Raw data corrections

A number of things can be done to "correct" or "pre-process" raw MWA data before it is ready for calibration (or other analysis). These tasks are handled by Birli, either as the Birli executable itself, or internally in hyperdrive. cotter used to perform these tasks but it has been superseded by Birli.

Geometric correction (a.k.a. phase tracking)

Many MWA observations do not apply a geometric correction despite having a desired phase centre. This correction applies

\[ e^{-2 \pi i w_f / \lambda} \]

to each visibility; note the dependence on baseline \( w \) and frequency.

Not performing the geometric correction can have a dramatically adverse effect on calibration!

PFB gains

The poly-phase filter bank used by the MWA affects visibilities before they get saved to disk. Over time, a number of "flavours" of these gains have been used:

  • "Jake Jones" (jake; 200 Hz)
  • "cotter 2014" (cotter2014; 10 kHz)
  • "RTS empirical" (empirical; 40 kHz)
  • "Alan Levine" (levine; 40 kHz)

When correcting raw data, the "Jake Jones" gains are used by default. For each flavour, the first item in the parentheses (e.g. cotter2014) indicates what should be supplied to hyperdrive if you want to use those gains instead. There is also a none "flavour" if you want to disable PFB gain correction.

In CHJ's experience, using different flavours have very little effect on calibration quality.

Some more information on the PFB can be found here.

Cable lengths

Each tile is connected by a cable, and that cable might have a different length to others. This correction aims to better align the signals of each tile.

Digital gains

todo!()