unhappy mwax zenith pointing

[PyxieLouStar]

There is a night of mwax data in which the zenith pointing is particularly unhappy compared to other pointings from the same night, as seen in the attached power spectra. The following tests are being run:

• cut observations which have ssins flagging due to tv
• additionally cut observations with flagging due to correlator errors
• additionally cut observations with flagging due to narrowband rfi
  
  
  
  

[PyxieLouStar]

The ssins for the observations going into each pointing are here
mwax_1_2_ssins.pdf
mwax_1_zenith_ssins.pdf
mwax_1_4_ssins.pdf
mwax_1_12_ssins.pdf

[PyxieLouStar]

I'm suspicious of the flagging in the zenith observation. The uvfits files going into fhd have flagged the first fine channel of every coarse channel. When I look at a Healpix cube from the 4 pointing, the weights are zero at the flagged fine channel locations, as I would expect. However, the weights of the zenith cube are only zero at the first fine channel of the frequency array.

To hunt this down, I looked at the nf_vis field stored in the obs structure for each observation going into the zenith pointing. I found the flagging I would expect in all but three observations, which had only a flag at the first fine channel. Those three observations are:
1321445344
1321445464
1321445584

I am doing an integration without these three observations to see how that impacts the power spectra.

[PyxieLouStar]

Things look much cheerier when these three obs are not included, as seen. I'm suspicious of the calibration stage, and am going to dig into calibration for the three problem obs to see if I can discover anything here.

[PyxieLouStar]

The issues in the model are especially apparent in the kpar plots. Old and new zenith plots are shown here, and plots of the other pointings are below.

[PyxieLouStar]

[PyxieLouStar]

The three problem obs went through the pipeline with incorrect flagging. Deleting all related fhd outputs and rerunning with the correct flagging resolved the issue.