Specular reflection vs diffuse reflection.

[daros]

The mixer test would be perfect if you would be allowed to use a fresnel function to blend the nodes. It cannot be a constant value (like it is in maxwell after version alpha) but a value which varies accordingly to the incidence angle.

It is pure non technical speculation, but as I understand :

• IOR controls the fresnel faloff
• mix shader normalizes reflected light between the two shaders

If this is true and I understand what you are explaining, playing with IOR and mix factor acts as a fresnel blend.

Here is a test with different IOR values for a glossy with black diffuse to see better the effect :


ClearcoatIOR.jpg


Higher the IOR and "flatter" are the reflections. However looking up for IOR references, IORs above 3 seem pretty rare, so it might not be physically sound to use it like this.

Correct, the higher ior trick is the same Maxwell users are forced to follow but it simply wrong. Things have to look realistic at real IORs.

[tokiop]

It seems "simple/single number IOR" are only valid for dielectrics, non-dielectics materials like metals requiring "Complex/full IOR" with more complex equation and additional components/values.

Maybe the "simple" real IORs that can be found in some list, for some metals for example, don't make sense because they are only one component of their complex IOR ? Therefore they might not be a better or "more real" approximation than using the "high" IOR trick ?

[pr0ggy]

Have been running into this same issue off and on again...I was able to create a Cycles dielectric node that seems to fix the problem (image/scene file attached). Overall light reflectance seems to drop very slightly but results are certainly comparable to the Principled/Disney node. I've also tested it in a "white furnace" scene and results appear to be accurate regarding energy conservation. I realize I'm posting this in a LuxCore forum, but wanted to reply as I thought perhaps it could be useful to other people running into this issue and maybe even useful to LuxCore devs somehow (though I'm really not sure how it would be...it's a basic diffuse/glossy mixed by roughness-dependent fresnel).

EDIT: Realized there was an issue with how I was calculating roughness-dependent fresnel in original upload...fixed screenshot and scene file attached

[daros]

Hi, toady i can resuscitate this older post.

Remember the white ball with the halo around i cannot get rid of in LUX?

Finally we have been able to make the Mitsuba 2.2.1 tests to see how mitsuba deals with fresnels.
I have to say it is almost perfect. I do not have any complain about this results and i think we are in front of an accurate simulation of reality.

But first let's take a look too some photos of the real materials samples we did in our sampler box.

Lime

Glazed ceramic

Red carpaint

Chrome

And this is Mitsuba

White matte

White ceramic

Red carpaint

Chrome

As you can see the mitsuba results catch almost perfectly the fresnel of our photos. The differences are related more to gamma settings and to light distribution since the mitstuba tests are done in path tracing.

I hope this helps to improve luxcore's material model.

Thanks to my friend Massimo Ciani for the tests!

here the original mitsuba files
https://drive.google.com/file/d/1Hp1H0q ... sp=sharing

[daros]

with some minor gamma and exposure correction i got this:



Left hemisphere Mitsuba, right hemisphere real photo.

Left hemisphere Mitsuba, right hemisphere real photo.

Left hemisphere Mitsuba, right hemisphere real photo.