GPU-Based Implementation of Small sizes DFT for Polyphase Channelizer
Amean Al-Safi · 2024
Polyphase filter bank channelizers are very important parts in different communication systems. They are responsible for signal separation and sample rate conversion. Efficient implementation of these subsystems depends heavily on the implementation of the Fast Fourier Transform (FFT)/ Discrete Fourier Transform (DFT) part. This paper presented a Graphics Processing Unit (GPU) based implementation of DFT using matrix-vector multiplication concept on different GPU platforms. The obtained results indicate a computation speed up of 9-12000 and 9-10000 compared to CUFFT library on Nvidia GeForce GT520M and Quadro 600 respectively within the DFT ranges of the polyphase filter bank channelizer.