Step 2.2: Validation of reference model construction

[TashaleeB]

The purpose of this notebook is to detail Validation Step 2.2.

Below is a picture illustrating the third test of the hera_cal scheme.
https://github.com/HERA-Team/hera_opm/blob/master/pipelines/h1c/idr2/v2/Analysis_Flowchart.pdf

GOAL

The overall goal is to validate the link between the externally calibrated visibilities from CASA and the Omnical calibrated visibilities.

SIMULATOR

RIMEz

SIMULATION COMPONENTS

• GLEAM (point sources)
• eGSM (diffuse sources)
• ATeam (bright sources not found in GLEAM)
• Built with only sources that have known spectrum
• 1024 channels, 100-200 MHz
• LST range of around 21~10 hr
• Noise (simulation has no noise but I will add)

There will be two parts to this test.

1. The first part involves adopting the work done by Josh Dillon (https://nbviewer.jupyter.org/github/HERA-Team/hera-validation/blob/test-2.1.0/test-series/2/test-2.0.0.ipynb). Instead of using data from pyuvsim, I will use data simulated with RIMEz and push it through the Redcal and Omnical pipeline to produce visibilities.

2. Then I will calibrate the RIMEz data with added artifacts using Nick Kern's CASA pipeline (https://github.com/HERA-Team/casa_imaging) and compare these externally calibrated visibilities to the ones made with Redcal.

I propose that this validation step consists of adding realistic gains (using hera_sim) to the simulation (eGSM + GLEAM + noise). I will use the tools available in hera_sim.sigchain to introduce gains (phase offsets, delay, bandpass) to the simulated visibilities. The resulting visibilities will be pushed through Redcal (redcal + omnical).

CASA requires that I provide it with a model to calibrate the visibilities to in the form of either a componentlist or a CASA .image file. Nick has a gleam FITS file that is fed into a model making script that makes an image cube. Because this model is only of GLEAM and not eGSM + GLEAM + noise this could affect things. The other option is to calibrate the simulation with gains added to the same simulation but with no gains added. These are the true visibilities.

Finally, I will compare the calibrated visibilities from Abscal to the calibrated visibilities from CASA.