shadow bug

[simatupang]

i found a bug while playing on jupiter. the shadow of it satellite on jupiter's surface is doubled when it is near the terminator (both east and west). the 'ghost' shadow move in opposite direction. if the real shadow emerge from western terminator, the ghost shadow appear starting at the terminator (splitted from the real shadow) and move in opposite direction. when the real shadow approching eastern terminator, the ghost shadow appear from behind and merge with the real shadow at the terminator. see the image:

[granthutchison]

That one has been fixed since version 1.3.0.

Grant

[Guest]

That one has been fixed since version 1.3.0.

Grant

o ya?
i'm using version 1.3.1, running on winxp. and the bug still there

[granthutchison]

You surprise me.
Before I posted, I checked my 1.2.5, 1.3.0 and 1.3.1 installations under XP and found the bug only under 1.2.5 (there was a handy shadow transit just finishing at 08:00 my time ). Chris also made a bug-fix announcement some time between 1.2.5 and 1.3.0.
Can you provide details of your graphics card?
Anyone else still seeing this bug?

Grant

[simatupang]

You surprise me.
Can you provide details of your graphics card?

this is my spec: nVidia GeForce4 460 Go, P4-M 2 GHz, WinXP

[JackHiggins]

It was definitely there in some of the 1.3.1 pre's. Are you sure you've got 1.3.1 final, simatupang?

[simatupang]

It was definitely there in some of the 1.3.1 pre's. Are you sure you've got 1.3.1 final, simatupang?

yup, it's v1.3.1 (i just check it from the help->about). but i dont know how to check if it 1.3.1 final or not.
i just upgrade it from v1.3.0 (my first celestia) a couple of days ago. the source (celestia-win32-1.3.1-1.exe) i downloaded from http://prdownloads.sourceforge.net/cele ... ror=heanet

[granthutchison]

Try using Control-V to toggle through the Render Path options supported by your graphics card. The bug still seems to be present in the "Basic" setting, but is fixed in all other settings on my machine.

Grant

[selden]

Of course, if you set "ambient light" to "none" you won't see the extra shadow. Non-physical lighting lets you see non-physical effects.

"Ring shadows on planet" has a similar bug.

If your card supports OpenGL vertex shaders, so you can see the ring shadows on the planet, the ring shadows can be seen to encircle the entire planet if you have ambient light turned up. This is easily seen on Saturn with today's date.

As I recall, it also has been noted that the ring shadows are much fainter than they should be. It'd be nice if that could be fixed, too.

[don]

i found a bug while playing on jupiter. the shadow of it satellite on jupiter's surface is doubled when it is near the terminator (both east and west).

Yes, I can confirm this under the following conditions:
* 1.3.1 final
* XP Pro
* ATI Radeon 9700 Pro
* Ambient Light: Medium
* Render path: Basic

The two eclipse shadows move towards each other like a mirror effect. The real one is on the left, moving to the right. (This is a brightness and contrast enhanced screen shot) ...



While trying to verify this, I found another interesting effect, but only in the Multitexture and OpenGL vertex program Render paths. Note how there is a definate line where the shadow is "eaten" and simply vanishes. The shadow is moving to the right in real-time. (This is a brightness and contrast enhanced screen shot) ...



Also found an ugly seam on Jupiter that I had not noticed before. This is the texture distributed with Celestia 1.3.1 final (screen shot not enhanced)...


-Don G.

[granthutchison]

Note how there is a definate line where the shadow is "eaten" and simply vanishes.

All of this, as Selden has already said, is visible because you have non-physical lighting turned on - the apparent edge is just the limit of the moon's shadow, and the beginning of Jupiter's own shadow, as you can confirm by turning ambient light off.

Grant

[don]

Howdy Grant,

Yes, I read what Selden wrote.

When a program has an Ambient Light setting, IMHO, I would expect it's rendering routines to take this setting into account and render accordingly. It appears that Celestia does not. It's rendering the shadow as if there is no Ambient Light at all, even though it's set to Low or Medium.

