Hi(gh)!

After fiddling around with the settings from rad_def.inc (and not 
getting near anything really looking good), I searched the Web for a 
tutorial on PoV-Ray radiosity and found this: 
http://www.simnet.is/hildurka/content/tut1page.htm

So I carefully worked through the steps described there - but when I 
came to the final setting (see code below), I noticed that my scene 
matched Mrs. Andresdottir's example scene as far as brightness of the 
shadowed parts is concerned - but still contained considerably more 
artifacts than her really clean scene (seventh from top, just above the 
version with area_light).

In her tutorial, Mrs. Andresdottir does not give the version of PoV-Ray 
she used, as her original code contains a switch for versions newer than 
3.1, I assume that it is 3.5 rather than 3.6 (which I use). Could it be 
that 3.5 uses different default radiosity settings than 3.6? Or is it 
because Mrs. Andresdottir places the whole scene inside an (monochrome) 
ambient 1 sphere rather than a gradiented sky_sphere?

Here is the code:

       radiosity
       {
         brightness 1.5
	count 300
	error_bound 0.05
         gray_threshold 0.8
         low_error_factor 0.2
         minimum_reuse 0.015
         nearest_count 10
	recursion_limit 5
         adc_bailout 0.01
         max_sample 0.9 // -1
         always_sample 1
	pretrace_start 0.08
	pretrace_end 0.01
         load_file "whatmough_residential.rad"
	save_file "whatmough_residential.rad"
       }

The only difference to Andresdottir's original code is that I use a 
gray_threshold of 0.8 rather than 0.

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download '2011-10-22 port whatmough residential building #1, take 188 - view from kitchen into atelier, cheap little beginner's ra' (25 KB)

Preview of image '2011-10-22 port whatmough residential building #1, take 188 - view from kitchen into atelier, cheap little beginner's ra'

=?ISO-8859-1?Q?J=F6rg_=27Yadgar=27_Bleimann?= <yaz### [at] gmxde> wrote:
> So I carefully worked through the steps described there - but when I
> came to the final setting (see code below), I noticed that my scene
> matched Mrs. Andresdottir's example scene as far as brightness of the
> shadowed parts is concerned - but still contained considerably more
> artifacts than her really clean scene (seventh from top, just above the
> version with area_light).

I cannot explain why Andresdottir's render is cleaner than yours.  I can say
that, in my experience, a really tiny pretrace_end is good for reducing
artifacts.

> In her tutorial, Mrs. Andresdottir does not give the version of PoV-Ray
> she used, as her original code contains a switch for versions newer than
> 3.1, I assume that it is 3.5 rather than 3.6 (which I use). Could it be
> that 3.5 uses different default radiosity settings than 3.6? Or is it
> because Mrs. Andresdottir places the whole scene inside an (monochrome)
> ambient 1 sphere rather than a gradiented sky_sphere?

I have no experience with 3.1, so I can't comment on that.  It is not a question
of any difference in the defaults between 3.5 and 3.6, as all the parameters are
specified.  I do not recall any difference in output between 3.5 and 3.6.  I
don't think there's any difference between a sphere and a sky_sphere, although
someone else will have to confirm this.

I am no expert at radiosity, but I question some of Andresdottir's settings:

>          brightness 1.5

This is physically unrealistic.  A high brightness value can compensate for a
low recursion_limit, but that's obviously not the case here.  I suppose that
this looks more "realistic" because a straight rendering does not capture the
dynamic range of real-life lighting.

>  error_bound 0.05

I've found that setting error_bound this low does nothing more than slow down
your render, and perhaps add more intractable artifacts.

>  recursion_limit 5

See the tutorial in the documentation.  There's no reason to set recursion_limit
this high, unless you enjoy waiting for stuff.

>  pretrace_end 0.01

I found that this works well in 3.6, but when you upgrade to 3.7, it won't be
small enough for many scenes.

>          load_file "whatmough_residential.rad"
>  save_file "whatmough_residential.rad"

I'm not sure what it means to load_file and save_file at the same time.  (I
vaguely recall that there's some trick that this effects, but it wasn't
fantastic enough for me to retain it.)

Good luck!

Post a reply to this message

Hi(gh)!

On 24.10.2011 06:41, Cousin Ricky wrote:

> I cannot explain why Andresdottir's render is cleaner than yours.  I can say
> that, in my experience, a really tiny pretrace_end is good for reducing
> artifacts.

So I changed the pretrace_end to 0.002...

> I've found that setting error_bound this low does nothing more than slow down
> your render, and perhaps add more intractable artifacts.

