Hi!

Is it mentioned anywhere in the docs how much the chosen scale affects 
accuracy? I am doing a simple room modeling and chosen an expected scale 
of 1 unit = 1 metre. I get a few artifacts  and it looks like if I 
multiply everything by 10 or 100 it can get rid of those. The smallest 
objects in my scene have dimension of about 0.005 units.

So what is the preferred scale or the smallest preferable dimension? How 
can one ensure that POV-Ray can be the most accurate?

Severi Salminen

Post a reply to this message

Hi Severi,

this is a problem I've suffered from too. For some
extreme examples (and some hints) see in povray.general:

   Intersection with quartic: bug?   3 Sep. 2002
   Small is NOT BEAUTIFUL           11 Feb. 2003

The reason for this behavior are some 'Epsilon' values in the
Pov-Ray code which are neccessary (in this way some overlap is
created at corners to make them 'watertight', without such overlap
it would  sometimes be possible to look inside objects at corners
due to limited numerical precision), but these 'Epsilons' are a
little large.

   Sputnik

Post a reply to this message

> this is a problem I've suffered from too. For some
> extreme examples (and some hints) see in povray.general:
> 
>    Intersection with quartic: bug?   3 Sep. 2002
>    Small is NOT BEAUTIFUL           11 Feb. 2003
> 
> The reason for this behavior are some 'Epsilon' values in the
> Pov-Ray code which are neccessary (in this way some overlap is
> created at corners to make them 'watertight', without such overlap
> it would  sometimes be possible to look inside objects at corners
> due to limited numerical precision), but these 'Epsilons' are a
> little large.

Thanks for the info. I actually tried and multiplied everything in my 
secene by 100 so that the smalles dimension is now about 0.5 units and 
largest is about 600 units and it looks like all the artifacts 
disappeared. I also noticed some very slight changes with radiosity but 
I believe the result is now more accurate.

In my opinion this is something that should be mentioned in the 
documents (if it is: good, then forget this) - if any rule of thumb can 
be given in this matter. People might spend too much time searching 
coincident surfaces when the problem is somewhere else.

Severi Salminen

Post a reply to this message

"Severi Salminen" <sev### [at] NOT_THISsibafi> wrote in message
news:3fd9da81$1@news.povray.org...
>
> In my opinion this is something that should be mentioned in the
> documents (if it is: good, then forget this) - if any rule of thumb can
> be given in this matter. People might spend too much time searching
> coincident surfaces when the problem is somewhere else.
>

for a similiar type of problem, try (modify the plane's translation to, say,
0.002, to fix it):

#version 3.5;

#include "colors.inc"

camera {
  location  <0.0, 1.0, -5.0>
  look_at   <0.0, 1.0,  0.0>
}

plane{y,-0.001 pigment{checker pigment{Black}, pigment{White}}}

sphere{0,1
  translate y*1
  pigment{rgbt 1}
}

Post a reply to this message

Severi Salminen <sev### [at] NOT_THISsibafi> reports:

> [...] I get a few artifacts  and it looks like if
> I multiply everything by 10 or 100 it can get rid of those. The
> smallest objects in my scene have dimension of about 0.005 units.

Don't make your objects too big, though.

Consider the following scene.  It should show nine spheres, each
ten times further away and ten times the size of its predecessor.

POV-Ray 3.5 says:
    "Scene contains 9 frame level objects; 5 infinite."

Not only does POV-Ray consider objects of size 10000 and above as
infinite.  If I render the scene, I see only seven spheres...

--------------------------------------------------------------
#declare R  =  1;
#declare A  =  -20;

#macro S()
  sphere {0,R
    texture {pigment {bumps} finish {ambient 1} translate y}
    translate 5*R*z
    rotate    A*y
  }
#end

light_source {<10,10,-100> color rgb 1}
camera       {location 0  look_at z}

#while (A <= 20)
  S()
  #declare R  = R * 10;
  #declare A  = A +  5;
#end
--------------------------------------------------------------

Some months ago I tried to add the moon to a scene, with "correct"
diameter and distance, given in metres.  It took me some time to figure
out why it wasn't visible at all....
-- 
Cheers,
haj

Post a reply to this message

Harald Joerg wrote:

> Severi Salminen <sev### [at] NOT_THISsibafi> reports:
> 
> 
>>[...] I get a few artifacts  and it looks like if
>>I multiply everything by 10 or 100 it can get rid of those. The
>>smallest objects in my scene have dimension of about 0.005 units.
> 
> 
> Don't make your objects too big, though.
> 
> Some months ago I tried to add the moon to a scene, with "correct"
> diameter and distance, given in metres.  It took me some time to figure
> out why it wasn't visible at all....

Exactly. It would be nice to know what are these bounds, and are they 
fixed or somehow relative to each other.

Severi

Post a reply to this message

In article <wk8### [at] oookmuunetde> , Harald Joerg 
<haj### [at] oookmuunetde>  wrote:

> Not only does POV-Ray consider objects of size 10000 and above as
> infinite.  If I render the scene, I see only seven spheres...

<http://tag.povray.org/povQandT/languageQandT.html#largescaleproblems>


____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde

Visit POV-Ray on the web: http://mac.povray.org

Post a reply to this message

In article <3fdaf2fb$1@news.povray.org> , Severi Salminen 
<sev### [at] NOT_THISsibafi>  wrote:

> Exactly. It would be nice to know what are these bounds, and are they
> fixed or somehow relative to each other.

This information would only confuse you, and some issues brought up in this
thread are completely unrelated anyway.  I would recommend not to jump to
any conclusions based on what has been said or mentioned in this thread...

    Thorsten

____________________________________________________
Thorsten Froehlich
e-mail: mac### [at] povrayorg

I am a member of the POV-Ray Team.
Visit POV-Ray on the web: http://mac.povray.org

Post a reply to this message

Thorsten Froehlich wrote:

>>Exactly. It would be nice to know what are these bounds, and are they
>>fixed or somehow relative to each other.
> 
> 
> This information would only confuse you, and some issues brought up in this
> thread are completely unrelated anyway.  I would recommend not to jump to
> any conclusions based on what has been said or mentioned in this thread...
> 

Oh, please don't make any guesses what would confuse me or what would 
not. I'd be very grateful if you could provide the information to 
clarify this matter. I might even be able to understand it :)

And in addition: how much is this a platform/system architecture 
dependent thing?

Severi Salminen

Post a reply to this message

In article <3fdb1a32$1@news.povray.org> , Severi Salminen 
<sev### [at] NOT_THISsibafi>  wrote:

>> This information would only confuse you, and some issues brought up in this
>> thread are completely unrelated anyway.  I would recommend not to jump to
>> any conclusions based on what has been said or mentioned in this thread...
>
> Oh, please don't make any guesses what would confuse me or what would
> not. I'd be very grateful if you could provide the information to
> clarify this matter. I might even be able to understand it :)

The very fact that you ask the question implies the probability you can
understand it is low simply because you lack the background.  This
information is not going to be made available in a general manner.  It will
only result in confusion by those who don't understand the implications of
the information or the background.  You can easily find out the details by
taking a look at the source code:  If you cannot find it there on your own,
you won't be able to make use of anything anybody can tell you about it
anyway.

    Thorsten

____________________________________________________
Thorsten Froehlich, Duisburg, Germany
e-mail: tho### [at] trfde

Visit POV-Ray on the web: http://mac.povray.org

Post a reply to this message