[Difx-users] difx2fits bug for a mixed pcal setup

[Leonid Petrov]

Dear colleagues, 

  I found a bug in difx2fits. Difx2fits silently did not put 
autocorrelations in FITS-IDI for one of the stations. 

  A small dataset is located at http://astrogeo.org/s0/corr_difx/vo0051_bug

  Depending on whether an affected station (in this case YJ) participate
in the 1st scan in the list or not, results for the 2nd scan are different

difx2fits --override-version -v --localdir --antpol --polxy2hv --difx-pcal-interval 10 \
   h_1068.input \
   h_1069.input \
/tmp/vo0051_01a.fits  # correct


difx2fits --override-version -v --localdir --antpol --polxy2hv --difx-pcal-interval 10 \
   h_1049.input \
   h_1069.input \
/tmp/vo0051_01b.fits # incorrect: autocorrelation of YJ is missing: records with autocorrelation are silently skipped.

My analysis showed that station YJ uses pcal with spacing 10MHz, while
other stations use pcal spacing 5MHz. toneSelection = all was not used
in v2d. Therefore, *input file command to extract two tones per IF.
When configs are merged, IFs of YJ and IFs of other stations are considered
different. If YJ participated in the first scan in the list, nothing wrong
happens. If YJ did not participate in the first scan, difx2fits considers
these IF are new... but then something wrong happens. It creates an additional
config, but does not assign freqID correctly. Records without a correct freqID
are skipped.
  
  I failed to fix this bug, although I found a way to circumvent it. See the next 
post. I added printing statistics of skipped records.

Leonid
2020.09.18_22:44:14