wcsaxes | wcsaxes has been merged into astropy | Data Visualization library

This issue stems from a subtlety in the way matplotlib handles WCS projections.

WCSAxes, which creates the WCS projection for maplotlib plots does not take into account SIP distortions in the WCS (because it uses wcs.wcs_world2pix rather than wcs.all_world2pix).

Often this doesn’t matter that much, however there are two factors that can make it matter a lot when displaying TESScut cutouts. The first is that the TPFs that TESScut produces come with the original WCS information from the full-frame image adjusted for the position of the cutout. This means that unlike the WCS information in TESS pipeline TPFs, the WCS information in TESScut TPFS includes a full complement of SIP distortions, which are being ignored by matplotlib. The second factor in how accurate the displayed coordinates will be is where on the TESS CCD the source falls. Because the TESS field of view is large, at the edges of a full-frame image the difference made by ignoring the SIP distortions can be on the order of 15 pixels.

This particular source is very much on the edge of that FFI so the SIP distortions will make quite a large difference:

There are two options for handling this problem:

1. Don't use the matplotlib projection option directly, and instead move everything into pixel space, manually calling wcs.all_world2pix or wcs.allpix2world as required.

2. Create a new WCS for the cutout that does not include SIP distortions (which will be fine because the cutout is small enough to not really need them). Here is a Jupyter notebook that describes one way to make this new WCS for the cutout: https://github.com/ceb8/tessworkshop_wcs_hack/blob/master/tesscut_wcs_hack.ipynb.

Following the error message, if the PCi_j format is used to describe the rotation matrix, the values should be normalised to 1, and any scaling factor encoded separately in the CDELT values. If one wants to include the scaling factor in the rotation matrix, the CDi_j format should be used, without any CDELT values. In the draft v4.0 FITS specification, the relevant section in the FITS standard is on p.29 around eqs. 9-11: https://fits.gsfc.nasa.gov/standard40/fits_standard40draft1.pdf

Two possible solutions exist. First, the scaling can be transferred to the CDELT values like so:

Oh, I see. I was overcomplicating it. Here's all I needed to do: