NRAO
 

Astronomical Image Processing System

[AIPS]

Problems and considerations in using AIPS


  • Considerations for 31DEC07
  • Considerations for 31DEC08
  • Considerations for 31DEC09
  • It is now possible to search the CHANGE.DOC files used by AIPS programmers to document the current changes being made to AIPS. This can be very useful if you want to keep up with what is happening to your favorite AIPS task or area of interest (e.g., VLBI). The current full content of the last three CHANGE.DOC files are available: 31DEC09, 31DEC08, and 31DEC07.


    Considerations for 31DEC07

    Under Considerations for 31DEC09 is discussed a change to the format and contents of AIPS antenna tables. This change makes an antenna table touched by 31DEC09 different from what earlier versions of AIPS expect. This difference probably does not matter for VLA and GMRT data. But for VLBI data, the difference is significant. Pre-processed VLBA data from the archive, processed after September 20, 2009 will be in the new format. A procedure called ANFIX8 is available from the 31DEC08 patch page which will convert the antenna file back into the older VLBI (left-handed) geometry.

    AIPS maintenance

    Note that, even if you are using an older AIPS, it is useful to occasionally run a midnight job or be sure you get any recent patches. If you have not done this on 31DEC07 since November 18, 2008 you do not have the fully corrected 31DEC07.

    There are eleven 31DEC07 patches:
    1. REBYTE did not handle tables with long rows (IM, possibly BP) correctly 2008-01-09
    2. FITLD did not translate WR tables correctly 2008-01-18
    3. DFT model division did not set weights correctly 2008-03-05
    4. FILLM did not scale and weight cross-hand data for some baselines correctly 2008-03-05
    5. VISDFT did not do multi-scale DFT model subtraction and division correctly 2008-04-29
    6. FILLM did not set the CORRCOEF keyword correctly for recent data 2008-06-19
    7. FILLM did not apply on-line flags correctly in modes 4, PA, PB, 2BC, and 2BD 2008-07-08
    8. GO verb limited the usage of GPOS and FPOS to less than some tasks require 2008-08-13
    9. FACSET got the wrong standard calibrator flux from the Clean Components file primarily for 3C286 2008-09-10
    10. The Mac OS/X version "leopard" requires changes to XAS and procedures START_AIPS and START_TVSERVERS 2008-09-26
    11. FILLM did not compute the shadowing test properly 2008-11-18

    Considerations for 31DEC08

    Under Considerations for 31DEC09 is discussed a change to the format and contents of AIPS antenna tables. This change makes an antenna table touched by 31DEC09 different from what earlier versions of AIPS expect. This difference probably does not matter for VLA and GMRT data. But for VLBI data, the difference is significant. Pre-processed VLBA data from the archive, processed after September 20, 2009 will be in the new format. A procedure called ANFIX8 is available from the 31DEC08 patch page which will convert the antenna file back into the older VLBI (left-handed) geometry.

    AIPS maintenance

    Note that, even if you are using an older AIPS, it is useful to occasionally run a midnight job or be sure you get any recent patches. These patches correct bugs that have been found that are believed to potentially cause significant problems for user's data. For example, recent patches for 31DEC08 AIPS have corrected errors in the use of the ICHANSEL adverb in BPASS and in the frequency put in the AN table by FITLD when rearranging the frequency order from correlator FITS IDI files.

    There are eleven 31DEC08 patches:
    1. CALIB could infinite loop if 1st NX record 0 length 2009-01-27
    2. TYAPL would apply the highest flag table without permission when it should just be copied 2009-02-04
    3. OFM files were not changed to the new TV intensity ranges 2009-04-02
    4. IMAGR TV option FORCE A FIELD caused all facets to be re-imaged; scratch files way too large under some circumstances 2009-04-22
    5. DBCON copied keyword MAXBLINE which confused later imaging routines 2009-04-27
    6. IMAGR had a limit test that was too small when restoring large numbers of CCs from 1 facet to another 2009-05-18
    7. IMAGR had a tendency to sort input data sets when it did not really have to do so 2009-06-19
    8. install.pl skipped the automatic tarball download when one was missing 2009-07-26
    9. FITLD did not correct the AN table frequency when rearranging frequency order for IDI-format data 2009-08-18
    10. BPASS did not translate the ICHANSEL adverb correctly, causing all channels to be averaged 2009-11-11
    11. ANFIX8 is a new RUN file and procedure to check ANtenna files for being in 31DEC09 format and fix them. 2009-11-30

    Considerations for 31DEC09

    We strongly recommend that you update your 31DEC09 with the so-called Midnight Job ("MNJ") on a regular basis. If you have run a MNJ since September 20, 2009, you should run another one now.

