[Radiance-general] rcontrib rtc options

[Lars O. Grobe]

Hi Nassif!

> I am trying to get stable and accuracy values using different parameters for rtrace options. I reached -ab 7 -ad 32768 -lw 2e-12.
> 
> But everytime i make higher parameters for ab or ad, the results are getting slightly higher for points with low illuminance but higher values for the ones with high illuminance. Should i proceed with higher parameters to get a saturate results?!

-ab X increases the number of bounces that will be considered in the diffuse-indirect calculation. So increasing it will lead to higher results, at any time. However, this increase will become less and, at a certain point, insignificant. It is up to you to decide when this point is reached. In a setup with dark surfaces and a lot of direct light reaching your sensor, -ab 1 may be enough, in a scene where light reaches the sensor only by reflections due to bright, diffuse surfaces, -ab 3 and more will be needed. Do a plot of the sensor value against the -ab setting to find where the increase becomes insignificant.

-ar X increases the number of sample directions when the indirect-diffuse calculation sends out random samples. Your setting is very high... and this makes the process probably very slow. Usually, -ad is set together with -as. Making use of genBSDF to replace e.g. detailed geometry (which would require to be "scanned" at a high angular resolution) in the indirect-diffuse calculation may allow you to use lower settings. Another good way to achieve quick results is to lower -ad and increase -c. This does basically the same as -ad does for the first ambient bounce (where the impact on the sensor tends to be highest). The lower -ad however speeds up subsequent bounces, and should help you decrease rendering times especially with high -ab settings. Again you should find out when results start to stabilize. This time it would be worth finding the level of noise against the -ad/-c value. Mark Stock suggested the file-size of the compressed images as an indicator for noise: http://markjstock.org/radmisc/aa0_ps1_test/final.html This works if you produce images, not single point sensors. Note that at that time, the -c parameter was not available yet, so he had to achieve oversampling by increasing the resolution. The histogram tools of Radiance may also be helpful.

I hope this somehow helps. Cheers, Lars.