> clipka<ano### [at] anonymousorg>  wrote:
>> I'd suggest radiosity with a very (/really/ very) high sample count,
>> setting any objects other than the planet to "no_radiosity" to avoid
>> wasting intersection tests on them for the sake of radiosity sampling.
>> Needs POV-Ray 3.7 though.
>
> Thanks a lot for your suggestions! With all of your help, I have achieved quite
> satisfying results in v. 3.6.1 using radiosity by increasing the objects'
> diffuse terms instead of the sample count. But I'm afraid things are about to
> get a lot more complex...
>
> The only remaining problem is, my scene includes an atmosphere around the planet
> using interior&  media. The moment I turn my atmosphere on, even if its diffuse,
> extinction,&  attenuation are all turned off and I have turned off media in the
> global radiosity settings, my "caustics" disappear.
>
> Do I need to follow your directions after all, about high sample counts in v.
> 3.7?
>
> For clarity, below is my ACTUAL scene code, including the atmosphere. Apologies
> for the bitmaps, which I cannot attach. The "atmos_layers" variable is used as a
> kind of quality setting, since a photo-ready number of layers takes too long to
> render while I'm actively working on the scene. Set it to zero to remove the
> atmosphere.
>
Replaced your images with pattern and colour_map having a good amount of 
transparency.

Your looks_loke is good. A looks_like object is located relative to it's 
light_source and moves with it if you change the coordinates.

Why is your sun and planets both at coordinate 0?

With 3.6.1, there was a bug in radiosity: The recursion level had to 
take transparent surfaces into consideration. You have 4 athmospheric 
layers and recursion_level default to 3. Then, you have two sets of 
layers plus the actual surface, for a total of 9.

That said, I realy think that you should use only one athmospheric shell 
containing only scattering media. You then modulate the density of that 
media as needed, using the spherical pattern scalled to the radius of 
the athmonsphre or some custom function.
You then use one or more cloud shells. Those are hollow and don't have 
their own media.

fade_power 10? for your athmosphere?
A realistic fade_power is 1 for substances to use linear fading, or 1001 
to use the exponential formula.
Test without fading gives beter results.
It's fade_power 2 for a fading light_source.

Using version 3.7RC3, the radiosity works as expected.

Keep in mind that radiosity can't illuminate your media.
media on will enable emissive media and directly illuminated scattering 
media to affect the radiosity illumination of nearby objects.



Alain

Post a reply to this message