Astronomical Image Processing System, developed by the National Radio Astronomy Observatory(NRAO), USA, is one of the various software packages available for reduction of radio interferometric data like that of the GMRT. AIPS is the standard data reduction package used at NCRA. Here you'll find a brief overview of AIPS setup at GMRT and NCRA, and also some help on getting started with it along with remedies to some of the commonly faced problems and recent issues specific to the GMRT data format. Its curve-of-growth to a large extent depends on your comments which may be e-mailed to firstname.lastname@example.org. Before getting into AIPS please make sure that you have used the latest version of listscan and gvfits to generate the GMRT data FITS files. The latest versions of these and various other utilities are available on GMRT and NCRA machines at /astro/bin/Linux and can be sourced using 'source /astro/RC' or 'source /astro/RC.csh'. If you have used gvfits version 1.11 or earlier to convert the GMRT data set to FITS then use 31DEC04 AIPS or earlier, have a look at the section FAQ and BUGS. This is critical for the datasets with negative frequency increments. Also, if you have been using SPLIT and SPLAT with two-sideband GMRT data, please pause to read the relevant item in the FAQ and BUGS section below.
AIPS setup at GMRT and NCRA
For the last couple of years, we have been following the practice of installing the latest frozen version AIPS on the general users' and visitors' machines at GMRT and NCRA, and also making available the binaries of the same for our local users for upgradation. We have now moved to the latest 31DEC10, which is updtaed everytime a patch mail arrives from Eric Greisen. Local users are encouraged to download the latest 31DEC10.tar.gz and install.pl from the NRAO-AIPS website for installation on their own desktops. All machines now in GMRT run the 31DEC10 version. The precompiled binaries of this version can be found here, for which README is available. For help on locally compiling the source code please have a look at the compiler notes available at NRAO AIPS Home Page. If you face difficulty in compiling AIPS on your computer, it is very likely because of incompatible compiler or missing libraries. Mail email@example.com for help if required.
Before starting AIPS it is required to define AIPS' data areas and also the environment variable pointing to the data area where user wants to read/write into/from AIPS. These can be done as following:
Defining AIPS' data areas
This can be done by running this shell script
and choosing the data areas interactively. This will create a file .dadevs in the home area of user say 'visitor' . Or else one may simply create the .dadevs contaning lines similar to
This will create three AIPS' disks ASTRO5_1, ASTRO5_2 and ASTRO5_3 in the area /analysis/visitor/ on the machine astro5.
However, if you have predefined AIPS data areas during compilation, the above step can be bypassed. For example, if you have installed AIPS in /home/aips/ the data areas are defined in the file /home/aips/DA00/DADEVS.LIST
Now start AIPS as following:
If the data areas have been created using the script (which is true for all the gtac and visitor accounts at GMRT) then one does not need to source LOGIN.SH or LOGIN.CSH as appropriate line has been added to .bashrc or .cshrc, respectively.
Getting started with AIPS
The description of various AIPS tasks and detailed instructions can be found in the AIPS COOKBOOK. Users unfamiliar with the AIPS environment may find it useful to first go through the easy guide to GMRT data analysis.
GVFITS and LISTSCAN
GVFITS and LISTSCAN are the utilities used to convert lta and ltb data files produced by GMRT into FITS format. Development of these, apart from fixing minor bugs, has largely been driven by efforts to incorporate new features/observing modes at GMRT and also to keep the FITS files compatible with any changes in AIPS. It is therefore important to use the latest versions of these. Here is some information on the recent gvfits' releases. There is no version called 1.15.
gvfits 1.17 : Merges the GSB and old correlator versions into one. This version of gvfits can therefore convert lta/ltb files from both GSB and the hardware correlator. The corresponding listscan is available too at astro0:/astro/Linux/bin at GMRT. The binaries of listscan and gvfits are available here.
gvfits 1.16 : Updated for leap-second correction on 31 Dec 2008. Internal header is now consistent with changes made in DAS-2.05 to improve fringe tracking.
gvfits 1.14 : Corrected bugs in the determination of the start date for the FLAG table. Also added option for honouring flags embeded in the lta data.
