Signal Processing and Linear Systems-B.P.Lathi copy

6 r ecursive f ilter d esign t he b ilinear t

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 29 7!- tan ( ¥s) = 8.3623 log[(10Ll - 1)/(1002 - I)J = 1.9405 2Iog(17.7696/8.3623) W e' H[zJ S2 In t he second step, we design a Butterworth filter with critical frequencies w/ = 8.3623 a nd ws' = 17.7696 with Op = - 2 dB and O. = - 11 dB. The value of n (order of t he filter) is obtained from Eq. (7.39): n= F ig. 1 2.14 A mplitude response of t he filter in Example 12.6. cies we' = Wi are transformed (prewarped) by the equation , 2 WiT i=1,2,···,m Wi = T tan - 2- ( 12.63a) T he p rewarped c utoff frequency we', d etermined b y u sing p rewarped c ritical frequencies, is u sed t o f ind t he p rewarped a nalog f ilter t ransfer f unction H a(s). F inally, we r eplace s w ith 1. z +1 i n H a(s) t o o btain t he d esired d igital f ilter t ransfer -1 Tz f unction H[z] ( 12.63b) - ----.., 12 Frequency R esponse a nd D igital F ilters 748 12.6 Recursive Filter design: T he Bilinear Transformation Method 749 w hich is identical t o t he r esult o btained e arlier. o A Simplified Procedure T he a bove p rocedure c an b e s implified by observing t hat t he s caling f actor ~ is irrelevant i n t his m anipulation a nd c an b e i gnored. I nstead o f u sing Eqs. (12.63a) a nd ( 12.63b), we c an u se t he s implified e quations w/ = w iT t an - - a nd (12.64a) i = 1 ,2,···, m 2 z -1 z+1 (12.64b) s=-- C omputer E xample C 12.5 Design a lowpass digital filter to meet the specifications in Example 12.6, using bilinear transformation with prewarping. We shall give here MATLAB functions to design the four types of approximations: Butterworth, Chebyshev, inverse Chebyshev, and elliptic. The input d ata asks for frequencies so normalized t hat t he sampling radian frequency is 2. This requirement means the sampling radian frequency, which is 2 n/T, m ust be normalized to 2. Therefore, all t he radian frequencies can be normalized by multiplying each of them by T /n. I n t he present case, T = n /35 so t hat T /n = 1 /35. Thus, to normalize any radian frequency, we j ust divide it by 35. T he normalized Wp and Ws a re 8 /35 a nd 1 5/35, respectively. W p=8/35;Ws=15/35;Gp=-2;Gs=-11;T=pi/35; T his s implification works b ecause t he f actor 2 /T i n E q. (12.63a) is a f requency s caling f actor, a nd i gnoring i t i n E q. (12.64a) r esults i n t he p retransformed f ilter t hat i s s caled by a factor 2 /T i n t he frequency scale. T his s caling is u ndone b y using Eq. (12.64b) i nstead o f E q. ( 12.63b). T o d emonstrate t he p rocedure, we shall redo E xample 12.6 using t his simplification. I n t he first s tep, we p rewarp t he c ritical frequencies W p a nd W s a ccording t o E q. ( 12.64a): wp' = t an ws' = t an % B uttervorth [ n,Wn)=buttord(Wp,Ws,-Gp,-Gs); [ b,a)=butter(n,Wn) %Chebyshev [ n,Wn)=cheblord(Wp,Ws,-Gp,-Gs); [ b,a)=chebyl(n,-Gp,Wn) ¥ = t an ( ¥s) = 0 .3753 % I nverse C hebyshev ¥ = t ane;;) = 0.7975 [ n,Wn)=cheb2ord(Wp,Ws,-Gp,-Gs); [ b,a)=cheby2(n,-Gs,Wn) I n t he s econd s tep, we design a B utterworth f ilter w ith c ritical frequencies a nd ws' = 0.79...
View Full Document

Ask a homework question - tutors are online