The problem seems to be that radiosity, like media, is distance-sensistive (i.e.
it performs differently at different scales). Apparently, the best distances
(meaning the size of the room for instance) are the ones used in the radiosity
example, so I use them as a basis. What I do is to give a general scale to the
whole scene (including the lights and the camera) and use this general scaling
factor to tune the picture for the best quality vs render time. I ran in exactly
the same problems as the ones you mention (plus the fact that a very low
error_bound seemed to crash the picture due to a lack of memory) and these
problems disappeared using the standard radiosity settings (in rad.inc)  with
the "right" distances. And the good news was that radiosity was not that slow. I
also noted that some radiosity artifacts are nicely smoothed by antialiasing.
I hope this helps.
Gilles

Nieminen Mika wrote:

>   I'm trying to find proper radiosity settings for an indoors scene.
> However, no matter what I do, I get one of the following:
>   1. Very blotchy illumination, like someone was throwing paint on the
> walls at random.
>   2. Quite smooth illumination but somewhat grainy and extremely annoying
> small dark spots in corners.
>
>   I get the second type of illumination with settings which take centuries
> to calculate (in my P-II 350MHz). If I use a little bit faster settings
> I get the first type of illumination.
>   Those dark spots in the second case appear no matter how small
> error_bound I specify (I have gone as small as error_bound 0.05).
>   The distance_maximum seems to be the key value which controls how much
> the image goes to the first or the second case. With large distance_maximum
> values the image gets blotchy while with small values it gets grainy.
> I have tried several values between 1000 (the room is about 200 units wide)
> and 1. With very small values the render time blows up and the walls look
> like they had a finish { crand .2 }.
>   The rendering time doesn't matter, but I just can't get rid of the
> graininess and the annoying dark spots in the corners. The only way I
> can do this is setting so bad values that the image gets blotchy, but that's
> not a very good solution.
>   Does anyone has any good advice?
>
>   It's funny that I got really good values for the radiosity test images
> in my web page, but I can't find them for this image (of course the settings
> for those radiosity test images do not work with this image).
>
> --
> main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
> ):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

Any idea how to rescale the whole scene (with thousand of objects)?

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

#declare sc=10;
//#declare sc=0.10;
union{
    all objects here
    scale sc
}
camera{
    location PointLoc*sc
    camera stuff
    look_at PointLook*sc
}

light_source{PointLight*sc color LightColor}

There may some other scaling involved if you use fog or fade_distance but it
should be straightforward. Media will be a problem, though.

G.

Nieminen Mika wrote:

>   Any idea how to rescale the whole scene (with thousand of objects)?
>
> --
> main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
> ):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

There should be a keyword in povray like:

global_settings { global_scale .1 }

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

Nieminen Mika wrote:
> 
>   There should be a keyword in povray like:
> 
> global_settings { global_scale .1 }

  It might help all of those people that keep trying to model the solar
system in real world units and then run into the problem of max epsilon
values being exceeded.

-- 
Ken Tyler

mailto://tylereng@pacbell.net

Post a reply to this message

why is there a max value anyways i wanted to make a solar system animation
:(

Ken wrote:

> Nieminen Mika wrote:
> >
> >   There should be a keyword in povray like:
> >
> > global_settings { global_scale .1 }
>
>   It might help all of those people that keep trying to model the solar
> system in real world units and then run into the problem of max epsilon
> values being exceeded.
>
> --
> Ken Tyler
>
> mailto://tylereng@pacbell.net

Post a reply to this message

Noah A <vip### [at] powersurfrcom> wrote:
: why is there a max value anyways i wanted to make a solar system animation
: :(

