DESIGN AND ANALYSIS OF INFINITE IMPULSE RESPONSE wry orerra.rutens__\ 455 \ 2, Specifications Required Inany filter design, following specifications are (i) The maximum tolerable passhy Gi) The maximum tolerable stoph, (ii) Passband edge frequency («,) Gv) Stopband edge frequency () required: andl ripple (6) rand ripple (6) 7.6 IR FILTER DESIGN FROM CONTINUOUS.TIME FILTERS As a matter of fact, the commonly used tech ce ques for the design of HR filters are based on transformations of continuous-time IIR systems into diserote-time IIR systems, "This is partly because continuous-time filter design was a highly advanced art before diserote-time filters were of interest and partly because of the difficulty of implementing a non-iterative direct design method for IIR. filters. In contrast, FIR filters are almost entirely restricted to discrete-time implementations. Consequently, the design techniques for FIR filters are based on directly approximating the desired frequency response of the discrete-time system. Furth rmore, most techniques for approximating the magnitude response of an FIR system assume a linear phase constraint, thereby avoiding the problem of spectrum factorization that complicates the direct design of IIR filters. The simplest method of FIR filter design is called the window method. This method generally begins with an ideal desired frequency response. As discussed earlier, in order to design the digital IIR filter, first analog IIR filter is designed. Then analog filter is converted into the digital fi Iter because of following two reasons: (i) The procedure to design analog filter is readily available and it is highly advanced. (ii) When we design digital filter using analog filter then the implementation becomes simple. 7.7 IMPULSE INVARIANT TRANSFORMATION: IMPULSE INVARIANT METHOD ]———— Eee In this method, the design starts from the specifications of analog filter. Here, we have to replace analog filter by digital filter. This is achieved if impulse response of digital filter resembles the sampled version of impulse response of analog filter. If impulse response of both, analog and digital filter matches then, both filters perform in a similar manner. In this method, we shall use the following different notations: h,(t) = Impulse response in time domain H,(s) = Transfer function of analog filter, here ‘s’ is Laplace operator h,(nT) = Sampled version of h,(t), obtained by replacing t by nT H(2) = z-transform of h(n). This is response of digital filter. = Analog frequency = Digital frequency 7.7.1 Transformation of Analog System Function H,(s) to Digital System Function H(z) Now, let the system transfer function of analog filter be H,(s). We can express H,(s) in terms of Partial fraction expansion. This means that: Therefore, wd 712) Here, A;= Aj, Ag... Ayy are the coefficients of partial fraction expansion.DIGITAL SIGNAL PROCESSING Phy Py «= Py are the poles. Here, ‘s’ is the Laplace operator. Hence, we can obtain impulse response of analo, e from H,(s)by taking inverse Laplace transform of H,(s). Therefore, using standard relate ier, hy Laplace transform, we obtain invey, N hw= YA e™ f a My} Now, unit impulse response for discrete structure is obtained by sampling hy(t). This mp. _ that h(n) can be obtained from h,(t) by replacing t by nT in equation (7.13). Mea | x, . nin? Thus. hin) = 2A oe ay Here, Tis the sampling time. The system transfer function