FS2004 Smoothing a surface...

[Here for the Beer]

... without adding polygons. Can it be done?

I'm making an AI model of the Dassault Falcon 900 and the top LOD is nearly done, bar the new mappings required when the shapes are finalised.

The model looks fine in Gmax but in-sim there are some oddities of shading around the tailplane even though the normals don't seem out of line.

The real Falcon 900's tail is a delightful thing to look at in many ways, the shapes flow from an S-duct nacelle through into the tailfin really nicely, and the fuselage "plinth" for the over-body nacelle is also pretty subtle and blends back until the tailfin. I couldn't figure any easy way to model this as a single object so I composed it as a selection of its fundamental shapes - from a tailfin, an S-duct nacelle, and a plinth - and booleaned all these objects together when their individual shapes seemed right.

The problem is shading (especially under angled lighting) and also smoothness of the parts as a whole. The building of the three sections has, despite my best efforts, left "seams" between the components that angled light shows up quite a bit but which look okay the rest of the time. The S-duct fairing tapers down towards the tailfin and should seamlessly segue into it, but it doesn't even though the surfaces at "boolean time" angled towards each other in a fairly nice way. I thought that by combining the objects and assigning them to the same smoothing group, I'd end up with a decent end result.

Short of going back and just retrying the whole procedure with slightly altered base components, which I don't think will change much in the end, is there any way that an uneven surface can be averaged across its surface? By that I mean that the area where the seams are could be physically smoothed (not just by creative shading with the smooth modifiers I know) while keeping the rest of the shapes untouched? In other words in my magic design program I'd select the polys and tell the program "the periphery of the selected area is sacrosanct, don't move nodes there, just make the physical angles between the selected faces flow better". I couldn't find that button in Gmax, but it would be a bit like selecting a contiguous group of polys to "make planar" - just in this case "make more sexy" rather than "make flat".

I will carry on experimenting. Short of that perfect solution - any advice as to how to achieve this with hard work is just fine!

 

[Here for the Beer]

I had a more focussed look at this surfaces problem. What has happened is that the boolean operation mustn't have been on solid objects, or that boolean doesn't work the way I think it does, because the engine nacelle surface actually passes beneath the tail surface - it doesn't end at the tail with the aircraft skin continuing aft as tail skin rather than nacelle skin. The boolean operation seems to have not noticed these two surfaces intersecting. The two surfaces are still seen as separate things which is why the smoothing is all ineffectual!

It seems I need to somehow get Gmax to recognise where these planes intersect and then create a join there. Gawd knows how... However I'm quitting for today! Brain full, too much fun - brain go pop.

 

[bob5568]

I think boolean is an error. I'd go back in time to before the boolean. You can "attach" the objects to have gmax recognize them as one part, and then fill the seams with polys you create. In 3dsmax I seem to think there is a cool "bridge" command that might do what you want, although its not in gmax.

 

[lionheart]

Booleans are very difficult to get them to work every time. You have to do a lot of messing with them to get them to work right.

* Make sure both parts are Editable Polys and of equal sophistication.
* Make sure curves do not intersect 'with' an edge. You want the curve to go through the middle of Polygons.
* Moving the plug cutter a bit might create a better Boolean cut.


Joining Nacelles into wings (for instance) can be daunting. My route was to move them together, then to manually cut lines through one of the parts, select the hidden Polys and delete them. Then do the same thing to the other side, cutting lines in the object all around at the intersection point and then delete the inner/hidden Polys. Time consuming and a test of patience.

I have never done this with Boolean.

You 'might' be able to delete large zones (not huge zones, but one square away zones) and 'bridge' the two parts together. This would work fast and clean, then cut a line along the middle (if needed) through the bridge zone and move that inward and perhaps chamfer it. Bridging is awesome and quick. Great for wing roots. Just create one single Polygon that joins the two parts (after you have 'Attached' the two together). With one join point, you can then use Border / Cap and cap a surface all around. Then use Edge / Create and create all your edges into the bridge zone (cap area) and Voila, done.



Bill

 

[Here for the Beer]

Thanks guys. I tried going back to first principles and building the tail again using solid shapes. I somehow thought that Boolean "Union" of these would thus create a solid object based around the three shapes being combined (all as editable polys). You'd think that such an operation on 3 solid shapes would do that, but where the engine fairs into the tail I end up with 2 surfaces, one behind the other, as though the union of the tail didn't work the same way as the union of the 2 other shapes (engine and "plinth") which do behave as one solid object. Needless to say the shading issue remains when surfaces behave that way - they can't be smoothed (well, shaded to look smoother) because there's no link between them.

