>Notice with the debug messages that three or four iterations should be
>plenty, but five or six should give you very, very good accuracy.  In the
>sample above, on my machine the accuracy maxes out at 10 significant digits
>after the first part of the fifth iteration.
>
>Like any of the formulas, there are probably scenarios that break this
>formula, causing divide by zero errors or whatever.  But in general, it is
>very, very fast.
>

Well, some combinations may require ten or fifteen iterations: if the
circles come close to touching, but at very oblique angles.

Post a reply to this message