of digital filter is denoted by H(z)". It is obtained by taking transform of hin). According to the definition of Z-transform for causal system, we have H@= Yhm)z* a0 Substituting equation (7.14) in equation (7.15), we get H()= x [3 on] azo Lisl AT No YA; Serat ze n=O This is the required transfer function of digital filter. ti Thus, comparing equations (7.12) and (7.17), we can say that the transfer function aie filter is obtained from the transfer function of analog filter by doing the following transform 18 7.7.2 Mapping of Poles | Equation (7.18) shows how the poles from analog domain are transferred into the digital do This transformation of poles is called as mapping of poles. BP eS ee —— ee F Basically, The art of continucus-Cimo IIR filter design is highly advanced and since useful resul Ire he achieved, its advantageous to-use the design procedures already developed foreontnue filters.DESIGN AND ANALYSIS OF INFINITE IMPULSE RESPONSE (IIR) DIGITAL FILTERS ‘We know that the poles of analog filters are located at s = p,, Ni ion (7.18), we can cay thatthe poles of digital filter, H(2) are located at, lad _- (7.19) ‘This equation eee that the poles of analog filter at s = p, are transformed into the poles of, digital filter’ atz= e?! T. Thus, the relationship between Laplace (s domain) and z-domain is given by, z= eT (7.20) Here, = p,and Tis the sampling time. Now, s is the Laplace operator and it is expressed as, s= 0 +jQ wAT21) Here, = attenuation factor and = analog frequency We know that z can be expressed in polar form as under: z= rele (7.22) Here, ris magnitude and ‘o’ is the digital frequency. Substituting equations (7.21) and (7.22) in equation (7.20), we obtain rel = @f0+i0/T — goT, git 2 AT.23) Separating real and imaginary parts of equation (7.23), we obtain reo wlT.24) and ei = oat ‘Thus, we have = OT (7.25) Now, we will find the relationship between s-plane and z-plane. Basically, plot in ‘s'-domain means that o is plotted on X-axis and j is plotted on Y-axis. Also, z-domain representation means that real z is plotted on X-axis and imaginary z is plotted on Y-axis. Let us consider equation (7.24), i.e., r=eT we shall discuss the following conditions: i) Ifo < 0, then r is equal to reciprocal of e raise to some constant. Hence, range of r will be Otolic,, o<050 0 then, r is equal to e raise to some constant. That means r > Li.e., o>0sr>1 Now, o > 0 indicates R.H.S. of s plane and r > 1 indicates exterior part of unit circle. Thus, RLS. of s-plane is mapped outside the unit circle.DIGITAL SIGNAL PROCESSING Combining all the above conditions, this mapping is shown in figure 7.10. I(2) jo Re(2) Fig. 7.10. Illustration of the mapping z = e°T. 7.7.4 Drawbacks of Impulse Invariance Method* x We know that @ is analog frequency and its range is from 7 to -7 While the digit rz frequency w varies from ~—r ton. This means that from 7 to ~ 7 ,@maps from —nton. Let i be any integer. Then, we can write the general range of @ as (i ~ bt to G+) £ However, for this range also, maps from ~ x to x. Hence, mapping from analog frequency Q to digital frequency w is termed as many-to-one. This mapping is not one-to-one. (Gi) Analog filters are not band limited so there will be aliasing due to the sampling process. Because of this aliasing, the frequency response of resulting digital filter will not be identical to the original frequency response of analog filter. (iii), The