Different Works in Progress

[FarbigeWelt]

Thanks very much for your systematic testing!
When I ran into the same problem I remembered your post, switched from Sobol to Metropolis and got OIDIN to work again.
Then I tried the random sampler and it worked as well!
Seems the scan lines are a Sobol-only issue?

Also I discovered that the problems OIDN has with volume scattering are solved by disabling Sobol.
[/quote]

Thank you much for your feedback and the completion with Random sampler. I forgot to have a look at it because I very like Metropolis. It is good to know than Random sampler is clean (stochastic well distributed) to work well with OIDN. I will use Random sampler for my current scattering volumes investigations.

[FarbigeWelt]

Before I can post some visualizations I have their quality in the way my observations are clearly visible.
Setup
World and camera volume is clear with IOR of air.
External volumes depend on objects‘ positions. The volumes are clear or homogeneous.
Internal volumes depend on objects‘ purpose. The volumes can be clear or homogeneous. Their priorities increase with every layer.
Observations
- visibility of scattered beam depends on start conditions
- visibility depends on bounce (first, in between or last bounce)
- visibility depends on render engine
- borders of scattering volume object (e.g. cube material null) must be set carefully (overlapping, separated by gap)
- external volumes must be considered carefully (same or different volumes and/or priorities)
Note
Although handling of volume is very logical there are traps.
Some observations (fading between reflections, brightness before versus between reflections) are unexpected and look like open issues.
To do next
Render and post images, especially to show open issues.

[FarbigeWelt]

Left: There is a intended gap between the two objects made from Glass-Null mix material of IOR 1.5 and 2.4.2.
Middle: The red object overlaps the blue slightly.
Right: The red objects overlaps the blue mostly.
The Right shows a strange effect like objects are mixed instead of overlapped. (All surface normals are okay. They show in a 90° angle away from the surface.)

Scattered Laser, null materials, openCL hybrid, 2800 samples

Additional Glass-Null mix material with homo. volume objects make rays visible.

Scattered Laser, homo boxes 2, Glass-Null materials, openCL hybrid, 2800 samples

Scattered Laser, homo boxes 2, Glass-Null materials, bidir 1400 samples

The Left puzzles me. This looks not as I expect it based on experiences with glass boxes and lasers. There happens this splitting of the rays and some rays are not refracted at all. Please correct me if I am wrong because I have a strong opinion that this splitting here is not physically correct.
EDIT: Meanwhile I know what is going on in the render above. The red and the blue object have a material mix of Glass and Null. The rays are split according to mix ratio. The same happens with Glass only and with opacity smaller than 1.0. Null material ignores refraction of internal volume completely. Glass renders refraction as expected but internal volume scattering is hardly visible even with reflection set to 0.0. Finally, this means I can only render what I want with n*10 kSamples.
Conclusion: Currently, If one likes to simulate correct ray casting with a laser pointing through diluted milk the only option is computing for a unreasonable time.

[FarbigeWelt]

After a dozen different set ups I found a solution close to what I wanted to illustrate. Aside of that it is the best looking image from a series of attempts.

Scattered Laser, 2, Glass rough materials, BiDir, 5600 samples, OIDN 0.9, zooms

There are three parts as described in the former post. Around two rough glass objects, with scattering internal homo. volume, is a box with null material and internal scattering homo. volume. The fine bright lines at the side of the glass objects result from the roughness. In contrast to pure glass laser beam inside the rough glass is faintly visible. Very nicely illustrated are refracting indices (IOR) and beam spreading, the latter due to roughness.

[FarbigeWelt]

In contrast to BiDir, openCL path hybrid (path plus light tracking) laser beams are not visible consistently. Unexpected is that the bright lines on the rough glass is only visible with the two separated glass objects and is not visible for the other glass objects, the overlapping ones. Also in contrast to BiDir laser beams are not visible before hitting a glass object and beams are not visible passing the rough glass objects.

Scattered Laser, 2, Glass rough materials, openCL hybrid, 5600 samples, OIDN 0.9, zooms

Actually I expected more similarities of render result of openCL path hybrid in comparison with BiDir result.

[FarbigeWelt]

Fine tuning of objects in Edit mode lead to some inexplicable observations. If an object is set Shading: Smooth in Object mode result was sometimes unexpected. If one deletes a face only and adds it again with Fill the shading of this face might be flat in some cases even if smooth shading is applied to object again.
Finally I found a solution to this reproducible but not foreseeable behavior: In Edit mode there is the feature Shading: Faces beside other shadings, i.e. edges and vertices.

Have a look at the visualization. The yellow curves point to flat shaded green faces. This faces show color has not the main influence on shading. Green and red faces are both glossy with default roughness. Spot the difference of smooth and flat faces.

Smooth or flat shaded faces

[FarbigeWelt]

Due to discussions about bump map I wondered how look like 'real' spherical bumps on a sphere.
The following pictures show influence on highlights and shadows.of a different numbers of light sources.
Light sources are suns set to distance, with radius 1. Object's material is default matte. Tone map is the same fix brightness for all renders.
Brightness is adjusted to maximum for all pictures with less than 6 lights as the only post process action.
Locations of the six lights are top, bottom, middle 45° rotated front and back - left and right.

1 light

2 lights

3 lights

4 lights

5 lights

6 lights

Did you guess highlights and shadows get this complexity? Okay there are six lights in the last scene and their hard highlights and shadows add from six positions. This sounds not exactly like a real case. Are you sure? What is with the number of fluorescent tubes in an average office. Or what is the number of spot lights in a bright shop?
The next picture shows influence of distant sun's radius on highlights and shadows.

Influences of distant Sun's Radius on Highlights and Shadows

As you may have observed bumps are best visible with small radius lights (like sun, point and spot lights) and hardly visible with large radius (like sphere, area and diffuse-indirect lights).

[FarbigeWelt]

In contrast to BiDir, see Hetero- vs. Homogeneous Volume, openCL hybrid with pgic indirect cache is capable of rendering brightness gradient in scattering homogeneous volume. Probably this is because of pgic since without indirect cache gradient is not rendered. Size of look up radius is good between 1/12 to 1/8 of volume‘s length or diameter. Is radius too small volume renders quite dark with spherical areas of different brightness inside.
Also in contrast to BiDir openCl hybrid pgic with indirect cache renders most appropriate images with asymmetry -1, the higher this value the darker the volume.

Sun, bright spot light and other spot lights

[FarbigeWelt]

A bit noisy, well it does not matter, this work is still in progress.

LASER through diluted Milk, openCL light tracking

[FarbigeWelt]

A bit noisy, well it does not matter, this work is still in progress.
LASER through diluted Milk, openCL light tracking.png

I was very glad to discover widening of laser beam inside scattering object with material Null. Unfortunately, I did not save the scene separately and after changing parameters widening effect was lost and since then I was not able to reproduce the result from last post.
I hope to regain correct settings because I had a few nice sets in mind for a short animation.