I am a PHD student who is using genBSDF to generate the BSDF of a CFS to analyze DGP. For convenience, I represented the entire louver system with a single louver. However, I also compared this to a BSDF generated using the entire louver system for accuracy. Intriguingly, when applying forward tracing, the discrepancy between the two BSDFs was substantial (possible light leakage?); conversely, with backward tracing, the discrepancy was notably smaller.
These findings lead me to inquire:
What could be the reason for this significant difference in outcomes between forward and backward tracing?
Is simplifying the louver system to a single louver a valid approach for accurately assessing DGP?
The differences you see are probably due to the geometry of your blinds. As one side is bended and you use a single blind, rays at grazing angles will miss your blind gemetry and thus report a rather high BSDF value. Tracing from the other side, rays will always hit the geometry. If you use the entire louver system, this only happens for the blinds at the border.
To avoid this, the suggestion is thus to use the entire system. Additionally, using the “-dim” option (which you already did), you can avoid the border effects by sampling a representative area at the center of the entire system.
I’m really happy to get your reply! Thanks for the advice and the welcome. I’ll use the whole system and the “-dim” option as you suggested. One more question: does the size of the sampling area influence the BSDF accuracy?
The area covered doesn’t matter so much as making sure that it spans an integral number of repeating structures. You can span 1, 2, 3, etc. up to but not including the full height of your system. The same rule applies to the width. The z-dimension is simply the full depth of your system.
