FITS headers conventions

[IanHeywood]

The reference frequency in the FITS headers produced by pfb seem to correspond to the lowest frequency in the data, e.g. for this UHF map:

CRVAL3  =        592376953.125

It's more standard for MFS images to have the central frequency, but if there's a good reason to have the lowest frequency instead it can be worked around.

[landmanbester]

Thanks for pointing this out. I'm not actually sure if my headers strictly adhere to the fits format (since the truth is actually encoded in the dds and fits is mainly for viewing). I set the headers up here

pfb-imaging/pfb/utils/fits.py

Line 33 in 7e3077c

def set_wcs(cell_x, cell_y, nx, ny, radec, freq,

if you spot anything else that needs to change please report it on this issue. I've renamed it to make it a bit more general

[landmanbester]

Just to note that there was also an issue with crpix2 viz. it was set to ny//2 instead of 1 + ny//2. I wouldn't call this a solved problem but I managed to get the WCS to agree with wsclean by flipping the v coordinate during gridding/degridding and not flipping any axes when writing fits files (and setting crpix2 = 1 + ny//2). This means that images will be upside down if you rerun the current master on tmp datasets using the old convention (the component models should still be fine, I hope)