CREASE EDGES FOR SUBDIVISION: THE BLOODY CORNER

mecons

Hi guys

I have a big question for you today.

I was trying to better understand the 3 edge-loops technique for hard surface modelling in 3ds Max when I discovered this kind of corner that in my little experience is not possible to subdivide without having noises on the surface unless another technique is used (nurbs modelling or creasing by smoothing-groups).
I can't really find a way of sorting it out and there isn't any specific tutorial on internet, I hope someone can give me the solution.



In this kind of shape the 90° angles subdivides correctly with the 3 edge-loops technique, the 270° angle gives problems as shown in the pictures.










The fold created is huge and not acceptalble on a rounded surface.





I've also tried to use Bevel tool in order to strengthen the creasing, but the result was even worst.






Any suggestion?

Thanks a lot in advance!

Bartalon

You should just be able to throw some support loops across all directions of the 270 degree corner piece:

mecons

Bartalon wrote: »

You should just be able to throw some support loops across all directions of the 270 degree corner piece:

Dear Bartalon, thanks a lot for your reply.
I'm aware of that, maybe it is better if i use a more appropriate example.

This is almost the same case but with a spherical surface, here the extra loops would cause 2 hard lines on the spherical part.
How could it be sorted out?

Will Faucher

In that case you're going to need more geo to maintain that curve you have going and avoid the hard edge along the curvature.

Bartalon

What Will said. The principle generally doesn't change, but if you need to support something along a curved surface you'll need a denser base to begin with.

Mark Dygert

That shape is going to make it hard to reinforce that corner edge. I personally would use a little more geometry to define the roundness of the shape which will give you more options to terminate loops.

The problem you seem to be having is that you're leaving the corner loop alone to fend for itself and not supporting it in some way.


This thread: http://www.polycount.com/forum/showthread.php?p=1393320#post1393320 has conquered this issue in quite a few posts so I would check it out when you get stuck.

Personally I would use a little more geometry to round out shape of the object which will give you some options to route that corner loop and it's supporting loops.

But there are probably different ways to handle it.

mecons

Thanks a lot Mark, Bartalon and Will, now I have much clear ideas: there is no magic trick, that case needs a proper topology and more geometry.

deolol

set different smoothing group to each edge loop and smooth, then you dont need supporting loop

Rooster128

What others have been saying is true- you need to place the loops adjacent to the actual corner to get the corner sharpened with smoothing. This can make a spherical or cylindrical object deform in unintended ways- but its possible to account for this by making sliding the edges in the loop that aren't needed for the corner outwards so they are more uniform with the other edges.

The last pic in Mark's last point kind of illustrates this- you can see the loops tight near the corner, but they eventually broaden out to help sell the outwards curve.

mecons

deolol wrote: »

set different smoothing group to each edge loop and smooth, then you dont need supporting loop

Yes My friend, it is true, but the subdivision by smoothing groups causes very sharp edges which is not always aesthetically correct, In my opinion the 3edgeloop technique is the most versatile and the one that allows you the best control on subdivision.
cheers :-)

mecons

Rooster128 wrote: »

What others have been saying is true- you need to place the loops adjacent to the actual corner to get the corner sharpened with smoothing. This can make a spherical or cylindrical object deform in unintended ways- but its possible to account for this by making sliding the edges in the loop that aren't needed for the corner outwards so they are more uniform with the other edges.

The last pic in Mark's last point kind of illustrates this- you can see the loops tight near the corner, but they eventually broaden out to help sell the outwards curve.

Thanks for your precious opinion Rooster, at this point I wonder if Nurbs are not the best way of doing mechanical modelling. I mean maybe it would be better to make a retopoly of a Nurbs model and then crease it's edges than making a subdivision hard surface model from scratch.

What you guys think about ? How is your mechanical modelling workflow?