Hi Michael,

I think one of your slogans for MoI should be "Nurbs that Work!" I recently had to do some nurbs modeling in Softimage and I'll tell you, it was a huge relief to get back to MoI! I really like what you've done in the new Beta! Keep up the fantastic work!

It's been a while but I'm starting another major project using MoI. It will be a similar theme as the B17 (http://moi3d.com/gallery/viewitem.php?id=99). This going to be a somewhat distant object so it won't be highly detailed but it's just a start.

As I've been working I've had a few thoughts on what might help the overall workflow. I've attached the models for you to look at if you get a chance.

1) There seems to be a bug in the mirror command. If I try to union the main body of the jet to the wing tips much of the model vanishes. This doesn't happen when I delete the outer starboard wing section (which is a mirror of the port wing section).

2) For some reason I can't union the body to the nose section, even though they share the same construction curves.

3) The engine proved to be a very difficult shape and it still isn't quite right. One idea I had that would make this and many other operations easier is this: Would it be possible to edit the boundary curves (not construction curves) of an object and have that represented in the surface? An example would be to trim a surface at a specific location and then use the resulting curve to further edit that surface. That would make MoI so much more flexible to work out non-symmetric surfaces. I have no idea if that's even possible but just a thought. Maybe there already is and I just don't know it!

4) When using tools such as (but not limited to) revolve, it would be nice to be able to specify some divisions (example: revolve with 6 or 8 divisions) and then be able to edit the curves at those divisions to modify the surface. This would have been great on the engine for example.

5) When exporting to Softimage using FBX or Obj formats (and perhaps others), the normals are always reversed. Not a big deal once you know but confusing at first.

6) An N-sided patch would be most welcome as would be a variable radius fillet.

7) When I try to create a network from the wing tip curves, it makes a mess, but if I do it 3 curves at a time and then join the surface, it's OK. Is it the order I'm picking them or is it just not possible to patch that whole area in one shot?

By the way, thank you so much for the link orthographic views function!

Thanks for the best nurbs modeler around!!!

Cheers!

Hi Michael,

You can scratch #3 and #4, I was doing it all wrong. I sure feel silly now! :) Seems I can show points and edit a surface just fine. It wasn't working for me before but I think it was because I was trying to show points on a joined surface. Silly me.

I did the engine again and it was so easy! Took about 2 minutes.

Cheers!

Just one more possible bug. I can't seem to trim the flap on this wing section. If I trim in different directions it sometimes works but if I go to trim the flap again, it mangles the object. I don't know if it's because of he way the surface was built or something I'm doing wrong. The same trim method on the wing section just beside it worked fine.

Point 3 and 4 could be solved in Rhino.

You can rebuild the surfaces to have less control points which allows you to tweak the shape of those surfaces by pulling points. MoI however doesn't have that option as most surfaces generated have so many control points that it is not practical to tweak them. This would certainly be a great feature to add to MoI though as it is often difficult to create some of the shapes required in a aircaft using default surfacing techniques and a bit of point pulling would overcome those problem areas.

If your using the November Beta Michael posted a patch for the mirror issue in this thred:

http://moi3d.com/forum/index.php?webtag=MOI&msg=2177.10

Hi Len!

> 1) There seems to be a bug in the mirror command.

Yeah there is a bug in the Nov-21 v2 beta which kind of messes up the directions of objects that are mirrored. This one should be fixed if you apply the patch from Burr's link above. You may want to delete your previous mirrored objects and re-mirror them once you're patched.

Also for the next v2 beta I have tuned things up some more beyond that so that the surface-normal direction on open surfaces will not flip direction when they are mirrored which will currently happen. However if you apply the patch the normal direction for solids should be fine currently even after mirroring.


> 2) For some reason I can't union the body to the nose section,
> even though they share the same construction curves.

Generally you might try using Edit/Join rather than boolean union when you have 2 surfaces that share a common edge already. Boolean union is sort of more intended for pieces that need to be intersected rather than already having touching edges... But join still has a problem with this one too, I'll need to debug into that some to see why.


