Op 20-1-2022 om 14:20 schreef Kenneth:
> I decided to go through all of the scene's include files-- turning off all of
> the light sources, correcting any out-of-gamut srgb colors to 1.0 max, and
> eliminating every object except the pentagram and the ground.
> 
> And I was still seeing the bright radiosity patches :-(
> 
> So I made my own VERY simplified scene-- posted below-- using just the bare
> necessities of Warren's code plus a few simplifications... and finally
> discovered what I think are the root causes of those bright patches. There are
> THREE reasons, as far as I can tell:
> 
> 1) The sky-sphere object is the primary cause. Comment-out that sphere, and the
> problem patches disappear.
> 
> 2) Radiosity's 'media ON' switch is also a primary problem; turn that OFF and
> the patches disappear.
> 
> 3) The media-pentagram is, I think, halfway embedded in the ground; that seems
> to definitely be adding to the problem, for some reason. Move it even slightly
> upward (like William P. mentioned), and the number of bright patches greatly
> decreases. Not completely, though, as 1) and 2) still produce some patches at
> random.
> 
> I don't know what is going on with rad's 'media on' switch; but the sky-sphere
> problem does not make any sense to me, as its colors are only 0.3 maximum...
> certainly not 'super bright'.
> 
> Maybe others here have some additional clues?
> 
[snip]

I have been very silent on this topic until now. A bit short on time.

I also came across your point (2) so thought it might (at least) be a 
radiosity/media problem. Then I thought that maybe the very thin value 
of the pentagram cylinders might play a role, so I made the cylinder 
radii much thicker while also setting intervals and samples to more 
default values and playing with sample values, but while renders where 
different, the (small) patches remained. I think that (3) might be an 
important issue here, but point (1) is really baffling.

Good work, Kenneth!

-- 
Thomas

Post a reply to this message

Le 2022-01-20 à 08:20, Kenneth a écrit :
> // This light is NOT the cause of the bright radiosity patches.
> light_source{<-2,4,-3 >*1000000 color srgb <0.5,0.5,0.6>*0.75 }
> 

WHY put that light so far away ? Why not use a much closer one, but with 
the parallel attribute ?

When an object or light is situated at a distance in the million units 
range, you can get unexpected behaviour.

Alternative light that will get the exact same result :
light_source{<-2,4,-3 >*1000 color srgb <0.5,0.5,0.6>*0.75 parallel}

Post a reply to this message

Thomas de Groot <tho### [at] degrootorg> wrote:
>
> ...but while renders were
> different, the (small) patches remained. I think that (3) might be an
> important issue here, but point (1) is really baffling.
>
> Good work, Kenneth!
>

Thanks. This has been my most intense 'detective work' in a long time-- and I'm
still thinking about all of it! That sphere problem really bugs me, but I think
I've discovered something interesting about it.

I LIKE a challenge (??) :-O

Post a reply to this message

Alain Martel <kua### [at] videotronca> wrote:
>
> WHY put that light so far away ? Why not use a much closer one...

I agree. And from more tests, I see that the sky-sphere's gigantic size has a
great deal to do with those super-bright radiosity patches-- which surprised me.

I ran an animation test using my posted 'simplified' code, while changing that
sphere scale from 8000 (which is not *too* large) down to 50. At 8000, the rad
patches are very bright; at 50, they are MUCH less bright. Currently, I don't
know why that should happen-- so I came up with yet another very simple test
scene of my own(!), which I post here. No light_sources in it. It produces the
same weird effect; change the sphere scale to see the difference. (I've also
attached an image example; if you can't see it here, it will nevertheless appear
in the 'image digest'.) Note that all of the scene's color values are purposely
very low.

What this tells me is that any scene with a media object in it, surrounded by
another larger rad-emitting object like that sphere, may show some unexpected
rad/media interactions somewhere, depending on the *scale* of the larger object
and the proximity of the media object to others. It's 'as if' the sphere is
causing brighter radiosity lighting from its colors, the larger it gets. I have
no other guess or explanation at this point. It makes me wonder if certain
radiosity values need tweaking, depending on how large a typical rad-emitting
'sky' sphere is (for example.)

Of course, one of the major reasons for the bright rad patches in all of the
posted scenes here is that the media objects are halfway-embedded in the
'ground'... so there's a proximity effect at work. But I wonder if subtle flaws
might also be occurring in 'typical' media/radiosity scenes that would be
similar to Warren's.

