Here's an example using service pack 2. Note how
the hidden edge directions differ from the way they originally were in Moi.
However in this example it does show service pack 2 handles smoothing better then before.

Maybe you are right tris and quads only work....

Hi Chris - sorry I still don't understand exactly what you're referring to there.

In your last screenshot you seem to be displaying "quads & triangles" type output from MoI, not n-gon output from MoI, is that correct?

If you output n-gons from MoI there is no triangulation stored in the file at all - that triangulation is instead calculated at import time by the Max FBX importer.

There are numerous different ways to triangulate an n-gon, it's not like there is only 1 single correct answer for it. So it's normal for different pieces of program code to produce different triangulations from the same original n-gon - that's not a bug all by itself as long as the triangulation is a valid one.

So don't worry if the hidden triangulation looks a little different between the one that the FBX importer generated when importing n-gons versus what MoI was going to produce when MoI was exporting to "Quads & Triangles" - that's not a bug that they have different triangles generated with different n-gon processing code.

If you see triangles leaking outside of the n-gon border or totally invalid triangulations then that's a much different matter.

But so far from your description I don't understand why you are concerned about it.


Why do you care if the hidden edges of the internal triangulation matches the triangulation that MoI was going to generate if you were exporting from MoI with non-n-gon output?


- Michael

Only care when they overlap a bunch in a model :( Means I have to get in there and poly edit them.
But I'm glad you stated the importer creates the
triangulation. I had wondered that. Not a programmer ;)

Hi Chris,

> Only care when they overlap a bunch in a model :( Means I
> have to get in there and poly edit them.

If you run into triangulation problems where the internal triangulation of the n-gon has overlapping triangles where some are starting to go in the wrong orientation, then that's a bug in the n-gon triangulation mechanism of the FBX importer.

That's in Autodesk code - there isn't anything I can do myself to fix up that particular problem, you will need to send a bug report in to them to get them to improve the triangulation algorithm so that it doesn't make bad triangles like that. Doing a really robust triangulator is a fairly tricky thing so problems similar to that are not too uncommon - it may help them if you can send them an example of the n-gon that does not triangulate properly so they can have a case to test with.

If that does not happen very frequently though, you can probably export from MoI using n-gons most of the time, and only export from MoI using "Output: Quads & Triangles" for cases where the FBX n-gon triangulation mechanism had some difficulty.

- Michael

What Michael says is right. The internal hidden triangulation is not a concern unless it causes artefacts. Don't forget that quads are internally triangulated as well. Non-planar polys can cause an issue but only for the Phong calculator. If the vertex normals are imported, then there should be no issue.

By the way you must turn off the Smoothing Groups option in the Geometry roll out. If this is turned on, the importer will discard vertex normals and calculate smoothing groups instead.

Here's another screen shot from Max 2012. This is the same model from MoI. I have added an edit Normal modifier so you can see the normals. This is showing the vertex normals of one of the filleted edges and they have been imported perfectly.

Sharp edges (not shown here) have 2 vertex normals, as they should.

Incidentally, the Max importer does import vertex normals but if the same fbx file is imported in to Modo, the Modo fbx importer discards the vertex normals. The problem there is with Luxology's implementation of the fbx importer. For Modo, LWO files import correctly with vertex normal data.