Hard Light Productions Forums
Modding, Mission Design, and Coding => FS2 Open Coding - The Source Code Project (SCP) => Topic started by: Scooby_Doo on October 26, 2007, 05:24:50 am
-
I just noticed this. If you look closely at the windows straight in front of you you will see the red nebula behind them. I noticed this also occured on the front winglets (can't see it on the pic though) It gets more noticable as you go further away.
(http://img.photobucket.com/albums/v356/Shodan_AI/problems.jpg)
-
They look like reflections to me - what do they look like close up?
-
is any sort of transparency being used on those textures?
-
There shouldn't be any transparency, although it looks like solid black is doing it.
You can get the model here if you want to http://scoobydoo.freespacemods.net/Finished_Mods/Aurora-10-26-07.rar (http://scoobydoo.freespacemods.net/Finished_Mods/Aurora-10-26-07.rar)
They look like reflections to me - what do they look like close up?
Behind me, the fighter, is mostly black space. When you get closer up the effect disappears.
-
I'm seeing some weird alpha map anomalies in some of your textures. See the attached image - that's the alpha channel of your texture there, and it's happening on a few of them. The alpha channel should be black.
It looks to me like the 1 bit alpha map of your DDS1 textures is somehow being used incorrectly - so you get these weird little groups of pixels that are black or white, and they're causing your transparency/env mapping (whichever it is) effect when they get blurred by the mipmaps. :)
[attachment deleted by ninja]
-
I'm not using channels with the psd file. Could the nvidia dds plugin be exporting the file incorrectly? How do I fix this?
-
You might not be exporting them in the right format. Last I remember it was DXT5 or something.
-
For diffuse maps and glowmaps? I thought they were DXT1 (no alpha)?
-
DXT1s have an alpha channel, but it's only 1bit (white/black or rather opaque/transparent) for each pixel. All it appears to be is that somehow the images managed to give themselves a non-uniform alpha map using the 1bit alpha channel that DXT1 files have.
I'd suggest you try adding an alpha channel in Photoshop and make it solid white. Then save as a DXT1c and that should give you a uniformly opaque alpha map.
And you're quite correct - DXT1c is used for the diffuse, glow and height map while DXT5 is used only for shine maps because they make use of the alpha channel. Don't use DXT5 for anything else unless you want the texture to be transparent, and DXT5_NM is only used for normal maps. :)
-
There is a DXT1 which has a 1-bit alpha channel, but even if it exists the game won't use it. As far as the game is concerned the alpha version of DXT1 simply uses more memory than the non-alpha version (the alpha version being 4:1 compression, the non-alpha being 6:1 compression).
I just agree with Vasudan Admiral (from his first post), it looks like reflection from the env. Turn off env and see if it goes away. If not then we'll know pretty easily. If it does go away then it's likely just a problem with the specmap, which would just need to be tweaked to fix the reflection issues on those parts of the map.
-
Oh the game won't use the 1 bit alpha?
In that case this is kinda odd then, since the abnormalities on the alpha maps appear in the same places as the 'phantom specular' bits do in his screenie. That said, I just checked it out in the ships lab and there doesn't appear to be anything wrong at all to me.
-
Oh the game won't use the 1 bit alpha?
Nope, it's treated as a regular RGB image, with no alpha channel. The memory usage is really the only difference as far as the code is concerned, and that is why I always tell people to use the "c" version, as the "a" version simply wastes memory.
-
I never use environment mapping, never really liked it.
However it appears the problem occurs with glowmapping.
BTW can shinemaps be DXT1 too? Because thats what i've always been using and it seems to work good.
-
However it appears the problem occurs with glowmapping.
What do you mean here? If you turn glowmaps off (ie, remove -glow from the command line), the effect goes away?
-
Yup
-
And what if you delete the alpha channel of 0-auro0-glow.dds in photoshop and resave as a DXT1c (no alpha map)?
-
Ya, that solved the problem... I mustn't have scrolled all the way back to DXT1 no alpha on the export combobox and it used the alpha version for it. Strange thing is, my psd files have no alpha channels, it must create one on the fly.
-
Strange, there must be a code bug somewhere then. I'll try and recreate the problem on my end and see if I can't work out why that happened. Quite a few of my own files are DXT1a, since I just never bothered to fix them, so I'll see about creating a similar effect with one of them and see what happens.
-
Strange, there must be a code bug somewhere then. I'll try and recreate the problem on my end and see if I can't work out why that happened. Quite a few of my own files are DXT1a, since I just never bothered to fix them, so I'll see about creating a similar effect with one of them and see what happens.
Well I meant, the DDS exporter must be creating one on the fly as I save the file.
-
The game code is supposed to treat DXT1c and DXT1a the same though, so it shouldn't matter whether there is an alpha layer or not. That's a code bug on our side. :)
-
Since it gets more noticable from further away, might it be a problem with the mipmaps somehow? I definitely didn't see it at all in the techroom using those textures yesterday.
Oh, and Scooby, which build are you using? If you haven't already, see if you get the same effect with Taylors new build here: http://www.hard-light.net/forums/index.php/topic,50258.0.html
-
Since it gets more noticable from further away, might it be a problem with the mipmaps somehow? I definitely didn't see it at all in the techroom using those textures yesterday.
Could be a mipmap related issue. But you can actually test that in the lab too, just by zooming in/out it will switch mipmap levels. Viewing it in the techroom may just use a mipmap level so small that you can't notice it (depends on the filter used to create the mipmap levels). In the lab you have a little bit more play to try and figure out where a map problem might be, but since the background is black, it's usually pretty difficult to actually see those problems.
The background problem in the lab is something that is on my todo list to deal with though, along with setting of light color/position, number of lights, and the envmap to use.
-
I haven't been able to replicate this problem so far, and I don't see anything really out of place in the code either. It's possible that this may just be a driver related bug. For now I'm going to push this to the bottom of my list and just wait and see if anyone else has a similar problem.