This was derived from Beckmann-Kircchoff-Spizichino model.

For mertal perfect mirrors the specular component doesn't exist and is equal to zero. The reflective component (mirror) is constant for all incidence angles.

For very rough metals the reflective (mirror) component does not exist and is equal to zero. The specular component is constant with a peak and then falling quickly to zero at 90 degrees of incidence.

For glossy metals that are in the middle of perfect mirrors and rough surfaces, both components do exist as is shown in the image.

The practical effect is that glossy metals at normal angle incidence present only the specular component and at glazing angles of incidence behaves as a mirror.

As specularity is not a separate physical phenomenon but just reflection which gets blurred on a rough surface, I am not surprized that it reduces at grazing angles. Did you have a question, anyway?

So all we need now is just another node to plug into the Alt_Specular channel. And perhaps a new Alt_Reflection channel as well. Poser 99? Poserpro 2092?

Or just a physically correct unbiased renderer, like LuxRender?

Now your talking reflections, it might be nice to have some attention for polarisation, which is angle-dependant in a similar way (and definitely a Fresnel thing). The human eye responds vey different to it than a camera, and no 3D renderer addresses it at all.

Specular component is what produce a highlight without producing a mirror image. It can be the normal or alt_specular channel.

Reflective component is what produce a mirror image, it is the reflection channel.

A blurred reflection is a combination of both or a surface that has not its normals smooth.

I don't know how to create this effect, you can multiply the specular value by (1 - X) and the reflective value by X, where X is a function of the incidence angle that has a value of zero at normal incidence and one at 90 degrees. The slope is a function of how glossy is the metal. Maybe Fresnel can work as the multiplying function. I repeat, this is not Fresnel, I am dealing with metals and not dielectrics.

LuxRender has not this effect and as I know no physically correct unbiased renderer is unable to produce it. In LuxRender you can simulate using a mixed material (glossy + mirror) multiplied by the function X, but from where do I get this function ?

For the moment I am not taking into account polarization.

ricardo, bill is working on a shader that normalises the blinn and reflection softness functions for specular and diffuse sources, in a version of poser which has a fresnel blend node to make the variation in the reflectivity from the edge to the centre of the posersurface, as viewed by the camera.

Something just went "pop" behind my left eye. ;)

As stated, reflection is the physical phenomenon which returns the light (and image), and surface roughness is the physical phenomenon which turns reflections into anything from specular to blurred reflection to full mirror.

So, in a physics based renderer, pick metal als the material of choise, adjust its surface roughness and that should do. If not, the metal shader for that renderer needs an upgrade. There should be no need for constructing metal from hard plastic (glossy) and mirror, as it does not work that way in reality either.

In Poser, you need a materials node that reduces specularity with grazing angle to plug into Alt_Specular. This node does not exist yet. And you need a node that reduces reflection with normal angle, which does not exist either, and above that: has no Alt_Reflection channel to be plugged in to. So, no way to get this effect done in FireFly - at least for now.

Is that a problem? Well, Firefly can come close to realistic looking results in a lot of cases, but will always fall short at the detailed level. This metallic specular / reflection thing is one of them. Polarisation is another. Specular and reflection in a no-light full-IDL render is another, and so on.

Will it get fixed? Perhaps one day, next to all other physical incorrect behaviour noted in other areas (cloth and hair dynamics, etc). Point is: SM doesn't see a market for these niche adjustments. So, forking out to another renderer is an option. Render with specular, render with reflection and merge in post is another option. Or building a math nodes construction altering Specular_Value and Reflection_Value with Angle (to camera? to light? to surface normal?) and some roughness and reflectivity parameters might be an option too.

My impression is that you're in search for the latter. Is it?

There are many models:

Phong or Blinn variant: Are very simple, can work fine for glossy or not so rough materials. The original were empirical and not physically correct, but there is a variant that makes them physically correct. These models don't work for rough materials and mirrors and are unable to produce the effect that is the subject of this thread. For plastics combined with Fresnel can work pretty well.

Schlick: Is the same thing as Phong or Blinn, but require less computation time. For very specular materials produce the same thing as Phong, but when the material becomes rough the behavior of Sclick is very poor, Phong is much better.

Cook-Sparrow-Torrance: Physically correct for rough and little glossy surfaces. Normally are used for metals. Fail to work with glossy and mirror like surfaces.

He-Torrance: The most complete physically correct model. It can deal with rough to mirror-like materials and polarization effects. It is derived from Beckmann-Spizichino model and is very computational expensive. As I know no rendering engine has implemented this model.

Lambert: The most simple and ideal model for diffuse illumination. Pure Lambertian objects do not exist in Nature.

Lommel-Seeliger: The other fundamental model for difusse illumination. Very few rendering engines has this model.

Oren-Nayar: Model of difusse illumination for very rough materials. It works fine with ceranics and many diffuse materials.

Hapke: Model for dusty surfaces, a must have for rendering planets, moon and vegetation.

And many others.....

What I am doing is working on the complete model of Kircchoff-Beckmann-Spizichino or He-Sillon-Torrance variant and trying to find an approximation that preserve all its features, but has a small computation time.

Sounds good. Keep on going.