SAGECal performance with large sky models

TL;DR

This paper evaluates the computational performance of the SAGECal calibration software when handling large sky models with up to 50,000 sources, including both point and extended sources, for advanced radio telescopes.

Contribution

It analyzes the compute load and performance implications of using extensive sky models in SAGECal calibration for modern radio astronomy.

Findings

Assessment of compute load for large sky models

Performance insights for calibration with 50,000 sources

Handling both point and extended sources in calibration

Abstract

As astronomical instruments become more sensitive, the requirements for the calibration software become more stringent; without accurate calibration solutions, thermal noise levels in images will not be reached and the scientific output of the instrument is degraded. Calibration requires bright sources with known properties, in particular with respect to their brightnesses as a function of frequency. However, for modern radio telescopes with a huge field of view, a single calibration source does not suffice; instead a sky model with tens of thousands of sources is needed. In this work, we investigate the compute load for such complicated sky models, with up to 50,000 sources, for the SAGECal calibration package. We have chosen half of the sources in these models to be point sources and half of them extended, which we represent by Gaussian profiles.