Table of Contents
Fetching ...

An Attempt to Search for Unintended Electromagnetic Radiation from Starlink Satellites with the 21 Centimeter Array: Methodology and RFI Characterization

Xupiao Yang, Qijun Zhi, Yanbin Yang, Quan Guo, Juhua Gu, Jianfeng Wang, Yan Huang, Yun Yu, Feiyu Zhao

Abstract

The rapid expansion of low-Earth-orbit (LEO) megaconstellations introduces new risks to radio astronomy from unintended electromagnetic radiation (UEMR). In this work, we present an attempt to search for UEMR from Starlink satellites using the 21 Centimeter Array (21CMA). Because the sensitivity of a single pod observation is limited, we focus on developing a robust observing and detection pipeline. Using Two-Line Element (TLE) data, we predict satellite transit times to guide the observations, and we define entry into the field of view (FoV) as an apparent declination greater than $85^{\circ}$ with respect to the 21CMA. We analyze the system equivalent flux density (SEFD) and the resulting single-pod sensitivity limits, which explain the detection of emission originating from the ORBCOMM satellites, rather than any detectable broadband UEMR in our dynamic spectra. To validate the methodology, we developed a Python package, orbdemod, to demodulate ORBCOMM downlink signals in our data. The recovered satellite ID agrees with the satellite predicted by our maximum-declination analysis, thereby validating the accuracy of our transit prediction and identification framework. Furthermore, via modulation power spectrum analysis, we show that the impulsive broadband bursts are produced by power line arcing near the array rather than by satellite UEMR.

An Attempt to Search for Unintended Electromagnetic Radiation from Starlink Satellites with the 21 Centimeter Array: Methodology and RFI Characterization

Abstract

The rapid expansion of low-Earth-orbit (LEO) megaconstellations introduces new risks to radio astronomy from unintended electromagnetic radiation (UEMR). In this work, we present an attempt to search for UEMR from Starlink satellites using the 21 Centimeter Array (21CMA). Because the sensitivity of a single pod observation is limited, we focus on developing a robust observing and detection pipeline. Using Two-Line Element (TLE) data, we predict satellite transit times to guide the observations, and we define entry into the field of view (FoV) as an apparent declination greater than with respect to the 21CMA. We analyze the system equivalent flux density (SEFD) and the resulting single-pod sensitivity limits, which explain the detection of emission originating from the ORBCOMM satellites, rather than any detectable broadband UEMR in our dynamic spectra. To validate the methodology, we developed a Python package, orbdemod, to demodulate ORBCOMM downlink signals in our data. The recovered satellite ID agrees with the satellite predicted by our maximum-declination analysis, thereby validating the accuracy of our transit prediction and identification framework. Furthermore, via modulation power spectrum analysis, we show that the impulsive broadband bursts are produced by power line arcing near the array rather than by satellite UEMR.
Paper Structure (8 sections, 3 equations, 6 figures, 2 tables)

This paper contains 8 sections, 3 equations, 6 figures, 2 tables.

Figures (6)

  • Figure 1: An example of the dynamic spectra of 21CMA single pod E8. The dynamic spectra is averaged by a factor of 1024 in the time domain, yielding a time resolution of $35~\mathrm{ms}$, and by 8 channels in the fr equency domain, resulting in a frequency resolution of $0.234~\mathrm{MHz}$. Within the frequency range of 110–188 MHz, there is no obvious broadband UEMR.
  • Figure 2: Several examples of power line arcing RFI. The horizontal axis spans 80–100 MHz with a frequency resolution of 29.3 kHz, while the vertical axis covers a duration of 100 ms with a time resolution of $34~\mu\text{s}$.
  • Figure 3: An example of modulation power spectrum of the integrated 60–120 MHz band. A dominant narrowband peak at $100~\text{Hz}$ and a secondary harmonic at $50~\text{Hz}$ are clearly identified. These features are characteristic fingerprints of power line arcing. .
  • Figure 4: Examples of successful recognitions of power line arcing in spectra sequences using SAM 2. For a more detailed exhibition, please refer to our video.
  • Figure 5: Dynamic spectrum recorded between 11:39:52 and 11:39:53 (UTC) on 2025-04-15. A cluster of intense narrowband RFI is clearly visible near $137~\text{MHz}$, identified as downlink transmissions from ORBCOMM satellites.
  • ...and 1 more figures