一种快速旋转器的VLSI硬件结构

VLSI architecture for fast rotators

计算复杂度太高极大地限制了“旋转算术”在数字信号处理领域中的应用，该文提出了快速实现Ｇｉｖｅｎｓ旋转运算的方法，与传统的Ｃｏｒｄｉｃ算法相比较，使“移位—加”运算的次数大大减少，从而为用“旋转算术”代替传统的“乘加算术”创造了条件。进一步给出了各种快速旋转运算的超大规模集成电路（ＶＬＳＩ）硬件结构，在此基础上发展出一种能够实现任意角度旋转的四级流水线快速旋转器，它可以在一类新型的使用Ｇｉｖｅｎｓ旋转序列实现各种正交变换的处理器中做为核心运算单元，使得其计算复杂度与传统的“乘加算术”可比。

High computing complexity greatly limits application of “rotation arithmetic” in digital signal processing. The methods for the fast realization of Givens rotation are presented. Compared with the traditional Cordic algorithm, they greatly decrease the number of “shift addition” so that it makes “rotation arithmetic” realistic to replace “multiplication addition arithmetic”. The very large scale integration (VLSI) architectures of various fast rotations are also given. Based on them, a four stage pipeline fast rotator is developed to realize the rotation of the arbitrary angle. It can be used as the core for the new VLSI DSP to perform various orthogonal transform by Givens rotation sequences. Its computing complexity can be comparable with the traditional “multipication addition arithmetic”.