TABASCAL II: Removing Multi-Satellite Interference from Point-Source Radio Astronomy Observations

TL;DR

TABASCAL II tackles satellite-based RFI in point-source radio interferometry by jointly estimating astronomical visibilities, antenna gains, and RFI signals within a Bayesian framework. It deploys Gaussian-process priors to model time variability and fringe-rate filters applied in multiple directions, leveraging antenna-based decomposition to separate correlated signals and account for fringe-winding effects. The method demonstrates near-perfect recovery of the uncontaminated signal in simulated MeerKAT data with up to 9 satellites, outperforming traditional flagging in imaging quality, source completeness, and flux accuracy, and enabling phase calibration that scales with RFI SNR. This approach offers a scalable, principled path toward mitigating pervasive RFI in current and future large arrays, with practical implications for maintaining survey speed and data quality in crowded radio skies.

Abstract

In the first TABASCAL paper we showed how to calibrate in the presence of Radio Frequency Interference (RFI) sources by simultaneously isolating the trajectories and signals of the RFI sources. Here we show that we can accurately remove RFI from simulated MeerKAT radio interferometry target data, for a single frequency channel, corrupted by up to 9 simultaneous satellites with average RFI amplitudes varying from weak to very strong (1 - 1000 Jy). Additionally, TABASCAL also manages to leverage the RFI signal-to-noise to phase calibrate the recovered astronomical signal. TABASCAL effectively performs a suitably phased up fringe filter for each RFI source which allows essentially perfect removal of RFI across all strengths. As a result, TABASCAL reaches image noises equivalent to the uncorrupted, no-RFI, case. Consequently, point-source science with TABASCAL almost matches the no-RFI case with near perfect completeness for all RFI amplitudes. In contrast the completeness of AOFlagger and idealised 3$σ$ flagging drops below 40% for strong RFI amplitudes where recovered flux errors are $\sim$10x-100x worse than those from TABASCAL. Finally we highlight that TABASCAL works for both static and varying astronomical sources.

Figures (11)

• Figure 1: Power spectrum prior used for our astronomical visibilities (orange). The faint black curves are the true calculated power spectrum for all baselines in one of our simulations. The blue curve shows the median power spectrum and the black dashed line shows the maximum expected fringe rate of the longest baseline in our simulation. The parameters for the example priors shown are $P_0=10^7$ Jy$^2$, $\eta_0=1$ mHz, and $\gamma=5$, 2, and 1.
• Figure 2: Antenna layout of the 32 MeerKAT antennas used in our simulations.
• Figure 3: Here we show some example predictions from tabascal for two different baselines (left and right) on an observation containing 6 satellites. The orange dotted curve shows the tabascal prediction and the blue curve shows the true value. The top panel shows this for the RFI visibility magnitude and the lower panels show the real part of the astronomical visibility. The top panel corresponds to a tabascal run where $\gamma=5$, as defined in Equation \ref{['eq:power_spec']}. The three lower panels show the results when varying the prior parameter $\gamma$ for the astronomical visibilities. $\gamma$ controls the smoothness of the solutions.
• Figure 4: The errors in the astronomical visibility predictions from tabascal and the other cases for comparison as the SNR of the RFI is varied. The black curve shows the distribution of the visibility noise used in the observations and therefore corresponds to the errors in the uncontaminated case. In each panel five bins of RFI strength are used where multiple observations are bundled together. The coloured curves show the Gaussian fit to the error distributions. (i) The top panel shows the errors in the tabascal predicted visibilities. (ii) The middle and bottom panels show the errors from the aoflagger and perfect $3\sigma$ flagging cases where flagged data is not included in the histograms.
• Figure 5: Images constructed from the same observation with our four different cases. Top Left: No RFI contamination. Top Right: tabascal fully removes the RFI and recovers the astronomical signal with comparable image noise to the uncontaminated data. Bottom Left and Right: The images from perfect $3\sigma$ flagging and aoflagger respectively, showing significant striping due to residual RFI contamination. The mean RFI SNR in this data was 1.8, showing that significant issues occur for traditional flagging methods even at weak RFI. Here perfect $3\sigma$ flagging means that any RFI with true amplitude greater than $3\times$ the noise is perfectly removed.