[daip] Re: AR504

[Eric Greisen]

Data Analysts writes:

 > Meri - thanks.  I experimented with the
 > online fillm, with whatever is running now, and
 > it seems fine.  if i understand correctly,
 > i could mess things up if i simply even
 > look at the data set while things are actually
 > filling.

     First - two questions: why do you think the analysts are AIPS
experts?  They are not There is a daip at nrao.edu address for aips
questions.  Second - it had been my impression that there were strong
limits placed on the use of on-line FILLM.  I suppose that if you are
filling a disk in the AOC you are within those limits.  I was
forbidden from including the code with AIPS as shipped.

Now your questions:

There is a possibility of destructive collisions.  If you read an
inactive output data set (or write with say UVPLT) and it tries to
activate while you are doing it, then it will fail to add a WRIT
status because of your READ or WRIT and FILLM will quit.  It the file
is currently being written, you may be able to open the UV file and
read it or file locking may get in your way and cause the read process
to just hang up.  You might ewxperiment with your current FILLM to see
since you do not care if it breaks.

 >    is this true?  or only if i
 > do anything specific to write?
 >    if i use the stop and restart option,
 > do i lose anything with dbconning later
 > on to put the whole day back together?

    The BREAK option is a good one.  It causes the next scans to go to
new files and properly closes the current ones without any FILLM
restarts.

 >    do i need to bring in any other external
 > files (gain, etc) as part of the fillm?

    FILLM creates all needed files.  The only difference between
real-time and off-line FILLM is the data source.

 >    do you recommend getting rid of shadowing
 > during fillm (i would probably get rid of
 > anything shadowed at all, so would just
 > leave cparm(4) 0.

    I strongly recommend the default shadowing when there are
subarrays.  Anything but the default will know only about those
antennas in the subarray and will not know about another antenna
shadowing one inside the subarray.  The default allows us to use an
on-line flag that knows about all antennas.

 >    [it's really unfortunate that different
 > language about shadowing is used in
 > observe and fillm. the former talks about
 > 'amount of shadowing' e.g., 2 meters,
 > and the latter talks about 'baseline
 > length', where presumably anything less
 > than 25 meters is shadowed.]

    Any degree of shadowing including near-miss kinds are likely to
suffer from cross-talk and other rfi-like problems.  I do not
recommend allowing slightly shadowed data into your data set.

Eric Greisen