A Comparative Study between FPGA and DSP for the Efficient Implementation of FIR Filters

In recent years, field-programmable gate arrays (FPGAs) have achieved a higher performance, higher number of gates, and lower power consumption. Therefore, potential applications of FPGAs have increased, including digital signal processing, which is traditionally performed using digital signal processors (DSPs). This study aims to reveal the advantages of using FPGAs for real-time digital signal processing systems over conventional DSPs. We implemented finite impulse response (FIR) filters with three different lengths to represent applications requiring different levels of filtering precision. These filters were implemented on a low-power Intel Cyclone 10 LP FPGA, large-scale Intel Stratix V GX FPGA, and Texas Instruments TMS320C6747 DSP. To further enrich the comparison, the filters were implemented on an AMD Zen 3 processor. For each platform listed above, we compared the output latency and power consumption data. The results showed that, in the best case, the implementation on FPGA achieved 1/500 the output latency and 1/2 the power consumption of that on DSP.