RFI editing

The data suffered from scintillations and intermittent 'radio frequency interference'(RFI). In addition to normal editing of the data, the scintillations affected data and channels affected due to RFI were identified using the tasks POSSM & TVFLG and the affected data was edited.

After calibration we focused on RFI removal. One way for looking for RFI affected data is running the task POSSM and look for each baseline (after bandpass calibration). But since it would be very time consuming to look at each baseline and edit out the RFI affected data. Another way is to use the AIPS task called FLGIT, which subtracts continuum from channels in $uv$-plane to determine entries in a flag table from scatter of residuals.

In the task FLGIT, one can set overall clipping for the data as well as remove data which fall too far off a linear fit. In practice, in removes most of the RFI which is strong enough to noticeably degrade the final image. We experimented with FLGIT parameters to find an acceptable balance between removing all the clearly bad data and keeping most of the unaffected data. Since the brightest source observed at 235 MHz is 3C147 and 3C147 has flux density 59.41 Jy, we clipped every baseline visibility above 100 Jy (flux density calibration and bandpass calibration need to done before setting these parameters.). As these clip levels are well above the total flux in each field, we are certain to have only removed corrupted data. This removed roughly around 20% to 30% data for each observed run.

For 235 MHz data analysis, the methodology was largely identical to previous section, except for some adverbs in the tasks. After the task BPASS, for the 235 MHz data, following procedure was followed:

Delete SN (ver 1) and CL (ver 2) tables, and run CALIB using: DOBAND=1; BCHAN=10; ECHAN=54; UVRANGE=0.5 50; SOLINT=1. DOBAND=1 was used to correct for the bandpass before averaging over the used bandwidth and improve signal-to-noise ($s/n$) ratio before setting the gain solutions. UVRANGE from 0 to 0.5 allowed us to neglect and avoid the short spacing RFI affected data (i.e., largely, central square antennas) which can affect the subsequent calibration. This will produce SN (ver 1) table.

The task CLCAL was run using; SAMPTYPE = 'MWF', BPARM = 1 1; DOBLANK = $-$1; DOBTWEEN = 1. This will generate CL (ver 2) table containing the median window filter solutions for the antenna gains, derived from the flux density (or primary) calibrator scan. BPARM 1 1, was used for the smoothing function for amplitude and phase for an interval time of 1 hour. DOBLANK = -1 was used for replace previously good values with more smoothed values. Or in other words, leave previously blank values blanked and DOBTWEEN=1 for smoothing all SN (ver 1) values regardless of the source.

Data affected by intermittent RFI needs to be flagged before the data is averaged in time. RFI on all the scans was identified using the task FLGIT. This task examines the data after subtracting a linear fit to the band shapes from individual baselines. A user specified set of channels is used to determine the linear fit. All data with residuals outside the user specified limits are then flagged. All channels were flagged for a given integration time containing bad data. This task was run using; BCHAN=10; ECHAN=54; DOCALIB=1; SOLINT=1;DOBAND=1; APARM= 100, 10, 10, 5, 5, 1; ICHAN=10, 20, 1, 1, 21, 30, 1, 1, 31, 40, 1, 1, 41, 54, 1, 1; ORDER=1.

The parameters for the APARM are APARM(1) maximum amplitude allowed in the spectrum before fitting. APARM(2) maximum residual flux allowed for channels within the baseline fitting regions for unity weight. APARM(3) maximum residual allowed in the non-baseline fitting regions of the spectrum. APARM(4) flag all the channels having residual flux $>$ aparm(4) $\times$ RMS. APARM(5) flag all the channels having real and imaginary parts $>$ APARM(5) $\times$ RMS. APARM(6) Maximum amplitude allowed in V polarization; any channel exceeding this is flagged in advance of the baseline fitting or medial filtering. The task INDXR was once again run on the newly generated UV-data, which was generated by the FLGIT task.

Channel averaging Channel averaging is done to reduce data volume and to increase signal-to-noise (snr) ratio. Also while doing averaging, if the RF band is split into frequency channels which smaller channel width the effect of bandwidth smearing can be reduced. The task SPLAT was used to apply the time and bandpass calibration (the CL and BP tables respectively) to the data and a calibrated database generated. The task SPLAT is also used to average the channel and reduce the data volume. To avoid the bandwidth smearing the final channel width $\sim$1 MHz was chosen.

For the 235 MHz data task SPLAT was performed in the following manner: DOCALIB=1; DOBAND=1; APARM(1)=3; BCHAN=1; ECHAN=45; ICHAN=0 ; CHANNEL=9; SOLINT=0. Aparm(1)=3 is for average each subsequent N frequency channels to diminish output data; where, N is the value used in the parameter CHANNEL, which is used in this case, i.e., CHANNEL=9, corresponding to a channel width of 1.125 MHz.

This produces a new visibility data averaged to form a 5 channel file in case of 235 MHz. As usual, the task INDXR was run after this task to obtain NX, etc. tables.

After averaging the data, the bad data was once again examined using tasks UVPLT, UVFND and such data was edited out, using the tasks UVFLG or CLIPM. Then, once again, a new iteration was done (i) of deleting the tables SN (ver 1) and CL (ver 2), and (ii) subsequently running tasks CALIB and CLCAL.