Arnold General Rendering Forum
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Diagram showing how diffuse rays are propagated in an Arnold render

Message 1 of 4
349 Views, 3 Replies

Diagram showing how diffuse rays are propagated in an Arnold render


 I was reading Arnold documentation of direct and indirect light sampling:


However, in the Diffuse Ray sampling section, I am confused with the diagram (attached) and the statement " Diffuse rays are sampled randomly over the hemisphere around the hit point."


The diagram shows one camera ray (sample, color in yellow) hitting a purely diffuse surface (I assume that is diffuse only) and further reflected in 4 different directions (color green). Now, my question is, if I shot only one sample per pixel, shouldn't the randomly directed diffuse reflected ray be also only one instead of 4? If we consider direct light sampling, then there might be a chance a shadow ray will try to hit the light (next event estimation). If I agree with this, then one randomly generated diffuse reflection ray, and one shadow ray- that make 2 rays in the figure. But it is showing 4 diffuse rays, could you please explain a bit?

Labels (1)
Message 2 of 4

That's what the Diffuse samples (in the Render Settings of whatever application you are using...since you are posting in the general forum, I won't assume) controls. By default, Diffuse (options.GI_diffuse_samples) is 2.


That means 2^2 diffuse rays are traced for every camera ray. On Arnold CPU.

On Arnold GPU, where there is no "ray splitting", there would be 1 diffuse ray per camera ray.

// Stephen Blair
// Arnold Renderer Support
Message 3 of 4

This is my first query in Arnold actually, so do not know the staff yet.


From the figure, I can see one primary (camera) ray, so I assume, there will be only one diffuse ray spawn in the scene. But those 4 diffuse rays actually confused me.


Message 4 of 4
in reply to: bipul_mohanto

There will be only one diffuse color sample getting back at that pixel from the evaluation of four diffuse samples that have been averaged (ie. added together and eventually divided by four).

Btw, the paragraph above the image is a bit misleading because is talking generally but then refers to a specific situation like a reflection ray from a mirror like surface.

At any intersection, a camera ray will be splitted into NxN secondary rays (controlled by global sample values). After that only a single ray will be fired for any other secondary intersection, being a reflection or diffuse ray makes no difference.

For completeness, reflection rays are still randomly fired from the intersection point exactly as diffuse rays however the randomness of that is spawned over the hemisphere (ie. kinda isotropic, practically cosine) where reflection rays have a different distribution around the reflection vector where this distribution is modelled over a roughness parameter where at zero is called a delta(Dirac) distribution (ie a single impulse, aka the reflection vector itself) and at 1 is almost isotropic (ie. very similar to diffuse).

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums