refraction attenuation

[fathom]

i haven't been able to crack this one in pbr -- aside from hacking the pathtracer.  anybody have a workflow in place to do atmospheric attenuation on refraction rays without having to add it everywhere you want to see if have an effect?

 

in micropoly or classic raytracing (non bsdf) you can use atmosphere shaders to do a one-time calcuation based on the incoming I and then tint the resulting Cf.  but since atmosphere shaders can't affect F, you're sol in pbr.  and since the standard mantra library is now using pbr for 2ndary rays, even raytracing fails.

 

my latest attempt has been to use a uniform volume with a colored opacity.  that works great for generating the correct shadows from lights, but loses color in refractions.  i recall colored opacity is not fully functional in pbr, which is a shame.  i guess i'm bumping into that here.

 

i'm really dying to figure this out, but i'm guessing it's a lost cause.  unless maybe somebody here has some insights?

 

 

[dennis.albus]

I didn't test this but isn't there a builtin solution for correct attenuation in H15?

 

As you said the problem is that we don't have access to I or at least the length of the incoming ray in the shader. This information is only accessible in the integrator.

In H14 my solution to this has also been to hack the pathtracer. I know it is not recommended, but as mine is so customized anyway I don't have a problem with that.

 

Cheers,

Dennis

[fathom]

the only thing i've seen in 15 is that they've broken up the surface model into a more modular approach and have created vops to compartmentalize.  so there's now a an "attenuate" vop but it's just the stuff from the surface model.

 

i'm curious in your hacked pathtracer, do you hijack the installed tracer or do you point your rops to your custom one?  that's an aspect of houdini i haven't felt like jumping into.

 

it occurs to me that i think i worked out a means to make things work in vanilla houdini by changing to the recursive pathtracer which computes Cf instead of F.  then i can use fog shaders again.  i'll have to look around for that approach again...