change in the value of sampling time (T) has no effect on the amount of aliasing. Standard Expressions Some standard formulae for transformation in impulse invariance method are as under: ~T eosbT] 2 = -aT, Tz) + e28t. 1-2e*" [eosbT]z"! +e Z 1-2e*" feosbT]z! +e 28", 7.7.6 Design Steps for Impulse Invariance Method not given, then, we obtain expression of H,(s) from the given specifications. (ii) If required, we expand H,(s) by using partial fraction expansion (PFE). ; (iii) Then, we obtain z-transform of each PFE term using impulse invariance transformatio? equation. : (iv) We obtain H(z), this is required digital IIR filter. Te The tradi to the design of discrete-time IIR filters involves: | } t | G) In numerical problems, analog frequency transfer function H,(s) is usually be given. Ifitis | f Te The nal appro’ ‘Valves the transformation of # continuous-time filter into a discrete-time filter meeting prescribed specifications. |DESIGN AND ANALYSIS OF INFINITE IMPULSE RESPONSE (IR) DIGITAL FILTERS. qal solved Examples Now, let us consider fow solved examples to illustrate the concept of impulse invariance method. £7.14 Determine H(z) using impul 1 tin Le pulse Invariance method at 5 Hz sampling frequency from tT? 2 (s+ 1)(542) Given analog transfer function is, Ti(s) = solution: ASS (ee tise 2) “id yet us expand H,(s) using partial fraction expansion, ic A As (set * HQ) Thus, we find that poles are at py =~ 1 and py H, Hts (i) Now, values of A, and A, can be calculated as under: 2 =P), <6 +). Gerd let Therefore, 35 . Also, Ay= (S~p,) Hls)|,. 9) =(5+2) Therefore, = —2 [aH Substituting values of A, and A, in equation (ii), we obtain 2 G+) +2) Now, let us obtain the z-transform using impulse invariance transformation equation. It is given by = . a sali) Here, T = Sampling time. Now, given sampling frequency is H,(s)= He, Also T = 0.2sec. We have poles at p, =~ 1 and p Hence, using equation (iv), we obtain oe oA) eel and _ — vi) a ive The transfer function of digital filter is given by x H@= Lyra it In this case, we obtain H@)DIGITAL SIGNAL PROCESSING Using equations (v) and (vi), ot 2 rr! 1 -OKIKe' 1-067! To convert each term into positive powors of y, let un multiply eneh torm by x to got 2 dy al ONT) a ONIN) ORIN 7-007 2 ORIRY- 0.07) Hi) Nir) = De? te 22" 100, 0.297 or Nay % “3 P0072 OI 0.61 2 1ABKZ 1 0.64 inital HR filter, Anu. This is the required transfer funetion for EXAMPLE 7.2 Determine H(2) uxing impulse invarinnce mothod for the following ayntem function, 1 (s4.0.5)(8" 4 0.684 2) Solution: The given transfer function is, Hs) = Hs) = y at (64 O.5)(82 4 0.58.4-2) af In the partial fraction expansion form, 11,(s) ean be written as und 1 A Bur BO) omit otne2) 408) * GE r0.552) m Let us obtain the values of A, B and C. 1 _ Als? 40.58 42) +(Bs + C)(8+0.5) (5+0.5)(62 40.5842) (84 0.5)(s" + 0.5842) or Als? + 0.58 + 2) + (Bs + Ci(s +0.5) = 1 or As? +A0.58 + 2A + Bs? + B0.68+ Cs +0.5 C= 1 4A +B) +8(0.5A+0.5B+C)+(2A+05C)= 1 ‘i Now, s? term is absent in RHS, therefore, we have A+Bz0 Aix) Similarly, s term is absent in RHS, thus, we have 0.5A+05B4+C=0 (w) and 2A+05 Now, let us shall solve equations (iv), (v) and (vi) to obtain the values of A, B, C. From equation (iv), we bart Be-A Substituting this value in equation (v), we obtain O.5A-05A+C= 0 or ce0 From equation (vi), we have 2A+0= 1 or Az 05 Since B =~ A, we obtain B Substituting these values in equation (ii), we shall have = 05 H(o= O5____0.58__ Avid 840.5 40.5942erect DESIGN AND ANALYSIS OF 1 INFINITE IMPULSE RESPONSE (IIR) 8 sna —Y— 8, we shall convert the second term on RLS. in the form |) goatee (stay+b?