Fine-tuning a narrowband SETI signal processing pipeline

A narrowband radio frequency Search for Extraterrestrial Intelligence (SETI) presumes a long-duration ($\approx$minutes), low bandwidth ($\approx \mathrm{Hz}$) beacon or incidental transmitted signal that would be unlikely to occur in the natural world: a technosignature. A constant-frequency signal at the transmitter will drift in frequency on reception due to relative accelerations between the transmitter and receiver. Drift rates will be proportional to center frequency, e.g. up to $\pm$0.44 Hz/s/GHz (Li et al. 2023). The typical signal processing pipeline involves creation of spectrograms over a wide bandwidth (up to 1 GHz) with Hz-level frequency resolution, followed by De-Doppler integration, which integrates energy over linearly-drifting tracks in the time–frequency plane.

Search capabilities have greatly expanded over recent years to include interferometric radio telescopes and aperture arrays (MeerKAT, VLA, MWA), commensal observing, and GPU-augmented server racks. However, the key detection algorithms in the dominant SETI search code used by Breakthrough Listen (BL) researchers have not been updated in spite of many works describing shortcomings. While updates by BL are in progress, it is important to review key design issues to be sure they are addressed.

In this paper, improvements to the BL algorithms and their implementation are discussed. Topics include: (1) an improved spectrogram normalization method for accurate threshold determination, (2) reducing spectrogram compression to avoid sensitivity losses, and (3) comparison of the commonly-used Taylor De-Doppler method to the fastDD algorithm (Houston, 2023). Reduced SNR losses on the order of 4–8 dB should be possible at high drift rates by optimizing averaging parameters. In addition, potential future post-detection pipeline enhancements are described, such as radio-frequency interference rejection, direction-of-arrival estimation, and waveform extraction.