GeorgeTirebiter

This is a common misconception, even among pilots, but it\'s totally untrue. A stalled wing still produces lift, just not as much as at a lower angle of attack. Here\'s two graphs of wind tunnel measurements of lift vs. AoA for the same airfoil. The first is for normal (unstalled) conditions: A stall happens when the aircraft\'s AoA exceeds that for maximum lift, which for this airfoil is between 12 and 15 degrees. Notice how the lift doesn\'t drop to zero beyond that point. That graph only had data up to twenty degrees AoA. Here\'s what happens if you keep increasing the AoA: The lift doesn\'t reach zero until the AoA is past 90 degrees—that is, the wing has to be nearly flat to the oncoming wind to produce no lift! The first graph is from Abbott & von Doenhoff, 1949, the second is from NASA CR-2008-215434.

Instead of using 1, 0, 0 for the Up X/Y/Z values, you would use cos(?), sin(?), 0 to rotate the part ? degrees about the Z-axis. There\'s also an example part made by Killerhurtz in this post.

I don\'t know if this would work, but you could try rotating the entire thing in Blender so that the hinge is aligned with the X-axis, and set the node_attach vector to the opposite of that rotation.

Select the object, press CTRL-A, choose 'apply rotation' and Blender will do this for you.

Yeah, that could be a problem. Try this: in the far left side of the menubar at the bottom of the 3D window there\'s a little button with a white cube and an up/down arrow. Click on that, and select 'UV/Image Editor' from the menu that pops up. You can also have the 3D and UV views open at the same time by splitting the window. In the upper-right and lower-left corner of the viewport is a diagonal 'grip' (when you hover over it the mouse cursor changes to a plus sign). Click-drag it into the viewport (as though you were peeling it away to reveal something hidden underneath) to divide the window, then change one side to the UV editor.

It looks like your UV mapping got reset somehow. Unfortunately, that means you\'ll probably need to unwrap the part again. Also, you should note that UV mapping is only one of many ways that Blender\'s renderer can assign textures (and it\'s not the default method), so a rendered image won\'t necessarily match how the part will look in KSP.

The 'auto smooth' option (like 'double sided' next to it) is only supposed to apply during rendering, so it\'s odd that it had any effect at all on the exported mesh. The Edge Split modifier really is the best method of controlling smoothing, and it doesn\'t actually change the mesh until you 'apply' it. I do that just before exporting and don\'t save the changed mesh so it\'s still easy to edit.

Looking at the file for the strut connector, there\'s actually several things that are causing problems. The default parts are built to a scale of 10 units/meter, but Blender\'s .dae importer forces the scene scale to 1 unit/meter and then applies individual scaling to each object in the file. This means you\'ll need to change the scene scale yourself. In the properties panel, select the Scene tab (the one with the icon that\'s a tiny sphere, cylinder, and light). Scroll down to the Units properties, set the Scale to 10 (i.e., the inverse of the scale setting in the part.cfg file), and make sure the units are set to Metric instead of None. This will correct the scaling for most of the default parts, but it\'s not enough for these strut connectors because of a bug in Blender\'s .dae exporter. The problem is that the bolts at each end of the strut are parented to a geometry-less object (called an 'Empty' in Blender). These empties also get scaled on import, and that scaling incorrectly gets applied to their children during export. There are a number of ways around this--I\'ll only describe one. In the 3D view, select one of the bolts. Remove the parent link by pressing ALT-P and choosing 'Clear Parent' (or look under Parent in the Object menu). Next, select the empty for that bolt (it\'s represented in the 3D viewport as the set of axes that\'s barely sticking out of the mesh) and press ALT-S to clear its scale. Finally, re-link the objects by first selecting the bolt, then hold SHIFT and click on the empty so they\'re both selected (in the correct order). Now press CTRL-P and choose 'Set Parent to Object.' Repeat for the other bolt. As for the rotation, it\'s wrong because KSP assumes animated parts are created with the Y-axis as the vertical axis, while Blender exports parts with the Z-axis as vertical (this doesn\'t seem to be an issue for non-animated parts). Fortunately, this is easy to fix. Start by rotating everything so the Y-axis is up (this is only necessary if you rotated the part after import or you\'re making a new part from scratch). Export the collada file, then open it with a text editor. Near the top of the file, find the line '<up_axis>Z_UP</up_axis>' and change it to Y_UP. Now your strut connector should work in KSP.

Blender\'s collada importer gets hung up on how textures are declared. Fortunately, it\'s easy to fix. Open the .dae file in a text editor. Look for a texture assignment like this: <diffuse> <texture texture='Map__1-image' texcoord='CHANNEL0'> <extra> <technique profile='MAYA'> <wrapU sid='wrapU0'>TRUE</wrapU> <wrapV sid='wrapV0'>TRUE</wrapV> <blend_mode>ADD</blend_mode> </technique> </extra> </texture> </diffuse> Delete everything inside the <diffuse> tags. Don\'t worry; this won\'t affect the part in KSP.

You don\'t need to manually triangulate the mesh, as Blender\'s .dae exporter takes care of that automatically. What it won\'t do for you is UV-mapping. A mesh without UV coordinates will also cause array index errors. If you don\'t know how to UV-map in Blender, here\'s a good introductory video tutorial: http://cgcookie.com/blender/2011/01/21/intro_uvmapping/

Smoothing in Blender works slightly differently than other software. Instead of assigning faces to smoothing groups, you add an 'edge split' modifier to the object and mark the edges on the perimeter of each smoothing group as sharp (or let Blender do it automatically based on the angle between faces). None of that matters, however, because Blender's .dae exporter is terrible and it ignores all modifiers.

The problem isn't that the NERVA model has too many polys, but that it has too many separate meshes. The default engine has two (the engine and the collision mesh), but the NERVA model has 34! I was able to get a major performance boost without deleting any polys just by merging those meshes and re-exporting.