Cycles: Principled BSDF: materials with Sepcular within close proximity to 12.5 are completely black #96719

New Issue

tamooz br · 2022-03-22T22:55:24+01:00

tamooz br commented

2022-03-22 22:55:24 +01:00

System Information
Operating system: Arch Linux with kernel 5.16.15-arch1-1.
Graphics card: Intel HD graphics 5500.

Blender Version
Broken: at least since the release of 2.79.
Worked: unknown, potentially never.

Short description of error

If the "Specular" value of a Principled BSDF is set to a value within a small margin of error from the value 12.5, then no specular reflection will occur.

Exact steps for others to reproduce the error
Set the "Specular" value of a Principled BSDF shader to 12.5, and render in cycles.

**System Information** Operating system: Arch Linux with kernel 5.16.15-arch1-1. Graphics card: Intel HD graphics 5500. **Blender Version** Broken: at least since the release of 2.79. Worked: unknown, potentially never. **Short description of error** If the "Specular" value of a Principled BSDF is set to a value within a small margin of error from the value 12.5, then no specular reflection will occur. **Exact steps for others to reproduce the error** Set the "Specular" value of a Principled BSDF shader to 12.5, and render in cycles.

tamooz br commented

2022-03-22 22:55:24 +01:00

Added subscriber: @Koopa512

tamooz br changed title from ~~Cycles: Principled BSDF: metallic materials with Sepcular within close proximity to 12.5 are completely black~~ to Cycles: Principled BSDF: materials with Sepcular within close proximity to 12.5 are completely black

2022-03-22 22:55:34 +01:00

Alaska commented

2022-03-22 23:10:45 +01:00

Added subscriber: @Alaska

Alaska commented

2022-03-22 23:10:45 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Alaska commented

2022-03-22 23:10:45 +01:00

I can confirm the issue.

System Information
Operating system: Linux-5.16.0-4-amd64-x86_64-with-glibc2.33 64 Bits
Graphics card: NVIDIA GeForce RTX 3060/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 510.54
Blender version: 3.2.0 Alpha, branch: master, commit date: 2022-03-22 07:07, hash: 47451af027

I can confirm the issue. **System Information** Operating system: Linux-5.16.0-4-amd64-x86_64-with-glibc2.33 64 Bits Graphics card: NVIDIA GeForce RTX 3060/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 510.54 Blender version: 3.2.0 Alpha, branch: master, commit date: 2022-03-22 07:07, hash: `47451af027`

Mikhail Matrosov commented

2022-03-23 22:26:18 +01:00

Added subscriber: @ktdfly

Riccardo Giovanetti commented

2022-03-24 20:13:38 +01:00

Added subscriber: @Harvester

Riccardo Giovanetti commented

2022-03-24 20:13:38 +01:00

Is this really a bug, or simply a misuse of this parameter in the Principled BSDF, being the Specular value a trick to play with the IOR within plausible values for dielectric materials? The Pixar paper on Principled BSDF, v3, states this for the Specular parameter: The specular parameter may be pushed beyond one to reach higher IOR values but should be done with caution.
A value of 12.5 seems far from being plausibile and within cautious values. Am I wrong?

Is this really a bug, or simply a misuse of this parameter in the Principled BSDF, being the Specular value a trick to play with the IOR within plausible values for dielectric materials? The Pixar paper on Principled BSDF, v3, states this for the Specular parameter: The specular parameter may be pushed beyond one to reach higher IOR values but should be done with caution. A value of 12.5 seems far from being plausibile and within cautious values. Am I wrong?

tamooz br commented

2022-03-24 21:15:55 +01:00

It is a bug. It doesn't appear black when using OSL nor EEVEE, and the "Specular" value to IOR converter returns 0, which is not physically possible, but does not return a specular value of 0, but rather a 1.

Jesse Yurkovich commented

2022-03-24 22:56:38 +01:00

Added subscriber: @deadpin

Jesse Yurkovich commented

2022-03-24 22:56:38 +01:00

Yeah, this is not really a bug. See the "Specular" section of https://docs.blender.org/manual/en/3.2/render/shader_nodes/shader/principled.html#inputs

A value of 12.5 has no solution to the equation listed there and is well outside the range of all known dielectric IOR values. Natural titanium dioxide peaks at ior=2.6 for a specular of ~2.5 and artificial meta-materials peak at ior=38.6 for a specular of 11.27. https://www.desmos.com/calculator/0tlhy6wymf

Perhaps we should hard-cap the slider at 12.

