I'm pretty sure the pillars are
not circular. Rather harlequin (diamond) shaped, rounded at two non-adjacent corner.
Just like GaryGB suggested, I'll go for one single model export
per pillar shape and size (Pillar_Height10.mdl, Pillar_Height12.mdl, Pillar_Height_08.mdl, Pillar_TerminalCorner.mdl, etc.) and compile those as bgl library (library resource, with one GUID per pillar) and place the produced bgl file and textures in a declared low level scenery library. Then
invoke those library models from another bgl in the actual airport scenery.
You have
two ways to do that (I understood there's a third one I don't know about called "graphic injection" for FSX and P3D only, that you can do with 3D objects and AI traffic)
- either use a bgl referencing each pillar in lat/lon, orientation
and altitude (0m above AGL, since you would have several pillars with different height) You can use several tools to do that like ObPlacer, SBuilder, probably MCX aswell...
- or use directly your 3D terminal scene, and
attach the library models directly within Sketchup (if possible) This will avoid you the hassle of looking for lat/lon coordinates of each pillar. Though, I'm not (yet) an user of Sketchup ! I use GMax and much less 3DS since I don't model for anything other than FS9 ATM. In GMax you can attach any library object you want by placing a small box, and tagging it with a specific name and the GUID of the model you want to attach. Then the compiler will do the job by actually converting that box to an
attach reference in the simulator code/language.
^^ that's the
instance concept of the simulator which is exactly the same as Sketchup
in an abstract way :
have an actual 3D source object (the mdl) then invoke as many ghost representations as required when rendering the scenery in view. However, neither the SDK compiler or MCX are able to convert/port/migrate/transform
Sketchup definitions of instances to
simulator definitions of instances. Instead, they actually
recreate each pillar as actual 3D objects with as many points (vertices) and polygons (triangles) as required
before exporting the scenery to FS native language (with all the extra drawcalls for materials/textures definitions)
Side note : Being ghost representation instances, or actual independant 3D objects, the amount of points and triangles to be processed by the graphic engine
will be the same. The main difference in the two concepts is the optimization made when you use ghost instances. The rendering pipeline is shorter because only one object and a few triangles are taken as reference to draw the others. Smaller source object boundingbox size and few amount of triangles means faster Z-index computations, so the rendering goes faster. By the way, the amount of memory required both in RAM and hard drive/swap/cache is much less.
So in the end, you'll definately need to
reduce the amount of polygons for a given pillar. Whatever number of polygons stated by Sketchup or other 3D modeling tool,
never believe them ! The simulators thinks in
triangles = 3 points. So, quads/square polygons, either shaded or not,
are not triangles. They are actually
two triangles. Double sided quads may be shown as one unique polygon, they are actually
four triangles. So a scenery containing 3000 quads is at least 6000 triangles.
I would suggest you to go for the mdl library, and :
- Make simplier shaped pillars (*) At least, don't make them that detailed. Since I think they are not that curved, you can save polygons where the faces would be rather flat. Applying a good texture and nice shading will surely soften the angular looking edges.
- Use LODs (Level Of Detail) which is a technique to make an object looking even simplier the farther from them you get (or the smaller they are on screen) That's the principle used for AI Aricraft models. You can create LODs on any scenery library model.
- The planar faces suggested by =rk= is a nice technique. They are extensively used in LODs. However, bear in mind that those planes are double sided, which doubles the number of polygons/triangles involved. Planar faces techniques works well with objects that requires a lot of polygons if they were to be designed as they are in reality. Objects like vegetation, fences, non cantilever bridges, lights beams, clouds/smoke, curved/circular objects (tires, modern cars, balls, San Francisco bridge...) Always make two samples of one object before deciding to go on with the planar technique. If you can't get less triangles in planar mode, choose the 3D volumic object version instead and drop the planar one.
(*) Basic, low detailed pillars will
indeed affect the general "realistic" look of the scenery. But bear in mind that users won't spend hours to contemplate one pillar and measure its height, but rather check the docking system, taxiway markings, watch out for moving vehicles, and like the nice blending between taxiway edges and grass. Furthermore, they'll ultimately hate
framerate losses on final, the critical moment of the flight, or an OOM (out of memory) exception when they load PMDG and 100% AI traffic on your airport... Especially those who use the simulator with WideFS and multiple monitors. The best scenery is the one that looks sooooo real and alive, while being able to sustain decent framerates whatever the maximum sliders settings the user have defined...
EDIT :
Just made an attempt at a rather very low poly one (GMax) :
This is a 32
triangles pillar.
The (unfinished) texture is a very basic one. Even not really convincing, you can actually use it as the
3rd or fourth LOD. From far, you won't notice the edges much. You can go for example up to 300 polygons for the first LOD. Each 6 or 7 pillars will then induce as much workload as one common AI aircraft. If you have 300 pillars, it would be like having an airport overview with
46 extra AI aircraft on screen (so 300 triangles could be a bit high. Would go for
200 max instead) Assuming your first attempt on the pillars has 28-sided cylinder base, with the curvature of the horn, you would end with
952 triangles !!! per pillar. Even if it's LOD-ed, 300 pillars at 952 LOD1 is like seeing at least 146 AI aircraft at the same time (much more in fact because a 952 triangles LOD1 pillar will have
slower polycount decrease in further LODs than a 200 or 300 triangles LOD1 version...)
Hope you understand the importance of balance between details/accuracy and performance. Quality is somwhere in between.
>.
), I have solved it like this:
) and enjoyed it very much. So that is my next project (for the coming years, unless I find somebody to help me out with this 'to-be-freeware' scenery in which case you could send me a PM if you are interested and have nothing else to do, thanks).
