search

EoRImaging / FHD

Fast Holographic Deconvolution
BSD 2-Clause "Simplified" License
20 stars 10 forks source link

2014 data --- mysterious behavior #107

Open nicholebarry opened 6 years ago

nicholebarry commented 6 years ago

I have run some 2014 MWA EoR0 data, and there are some weird quirks. I ran 100 zenith observations, and the power spectrum is much more contaminated throughout the window, even for the model. fhd_nb_2014longrun_autocal_1-100_averemove_bh_kperplambda10-50_1dkpower And weird behavior can be seen in a kcube slice...almost as if there is an aliasing occurring at a constant ky. Below is a slice from the weights. kzslice_weights_005 The only real difference that I've seen between 2014 and 2013 data is that some tiles are not reported at all in 2014. In 2013 data, if a tile was flagged and/or was missing, there were zeros as placeholders in the visibilities. In 2014 data, some tiles are skipped all together. After investigating, it does appear as though FHD is handling this correctly; the params and the obs structure reflect the fact that some tiles are skipped (i.e. tile_a and tile_b record the correct tiles in the visibility array). However, I can't help but wonder if this is the cause. firstpass.o87773.85.txt 1090353176_settings.txt I've labeled this as an FHD problem, but it could be eppsilon too. Any help is greatly appreciated!

miguelfmorales commented 6 years ago

Sorry for the slow reply, this is really weird.

Does the edge in the weights happen for a single obs? I just want to make sure this is isn't an integration issue (somehow adding data from a different pointing, or something that appears in only some observations) and is something that is generic to the 2014 EoR0 data

bhazelton commented 6 years ago

It seems that eppsilon is using a bigger uv plane than it should. This is usually set by values in the obs structure from FHD (i.e. the maximum gridded baseline). We should check how those values are set in the obs structure.