Next: Examples Up: Automated Editing of Radio Previous: Application of flags to

Limitations

Although Pieflag can deal with a variety of data sets, it should not be used thoughtlessly.

If the amplitude on a baseline varies rapidly due to source structure, then even the rms-based flagging cannot detect bad points. Either the amplitude variations within each section will be high, making exceedingly large, or is not representative because the timescale of the amplitude variations changes largely throughout the experiment.

The amplitude-based flagging assumes that the contribution of astronomical sources to the signal is the same for all channels. This assumption is not valid if the signal is dominated by sources with large spectral indices. However, the rms-based flagging will be unaffected by this.

However, both of these effects can be avoided if a suitable source model, derived from the data in channels which are essentially free of RFI, is subtracted from the data before Pieflag is used.

Pieflag should be used with caution in spectral line observations, when the observed lines are strong. The channels containing line emission would be flagged if compared to a line-free reference channel. In spectral line detection experiments, however, Pieflag can be used because the data are dominated by receiver noise. It should be noted that spectral line observations are frequently much less affected by terrestrial RFI than continuum observations, because the bandwidths are much smaller and may be within protected parts of the electromagnetic spectrum. A further downside of using Pieflag on spectral line data sets can be that the computing time increases linearly with the number of channels. As a guideline, the computing time for a 12h observation with all six ATCA antennae and a correlator cycle time of 10s (the default mode) was found to be 6s per channel, using an otherwise idle 2.8GHz PC running Linux.

Next: Examples Up: Automated Editing of Radio Previous: Application of flags to