Hi, I have been playing with querying the images associated with various diaSources and I have noticed some offsets in the pixel coordinates for a given sky coordinate depending on which WCS I use.
The ExposureF images have a wcs attribute which I understand to be the most accurate representation of the image WCS. I have found examples of how this WCS can be approximated in FITS file format (e.g. by converting the ExposureF WCS or when saving a cutout to a FITS file). It is stated that the FITS approximation is of course not exact but that it should be accurate to within 0.05 pixels.
I might be doing something wrong but I’ve tested a hnadful of visits and I’m finding positional offset of ~5 pixels when I try the two different WCS methods, which is much bigger than I was expecting. Any comments or advice? Here is a demo notebook: https://github.com/jrob93/lsst_nbs/blob/main/dp1/diaSource_WCS_analysis.ipynb