Optimising the FRB Search Pipeline for the Northern Cross Radio Telescope

Abstract

FRB search pipelines are being developed to operate under strict real-time constraints while maintaining sensitivity to short-duration transient signals. In incoherent dedispersion based pipelines such as Heimdall, apart from observation bandwidth and number of beams, detection performance and computational throughput are strongly dependent on the choice of processing parameters, which are often selected heuristically. In this work, we present a systematic evaluation of key dedispersion and matched filtering parameters and quantify their impact on both detection accuracy and runtime performance. A controlled synthetic injection framework is developed in which artificial FRB pulses with known DMs, SNRs, and pulse widths are embedded into realistic filterbank data containing instrumental noise representative of observations from the Northern Cross radio telescope. Using this framework, a grid of Heimdall configurations is explored, spanning DM tolerance, boxcar filter width, and processing gulp size. Detection performance is assessed by comparing recovered and injected signal properties, while computational performance is evaluated through end-to-end processing time measurements. The results reveal clear trade-offs between sensitivity and throughput across parameter choices. We identify an empirically optimal configuration that provides burst recovery while maintaining processing speeds exceeding real-time requirements. While the specific optimal parameters are derived for the Northern Cross, the methodology and findings are broadly applicable to any real-time transient detection pipeline employing matched-filtering and dedispersion, and are particularly relevant for low-frequency radio telescopes with similar observing configurations. These findings demonstrate the value of data-driven parameter evaluation for improving the performance of real-time transient detection pipelines.

Figures (8)

• Figure 1: Pulse profiles for PSR B1831-03 observed at five different frequencies with the Lovell telescope and the GMRT, clearly showing the increasing effect of scattering at lower frequencies. The solid lines show exponential fits to the data. Figure extracted from lorimer05.
• Figure 2: Top view of the Northern Cross radio telescope with the two perpendicular arms along the East-West and North-South direction. Figure extracted from debarro25.
• Figure 3: Flow chart of the key processing operations in the pipeline. Heimdall is the name of the main GPU-based pipeline implementation. Adapted from barsdell12.
• Figure 4: Impact of the dm_tol parameter on DM trial spacing and recovery accuracy for Heimdall dedispersion. The left panel shows the local DM trial step size ($\Delta$DM) as a function of dispersion measure, with red dashed lines marking the injected pulse DM positions. The right panel shows the mean absolute offset between detected and injected DM values $(\pm 1\sigma)$ for each dm_tol setting, where all values successfully recover injections but with varying accuracy, demonstrating that parameter choice affects signal recovery fidelity rather than detection alone. Results shown are from a representative file drawn from the test dataset; the trends are consistent across the full dataset.
• Figure 5: 2-D projection of t-SNE dimensionality reduction on the data; where the top figure is overlayed with a heatmap representing percentage accuracy for SNR (cool colours = lower accuracy; warm colours = higher accuracy, up to 100%) and the bottom figure is overlayed with colours which indicate cluster labels returned by HDBSCAN; label –1 marks points classified as noise/outliers. Each point represents a parameter-file outcome embedded into two dimensions by t-SNE (axes are unitless and not directly interpretable). High-density regions indicate many outcomes with very similar feature profiles (locally preserved neighbourhoods), i.e., parameter combinations that obtained similar performance and results.