TransientX: A high performance single pulse search package (2401.13834v1)

Abstract: Radio interferometers composed of a large array of small antennas posses large fields of view, coupled with high sensitivities. For example, the Karoo Array Telescope (MeerKAT), achieves a gain of up to 2.8 K/Jy across its $>1\,\mathrm{deg}^2$ field of view. This capability significantly enhances the survey speed for pulsars and fast transients. Nevertheless, this also introduces challenges related to the high data rate, reaching a few Tb/s for MeerKAT, and substantial computing power requirements. To handle the large data rate of surveys, we have developed a high-performance single-pulse search software called "TransientX". This software integrates multiple processes into one pipeline, which includes radio frequency interference mitigation, de-dispersion, matched filtering, clustering, and candidate plotting. In TransientX, we have developed an efficient CPU-based de-dispersion implementation using the sub-band de-dispersion algorithm. Additionally, TransientX employs the density-based spatial clustering of applications with noise (DBSCAN) algorithm to eliminate duplicate candidates, utilizing an efficient implementation based on the kd-tree data structure. We also calculate the signal-to-noise ratio loss resulting from dispersion measure, boxcar width, spectral index and pulse shape mismatches. Remarkably, we find that the signal-to-noise ratio loss resulting from the mismatch between a boxcar-shaped template and a Gaussian-shaped pulse with scattering remains relatively small, at approximately 9%, even when the scattering timescale is 10 times that of the pulse width. Additionally, the S/N decrease resulting from the spectra index mismatch becomes significant with multi-octave receivers. We have benchmarked the individual processes, including de-dispersion, matched filtering, and clustering. TransientX offers the capability for efficient CPU-only real-time single pulse searching.