= [wiki:Meetings]/20081212 =

Warwick has gotten things to a point where [http://www.narrabri.atnf.csiro.au/observing/CABB.html images can be made from CABB data.]
We need to get ready to start making changes to code and testing them.

== Agenda ==
 change of project scientist::
   Emil steps in for Naomi
 what role for Jamie Stevens?::
   Jamie starts ... January 12th?
 state of MIRIAD code changes::
   We have a simple python script, is it
    * still working?
    * in RCS yet?
 availability of test versions of programs::
   Vince to report
 plan for January::
   There'll be pressure to get code ready. Who's away, when?

----

== Minutes ==

Warwick gave a report on the status of CABB, which has advanced a lot now.
The system has produced images (from simultaneous observations) that look very close to the old system.
 * polarisation images differ. It may be just the error response that differs, i.e. it could be that the differences are not significant.
 * relative amplitudes and phases (of the antennae?) appear to be wrong.
  * needs a change to Tsys output formatting (right data being written to wrong record)
  * need to investigate how the XY phase is being formed and how it is being recorded
 * continuum output available only (no zooms)
 * already writing autocorrelation data, though in a potentially wasteful way.
  * the data are being written as two fake 'timing bins'. ugh.
  * the way they are being written means there is a lot of redundant data.[[BR]]
    !MarkW suggested 'packing' the correlations in some way, after discussion this was judged not worth the effort.
 * Coping with hardware failures: if a module dies, lots of channels will disappear from the bandpass. However these will not be in a contiguous chunk. We'll need a way to deal with this. The discussion suggested the 'channel weight' flags can be used for the continuum case at least.
  * In the case of zoom nodes, each central frequency is a separate (FITS binary) table with all baselines in it (including the n-n "autocorrelation baselines"). The spectral weights feature is in the RPFITS library already, but it writes all baselines for every visibility; this leads to a 50% overhead in the raw data files. It's unclear how to deal with this.

Warwick sees two main issues for the software side:
 1. using information in the autocorrelations to calibrate on a per-channel basis
 1. dealing with the spectral weights issue.


Xmas absences:

  || Who   ||Last Day ||Back In||
  ||Dave   ||19 Dec   ||05 Jan||
  ||Dave   || ? Jan   || ?+4 weeks||
  ||Emil   ||19 Dec   ||12 Jan||
  ||Jamie  || N/A     ||12 Jan||
  ||MarkC ||24 Dec   ||12 Jan||
  ||MarkW || ? || ? ||
  ||Naomi  ||19 Dec   ||late 2009||
  ||Vince  ||24 Dec   ||05 Jan||
  ||Warwick||24 Dec   ||05 Jan||



=== Actions ===
 * Emil becomes PS as of today.
 * Jamie will be dealing with "CABB issues" from day 1. He will be doing calibrator observations and reductions, for example, which may well turn up problems.
 * Emil, in collaboration with Jim Caswell, to select some test fields - for repeated observations as the system implementation goes forward.
 * Vince to work with !MarkC on implementing a 'test version' of all the programs.
 * Next meeting 2009-01-09.

== Participants ==
 * Dave
 * MarkC
 * MarkW
 * Emil
 * Naomi
 * Warwick

[wiki:Meetings Meetings Home]