Page 1 of 1
[bug] strange rendered model
Posted: Tue Mar 27, 2007 6:34 am
by Sebastian
when i rendered the folloing model I recognizez a bug !!
but watch yourselves !
I turned the Normals to the right side.
The Material is Phong with some exponent and IRO to 1.5 .
Posted: Tue Mar 27, 2007 6:47 am
by Kram1032
sorry, I can't really see the bug...
where there shouls be Phong?
how have you lit the scene?
Can you post a pic, where you can see the normals?
Which app + Exporter + Version?
Posted: Tue Mar 27, 2007 9:00 am
by Sebastian
App: Blender
Indigo 0.7 t5
Blendigo Exporter
Lit by a Meshlight
Here a Image where I've marked the strage areas.
Look at the preview and compare these and u'll see that there is a fault !
Posted: Tue Mar 27, 2007 10:05 am
by CTZn
The normals
are not right

edit: I think they are reversed in the toroidal shape (the round border).
Someone wrote once, that often, the bug was between the screen and the chair

Posted: Tue Mar 27, 2007 10:21 am
by BbB
Blender does weird things to normals, sometimes utterly un-understandable. Better export and re-import your mesh as an .obj file. It generally gets rid of the curse, as a famous Docteur once advised.
Posted: Tue Mar 27, 2007 10:22 am
by BbB
Oops. Actually it was CTZn's advice. Apologies

Posted: Tue Mar 27, 2007 10:30 am
by CTZn
no harm

Posted: Wed Mar 28, 2007 5:42 am
by Kram1032
I didn't have problems with normals, in blender, yet, except when there have been intersecting faces... they only would be used correctly, if normals would be pointing in both directions...
Can anyone explain me, why (choose one of these):
- BiDirectional Normals aren't possible
- BiDirectional Normals aren't a good idea
- or: noone thought about that
Posted: Wed Mar 28, 2007 7:18 am
by CTZn
- At least it's not a good idea because if there was no normals (or BiDirectional normals as you call them) it would be much expansive for a raytracer to check wether a ray is currently inside or outside of a transparent, closed volume. Basically when a ray hits a transparent surface, if the normal is pointing outside
the renderer knows it is entering a volume. Quick and easy, there is nothing to compute because the information is already embedded in the geometry.
The other way, if there exists no normal data, is to count the numbers of surfaces the ray traversed (?), wich in some cases (that is, very often) lead to errors. boooh. You want normals to be pointing in one direction only.
- So BiDirectional Normals are possible. Thats the very same that
no normals data. Showing only one side of the surface is a technical choice.
- You, think about it again

Posted: Wed Mar 28, 2007 7:25 am
by Kram1032
ok....

Posted: Wed Mar 28, 2007 8:22 pm
by mrCarnivore
I also have thought about it before and I like bidirections normals if there are also bi-material faces. That way you could easily model a fluid in a glass and the faces that are fluid on one side and glass on the other would simply get assigned two different materials (one at each side).
That would solve that problem with fluids and glass that they are either intersecting and lead to strange artefacts or have "air" between them which also leads to unwanted results...
However, thanks to the precedence system in indigo there is a neat way around this problem that works really nice.
Posted: Thu Mar 29, 2007 7:54 am
by Sebastian
I tried it with deactivating "SetSmoth" and it works !!
Now I'm confused

Posted: Thu Mar 29, 2007 8:05 am
by zsouthboy
Because "SetSmooth" essentially turns on normal smoothing (to both of the Blender exporters I know of).
Normal smoothing means that the normals of surrounding tris are averaged and interpolated across the face of the first tri - it prevents the common "blocky" looking 3d models by smoothing the lighting.
Posted: Thu Mar 29, 2007 8:11 am
by zuegs
@Sebastian:
please check your objects to be uniform-scaled (Scale-X should be equal to Scale-Y and -Z). If not just apply transformation to mesh by hitting Ctrl-A (i think

).
Also please check concole after export if exporter has noted some warnings.