Am 22.09.2012 22:16, schrieb Christian Froeschlin:
> I tried the subsurface feature for the first time (RC6) and noticed
> that normals no longer seemed to have a visible effect on lighting, e.g.
> testing with an object such as
>
> sphere
> {
>    y,1
>    pigment {color White}
>    normal  {bozo 100 scale 0.2}
>    // finish  {subsurface {translucency 0.2}}
> }
>
> Is this a known limitation? If so it should probably be documented
> in the warning section of the subsurface reference.

I never paid attention to this before, but you're right: Currently the 
algorithm doesn't properly respect perturbed normals. The reason is that 
this would require extra effort (both in terms of speed, but also and 
especially in terms of programming) to handle averaged and layered 
textures. There'll come a time to dig into this, but not yet.

> Also, if I understand correctly SSLT completely replaces the lighting
> model. I would have found it to be more intuitive if it were just some
> additional effect (similar to a reflection block, at least without
> conserve_energy). Possibly by defaulting the SSLT contribution with
> a diffuse of 0 unless an explicit diffuse appears in the subsurface
> block (in which case it would be physically correct, but optional, to
> use diffuse 0 for the old model. Not sure if this is feasible, though.

 From the math behind SSLT, there is no such thing as an "SSLT 
contribution" - the formulae totally /replace/ the conventional diffuse 
term.

To obtain a pure SSLT-to-diffuse "delta" image, you could render the 
scene once with only regular diffuse, and once with SSLT enabled, and 
from the resulting images generate a difference. (Note though that 
SSLT's light "bleeding" feature does not only brighten up dark areas, 
but also darken bright ones in turn; so the delta may be negative, and 
therefore OpenEXR is highly recommended as file format of choice.)

Post a reply to this message