...and increased error_bound to 0.1 - aside from the scene now rendering 
somewhat faster, I found no advantages, there are still the same 
annoying artifacts (see below)! Perhaps I really should try the ambient 
sphere...

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download '2011-10-25 port whatmough residential building #1, take 190 - view from kitchen into atelier, modified andresdottir radi' (20 KB)

Preview of image '2011-10-25 port whatmough residential building #1, take 190 - view from kitchen into atelier, modified andresdottir radi'

Hi(gh)!



> Perhaps I really should try the ambient
> sphere...

I tried this:

   sphere
   {
     0, 5000000
     texture
     {
       pigment { color rgb <0.5, 0.7, 1> }
       finish { diffuse 0 ambient 1 }
     }
     hollow
   }

...but it did not help either (see below)!

Note that I obviously made the sphere too large to be still visible 
(black sky), but at least radiosity was calculated...

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download '2011-10-25 port whatmough residential building #1, take 191 - view from kitchen into atelier, modified andresdottir radi' (22 KB)

Preview of image '2011-10-25 port whatmough residential building #1, take 191 - view from kitchen into atelier, modified andresdottir radi'

> So I changed the pretrace_end to 0.002...
...
> ....and increased error_bound to 0.1 - aside from the scene now
> rendering somewhat faster, I found no advantages, there are still the
> same annoying artifacts (see below)! Perhaps I really should try the
> ambient sphere...

What you really should try is POV-Ray 3.7 - radiosity has changed a lot 
since 3.6.

You'll probably need to increase the radiosity count parameter at any 
rate though; the irregularities seen on the walls indicate that POV-Ray 
doesn't "catch" the bright patches of sunlight on the floor reliably.

In case render times increase unacceptably, there are some ways in 3.7 
to counter this increase to some degree. (In particular, you can define 
fully transparent objects to "mark" the patches of sunlight, set to 
"radiosity { importance 1.0 } no_image no_shadow no_reflection", while 
you set the walls to "radiosity { importance 0.1 }" or some such.)

Post a reply to this message

High!

On 25.10.2011 22:07, clipka wrote:

> What you really should try is POV-Ray 3.7

O.k., I downloaded the source code package from www.povray.org - but 
then, I failed already after I unpacked the zip file. There is no such 
thing as "configure" on my system!

Perhaps I'm just too stupid to compile complex C++ code under Linux...

See you in Khyberspace!

Yadgar

Post a reply to this message

> High!
>
> On 25.10.2011 22:07, clipka wrote:
>
>> What you really should try is POV-Ray 3.7
>
> O.k., I downloaded the source code package from www.povray.org - but
> then, I failed already after I unpacked the zip file.

That should be a *.tar.bz2 file actually. You won't get far with the 
Windows zip file on a Linux machine :-P

> There is no such thing as "configure" on my system!

There is - in the folder you extracted the POV-Ray sources to ;-) Just 
invoke "./configure" in that directory.

> Perhaps I'm just too stupid to compile complex C++ code under Linux...

... or you're simply not familiar with the typical procedures (yet).

Post a reply to this message

Hi(gh)!

On 26.10.2011 00:32, clipka wrote:

> That should be a *.tar.bz2 file actually. You won't get far with the
> Windows zip file on a Linux machine :-P

In fact it is! But I'm not even able to unpack the archive with ark 
correctly - all I get is an error message saying "libarchive is not able 
to process the archive file"!

See you in Khyberspace!

Yadgar

Post a reply to this message

Compare with this scene I happen to work on at this moment and rendered 
with version 3.7...

Thomas

Post a reply to this message

Attachments:
Download 'at the jackal_03.png' (683 KB)

Preview of image 'at the jackal_03.png'

> Hi(gh)!
>
> On 26.10.2011 00:32, clipka wrote:
>
>> That should be a *.tar.bz2 file actually. You won't get far with the
>> Windows zip file on a Linux machine :-P
>
> In fact it is! But I'm not even able to unpack the archive with ark
> correctly - all I get is an error message saying "libarchive is not able
> to process the archive file"!

I don't know that "ark" you're mentioning; do you mean arc? arj? ar? 
 From what I know, each of these three is specialized on one special 
type of compressed file respectively, none of which is the bzip2 format.

The traditional official way to extract the contents should be:

     bunzip2 povray-3.7.0.RC3.tar.bz2
     tar -xf povray-3.7.0.RC3.tar

The following single command should also do the job:

     tar -xjf povray-3.7.0.RC3.tar.bz2

Post a reply to this message