Ray Visibility (glossy, diffuse, transmission) on light sources generate a different result when using hit by the BSDF or next-event estimation #103977

New Issue

Sebastian Herholz · 2023-01-18T18:53:51+01:00

Sebastian Herholz commented

2023-01-18 18:53:51 +01:00

System Information
Operating system:
Graphics card:

Blender Version
Broken: (example: 2.80, edbf15d3c0, master, 2018-11-28, as found on the splash screen)
Worked: (newest version of Blender that worked as expected)

Short description of error
When changing/disabling the ray visibilities on a light source (glossy, diffuse, or transmission), the result becomes
inconsistent when using only BSDF sampling, next-event estimation, or MIS between both.

The reason for that is that the contribution of a light source is completely neglected when the type of the BSDF closure used for sampling the direction is disabled on the light source.
During next-event estimation, on the other hand, the light source ALWAYS contributes. Only the BSDF evaluation components multiplied with the light contribution differ [ eg. instead: Lglossy + Ldiffuse -> L*diffuse if glossy is deactivated].

The example shows a scene with one big rectangular light at the ceiling, where glossy rays are disabled on.
As you can see, the BSDF image is darker compared to MIS, and the NEE version is a little bit brighter.

The MIS image is more or less a blended version of BSDF and NEE.
Because both BSDF and NEE do not converge to the same result, the MIS one is inconsistent/biased.
Since the blending factors/mis weights depend on the sampling strategies, the result changes when using path guiding or the light tree.
This makes these optimized strategies unpredictable and can confuse the artists.
In the worst case, they report bugs on for the light tree and path guiding.

A WIP fix is proposed in D16965

Exact steps for others to reproduce the error
Based on the default startup or an attached .blend file (as simple as possible).

**System Information** Operating system: Graphics card: **Blender Version** Broken: (example: 2.80, edbf15d3c044, master, 2018-11-28, as found on the splash screen) Worked: (newest version of Blender that worked as expected) **Short description of error** When changing/disabling the ray visibilities on a light source (glossy, diffuse, or transmission), the result becomes inconsistent when using only BSDF sampling, next-event estimation, or MIS between both. The reason for that is that the contribution of a light source is completely neglected when the type of the BSDF closure used for sampling the direction is disabled on the light source. During next-event estimation, on the other hand, the light source ALWAYS contributes. Only the BSDF evaluation components multiplied with the light contribution differ [ eg. instead: L*glossy + L*diffuse -> L*diffuse if glossy is deactivated]. The example shows a scene with one big rectangular light at the ceiling, where glossy rays are disabled on. As you can see, the BSDF image is darker compared to MIS, and the NEE version is a little bit brighter. ![DarkInterior01-vf0.png](https://archive.blender.org/developer/F14172396/DarkInterior01-vf0.png) The MIS image is more or less a blended version of BSDF and NEE. Because both BSDF and NEE do not converge to the same result, the MIS one is inconsistent/biased. Since the blending factors/mis weights depend on the sampling strategies, the result changes when using path guiding or the light tree. This makes these optimized strategies unpredictable and can confuse the artists. In the worst case, they report bugs on for the light tree and path guiding. A WIP fix is proposed in [D16965](https://archive.blender.org/developer/D16965) **Exact steps for others to reproduce the error** Based on the default startup or an attached .blend file (as simple as possible).

Sebastian Herholz commented

2023-01-18 18:53:51 +01:00

Added subscriber: @sherholz

Sebastian Herholz commented

2023-01-18 19:03:49 +01:00

Added subscribers: @LukasStockner, @Christophe-Hery, @brecht, @Stefan_Werner

Pratik Borhade commented

2023-01-19 05:59:33 +01:00

Added subscriber: @PratikPB2123

Pratik Borhade commented

2023-01-19 05:59:33 +01:00

Changed status from 'Needs Triage' to: 'Needs Developer To Reproduce'

Pratik Borhade commented

2023-01-19 05:59:33 +01:00

Hi @sherholz, it'd be really nice if you fill out steps to reproduce field from the bug report template.

Hi @sherholz, it'd be really nice if you fill out `steps to reproduce` field from the bug report template.

Christophe Hery commented

2023-01-19 07:02:35 +01:00

On first run through the code, the changes make sense to me.
There are always complications and subtleties stemming from multi-lobe MIS and LPEs (here ray visibility flags).

On first run through the code, the changes make sense to me. There are always complications and subtleties stemming from multi-lobe MIS and LPEs (here ray visibility flags).

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Ray Visibility (glossy, diffuse, transmission) on light sources generate a different result when using hit by the BSDF or next-event estimation #103977