Bump Mapping - Loss of detail on planar surfaces

(onu) #1

I’m experiencing a severe loss of detail for bump maps which appear on planar surfaces. If a face isn’t aligned with two of the three axis’s it’s perfectly happy to deliver spectacular detail but if it’s planar, it’s blurry. Observe:

Befor anyone asks, here’s the similarities between the two pictures:[ul][li]Same room
[/li][li]Same compile
[/li][li]Same Shader
[/li][li]Same lights
[/li][li]Both are func_groups (though seperate)
[/li][li]Both are detail brushes[/ul]This map was created to observe this phenomenon. Thus there is only one variable between the two groups of brushes. The surfaces are either planar or they are not.
[/li]
Why does it do that?
How can I fix it?

(Emon) #2

Just to let you know, a planar surface means the surface lies on a single plane, and is entirely flat.

As for your problem, I’m not entirely sure. I know others have had problems with the amount of detail they get with bumpmapping, and using thresh in their compile line fixed it. Although I don’t get why a slanted surface would be any different from one that is flush…

(ydnar) #3

The coplanar brush faces are either being merged into a single drawsurface, or the light stage is mapping a lightmap across the adjacent surfaces.

To prevent this from happening, put the faces in different func_group entities.

y

(onu) #4

Alright, thanks a lot. Worked precisely as you predicted.

(Matt) #5

Your example looks like soundproofing in a studio. :slight_smile:

(onu) #6

That’s not the actual normal map, but just a test version. I decided a prism look would give me the best example of how q3map2 interprits the normal map. And I have to say, it does a ^@&# good job.

I should probaby make another version that includes height information (instead of just angle) so that I can see if q3map2 even pays any attention to that. From the few unrelated tests I’ve done it appears as though q3map2 ignores height completely and just focuses on shading angles correctly. However I haven’t created a normal map to test this hypothesis yet so don’t go quoting me or anything.