12
$\begingroup$

The top[ has some weird deformations

My problem is simple. I think that the strange deformations on the top are there because there are so many vertices in one face. How do I fix this while maintaining nice topology?

Although I triangulated the top polygon, the problem still persists.

enter image description here

$\endgroup$
2

3 Answers 3

7
$\begingroup$

Not to get too technical, this is because there is no supporting geometry at the edges where the vertices transition into the larger face (the ngon) and the subsurf modifier will stretch these (for lack of better words) giving the bumpy look.

Two simple fixes are:

Extrude and scale the top edge ring in more to the center

Just select the top ring, hit E and scale it in a bit to the center.

enter image description here

OR

Merge all the vertices on the top edge ring at the center

Select the top edge ring or the top face (the ngon) and hit AltP. You still need to add supporting geometry or the problem will persist. An edge loop near the top edge ring should suffice. It's a flat hard surface so merging to tris should be fine.

enter image description here

$\endgroup$
1
  • $\begingroup$ Both answer are good and should be merged. $\endgroup$
    – user320
    Oct 6, 2013 at 19:51
8
$\begingroup$

Rule of successful subsurfing: surround corners with a row of faces on each side (like the green cylinder) so you always have one line of faces that connects to the flat part at 180°. That way you disconnect the subsurf from spreading too far.

subsurf examples

The subsurf won't change a vertex's height if this vertex lies on a plane with all verts of all faces it is connected to. And then it doesn't matter how you fill the flat part, be it Ngon, Triangle Fan or a map of Louisiana.

At least almost. All the inner verts will keep their height, but they are still moved in the other two dimensions according to the algorithm. While the orange part of the mesh is flat, everything green starts to bend down immediately so it's not necessarily a circular coast line.

In order to get a microscopically clean solution (which might be needed for reflections), a Triangle Fan or yet another additional face line seems best.

Louisiana

Take a look at the horizontal edge line to the right. Since the inner verts have the same height as the corner verts, they keep their height and don't get pushed around (horizontally, but not vertically) which in turn means, the edge between them doesn't change either. Only the corners get bent.

subsurf explanation

$\endgroup$
7
  • $\begingroup$ what does "these stay out" mean, are the other ones moving? $\endgroup$
    – user320
    Oct 6, 2013 at 20:16
  • $\begingroup$ Perhaps add a note about Mean crease too? $\endgroup$
    – gandalf3
    Oct 6, 2013 at 20:23
  • $\begingroup$ mean crease a whole set of new problems. but yes it should be mentioned $\endgroup$
    – user320
    Oct 6, 2013 at 20:26
  • $\begingroup$ @user320 They don't stay out, they stay put. They don't move. When you use a Catmull Clark subsurf, most of your surface gets changed, rounded. Verts on a flat surface don't change if all faces that are connected to that vert lie on a plane. $\endgroup$ Oct 6, 2013 at 20:29
  • $\begingroup$ Oh so once the Sub-D is applied they will be in the same 3d space? $\endgroup$
    – user320
    Oct 6, 2013 at 20:31
3
$\begingroup$

I see two solutions for this. The first is grid fill. It looks like this:

grid fill

Nice quads all around. Make it like this. Delete the center ngon. Select the same amount of edges on opposite sides of the circle. Ideal is to select 1/4 of the total around the ring. Hit space and type grid fill.

grid fill howto

The other way is to inset the center ngon almost to the center, like this. Some problems will probably stay, but they will be very tiny.

tiny center ngon

$\endgroup$
2
  • $\begingroup$ Grid fill is also in Ctrl+F > Faces menu $\endgroup$
    – gandalf3
    Oct 6, 2013 at 18:28
  • $\begingroup$ Grid Fill adds alot of unnecessary geometry, I wouldn't recommend it in this case. $\endgroup$
    – iKlsR
    Oct 6, 2013 at 18:32

You must log in to answer this question.