* . i Lat us consider the term s? + 0.5 +2. 1tean be expressed as und i F058 +2 (4 058 + 0.0625) 6 (L097) i cr) $740.55 +2 = (8+ 0.28) + (1.39)? i Substituting this value in equation (vi), we got 0.5 0. 7 fee os 0250.25 yor 5 Oe fn as-o.5 40.5 7 (640.25) (L992 > S008 [canta | . a (540.2 9 Lis v0.a5) + a.a9" Now, we want the numerator of third term equal to be 1.39. It can be arranged as under, Hoe -os{ seas 28K095 [1.89 -| (s+0.25)" + 1.39)" 1.39 [(9+0.25)* +(1.39)" } 0. 840.25 Hys)= -05| > _. — : or LS) = 05 of eosseape | some eats 397 7 Let us recall the standard transformation formulae, ie, 1 1 @® —>>5r sp 1-eR Tet sta, -T oosbT] 2” (stay+b? 1-2 [cosbT]z* +e~ eT fsinbT]z 1-26" feosbT]z™ +e Using these formulae, equation of H(z) can be obtained from equation (viii) as follows: 0257 feo81.39T]2 1-2¢ 5 [e9s 1.39T]z 1 +e"* -05 +0.089 eT eos 1.39T)2t +e MT 2 kept Ans. This is the required transfer function. Note that the value of Tis not given in the problem, hence, Tis asitis, EXAMPLE 7.3, The system function of an analog filter is expressed as follows: +02 HO) = 640.2)" +9 , a / convert this analog filter into a digital filter by making use of the impulse invariant techniques, sume T= 1s, ; Solution: Here note that the given system response of the analog filter is ofthe standard form sta HA) = Gray+ 3, We can obtain system resp Where we are given a = 0.2 and b= jonse of the digital filter as under: na =0: = q-e*"cosb) 27 = =, at A HO) = TroeTeosbT) at +e 2DIGITAL SIGNAL PROCESSING Putting values of a and b, we shall get 1-°°2Tcos3T) tee AT con 7) 72 HG) = [age (eosaT)z +e" or Taking T= 1s. we have 1-(0,8187)(-0.99) z HO) = 79(9.8187)(-0.99) 2 ' + 0.67032" ‘Simplifying. we get 14 (0.8105) 2 a H@ = Tyze2102"+0.670827 or EXAMPLE 74. Given an analog transfer funetion as follows: 1 HAS) = GHE+D Obtain H() using impulse invariant design method. Take T= 1s: Solution: Using partial fractions, H,(s) ean be expressed as fi HAG) = GanE+D ~ AG +2)+Bis+)) (s+ D(S+2) or H,) Comparing, we have 1=A(+2)+ Bost) we obtain B we get A= 1. Letting s and letting s = ‘Therefore, equation (i) because H,) #1 ‘The system function of the digital filter can be obtained as under: H@) = Because T = Is, therefore 0.2326 27" Trososzzt +0008 “T H@) = EXAMPLE 7:! i i i invari: is 4 eich is gl aoa H(z) using the impulse invariant technique for an analog system function | H - : a(S) = G40.8)6? + 0.5542) . We have Bs+C ; (6 +0.58+2) 6) GHOSE 405842) +O. Therefore, we write A(s? + 0.58 + 2) + (Bs + ©) ( +0. Now, if we compare the co- 2, aan efficients of s2, s and the constants on either side of the above expressio™ we. A+BDESIGN AND ANALYSIS oF INFINITE IMPULSE RESPONSE (IR) DIGITAL FILTERS 0.5A + 0.5B+¢ 0 2A+0.50 = 1 Solving the above simultaneous o written ag: 0, 0.5 He) 540.5 840.6842 ~ sep ees -05| 8+0.25 5705] —— $5025 __| 6.089 805 Le roaR E+ aaniae |* 0.0898 Again, we have H@) 0 +0. wo EXAMPLE 7.6, The system function of the analog filter is given as s+01 HAS) = Gyoaye+9 (s+ 0.25)" + (1.3919) 1- 2c" (cos1 B919T\21 + eT? ye al mol ens 026 (8+0.25)? + (.3919)" (6+ 0.25)? + (3919 | | atations, we obtai Thus, the aystem response can be whither eit A = 85, B= = 0.5 and C= 0, | | T(sin1.3919T) (cos1.3919T)z"* +e” 0.76632" ] (= 0.277 2°" + 0.606 2°? Obtain the system function of the IIR digital filter by using impulse invariance method. Solution: The denominator of H,(s) has roots at py = -0.1+)3 py = 0.1 -j3 +01 ‘Therefore, H,(s) = G+01-j9) +0153) Let us expand H,(s) in partial fractions as under: _ AL Ap HA) = [y01-j9 st0.1+8 values of A, and A, may be obtained as under: A, = 6401-9)-HeO),--o16i9 s+0.1 0.14 j3+0.1 AL = 5401+ 78|,.