DXT5NormalMap "shadow" bug

[john71]

I started to experiment with DXT5nm textures.

I used DDS Converter 1.4 and Gimp, but there is a strange graphic bug which just won't go away, no matter how I manipulate the normalmap texture.

ddsnormalproblem.png

There is this strange shadow...what is it?

Have you ever seen such a shadow?

I use Celestia 1.7 64 bit in Windows 10 pro.

What can be the problem?

Added after 12 minutes 50 seconds:
I think I solved it. I renamed the *.dds file to *.dxt5nm.

It can't be that simple.

[John Van Vliet]

how big is the single texture normalmap ( that is NOT corrected for a sphere)

this is normally caused by using a map that is too big for your 3d card
2k and 4k are what most cards can use

also if the normalmap has a alpha channel this can happen

did you "flatten" it in gimp ?

a normal dds normal is blue
a "dxt5nm" normal map is orange

[john71]

I simply renamed the file from *.dds to *.dxt5nm and the problem disappeared.

I don't understand it.

Dxt5nm is not a special dds file? How can it be a separate file type?

Added after 1 minute 17 seconds:
The color is OK.

By the way, this utility is a batch converter and fantastic.

https://vvvv.org/contribution/dds-converter

[catwrings]

Dxt5nm is not a special dds file? How can it be a separate file type?

I encountered the same "shadow" problem when using DXT5nm for the first time, I suppose that it is just Celestia that interprets a file as DXT5nm only if it has the extension dxt5nm, otherwise she falls back to another default interpretation (DXT3? DXT1nm? Don't know). The image interpreted as if it has an other format of course is nonsense. It seems as if the normal map had a negative sign so that shadow and light are inverted; the dark side becomes dark anyway because that is computed on the surface map, so there remains an illuminated stripe along the terminator.