----
#version 3.8;
#declare RAD_ON = 1;
#declare P_start = 64 / image_width;
#declare P_end_final = 4 / image_width;

global_settings{
assumed_gamma 1.0

#if(RAD_ON)
radiosity{ // mostly Warren's settings
pretrace_start P_start
pretrace_end   P_end_final
error_bound 0.2
minimum_reuse 0.15 // Ken
maximum_reuse 0.151 // Ken
nearest_count 9
count 50
recursion_limit 1
always_sample off
//gray_threshold 0.6
brightness 1
adc_bailout 0.01/2
normal on
media on // [If OFF, it ELIMINATES the super-bright radiosity patches.]
         }
#end
}

camera {
  perspective
  location  <15, 70, -140>
  look_at   <0, 0,  -10>
  right     x*image_width/image_height
  angle 40
}

// the large sky-sphere
sphere{0, 1
pigment{rgb .1}
    finish{ambient 0 diffuse 0 emission 1}
    scale 8000 // change to 50
   // hollow on // does not seem to be necessary-- the media objects and
                // rad patches show up anyway
    no_image
}

box{0,1 translate -.5 scale <100,.01,100>
pigment{rgb <.3,.3,1>}
}

merge{ // or union
cylinder{-45*x,45*x,.3}
cylinder{-45*x,45*x,.3 rotate 120*y}
cylinder{-45*x,45*x,.3 rotate 240*y}
hollow
pigment{rgbt 1}
interior{
 media{
  emission .003*<.1,1,.1>
  method 3
  intervals 1
  samples 30
  }
 }
 //translate ... *y
}

Post a reply to this message

Attachments:
Download 'rad_sphere_scale_test_comp.jpg' (28 KB)

Preview of image 'rad_sphere_scale_test_comp.jpg'

Op 21-1-2022 om 15:52 schreef Kenneth:
> Alain Martel <kua### [at] videotronca> wrote:
>>
>> WHY put that light so far away ? Why not use a much closer one...
> 
> I agree. And from more tests, I see that the sky-sphere's gigantic size has a
> great deal to do with those super-bright radiosity patches-- which surprised me.
> 
> I ran an animation test using my posted 'simplified' code, while changing that
> sphere scale from 8000 (which is not *too* large) down to 50. At 8000, the rad
> patches are very bright; at 50, they are MUCH less bright. Currently, I don't
> know why that should happen-- so I came up with yet another very simple test
> scene of my own(!), which I post here. No light_sources in it. It produces the
> same weird effect; change the sphere scale to see the difference. (I've also
> attached an image example; if you can't see it here, it will nevertheless appear
> in the 'image digest'.) Note that all of the scene's color values are purposely
> very low.
> 
> What this tells me is that any scene with a media object in it, surrounded by
> another larger rad-emitting object like that sphere, may show some unexpected
> rad/media interactions somewhere, depending on the *scale* of the larger object
> and the proximity of the media object to others. It's 'as if' the sphere is
> causing brighter radiosity lighting from its colors, the larger it gets. I have
> no other guess or explanation at this point. It makes me wonder if certain
> radiosity values need tweaking, depending on how large a typical rad-emitting
> 'sky' sphere is (for example.)
> 
> Of course, one of the major reasons for the bright rad patches in all of the
> posted scenes here is that the media objects are halfway-embedded in the
> 'ground'... so there's a proximity effect at work. But I wonder if subtle flaws
> might also be occurring in 'typical' media/radiosity scenes that would be
> similar to Warren's.
> 
> ----
> #version 3.8;
> #declare RAD_ON = 1;
> #declare P_start = 64 / image_width;
> #declare P_end_final = 4 / image_width;
> 
> global_settings{
> assumed_gamma 1.0
> 
> #if(RAD_ON)
> radiosity{ // mostly Warren's settings
> pretrace_start P_start
> pretrace_end   P_end_final
> error_bound 0.2
> minimum_reuse 0.15 // Ken
> maximum_reuse 0.151 // Ken
> nearest_count 9
> count 50
> recursion_limit 1
> always_sample off
> //gray_threshold 0.6
> brightness 1
> adc_bailout 0.01/2
> normal on
> media on // [If OFF, it ELIMINATES the super-bright radiosity patches.]
>           }
> #end
> }
> 
> camera {
>    perspective
>    location  <15, 70, -140>
>    look_at   <0, 0,  -10>
>    right     x*image_width/image_height
>    angle 40
> }
> 
> // the large sky-sphere
> sphere{0, 1
> pigment{rgb .1}
>      finish{ambient 0 diffuse 0 emission 1}
>      scale 8000 // change to 50
>     // hollow on // does not seem to be necessary-- the media objects and
>                  // rad patches show up anyway
>      no_image
> }
> 
> box{0,1 translate -.5 scale <100,.01,100>
> pigment{rgb <.3,.3,1>}
> }
> 
> merge{ // or union
> cylinder{-45*x,45*x,.3}
> cylinder{-45*x,45*x,.3 rotate 120*y}
> cylinder{-45*x,45*x,.3 rotate 240*y}
> hollow
> pigment{rgbt 1}
> interior{
>   media{
>    emission .003*<.1,1,.1>
>    method 3
>    intervals 1
>    samples 30
>    }
>   }
>   //translate ... *y
> }
> 

Is there a reason why the sky sphere has emission 1 ? It is a night 
scene, so I would use emission 0 or emission0.01......

