Home › Technical Talk
# so ... can anyone here do subdivision in their head?

AKA, can you look at a given section of a control cage and work out in your head how it's going to subdivide based on Catmull-Clark subd? Because if you can, you're either a savant or a robot. Pretty sure Perna is a robot, so the answer for him is yes he can, but what about the rest of you?

I'm not talking about having amassed a collection of subd tricks and mental shortcuts to achieve predictable smoothing, I'm talking about being able to look at an arbitrary combination of [x]sided polygons and [x]poles at arbitrary angles to one another and be able to figure out resultant facepoint, edgepoint, and vert positions:

Seriously, good luck figuring out isoline verts. I know what each of those variables is, but I sure as hell can't do that in my head. I'm not even sure how to simplify it into a rule of thumb (the higher the valance of the control cage vert, the closer the resultant isoline vert is to it in space?? I don't know.) Except, if you can't do the calculations, how can you predict things like the following?

(That's an old experiment based on a Perna example. FYI, the answer to eliminating that bump in the subd is to make the edge highlighted in blue shorter.)

In general, I'm assuming "learn subd" REALLY means "learn a bunch of tricks and mental shortcuts and rely on trial and error when you have to." Because the above math is impossible for most humans modelling things on deadline.

I still need to go through the subd page on the wiki here to learn more tricks and general principles abstracted from the math, because the math on its own seems like a dead end to me. Am I wrong?

I'm not talking about having amassed a collection of subd tricks and mental shortcuts to achieve predictable smoothing, I'm talking about being able to look at an arbitrary combination of [x]sided polygons and [x]poles at arbitrary angles to one another and be able to figure out resultant facepoint, edgepoint, and vert positions:

Seriously, good luck figuring out isoline verts. I know what each of those variables is, but I sure as hell can't do that in my head. I'm not even sure how to simplify it into a rule of thumb (the higher the valance of the control cage vert, the closer the resultant isoline vert is to it in space?? I don't know.) Except, if you can't do the calculations, how can you predict things like the following?

(That's an old experiment based on a Perna example. FYI, the answer to eliminating that bump in the subd is to make the edge highlighted in blue shorter.)

In general, I'm assuming "learn subd" REALLY means "learn a bunch of tricks and mental shortcuts and rely on trial and error when you have to." Because the above math is impossible for most humans modelling things on deadline.

I still need to go through the subd page on the wiki here to learn more tricks and general principles abstracted from the math, because the math on its own seems like a dead end to me. Am I wrong?

## Replies