since the bright fresnel reflection ,at grazing angles with high roughness values,doesent looks right to me.i have searched in the net for a workaround.and i think i have found a answer.some smart guys have allready found a way to manage this problem.the fresnel falloff seems allready patched and used in many known renderengines such as,Vray ,Fstorm and Corona to name a few wich have all in common the same fresnel falloff curve.
here are pictures from a guy who measured the falloff curves in the meantioned renderengines.
the reason for open a new is to not spam allways moonys threats with my own testrenders.
btt .i dont know if this is posted before,but in moony threat came the question, if we can have a simple curve for the fresnel falloff.and now at least 3 big renderengines uses the same falloff curve.
if i am not wrong,then this would be the standart curve behavior for PBR renderings with roughness falloff.
edit,here a testrender with PBR mat from this site
the math of falloff fac is based on : F0 fresnel reflection as base and the roughness value as exponent in the power node,and works well imho, with the tested IOR of 1.5 ,as you can see in my first testrender above.
and additional the roughness value need to be scuares, at least to match the posted testrender,depents ofcourse of the roughness maps.you can use the roughness² output for this.
for interested users ,here are the link, with the disscusion about the fresnel falloff,and where i found the measured testrender.
i have found measurements from real metals.
at pic a and b,you see the absorptance on a copper surface (at green 532nm) with different roughness values (microns/mikrometer um) plotted,at angles of incident from 0-90.
(100-Absorptance=Reflectance)
pic a shows 0.01 -0.5 microns,and pic b shows 0.5-2 microns roughness Absorptance curve
interesting to see with 0.5- 2 microns the fallofff at 90 is allmost down to F0.
