w[n] I(B) 0 M Kaiser Window Given a set of filter specifications,the values of M and B needed can be determined numerically. As in OSB Figure 7.23,define: Transition bandwidth:Aw =ws-wp Peak approximation error:6 (passband:1+6 stopband:+6) Stopband attenuation:A =-20log10 6 Then the values of B,M needed to achieve A are 0.1102(A-8.7) A>50 0.5842(A-21)0.4+0.07886(A-21)21≤A≤50, A-8 3= 0 A<21 M22.285AW See OSB Section 7.3 for examples of the Kaiser window method. Optimum FIR Filter Approximation: The Parks-McClellan Algorithm Although the rectangular windowing method provides the best mean-square approximation to a desired frequency response,just because it is optimal does not mean it is good.FIR filters designed with windows exhibit oscillatory behavior around the discontinuity of the ideal fre- quency response and does not allow separate control of the passband and stopband ripples.An alternative FIR design technique is the Parks-McClellan algorithm which is based on polynomial approximations. Filter Design as Polynomial Approximation Consider the DTFT of a causal FIR system of length M+1: M M H(eu)=∑hmle-m=∑hinl"I=e-扣 n=0 3Kaiser Window Given a set of filter specifications, the values of M and β needed can be determined numerically. As in OSB Figure 7.23, define: Transition bandwidth: Δω = ωs − ωp Peak approximation error: δ (passband: 1 ± δ stopband: ± δ) Stopband attenuation: A = −20 log10 δ Then the values of β, M needed to achieve A are β = ⎧ ⎨ ⎩ 0.1102(A − 8.7) A > 50 0.5842(A − 21)0.4 + 0.07886(A − 21) 21 ≤ A ≤ 50 , A − 8 M ≥ 2.285Δω A < 21 . 0 See OSB Section 7.3 for examples of the Kaiser window method. Optimum FIR Filter Approximation: The Parks-McClellan Algorithm Although the rectangular windowing method provides the best mean-square approximation to a desired frequency response, just because it is optimal does not mean it is good. FIR filters designed with windows exhibit oscillatory behavior around the discontinuity of the ideal frequency response and does not allow separate control of the passband and stopband ripples. An alternative FIR design technique is the Parks-McClellan algorithm which is based on polynomial approximations. Filter Design as Polynomial Approximation Consider the DTFT of a causal FIR system of length M + 1: � M h e−jωn [n] = � M H(ejω) = h n [n]x |x=e−jω . n=0 n=0 3