An Introduction to GMRT Continuum Data Analysis in AIPS

Introduction

The purpose of this document is to introduce the process of data analysis for observation done with GMRT, with Astronomical Image Processing System (AIPS) for beginners. Some knowledge of AIPS is assumed, for example, how to load the data, the syntax of running AIPS tasks, etc. There are multitudinal ways of accomplishing any particular objective by making use of the various `tasks' available in AIPS, this document is meant to reduce the steepness of the learning curve of the beginner, and once the basic processes are understood and internalised one can devise one's own stratagem and algorithms to swim through the data according to one's whims and fancies following one's thread of natural thought processes!


The GMRT data format

The GMRT observations are done in the following sequence: flux calibrator, phase calibrator, source, phase calibrator, source,.......phase calibrator, and (optional) flux calibrator at the end. The data is obtained in the GMRT raw format of Long Time Accumulator or lta format. Till the end of 2010, the GMRT hardware backend (GHB) was the main backend and would normally produce two separate files for the Upper Side Band (USB) correlator and Lower Side Band (LSB) correlator, with the extension to the filename as .lta and .ltb respectively. Each file has 128 spectral channels which is required to address the bandwidth smearing issues at low radio frequencies and is also useful to edit the narrow band Radio Frequency Interference (RFI). The data can be converted to FITS format in most machines. Go to the directory which has enough disk space (usually /DATA/ will have 10 to 20 GB free space). At GMRT Khodad, on any astro machine, initialise the path (eg. for C shell): source /astro/RC.csh (OR RC.sh for bash).

Since the beginning of 2011, the GHB is shutdown and the new GMRT Software Backend (GSB) was made the default backend. This produces single lta file and can have 256 or 512 channels, as desired by the user.

Here is how to convert lta to FITS format:-

astro101> listscan
Usage: listscan filename

The file name is the lta or ltb file from your observation. This will create filename.log. Edit this log file to suit your requirements, like output file name, fully dead antennas, link to ONLINE LOGFLAGS file, etc.

astro102> gvfits
Usage: gvfits input_param_file

The input is the above log file obtained from listscan. After the FITS file is obtained, we are ready to get into AIPS for analysis. Backup this FITS file for future use.

The basic logical steps

The basic logical steps amount to the following:-

Before getting the data to AIPS,  set the environment for the data, in order to let AIPS know location in the directory tree. This depends on the shell you are working (eg. for csh):- setenv DIR=<full path to the directory where the FITS file is kept>. It is recommended that one or two such environment for a directory should be permanently defined in your .cshrc or .bashrc

Load the data file into AIPS using the task FITLD

Run the task INDXR for creating the index and initial CL table (use cparm 0 0 1 0 for CL table increment of 1m).

Run LISTR to list the sources observed. Save the listr output in a txt file for future use.
Run the task QUACK to remove the first 0.25 minute of data from each scan (first visibility from each scan).

Set the flux of the flux calibrator using the AIPS task SETJY.

The main job is to look out for bad data and flag them. Inorder to flag bad data, choose a representative channel, somewhere in the central band. For example, in the 256 channel data, choose a channel between 50 to 200.

There are many ways to identify the bad data, mainly TVFLG and VPLOT. Since there are large number of baselines in GMRT, VPLOT is not suited best for this purpose. Instead, TVFLG can be used to view all the data in one screen. Use phase difference primarily to identify bad antennas and baselines on flux and phase calibrator.

Flag the antennae, baselines which are deemed bad, first for the flux calibrator, followed by the phase calibrator. If an antenna is dead for flux and phase calibrator, it should be flagged from the entire data. This step is a detailed one and needs some amount of 'feel' to accomplish it. When the data is free from RFI, it is reasonable to assume that an antenna dead in one channel is dead for all channels. However, if the data has some RFI, you have to careful in choosing the 'representative channel'. Data should NOT be flagged using TVFLG. The AIPS task for flagging is UVFLG for the bad data  identified using the AIPS task TVFLG or VPLOT. It is also possible to flag using TVFLG, however it requires online change of some of the parameters like polarization, channel, source details etc, which may lead to error. Hence it is recommended to use TVFLG only to identify the bad data and put these in a FLAG file which is given as input to UVFLG.