Yeah, this is not really a bug. See the "Specular" section of https://docs.blender.org/manual/en/3.2/render/shader_nodes/shader/principled.html#inputs A value of 12.5 has no solution to the equation listed there and is well outside the range of all known dielectric IOR values. Natural titanium dioxide peaks at ior=2.6 for a specular of ~2.5 and artificial meta-materials peak at ior=38.6 for a specular of 11.27. https://www.desmos.com/calculator/0tlhy6wymf Perhaps we should hard-cap the slider at 12.

Jesse Yurkovich commented

2022-03-25 01:45:51 +01:00

For completeness, yeah there does seem to be a difference between OSL and not. Though it's by accident it looks like. Both are kind of wrong really.

OSL's kernel\osl\shaders\node_principled_bsdf.osl:93 calculation of (2.0 / (1.0 - sqrt(0.08 * Specular))) - 1.0, is different to the below (can't debug through to get actual value right now).
CPU's kernel\svm\closure.h:313 calculation of bsdf->ior = (2.0f / (1.0f - safe_sqrtf(0.08f * specular))) - 1.0f produces a floating point "inf" value due to the div by 0

For completeness, yeah there does seem to be a difference between OSL and not. Though it's by accident it looks like. Both are kind of wrong really. OSL's `kernel\osl\shaders\node_principled_bsdf.osl:93` calculation of `(2.0 / (1.0 - sqrt(0.08 * Specular))) - 1.0,` is different to the below (can't debug through to get actual value right now). CPU's `kernel\svm\closure.h:313` calculation of `bsdf->ior = (2.0f / (1.0f - safe_sqrtf(0.08f * specular))) - 1.0f` produces a floating point "inf" value due to the div by 0

tamooz br commented

2022-03-25 10:22:43 +01:00

Both are (2/(1-√(0.08 * specular) - 1, so the difference is how they handle division by 0.
BTW with the IOR to specular formula with IOR 0, we get ((0-1)/(0+1))**2*0.08, which is 12.5, and by removing the *0.08 part, which yeilds an f0 value, we get of 1.

Both are (2/(1-√(0.08 * specular) - 1, so the difference is how they handle division by 0. BTW with the IOR to specular formula with IOR 0, we get ((0-1)/(0+1))**2*0.08, which is 12.5, and by removing the `*0.08` part, which yeilds an f0 value, we get of 1.

Riccardo Giovanetti commented

2022-03-25 16:09:17 +01:00

Refractive index of a transparent medium is always greater than 1 (it cannot be less than 1) because speed of light in any medium is always less than that in vacuum. IOR 1,00028 (approx) which is for Air should be the lowest value to be used.

Lukas Stockner commented

2022-05-07 23:31:33 +02:00

Added subscriber: @LukasStockner

Lukas Stockner commented

2022-05-07 23:31:33 +02:00

I wouldn't call this a bug, specular is supposed to be treated as a fraction. Going a bit beyond 1 for artistic control might make sense in some cases, but 12.5 is so far outside the intended range that I'm not even sure what the behavior should be.

Generally, the principled BSDF is intended for perceptually uniform, intuitive control. If you find yourself manually inverting the IOR mapping function to find the appropriate input for a given material, you probably don't want to be using the Principled BSDF in the first place.

It's probably best to just hard-clamp the input at something like 5 or even 2.

I wouldn't call this a bug, `specular` is supposed to be treated as a fraction. Going a bit beyond 1 for artistic control might make sense in some cases, but 12.5 is so far outside the intended range that I'm not even sure what the behavior should be. Generally, the principled BSDF is intended for perceptually uniform, intuitive control. If you find yourself manually inverting the IOR mapping function to find the appropriate input for a given material, you probably don't want to be using the Principled BSDF in the first place. It's probably best to just hard-clamp the input at something like 5 or even 2.

tamooz br commented

2022-08-15 09:47:00 +02:00

this should be closed, it's a clone of #63938

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Cycles: Principled BSDF: materials with Sepcular within close proximity to 12.5 are completely black #96719