In what specific circumstances are tris bad for deformation, and why?
polycounter lvl 6
A lot of these sorts of (usually entry-level) questions about triangle usage culminate in the following understanding:
1) The notion that one should never use tris/ngons is reductive, and there are times when tris are actually beneficial to deformation.
2) Quads get turned into tris at the end of the process, anyway.
I'd like to explore this in a bit more depth, however. Clearly, when people talk about tris being bad for deformation, they're referring to specific tris, i.e. tris that are out of place among the ones created from quads, and not fit to purpose.
I'm aware that tris are sometimes used to act as hinges for joints, for example, so I get the sense that if tris are placed sort of parallel to an axis of rotation they can be beneficial to deformation.
Conversely, can anyone point to some general-case topology examples that use triangles that are demonstrably bad for deformation? And can it be demonstrated from first principles exactly why that's the case? I'd like to get a more robust understanding of when (or when not) to use tris, but most of the time the discussion of theory doesn't go that deep.
I'd like to explore this in a bit more depth, however. Clearly, when people talk about tris being bad for deformation, they're referring to specific tris, i.e. tris that are out of place among the ones created from quads, and not fit to purpose.
I'm aware that tris are sometimes used to act as hinges for joints, for example, so I get the sense that if tris are placed sort of parallel to an axis of rotation they can be beneficial to deformation.
Conversely, can anyone point to some general-case topology examples that use triangles that are demonstrably bad for deformation? And can it be demonstrated from first principles exactly why that's the case? I'd like to get a more robust understanding of when (or when not) to use tris, but most of the time the discussion of theory doesn't go that deep.
Replies
Most of the time, no.
Triangles are mostly bad for subdivision surface modeling. Yes tris are ok for deformation, but it's totally dependent on specific topology. Which way an edge flows with a quad, determines how the quad will pinch.
Same quad, different internal edge direction, so you get different deformation.
We have some examples in the wiki if you search for Limb Topology. For example
The best way to learn if by doing. Rig your mesh and animate it, see where the deformation happens and why. See how massaging weights won't always solve the problem... often the mesh has to be edited to flow better.
It makes a big impact when you tell a beginner "don't do this. It's a rule." Because beginners want to do their best, and show that they understand the rules and can conform. So they carry this false information for a long time, and it lulls them into believing they have understanding when they have the opposite.
The good thing is, if you stay busy working, doing a variety of projects, you'll inevitably come to points where breaking the rules seems like a good idea, and you'll do it, and "hey whaddoya know," everything works great. So, be a bit contrarian, especially if somebody is touting hard rules. It's not a big deal to try new things in digital work -- that's why it's so much fun.