures. Does anyone know the similar conclusion regarding three phase method?
Very curious. It is true that 3pm can produce patchy results for higher sky subdivisions if the ambient divisions (ad) are set too low or the ray weight limit (lw) is not low enough. You need to make sure that each sky patch can be reached by multiple rays, and if the number of rays per patch is low and the patch containing the sun is hit by one more or one fewer ray than the patches surrounding it, the results will be off. Perhaps this is what your reviewer meant. This is unbiassed error, and averaged over an entire room, it is usually not a concern. I’ve never heard it mentioned as reason not to use higher Reinhart sky subdivisions.
Thanks for your insight.
My ray weight limit is 1e-5, which should be enough low. But my ambient division is 1000, which may be a little bit low. The value for the -c option is 1000, producing 1000 sample rays per Klems division.
Hi @1112 - @sarith ran some parametric studies for the two-phase method that is documented here. The parameters in play are
c for sampling.
Thank you, Mostapha! This information is very helpful!