Subd pinching problem
polycounter lvl 8
So I decided to learn subd modeling. However I've run into a problem with pinching as seen on the image below:
I assume that it is the pole that is causing the problem and I tried to do this in two ways but I ended up getting the same pinching error.
First I tried to just bevel/extrude and just add an edge loop around the center to get a more squared look (top right image) but it caused a lot of pinching. So I browsed the forums and other sites and people seemed to suggest to just bevel the edges of interest. I tried that (bottom left) but alas the pinching is still there.
I've checked the wiki, searched the forums as well as any external links I could find. But I can't seem to find a solution.
Now I've run out of ideas and I feel really dumb cause this must be one of the most rudimentary problems out there.
I assume that it is the pole that is causing the problem and I tried to do this in two ways but I ended up getting the same pinching error.
First I tried to just bevel/extrude and just add an edge loop around the center to get a more squared look (top right image) but it caused a lot of pinching. So I browsed the forums and other sites and people seemed to suggest to just bevel the edges of interest. I tried that (bottom left) but alas the pinching is still there.
I've checked the wiki, searched the forums as well as any external links I could find. But I can't seem to find a solution.
Now I've run out of ideas and I feel really dumb cause this must be one of the most rudimentary problems out there.
Replies
Hopefully that actually made sense.
I made another picture to show the problem:
It is all quads except for a few ngons (cylinder caps and the inside of the insets) but they shouldn't be affecting much.
I mainly ask this because your error is happening where I assume you've welded together duplicated instances of the same original piece (the inset).
As for what Cojax says about triangles in sub-d meshes, that's not true. Use any sort of n-gons you want, obviously quads will subdivide more "predictably", but you can easily use 5,6,7+ sided polys if you want, and triangles if you need. The main thing is what the subdivided result looks like, doesn't really matter how you construct it.
The main reason for keeping quads other than predictable subdivision results is so that you can use loop/ring selecting tools, so obviously it is a good idea to work mainly in quads, but there's no point spending time trying to get rid of tris/n-gons if the resulting subdivided mesh looks good.
Basically to fix this you will need to have more divisions running down your main cylinder piece.
The pole (star geometry) isn't helping either, since the more edges that are connected to a vertex, the more "weight" the vertex will have (retaining its original shape) when subdivided.
Similarly, the closer together verts/edges are, the more "weight" they will have when subdivided, which is why if you want a "sharp" edge, you bevel it or put loops on either side.
Thankyou.
I also tried Microneezia's suggestion and that didn't work at all.
If anyone could be kind and show how they would do this I think it'd be really helpful
Bah...good to know, having issues at home on a mechanical model I'm trying to do.
Like a lot of solutions to complicated mesh with sub-ds, sometimes the answer is just to throw more geometry at it! since you want to retain those curves you have, the # of sides shouldn't = the inset features, it should be more, in this case i used 3 sides per indent, then added in some padding edges around each indent, then beveled padding edges around the shape, and added edges to retain the shape etc. And then finally, remove the excess loops so you dont have sharp edgse every 3 sides. Thus creating NGONS!!! MWUAHAHAHA
also http://johnyontehspot.com/pix/shapething.obj
You forgot the triangles I believe