[daip] Upadates of FITLD/SPLAT?

[Eric Greisen]

Ciriaco Goddi writes:
 > Hello Eric,
 > 
 > I am running the 31DEC07 aips version and I found some
 > problems/inconsistencies running FITLD and SPLAT lately (the last few
 > weeks).
 > More in detail, I've been working on a multi-epoch VLBA dataset of SiO
 > masers, consisting of two IFs (v_rest=42.8 and 43.1 GHz). For the most
 > recent datset (loaded on disk on Oct 29 with FITLD), I had some problems
 > creating images of both IFs after the calibration, in particular I
 > obtained a good map for IF1 but only strange wraps for IF2 w/ signal (the
 > uv-data looked good for both IFs, though). Trying to figure out what went
 > wrong in the calibration, I did a simple test, I tried to produce channel
 > maps WITHOUT applying any calibration, that is: run SPLAT
 > on the FITLD file choosing a source and a specific channel, then CALIB on
 > the single-source, single-channel SPLAT
 > file, and finally IMAGR on the output from CALIB.
 > Again, I was able to produce reasonably good maps for IF1 but not for IF2.
 > Trying the same procedures for previous epochs, I obtained good results
 > for both IFs.

    Your test therefore clearly shows that the data for IF2 does not
calibrate correctly for some reason.  It is indeed true that I have
modified FITLD recently - in the UV tables section appropriate to
reading FITAB output.  If you downloaded processed UV data from the
archive it may well have been in that form.  FITLD was changed on the
15th of October and corrected on the 25th.  The errors that were
corrected were generally fatal or very damaging.  Your data, if
affected, would not have "looked good".  Have you checked calibration
tables and run UVPRT to look at the UV data (w and w/o calibration)?

What sort of file did you download - raw visibilities or ones that
have passed through AIPS' pipeline?

 > 
 > As a further test, today I tried to load with FITLD a dataset
 > corresponding to an old epoch where previously I did not encounter any
 > problem, and now I notice these error messages in FITLD:
 > FITLD2: PHINI: 'PC_FREQ_1' COLUMN HAS NO VALUES
 > FITLD2: PHINI: 'PC_REAL_1' COLUMN HAS NO VALUES
 > FITLD2: PHINI: 'PC_IMAG_1' COLUMN HAS NO VALUES
 > FITLD2: PHINI: 'PC_RATE_1' COLUMN HAS NO VALUES
 > Trying to create a channel map as described above for this datset, I found
 > that now IF1
 > shows weird wraps and no signal in the map whereas IF2 looks good.
 > Applying the same procedure to a file loaded with FITLD on OCT 23,
 > everything looks good.
 > Has FITLD changed since OCT 23?
 > Is that possible that FITLD and other tasks like SPLAT read/write the
 > header info (e.g. freq) in a different way?

SPLAT has also been corrected.  It had troubles if you ran in the mode
that reduced the number of spectral channels by averaging.  Again it
did not get the data wrong, but it labeled the frequencies wrongly.

I will ask our VLBI expert to check all this.

Eric Greisen