Celestia Forums

In Celestia if I go very close to earth at the daylight, and the galaxy (and dim stars) becomes more visible , which doesn't happen in real life.
I guess(i'll check the sources in future, I didn't do it ) that is done by addition like "the illumination"= "the light of the galaxy"+"the daylight". I think that it should be like this: "the illumination"=maximum of ("the light of the galaxy", "the daylight").
This error(?) happens also if "I go" very close to a star.

Could've sworn I'd replied to this earlier.... oh well, must be an error somewhere...

I've written about this particular glitch before. It seems the glare effects are reversed - stars iluminate distant stars rather than overwhelming them with their own light. Realistic glare effects would be a nice touch for a future version - it would be good to see the number of stars change depending upon your direction of view, proximity to the closest star, and whether or not you were in the shadow of a planet.... to get all the details right would be a major challenge. And the realism would, of course, never be perfect since the user has the option to control how many stars are visible. Perhaps there could be an option to select the maximum magnitude visible, which would be affected by those conditions. Could any of this be done ?

(Mad Boris)

paulnasca wrote:I guess(i'll check the sources in future, I didn't do it ) that is done by addition like "the illumination"= "the light of the galaxy"+"the daylight". I think that it should be like this: "the illumination"=maximum of ("the light of the galaxy", "the daylight").

You're right. When adding brightness of two pixels, if p_1 is the pixelvalue of pixel no 1 and p_2 of number 2, then the calculated pixel value should be

p_tot = log(exp(p_1) + exp(p_2)).

This is much better approximated by p_tot = max(p_1, p_2) than p_tot = p_1 + p_2; however, even with this simplification, you have to think out a clever way to implement it (it's not straightforward to do it efficiently).

/Alexis

p_tot = log(exp(p_1) + exp(p_2)).

This is much better approximated by p_tot = max(p_1, p_2) than p_tot = p_1 + p_2; however, even with this simplification, you have to think out a clever way to implement it (it's not straightforward to do it efficiently).

It's actually not too bad . . . The OpenGL extension EXT_blend_minmax enables max(a, b) and min(a, b) blending. It's not available on all cards, but I know that nVidia chipsets from the TNT on up support it. So does the Matrox G400. I'm experimenting with it in Celestia and it will probably appear in 1.2.5.

--Chris

chris wrote:The OpenGL extension EXT_blend_minmax enables max(a, b) and min(a, b) blending. It's not available on all cards, but I know that nVidia chipsets from the TNT on up support it.

Oh, I didn't know that. That's excellent news! Crowded star-fields, daylight, eclipses... lots of things will be much more realisticly rendered.

/Alexis

It's especially helpful with double stars . . . Notice how in Celestia 1.2.4 Alpha Centauri appears abnormally bright because the two components are rendered additively.

--Chris