Re: #3 / #4 - it sounds like you got these figured out now?
Did you find the info here: Q: Why does show points work for some objects but not others?


> it would be nice to be able to specify some divisions (example:
> revolve with 6 or 8 divisions)

I would like to set that up as a choice, but right now you can set that up by using Transform/Array/Circular to make that radial arrangement of curves and then do a Loft through them with the Closed option. Then you should be able to edit those curves and watch the Loft update.


> 5) When exporting to Softimage using FBX or Obj formats
> (and perhaps others), the normals are always reversed. Not a
> big deal once you know but confusing at first.

On everything including closed solids? Or are you talking about on open surfaces?

MoI doesn't automatically know which side of an open surface should be the "outside" so things like that may have either normal orientation. But if you have a solid MoI should be able to automatically determine the "outside" direction of that and point the polygon normals in that direction.

Can you please do a test with some basic solids like a box and a sphere and see if those are reversed for you or not?


> 6) An N-sided patch would be most welcome as would be a variable radius fillet.

Yeah I do expect to get to both of these eventually. They are each kind of tricky though.

Variable radius fillet is mostly a matter of getting a good UI done for it. But good UI design tends to be a time consuming area of work as well.

The N-sided patch is a bit harder yet then that, it is something that tends to have oscillations in the generated surface unless it is done pretty carefully.


> re: #7 - network surface

Basically Network needs to have a curve arrangement that can be unwrapped to a 2D grid pretty much the same as how a globe of the earth with latitude and longitude lines can be unwrapped to a 2D map.

So similar to a globe, you can only have 2 "pole" regions where endpoints of more than 2 curves all collapse together.

In your case if I separate the curves a bit for illustration, you can see you've got 5 poles:



So basically that current arrangement is not right for Network all in one shot.

To do it all in one shot probably the easiest arrangment is to have the curve on the left here:



be joined together all as one long single curve rather than as 4 separate curve fragments. Then that will have an arrangement with that long curve similar to the equator line of a globe with the other ones coming to touch at one "pole" at their other ends.

Does that make sense?

Thanks for posting the models, that really helps a lot in analyzing what is happening! I'll see if I can find out about that nose cone joining problem.

- Michael

Hi Len - the Join/Union won't work because those edges are 0.007 units apart from one another and Join will only glue edge together that are up to 0.005 units apart from one another.

The root cause of the gap is that Network uses a lesser tolerance than other things to try and reduce the density and complexity of the generated surface. But in this case that reduction of complexity caused just enough sagging away from the original curves to cause a join problem.

At the moment a workaround is to select those objects, and scale them down by 1/10 in size (run Transform/Scale, type 0 [enter] to pick the 0,0,0 origin point as the scale origin and type in 0.1 [enter] to specify a 1/10 scale factor) That also shrinks the gap between them so that Join will work on those objects, then scale up after they are joined.

I'll think about what to do to help fix this up, I could try to tighten the tolerance on the network, or maybe I should create the result of Network with the original input curves as the trim curve boundaries rather than using the natural surface boundaries of the generated surface like it does now.... I'm not quite sure yet.

- Michael

Hi Michael,

Thank you so much for your very insightful explanations. That really helps!

I will do some tests with Softimage and the normals. Initially, it seems that ever object is reversed except objects that have been mirrored, they seem OK but that's just a quick observation. I'm not that competent in Softimage yet having spent years on LW and only recently used Softimage so it might be something to do with Softimage.

Thanks again!

Yes kevjon, a rebuild surface with less points would be helpful in many cases. Hopefully in the future.

Thanks for the link BurrMan. This is a great forum!

Cheers!

Hi Michael,

I got a chance to test some normals with Softimage. It seems that basic objects such as cubes, cylinders, spheres, planes, cones and so on are just fine. Somehow the normals got reversed on this aircraft model (I assume somewhere in the construction phase) but I have no idea where.

I'll let you know if I figure it out.

Cheers!