I am currently modelling a xenon solar simulator with a parabolic specular reflector.
Without photon mapping, the parabolic reflector does not contribute to the irradiance at the target. Therefore, I am using photon mapping to simulate the light path from the xenon source via the specular reflector to the target, which is located 4 m away.
I also want to calculate the spectral irradiance at the target. The xenon lamp has a measured spectrum, and later I would like to include more spectral properties such as the spectral reflectance of the reflector, glass transmission, and the spectral reflectance of surrounding surfaces.
I tried combining photon mapping with 24 spectral channels (-cs 24), but rtrace gives the following error:
fatal - hyperspectral photon-mapping currently unsupported
What would be the recommended approach in Radiance 6 for this type of problem?
Would I need to perform separate photon-mapping simulations for different wavelength bands, or is there another recommended way to combine photon mapping with spectral calculations?
It is unlikely that the photon-mapping extension will support hyperspectral calculations anytime soon. The changes needed are extensive, including storage and retrieval methods for hyperspectral photon maps. Roland Schregle would have to be involved, and we have been unable to secure funding for this effort, though we have tried a number of avenues.
If you absolutely need the photon map, you can fall back on traditional multi-pass methods for spectral rendering, where the R, G, and B channels are assigned different wavelength ranges in each pass. For a 24-band simulation, that would mean 8 passes, though most people use just 3 or 4 passes to get 9 or 12 bands for the sake of efficiency. You might also consider a higher spectral resolution in the blue-violet range if that’s what changes most in terms of mirror reflectance and glass transmittance.
A simpler alternative is to compute the irradiance of your system using an RGB calculation, then apply the spectral transmittance of the glass (since all rays make a single pass through it) and the mirror (since 95%+ rays reflect once in such systems) as a post-process. I think this will give you an accurate result with the least trouble and expense.
Thank you for the detailed explanation and for outlining the possible alternatives. The proposed approaches sound like useful options I will have to consider for the solar simulator. Thank you again for your help!
apologies for my very late reply, I check discourse less regularly than I used to these days. Shame on me. ;^)
Photon mapping is ideal for your use case with the parabolic reflector. But as Greg mentioned, photon mapping indeed requires some fundamental changes to support hyperspectral rendering. Beyond the (manageable) implementation issues, the lack of funding is the real show-stopper. I’m still exploring options to finally make this happen, but no ETA on that, I’m sorry to say. Plus I can only devote part-time to it alongside my regular (totally-non-Radiance) job.
So yes, unfortunately multipass rendering is the way to go with photon mapping at the moment. Fingers crossed this changes soon!