gvfits 1.13 : Fixes a bug in which the uvw co-ordinates were not computed at the centre of the integration. Also added some more information in the history file.
gvfits 1.12 : Fixes a bug which caused 31DECO5 version of AIPS to get confused when the frequency decreased with channel no.
gvfits 1.11 : Allows user to label the polarisations as XX/YY or RR/LL (By default they are labelled RR/LL). Also generates the correct FQ table when more than one IF is present.
gvfits 1.10 : Modifications to support the polar and other new modes of the GMRT correlator.
gvfits and listscan are special utilities that are specific to the LTA data format of the GMRT. It is recommended that GTAC users external to NCRA convert their LTA/LTB files into FITS files before leaving the observatory or NCRA, using these utilities.
If you are logged in to one of the GTAC machines at GMRT, follow these steps in sequence
After logging into one of gtac1/2/3/4 or astro1/2/3/4, at the prompt run source /astro/RC or source /astro/RC.csh depending on whether you are working with bash or tcsh shell respectively.
Type listscan and gvfits to see their usage; by default these are the latest versions.
Run listscan on your lta/ltb files to generate log and plan files.
Edit the log files appropriately.
Run gvfits on the edited log files to obtain FITS files.
If you want combined sideband data, then combine the USB and LSB data using the utility ltamerge before running listscan and gvfits. However, please be aware of the recent bugs in the AIPS tasks SPLIT and SPLAT.
Refer to FAQ and BUGS for more light.
However, finally if you want to install listscan and gvfits locally on your laptop, you could get them from herealong with the shell scripts required to set the environment variables. You may have to change the paths defined in the shell script to your local settings for listscan and gvfits to work!
NEW AIPS TASKS
This section intends to enlist the tasks that are not part of the official AIPS package but can be installed and used as normal AIPS tasks.
UVFXT: Task to fix timestamps on GMRT uv-data files within AIPS.
UVFLP: Task to to re-order channels in GMRT LSB uv-data files within AIPS.
Both these tasks have been written by Dave Green of Mullard Radio Astronomy Observatory, Cavendish Laboratory, Cambridge and can be obtained here. Task UVFXT is useful if you have GMRT data FITS file generated using listscan and gvfits version 1.12 or earlier and want to correct for the error that uvw co-ordinates were not computed at the centre of the integration. UVFLP (along with AIPS tasks BLOAT and DBCON) can be used to combine lta and ltb (i.e. the two IFs) data inside AIPS. Although the program `ltamerge' can be used to do the same with lta and ltb files it may be more convenient to merge data right at the end, i.e. just before imaging. This does make sense since the data editing as well as calibration for the two IFs are anyway done independently. Detailed help and guidelines to install and use these new tasks by Dave Green are available at the above link. However, for the convenience of users at GMRT we have made available both these tasks on all the general user machines. See FAQ for more.
Right click on the links and "Save Target as" to download the LTA utilities you may require. These are binaries should work on bash/tcsh. The LTA utilities are now called "lute" - it has now reached lute-1.17. This version merges the source codes of the GSB and hardware correlator versions into one common utility for data originating from either. The changes and updates to lute follow those to gvfits. Identically numbered versions of lute and gvfits are tied to each other. All the lute programs are available on astro0:/astro/Linux/bin at GMRT. The most important lutilities are:
ltahdr : Utility to list the scans in the LTA/LTB file
ltaedit : Utility to edit parameters in the LTA/LTB file
ltamerge : Utility to merge the LTA and LTB files. LTA and LTB are two different IFs in default observing modes, and correspond to two polarizations in the HIGHRES mode.
ltacomb : Utility to combine/concatenate LTA files. LTA1:LTA2:LTA3:.... or LTB1:LTB2:LTB3...
Files can be combined in time order also after merging LTA/LTB files.
FAQ and (historical)bugs
How do I run aips on machine astro1 from astro2?
This can be done as following:
>ssh -Y astro1 (from astro2)
For those who still swear by 31DEC04 / 31DEC05 and stoutly refuse to move to the latest 31DEC10 version, the following four questions are relevant:
Why I am not able to remote mount the tape from a machine having having AIPS 31DEC04 (or later) on to a mchine having version 31DEC03 (or earlier)?
