v2.04

2013-04-06 André Offringa <of…nga@gmail.com>

AOFlagger changes for version 2.4.0

This version took longer than usual, but this is reflected by many large changes, most prominently: JVLA compatibility due to multi-SPW and multi-field support, improved Parkes compatibility due to SDFITS fixes, automatic loading of a strategy that matches the telescope and observation properties and switching between plots in the RFIGui.

Changes

  • Added a simple Parkes strategy (#19) – not extensively optimized, but seems to do a reasonable job (…much better than simple thresholding)
  • Measurement sets with multiple spectral windows are now fully supported (#11). All [reading modes](reading mode) will correctly read and write such measurement sets.
  • Measurement sets with multiple fields are now fully supported (#18). This and the previous feature has been requested many times, in particular to be able to easily flag JVLA data without ‘split()’-ting off the bands or sources. As with the previous feature, all reading modes will now correctly read and write multi-field measurement sets. Each sequence of correlations on a single source will be flagged individually. Thanks to R. van Weeren for providing test sets.
  • The RFIGui and aoflagger, when not specified otherwise, will now load an optimized strategy when an observation is opened. (#10) Telescopes for which this is currently supported are LOFAR, MWA, Parkes, JVLA (beta) and WSRT. In other cases, a “generic” strategy is loaded, which is the LOFAR strategy (which is actually not very generic – I hope to improve that in the next version).
  • An initial JVLA strategy was added, that I’ve optimized for a ~1.1 GHz observation that I’ve received from R. van Weeren. Together with the other 3 previous features, this allows to simply execute “aoflagger my_jvla_set.ms” on a standard JVLA multi-source/spw file and get good results (if your sets look like the set I got) without any further tweaking. If you have, send me “typical” observations of other frequencies so I can optimize this further.
  • The default WSRT strategy has been tweaked.
  • The strategy wizard has been improved: a “fast” option was added and the sensitivity can be changed. (#12)
  • The plotting window in the RFIGui now shows a list of all previous plots, which can be quickly reselected and its style can be changed (#9)
  • The main window in the rfigui now shows the selected correlation name (e.g. RT0 x RT1). Can also be turned off and/or changed. (#8)
  • The aoflagger console application will now automatically select the optimal number of threads when not overriden on cmd line (#6)
  • The [indirect reader](reading mode) has been made a lot more efficient. It will “preallocate” the files it will write (if [your filesystem supports it](http://linux.die.net/man/1/fallocate)) and will no longer create a file for every correlation, but rather uses only one for data and one for flags. This overcomes problems with “too many open files” on some environments.
  • When properly installed, the RFIGui and aoqplot applications will have a application/taskbar icon and show up in the menu of common desktop environments. (#14)
  • Toolbar has been cleaned up, the more commonly used buttons are now there & some icons were added. (#13)
  • Several menu items now got accelerator key, e.g. “F9” is “run strategy”.
  • Bugfix: G van Diepen send me a patch to fix variable length arrays and class/struct mixing, which caused problems on the Mac.
  • Bugfix: Fixed a bug in the SDFITS reader/writer. With some help from J. Delhaize, this version is able to open and flag Parkes SDFITS file both with the RFIGui and the aoflagger console application. (#21)
  • Bugfix: MWA strategy could silently fail
  • Bugfix: aoqplot will ask for filename if none given (#22)
  • Bugfix: Deadlock when showing error dialog box during execution of strategy
  • Bugfix: fixing rare bug in RFIGui when a set contains invalid times
  • Refactoring: The main RFIGUI class is huge, major efforts to clean it up and split some things in new classes in a model-view-control way.