CubiCal - Fast radio interferometric calibration suite exploiting complex optimisation

TL;DR

CubiCal is a high-performance Python package for radio interferometric calibration that leverages complex optimisation techniques to efficiently handle chains of Jones terms, outperforming existing tools in speed and flexibility.

Contribution

The paper introduces specialized solvers for complex gains parameterized by real values and extends complex optimisation methods to chains of Jones terms, enhancing calibration efficiency.

Findings

CubiCal achieves faster calibration compared to existing tools.

It effectively handles both simulated and real data.

The package supports direction-dependent and independent calibration.

Abstract

It has recently been shown that radio interferometric gain calibration can be expressed succinctly in the language of complex optimisation. In addition to providing an elegant framework for further development, it exposes properties of the calibration problem which can be exploited to accelerate traditional non-linear least squares solvers such as Gauss-Newton and Levenberg-Marquardt. We extend existing derivations to chains of Jones terms: products of several gains which model different aberrant effects. In doing so, we find that the useful properties found in the single term case still hold. We also develop several specialised solvers which deal with complex gains parameterised by real values. The newly developed solvers have been implemented in a Python package called CubiCal, which uses a combination of Cython, multiprocessing and shared memory to leverage the power of modern hardware. We apply CubiCal to both simulated and real data, and perform both direction-independent and direction-dependent self-calibration. Finally, we present the results of some rudimentary profiling to show that CubiCal is competitive with respect to existing calibration tools such as MeqTrees.