Right now we use a single exposure dependent CRPIX/CRVAL pair for the astrometry of all sources, even though we use source and exposure dependent scale matrices. In the presence of distortions this will cause errors in pixel positions. As long as the distortions are small across a patch this can be mitigated by using a locally calculated CRPIX/CRVAl pair for each exposure; but in this case we could also use source independent scale matrices.
We should either use source and exposure dependent CRPIX/CRVAL pairs, or make the scale matrices source independent.
If we need extra room in memory for this info, we could probably save space by using a single scale matrix, since the scale matrices are simply related by a cos(dec) multiplication
bd-j
commented
3 years ago
Right now we use a single exposure dependent CRPIX/CRVAL pair for the astrometry of all sources, even though we use source and exposure dependent scale matrices. In the presence of distortions this will cause errors in pixel positions. As long as the distortions are small across a patch this can be mitigated by using a locally calculated CRPIX/CRVAl pair for each exposure; but in this case we could also use source independent scale matrices.
We should either use source and exposure dependent CRPIX/CRVAL pairs, or make the scale matrices source independent.
If we need extra room in memory for this info, we could probably save space by using a single scale matrix, since the scale matrices are simply related by a cos(dec) multiplication