Regression: Lighting difference between 3.5 and 3.6 with Principled BSDF #109494
Labels
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset System
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Viewport & EEVEE
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Asset Browser Project
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Module
Viewport & EEVEE
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Severity
High
Severity
Low
Severity
Normal
Severity
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
3 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: blender/blender#109494
Loading…
Reference in New Issue
Block a user
No description provided.
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
System Information
Operating system: Windows-10-10.0.19045-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3080/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 531.61
Blender Version
Broken: version: 3.6.0, branch: blender-v3.6-release, commit date: 2023-06-27 08:08, hash:
c7fc78b81ecbWorked: 3.5.
Short description of error
I'm observing a difference in lighting on some surfaces, and I'm wondering if it's expected - or whether I need to investigate further.
I'm guessing it could be related to this.
"Improved Fresnel handling of the Glass BSDF for better energy preservation and accuracy of results at high roughness."
https://archive.blender.org/developer/D17149
My surfaces are high roughness, but they are using Principled BSDF with a fairly standard setup. One tree is using displacement.
The lighting is harsh, one Sun with an angle of 0.5 degrees, strength 2.5, and another Sun with an angle of 5 degrees and strength 0.25. (Note I've turned off the fog, and the difference is still observed.)
Exact steps for others to reproduce the error
I'll attach some images. You probably need to load the images in separate windows so you can switch between them.
Please advise if you need me to investigate and break it down into a simple test scene.
(If this is just an expected difference for Principled BSDF - then that's fine.)
The main difference here is the upper part of the trunk on the right:
This one the tree on the left is the difference:
In both cases I guess it's the glancing angles of the light that are darker in 3.6.
I should mention I'm not using Light Trees - and yes this is with Cycles.
#109404 also included some changes to lighting. #108996 might be related, but considering you have such a low angle, it's not that likely.
To me the result seems more or less passable.
I don't disagree with that.
However, I've done some more research and created a test file attached that highlights the differences.
Issue 109494 DisplacementBumpOnlyLighting.blend
The scene has a cylinder for the tree trunk, a moon (icosphere), and a sun light where the moon is. (The moon is only to indicate the position of the light and does not illuminate or cast shadows - it can be deleted.)
When I created the cylinder I made a mistake that actually accentuates the problem. I turned on Smooth, but forgot to set an Auto Smooth angle. Hence the cylinder looks like this:
So, rendering the attached file, generates this image:
Blender 3.6 is brighter, and the lighting wraps further around the cylinder.
Blender 3.5 appears to not have smoothing applied - you can clearly make out the faces of the cylinder.
Blender 3.6 looks much better, though it is perhaps questionable that the light should wrap around that far - but then the normals are messed up due to not using auto-smooth.
With regard to the lighting difference, sampling a random pixel on the brightest part of the cylinder shows the luminosity as 0.4323 (Blender 3.5.1) and 0.4819 (Blender 3.6.0) - about 11% brighter.
Now, what's special about this file is the material has a Noise Texture applied to the Displacement. However Displacement is configured as Bump Only.
I'll disconnect the Displacement node from the Material Output:
The two images are much closer.
The Blender 3.6.0 image is about 4% brighter (sample of 0.4829 vs 0.4649).
The main difference is just the top of the image where the smoothing/shadowing appears to be handled differently.
So, what if I use a normal map, instead of the Displacement/BumpOnly thing? I've loaded a normal map from Poly Haven (bark_brown_02_nor_gl_2k.exr), but you should be able to use anything.
The only difference is a 13% difference in brightness (0.4651 vs 0.4102).
Finally, going back to the original configuration with the Displacement Noise texture (no Normal Map). I will turn on Auto-Smooth at 30 degrees.
The only difference is an 8% difference in brightness (0.4428 vs 0.4114). Now only the first face of the cylinder is illuminated, instead of the light wrapping further around.
Anyway, I guess I'll leave it at that.
I'm not concerned specifically about the change in brightness. I'm assuming the approach now is more accurate.
While the smoothing applied to the (capped) cylinder was a bit dodgy. It still could be technically valid in some circumstance. There's definitely something a bit strange happening - but only when using Displacement. Note that Displacement Only doesn't do much in this case because there's no geometry, but the same anomaly is seen with Displacement+Bump. And I would certainly agree that Blender 3.6 looks better than Blender 3.5.
(Also FWIW, the tree trunk in the original scene used an un-capped cylinder - so was not subject to any strange smoothing normals.)
Lighting difference between Blender 3.5 and Blender 3.6 with Principled BSDFto Regression: Lighting difference between Blender 3.5 and Blender 3.6 with Principled BSDFRegression: Lighting difference between Blender 3.5 and Blender 3.6 with Principled BSDFto Regression: Lighting difference between 3.5 and 3.6 with Principled BSDFThanks for investigating this. This is due to intentional changes to bump mapping from #105776. That was not mentioned in the release notes, I've added it now.