"Shay" <sah### [at] simcopartscom> wrote in message
news:41b4df6d$1@news.povray.org
> Slime wrote:
>>> Had to try this again. Probably not interesting unless you've
>>> experimented with this type of thing yourself, but my new
>>> edge-rounding algorithm is very sophisticated.
>>
>>
>> Maybe I missed this before, but what is this algorithm? Is it an
>> SDL thing or some sort of patch?
>>
>
> Just as SDL thing. I posted a similar shape a while ago. The edges
> looked *decent*, but only because I made the curves very small. The
> whole thing got me thinking about the odd parts of rounding a shape
> like this. There are no obvious solutions. Compromises have to be
> made, and I did some experiments to find what arrangement of
> compromises made the most attractive shape.
>
> The corner bends are all perfectly cylindrical, like a CSG rounded
> cube, but the corners won't meet in a sphere like a cube's corners.
> The bends are all different radii, but the width of the bends is
> perfectly equal between edges. This is the best solution IMO (equal
> radii curves of different widths being another possibility), but it
> is possibly flawed. I've failed to think of a better way to join a
> concave curved edge with two convex curved edges of a smaller
> radius. What I've got here might be as good as that situation can
> possibly look, and, looking at some of the machined parts around
> here, looks like the same compromise made by some manufacturers and
> welders.
> There *are* differences, however subtle, between this and what would
> come out of a modeler.

Surely it depends on the modeller?  In something like ProEngineer you have 
lots of complex options to specify exactly what it does in the situations 
you mention, far more than I understand or have needed to use so far :-)

Post a reply to this message