Hi Rogurt,

> Here´s the scene file

If you turn on surface control points it should help you to see what's going on there:



So note there that the control points are not all one parallel plane slices - the middle row of 3 control points are on a plane but their plane is rotated a bit. It's often that kind of rotation or wobbling of the internal structure of the surface that makes for a bit of a bulging shape where it ends. The same effect does not really create as much problems when you make an extended larger shape instead of separate small patches like this.

The canonical way to build a sphere shape in NURBS is as a surface of revolution and not by Network.

Also working with 3 sided areas can add in more difficulties, since NURBS surfaces are inherently 4 sided things, so to make a 3 sided patch one side is compressed to a point and that often makes for some slightly different shaping behavior in the direction that collapses down like that (although not with Revolve).

It is actually possible to surface your sphere here patch by patch if you use Sweep instead of Network - if you select this curve here as the profile for the sweep:



Then run Construct > Sweep and pick the other 2 adjacent curves for the rails. This will initially generate the same kind of slightly bulged thing like this:



But with sweep for some particular cases like this you can enable the "Maintain tangent" option which will construct the sweep all out of parallel planar sections, which basically avoids wobbling the shapes as the travel along the rails. That will generate the kind of result you were looking for:




If you turn on surface control points for this result you can see the difference:




Also note though that the surface is a lot more complex and dense, with the sections restrained to only be on parallel planes it usually requires a lot more of them to be used to generate a surface that matches close enough to the rails.

So it's not a way that I'd recommend for actually building a sphere - a surface of revolution is more accurate and also lighter at the same time.

But I hope that this may illustrate to you why you get the results that you were confused about - when the surface is marching along in 3D in a totally generic way like with network there is no special restriction for its structure to be all on parallel planes and that kind of general rotation is what causes the kind of bulged shapes at the edges that you were seeing.

I've attached a new version of your 3DM file that was built in octants using this particular sweep method instead of Network.

- Michael