Sunlight and WorldRadius

[ThomasR]

Hello,

I'm using the sun and sky lights, and it appears that their radiance is growing proportionnaly to the world radius.

I don't understand the reason of that, why would the sun radiance double if the world radius doubles ???

Is this effect normal ? physically argumented ?

Thnaks for helping...
Thomas

[Dade]

I don't understand the reason of that, why would the sun radiance double if the world radius doubles ???

World bounding sphere is used in the computation of infinite light sources: http://src.luxrender.net/lux/file/5eb01 ... y.cpp#l300

The idea is: a larger world requires a larger amount of radiance to keep the incoming radiance on a point constant. However, I have not understood if you have found a bug or not (i.e. the absolute incoming radiance on a point varies with the world radius).

[ThomasR]

Here are a little more details, so that we can try to make everything clear...

I'm using path tracing integrator, on a very simple scene :
I'm looking at a matte plane at normal incidence with a perspective camera.

I attached my input files... I have two lightgroups : lightGroupCount=0 will be the gaussian spot light, lightGroupCount=1 the sunsky lightgroup.

I modified path.cpp in order to extract the computed values for L[1] here in the code :

path.cpp (lines 186-193)

Code: Select all

         
nrContribs += hints.SampleLights(scene, sample, p, n,   wo, bsdf, pathLength, pathThroughput, Ld, &Vd);
for (u_int i = 0; i < lightGroupCount; ++i) {
   L[i] += Ld[i];
   V[i] += Vd[i] * VContrib;
}
// My edit
sunContribution = L[1];
ofstream outfile;
outfile.open("sun_data.txt",ios_base::app);
outfile << L[1] << endl;
outfile.close();
// end


And here is the result : the average value of L[i] for the sunsky lightgroup grows linearly with the world radius...
I don't undrestand :
- why it grows
- why it grows linearly, all dependencies to the variable worldRadius are squared in the code you just sent me...

Thanks a lot for helping me understand !
Thomas

[ThomasR]

No news about this problem ?

[Dade]

outfile << L[1] << endl;

L[1] is a SWCSpectrum, I would try to output "L[1].Y(sw)": http://src.luxrender.net/lux/file/47bb4 ... um.cpp#l32

Were you assuming L[1] was an RGB or XYZ color ?

[ThomasR]

I knew that L[1] was a SWCSpectrum, I edited spectrum.h so that there is only one sample in the narrow spectral band I'm interested in :

Code: Select all

#define WAVELENGTH_SAMPLES 1
#define WAVELENGTH_START 549.5f
#define WAVELENGTH_END   550.5f

do you think i still need to use L[1].Y(sw) ?

Thanks again