As @niklasharrysson notes in his review of #2998, the OpenPBR specification defines the thin-walled subsurface lobes as unit-albedo diffuse reflection and transmission lobes, scaled by a single factor of subsurface_color and split by the anisotropy weights (1 -/+ g).
In our nodegraph implementation, subsurface_color is applied both as the color input of each diffuse lobe and as an external multiplier alongside the anisotropy weights, so the total albedo of the thin-walled subsurface term is squared. At the default zero anisotropy, a 0.5 gray subsurface_color yields a total albedo of 0.25 rather than the specified 0.5, darkening and saturating subsurface colors in thin-walled mode.
One proposed fix would be to remove subsurface_color from the external multiplier factors, leaving the color input of each lobe as its single application.
The reference implementation in the OpenPBR repository shares this construction, and we would propagate any fix made here to that project as well.
As @niklasharrysson notes in his review of #2998, the OpenPBR specification defines the thin-walled subsurface lobes as unit-albedo diffuse reflection and transmission lobes, scaled by a single factor of
subsurface_colorand split by the anisotropy weights(1 -/+ g).In our nodegraph implementation,
subsurface_coloris applied both as thecolorinput of each diffuse lobe and as an external multiplier alongside the anisotropy weights, so the total albedo of the thin-walled subsurface term is squared. At the default zero anisotropy, a 0.5 graysubsurface_coloryields a total albedo of 0.25 rather than the specified 0.5, darkening and saturating subsurface colors in thin-walled mode.One proposed fix would be to remove
subsurface_colorfrom the external multiplier factors, leaving thecolorinput of each lobe as its single application.The reference implementation in the OpenPBR repository shares this construction, and we would propagate any fix made here to that project as well.