-- 
Thomas

Post a reply to this message

Le 2022-01-21 à 11:10, Thomas de Groot a écrit :

> Is there a reason why the sky sphere has emission 1 ? It is a night 
> scene, so I would use emission 0 or emission0.01......
> 

Emission 1 and rgb value of 0.1.
If it was rgb 1, then, a small emission value would be required.

Post a reply to this message

Le 2022-01-21 à 09:52, Kenneth a écrit :
> // the large sky-sphere
> sphere{0, 1
> pigment{rgb .1}
>      finish{ambient 0 diffuse 0 emission 1}
>      scale 8000 // change to 50
>     // hollow on // does not seem to be necessary-- the media objects and
>                  // rad patches show up anyway
>      no_image
> }

Have you tried to replace that large sphere with a sky_sphere ? Or even 
a simple background statement if the pigment is solid ?
Some options :

background{ rgb 0.1 }

sky_sphere{
	pigment{ rgb 0.1 }
   }

sky_sphere{
  pigment{plnar{
	[0 rgb 0.1]
	[1 rgb 0]
	}
    }

Post a reply to this message

Alain Martel <kua### [at] videotronca> wrote:

> > // the large sky-sphere
> > sphere{0, 1
> > pigment{rgb .1}
> >      finish{ambient 0 diffuse 0 emission 1}
> >      scale 8000 // change to 50
> >     // hollow on // does not seem to be necessary-- the media objects and
> >                  // rad patches show up anyway
> >      no_image
> > }
>
> Have you tried to replace that large sphere with a sky_sphere ? Or even
> a simple background statement if the pigment is solid ?
> Some options :
>
> background{ rgb 0.1 }
>
> sky_sphere{
>  pigment{ rgb 0.1 }
>    }
>

Good suggestions. I tried both of those as a substitute for the large sphere--
and now there are NO bright rad-patch artifacts, in either case! So that sphere
(and its scale) seems to be causing a real/mysterious problem.

I tried one more simple test of that sphere: *creating* it with a radius of 8000
vs. *scaling* it to be so. But that didn't solve the problem.

Thanks Alain; this result *might* indicate a bug, regarding use of such an
actual sphere along with radiosity's 'media ON' switch.

Post a reply to this message

> Alain Martel <kua### [at] videotronca> wrote:
> >
> > Have you tried to replace that large sphere with a sky_sphere ? Or even
> > a simple background statement if the pigment is solid ?
>
> Good suggestions. I tried both of those as a substitute for the large sphere--
> and now there are NO bright rad-patch artifacts, in either case!

Well, I spoke too soon. The reason I didn't see any patches-- or any rad/media
interaction at all-- was because my sky_sphere color and emission color were
just too dim to produce them. In my most recent test code posted earlier, use
the sky_sphere at rgb 0.1, and try emission 1*<0.1,1.0,0.1> to see the media
cylinders, then emission 250*<0.1,1.0,0.1> to see the bright patches again.

So that media object being half-embedded seems to be the real problem... along
with rad's 'media ON' switch.

But I find it strange that the sky_sphere produces its own particular radiosity
brightness, versus the enclosing-sphere object imparting its *variable* rad
brightness depending on its size.

A mystery for now.

Post a reply to this message

Le 2022-01-24 à 03:05, Kenneth a écrit :
>> Alain Martel <kua### [at] videotronca> wrote:
>>>
>>> Have you tried to replace that large sphere with a sky_sphere ? Or even
>>> a simple background statement if the pigment is solid ?
>>
>> Good suggestions. I tried both of those as a substitute for the large sphere--
>> and now there are NO bright rad-patch artifacts, in either case!
> 
> Well, I spoke too soon. The reason I didn't see any patches-- or any rad/media
> interaction at all-- was because my sky_sphere color and emission color were
> just too dim to produce them. In my most recent test code posted earlier, use
> the sky_sphere at rgb 0.1, and try emission 1*<0.1,1.0,0.1> to see the media
> cylinders, then emission 250*<0.1,1.0,0.1> to see the bright patches again.
> 
> So that media object being half-embedded seems to be the real problem... along
> with rad's 'media ON' switch.
> 
> But I find it strange that the sky_sphere produces its own particular radiosity
> brightness, versus the enclosing-sphere object imparting its *variable* rad
> brightness depending on its size.
> 
> A mystery for now.
> 
> 
> 
> 

New suggestion :
Scale your media container by <0.5, 1, 1> and move them up so that it is 
entirely above the surface. Even leave a tiny gap between it and the 
surface.

Post a reply to this message