After delving deeper into 3-D object modding/making, I've found that in a great many instances the original authors of cockpits could have profitably departed from the usual practice of defining polygon surface normals as always perpendicular to their surfaces. For instance, a horizontal surface in the X/Y plane, with the Z axis vertical, will usually have the surface normal set as 0, 0, 1. The poly would be most brightly lit when the light source is overhead, in the direction of the Z axis.
But in a cockpit there are often enough surfaces which have their illumination arriving from a limited range of angle. A good example would be the chutes feeding shells through the cowl mounted MGs in the Zero, where they are located well forward, under the hood, and being illuminated only by light entering from the rear direction. They cannot be illuminated from the forward direction. And so it's not only distracting, but downright ludicrous to see such parts being lit up by the sun as though the aircraft skin were transparent. Flying toward the sun has quite bright illumination and a specular component that could only occur if the nose of the plane were fully removed, with the innards exposed.
My solution in such cases is to make ALL surface normals--irrespective of the orientations of the various plane surfaces--point rearward out through the opening under the hood. This nicely eliminates the ridiculous 'transparent skin' effect, and looks FAR more realistic. For instance, if the Y axis points toward the rear, all surface normals would be set to 0, 1, 0.
It can be useful also to reduce the values somewhat for the .mat file parameters, Ambient and Diffuse. For instance, if more exposed cockpit object textures have these parameters set to 1, the textures for the more shaded parts could be set to 0.7 or 0.8. Of course, this can only be done when a separate texture is used, or if the different objects sharing the same texture are separated in the mesh so that they're using different .mat files.
Additional texture painting can be performed so as to add the appropriate shading. For instance, any surfaces facing the opening directly--and particularly if not blocked by other stuff--would be more brightly lit at all times, and so should be made lighter accordingly. Surfaces not facing the light source, or which are otherwise masked to some degree, should be painted darker.
I cannot stress too strongly how very much superior this approach is as compared to the usual 'default' condition of surface normals being set always perpendicular to the surfaces/vertices.
Another object which benefits very notably from a uniform series of surface normals is the gun sight collimating lens. Differing surface normals tend to make the lens look like cut glass, or a faceted jewel. More so if specularity is set high, whereby the splodge of reflected light from any one polygon is more tightly constrained. I always make all lens poly surface normals point in the same direction, as though the lens were simply a flat plate. This completely avoids the 'faceting' effect, and makes the lens--even if quite strongly curved--appear as a nicely smooth surface illumination-wise.
Finally, I've discovered that in no small number of cases there are meshes which are riven throughout with utterly bad surface normals, the values supplied strongly suggesting unfinished work. Such as, great swaths of similar values, as, e.g., "0 0 1", where it's clear that such simply does not and cannot apply. One stock plane badly afflicted is the F2A-2, or the USN/USMC variant of the Buffalo. I spent a couple of days fixing the more egregiously flagrant instances.
The amount of tweaking that could be undertaken is monumental....
Author
Topic: Some thoughts on polygon surface normals (Read 1002 times)
