w*(<pl1 -
149
!Xh Xi| e/cos 91
(3-1.8)
If we compute the geodesic with (3.1.3) then we get a point P\
with the coordinates cph, Xi, see figure 5.
The ellipsoidal triangle is also replaced by a spherical one and
dA can be computed in this spherical triangle with the formula
(3-I-9).
Figure 5.
tan dA
sin Ah
sin a
tan Acp
in which:
cos A1! cos a
a s'/a
(3.1.9)
(3-I.IO)
With this approximation of d.4 we compute a new approximation
A2o of Ao with A2o A1o dA. With this azimuth A2o the geodesic
can be recomputed. We repeat this process until the difference
between Pi and Ph is less than s, see (3.1.11).
|<pi 9 l|
(3-i.ii)
For this computation an algol procedure has been written (see
appendix).
3.2 Some technical aspects of the computation.
The method described in 3.1 is an iterative process, and the
question arises in how far this process is convergent. An exact
analytical proof of the convergence has not been found by the
author, but several test computations have indicated a very good
convergence. In most cases the required accuracy is reached in
2 to 4 iterations.