Hi all, could someone run the code below and tell me if they see large
artifacts? A brief description of the problem, and the code is below:

Rendering scenes containing sphere sweep objects which are either very large
or very small leads to large artifacts. Different artifacts are seen
depending on the size of the object, and the type of spline used (linear,
cubic etc). A simple test scene is reproduced below.

The parameter 'myscale' scales the object and the cameras position
proportionally. For values of 'myscale' less then 0.01, and greater then
10, various artifacts are seen. The artifacts cannot be removed by varying
the value of 'tolerance'.

#include "colors.inc"

#declare num_phi1=10;
#declare myscale=0.01;

sphere_sweep {
    cubic_spline
    num_phi1+1,
    #declare _phi1 = 0;
    #while (_phi1<=6)
         #declare p1 =<sin(_phi1),-cos(_phi1),0>*myscale;
         p1,0.1*myscale
         #declare _phi1 = _phi1 + 6/num_phi1;
    #end
    tolerance 0.01
    pigment {Red}
}

background{White}

camera {
    angle 0
    location <0.0, 0, -1>*3*myscale
    look_at <0,0,0>
}

light_source { <0, 0, 0> color White }

Post a reply to this message

Could someone run the code and let me know what they see?

Post a reply to this message

Much delayed response from me, sorry, I just now took a look at this. Not 
sure why no one else has replied yet, but could be because the sphere_sweep 
has some peculiarities (problems?) and this only seems like another example.

I have no idea about it. Does appear perfectly okay here, using both version 
3.6.1b and the 3.7 beta 13, when the scale number is between 0.013 and 0.34.

Just letting you know I've given it a look even if I can't explain the 
trouble with it. Don't know what 'angle 0' was doing in there, I've never 
used zero for that (not that I can remember anyway), but I got the same 
results with it in or out of the camera.

-- 
Bob H  www.3digitaleyes.com
http://3digitaleyes.com/imagery/

Post a reply to this message

Thanks Bob, it's good to hear someone else can see this problem. Sorry about
the 'angle 0' red herring, however as you say it doesn't affect the
existense of the artifacts. It's a shame about the problems with sphere
sweep, as it is quite useful.

For the record I have seen this on the latest stable and beta releases on
Windows.

Can anyone tell me how many independant confirmations are required for this
to be registered as a bug?

Post a reply to this message

I can confirm your observations: Running the scene on my Gentoo Linux 
Athlon XP2600+ 1Gig RAM box I get the mentioned artifacts with the 
official 3.6 binaries.

With a self-compiled (from unmodified sources) version of 3.5 there are 
*no* artifacts, so you may check out what you get using an official 3.5 
binary.

With 3 confirmations (counting yourself in) from different platforms 
it's IMHO fine to file a bug report. Just make sure there hasn't been 
one filed for the same issue - the sphere sweep artifact topic has come 
up often in the past. So search povray.bug-reports before posting ;)


HTH,
Florian

Post a reply to this message

Thanks Florian, especially for the 3.5 tip; at least I can now see my image
how it was suppossed to be. I'll file a bug report soon.

Thanks all.

Post a reply to this message

"scam" <sca### [at] mailusydeduau> wrote:
> Hi all, could someone run the code below and tell me if they see large
> artifacts?

I'm coming late to the discussion, sorry.  I was testing out your code,
which is a good one for producing these artifacts. This is a problem I've
been giving some thought to lately as well.

On my own system (a Pentium II 400MHZ, running Windows version of POV-Ray v
3.6.1) I'm actually seeing artifacts at MOST values of myscale -- the only
artifact-free "zone" I can get is from 0.5 down to 0.013 --and then CRAZY
artifacts from 0.0127 on down (possibly floating-point errors there? All
the scene's values are very small in that domain, and the camera is getting
quite close to the origin and your object.)

Yep, it does seem like a bug.

I have a theory: From a purely visual analysis of your scene, it seems that
the artifacts are at their worst when the camera is perpendicular to the
sphere sweep object (or its points.)  That is, imagine each point in your
sphere_sweep having a plane slicing right through it from front to back, at
right-angles to the surface there. Like a knife blade slicing the
sphere_sweep into perfect tube segments at each point. If the camera
happens to be looking RIGHT AT those imaginary planes, edge-on (as in your
scene), it sees the most artifacts. Moving the camera off-axis of those
planes (all of them, not an easy or practical thing to do) eliminates most
if not all of the artifacts. That's borne out by experimentation I've done.
Which could also explain something I've noticed in my own scenes, why longer
sphere_sweep objects tend to show more problems: There are more points--and
thus more imaginary planes to have to avoid.  Close to impossible to do, of
course, with a long, sinuously curving sphere_sweep object. What's
surprising (to me) is that an orthographic camera shows the artifacts in
your scene in a different way--no longer as radiating spikes. So my theory
could be wrong. But it does help me in trying to understand what's going
on.

Ken W.

Post a reply to this message