Hybrid Back/Forward path tracing (aka BiDir without MIS)

[Dade]

I have a scene that represents a common lighting scenario.
A row of LED illuminates a volume diffuse, but mostly polished, lens.
All this is built in a housing and closed by a clear cover lens.

It is frankly a scene to cryptic to analyze without knowing exactly what is supposed to be. Try to reduce the problem to a test scene easy to check (a single light source, not 16, and less volumes and materials as possible).

I have no idea of what is supposed to be and how it is supposed to look.

[Sharlybg]

Any possibility to see caustics in the viewport (No Bidir of course light tracing).

[Dade]

Light tracing doesn't work with Camera clipping plan :

I should have fixed the problem.

[Dade]

Any possibility to see caustics in the viewport (No Bidir of course light tracing).

The support for BiDir pre-view should cover that topic

[Sharlybg]

The support for BiDir pre-view should cover that topic

But light cpu will help in complexe and heavy scenario case as it converge faster than Bidir.

[FarbigeWelt]

The support for BiDir pre-view should cover that topic

But light cpu will help in complexe and heavy scenario case as it converge faster than Bidir.

Hmm, and why do you think that?

[Sharlybg]

Hmm, and why do you think that?

I mean that the hybrid approach is already faster in final frame rendering. So why it will be slower in the viewport ?

[provisory]

I mean that the hybrid approach is already faster in final frame rendering...

It seems to me too (compared on CPU only).

(But unfortunately the result isn't the same.)

[lighting_freak]

Hi

I have no idea of what is supposed to be and how it is supposed to look.

Let's assume it's a car's tail lamp, already simplified a lot!
The complete surface of the diffuser inner lens should glow, like in the BiDir example (this is the correct result).
With less noise and more details visible some tiny inhomogenties may occur.

That's exactly what the rendering result should show...
That's, by the way, exactly the way I'm used to use a physics based rendering system... it not tuning an image until it looks like I want it.

BR

[epilectrolytics]

I have a scene that represents a common lighting scenario.
A row of LED illuminates a volume diffuse, but mostly polished, lens.
All this is built in a housing and closed by a clear cover lens.

Do you believe I'm doing something wrong, or is CPU only not dead yet?

I guess it's an SDS-path issue.
Remember, BiDir can render SDS-paths with mesh lights like in your scene, albeit slowly.

But light tracing cannot.

Light tracing works like this:
A ray originates from a light source, hitting some object in the scene.
If it is a specular surface, the ray will be reflected (or split and refracted too) according to optical laws and continue until it hits another surface.
If it is a diffuse surface, a test will be made if the camera is in line of sight and if so, the shading colour of the surface will be transmitted to the corresponding pixel.

This means that light tracing can only connect to the camera if the last hitpoint was a diffuse surface, all specular surfaces will appear black because the light rays from there have to follow optical laws and will never hit the camera which is an infinitesimal point.

In your scene a clear glass covers the whole front of the device and shuts off all light rays to the camera.

Edit: I guess scattering volumes work like diffuse surfaces too.