You are saying the shadow should *not* be cast until it becomes a line, in this particular case, but rather, be cut-off (as shown) as if there was no Ambient Light. To me, this is incorrect, which is why I reported it.

Cheers,

-Don G.

[selden]

Don,

I don't think either of us disagrees with the bug report. We're just making excuses

[granthutchison]

I don't think either of us disagrees with the bug report.

No no, hang on, I am disagreeing with the bug report, or at least the part of it involving the sharp shadow edge that Don is commenting on (the persistence of the double shadow in basic rendering is of course a problem). I think the sharp cut-off is entirely correct rendering, which has been made to look odd by the unreal ambient light on the night side of Jupiter. (Non-physical illumination, non-physical effect, as Selden says.) To cast the moon shadow across the terminator on to Jupiter's night side, in order to avoid a sharp cut-off, makes no sense at all - it is, in fact, exactly the bug that causes the double-shadow effect that started this thread.
I guess you might make a feature request for ambient light to be extended to eclipse shadows, so that the hard cut-off between realistic shadow and unrealistic shadow was abolished. But that would bring its own problems: people like shadows on their planets (there's a constant level of complaint from those who can't see ring shadows), and extending ambient light to eclipse shadows would immediately make eclipses and shadow transits difficult to see.
Perhaps a set of ambient-light check-boxes would be the way to go: allowing the user to choose which kinds of shadows are illuminated by ambient light and which aren't. Something like:

Night side of planets/moons: X
Eclipse shadows: X
Planet shadows on rings: X
Ring shadows on planets: X

Grant

[don]

I'll let you PROS figure out how to make it work. I am not a graphics programmer and do not understand the OpenGL or graphics cards routines used in Celestia.

All I was saying is that if Ambient Light is ON, then the rendering routines should take it's setting into effect (if possible), at whatever level of ON it is set to. If people do not want Ambient Light ON, yes, they can turn it OFF. Simple as that.

Rendering shadows with it OFF works fine. Rendering shadows with it ON produces undesirable effects. Turning it OFF is not the answer. Fixing the undesirable effects is the answer. If this is an OpenGL limitation, that's fine too.

Cheers,

-Don G.

[granthutchison]

Don:
It's not a hardware or software problem. It's just geometry - in the real world of our solar system an eclipse shadow cannot extend beyond the sunset/sunrise line. (Otherwise you'd be able to see the shadow of a solar eclipse passing beneath your feet during the night.) The imaginary "ambient" light illuminating the night side of Jupiter is just drawing your attention to this fundamental fact in a way you can't see in the real world.

Grant

[maxim]

It's a problem of illumination. If your nightside is lightend by ambient sources, your eclipse shadow should be lighted same way.

maxim

[granthutchison]

If your nightside is lightend by ambient sources, your eclipse shadow should be lighted same way.

That would get rid of the cut-off at the terminator, but at the expense of making eclipse shadows, particularly penumbral shadows, much harder to see when ambient light was turned on. I'd really hate to see that implemented, since many users could just lose the beauty of shadow transits without realizing what they were missing.
It's why I suggested allowing the user to choose the areas ambient light would affect. I'd also suggest that the best default setting for these options would be the current behaviour of Celestia, which maximizes visibility by illuminating the night sides of planets but leaving other shadows dark enough to stand out clearly.

Grant

[maxim]

Users should (hopefully) be aware of what they're doing when pushing up ambient lighting. If not :schulterzuck:, I don't know if more option settings would help. But of course it would be okay having these fine tuning possibilities.

I wouldn't expect a behavior as it is now. What I'm seeing is that it's a problem for Jupiter moons and for other minor moons, as they are regulary fully eclipsed by their planets. This leads to curious visions, especially when the moons are 3D models.

maxim

[granthutchison]

:schulterzuck:

I know enough German to recognize "shrug", which is certainly one approach to the problem.

:shrug:

Grant