Hi I’ve been doing some box model testing with a daylight factor analysis and have found that the DF result is inconsistent between runs with the exact same settings. For example, I am changing only the shade depth between 4 distances and record the DF for each scenario. Then I re-run the analysis and the results changed quite drastically for some depths, and remained for others (92-76%). I have been re-running the analysis many times now making sure only to change the shading depth using the same 4 distances, and the inconsistent results persists. Usually the result differs only by a couple % points, but every now and then it will jump by 10% or more.
The inconsistency of a couple % points is something I have observed in the past (run analysis, recompute grasshopper, slightly different results), but didn’t worry much. A difference of 10% is quite significant though when the settings are identical, so would like to understand the cause/solution for this.
I am using v1.9/8 LBT components, but my recipe is 1.9.0.
I have tested this with the DF recipe and the PIT recipe with a 10000Lux uniform sky and I’m getting the same type of inconsistency.
Thanks for your response. I did a bit more testing to try figure out what the issue was.
The radiance parameters I had set for those results were:
aa - 0.1, ab - 3, ad - 2048, ar - 64, and my analysis grid has 400 points (10x10m room with a 0.5m grid size).
I’ve taken a look online (SETTING RENDERING OPTIONS) and these settings seem to be within the band for “accurate” results or better, so I’m still unsure why this is happening. Regardless, using the above link, I’ve set all my rad par to equal the accurate column (-aa 0.15 -ab 2 -ad 512 -ar 128 -as 256 -dc 0.5 -dj 0.7 -dp 512 -dr 3 -ds 0.15 -dt 0.05 -lr 8 -lw 0.002 -ss 0.0 -st 0.15) and the results are consistent now, but considerably lower than previously.
Further to that, I have continued to explicitly set all of those rad par and changed only the ab to 3 brings the results back up to what I’d expect, and more importantly the results remain consistent.
If I now go back to my original situation where I only explicitly set the rad par that were called out as defaults in the PIT recipe (ab, aa, ad, ar), I also had only increased my ab from the default of 2 to 3, but that resulted in the inconsistent results I flagged at the start.
My question is what the other rad par are set to by default in the PIT recipe (if they are), and I’m curious about the rationale behind the default values (-ab 2 -aa 0.1 -ad 2048 -ar 64) as these seem to be source of inconsistent results?
Try aa -0.05, -ab 6 -ad 2048 -ar 2048. The direct parameters like -dc -dj and -dt etc are irrelevant since there are no direct light sources to deterministically trace shadow rays to (you mentioned you are running DF calcs).
There is no one-size-fits-all solution for stochastic simulations. Assuming your ceiling is less than 4m and you have a window in only one direction, it is likely that the points deeper into your room are actually not able to “find” the sky.
