"Burki" <nomail@nomail> wrote:
> Hi!
>
> I agree with Mike:
>
>
> >My guess is that they might possibly be real effects. The particular
> >shape of the superellipsoid causing infinite internal reflections.
>
> When I set
> max_trace_level 10
> color instead if rgbt 1 to 0.96
> and in the ior: fade_power 1001
> then also black dots appear:
>
> Bu.

I just made another try with simplified scene (no carving, no photons, one
single light source), scaling the isosurface <0.6, 1, 0.45> instead of
<0.6, 1, 0.25> and lowered the 'east' and 'north' parameters (curvature
radii of the superellipsoid) down to 0.005 (almost a box), and here is what
I get. Rem: to avoid any undesirable interaction with the ground, the shape
is translated vertically 0.1*y.

The bottom of the stella is dark (adding a sky_sphere makes it look better)
because it reflects the black sky.

With higher values of the radii than my 0.005 (original 0.25 value for
example), the top seems to reflect that darkness and so on. The probem is
that these dark areas are irregular. If they were more regular, I would
think it is a normal and natural effect. So what? Numerical artifacts?

With east=north = 1, we get an ovoid that looks quite OK. With a box, it is
perfect. So, I think the problem comes from the superellisoid function
which maybe implements epsilons or whatsoever that produce those artifacts.

A last try with a sphere, and we have the same aspect than the
superellipsoid with radii=1.0.

My last remark for that post is: in the original scene, it would be cautious
to raise the shape enough along y-axis in order to make sure no point of
the ellipsoid touches the ground because the sides are also curved!

   See you soon

       Bruno

Post a reply to this message

Attachments:
Download 'stellas.png' (381 KB)

Preview of image 'stellas.png'

"Burki" <nomail@nomail> wrote:

>
> Any idea how to get rid of the silly dots on top of the stela?

This is quite a thorny little problem!

Spent several hours with your code, trying just about every conceivable
combination of things, eliminating this, tweaking that, moving the object
here and there, yet the problem wouldn't go away.

Moving the camera around, I could see that the dark areas were confined to
the rounded edges of the shape and nowhere else (but visually distorted by
the high IOR to appear as odd blobs and speckles, at odd
locations...depending on where I placed the camera.)

So I added a sky_sphere, with a single gray rgb .5 value, and that DID
eliminate most of the problems (as has already been noted.)  Well, the
major dark areas were now "filled in" with gray (filtered by the cube's
redness, of course), so they blended better with the rest of the shape's
colors. But SOME actual dark areas still remained, at the object's "edge
centers."

I began to think that the problem had something to do with your scaling of
the object--that its IOR was also being scaled in some weird way,
non-uniformly.  But reading the POV docs, IOR isn't scaled (nor is anything
else in an INTERIOR block.) Eliminating your interior's light fading didn't
help either.

In frustration, I replaced your isosurface object with a regular, non-scaled
superellipsoid, as you did...which also didn't help.

Thinking your plane's marble texture was contributing something odd,  I
changed it to a simple rgb .5 value.  STILL no good.  Eliminating the
object's shadow didn't help either. MADDENING!

(In my frustration, I even began to imagine that the problem might be
lightwave interference inside the object...one wave canceling out another
and creating dark areas. But POV doesn't use "lightwaves".)

Since the MAX max_trace_level of 256 was now being exceeded, I was ready to
give up, and blame the problem on that.  (BTW, if you plug in a really high
max_trace_level value, POV simply limits it to 256... something I didn't
know until now. So 1000000 or whatever is the same as 256.)

Then it dawned on me: Your scene's lights are spotlights...which means,
somewhere outside the scene's visible area, the plane was going to
near-black, outside the spotlight cones. So I changed the plane's finish to
ambient 1 diffuse 0 to eliminate the lights' effects, and VOILA!!  The
remaining "problems" disappeared!! Your original object was just picking up
those remaining dark areas and distorting them into unrecognizable shapes.
And even though max_trace_level was being exceeded (with my "simpler"
superellipsoid), I couldn't see any trace of that.

Of course, now my superellipsoid justs looks like a big red cube!  Nothing
of visual interest outside it to give its IOR something to work with. So it
does needs *something* darker or brighter outside, to show its optical
effects...which may end up looking like artifacts.

This was a fun little scene to play around with and learn from.

Ken

Post a reply to this message

Hi all !

First thank you all for your contributions! Especially Bruno and Ken for
their experiments. They are very interesting.

Now returned from my computer-free Christmas holidays I applied Ken's idea
of adding ambience to the floor plate (but 0.3, not 1.0) plus a sky_sphere.

And: I gave the max_trace level a last change: In LOWERING it !

Since the dots may result from multiple reflections on the bent surface
inside the stela, I cut them down by lowering the max_trace value.
Bruno showed this reflections won't show in a more regular object (as in the
cube). I got the impression that all this reflections in a slightly curved
object cause this dots.

The black areas however are still there but I don't feel that they are a
problem. In fact I think they are a result of the high ior and have to be
there.

So now I present the outcome of some lower max_trace levels. The depth of
the carving is not changed though and will be my next PTS (*)




Yours,
Bu.


(*) PTS = problem to solve :-)

Post a reply to this message

Attachments:
Download 'dots_b_gone.jpg' (77 KB)

Preview of image 'dots_b_gone.jpg'