一种基于Schur分解数字滤波器有限精度实现的高效结构  

A Schur-form based efficient structure for finite precision implementation of digital filters

作　　者：朱广信[1] 王匡[1] 徐红[2] 

机构地区：[1]浙江大学信息与电子工程系,浙江杭州310027 [2]浙江工业大学信息工程学院,浙江杭州310032

出　　处：《电路与系统学报》2010年第2期1-6,共6页Journal of Circuits and Systems

摘　　要：本文基于Schur分解提出了一种用于数字滤波器正规实现的高效结构。对于一个N阶滤波器,该结构中所需的乘法器数量约为3.5N。本文对该结构的性能进行了分析,证明采用该结构的滤波器能避免产生自激振荡,并推导得到相应的舍入噪声增益表达式。相对于一些熟知滤波器结构,本文所提出的稀疏结构在最小化有限精度影响和降低实现复杂度方面具有更优越的性能,文章通过两个仿真实例对该结构设计方法及性能作了说明。In this paper, an efficient structure based on a Schur-form normal realizations is derived for digital filters. For an Nth order filter, the number of multiplies in such a structure is of an order about 3.5N. It is shown that the proposed structure is free of self-sustained oscillations. The corresponding expression of roundoff noise gain is obtained. Two design examples are presented to demonstrate the superior performance of the proposed sparse structure to several well-known realizations in terms of minimizing the finite precision effects that ultimately can not be avoided in real-time applications and reducing filter implementation complexity.

关 键 词：数字滤波器 状态空间实现 正规实现 有限字长效应 

分 类 号：TN911.72[电子电信—通信与信息系统]