Hi Michael,

It has been a long time since I've seen such some many flipped normal when exporting a model (using ngones, anlge 3, welded)
Can you confirm that's 100% modo bug ?

Hi PaQ - are you certain they're flipped? Or did it just fail to triangulate those n-gons which is needed to get the n-gon to actually display. I've seen a few cases before where LW/Modo's triangulator fails with a few kinds of point arrangements, it may not deal well with some of those vertices that are all in the same line.

What happens if you do the "flip twice" trick inside of Modo - does that make it better or not?

- Michael

Hi PaQ,

> Can you confirm that's 100% modo bug ?

Well, the first thing is to see if it's a poly normal flipped problem or if it is a n-gon triangulation problem.

Maybe try deleting the back half of the object - then when you rotate around to the other side do you see those n-gons or are they still missing when viewed from either side? If missing on either side then it's a triangulation problem.

Then the other thing that I'd do is to move some points around on one of the bad n-gons and see if there are any surprises in there like points stacked up where you wouldn't expect possibly making a small self-intersection area or something like that.

Is there any way in Modo to display which point is the starting point in the list of vertices in an n-gon? Modo and LightWave are sensitive to which point is set as the starting point because they determine front face/back face by only looking at the directions of the edges from that first point, so if that first point is in a degenerate area that could be a problem. The double-flip method also sets the first point of the n-gon to a spot that it thinks is good.

I'll take a look at it some too.

- Michael

Hi Michael,

No the flip/flip dont resolve the problem.
Something weird is that the ngone that looks like a triangle has 6 points. In fact it's a triangle, but the vertices are duplicated. I can merge the vertices in modo (with 0 distance radius), and it resolves the problem.

I'm wondering if its maybe a side effect of the special exporter for creating this famous wireframe hack.

Hi PaQ,

> No the flip/flip dont resolve the problem.

What about if you go and look at it from the other side, can you see it on the other side or is missing from both sides?

Missing from both sides most likely means triangulation failure due to some degenerate too-close-together points in the n-gon.


> I'm wondering if its maybe a side effect of the special
> exporter for creating this famous wireframe hack.

No, it shouldn't be related to that - that segment generator thing didn't change anything about how mesh vertices are created.

You've just got a case here where the points from the trimming edge ended up close to the points from the untrimmed underlying surface refinement but not quite close enough for them to be get merged into a single point. It's probably something I could tune up a bit, but it's a pretty delicate area because glomming points together too much can mess things up also.

- Michael

»» What about if you go and look at it from the other side, can you see it on the other side or is missing from both sides?
Yup, the poly is missing in the booth side.


Ok I have replace moi_lib.dll with the original one, and restore the .ini. (sorry you reply too fast :))

The problem is still there ... there are 72 overlapping vertex in the object. After the 0 radius merging, and a double flip, only one triangle if still missing.
Looks like the model generate a lot of small radius overlapping vertex. (an other merge with 0.1 mm radius remove an other 138 vertices :S)

If they're exactly stacked on top of each other, that's something that I could probably be able to tune up better without any bad side effects though.

- Michael

»» If they're exactly stacked on top of each other, that's something that I could probably be able to tune up better without any bad side effects though.

Well it's up to you, I have to admit it's probably not the kind of geometry I'll produce normally, I was trying to bug track something different in modo :)

> Looks like the model generate a lot of small radius
> overlapping vertex.

There's basically two kinds of refinement that happens when the mesh gets created - the underlying surface gets broken down into quads, and then apart from that trim curves get broken down into some little edges as well.

In the pieces here where you've got a surface cut at like a 45 degree angle, some of the vertices from underlying surface quad refinement are really close to the vertices from the trim edge refinement - something along those lines anyway.

- Michael

Hi PaQ,

> Well it's up to you, I have to admit it's probably not the
> kind of geometry I'll produce normally,

Well, if I can produce better mesh output without any bad side effects from a change, then I'd be all for digging into it.

But if your immediate problem is solved by merging vertices in Modo, I'll probably put it on my list of stuff instead of worrying about it right this moment.

Does the merging step in Modo have any bad side effects like throwing out vertex normals?

- Michael

Hi Michael,

Yes I did a final test, merging the model at 100mu, and everything looks right in one step.
So the problem is solved from my side.