离散时间信号处理（英文版）chap7-第7讲.pptVIP

7.4 Optimum Approximations of FIR Filters Properties of the optimal type I LPF 7.4.3 The Parks-McClellan Algorithm (Remez) Steps of Steps of designing FIR filter by Chebyshev approximation 7.4.4 Characteristics of Optimum FIR filters 7.6 Comments on IIR and FIR D-T Filters minimum mean-square error criterion minimax criterion (Chebyshev equiripple approximation ) Optimum criterion Minimum mean-square error criterion constant So, The rectangular windowing result must satisfy the minimum mean-square error criterion error function: mean-square error : minimize: Minimax criterion Weighted error function: weighting function Minimax criterion: (not include transition band) —— equiripple —— Parks-McClellan algorithm are fixed, adjust Reformulate the filter design problem as a problem in polynomial approximation. The Theory of the Chebyshev Equiripple Approximation Type I linear-phase FIR filter: Lth-order polynomial Let nth-order Chebyshev polynomial Weighted approximation error: Minimax criterion: So, Alternation Theorem (交错定理) Let i.e., such that A necessary and sufficient condition that be the unique Lth-order polynomial that minimizes is: exhibit at least (L+2) alternations; there must exist at least (L+2) values and , 7.4.1 Optimal Type I Lowpass Filters —— equiripple approximation (L=7) (L+2)=9 alternations (L+2)=9 alternations —— equiripple approximation If alternations = (L+2) If alternations = (L+3) (L+3) =10 alternations —— extraripple approximation Note: the case of (L+3) alternations is true only when an alternation occurs at all band edges, that is, include 0 and The maximum possible number of alternations of the error is (L+3). (ii) Alternations will always occur at and (iii) All points with zero slope inside the passband and the stopband ( )will be alternations; i.e., the filter will be equiripple, except possi