Detection of Signals from Cosmic Reionization using Radio Interferometric Signal Processing

TL;DR

This paper analyzes the requirements for removing bright extragalactic sources in radio interferometry to detect the cosmic reionization signal, focusing on calibration accuracy and source removal precision.

Contribution

It provides specific constraints on positional and calibration errors necessary for effective foreground removal in upcoming low-frequency radio arrays.

Findings

Bright sources must be removed with ~0.1 arc-second positional accuracy.

Residual calibration errors should be less than 0.2% in amplitude or 0.2 degrees in phase.

Effective foreground removal is feasible with current strategies if calibration errors are within specified limits.

Abstract

Observations of the HI 21cm transition line promises to be an important probe into the cosmic dark ages and epoch of reionization. One of the challenges for the detection of this signal is the accuracy of the foreground source removal. This paper investigates the extragalactic point source contamination and how accurately the bright sources ($\gtrsim 1$ ~Jy) should be removed in order to reach the desired RMS noise and be able to detect the 21cm transition line. Here, we consider position and flux errors in the global sky-model for these bright sources as well as the frequency independent residual calibration errors. The synthesized beam is the only frequency dependent term included here. This work determines the level of accuracy for the calibration and source removal schemes and puts forward constraints for the design of the cosmic reionization data reduction scheme for the upcoming low frequency arrays like MWA,PAPER, etc. We show that in order to detect the reionization signal the bright sources need to be removed from the data-sets with a positional accuracy of $\sim 0.1$ arc-second. Our results also demonstrate that the efficient foreground source removal strategies can only tolerate a frequency independent antenna based mean residual calibration error of $\lesssim 0.2 %$ in amplitude or $\lesssim 0.2$ degree in phase, if they are constant over each days of observations (6 hours). In future papers we will extend this analysis to the power spectral domain and also include the frequency dependent calibration errors and direction dependent errors (ionosphere, primary beam, etc).