A high-speed processor for finely-spaced Fourier transform via chirp z-transform

Abstract

This paper deals with the design of a processor for very-finely spaced spectral analysis over a narrow band of the available spectrum. The processor implements the chirp z-transform (CZT) algorithm, and exploits a fully-parallel architecture in order to address real-time applications with very-high throughput. The internal data-path is optimized as a trade-off between fixed-point accuracy and implementation complexity. The proposed architecture has been customized for the case study of a 64-point transform in a sub-band of 10% of the available spectrum, and has reached the astonishing throughput of 3.2 Gs/s on a Xilinx Virtex-IV FPGA. Also, compared with a customary approach based on FFT, a remarkable saving in complexity is shown.