[daip] IF handling

[Eric Greisen]

Orla Aaquist writes:

 > I am a little confused how the two IFs are handled by IMAGR (AIPS version
 > 15APR98).

        This is an antique.  You really should update.  See
            http://www.cv.nrao.edu/aips
        for information.

 > 
 > I am in the process of comparing, via UVSUB, two data sets form poorly
 > resolved planetary nebulae.  I am looking for expansion across the minor
 > axis (cf. Collin Masson, 1996).  Epoch 1 is from before 1987 and Epoch 2
 > data is from 1998.  When we originally handled the Epoch 1 data, we combined
 > the two IF's using the aips task DBCON (after initial calibration), and the
 > lower IF (14.9149 GHz) was usually shifted to the upper IF (14.9649 GHz)in
 > the concatenation.  Now, when imaging this old DBCONed data using the task
 > IMAGR, the map frequency shown in the image header (using IMHEAD) is 14.9649
 > GHz.

        AIPS stores the uv values in wavelengths.  DBCON just copies
the two data sets together - it does not "shift" the frequency of one
set to the other.  That shift would require new measurements in the uv
plane!  The first of the two data sets controls the header frequency
and that frequency is simply copied by IMAGR.

  However, when I use IMAGR on our latest uv-data, the frequency in the
 > image header is the average of the two IFs (14.9399 GHz).  Strangely,
 > however, the corresponding (new) uv-data set reports only the upper IF in
 > the header.  Why is this?

        I have no idea why the UV data set and IMAGR on that data set
should disagree on the frequency.  IMAGR on a data set with two IFs
uses the FQ table to tell it what the 2 frequencies actually are and
images them correctly.  In this case the header frequency gives the
frequency to which the uv values are computed and IMAGR uses the FQ
values plus the header to adjust the uv before gridding.

 > I have the impression that a change (decrease) in frequency will created a
 > slight, self-similar expansion of the image.  This worries me because a
 > small expansion due to a frequency shift may give rise to a false
 > interpretation of the data.  Is there some way to shift the frequency of the
 > lower IF to the value of the upper IF in the new data so that the data at
 > both epoch are being reported as the same frequency?  I could restrict
 > myself to only one IF, however I would like to know if there is another
 > solution, or if I am worrying needlessly.

     Unless your data have been screwed up somehow, IMAGR will handle
the frequencies correctly.  If you want me to examine things in more
detail, I will need to see the outputs of IMHEADER on all files and of
PRTAB on the FQ tables of each UV file.

 > 
 > I have tried to use UVFIX, but it shifts the coordinates to Epoch 2000, and
 > the image I obtain is smeared.

      UVFIX in your old version has errors I believe.  The worst was a
problem with subarrays.  If your data do not have subarrays then I am
suprised that things are "smeared".

 > 
 > One final, probably related, question.  The number of visibilities reported
 > in the image header of the old data is the sum of the visibilities in the
 > two concatenated files, whereas the number of visibilities reported in the
 > header of the new data seems to show only the number collected by one IF.
 > Is this correct, or am I miscounting?

   Read the output of IMHEAD - it will show an IF axis with 2 points
on it.  The separated (old) data counts each IF separately, the new
only counts each (dual) sample.

Eric Greisen