Abstract
The visibilities measured by radio astronomical interferometers include non-astronomical correlated signals that arise from the local environment of the array. These correlated signals are especially important in compact arrays such as those under development for 21 cm intensity mapping. The amplitudes of the contaminated visibilities can exceed the expected 21 cm signal and represent a significant systematic effect. We study the receiver noise radiated by antennas in compact arrays and develop a model for how it couples to other antennas. We apply the model to the Tianlai Dish Pathfinder Array (TDPA), a compact array of 16, 6-m dish antennas. The coupling model includes electromagnetic simulations, measurements with a network analyzer, and measurements of the noise of the receivers. We compare the model to drift-scan observations with the array and set requirements on the level of antenna cross-coupling for 21 cm intensity mapping instruments. We find that for the TDPA, cross-coupling would have to be reduced by several orders of magnitude in order to contribute negligibly to the visibilities.
Introduction
The goal of 21 cm intensity mapping is to create a 3D map of neutral hydrogen in the distant universe, providing a window into the "Cosmic Dawn." However, the signal is incredibly faint. To achieve the necessary sensitivity, astronomers use "compact arrays"—groups of antennas placed very close together. This proximity creates a significant systematic challenge: cross-coupling.
Every radio receiver generates its own internal thermal noise. In a compact array, some of this noise can travel backward out of the antenna, radiate through the air, and be picked up by a neighboring antenna. Because this noise is "correlated" (the same signal appearing in two places), the interferometer's correlator mistakes it for a real signal from the sky. This paper quantifies exactly how much this "fake" signal contaminates the data from the Tianlai Dish Pathfinder Array (TDPA).
Methodology
The researchers employed a three-pronged approach to characterize the coupling in the 16-dish Tianlai array: mathematical modeling, electromagnetic simulations, and field measurements.
Array Configuration and Modeling
The TDPA consists of 16 six-meter dish antennas arranged in a specific compact layout to maximize sensitivity to large-scale structures.
To describe the noise leakage, the authors used a "noise wave model." This model describes how noise travels forward into the electronics and backward out toward the dish, where it can be radiated and subsequently received by adjacent ports.
Simulations and Measurements
Using CST Studio Suite, the team performed massive electromagnetic simulations of the entire 16-dish array. This computationally intensive task required roughly 300 GB of RAM to simulate the dishes at various pointing angles, including Zenith and the North Celestial Pole (NCP).
Field measurements were conducted at the telescope site in Xinjiang, China, using a Vector Network Analyzer (VNA). These measurements provided direct hardware data on signal loss (S-parameters) between pairs of dishes. Finally, the model was validated against 210 hours of "drift-scan" observations collected in 2018.
Results and Discussion
The study found that cross-coupling in the TDPA is roughly -80 to -100 dB. While this sounds small, it results in a "fake" signal of about 10 mK, which is significantly stronger than the cosmological signals of interest.
Key Insights
- The "Shadowing" Effect: Coupling is not solely a function of distance. When dishes point toward the NCP (tilted at 45.1°), the physical structure of one dish can block the line-of-sight to another, reducing coupling compared to Zenith-pointing configurations.
- High Chromaticity: The cross-coupling signal is "chromatic," meaning it fluctuates rapidly across different radio frequencies. This is problematic because standard foreground removal methods rely on signals being spectrally smooth.
- The Delay Spectrum: Analysis of signal travel times (delays) between dishes revealed clear peaks corresponding to physical distances, as well as secondary peaks caused by waves bouncing off dish edges.
Conclusion
To make cross-coupling noise negligible for 21 cm cosmology, isolation must be improved to roughly -140 dB—several orders of magnitude beyond current performance. Furthermore, the model under-predicted total systematic noise, suggesting that ground pickup (thermal noise from the earth) is an even larger source of error. The authors conclude that mitigation must be achieved "by design" through better-shielded feeds or optimized geometries, as post-processing cleaning of these complex signals is likely impossible.
Summary updated