skip to main content

Title: Optimizing Sparse RFI Prediction using Deep Learning
Abstract Radio Frequency Interference (RFI) is an ever-present limiting factor among radio telescopes even in the most remote observing locations. When looking to retain the maximum amount of sensitivity and reduce contamination for Epoch of Reionization studies, the identification and removal of RFI is especially important. In addition to improved RFI identification, we must also take into account computational efficiency of the RFI-Identification algorithm as radio interferometer arrays such as the Hydrogen Epoch of Reionization Array grow larger in number of receivers. To address this, we present a Deep Fully Convolutional Neural Network (DFCN) that is comprehensive in its use of interferometric data, where both amplitude and phase information are used jointly for identifying RFI. We train the network using simulated HERA visibilities containing mock RFI, yielding a known “ground truth” dataset for evaluating the accuracy of various RFI algorithms. Evaluation of the DFCN model is performed on observations from the 67 dish build-out, HERA-67, and achieves a data throughput of 1.6× 105 HERA time-ordered 1024 channeled visibilities per hour per GPU. We determine that relative to an amplitude only network including visibility phase adds important adjacent time-frequency context which increases discrimination between RFI and Non-RFI. The inclusion of phase more » when predicting achieves a Recall of 0.81, Precision of 0.58, and F2 score of 0.75 as applied to our HERA-67 observations. « less
; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; more » ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; « less
Award ID(s):
Publication Date:
Journal Name:
Monthly Notices of the Royal Astronomical Society
Sponsoring Org:
National Science Foundation
More Like this

    Combining the visibilities measured by an interferometer to form a cosmological power spectrum is a complicated process. In a delay-based analysis, the mapping between instrumental and cosmological space is not a one-to-one relation. Instead, neighbouring modes contribute to the power measured at one point, with their respective contributions encoded in the window functions. To better understand the power measured by an interferometer, we assess the impact of instrument characteristics and analysis choices on these window functions. Focusing on the Hydrogen Epoch of Reionization Array (HERA) as a case study, we find that long-baseline observations correspond to enhanced low-k tails of the window functions, which facilitate foreground leakage, whilst an informed choice of bandwidth and frequency taper can reduce said tails. With simple test cases and realistic simulations, we show that, apart from tracing mode mixing, the window functions help accurately reconstruct the power spectrum estimator of simulated visibilities. The window functions depend strongly on the beam chromaticity and less on its spatial structure – a Gaussian approximation, ignoring side lobes, is sufficient. Finally, we investigate the potential of asymmetric window functions, down-weighting the contribution of low-k power to avoid foreground leakage. The window functions presented here correspond to themore »latest HERA upper limits for the full Phase I data. They allow an accurate reconstruction of the power spectrum measured by the instrument and will be used in future analyses to confront theoretical models and data directly in cylindrical space.

    « less
  2. Abstract We report upper limits on the Epoch of Reionization 21 cm power spectrum at redshifts 7.9 and 10.4 with 18 nights of data (∼36 hr of integration) from Phase I of the Hydrogen Epoch of Reionization Array (HERA). The Phase I data show evidence for systematics that can be largely suppressed with systematic models down to a dynamic range of ∼10 9 with respect to the peak foreground power. This yields a 95% confidence upper limit on the 21 cm power spectrum of Δ 21 2 ≤ ( 30.76 ) 2 mK 2 at k = 0.192 h Mpc −1 at z = 7.9, and also Δ 21 2 ≤ ( 95.74 ) 2 mK 2 at k = 0.256 h Mpc −1 at z = 10.4. At z = 7.9, these limits are the most sensitive to date by over an order of magnitude. While we find evidence for residual systematics at low line-of-sight Fourier k ∥ modes, at high k ∥ modes we find our data to be largely consistent with thermal noise, an indicator that the system could benefit from deeper integrations. The observed systematics could be due to radio frequency interference, cable subreflections, or residualmore »instrumental cross-coupling, and warrant further study. This analysis emphasizes algorithms that have minimal inherent signal loss, although we do perform a careful accounting in a companion paper of the small forms of loss or bias associated with the pipeline. Overall, these results are a promising first step in the development of a tuned, instrument-specific analysis pipeline for HERA, particularly as Phase II construction is completed en route to reaching the full sensitivity of the experiment.« less
  3. Abstract The detection of the Epoch of Reionization (EoR) delay power spectrum using a ”foreground avoidance method” highly depends on the instrument chromaticity. The systematic effects induced by the radio-telescope spread the foreground signal in the delay domain, which contaminates the EoR window theoretically observable. Applied to the Hydrogen Epoch of Reionization Array (HERA), this paper combines detailed electromagnetic and electrical simulations in order to model the chromatic effects of the instrument, and quantify its frequency and time responses. In particular, the effects of the analogue receiver, transmission cables, and mutual coupling are included. These simulations are able to accurately predict the intensity of the reflections occurring in the 150-m cable which links the antenna to the back-end. They also show that electromagnetic waves can propagate from one dish to another one through large sections of the array due to mutual coupling. The simulated system time response is attenuated by a factor 104 after a characteristic delay which depends on the size of the array and on the antenna position. Ultimately, the system response is attenuated by a factor 105 after 1400 ns because of the reflections in the cable, which corresponds to characterizable k∥-modes above 0.7 $h\,\,\rm {Mpc}^{-1}$ at 150 MHz.more »Thus, this new study shows that the detection of the EoR signal with HERA Phase I will be more challenging than expected. On the other hand, it improves our understanding of the telescope, which is essential to mitigate the instrument chromaticity.« less
  4. ABSTRACT We quantify the effect of radio frequency interference (RFI) on measurements of the 21-cm power spectrum during the Epoch of Reionization (EoR). Specifically, we investigate how the frequency structure of RFI source emission generates contamination in higher order wave modes, which is much more problematic than smooth-spectrum foreground sources. Using a relatively optimistic EoR model, we find that even a single relatively dim RFI source can overwhelm the EoR power spectrum signal of $\sim 10\, {\rm mK}^2$ for modes $0.1 \ \lt k \lt 2 \, h\, {\rm Mpc}^{-1}$. If the total apparent RFI flux density in the final power spectrum integration is kept below 1 mJy, an EoR signal resembling this optimistic model should be detectable for modes $k \lt 0.9\, h\, {\rm Mpc}^{-1}$, given no other systematic contaminants and an error tolerance as high as 10 per cent. More pessimistic models will be more restrictive. These results emphasize the need for highly effective RFI mitigation strategies for telescopes used to search for the EoR.
  5. Foregrounds with polarization states that are not smooth functions of frequency present a challenge to HI Epoch of Reionization (EoR) power spectrum measurements if they are not cleanly separated from the desired Stokes I signal. The intrinsic polarization impurity of an antenna's electromagnetic response limits the degree to which components of the polarization state on the sky can be separated from one another, leading to the possibility that this frequency structure could be confused for HI emission. We investigate the potential of Faraday rotation by the Earth's ionosphere to provide a mechanism for both mitigation of, and systematic tests for, this contamination. Specifically, we consider the delay power spectrum estimator, which relies on the expectation that foregrounds will be separated from the cosmological signal by a clearly demarcated boundary in Fourier space, and is being used by the Hydrogen Epoch of Reionization Array (HERA) experiment. Through simulations of visibility measurements which include the ionospheric Faraday rotation calculated from real historical ionospheric plasma density data, we find that the incoherent averaging of the polarization state over repeated observations of the sky may attenuate polarization leakage in the power spectrum by a factor of 10 or more. Additionally, this effect provides amore »way to test for the presence of polarized foreground contamination in the EoR power spectrum estimate.« less