Another evidence I think that we are allowed to use some metallicness for dielectrics; We cannot expect to build the micro surface with geometrie.
“We cannot expect to build the micro surface with geometrie.”
That’s what most various shader types in a material node setup does; mimics how various microsurface categories react to light and surroundings. A pure Lambertian diffuse (fully view independent) is hard to obtain in the real world. Similarly most materials are also not pure specular (fully view dependent), but a mix of the two (or more) - or more specifically, it’s look can be approximated using a mix of the two (or more). All dielectrics are absorbent to varying extent, and will scatter back some of the light that passes through its surface, meaning theoretically they should all have some kind of SSS or translucent effect. But we don’t due to the cost. Even Disney use a fake SSS approach rather than the real deal.
“some metallicness”
No. This is a workflow descriptor, and solely describes if a material is a dielectric or a metal. Other than for antialiasing purposes, it is advised to use either full one or the other rather than some weird 50% metallicness. Disney have in their PBR approach sheen and sheen tint controls which are separate from glossy and glossy tint, but I’m not sure what shader type is used. Could be some velvet shader variant. I don’t recall if the Cycles velvet shader is view dependent like a glossy or view independent like a diffuse. But the name velvet shader alone suggest this is what you should try to use rather than colored gloss. I’m no expert though so I could be wrong. Are your own observations view dependent or independent?
Since it seems a bit controversial since all evidence seems to suggest something else, I would be care to put such a material in a library as a reference thing. However, if it looks okay in the current project, I guess nothing else really matters. I don’t hesitate using pure diffuse if it gets my rendertimes down to manageable levels, even if I know nothing out there behaves that way. Unless a test with glossy reveals significant improvements in important parts of the frame.
Thank you for your detailed reply. It feels like a mentor relevates the way I see things. I realize now (again) that the shaders already mimic a micro structure.
When I look at my real life satin curtain, I see the colored gloss view depended. But since rough gloss reduces the fresnel effect, I wonder if I should use fresnel or not. It is so rough (0.35) that it doesn’t matter much I think. My curtain shows a strong sheen I think. It is quite hard to judge it; for that I need to bring the curtains outside when it is cloudy so I have a more diffuse light. Sometimes I don’t know if I look at a shadow, sheen, or dark reflection.
Here some collage with pictures of the curtain. Sometimes you don’t see any “Metallicness” effect, and sometimes it is strong. I put some green arrows where there I find the effect strong.
No image? Check:
http://www.pasteall.org/pic/show.php?id=104537