Ultra-long-point high-performance FFT calculation device

Abstract

The invention relates to the field of digital signal processing, in particular to an ultra-long-point ultra-high-performance FFT calculation device. The ultra-long-point ultra-high-performance FFT calculation device is characterized by comprising two FFT calculation modules, a control unit, two three-dimensional transposed memories and a two-dimensional transposed memory, wherein the two FFT calculation modules are the calculation module I and the calculation module II respectively; a data input end of the three-dimensional transposed memory I is connected with a main memory, and a data outputend of the same is connected with the calculation module I; and a data input end of the two-dimensional transposed memory is connected with the calculation module I, and a data output end of the two-dimensional transposed memory is connected with the calculation module II. According to the ultra-long-point ultra-high-performance FFT calculation device, data smoothness is improved through a brand-new data transposition processing mode, so that the data bandwidth utilization rate in the FFT data processing process is high, and the processing efficiency is greatly improved.

Description

technical field [0001] The invention relates to the field of digital signal processing, in particular to a high-performance FFT computing device with super long points. Background technique [0002] Discrete Fourier Transform (DFT) is one of the core technologies of digital signal processing, widely used in digital communication, speech processing, image processing, spectrum estimation, simulation, system analysis, radar, optics, medical imaging, astronomy and other application fields. The fast Fourier transform (Fast Fourier Transform, FFT) is based on the DFT, using the coefficient characteristics of the DFT, merging some items in the DFT operation, and changing the long sequence DFT into a short sequence DFT, thereby reducing its calculation volume. FFT is the most common digital signal processing algorithm, plays an important role in various digital signal processing systems, and is the basic operation of time domain and frequency domain conversion. For example, in el...

AI Technical Summary

Problems solved by technology

[0006] First of all, even with the two-dimensional cascade processing of one-dimensional FFT, the number of points that can be processed is still limited

[0007] Secondly, when the number of FFT points exceeds the processing capacity of the computing device, multiple reads and writes of data involve a large number of buffers and interconnect structures, and the efficiency is greatly impaired

[0008] Third, the rotation factor and the hinge factor need to be used in this process, and the generation of the rotation factor and the hinge factor needs to use the CORDIC algorithm, that is, the coordinate rotation digital calculation method. In this process, the area overhead and the circuit delay are relatively large.

Images