Thanks Hummel!!!

Btw, I have just a rough idea what spherical harmonics are and why and for what they are used for -- but can YOU recommend a paper/article/whatelse, that describes how to sample e.g. from an illuminated sphere a SH and how to calculate the lighting terms in a shader?

I want to try out two applications:

1.) placing small dummy-spheres as lightprobes in a level, so that they bake the (radiosity) lighting at that world position into a texture - which I in return convert into a leightweight SH, to be used ingame

2.) editing a SH for the whole ambient lighting in realtime by adjusting 6 directional lights on a virtual sphere by their intensity and color... question is then how to translate that into a SH. Realtime baking would be possible, too, so that I could simply apply the technique for case 1.)

Regards,
-Christian

I added some papers related to SHs just now, but I would strongly recommend you to check the section about SHs in "Real-Time Rendering" (always check this one if you come up with questions about fundamental real-time cg stuff) and "Physically Based Rendering" (if you got this one too in the meantime). Badly I need to get productive regarding our Uni software project so I've not time to give you a personal explanation atm.

Does someone knows a good paper about the advantages and disadvantaes about post-processing distance fog?

Isn't distance fog always pp? or is there another technique to render it?

Distance fog is relatively cheap. You just calculate a blending factor based on the distance between the camera and the pixel and two distance thresholds (no fog and total fog). That is so cheap that it has virtually no impact on any framerate on the planet.

Well you can combine it with the object shader to improve fps.
If the pixel is "deeper" than fog_far - so it's fully covered in fog - then only the fog color has to be returned without calculating all the lighting / normalmapping.

For very complex object shaders this CAN be good but in this case I suggest to combine the shaders with the LOD-system and "turn off" the fancy shading for objects that are far away.

Besides that total-fog doesn't look good at all for 'normal conditions'.

Clipping away pixels if they are beyond the far-fog plane is a must for object shaders; besides, it makes sense to equalize the far-clip plane with the far-fog plane. --- When I think about it, there are actually some disadvantages with PP-based distance fog in some cases.

1) you need a depthmap
2) if you have alpha-objects, you are screwed

Well, if you have other distance-based PP effects (like DOF) running, 1) might be no problem at all, but if you do distance fog only for the sake of "doing it PP", it makes no sense.

2) is of a general concern and can be overcome if you are treating alpha transparency in general with mechanics like alpha-to-coverage.

Well, if you have large amounts of serious alpha-objects onscreen it might be wise to calculate the fogging on a per-pixel basis because it might look better in the end. But if you have mostly solid or cutout objects onscreen (and another effect that uses a depthmap, like DOF), do it PP.

The other technique would be the technique Acknex uses nativly.

I have the feeling that with post-processing distance fog you can get better looking results (if I compare with AAA engines) than with the current Acknex technique. But I didn't find any good papers about it yet.

but it should work if the fog is calculated in the object shader?