Real-time calibration of the AARTFAAC array for transient detection

¹ Anton Pannekoek Institute, University of Amsterdam, Postbus 94249, 1090 GE Amsterdam, The Netherlands
e-mail: F.Huizinga@uva.nl; r.a.m.j.wijers@uva.nl
² ASTRON, The Netherlands Institute for Radio Astronomy, PO Box 2, 7990 AA Dwingeloo, The Netherlands
e-mail: wijnholds@astron.nl; wijnholds@astron.nl

Received: 18 February 2014
Accepted: 5 May 2014

Abstract

The search for transient phenomena at low radio frequencies is now coming of age with the development of radio sky monitors with a large field of view, which are made feasible by new developments in calibration algorithms and computing. However, accurate calibration of such arrays is challenging, especially within the latency requirements of near real-time transient monitors, and is the main cause of limiting their sensitivities. This paper describes a strategy for real-time, wide-field direction-dependent calibration of the Amsterdam-ASTRON Radio Transients Facility and Analysis Center (AARTFAAC) array, which is a sensitive, continuously available all-sky monitor based on the LOw Frequency ARray (LOFAR). The monitor operates in a zenith pointing, snapshot imaging mode for image plane detection of bright radio transients. We show that a tracking calibration approach with solution propagation satisfies our latency, computing, and calibration accuracy constraints. We characterize the instrument and verify the calibration strategy under a variety of observing conditions. This brings out several phenomena, which can bias the calibration. The real-time nature of the application further imposes strict latency and computational constraints. We find that although ionosphere-induced phase errors present a major impediment to accurate calibration, these can be corrected in the direction of the brightest few sources to significantly improve image quality. Our real-time calibration pipeline implementation processes a single spectral channel of a snapshot observation in ~0.2 s on test hardware, which is well within its latency budget. Autonomously calibrating and imaging one second snapshots, our approach leads to a typical image noise of ~10 Jy for a ~90 kHz channel, reaching dynamic ranges of ~2000:1. We also show that difference imaging allows thermal-noise limited transient detection, despite the instrument being confusion-noise limited.