[tuhopuu-devel] Re: Minnaert diffuse shader

[Jonathan Merritt]

Hi Everyone,

I have committed a slightly modified version of Jorge's Minnaert shader 
patch.  So far, it implements both the "original" and the "nVIDIA" 
versions of the shader.  However, I'm not 100% happy with either of them 
at the moment.  I'll expand on this... :-)

Basically, the two versions of the shader implement the same kind of 
limb darkening effect, which is an extra coefficient multiplied by the 
standard Lambertian shading model.  This coefficient is <= 1.0, and 
hence selectively darkens certain parts of the surface.  The magnitude 
of this limb darkening is controlled by a "darkening" parameter.  The 
main difference between the two implementations (original vs nVIDIA) is 
the range of the "darkening" parameter, and the direction in which this 
parameter should move to introduce a greater darkening effect:

1. Original:  "darkening" should be in the range [1,2], with increasing 
values showing increasing limb darkening.

2. nVIDIA:  "darkening" should be in the range [0,1], with decreasing 
values showing increasing limb darkening.

I'm personally not happy that the "darkening" parameter is as intuitive 
as it could be for either of these cases.  What I would suggest is a 
"Blender" version of this shader, in which "darkening" lies in the range 
[0,1], with increasing values showing increasing limb darkening.  I 
think that would be a range that a user might expect for this parameter, 
and certainly having "increased darkening" ought to result in an 
increase of the darkening effect, wouldn't you think?  This is unlike 
the nVIDIA version, where "decreased darkening" results in an increase 
of the darkening effect.

To implement this "Blender" version, I would propose a straightforward 
modification of the current nVIDIA model (refer to 
source/blender/render/intern/source/rendercore.c):
    i = nl * pow( nl, 1-k ) * pow( (1.001-nv), k )
 o  "darkening" = k, and is in the range [0,1]
 o  increasing k increases the limb darkening effect.
 o  (NB: I'm not sure 100% how Alfredo's energy conservation term fits 
into all of this yet.)

What does everyone think?  Have I messed anything up really badly here? 
:-)  If people are happy about using this new definition for darkening 
then I'll go ahead and implement it.

Jonathan Merritt.