|
In lugnet.cad.dat.parts, Steve Bliss wrote:
|
In lugnet.cad.dat.parts, Ross Crawford wrote:
|
|
If the top points are A1-5 and the bottom points are B1-5, keep
the A2-B1, A3-B2 as currently, but instead of A2-B2 etc, use
A3-B1, A4-B2, etc. I think that would make all the edges inside
curves instead of outside, and eliminate the need for
conditionals. But Im not sure. And I dunno if it would look
right.
|
OK, I tried it and realise now that it doesnt solve the problem at
all, just moves the outside curves to a different spot. I think Don
is right, a more complex torus-like curve is necessary to represent
it properly.
|
Im pretty sure thats going to be the problem with *any* partition
of this surface into polygons. Having a mathematical description of
the surface would be good, because then we could potentially create
a primitive for the surface, and rendering programs could substitute
a smooth-curve construct for the primitive.
In the LDraw code, I believe the appropriate fix will be to replace
the current coarse mesh with a mesh thats fine enough so that the
potential error (ie, the distance from the ideal brick surface to
the LDraw conditional edge) is less than 1 pixel (at reasonable
magnifications). Im not sure how fine such a mesh will be, we
might have to settle for coming within a couple of pixels...
|
Oh, I dont know about that 1 pixel business. The problem with the
surface as its currently written, is that it flip-flops between
convex and concave every at every triangle. But if you ignore the
triangles and spin the surface around a bit, it really looks like
an inner torus surface oriented diagonally. We know how to do that
without making a lumpy mess of it. The challenge is in getting
a diamond shaped slice of torus instead of a rectangular slice.
Enjoy,
Don
|
|
|
In lugnet.cad.dat.parts, Don Heyse wrote:
|
Oh, I dont know about that 1 pixel business. The problem with the
surface as its currently written, is that it flip-flops between
convex and concave every at every triangle. But if you ignore the
triangles and spin the surface around a bit, it really looks like
an inner torus surface oriented diagonally. We know how to do that
without making a lumpy mess of it. The challenge is in getting
a diamond shaped slice of torus instead of a rectangular slice.
|
I think the key difference is the twisted nature of this surface. I think the
twist makes it impossible to partition the surface into polygons without having
a mix of convex and concave seams. Id like to be proven wrong.
I dont have one of the pieces handy, Ill have to look at some tonight, at
home.
Steve
|
|
|