wokxy2radec and radec2wokxy don't agree

[blanton144]

The code below SHOULD produce an ra2, dec2 that is the same as the original ra, dec, but it is off by a fraction of an arcsec. This happens for most examples I try.

Note that getting the reverse transform (x,y -> ra/dec) of the radial distortions is often tricky; not sure how this is being done in the code. Usually I implement that as an optimization problem, finding the ra,dec that best predict the given x,y using the forward (ra/dec -> x,y) transform.

ra = np.array([12.30468545] * 2, dtype=np.float64)
dec = np.array([52.58780416] * 2, dtype=np.float64)
racen = 14.289097510524789 
deccen = 51.98826274680977
jd = 2459491.7920949073
pa = 41.42279
 
x, y, warn, ha, para = coordio.utils.radec2wokxy(ra, dec, jd, 'Apogee',
                                               racen, deccen, pa, 'APO', jd)

ra2, dec2, warn2 = coordio.utils.wokxy2radec(x, y, 'Apogee',
                                             racen, deccen,  pa,
                                             'APO', jd)

print(ra, dec)
print(ra2, dec2)
print((ra - ra2) * 3600., (dec - dec2) * 3600.)