Hello friends.

When I first rendered my povlab scene with radiosity, I noticed the dirt
layers of textures (fire place, casks) are not very well rendered: it has a
binary (all/nothing) aspect. In the code, the dirt layers are progressive,
and when rendered without radiosity it is OK, but when rendered with
radiosity, they are not OK for the objects-indirect lighting (but seems OK
for light_sources).

Is it a normal radiosity side effect?

In order to illustrate, here is the cask used in my scene without and with
radiosity. I also just noticed that specularity does not show. In the
radiosity version I replaced the light_source by a sphere with an ambient
of 100 and set the count parameter to 1000 in order to avoid too much
splotchiness.


    Bruno

Post a reply to this message

Attachments:
Download 'cask_radio_pb.jpg' (24 KB)

Preview of image 'cask_radio_pb.jpg'

"Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> Is it a normal radiosity side effect?
Is your dirt a layered texture? I think there's a separate max_trace_level
setting in the radiosity block which would need to be the same as the
global one. I haven't tried this, though, so I could be wrong...

This would also explain why transparent objects seem to block radiosity
for some people (there was a post to this effect recently).

> In order to illustrate, here is the cask used in my scene without and with
> radiosity. I also just noticed that specularity does not show.
Specular and phong highlights are calculated from light sources alone, I
think. If you removed the light sources and lit the scene using only
radiosity then this is to be expected.

I did notice the granularity in your original image, but I wasn't sure if it
was the desired effect or not!

Bill

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> "Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> > Is it a normal radiosity side effect?
> Is your dirt a layered texture? I think there's a separate max_trace_level
> setting in the radiosity block which would need to be the same as the
> global one. I haven't tried this, though, so I could be wrong...
>
> This would also explain why transparent objects seem to block radiosity
> for some people (there was a post to this effect recently).
>
> > In order to illustrate, here is the cask used in my scene without and with
> > radiosity. I also just noticed that specularity does not show.
> Specular and phong highlights are calculated from light sources alone, I
> think. If you removed the light sources and lit the scene using only
> radiosity then this is to be expected.
>
> I did notice the granularity in your original image, but I wasn't sure if it
> was the desired effect or not!
>
> Bill

Hi Bill!

The texture of the cask is quite complex: inside/outside with several layers
(base wood (when brand new) + 'wheathering' (old aspect) + dirt). The dirty
outside texture is:

// Base wood pigment (brand new cask)
#declare P_CaskBodyBase =  pigment
{
    P_WoodGrain14A
    color_map {M_Wood7A}
    rotate 90*x
    scale <0.2, 1, 0.2>
}


// Base texture for outsides
#declare T_CaskBodyBaseOut = texture
{
    pigment {P_CaskBodyBase}
    normal {granite 1 scale <0.04,1,0.04>*4} // Some relief in the wood
}
texture // Wheathering texture
{
    pigment
    {
        granite
        scale <0.04,1,0.04>*4
        color_map
        {
            [0 color 0.3*ForestGreen transmit 0.1]
            [0.3 color 0.4*DarkWood transmit 0.2]
            [0.7 color 0.7*DarkWood transmit 0.3]
            [1 transmit 0.8]
        }
    }
    finish {specular 0.2 roughness 0.04 } // Somewhat polished by
manipulations
}

// Dirt layer at both ends
#declare T_CaskBodyDirt = texture
{
    pigment
    {
        gradient y
        scale -2*CASK_LOW_THRESHOLD // Cask = differenced elongated spheres,
intersected with 2 planes
        translate -CASK_LOW_THRESHOLD*y
        turbulence 0.1 omega 0.65 octaves 8
        scale 1.1*<0.15,1,0.15>
        color_map
        {
            [0   transmit 0.05]
            [0.2 transmit 0.1]
            [0.5 transmit 0.3]
            [0.8 transmit 0.1]
            [1   transmit 0.05]
        }
    }
}


#declare T_CaskBodyBaseOutDirt =
  texture {T_CaskBodyBaseOut}
  texture {T_CaskBodyDirt}


The dirt layer in the first encountered by the rays (like in real life). I
read the documentation and I did not see a specific max_trace_level in the
radiosity section, but rather adc_bailout, which is closely related. I
tried to tweak this parameter from 1 to 1e-5), but the problem remains.

