How to set parameters like ab, ad

Hi list,
I am puzzled about how to set the value of some critical parameters like ab and ad. Though specific value can be refered to in the literature, higher value tends to cause the longer simulation time. Now there exists “pixel-noise” in the image and @Lars_Grobe pointed out that higher ad can solve this problem:

The parameters that I have used is mostly based on the paper Experimental validation from @Nathaniel_Jones and Benchmark of @Taoning_Wang1.

  1. If I wanna control the annual simulation time, will this pixel-noise influence the glare analysis by evalglare @Jan_Wienold?

  2. Or is there any other method to decrease or avoid the “pixel-noise” when lower parameters are used?

  3. To get a tradeoff between efficiency and quality, the suitable parameters can only be obtained from lots of test of simulation?

Any advice or experience on how to set these parameters is much appreciated!

1 Like

Hi @lee,

Are you running individual point-in-time simulations, or are you using something like the five-phase method for rendering your annual glare analysis? The parameters I came up with in Experimental validation were for point-in-time simulations using ambient irradiance caching. If you are using rcontrib, I don’t expect they will work for you, and I agree with @Lars_Grobe that you need a higher ad.

That said, depending on the settings you use in evalglare to lump glare sources, your pixel noise might not matter. I think you will need to do some sensitivity analysis of the evalglare parameters to find out if the pixel noise is an issue or not.


Hi @Nathaniel_Jones,
Thanks for your reply! I am using the five-phase method for annual glare anlysis.

I am using rfluxmtx and @Lars_Grobe thought rfluxmtx always disables the ambient cache.

This means if the settings in evalglare are not sensitive to the pixel-noise, then lower ad is acceptable.
And specific value of ab, ad … is determined by plenty of tests or your experience?

Hi Lee,

I haven’t tried it out myself, but I know that there are AI algorithms out there that can remove pixel noise quite well. A quick googling should yield a few examples. Topaz Denoiser is quite well but a paid product.

I’m saying this because computationally it could be quicker to accept the noise in the radiance simulation and denoise later.

Try this one