Quote:
Tesselating a tetraheron and pushing the vertices out to unit length won't give me regular triangles. I tesselate by dividing all edges and creating 4 faces.

i don't quite get what you want to do. What you mean by "regular triangles"? Equilateral triangles?

Quote:
Since the distances from the origin to the original vertices and the distances to the centers of the edges are all the same i think my sphere should end up with uniform sized faces.

Where origin is? in center of tetrahedron?
What mean "sized"? Have equal area? "the distances to the centers of the edges" from what?
Do you set your vetrices after tesselation to be unit-length?

If you'll tesselate tetrahedron and then set all vertice vectors to be unit length, i'm almost sure you'll get a polyhedra that will have 4 bigger equilateral triangles and 12 smaller non-equilateral triangles, 16 triangles total.


What rules you want it to obey?


anyway, if you want to make polyhedron with vertices placed at equal distances from center, and composed of equal triangles, you can try
1:icosahedron, it have 20 equal faces(more than your 16-face polyhedra), and all edges is equal too, it'e regular polyhedron.

2:dodecahedron with additional vertice at center of each face, so each face is splitten into 5 triangles. Center vertices must be renormalized. So you'll have 60 equal triangles.(it will be slightly non-equilateral triangles)
edit: if by above rules you mean that thing about face edges, that 60-face thing made from dodecahedron probably will not work.

If you start with a regular tetrahedron, each of the four faces is an equilateral triangle like this:

____________ .
\ / .
\ / .
\ / .
\ / .
\ / .
\/ .
[/code

If I understand correctly, you want to subdivide the four faces each into four smaller equilateral triangles, like this:

____________ .
\ /\ / .
\ / \ / .
\/____\/ .
\ / .
\ / .
\/ .
[/code

And the problem is that when you blow out the new vertices that bisect each of the edges of the original triangle, your new sub-triangles aren't all the same size.

This is inevitable. The middle face will always be bigger because you're moving apart all three of its vertices, but the corner triangles only are only having two of their vertices moved apart. (By moving apart, I mean moving out to lie on the surface of the sphere on which the original four vertices lie, which also lengthens the distance between them in the direction of the line connecting them).

If you want completely regular, equal angle and equal length sided faces, you need to use a Platonic solid (such as the regular tetrahedron or icosahedron). But you can't keep adding more faces by subdividing the existing ones if you want to keep all faces equal shape and size.

... previous was me.


If you start with a regular tetrahedron, each of the four faces is an equilateral triangle like this:

____________   .\          /   . \        /    .  \      /     .   \    /      .    \  /       .     \/        .

If I understand correctly, you want to subdivide the four faces each into four smaller equilateral triangles, like this:

____________   .\    /\    /   . \  /  \  /    .  \/____\/     .   \    /      .    \  /       .     \/        .

And the problem is that when you blow out the new vertices that bisect each of the edges of the original triangle, your new sub-triangles aren't all the same size.

This is inevitable. The middle face will always be bigger because you're moving apart all three of its vertices, but the corner triangles only are only having two of their vertices moved apart. (By moving apart, I mean moving out to lie on the surface of the sphere on which the original four vertices lie, which also lengthens the distance between them in the direction of the line connecting them).

If you want completely regular, equal angle and equal length sided faces, you need to use a Platonic solid (such as the regular tetrahedron or icosahedron). But you can't keep adding more faces by subdividing the existing ones if you want to keep all faces equal shape and size.

i think your neighbor rule is only conserved on primitives with an integer multiple of 3 triangles around each vertex: this only goes with the 4-sided primitive.

no subdivision method will create equal triangles, however, you can correct a lot of the distortion either by making the edges springs, and the verts masses and simulate this mass spring system, or some algebraic correction. as a rule of thumb though: the more tris you start with, the more uniform the result will be.

also, am i wrong in assuming your edge rule isnt really needed, but merely a convenience? not saying conveniences are a bad thing, certainly since your code can be quite a bit shorter and more elegant with this contraint, but a 4-sided primitive really leaves to be desired...

well,and how i woulda do it:
Use a icosohedron as base shape.Make a sphere out of it by non-recursively tesselating faces(well,you can recursively tesselate but it's needed to be possible to do that w/o recursion), and THEN renormalizing vertices of faces. Let each face have associated surface coordinate system with axises along edges of triangle.Let's call it, say, u,v coordinates . So coordinates of vertices is u=0,v=0;u=1,v=0;v=1,u=0 .

So,for example, finding vertices of subdivised face it's simply
for(i=0;i<N;i++)
for(j=0;j<=i;j++){u=i;v=j; WorldCoordFrom_uvCoords=Corner+u*Edge1+v*Edge2;}
(draw it on paper)

Algo:
Find face of icosohedron that is hit by beam(say,you can do it by finding largest positive dot product of icosohedron face normal and beam. It's enough to do 10 normals(symmetry) and find smallest negative dot product as well and then select one that have larger absolute value.
You can avoid checking unnecessary normals using sign of x, y , and z components of the vectors, or using values of dot products with some normal to find if it needed to check other normals.
)

Then note that intersection of beam with sphere surface corresponds to intersection of beam with face prior to renormalizing. So find "u,v" coordinates of intersection of beam with icosohedron face. Then find subdivised face it have intersected from "u,v" coordinates, analitically, no recursion needed, no rules needed, no strange simplifications needed....