Validation of the HERA Phase I Epoch of Reionization 21 cm Power Spectrum Software Pipeline

TL;DR

This paper validates the HERA Phase I software pipeline for 21 cm power spectrum analysis using synthetic data, demonstrating its accuracy and identifying areas for improvement in future data analysis efforts.

Contribution

It presents a comprehensive validation framework for the HERA pipeline, including modular tests and end-to-end simulations, to ensure reliable power spectrum estimation from synthetic data.

Findings

Pipeline accurately recovers known input spectra within thermal noise levels.

Successfully detects amplified signals with high foreground suppression.

Identifies and characterizes sources of signal loss in the analysis pipeline.

Abstract

We describe the validation of the HERA Phase I software pipeline by a series of modular tests, building up to an end-to-end simulation. The philosophy of this approach is to validate the software and algorithms used in the Phase I upper limit analysis on wholly synthetic data satisfying the assumptions of that analysis, not addressing whether the actual data meet these assumptions. We discuss the organization of this validation approach, the specific modular tests performed, and the construction of the end-to-end simulations. We explicitly discuss the limitations in scope of the current simulation effort. With mock visibility data generated from a known analytic power spectrum and a wide range of realistic instrumental effects and foregrounds, we demonstrate that the current pipeline produces power spectrum estimates that are consistent with known analytic inputs to within thermal noise levels (at the 2 sigma level) for k > 0.2 h/Mpc for both bands and fields considered. Our input spectrum is intentionally amplified to enable a strong `detection' at k ~0.2 h/Mpc -- at the level of ~25 sigma -- with foregrounds dominating on larger scales, and thermal noise dominating at smaller scales. Our pipeline is able to detect this amplified input signal after suppressing foregrounds with a dynamic range (foreground to noise ratio) of > 10^7. Our validation test suite uncovered several sources of scale-independent signal loss throughout the pipeline, whose amplitude is well-characterized and accounted for in the final estimates. We conclude with a discussion of the steps required for the next round of data analysis.