    Bruno

Post a reply to this message

"Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> read the documentation and I did not see a specific max_trace_level in the
> radiosity section, but rather adc_bailout, which is closely related. I
> tried to tweak this parameter from 1 to 1e-5), but the problem remains.

Yes, despite some obscure Google references to max_trace_level in
radiosity{}, it appears that I am in fact talking out of my backside. There
is no such setting.

However, I conducted some experiments with an upturned glass bowl and
radiosity lighting. It would appear that both max_trace_level and
recursion_limit must be set high enough for the lighting to behave
correctly. Try adjusting the latter upwards...

Bill

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> "Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> > read the documentation and I did not see a specific max_trace_level in the
> > radiosity section, but rather adc_bailout, which is closely related. I
> > tried to tweak this parameter from 1 to 1e-5), but the problem remains.
>
> Yes, despite some obscure Google references to max_trace_level in
> radiosity{}, it appears that I am in fact talking out of my backside. There
> is no such setting.
>
> However, I conducted some experiments with an upturned glass bowl and
> radiosity lighting. It would appear that both max_trace_level and
> recursion_limit must be set high enough for the lighting to behave
> correctly. Try adjusting the latter upwards...
>
> Bill

I am currently looking at the POV source code (radiosit.cpp) and trying to
understand what may happen (presence of thresholds ...). There is indeed a
Radiosity_Trace_Level variable in radiosit.h which seems to be initialized
to 1.

My recursion_limit is 3 and max_trace_level is 20, so it should be OK. It
must be something else... Still looking and experimenting. Thanks for your
attention!

    Bruno

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> "Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
> > read the documentation and I did not see a specific max_trace_level in the
> > radiosity section, but rather adc_bailout, which is closely related. I
> > tried to tweak this parameter from 1 to 1e-5), but the problem remains.
>
> Yes, despite some obscure Google references to max_trace_level in
> radiosity{}, it appears that I am in fact talking out of my backside. There
> is no such setting.
>
> However, I conducted some experiments with an upturned glass bowl and
> radiosity lighting. It would appear that both max_trace_level and
> recursion_limit must be set high enough for the lighting to behave
> correctly. Try adjusting the latter upwards...
>
> Bill

I am currently looking at the POV source code (radiosit.cpp) and trying to
understand what may happen (presence of thresholds ...). There is indeed a
Radiosity_Trace_Level variable in radiosit.h which seems to be initialized
to 1.

My recursion_limit is 3 and max_trace_level is 20, so it should be OK. It
must be something else... Still looking and experimenting. Thanks for your
attention!

    Bruno

Post a reply to this message

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Bill Pragnell wrote:
> "Bruno Cabasson" <bru### [at] alcatelaleniaspacefr> wrote:
>> In order to illustrate, here is the cask used in my scene without and with
>> radiosity. I also just noticed that specularity does not show.
> Specular and phong highlights are calculated from light sources alone, I
> think. If you removed the light sources and lit the scene using only
> radiosity then this is to be expected.
> 
	To be more precise, in real life, highlights are caused by blurred
reflection of the light-emitting object. Since light-sources (even
area lights) are single points, no amount of reflection blurred or
not would show them. Hence, the phong and specular keywords that
intend to fake this effect.

	When you use a radiosity light emitting object, this object is
large enough and bright enough that you can do the physically
accurate thing: add a slight amount of blurred reflection to your
object and the highlight should appear.

		Jerome
- --
+------------------------- Jerome M. BERGER ---------------------+
|    mailto:jeb### [at] freefr      | ICQ:    238062172            |
|    http://jeberger.free.fr/     | Jabber: jeb### [at] jabberfr   |
+---------------------------------+------------------------------+
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFF7dB+d0kWM4JG3k8RAgvKAJ458gMAWrUUeSy8wjtmlVo+jNLoEQCcD10W
Io5ctFPyt1RXtWspClvH3+w=
=R1R4
-----END PGP SIGNATURE-----

Post a reply to this message