# cross check with rcontrib -V+ -I

Hi

I did a simple test to know if I understood correctly what "rcontrib -V+
-I" computes.

I created a scene with a light, the shape of a ring of radius *r*.
void light windowglow 0 0 3 1 1 1
windowglow ring window_0
0 0 8
0 0 0
0 1 0
0 *r*

And I run the following command :
echo 0 *y* 0 0 -1 0 | rcontrib -V+ -I -st 0.15 -dj 0.7 -ds 0.15 -dt 0.05
-dc 0.5 -dr 3 -dp 512 -ab 2 -ad 4096 -ar 128 -aa 0.15 -av 0 0 0 -as 1024
-lw 1e-9 -m windowglow -n 4 scene.oct

By hand I computed that the resulting coefficient to transform the constant
radiance of the source (1 W/m2/sr) to the irradiance at the point (looking
towards the source) should be *pi * sin(a).^2* where *a = atan(r/y)*

So I do a variation of the angle *a*. I keep *y* constant (10m) and I get
this results for 2000 points.

What is wrong with the simulation or my calculus ? Why do the points spread
out at large angle ? Why do the coefficients becomes small at very large
angle ? Does the source follow the Lambert's cosine law ?

Sorry for my english,

Mario Geiger

try -dj 0
it may be that
G

···

On 20 Aug 2014, at 14:22, Mario Geiger <[email protected]> wrote:

0.7

Mario,
I think 'G' is right, -dj 0 should fix it.
Also, for large area sources it is better to use the glow instead of light
material partly because of the same sampling problems you're experiencing.
But also because if there was any geometry between your sampling point and
large area light source the results would be crazy unstable.
Andy

···

On Wed, Aug 20, 2014 at 1:30 PM, G <[email protected]> wrote:

try -dj 0
it may be that
G

On 20 Aug 2014, at 14:22, Mario Geiger <[email protected]> wrote:

0.7

_______________________________________________