[daip] AIPS visibility data export / task writing

[Eric Greisen]

Stephan Wenger wrote:
> Dear Eric Greisen,
> 
> I am currently working on an image reconstruction algorithm for radio
> interferometry together with Ylva Pihlström from UNM. We hope to be able
> to reduce the amount of manual work that goes into creating a
> representable image, and would like to integrate our algorithm with
> AIPS. Dr. Pihlström told me that you might be able to resolve some of my
> AIPS-related questions, and I would be very glad if you would find the
> time to give me a hint on the two following subjects.
> 
> Firstly, in order to test the algorithm on actual data, I need to read
> the calibrated visibility data as well as the baselines (in wavelengths)
> and maybe the corresponding weights from AIPS. I had a look at some of
> the FITS files that can be exported from AIPS but wasn't able to locate
> this information. Is there a way to export the preprocessed data into
> plain text or FITS files?
> 
> Then, if the algorithm turns out to give satisfactory results (the
> output on simulated measurements looks promising), I would be pleased to
> make it publicly available as an AIPS task. Can you tell me where I can
> find information on how to write a good AIPS task? Unfortunately, my
> AIPS experience is still very limited.

I will try to answer although the questions are rather broad or vague
in what will be interesting details.

1. AIPS exports UV data in FITS format through tasks FITTP and FITAB.
The former is described in some detail in the 2 volume text known as 
Going AIPS.  Chapter 14 in particular describes the output of FITTP.
FITAB puts the UV data into FITS tables instead of random groups.
The actual binary portion of those data are identical to the random 
groups if the data are not compressed.

2. The text about AIPS tasks is all of Going AIPS.  Chapters 1 and 2 
provide a good introduction - esp 2 which talks about what we call 
paraform tasks.  These are tasks in which the portions someone might 
wish to change to implement a new algorithm are clearly identified.
FUDGE has been widely used - by the aips group and outsiders - to
implement tasks.  It has a subroutine DIDDLE which is fed UV data one 
sample at a time and which takes back the user's modified data and 
writes it into an aips output file.  FUDGE hides (in plain sight but you 
are told to ignore those subroutines) all the hard details.

AIPS has only limited ability to export data in text form - tasks like 
PRTUV, UVPRT, etc will print UV data but the number of characters are 
limited so the accuracy is limited as are the number of IFs, spectral 
channels etc.  You can set DOCRT=-3 or so and leave out the page breaks 
and most header info and write to an OUTPRINT text file.

You might find it easier to implement your algorithm inside aips using 
FUDGE than having to read the FITTP output.  Note that GOING AIPS is 
pretty old but much of what it says is okay although the details should 
be checked carefully using the current code.

Locally developed aips tasks can only be worked upon if you have done a 
text (local compilation) installation of aips or have the Intel compiler 
that matches the one we use (for binary installations).

Going aips can be accessed a chapter at a time from the web

http://www.aips.nrao.edu/goaips.html

and your local aips installation has it all as

$AIPS_ROOT/TEXT/PUBL/CHAPn.PS

Eric Greisen