View Full Version : 3D Theory: Understanding Subpatch & Catmull-Clark

02-10-2011, 06:58 PM
Greetings everyone.

I'm working through the training material I have for LW9.6. The other day I completed Dan Ablan's 3D Garage course, and today I have started to work through WordWare's Essential LightWave v9.

In Dan's course, and now on page 61 of the ELW text, I am introduced to Subpatches and Catmull-Clark. I can readily see what both do, but as I am new to CGI, do not understand when I might want to use one rather than the other.

I was wondering if there might be gallery examples that helps me understand, or perhaps something else?

02-10-2011, 07:28 PM
In Dan's course, and now on page 61 of the ELW text, I am introduced to Subpatches and Catmull-Clark. I can readily see what both do, but as I am new to CGI, do not understand when I might want to use one rather than the other.

I believe it's the context in which you use them, and this especially important because you're using LW (not some other app's implementation of CC). I generally do not use CC either way, in Maya nor in LW for varying reasons. First, because I move between multiple applications, CC is not the same across, and some apps, like Mudbox, doesn't really use CC to a degree that's readily usable in production. So despite having advantages, such as ngons, it doesn't translate well in the bigger scheme of things. In LW, CC is also very slow.

In contrast LW subpatches are identical to other smoothing functions across different apps and they all support quads and tris. So consistency across apps is maintained. Subpatches are fast too, and UV textures are solid because it forces you to keep to quads or tris.

02-10-2011, 07:43 PM
In LW typically you'll want to stick with subpatch because it is quicker and has less issues and problems (with things like UVs). In an ideal world I would just use CC all the time but I usually use sp.

CC smooths a little rounder and nicer IMO but most of the time it's really hard to see the difference and you can use ngons in CC models. CC also has edge weighting to help control edge creasing/sharpness but especially for someone just learning I think it's important to know how to do it without edge weights by using extra control loops first. You can also work in CC and then when you have time to turn all the ngons into quads (and/or tris) you can convert to SP.

Also the amount of subdividing is much different for the two methods at the same level. CC is much higher.

But a model will usually look the same if it's CC or SP at the appropriate level of divisions.

faulknermano beat me to it... :)

02-11-2011, 01:10 PM
Also note that if you add edges after setting edge weights in CC, your model goes crazy and you have to reset all edge weights to 0 and then set them back up. This makes modeling with CC and edge weights pretty much useless.

02-11-2011, 01:51 PM
Thanks for the replies. It looks like I'll be learning more about Catmull-Clark in ch. 15.

Thanks again.

02-11-2011, 04:15 PM
The main difference is that CC works fine on polygons with more than 4 vertex while subpatch just works on 3 and 4 vertex polygons.
Also CC is slower to render but the result is smoother...

02-11-2011, 10:40 PM
Oh thats was helpful guys.

I just started to learn subpatching this year, and its nice to know the difference between the two.


02-13-2011, 03:51 PM

With Subpatches you can have better control of the number of divisions at higher levels, since Catmull-Clark will go from 16 to 32 to 128, while with Subpatches you can fine tune it to something like 18 or 24, for example, depending on how much you want to push your machine.

Catmull-Clark will always subdivide to quads (poles), while (Doo-Sabin) Subpatches will subdivide quads to quads (poles) and triangles to triangles (no poles)

Catmull-Clark steeper change in the division number with the level number is quite good for steeper changes in APS (Adaptive Pixel Subdivision, per Polygon Level)

Catmull-Clark Edge Weight is tricky in LW and better done as a last step after the whole modeling is complete. (sam rule applies for making UV maps in LW, they are tricky and can get messy when the mesh is edited.)

Catmull-Clark UV will be always Linear (not true for other softwares). With Subpatches you have more interpolation options.

While Subpatches Point Weight will push divisions to the edges, Catmull Edge Weight will keep divisions evenly distributed, like so:

Anyway it will depend of what you are looking for.

I would say that even if you will end up using Subpatches in the end, usign Catmull-Clark in the work in progress makes more sense, since you won't have polygons snapping to Faces state because they have become N-Gons in the modeling process.


02-13-2011, 07:10 PM
Subdivs aren't implemented equally on all programs. RenderMan for instance use Catmull for obvious reasons, but the subdivision depends on the size of the micropixels, so it changes with the distance of the object to the camera.