In 31DEC04 AIPS, the reserved tape Logical Unit Numbers (LUNs) have been changed. Due to this, tasks of earlier versions cannot talk to TPMONs of 31DEC03 (or earlier) and viceversa. An effort has been made to keep the tape drives on machines that have AIPS version 31DEC04 or later. For smooth functioning, local users should upgrade their machines if they have 31DEC03 or older versions.
Why is it that different AIPS' tasks with version 31DEC05 are not able to interpret the sign of frequency increment correctly (and consistently) from the GMRT datasets?
We have noticed that for 31DEC05 some of the AIPS tasks (like IMAGR) are goofing up in reading the correct sign of frequency increment from our FITS files. The problem is due to change of keyword 'SIDE BAND' in FQ table for 31DEC05 to 'SIDEBAND'. According to Eric Greisen, this has been done to make FQ table more similar to other table types in AIPS. The lateset version of listscan and gvfits i.e. version 1.12 have taken this into account and must be used for generating FITS files for GMRT data. These files will work with 31DEC05 as well as earlier version of AIPS. Therefore to avoid this problem one should really be using listscan and gvfits versions 1.12 or later. These versions are installed at /astro/bin/Linux/ and can be sourced by 'source /astro/RC' or 'source /astro/RC.csh'.
Can one still use the FITS files generated using listscan and gvfits versions earlier than 1.12 with AIPS 31DEC05?
AIPS 31DEC05 can still be used for such GMRT datasets by changing the keyword 'SIDE BAND' to 'SIDEBAND' in FQ table. This may be done in following two steps:
Using AIPS' task TBOUT read the table outside AIPS system and using a text editor modify the keyword 'SIDE BAND' to 'SIDEBAND'. Then replace this modified FQ table with the earlier FQ table in the dataset. This can be done using TBIN.
Now, use the task TABED to assign the appropriate value to 'SIDEBAND'. For example, 'SIDEBAND' can be set to -1 using following values of the adverbs for TABED: OPTYPE 'REPL'; APARM 5 0; KEYWORD 'SIDEBAND'; KEYVALUE -1 0.
How do I use the new tasks UVFLP and UVFXT at GMRT machines?
For this type `setenv MYAIPS /usr/aips/MORETASKS/' for cshell or `export MYAIPS=/usr/aips/MORETASKS/' for bash before starting AIPS. Then inside AIPS say: version 'MYAIPS'. Saying version 'CDTST' inside AIPS will make these unavailable i.e. restore to default AIPS as obtained from NRAO site.
Frequency Increment problems with SPLIT and SPLAT
This bug has been recently reported by Dave Green of MRAO, Cavendish Laboratory, Cambridge. If you are combining USB and LSB GMRT data into the same FITS file as two IFs, make sure you are using the most recent version of AIPS 31DEC06 or 31DEC07.
When using FITS files with two IFs corresponding to combined USB and LSB data, in the 31DEC06 and 31DEC07 versions of SPLAT, the FQ table contained wrong entries for the frequency increment for the second IF; this was because the frequency increment(sense) is different for the two GMRT sidebands. The problem was the opposite in 31DEC05 version of AIPS - the first IF had wrong entries. The problem was identified as a bug in SPLAT and SPLIT as well. However, patches have been applied to both these tasks.
A second problem with SPLIT cropped up when trying to fix the above problem: the output frequencies were wrong if BCHAN was not 1. This has been patched as well in the 31DEC06 and 31DEC07 AIPS on 23 October 2007. See http://listmgr.cv.nrao.edu/pipermail/bananas/2007-October.txt and http://www.aips.nrao.edu/31DEC06/patches.html#018 for patch details. Also see these emails for a bit of background.
Update your SPLAT and SPLIT tasks if you are using 31DEC06 and 31DEC07 versions of AIPS. However, the FQ table can also be edited manually(!).
More problems with SPLIT and SPLAT !
The number of correlators was too high by a factor of three, causing array overruns and less than optimal scaling of compressed data. The relevance of this bug to GMRT is not known; nevertheless it is recommended to remain up-to-date. The mails from Eric Greisen, as on 07 Nov 2007 can be found here.
Last revised: 18 Mar 2010