 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
56 hour render time on this, 46 minutes just for the photon map - a record.
I used a saved photon map from a much smaller test render. It's hard to
experiment
when the render speed drops under 1 PPS. My biggest mistake was not
setting the finish of the background/foreground to ambient 0 and letting
the sky sphere light it, which, combined with the media, led to a washed
out appearance. The projected caustics were a suprise at this resolution.
In small test renders they looked quite soft, not at all like the hot-spots
you see here. I hand bounded the sphere with a sphere the same size the
visible one was cut from, but did so after I had applied the texture, which
may have hurt rather than helped the render time. Also, I neglected to
enable media in radiosity. Radiosity has so many contols to tweak I am
beginning to believe I should start with it first, and get the scene to
look right before adding media and photons. I'm planning to use this as
the center sphere of an armillary I am designing. Comments?
Post a reply to this message
Attachments:
Download 'orb.jpg' (51 KB)
Preview of image 'orb.jpg'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Lonnie spake:
> 56 hour render time on this, 46 minutes just for the photon map - a
> record.
> I used a saved photon map from a much smaller test render. It's hard to
> experiment
> when the render speed drops under 1 PPS. My biggest mistake was not
> setting the finish of the background/foreground to ambient 0 and letting
> the sky sphere light it, which, combined with the media, led to a washed
> out appearance. The projected caustics were a suprise at this resolution.
> In small test renders they looked quite soft, not at all like the
> hot-spots
> you see here. I hand bounded the sphere with a sphere the same size the
> visible one was cut from, but did so after I had applied the texture,
> which
> may have hurt rather than helped the render time. Also, I neglected to
> enable media in radiosity. Radiosity has so many contols to tweak I am
> beginning to believe I should start with it first, and get the scene to
> look right before adding media and photons. I'm planning to use this as
> the center sphere of an armillary I am designing. Comments?
Stupendous! Very nice, I especially like the caustics under the orb - they
look very realistic to me. How did you get the faceted look? Bound with
looped cubes?
Very nice - I like!
--
Stefan Viljoen
Software Support Technician / Programmer
Polar Design Solutions
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Lonnie" <lon### [at] yahoo com> wrote:
> 56 hour render time on this, 46 minutes just for the photon map - a record.
> I used a saved photon map from a much smaller test render. It's hard to
> experiment when the render speed drops under 1 PPS.
I think a few us here can relate to that, Lonnie. :)
> My biggest mistake was not
> setting the finish of the background/foreground to ambient 0 and letting
> the sky sphere light it, which, combined with the media, led to a washed
> out appearance.
Nonzero ambient backgrounds with radiosity are generally a no-no for that
reason, but I like the mood it creates in this image.
>The projected caustics were a suprise at this resolution.
> In small test renders they looked quite soft, not at all like the hot-spots
> you see here.
They look nice, though. The spectra look fairly smooth. How many levels of
dispersion are you using? I think the disc of tiny bright points near the
centre of the Arkenstone's shadow is an area lights artefact/bug.
>I hand bounded the sphere with a sphere the same size the
> visible one was cut from, but did so after I had applied the texture, which
> may have hurt rather than helped the render time.
I generally apply bounding to untextured shapes (because the bounding object
must be untextured anyway), but I don't think doing it your way will affect
rendering speed at all.
>Also, I neglected to enable media in radiosity.
This would give lovely coloured beams through the media, but it will also
slow things down even more.
> Radiosity has so many contols to tweak I am beginning to believe I should
> start with it first, and get the scene to look right before adding media
> and photons.
I'm pretty much a novice with media, so I'll keep quiet on that aspect, but
I agree that it's best to get radiosity looking approximately right before
adding photons, since using radiosity generally affects the overall
lighting of a scene more than using photons does.
> I'm planning to use this as the center sphere of an armillary I am
> designing. Comments?
It should look brilliant! Have you done any other sundials? A few months
back I modified the sundial example scene to work for both Hemispheres, and
added a little bit of decoration as well as East & West dial faces.
Sundial renderings need very high antialias to get smooth, accurate shadows.
The scene below is intended to be accurate, but not photorealistic by any
stretch of the imagination. :)
Post a reply to this message
Attachments:
Download 'sundialf1.jpg' (97 KB)
Preview of image 'sundialf1.jpg'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Fantastic render and amazing patience! ...and I think you are right
about everything on your post-analysis, specially about the washed out
appearance.