-o1 +19 an A, = 640.14 9HLO),DIGITAL SIGNAL PROCESSING Hence, equation (i) becomes ve, v2 WO) = yor eros transformation is given by 1 ' ally Further, we know that impulse invariance 8-Mm Ie evatom fanction for digital filter From equation (i) ns under Using this relation, we can obtain the Ve V2 Wey = oT T ay IT TT or as follows: 1 ‘This system function ean be simplified furthi 1 (0? TeosstT yz We) = a OAT egsaTye | +0 onse of digital filter. When the digital Key Note: We are always required to find frequency respo response is related to that of analog filter is designed using impulse invariance its frequency filter by following relation: or T -nsosTn 1 He) = Hal EXAMPLE 7.7. If FH, (9)= [yyy ¢qag) ‘find the corresponding H(z) using impulse invariance method for sampling frequency of 5 samples / sec. Solution: Let us first expand H,(s) in partial fractions as under: a, H@) = 1, then o > 0. Hence, the left-half of the s-plane maps onto points inside the unit circle in the z-plane and the transformation results in a stable digital system. Let us consider equation (7.50), for unity magnitude (r = 1), 6 is zero. In this case, we have g=2/_sno_)_ 2sinw/2 cosw/2 Ti+ cos0 cos” /2+ sin” w/2 + cos” w/2- sin” 0/2 = tan” . Q= pany (7.51) or equivalently, we have 2 -19T @ = 2tan"!—— asd Important Note: Equation (7.51) provides the relationship between the frequencies in the cans aa this has been shown in figure 7.13. It can be noted that the entire range in 2 is mapped only once into the range —m $ @ <7, But, as seen in figure 7.13, the mapping is nom ea uate frequencies in analog domain are expanded in the digital domain, wheres* ire compressed, is i i i : function and usually known as frequency Senes sd Bo peste of the are aneDESIGN AND ANALYSIS oF INFINITE IMPULSE No aliasing in Bilinear Transformation he main difference betweon, impulse andbilinear transformation is that then, heat ofist in bilinear transformation. Infact thisiene ™ gavantage of bilinear transformation Observe that fomplete aris is Mapped on the Unit cinelo pat in impulse invariance tne invariance nior that the only once. seamonts Vn 2k 41) RAVE cg ERA rig agi ee i niteiree repeatedly, Thus the t tone. Hence, problem of aliasing are all mapped on nsformation is many takes place in impulse invariance method The problem with bilinenr Fig. 7.13. Mapping between '( and w in aransformation is that the frequency relationshipienen Binge transformation linest Key Point: As a matter of fact, the bilinear transfor if i : i rmation applied to a continuous-time éiferentiator will not resalt in 9 diseretestime differentintor Hen immpulee investaned applied to an appropriately bandlimited continuous-time differentiator willamealt a yee time differentiator. 