1.5 Odd light drop-off in OpenGL 2.0

[Kolano]

The drop-off from day to night seems a bit odd on the OpenGL 2.0 path (using Phoenix's most recent CVS build from 10-21-06). There is a bright ring of light right before it starts to fade out.

I can post some shots if what I mean is not immediately apparent.

[chris]

Yes, please post some shots. I'm not sure what effect you are referring to.

--Chris

[Kolano]

OpenGL /w Nvidia Combiners:


OpenGL 2.0


Note in the second image just before the fade to dark there is a slightly brighter ring around the planet. It intersects the large spot (labeled Lyot now that I check what it's called.) visible in the upper portion of the image.

Also note the weird seam that goes all the way acoss the planet along its longitudes. There are a number of these across the planet. Though the shading makes this seem to be in the normal map, this actually appears/disapears depending on if cloud shadows are turned on.

My Mars ssc is...

Modify "Mars" "Sol"
{
Texture "m46shaded_dds.ctx"
NormalMap "m46normal_dds.ctx"

Atmosphere {
Height 30
Lower [ 0.8 0.6 0.6 ]
Upper [ 0.7 0.3 0.3 ]
Sky [ 0.83 0.75 0.65 ]
Sunset [ 0.7 0.7 0.8 ]
#Mie 0.001
#MieAsymmetry -0.15
#Rayleigh [ 0.006 0.0025 0.002 ]
Absorption [ 0.0002 0.0004 0.0008 ]
#MieScaleHeight 8
#CloudHeight 7
#CloudSpeed 0
#CloudMap "MarsClouds.png"
Mie 0.00025
MieAsymmetry 0.15
Rayleigh [ 0.00175 0.000875 0.00075 ]
#Absorption [ 0 0 0 ]
MieScaleHeight 14
CloudHeight 7
CloudSpeed 0


CloudMap "mars-clouds.dds"
}
}

[chris]

I notice that you're using a DDS normal map . . . When I added the compressed normal map feature to Celestia, I made Celestia incompatible with DDS normal maps that didn't use the dxt5nm color channel assignments (which includes every DDS normal map on the Motherlode.) Yesterday, I changed how Celestia determines whether a normal map is compressed and restored compatibility with existing DDS normal maps. I see that you have a very recent build from phoenix, but it still may not include the normal map fix.

--Chris

[phoenix]

I've just made a new build but still can see this bright ring as seen above on earth using "earthAlphaBumped_8k.dds" where the alpha-channel contains the normalmap

[Kolano]

I am fairly sure I am using the new code that requires the "dxt5nm" renames. At least Phoenix's build had a date that was post the date of your announcement about that.

Prior to that update, Mars looked even weirder, basically it was totally dark except for a small slice. Basically it had looked like it was being illuminated by a flashlight held very close to it.