Beam normalization is wrong

[bennahugo]

@kmbasad @SpheMakh @o-smirnov

Ok there is something wrong with the way the final products have been normalized:

It appears to me that the boresight is normalized to [[1 0], [0 1]] while the off axis is left unnormalized, because the crosshand phase shows extreme ramps even just barely off boresight (<1'). My idea was to use unnormalized beams which contain the residual electronic OMT and digitizer ripples and impedances at boresight, and absorb them into diagonal and offdiagonal CASA solutions while calibrating. However this information is now effectively destroyed. Here are the two side by side (yaxis degrees, range +/- 90 deg):

Looking at the code (as far as I can tell) it seems the normalization routines (normalise_multifreq) are no longer called and the errors are baked into the zernike coefficients. I need to track down where this is going wrong - in the way you process them for fitting or the way Mathieu normalizes. Where can I find the original measured fits files from which these coefficients were derived?

[bennahugo]

This is why circular polarization is not corrected when applying the beams in DDFacet as it stands.

[SpheMakh]

As I said earlier, both sets of beams go through the same treatment in eidos. So the problem probably lies in the input holography beams. In the meantime, we should all use the EM simulations

[kmbasad]

Sorry for such long delay, but I will start maintaining EIDOS again from this month. Do you still have this problem?