    It is now possible to search the CHANGE.DOC files used by AIPS programmers to document the current changes being made to AIPS. This can be very useful if you want to keep up with what is happening to your favorite AIPS task or area of interest (e.g., VLBI). The current full content of CHANGE.DOC for 31DEC09 is also available.

    Change of Terrestrial Reference Frame

    We thank Craig Walker for providing the text below:

    The change is a move to a right-handed coordinate system for stations in all UV data sets. The puts AIPS in agreement with the ITRF rather than an ancient left-handed system (left over from MarkII days?) and is therefore the right thing to do. But it does introduce compatibility issues during the transition. Note that this applies to VLBI data, the VLA and GMRT have always been right handed. In practice, the change means that the sign of BY in the AN table is reversed. This change was made in the 31DEC09 version of AIPS on September 20, 2009. If the 31DEC09 version of AIPS sees an AN table of the old type, identified by knowing the approximately correct coordinates of many stations, it converts the table to the new convention so backward compatibility is good. The problem comes when a data set with the ITRF convention is read by an older version of AIPS, be it 31DEC08 or earlier, or even a 31DEC09 version that has not had a midnight job run since Sept. 20. The older versions will not recognize that the coordinates are different and any task that needs station elevations will get them wrong. This includes UVFIX, PLOT routines doing elevation or parallactic angle, CLCOR "POLR", "PANG", "ANTP", "IONS", "ATMO", et al. corrections, UVFLG elevation flagging, CALIB elevation cutoff for gain normalization, TECOR, ANTAB with opacity corrections, DELZN, PCAL, LPCAL, CVEL and probably others. In other words, it will almost certainly affect your processing.

    Users can get into trouble during the transition in a couple of ways. If they use both new and old versions of AIPS in their processing, the recent version will convert the AN table, but the old version won't know about it. So, for now, it is best not to mix versions. The processed FITS files in the VLBA archive are produced using the latest AIPS. Therefore, files made since around Sept 20 will have the new antenna convention. If those are downloaded and processed in an older AIPS, there will be problems. A procedure, ANFIX8 is now available from a midnight job for 31DEC08 or from the 31DEC08 patch INFO patch info on the web that will convert back to the frame needed by the 31DEC08 AIPS. Alternatively, one can download the individual job files (the FITS IDI format files) and run the procedures (VLBALOAD, VLBAFIX) to read and combine them.

    Antenna coordinate bug in FITLD for FITS-IDI data

    There was a bug in the implementation of the above change that caused the old convention to be used for any station that did not appear in the first correlator job file. This will affect observations for which some station(s), typically Mauna Kea, start late. The archive files are subject to this bug. Users should check their AN tables carefully to see if a fix is needed. For example, all VLBA antennas should have the same sign of BY. A procedure to fix affected data (ANCHECK) will be available with the midnight job after Dec 1. Processed files in the archive are subject to this bug. They will be fixed eventually, but meanwhile be careful with such files. If you schedule calibrators on all antennas at the start of your run, you may not be subject to the problem. This bug affects AIPS 31DEC09 with midnight jobs run between Sept. 20 and Nov. 26. Processed archive files made during this date range are suspect. The RUN file and procedure ANCHECK in 31DEC09 will allow you to check your VLBI antenna files.

    Polarization calibration bug

    The application of polarization calibration (DOPOL=1) was entirely broken in 31DEC09 AIPS versions updated between Sept. 20 and Nov. 17. The data are modified, but not correctly. Any application of polarization calibration done with an affected AIPS should be redone after running a midnight job. Note that the derivation of the D terms is probably ok if your antenna coordinates are ok.

    Polarization EVPA correction bug

    The POLR option in CLCOR, used to adjust the absolute polarization position angle, was broken between Sept. 20 and Nov 26 in the 31DEC09 AIPS. The CL table is corrected properly, but the AN table may not be. The task made an error in determining the number of IFs in the table and might only modify some, but not all IFs. This leads to IF-dependent errors in adjusting the angle when the polarization rotation is applied. The fix is to update AIPS and re-apply the rotations with CLCOR. This requires zeroing the angle in the AN table or deleting the AN table and copying an unmodified one, that hopefully you saved early in your processing, in its place. Then rerun the polarization calibration.


Modified on $Date: 2009/12/01 18:22:39 $
Eric W. Greisen