  The computer can handle only finitely large numbers.
  In a typical 32-bit computer (like PC) a floating point value takes 32
bits and a double value 64 bits. As you can see, they can only represent
certain amount of different values (which should go from as small values
as possible to as large values as possible with as small steps as possible).
  With a 64-bit value you can represent pretty large amount of numbers, but
only a finite amount of them (and sometimes a too little amount). With
these same 64 bits it should be possible to represent numbers like
0.000000001 and 100000000.
  Also the resolution has to be limited. You may be able to represent the
number 100000000 and the number 100000002 but not the number 100000001
because there's not enough bits for such a detailed precision.
  This precision problem is the reason why we need an epsilon value. Suppose
the previous case. If we have a=100000000 and b=1 and we calculate their
addition a+b, the result is not 100000001 because it can't be represented
in the floating point format, but the result is 100000000 instead.
Because of this comparing two floating point numbers together may be
pretty inaccurate. So what we do is making the comparation with certain
precision. This means that if two numbers are close enough to each other,
they are considered the same. The epsilon value states this precision.

  There's absolutely no need to represent the solar system taking a metre
as the measure unit. You can perfectly use, for example 1000 km as the unit
or even 1 AU (Astronomical Unit, ie. the distance from the Sun to the Earth).
  Actually it doesn't make sense to use a too small unit because you don't
have the distances between planets at that precision. You will only be
wasting valuable bits. It's like taking the micrometre as measure unit when
all the distances are measured in metres. There will only be six useless
zeros after each value.

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

One day I'll actually look at the source myself (I've got it lying around on
some partition somewhere), 'till then, can anyone tell me what precise
format POV does use? Is it 32bit, 64bit or something else?  Does this vary
from platform to platform?  If it is less than 64bit, when will this be
changed?

Post a reply to this message

Edward C. <edw### [at] hotmailcom> wrote:
: One day I'll actually look at the source myself (I've got it lying around on
: some partition somewhere), 'till then, can anyone tell me what precise
: format POV does use? Is it 32bit, 64bit or something else?  Does this vary
: from platform to platform?  If it is less than 64bit, when will this be
: changed?

  AFAIK povray uses the 'double' type for most of its floating point
calculations. This type is 64 bits in most computers.

-- 
main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message

You might try upping the count value, as this will improve the quality. Also,
I've found that a good trick for setting distance_maximum is to use one third of
the distance between the camera location and the look_at. Finally, if you add a
bit of normal to your wall, like a bozo .2, the shadow banding is much less
noticable.

Nieminen Mika wrote:

>   I'm trying to find proper radiosity settings for an indoors scene.
> However, no matter what I do, I get one of the following:
>   1. Very blotchy illumination, like someone was throwing paint on the
> walls at random.
>   2. Quite smooth illumination but somewhat grainy and extremely annoying
> small dark spots in corners.
>
>   I get the second type of illumination with settings which take centuries
> to calculate (in my P-II 350MHz). If I use a little bit faster settings
> I get the first type of illumination.
>   Those dark spots in the second case appear no matter how small
> error_bound I specify (I have gone as small as error_bound 0.05).
>   The distance_maximum seems to be the key value which controls how much
> the image goes to the first or the second case. With large distance_maximum
> values the image gets blotchy while with small values it gets grainy.
> I have tried several values between 1000 (the room is about 200 units wide)
> and 1. With very small values the render time blows up and the walls look
> like they had a finish { crand .2 }.
>   The rendering time doesn't matter, but I just can't get rid of the
> graininess and the annoying dark spots in the corners. The only way I
> can do this is setting so bad values that the image gets blotchy, but that's
> not a very good solution.
>   Does anyone has any good advice?
>
>   It's funny that I got really good values for the radiosity test images
> in my web page, but I can't find them for this image (of course the settings
> for those radiosity test images do not work with this image).
>
> --
> main(i,_){for(_?--i,main(i+2,"FhhQHFIJD|FQTITFN]zRFHhhTBFHhhTBFysdB"[i]
> ):5;i&&_>1;printf("%s",_-70?_&1?"[]":" ":(_=0,"\n")),_/=2);} /*- Warp -*/

Post a reply to this message