论文部分内容阅读
提出了一种基于时间抽取原位计算的高效并行的二维矢量基2×2快速傅里叶变换的硬件实现结构.该算法结构将N×N点数据分解为4个独立存储的部分来实现矢量基2×2蝶形计算单元4个操作数的并行访问,仅用一个二维分裂基蝶形运算单元对这4块数据进行二维矢量基快速傅里叶变换,利用无冲突访问方法完成对存储器的并行访问.推导出了该算法硬件实现结构下的各存储器数据地址存取公式和旋转因子的产生方法,并利用CORDIC算法实现旋转因子的产生来减少存储器的使用.该算法对N×N点数据进行二维离散傅里叶变换处理的时间仅为(N2/2)(lb N-1)个时钟周期,与以往算法计算时间的比较结果表明了该设计的有效性.
A hardware implementation of two-dimensional vector-based 2 × 2 fast Fourier transform with efficient parallelization based on time-in-place calculation is proposed.The structure of this algorithm is realized by decomposing N × N points into four independent parts Vector-based 2 × 2 butterfly computing unit 4 operands parallel access, only a two-dimensional split butterfly operation unit of these four data two-dimensional vector-based Fast Fourier Transform, the use of conflict-free access method to complete And accesses to the memory in parallel.It deduces the memory data address access formula and twiddle factor generation method under the hardware implementation structure of the algorithm and uses the CORDIC algorithm to reduce the memory usage by using the twiddle factor generation algorithm.The algorithm reduces the memory size of Nx The N-point data is only (N2 / 2) (lb N-1) clock cycles for two-dimensional discrete Fourier transform. The comparison with the previous algorithm shows the validity of the design.