Page 1 of 2
[BUG] No specular reflections on glass from sun
Posted: Thu May 17, 2007 8:49 am
by Caronte
I think this can be a bug (or I don't understand) because in real life I can see it
The only way I was found to get specular reflections over glass from the sun, are using a "Spec type=glossy transparent" and set the exponent value to something like 1500 but thi's not a solution because the glass becomes frosted.
From this post:
http://www2.indigorenderer.com/joomla/f ... highlight=
Posted: Thu May 17, 2007 10:42 am
by neepneep
This isn't a bug but I am guessing the problem is that the sun in the sky model is a little small.
What you can do is create a blend material of both the glossy and the regular specular materials and then fiddle with the blend factor until you get a better result...
eg.
Code: Select all
<medium>
<name>Spec</name>
<precedence>10</precedence>
<basic>
<ior>1.5</ior>
<cauchy_b_coeff>0</cauchy_b_coeff>
<absorption_coefficient_spectrum>
<rgb>
<rgb>0.001 0.001 0.001</rgb>
</rgb>
</absorption_coefficient_spectrum>
</basic>
</medium>
<material>
<name>Spec</name>
<specular>
<transparent>true</transparent>
<internal_medium_name>Spec</internal_medium_name>
</specular>
</material>
<medium>
<name>Gloss</name>
<precedence>10</precedence>
<basic>
<ior>1.5</ior>
<cauchy_b_coeff>0</cauchy_b_coeff>
<absorption_coefficient_spectrum>
<rgb>
<rgb>0.001 0.001 0.001</rgb>
</rgb>
</absorption_coefficient_spectrum>
</basic>
</medium>
<material>
<name>Gloss</name>
<glossy_transparent>
<exponent>2000</exponent>
<internal_medium_name>Gloss</internal_medium_name>
</glossy_transparent>
</material>
<material>
<name>blend</name>
<blend>
<a_name>Gloss</a_name>
<b_name>Spec</b_name>
<blend_factor>0.85</blend_factor>
</blend>
</material>
Posted: Thu May 17, 2007 11:03 am
by zsouthboy
Your scene scale is wrong.
Posted: Thu May 17, 2007 12:53 pm
by CTZn
1 unit = 1 meter
Just in case
zsouthboy is right
Posted: Thu May 17, 2007 3:04 pm
by OnoSendai
Hi,
While not necessarily a bug, this is definitely an 'issue' with Indigo.
The sun is the correct size, the scene is modelled just fine.
The problem is that the sun subtends a very small fraction of the sky. Combined with the curvature of the sphere, the result is that the sun image covers only a small fraction of a pixel.
I guess the underlying divergence from 'real life', is that in the human eye, or a camera, there will be a number of mechanisms by which the sun image is spread out a bit over the rest of the image - diffraction effects, scattering off particles in the eye / camera, imperfections on the lenses etc..
There is also a biological 'veiling glare' effect due to retinal neurons getting overloaded, and an analogous CCD 'overflow' phenomenon.
I'm still thinking about the best way to handle this issue. For example, see rerdavies' threads in the Violet dev section as one possible approach that could be integrated into the Indigo core.
In the mean time, I suggest you save out a HDR image as an .igi. Then you can experiment with Gaussian bloom, glare, and diffractive bloom, which should make the sun specular reflections a lot more visible
Posted: Thu May 17, 2007 4:25 pm
by Kram1032
there is another issue with the sun, that I already pointed out, once...
if you point the cam directly on the sun, you can see a black pixel line arround it...
Posted: Thu May 17, 2007 8:33 pm
by CTZn
While not necessarily a bug, this is definitely an 'issue' with Indigo.
The sun is the correct size, the scene is modelled just fine.
In that sense I don't see a reason why to "fix" that. Burn parameter should be set accordingly by users, this is related with tonemapping isn't it ?
Posted: Thu May 17, 2007 8:49 pm
by OnoSendai
CTZn wrote:While not necessarily a bug, this is definitely an 'issue' with Indigo.
The sun is the correct size, the scene is modelled just fine.
In that sense I don't see a reason why to "fix" that. Burn parameter should be set accordingly by users, this is related with tonemapping isn't it ?
It's an issue with Indigo because the goal of Indigo is photorealistic graphics.
And in a photo the sun image will be larger than 1 pixel.
Posted: Thu May 17, 2007 9:32 pm
by crsrma
Hmm...what's this biological talk, when most everything in the XML/exporters is based on shutters & f-stop?
edit: Wouldn't experimenting with creating more pronounced specular [highlights] create unwanted hot spots as well?
Posted: Thu May 17, 2007 9:49 pm
by Kram1032
the sun would look larger... I don't think it would affect all the highlights, 'cause it's brightness based...
Dunno, if that's right...
Posted: Thu May 17, 2007 10:04 pm
by crsrma
Maybe sky turbidity would matter as well..?
I've had instances of this myself, basically in glassy/glossy scenes.
Posted: Thu May 17, 2007 10:10 pm
by Kram1032
not really...
sky turbity won't upsize the sun appearance. It just changes the colour of the sky.
And the physically accurate range of turbity is 2-6. Lower or higher values will give unrealistic skys.
But a new skymodel is comming
(I don't write soon, just that it's comming)
Posted: Fri May 18, 2007 2:55 am
by Caronte
OnoSendai wrote:It's an issue with Indigo because the goal of Indigo is photorealistic graphics.
And in a photo the sun image will be larger than 1 pixel.
¡Exactly!
I think the goal of a realistic render engine would be to produce the same results than we can see with our eyes or capture with our camera.
OnoSendai, I tryed Bloom, Glare, etc... but thi's not god, because it affect another image zones.
I hope you find a good solution to solve it (may be doing a diferent treatment to the sun because thi's the only problematic lightsource?).
Thanks by your time
Posted: Fri May 18, 2007 7:05 am
by Maluminas
Sky turbidity is important, its what is responsible for the sun bloom in real life. The turbidity scatters the light and thats why the sun can be seen in real life as a big painfully bright blotch. I would be interested to see what would happen if someone made a 300m (indigo world size?) wide sphere with a very low scattering medium, im pretty sure we'd get sun bloom on the sky and on reflections... I might try it if i find the time later today.
Posted: Fri May 18, 2007 7:08 am
by Kram1032
nope. not indigo world size
that was once.
now, it's everytime 100*scene size.
So, if that overall fog-cod doesn't come, a perfect matching worldsized fog sphere isn't possible.
All this may come with the new skymodel (I hope)