MWA Project Meeting - 2026

Identifying “bad” channels from calibration solutions

Not scheduled

20m

Main sessions: Room 190, Break-out sessions: Room 125 (Engineering Research Center, Brown University)

Presentation High Performance Computing

Speaker

Mr Kenil Ajudiya (Curtin Institute of Radio Astronomy)

Description

Inyarrimanha Ilgari Bundara, CSIRO’s Murchison Radiastronomy Observatory, is a radio quiet zone that reduces “local” RFI observed by the MWA. However, it is vulnerable to several types of RFI - satellite signals, RFI ducted by the ionosphere, television signals from far away, etc. The effect of RFI is to increase the background noise level, decreasing the S/N ratio of astronomical signals of interest, effectively reducing the sensitivity of the telescope. Much of the RFI that the MWA is exposed to is persistent and band-limited. We present a working approach to identify and flag “bad” (RFI-contaminated) channels in a very inexpensive way using calibration solutions derived using hyperdrive which can be applied directly to the visibilities, without requiring to flag them in the measurement set, avoiding a significant amount of disk I/O operation. The gain amplitudes output by hyperdrive are proportional to the visibility amplitudes. Since RFI is almost always the most dominant contributor to linearly polarised signals that are observed by all the baselines (astronomical transients are well localised in the sky), it is easier to identify bad channels using use only the sum of the calibration gain amplitudes of XY and YX polarisations from all the tiles. We show that simple, statistical outlier rejection algorithms are very effective in identifying the bad channels, leading to improvements in the image quality, for example, an image in which RFI dominated the total flux and only bright sources were detected could not be easily distinguished from an image that one would get from an RFI-free observation post flagging and re-imaging. The results are particularly useful for searching long period transients since the background noise levels of the images drop significantly. Such an approach is also suitable for science goals like studying the epoch of reionisation, where data quality is more important than data quantity (i.e. systematics dominate).