Hi gibs,

> eventhough it looks clean in the viewport.

Yeah, it can be deceptive.

It's possible for certain kinds of problems to be localized into a really pretty small area at the end of the surface which can make it difficult to see. But Filleting will still be sensitive to such things.

Here's a kind of more gradual zoom in to one juncture you have there, you can kind of see here that you've got a swoop or cusp in there but at a pretty small level. Notice how it looks like it would be smooth from a further out zoom level but it actually is not:




So one of the ways NURBS surfaces work is that a NURBS surface is a kind of quadrilateral "net" of curves, with one set of curves running in the "U" direction and another set running in the "V" direction.

The normal of a surface is determined by the perpendicular taken from both of those directions.

When you have a corner of a surface where the corner is "degenerate" (both U and V directions going in the same direction at that point) it can cause an ambiguity in how the normal is defined there.

So when you want to build a surface with something like Sweep or Network, you need to give a kind of semi-quadrilateral layout with more distinct corners.

So for example this is a clean distinct corner:




On the other hand, this kind of corner is not good for the underlying NURBS surface:



That's because the U and V curves in the corner do not have distinct directions, they have parallel tangents there.


Another kind of problem area is when the U and V directions are parallel but pointing exactly opposite, which is like this:




Now, none of this applies to the arrangement of "trim curves" on the surface, only to the structure of the underlying surface.

In your case, you do have one of these kinds of corners in this area:



So that's kind of another general problem area with that kind of shape.

- Michael