View Full Version : confused on the whole subdivision surfuce modeling

09 September 2003, 08:27 PM
IS subd models just a smoothed version of polygon models.
Im confused cause i know with wings3d if you hit shift-tab it subdivides the model. but does it just do the same as say if i selected the model and just smooth?
how about in maya or realsoft.
i know im maya you need to convert thepolygon model to a subdivision surface. To get teh smooth surface ( supposedly like a nurbs surface with no edges) I know realsoft3d is likethis also.
now with silo to subdivide it looks like if im just hitting smooth (in maya or wings) may times.
So what does it mean by a subdivision surfaces modeling. thanks

09 September 2003, 08:51 PM
The implementation of Sub-D's varies from program to program but generally speakign the tools are often similair or the sam etools you'd use to edit polys. The main difference between polys and sub-d's is that Sub-D's can have adpative resoltuin unlike polys which are fixed. You can smooth polys but to my knowledge there is not a program that can do it adaptively. This aspect of Sub-d's make them similair to nurbs in that you cna have the tesselation adjust based on the proximity of the camera to the surface or based on the amount of curvature in a specifc area. Most poly smoothing happens globally to an object where may create more faces for an area than it needs to look smooth.

In maya you have two methods that are very similair but one produces a poly mesh and the other a sub-d mesh but both can use a poly cage to manipulate, shape and aniamte a higher rez mesh. Smooth poly proxy uses a low res poly cage to create a "smoothed" shape. Likewise with a Sub-d object you could use varioud levles of divisions to create low res cage andhave higher detail on the lower levels. You can also use a poly proxy for the sub-d which is the most like a smooth poly proxy but is is slow. The difference between these and a smoothed object is that you can easily use the low res cage to animate the high res which is usually a very effective and fast way to work with smoothed or sub-d meshes.

Depending on what program you are working with .. if you do your modelling in Wings3d you may want to export the low res unsmoothed cage and utilize special features of the animation program rather than be locked into working with a high rez and potentially unwieldy object.

09 September 2003, 08:56 PM
Everybody talks about subdivision, but there is actually two types.

The first type is 'true' hierarchical subidivision. You start with a low poly cage, you subdivide it one time, you tweak the result, you can subdivide it again and tweak it more.
When you go back to the low-poly cage you can change the shape, and the end result will follow.
The advantage is that you can add more detail without having to resort to a very dense cage.
The disadvantage is that true subdivisions are slower to work with (takes up more CPU resources).

Most applications have a the other type of subdivisions, which is one lowpoly cage that is subdivided (multiple times) There is no option to tweak the subdivided result.
This type is called 'NURMS' meshsmooth in 3DS Max, and 'Smooth Proxy' in Maya, and MetaNurbs in Lightwave (if I'm correct).

As far as I know most people are using the second type (the non hierarchical version) to make their models.

Gmask: hehe, 100 points for the double post COMBO!

09 September 2003, 09:04 PM
Marcel and gmask explained subdivision surfaces nicely. I just want to add that usually you can add crease and pinch factors on a subdivisional mesh. Well atleast, those are optional tags listed in the Renderman 3.2 specification. Those type of subdivisional surfaces use a Catmull-Rom subdivisional method. In many modeling programs you can do something similar. You select an edge or point and you can specify how sharp you want it to be after the subdivision operation.

09 September 2003, 09:11 PM
ok now if your going to render out ribs in may and your model is already a subdivisional surface so you still need to select make subdivision surface. cause i know if you have a poly and select make subdivision it renders out a smooth object. no matter how close you are... well at least its supposed to.... so when a render says it supports subd does it mean that or actually render the subd surface in maya. Since maya you need to convert the poly to subd.

09 September 2003, 07:37 PM
ok i think i get it. Thanks guys.

09 September 2003, 04:32 PM
Hierarchical editing has absolutely nothing to do with "true subdivs". This whole true / not true discussion is close to driving me crazy anyway.

A subdivision surface is a type of parametric surface, which is infinitely smooth. It can be tesselated into triangles for rendering (or in the case of Renderman, to micropolygons). It is generated from an arbitrary polygonal mesh via an iterative refining process. The result after an infinite number of subdivisions is called the limit surface.

AFAIK only Maya has support for this type of geometry within the package. PRMan supports subdivs as a type of surface.

Smoothed polygons behave very close to subdivision surfaces: they usually apply the same iterative refinement to the polygonal control cage and the UV coordinates. The only difference is that you have a limited number of iterations and thus you won't reach the limit surface; however a few refinements are usually enough for a very close approximation.

Maya, Max and XSI have such a polygon smoothing algorythm. Wings, Blender and Mirai too.

Lightwave's method is a bit different, producing very similar results for quad polymeshes.

You can see that the only important difference is in rendering, where you can have an adaptive tesselation for a subdivision surface, and a fixed parametric tesselation for the polymesh.
But in terms of modeling, texturing, and animation (skinning, deformations etc.) subdivision surfaces and smoothed polymeshes are practically the same.

Maya's implementation of subdivision surfaces introduced a hierarchical editing option, described above. Let me also note that Max has support (although not too good) for this with the HSDS modifier, too. But this is just an addition to the original specification, which is several decades old by the way.

There's some quite nice work displayed with hierarchical subdivs, most notably that of Steve Stahlberg - but I personally think that it will be replaced by extraction detail modeling. Which means that you smooth the low-res polymesh several times, use some 3D sculpting tools to add details (similar to the hierarchical editing but a lot faster and more intuitive) and then generate displacement maps for the low-res model using this high-res mesh. Results in better interactivity (no hierarchies to store and display), lower scene sizes, better models. The only requirement is support for fast and robust subpixel displacement mapping - Mental Ray looks like a good contender, and it is going to be included in 3 of the main apps once 3ds max R6 is out.

09 September 2003, 11:59 AM
I have to come back on my earlier statement then Laa-Yosh, I really thought the hierarchical subdivisions were the original specification. Thanks for clearing that up :)

09 September 2003, 07:42 AM

Polygons= easy to model -- faceted edges when you zoom in.

Nurbs = difficult to model -- smooth edges irregardless of zoom.

Sub D's = model like polygens -- smooth edges like nurbs.

Its almost like Polygons are raster images and Sub D's/Nurbs are vector images in terms of ability to zoom in.

CGTalk Moderation
01 January 2006, 06:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.