Calibrate the flux calibrator using the AIPS task CALIB. This task gives the conversion factor between the observed raw flux of the flux calibrator and its absolute flux. The task should be run for a single channel.

Rerun this for phase calibrator by appropriately taking care of UV limits, if any.

Check the solutions using AIPS task SNPLT. If not satisfactory, continue flagging.

Get the flux of phase calibrator using the AIPS task GETJY. 

Apply the calibration to the flux cal and phase cal using the AIPS task CLCAL.

Run AIPS task UVPLT for flux and phase cal separately to look for any amplitude and phase deviations. If any, run AIPS task TVFLG with docal=1 for these calibrators to identify the bad data. Flag them; delete older SN tables and CL table 2; and run CALIB, GETJY and CLCAL sequence.

Run AIPS task CLCAL for the sources with respect to corresponding phase calibrator. 

Additional flagging, especially in the time and frequency domain, may be needed in case of RFIs.

Get the bandpass averaged table for the sources using the AIPS task BPASS. Check the bandpass using the AIPS task POSSM. Flag any bad channel found (channel 116 is bad in most observations).

Once you are satisfied with the flagging, the data is ready for channel collapsing. While collapsing, the effects of bandwidth smearing should be considered. At 1 GHz and above, it is safe to collapse to 4 or 5 MHz, and in each side band, you may have to create three channels each of 4 or 5 MHz bandwidth. However, at lower frequency such averaging will introduce bandwidth smearing. To improve the noise more channels should be collapsed, while to reduce bandwidth smearing, lesser number of channels should be averaged. Taking a balance between these two, at 610 MHz, the data should be averaged in steps of 2 MHz and at 325 MHz and 235 MHz, it should be in steps of 1 MHz or less. The AIPS task used for this purpose is SPLAT. 

At this stage, check the data obtained after SPLAT using UVPLT. If any bad data is found, it should be flagged from the original UVDATA and the should be channel collapsed again. 

Run the AIPS task IMAGR which makes the map. At 1 GHz and above 3-D imaging should be used depending on the strength of the sources at the edge of the primary beam. However at 610 MHz and below the effect of non-co-planarity is significant and 3-D imaging should be mandatory. Consult AIPS expert on this.

The tasks used in AIPS The main list of AIPS task that need to be run (in more or less the chronological order) is given as follows:-

FITLD: Loads the data file into AIPS.

INDXR: Indexes the tables in the data files loaded in AIPS. Recommended parameters are cparm 0 0 1 0 for CL table increment interval time of 1m. This creates extension tables CL version 1 and NX.

LISTR: Lists the sources observed. Save the listr output in a txt file for future use.

SETJY: Sets the absolute flux level of the flux calibrator.

QUACK: Flag the first visibility of each scan using this task. No need to flag half a min, only the first visibility will do.

TVFLG: Presents grey-scale plots of amplitude and phase for all baselines as a function of time on TV screen.  Check the data for flux and phase calibrator. Most commonly used parameter is PHASE DIFFERENCE and AMPLITUDE VECTOR DIFFERENCE.  Load only one "good" channel and assume that the data which is bad in this channel is also bad in ALL channels. Avoid flagging directly from TVFLG. Create the extrenal table to be called in UVFLG and flag the data using UVFLG.

UVFLG: Used for flagging the antennae and base lines identified from the output of TVFLG.

CALIB: Calibrate the observed flux with the theoretical flux.

SNPLT: Plots the SN table produced by CALIB.

EXTDES: Destroys the extension tables (to avoid confusion we need to destroy improper tables))

GETJY: Obtains the flux level of the phase calibrator. using the SN tables obtained from CALIB.

CLCAL: Actually applies the calibration.

UVPLT: Plots the UV data in various ways. The next step is not taken until the results of UVPLT is clean and satisfactory.

BPASS: Creates the band pass table.

POSSM: Displays the results of the band pass table. Used to identify the bad channels.

SPLAT: Applies the bandpass and creates the band pass averaged `FITS' file for the whole observation.

SPLIT: Splits the original data into individual sources

IMAGR: Creates the map.