What I DO have now is a far simpler object to work with, though (this is just AI, remember, so we're not taking of a huge amount of vertices etc. Since I do lots of incremental saves I can revert back to the step where all the shapes were right and ready to be booleaned, and use a copy of each one as a cutter for the two others, then stitch them all together.

What I may do, in fact, is boolean merge the engine part, then use a copy of the tail to cut a tail-shaped slot in the rear of that structure, then use a copy of that structure to cut another copy of the tailfin to omit the part that extends into the engine area, so each piece would be a "negative" of the other, then just stitch the two together removing any interior vertices. I think that's how this will finally go down.

@Linheart - I did make the mistake of aligning the vertices of the objects rather than using the offsets you suggested, but since there were so few vertices anyway, and I'd aligned them beforehand, I ende up with just a couple of small triangles that were easy enough to fix. Then of course I discovered that the fin and the engine still have their own surfaces overlaying/underlying each other, so the shading problem is still a problem.

Its the fact that the boolean A+B operation actually doesn't see these things as separate shapes that intrigues me although, perhaps significantly, the fin leading edge does extend slightly below the engine structure and perhaps the boolean saw this and behaved differently. Perhaps if the leading edge bottom vertex was inside the engine mount it would behave differently... hmm, thinking aloud. I might try that one.

After that it'll be the "bridge" process you both suggested. For me booleans are nice and quick until, suddenly, they aren't! This is one of those times. At least I'm getting a handle on the real problem now. Its like doing the lottery, only with worse odds sometimes.

 

[lionheart]

Here is a little trick I used to do with Windows cutouts in fuselages.


I would take the cutter and the fuselage and create a secondary 'scene' save. This new scene would be renamed, something like windows. I then would cut the windows out of one scene, then cut the fuselage out of the secondary scene. If they both cut correctly, I would then merge the windows into the fuselage scene and carry on. I could always backstep to a previous scene in order to redo the Boolean if needed.


Hope that helps.

Sounds like a difficult path to traverse. :S I do remember though that cutting windows out was easier this way, so I guess I can see your point.

 

[Here for the Beer]

I've ended up getting a decent shape using the unskilled manual labour approach rather than looking for a Gmax trick. Sometimes there are no tricks of the trade that will beat Gmax into submission. I carried on yesterday bumping against that brick wall of boolean behaviour, I made a Saturn-like planet (sphere and ring) and booleaned them together just to remind myself that this was how "union" is supposed to work, the ring stops at the sphere and creates a load of vertices around the sphere's surface while losing its central vertex - it was very pretty, and solid.

Spent the rest of the day arguing with boolean, Slept on it. Woke up. Thought "why not just get rid of the huge polygon that slides down behind the engine intake polys and just rebuild one from scratch in the hole, and which abuts up to the engine polys?", so that's what I did - just drew the accursed things in. It ended up being more than just one hand-built poly, of course (this is Gmax modelling after all...). Presently the result doesn't look a whole lot better than it did previously (fine most of the time but when the incidence of the light is an acute angle the imperfections really show up).

Now I can at least smooth it as a complete structure since it has one contiguous surface, and I can control that surface by subdividing it and adding new polys that I can use to recontour the fairing/fin merge. If you look at the Falcon 900 tail/engine fairing it has a deceptively simple look until you see the curves by looking along the structure. Because all Falcons use some serious area-ruling around the back end, there are no two panels that share the same line, everything is constantly curving to meet something else. This is AI, so I make compromises, but the trouble is that the look is so specific that when it is wrong enough, it really shouts out. It's like the De Havilland tail fin on the Tiger Moth, or the Spitfire wing - draw one even slightly "off" from reality and it really screams "this is so wrong!". Some shapes are very intolerant of inaccuracy.

Some images. CS-DFB is the newer model.

Old tail - lighting from behind really shows up the change of plane from engine surface to tail surface. I'd call it a seam except that it doesn't join those two surfaces together.


Work-in-progress tail from this a.m. with initial joined and smoothed panels. Still not great, but better. I will cut the tail (slice) from the crook of the join at the top of the engine nacelle rearwards and downwards taking an angle more towards the tail. I can then shift the new vertices created, only about 8 or 10 slightly apart to each side to create a gentler transition from engine to tail and smooth the shadows even more.


Light from the front showing up the slightly nicer look of the new engine fairing.


One day I'll get my own picture host for these projects... presently I borrow someone else's!