It’s been a year since first asking why triangles are “bad” and all quad encouraged. Even for sub-d it doesn’t seem any more obvious. Keep seeing people use triangle step-ups, but reading that all quad topology is better.
Looking at what happens under subdivision doesn’t make it clearer either. Example:
The triangle step-up seems much more crisp and there is one more vertex to control the shape. “tris cause pinching!” hasn’t really been helpful as an answer. What if you want it to pinch to create a crease or sharp edge?
Still looking for the answer to these questions:
Is there some other reason to fervently avoid tris? morphs? sculpting? something to do with the normals or baking?
How does one replicate the properties of a triangle in all quad modeling?
You’re forgetting that triangles stop edge flow. A triangle would be a stooping point for an additional edge loop that would transport that triangle into a quad. When subdividing your mesh, triangles that you have would convert into 3 quads.
In my opinion triangles are not that bad especially if they’re on a flat surface or you’re just using them when modeling a low poly mesh that you will later subdivide and correct it’s mistake. But trust me on this one, you must avoid them in most cases, they do cause pinching because they make a 3 sided pole that cause pinching. I know, I know, you’ve heard this a bunch of times. But it is as how it is.
Hell, we could talk about this all day. Here’s a clearer example why they’re not good: .blend
As for your second question I really don’t know what you mean by saying “replicate properties of a triangle in all quad modeling”? If you want pinching, you can use edge crease instead of a triangle.
The 3pole on a tri is like any other 3pole when it comes to deformation and sculpting? That would mean it’d make creating good morphs harder by extension? The all quad with 5poles takes many more faces, but offers comparable or greater control and doesn’t create new poles along the “inside”?
Watching people model in reality they use that M-shaped triangle configuration. even for sub-d. Generally in clothes, armour or hair; sometimes in hands, feet, or parts of the armpit or inner thigh. In clothing and armour it seems to be for wrinkles, creases and keeping very sharp edges. How does one replicate this in all quad modeling?
Avoiding triangles in subsurf modeling has more to do with not making your model out of triangles and using good topology than “if you use triangles you’re doing it wrong”
At run time your mesh will be converted to tris when you render. BUT!!!
There are things that happen before render time. Tris have some issues when it comes to applying a subsurf to them and when it comes to deformation artifacts ((If used poorly)) Mostly because it makes poles that are placed in less then strategic locations. When you see poles being used it normally is in a place where you have 3 surface planes meeting. And this is where you need to be mindful of your edge and loop flow!
When you need to do something like the above, Consider if you can just turn your mesh around and make a loop. With a loop you have the ability to add significantly more detail where it is needed most and it tends to give a geometry that is more friendly to use where the tris would of been ((see above on pole placement))
Look here in the last one. I’m not saying that ether of them produce a bad end or a good end. Just know what you are using it for and be aware of the results it will give you.
Could you elaborate?
All Nurbs surfaces need to get tesselated to be visible in openGL (even if it’s realtime meshing can’t get edited in all apps). Building the simple recessed shape by the OP with Nurbs would be done by combining several surfaces. Each of them gets represented by mesh patches. These again get sewn together with (lots of) triangles at their borders, in Blender, in Rhino, in Solidworks, everywhere.
