Unconsistant daylight results using point in time daylight recipe

Hi Pollination team,

I did a daylight factor study with the point-in-time recipe to simulate the effect of different glazing ratios on daylight. However, the results seem to be inconsistent sometimes! For example, as you can see in the images below the results show more light penetration for the variant with 50% WWR compared to the 60% one.


When I reran the same 50% variant with the same model that I uploaded on the pollination, the results changed very much
showing lower penetration of light as you can see below.

I know that a 10% difference in radiance results is expected but this was more than 40% difference.
What could be the reason?

My radiance settings were: -aa 0.2 -ab 5 -ad 4000 -ar 128 -as 4096 -dc 0.75 -dj 1.0 -dp 512 -dr 3 -ds 0.05 -dt 0.15 -lr 8 -lw 0.001 -ss 1.0 -st 0.15

Hi, @hshahriari - Thank you for documenting this. Can you send us the run ids so we can have a closer look at the inputs and outputs?

Is it possible that you have a large context in the model? For the size that I see you need a lower value for -lw and a higher value for -ad. That said, the difference looks larger than expected. The first step is to compare the inputs and make sure they match. We can start from there.

UPDATE: I missed that this is a daylight factor simulation and not an annual daylight simulation. The -aa and -ar values are the important ones, in this case, having a large scene. @mikkel will have a closer look and will let you know.

Hi @hshahriari,

I tried to recreate the inconsistency with a small room, but with the same context from your model, and your Radiance parameters. Ideally, you should increase -ar and/or decrease -aa, but I don’t think those settings are causing the issue here.

Is the simulation of the last image run locally in Grasshopper or on Pollination?

1 Like

Hi @mikkel ,

Thanks for taking a look at it.
The last image I did locally.

I was a bit in a hurry when writing that answer, so I left out why I initially thought the above. Anyway, I now managed to recreate the issue, and I think -aa and -ar may very well be the issue - perhaps I made a mistake in my tests earlier.

With your model size, and -aa 0.2, I would try -ar 330 based on the calculation by Ambient Resolution*. Or -aa 0.1 and -ar 165. I got consistent results with either of these -aa/-ar combinations. For the other parameters I used lower settings than those you shared in order to speed up the simulation.

image

*The calculation is based on formula 6.12, Chapter 6: Daylight Simulation (by John Mardaljevic) in Rendering with Radiance. The whole chapter is by the way definitely worth a read.

2 Likes

Hey @mikkel,

Thanks so much for taking a look at it. I will read the chapter and will try the study with the new settings and will update you.

Thanks

You should also set lw to slightly above 1/ad

Typically I’d set ad to 4096 and lw to 0.00025 or 0.0003. setting lw too high will make simulation faster but more inaccurate.

/Mathias Sønderskov

1 Like