Poster Presentation

Research on Channelization Techniques Based on Oversampled Polyphase Filter Bank for Fast Radio Burst Detection: Implementation on Xilinx RFSoC 4x2 Platform

Author(s): Ke-Jyun Li (National Ilan University), Ying-Ren Chien (National Taipei University of Technology)

Presenter: Ke-Jyun Li (National Ilan University)

Fast Radio Bursts (FRBs), highly energetic radio transients of millisecond duration originating from extragalactic sources, necessitate digital channelization techniques capable of wideband operation, high temporal resolution, and fine frequency granularity to enable reliable detection and analysis. This study introduces a 1,600 MHz real-valued input, 256-branch, 4-tap, 2× oversampled polyphase channelizer that achieves spectral image suppression exceeding −60 dB. Each coarse channel is further subdivided into 512 finer sub-channels through zoom-in processing. The proposed channelizer is implemented on the Xilinx RFSoC 4x2 platform utilizing MATLAB/Simulink and DSP IP cores provided by the CASPER collaboration, with testing conducted via a PYNQ overlay. Compared to conventional critically sampled polyphase filter banks, the oversampled design significantly mitigates signal attenuation at sub-band edges, suppresses aliasing artifacts, and reduces spectral leakage across sub-bands. By extracting and recombining the central 50% bandwidth of each sub-band, the system constructs a flat spectrum comprising 32,768 frequency channels with a spectral resolution of 24.4 kHz. This advancement enhances the performance of real-time FRB detection systems by improving spectral fidelity and analytical precision.