Rederived UHF and L-band model products: B1934

[bennahugo]

As discussed via email the previous attempt to integrate off-axis model introduced astrometrical errors. To this end I have rederived subsets. I have carefully checked that these SP models subtracts out the primary calibrator and some of the brightest flux in the field after accounting for smeared response.

The UHF components are selected from across the field at 600MHz (where the phase errors are worst). The middle plot is truncated to the 75th quartile and includes the primary calibrator collapsed to a single component

The 90 component model quotients with hand-calibrated data performs substantially better in phase than using a single component: vs. a single component model:

The residual amplitude skewness is only corrected once a vastly more detailed model is used

I've done the same for L-band although for B1934 there is essentially no problem at L-band, because the primary is substantially brighter and the beam modulated off-axis AGN population is substantially fainter

I have carefully checked that the flux scale of the primary source is contiguous in the usable part of the bandwidth

[bennahugo]

The same has to be done for B0407, however I have no calibrated products to test against, so this will need to go onto the backburner for now. I suggest that until that can be carefully tested we use a single delta component for that field.

[bennahugo]

Please also note that these models are only valid for the current homogeneous 64 dish layout and across the bandwidth that I indicated. Even the central component starts breaking down at the band rolloff

[bennahugo]

@KimMcAlpine can you please review these changes. I will check and fix 0407 in another PR. It will take a bit more time.

[bennahugo]

Done changed to the usable L-band bandwidth

[bennahugo]

Ok I've also redone B0407. The spread here is slightly bigger because I have done a very basic calibration of the field to check that subtracts. The previous model also had astrometry which was slightly off. The phase spread also reduced with this model:

[bennahugo]

(just for UHF for now. I still need to recalibrate some L-band data to check the L-band model. However it is not as important as getting the UHF system going).