Changeset 8241 for trunk/Ohana/src/gastro2/src/gastro2.c

Timestamp:

Aug 8, 2006, 3:50:18 PM (20 years ago)

Author:

eugene

Message:

substantial work to correctly handle higher order polynomial.
conversions between the internal fitting model (R = sum (Aij x^{i y}j))
and the header representation were incorrect: they did not correctly
handle the shift to the CRPIXi value (ie, x -> (x - xo)). i fixed this
by making the fit explicitly work on the fit to the tangent-plane
coordinates, instead of fitting corrections, and calculating the correct
change of polynomial coefficients. this is not as simple as it seems at
first glance: the header version of the coefficients does not have a
zero-order term, and instead has the CRPIX value. to convert from one to
the other requires solving the equation L(Xo,Yo),M(Xo,Yo) = (0,0). I do
this using the Newton-Raphson method.

File:

• trunk/Ohana/src/gastro2/src/gastro2.c (modified) (1 diff)

trunk/Ohana/src/gastro2/src/gastro2.c

@@ -30 +30 @@
   }
 
-  for (i = 0; i < 3; i++) {
+  for (i = 0; i < 5; i++) {
     gfit (&Target, &Ref, MIN (MAX (1, NPOLYTERMS), 3));
     fprintf (stderr, "precision: %f\n", Target.answer.dR / sqrt(Target.answer.N));