Getting numerical errors while using secondary light sources...

Dear friends,

I'm trying to make use of advantages of the "secondary" light sources
to speed up the indoor irradiance calculations.

In the test below we have compared a daylight factors in a room,
calculated along the central depth axis of the floor.

The room parameters are:
3.2 meters - width,
5.6 meters - depth,
2.5 meters - height.

The window is 1.92 meters width and 1.5 meters height 0.8 meters
altitude (above the floor).
The room is located in the building having external obstruction
(another building) on the opposite side of the street.

The uniform sky model was used.

I run Radiance (3.7.2 and 3.7.3) with different parameters and here
are the most valuable of them:
http://www.audice.com/radiance/Graphs_3.png

Where:

"Radiance Illum 1" and "Radiance Illum 2" - the scene was divided into
an outdoor and indoor parts. At the first pass the distribution of light
from the window was calculated in the outdoor scene. At the second
pass the daylight factors were calculated in the indoor scene.

For window material I used:

void illum window_mat
0
0
3 0 0 0

I didn't use "glass" material, because I needed rather empty opening
than the glass.

"Radiance Illum 1"
Parameters I used for mkillum:
-ab 4 -aa .2 -ar 128 -ad 512 -as 256 -dp 256 -ds .15 -lr 8

Comment in scene .rad file:
#@mkillum i=window_mat d=64 s=24 m=wind_illum f=./dist/window c=n b=0

Parameters I used for rtrace:
-ab 8 -aa .1 -ar 128 -ad 512 -as 256 -dp 256 -ds .15 -lr 8

···

--
"Radiance Illum 2"
Parameters I used for mkillum:
-ab 8 -aa .1 -ar 128 -ad 512 -as 256 -dp 256 -ds .15 -lr 8

Comment in scene .rad file:
#@mkillum i=window_mat d=256 s=128 m=wind_illum f=./dist/window c=n b=0

Parameters I used for rtrace:
-ab 8 -aa .1 -ar 128 -ad 512 -as 256 -dp 256 -ds .15 -lr 8

--
"Radiance Standard" - no illum surfaces were used. The model was not
divided into indoor / outdoor parts so the interreflection calculation
was used to compute the contribution from the window.

rtrace parameters used:
-ab 8 -aa .032 -ar 512 -ad 2048 -as 1024 -dt .05 -dr 8 -dp 2048 -ds .1
-dc .6 -lr 8

--
What causes more than 6% error for "Radiance Illum 1" and "Radiance
Illum 2" in the "pink" zone? Is there any way to get more accurate
results using secondary light sources?

--
Best Regards,
Ilya A. Zimnovich
mailto:[email protected]

Hi Ilya,

My guess is that you are seeing sampling areas associated with your exterior obstruction, which will be translated into the window's output distribution by mkillum as a fairly sharp cut-off since it is on the other side of the street. Basically, I recommend reducing your -ds setting even further (to 0.1 or 0.08) in your final rtrace, in hopes of reducing the parallax error of the window sampling. Even better would be to set -dj 0.9 and give the same point to rtrace many times, then average the result in a post-process. This will eliminate any direct sampling errors, which I suspect are the problem in your particular case.

-Greg

P.S. Very nice analysis and presentation of your results, by the way.

···

From: [email protected]
Date: December 16, 2005 11:18:10 AM PST

Dear friends,

I'm trying to make use of advantages of the "secondary" light sources
to speed up the indoor irradiance calculations.

In the test below we have compared a daylight factors in a room,
calculated along the central depth axis of the floor.

The room parameters are:
3.2 meters - width,
5.6 meters - depth,
2.5 meters - height.

...