"Trevor G Quayle" <Tin### [at] hotmailcom> wrote:
> Warp <war### [at] tagpovrayorg> wrote:
> > Stephen <mcavoys_AT_aolDOT.com> wrote:
> > > Don?t use Merge, use Difference to subtract the spheres for the cornea. Then
> > > Union a sphere slightly smaller than the inner surface of the cornea for the
> > > Vitreous Humour. (To avoid coincidental surfaces)
> >
> >   Are you sure it makes a difference to use an outer layer consisting of
> > the difference of two spheres, with a third inner sphere inside it,
> > compared to just putting one sphere inside the other?
> >
> > --
> >                                                           - Warp
>
> There seems to be a slight difference in the refraction at the edges of the
> inner sphere.  Which one is correct, I don't know for certain, but if you
> try with two spheres of the same ior, the one without the difference looks
> correct, i.e., looks like a solid sphere of continuous ior, while the
> differenced one has refraction that is inconsistent.  When using fresnel
> reflection, the differenced one will also give incorrect reflections on the
> inner sphere, i.e., too intense.  Based on these observations, I think that
> it is more correct to 'embed' the second sphere inside the first without
> differencing.  Perhaps someone who knows the inner working of how POV
> handles refraction has some better information on this.
>
> -tgq

I think I have found the answer to why these are different.  In general, the
refraction path is almost identical:
ior: 1.0-1.5-1.5-1.5-1.0 for embedded vs
ior: 1.0-1.5-1.0-1.5-1.0-1.5-1.0 for differenced
at each object interface.  In the differenced one, the extra 1.5-1.0-1.5
transition takes place over a very small distance and can be neglected,
effectively giving it the same properties as the embedded.  However, the
problem arises when the 1.5-1.0 transition is beyond the total internal
refraction angle.  The ray erroneously gets reflected and remains inside
the outer object and it never passes through the tiny 'gap' into the inner
object.

Now, both these are valid situations.  If the objects are such that there
actually is a tiny air gap, then the differenced one is correct, however,
for objects without the gap, where the transition is directly from the
outer to the inner material (i.e., they are in contact), then the embedded
one is correct.
Consider two pieces of glass lying on top of one another in real life.  The
interface is quite noticeable if the glass is dry (i.e., have a gap) vs if
there is a thin layer of water essentially making them in direct contact.

Most glazed or embedded objects with act like the second situation where the
materials are in direct contact.  This is especially true when simulating a
varying ior via layers of objects like the eye model, you definitely do not
want to introduce the air gap.

-tgq

Post a reply to this message