VRayMtl Anisotropy color problems

[eric_chadwick]

We are investigating the conversion of V-Ray anisotropy into glTF anisotropy, and we have noticed strange color behavior in V-Ray. Is this the expected result?

For all materials, Diffuse color is (250, 208, 192), and Metallic is 1.0.



Model is an edited version of https://github.com/KhronosGroup/glTF...pyStrengthTest
© 2023 Ed Mackey, Public. CC0 1.0 Universal

HDR is from https://github.com/BabylonJS/Assets/...a_radiance.hdr
CC BY 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode
Babylon.js


3ds Max 2024.2
V-Ray 6.10.08
Windows 11 Home 64-bit (10.0, Build 22631)
Intel i7-11700K @ 3.60GHz
GeForce RTX 3080 Ti
64GB RAM

[eric_chadwick]

It also seems odd that setting anisotropy to 1.0 causes anisotropic blurring/stretching to disappear completely. Shouldn't the highest setting cause the strongest anisotropy?

VRayMtl applied to the 2nd sphere in the top row:

[hermit.crab]

Values 1 and -1 do not have an effect. It's probably a good idea to write this in the docs and set it as a limit to the value box.

[eric_chadwick]

Thanks for the input, that helps with the 1.0 setting.

How about the odd color behavior in the rest of the grid? I get very different results with glTF rendering of anisotropy. Admittedly it's being shown in wholly different renderers, and they are rasterizers. However i would hope for roughly equivalent results in V-Ray. It almost looks like there could be shader errors in V-Ray?


AnisotropyStrengthTest_Copper_upper-range.glb is attached. Note: the lower end of the Anisotropy Strength is not 0.0 as the label suggests, instead in this example it is 0.94, and the values increase by 0.01 as you get higher:

[hermit.crab]

Just so you know, I've troubleshoot the issue and logged it (internal bug-tracker id: VMAX-14010) for dev investigation. I'll write back whenever there is more information.

[asen.atanasov]

Hi,

I am looking into these issues. Here's some information:

1. The anisotropy approaching 1/-1 does not correspond to any physical material, because it approaches infinite stretching-compressing. I will fix it so that this is not visible to the users.
2. Regarding the darkening - this isn't a bug, but it is actually a feature called "color sharpening". As the roughness increases, the multiple scattering inside the microstructure (which we do not simulate, but only approximate) leads to accumulation of Fresnel terms. However, it seems counterintuitive to our users, thus we consider to remove that for new scenes and leave a possibly hidden option to turn it on.

Best regards,
Asen

[eric_chadwick]

Thanks for the investigation and information! I am wondering if energy compensation might be beneficial in this context? For example see Revisiting Physically Based Shading at Imageworks, by Christopher Kulla & Alejandro Conty, from SIGGRAPH 2017: https://blog.selfshadow.com/publicat..._slides_v2.pdf