Am 17.05.2013 03:51, schrieb OJD:
>   I am looking for some clarity on what the 'samples' setting actually controls
> for subsurface in global settings.
>
> version 3.7 RC7
>
> The documentation (3.4.6.3.3.4 Subsurface Light Transport) states the setting
> controls "the number of samples for the diffuse scattering and single-scattering
> approximation". What is the difference in diffuse scattering and single
> scattering??
>
> Now according to what I have read on the theory behind subsurface light
> scattering there are so many bounces going on that regardless of the material
> you will always get an isotropic scattering effect. So this would account for
> the diffuse setting. I would assume this would be a similar setting to
> scattering media type 1 and increasing samples will control how evenly light is
> distributed.
>
>   So I am puzzled at what is meant for single scattering and what that setting
> will effect/control.

To elaborate on what Alain already mentioned:

The math behind the subsurface light scattering as implemented in 
POV-Ray is based on scientific research on the matter, and - within 
certain limitations - simulates the whole thing. But its core is a model 
that only accounts for light scattered at least twice in the material, 
while in reality some light rays bounce right out after a single 
scattering event; so the algorithm is augmented by another - quite 
different - method that takes care exclusively of the portion of the 
light that experiences a single scattering event. The two contributions 
are then simply added. (And, in POV-Ray, yet another term is added to 
account for light that passes through entirely unscattered, to also get 
acceptable results for very thin or highly translucent objects that 
border on being transparent.)

Each of the two algorithms take their own set of samples (there's really 
no way either of the algorithms could make use of the other one's 
samples), and the number of those samples can be controlled independently.

With the single-scattering contribution being typically much lower than 
the dual-or-more-scattering contribution, it makes sense to have POV-Ray 
spend less time on that one by choosing a lower number of samples for 
the former than for the latter. I typically use a factor of 10.

Post a reply to this message