Spectral path tracing

[Rob Burns]

I've started playing with the BDPtracer in vray and it's opened up a whole new world of info about unbiased renderers and some of the techniques they use. One thing that I've always been curious about is the spectral sampling some rendering engines use. I'm a real noob to this area of rendering so please excuse me if this has already been discussed a lot. From what I understand using spectral sampling instead of RGB for samples can lead to more realism in renderings and that certain phenomena just aren't possible to simulate without it (like iridescence and thin films).

What I'm wondering is if there is a way to use or enable spectral sampling in Vray or if it will be used in the future? What are the main advantages of Spectral sampling - would it really make a difference in the look of architectural renderings to use it (even if it is subtle)? Also does using spectral vs RGB sampling necessarily mean longer render times? I would assume yes since instead of sampling at 3 color wavelengths (RGB) you need to sample at many more wavelengths and it seems like for certain materials there are spikes in the absorption and IOR at certain light wavelengths and those spikes would likely need extra samples.

[vlado]

What are the main advantages of Spectral sampling - would it really make a difference in the look of architectural renderings to use it (even if it is subtle)?

No difference whatsoever, unless you have dispersion effects.

Also does using spectral vs RGB sampling necessarily mean longer render times?

It depends on how its implemented. One possible solution is to use more wavelength samples, so instead of 3 components, colors would have f.e. 7, 11 or more components. In that case, processing these components will require extra calculations, but on the other hand, each camera ray will compute all those components in one go. In many cases though, these extra calculations are just wasted, as very few material effects require spectral sampling and for many images, there won't be any difference at all in the final output. Another solution is to assign just one wavelength to each camera ray, but to vary the wavelength from one camera ray to the next. In that case, each ray has just one color component, but that means that you would need more camera rays in order to clean up the color noise that you get. V-Ray uses a hybrid approach, where normally camera rays compute the three RGB components in one go, but for dispersive materials, each dispersion ray is assigned a single wavelength from the point of split onward.

Best regards,
Vlado

[Rob Burns]

great, thanks for the reply, good to know we're not missing out on much