Hi,

in my last image there are small artifacts, when 13 media containers meet each
other. It didn't disturb too much, because they aren't visible at smaller
scale.

Now I wanted to find a solution, but all of my usual tricks failed (hollow,
merge, max_trace_level, same ior, full transparency, samples/intervals).

IIRC clipca said, in 3.7 fully transparent objects are handled differently as in
older versions.


Now for the experts, here is a minimal scene -


#version 3.7;

global_settings {
        assumed_gamma 1
        max_trace_level 255
}

camera {
        location <0.6,2,-5>
        look_at 0
        angle 1.2
}

#declare MPipe =
material {
        texture {
                pigment {color rgbt 1}
        }
        interior {
                ior 1.0
                media {
                        emission 1
                        samples 80
                }
        }
}

#declare m = 0;
#while (m < 5)
        cylinder {
                y*-5, y*5, 0.02
                hollow
                material {MPipe}
                rotate <90,m*72,0>
        }
        #declare m = m+1;
#end

// end

Norbert

Post a reply to this message

Attachments:
Download 'media_issue.jpg' (317 KB)

Preview of image 'media_issue.jpg'

Le 24/03/2016 15:37, Norbert Kern a écrit :
> Hi,
> 
> in my last image there are small artifacts, when 13 media containers meet each
> other. It didn't disturb too much, because they aren't visible at smaller
> scale.
> 
> Now I wanted to find a solution, but all of my usual tricks failed (hollow,
> merge, max_trace_level, same ior, full transparency, samples/intervals).
> 
> IIRC clipca said, in 3.7 fully transparent objects are handled differently as in
> older versions.
> 
> 
> Now for the experts, here is a minimal scene -
> 
Good scene.

artefacts appear in:
* stable
* master

artefacts *do* *not* appear in:
* hgpovray (which is based on previous of my own master)
* my own master (which is late behind master)

Does not seems tied to the compiler either.

Of interesting differences between official and mine, I would consider the fix to
fs#324 (2014-06-07):

MIN_ISECT_DEPTH (configbackend.h) was moved from 1.0e-4 to 0.0
*with* a change in trace.cpp and objects.cpp : instead of tmpDepth >= MIN_ISECT_DEPTH,
it is tmpDepth > MIN_ISECT_DEPTH
(not equal, strict inequality)

and in object.cpp, MIN_ISECT_DEPTH is replaced by SMALL_TOLERANCE for intersection
with bounding box.



https://bitbucket.org/LeForgeron/povray/commits/948a24f32f328e3d8ff043ab177c94501814c7e2?at=default


http://bugs.povray.org/task/324?project=2&order=id&sort=desc

Post a reply to this message