--
Jaime
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thanks to everyone for their kind words and recommendations. The orb itself
was quite simple to construct:
difference{
sphere{
0,1
}
#declare l2=-80;
#while (l2<90)
#declare l1=0;
#while (l1<360)
box{
<.95,-.4,-.4>,<1.1,.4,.4>
rotate <0,0,l2>
rotate <0,l1,0>
}
#declare l1=l1+22.5;
#end
#declare l2=l2+20;
#end
texture{
pigment { rgbt < 1,1,1,1 > }
finish { specular .1 diffuse .1 reflection .4}
}
interior{
ior 2.4
dispersion .9
dispersion_samples 50
}
photons{
target 1.0
refraction on
reflection on
}
bounded_by{
sphere{
0,1
}
}
rotate <-12,0,18>
}
Note that I did not set the ambient level to 0, so it becomes a light source
itself with radiosity.
I am quite certain that the strange disk of points in the orb's shadow is at
least an artefact, it's suprising that there weren't more of them in a
scene with millions of bouncing photons. I got a warning that the
max_trace_level had been reached, even though it had been set (in the
global photon block) to what is to me an incredible 41. It would be nice
if POV could pop up a dialog box when this happens, allowing you the option
of increasing it. While this was rendering, windows increased the size of
my page file to over one gig, no suprise considering the huge photon map.
What was a suprise at first was the erratic CPU usage. While doing the
area above the orb the CPU ran flat out, with a render speed well over 100
PPS. When it started on the orb, the speed slowly dropped, eventually to
under 1 PPS near the center. CPU usage dropped dramatically, often to just
an idle with occasional spikes. A friend pointed out that the hard disk
was in constant use, even though no other major apps were running. I would
guess than in situations like this the speed of the CPU is not the
bottleneck, hard disk access speed is. A possible solution I'll try is
relocating the swap file on a second hard drive.
A question that comes to mind is how random is dispersion without jitter?
Suppose I am doing an animation (I truly am a cruel task master) where the
orb spins. The facets are 22.5 degrees apart. If I use no anti-alias and
no photon jitter the orb rotated 22.5 degrees should produce exactly the
same scene as the orb with no rotation, but I have a sinking feeling this
will not be true. More tests on the way - I'll let you know.
An armillary (or spherical astrolabe) makes an excellent sundial but is by
no means limited to that. Their history goes back to ancient times and
they have been used for real research, teaching, garden ornamants (very
popular now, often totally non-functional) and as VERY elaborate fakes
intended
to fill the pockets of their makers and empty the pockets of his un-educated
patron.
My model will be as functional and astronomically accurate as I can make it,
with rings to show celestial equator and poles, latitude, longitude, the
tropics, the arctic and antarctic circles, and the plane of the ecliptic
with the zodiac constellations. I promise lots of brass :)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Lonnie nous apporta ses lumieres en ce 2005-07-22 08:19:
> I am quite certain that the strange disk of points in the orb's shadow is at
> least an artefact, it's suprising that there weren't more of them in a
> scene with millions of bouncing photons. I got a warning that the
> max_trace_level had been reached, even though it had been set (in the
> global photon block) to what is to me an incredible 41. It would be nice
> if POV could pop up a dialog box when this happens, allowing you the option
> of increasing it. While this was rendering, windows increased the size of
> my page file to over one gig, no suprise considering the huge photon map.
> What was a suprise at first was the erratic CPU usage. While doing the
> area above the orb the CPU ran flat out, with a render speed well over 100
> PPS. When it started on the orb, the speed slowly dropped, eventually to
> under 1 PPS near the center. CPU usage dropped dramatically, often to just
> an idle with occasional spikes. A friend pointed out that the hard disk
> was in constant use, even though no other major apps were running. I would
> guess than in situations like this the speed of the CPU is not the
> bottleneck, hard disk access speed is. A possible solution I'll try is
> relocating the swap file on a second hard drive.
>
> A question that comes to mind is how random is dispersion without jitter?
> Suppose I am doing an animation (I truly am a cruel task master) where the
> orb spins. The facets are 22.5 degrees apart. If I use no anti-alias and
> no photon jitter the orb rotated 22.5 degrees should produce exactly the
> same scene as the orb with no rotation, but I have a sinking feeling this
> will not be true. More tests on the way - I'll let you know.
>
When you start to swap, the CPU is always waiting for the drive. Spliting your swap on
2 physical
drives may accelerate things some. It can be done with Windows 2000, and probably with
XP. Using a
raid controler could help more.
Without jitter, dispersion is NOT random. The true random bits in POV Ray are: crand
and jitter,
nothing else.
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
