Thanks Kelvin, for investigating this for us ... > I can see when the program startup, there is two small > black line in the initial orientation. Moving the cylinder > around will sometime see the artifact (under windows). > It's much worse here, as there are perhaps 50 unwanted dots / dotted lines, although some orientations are better (Windows 98; i810; OpenGL 1.1 or DirectX 8, Java3D 1.2.1). > Under solaris, it works fine. Probably because > a combination of numerical precision, opengl and > graphics card give more precise location where > the disc render. > It's good news that it works somewhere, as it gives us something to work with. I think that you're suggesting that it isn't Java3D's issue, in that it's an OpenGL/DirectX/graphics card problem? I guess this means that the lighting calculations occur outside of Java3D, and that Java3D just passes the geometry coordinates and vertex normals on. I'm showing my lack of understanding ... Java3D is my first contact with 3D graphics development, and OpenGL and DirectX are currently 'black magic' to me. > To workaround this numerical problem, you > can add an small offset to create a disc > with longer length. Then add polygon offset between > alternative disc to avoid z buffer fighting for > the small overlap portion. > I had tried this but it's no help. Actually, we've discovered that by removing the ends of the joining shapes then that works (unattractive but it's a possible option). Thanks again, Geoff. =========================================================================== To unsubscribe, send email to [EMAIL PROTECTED] and include in the body of the message "signoff JAVA3D-INTEREST". For general help, send email to [EMAIL PROTECTED] and include in the body of the message "help".
