yea, i've tried the other methods, but the results are not good. i could redo the whole shape in order to make it work with network, but this seems like it might be a bug or something. which curves can't you see? maybe i uploaded an old version of the file. heres another one

Hi jehan, thanks for posting the file to look at.

There are a few different things going on here that make things difficult for sweep - the biggest one is that your rails don't touch at their ends, there is a slight overlap between them:



Here I moved one off to the side slightly so you can see the overlap:



Having those rails overlapping each other instead of ending at the same point is causing the worst problem - that creates a really crazy type sweep situation in that area where the generated surface will try to fold back inside of itself.

It looks like this may have come from doing a mirror where the mirror axis was not quite in the right place, like maybe the mirror axis got snapped on to some other object instead of the endpoint of this curve like you wanted.


Then actually on top of that there are a few other difficult things, such as having non-planar rails which will cause the profile curves to kind of tilt forward or backward as they move through the sweep. You may need to add an additional profile down towards the end of the sweep to try and keep control of the profile's "up" orientation as it moves along these rail curves that are curved in 3D.

Then also the dense rail curves are ok for some situations, but tend to cause problems in the area approaching the tip which is usually a kind of a problem area.

You may need to redraw those with a fewer number of points, there seems to be some slight wiggling in the ends where the pieces come together, that is getting exaggerated by the sweep and making wiggles in the sweep near that end.

I think that with those very slight wiggles in the rail curves in that area, that tip is just not going to work well with these rail curves.

- Michael

wow can't believe i didn't notice that- thanks. i've really got to tighten up my modeling technique. maybe it would be worth it (after the million other things you have to do of course) to put in some sort of "auto snap" that would correct these tiny discrepancies. actually what might be better is to just have the program highlight or draw a circle around places where stuff comes really close but doesn't touch. that way you could correct it yourself if you wanted. of course this kind of stuff is no substitute for clean modeling practices, but it might make moi more artistic and intuitive to use.

jehan,

OK. Your biggest problem was as Michael pointed out.

Correct them first, then loft three curves using profiles and scaling curve. This gives you a close match
(now we need match Michael !!)...ok then sweep the surface as you did originally. Because you have kinks in the
curves it builds a polysurface which is rather handy....split that polysurface and also the first one and join together.

Hope that makes sense for a quickie ?

Hi jehan,

> maybe it would be worth it (after the million other
> things you have to do of course) to put in some sort of
> "auto snap" that would correct these tiny discrepancies.

It's definitely a good idea! But unfortunately in practice there are many difficult things in implementing this.

One big problem is that it could be pretty time consuming to try to examine collisions between all different kinds of shapes in a complex model, it could really slow things down a lot for all cases even if the final check did not detect any problems...

It is also difficult to analyze the kind of "chain" of behavior, like for example in this case with mirror you could actually be snapping the mirror axis line on to some other object and not even nearby that end point... So like in that kind of case it is hard to adjust the snap point of the mirror line to take into account the result of the mirror instead of just the points you are actually snapped on to....

There are just a lot of difficulties similar to that...

- Michael