FIR Filter Design Based on the OLS Method

Finite Impulse Response (FIR) filters, with linear phase and stability, are widely used in communications and biomedical applications. Traditional time-domain convolution faces high computational complexity O(NM) and resource conflicts in long-tap filters (e.g., >1,000 taps). This paper proposes an optimized Overlap-Save (OLS) method. MATLAB/Vivado simulations and Xilinx Zynq-7020 tests show that a 64th-order FIR design using OLS requires only 16 DSP units and 8 BRAM blocks, cutting hardware use by 75% and power by 33% compared to time-domain convolution, achieving 98 MSPS throughput. Frequency-domain block processing and invalid data truncation balance efficiency, cost, and fidelity, providing a cost-effective solution for real-time 5G and medical signal processing.