
on 28.08.2009 at 19:17 Michael Fawcett wrote :
Well, I'm certainly interested. Can you prove yours is more accurate? well, will you believe my word?
i don't know about other implementations but the implementation (vincenty) that i saw truncates the series to the order 6 say (don't remeber exactly) while i save some extra terms (say of order 8) i also compared this solution to a higher order solution (which provides ~12 significant digits but which is overweighted with coeffs) if you are interested i can post a paper of formulas' derivation also there is another approach (to direct problem at least) whole thing is to evaluate an elliptic integral while i do the series expansion (to build inverse routine later) one can directly evaluate the integral with one of special function evaluation techniques (which are widely available, e.g. in numeric recipies) i didn't try this (no time) and can not say if it is more accurate or faster (however it might be) -- Pavel