7.9.4 Advantages of Bilinear Transformation Method i aeistormation Method () There is one-to-one transformation from the (i) The mapping is one-to-one. (ii) There is no aliasing effect. c s-domain to the z-domain. ) Stable analog filter is transformed into the stable digital filter 7.9.2 Disadvantages of Bilinear Transformation Method ee ONO MBHIOG The mapping is non-lines and because of thi the frequency warping effect takes place 7.8.3 Comparison between Impulse Invariance Method and Bilinear Transformation Method |S.No. Impulse Invariance Method ion Method | 1 | Poles are transferred by using the | — Polesure transferred by using the expression | | expression: | 14 G ae 7 2 | Mapping is many-to-one. Mapping is one-to-one | 3. | Aliasing effect is present. Aliasing effect is not present | 4 | Ttis not suitable to design high-pass filter | jjigh-pass filter and band reject filter ean | | and band rejeet filter. be designed | 5, Only poles of the system can be mapped. Poles as well as eros can be mapped, | 6 | No frequency warping effect. Frequency warping effect 1s present, | EXAMPLE 7.12. The system function of the analog filter is given 1 (540.1) +16 H,(s)DIGITAL SIGNAL PROCESSING filter using bilinear transformation which is reso, lant Obtain the system function of the digital denominator of H,(@), we ean verte the poles of analog filter as under: +0.1)2 +16 = (+01 jd) @tO1 +j4) 2-01 +jdands=-01-i4 poles, We know that 8 =6 + jQ. Hence, the values of ‘o' and ‘Q", or Solution: From the @ or ‘There are the these two poles wi the two complex conjugate il be o =-O0.1andQ=44 at its poles. Therefore, H,(s) will be resonant at, dl to be resonant s=-014)4 will be resonant at 2= 4. A funetion i: Itis needed that the digital filter muy In other words, we can say that H,(s) bilinear transformation must map Q= 4 into o, 5 We ky k be resonant at 0, = =. This means that the that the relationship between ‘0 and ‘a’ is given by: Q= Stane Te 2,2 2,0 | or T= gtan> Substituting for 2= 4 and @, | | | 1-2" AL get 2 te ‘Substituting for this value of s in H,(s), we get H(2) ie, HG) = "Uh (22) = [Ee On simplifying the above expression, we get 0.128 + 0.0062"! - 0.1222" 1282? + 0.0062 14-0.00062"" ine ‘The roots of de sae : _onttg sat denanieator of H(2) are poles of H(z). They are | Sareeaaiee r converting these poles to their polar valtee te 0.0008 +0.8874208 aa 2 = 0.9874208 e +i 16711001 , H@ Important Point: We know that z = rel®, hence, r = 0,9874208 and @ = + 15711001 . ind w= + 1, 2 Hence, the ty ji 1e two complex conjugate poles are located at a = += 3. Therefore, H(z) will be resonant ato= = 2°PLE 7.13 It is required ¢ “nalo® filter having the follow: a design a di nit system inetiene with a7 AR bandwidth of 0.26 from an His) = Pe 840, Using bilinear transformation, obtain ts) 2 Qo = Sinn Glam = Stat. t2hn = a ayant golution: We have The system response ofthe digital filter is given ax HO) = His 20-1 = O, oo 2-1) Ts) +a Tory % 0.828 He) = 5 O.826(2 +1) a 2a), 27-1) +0.828(2 +1) Ti+) Simplifying further, we get Lea 3aid- Lana? Ans: EXAMPLE 7.14, Make use of bilinear transformation to obtain H(2) it is given that Hs) = and T Os. (s+1) Solution: For the bilinear transformation, we have H@) = H,(6)), 2 Putting T= 0.1 s in last equation, we have 4 1 (es? (2) @u=19? “pF Gke-197 |e Day (41) Simplifying, we obtain = 0.047601 +2"); BO (0.9088 Ans, 1 Ee al EXAMPLE 7.15. An analog filter has the following transfer function H,(s) = -. Using bilinear transformation technique, determine the transfer function of digital filter H(z) and also write the difference equation of digital filter. Solution: The given transfer function is: a Ai) HY = <5 Inbilinear transformation, H(z) is obtained by substituting, 2 [ect 8= plael Here, Tis the sampling time; which is not given. Hence, let us assum: see,DIGITAL SIGNAL PROCESSING » ‘| nal ng this value in equntion (i), we 1 _ L Mare a? OE ) Therefore, obtain Substitutin n+l O74 lt da Wa val ‘his is the transfor finction of digital filter. Now, we have Yo) _ 241 Hin) = c X@) Bet Converting into negative power of 7, we obtain lia! = a-2 or Yi2n3 = 271) = XML +2) or 8Y(2)— 271. Viz) = X(2) + 2-1 X(2) Taking inverse z-transform (IZT) of both sides, we obtain 3yin) ~ yin = 1) = x(n) + x(n 1) Sy(n) = x(n) + x(n 1) + y(n 1) or 1 1 1 or yin)= Z x(n) + 3x1) + 3 yn-D This is the difference equation of digital filter. EXAMPLE 7.16. The transfer function of analog filter is, 3 . HY) = GHerH With T= 0.1 see. Design the digital IIR filter using BLT. Solution: The given transfer function is oe (s+ 2)(s+3) In bilinear transformation (BLT) H(z) is obtained by substituting 2fz-1 s= =|40 fle] Here, ‘T= sampling time = 0.1 sec. H(s)= Substituting this value in equation (i), we get H@j= ————_ 3 S 3 1) soloof 4-1 ),.4 202-20.) 202-20.) Jeetas) Paes] a4 We) B+)? or H(z) = ———_Sa ves os a TY (202.~20 + 22.4 2\(202~20+32+8) ~ (222-18)(232-17) Gratitude is the memory of the heart. iy yeaa |= 37 4 29 72 tev) . “Aan + 06 7 BOG? 7 a smth Thisis the required transfor Function fry Un iter, M8 R208 al IR filter 17. Design a singh EXAMPLE 7. in 8 single pole lows pilinenr transformation applied to the halon filter with ag ap bandwidth of 0.2 t by use of GQ Ryw = —2 a S4Q,¢ where Q, is the 3 AB bandyw: th of analog file Solution: The given transfor function Seer of analog filter ig Q Hys)= nals Q, _ The cut-off frequency of digital filter (o,)is given, I ig = 0.2% We know that in BLT, Q, ii) and @, are related as under: Q, 2 [22 Joop tan 10.tn) or Q Substituting this value in equation (i), we obtain a T . H@)= —tq ivy st 7 Thus, we have obtained the transfer function of analog filter in the appropriate form, to obtain H(z) we have to put, 2h Therefore Ha) = = 2 iA oO. Testy Multiplying numerator and denominator by T, we obtain 0.65(e +1 240.652 40.65 7 Dividing numerator and denominator by 2.6 0.245(4.41) He) eee iw) = 0.509 This is the required transfer function for digital filter, It may be noted that there is a single pole at P, = 0.509, . Fi Now, in the problem, frequency of digital filter is given which is 0, = 0.2n and bandwidth is 3 dB. We shall check th frequency response. The frequency r uation of H(z). Hence, we get, of digital filter is obtained by substituting 2 = 6 in the 7 y depends upon the character of its people rather than the form ofits aa happiness of every country depends upon the —Thomas C. Haliburton ‘rnment.& DIGITAL SIGNAL PROCESSING. o.2an(e" 11) Teo) = Had}, «ol i 9.509 Ot) Tee = 0, then equation (vi) becomes. O.2aKe 7 ono al HOl= “eTanog O48 Thos. at = 0. the magnitude [Hla] = 1 Now, at «= O.2x, we obtain 2ga(e™2* 41) _ 0.245 fcos(0.2n) + jsin(0.2n} Mio) =" 03* —g.ny ~ cos(00.2a) + jsin(0.2n) - 0.509 0.245 [0.809 4 j0.587 41] _ 0.245(1 809 + j0.589) as Meo) =) $09 + 0.587 - 0.509 + 0.587 or [Ho aB Hence, the desired response is obtained Ans. { EXAMPLE 7.18 The system transfer function of analog filter is given b; | +01 f B= Const Obtain the system transfer function of digital filter using BLT which is resonant at as: Solution: The given transfer function is, s+0.1 s+0. (+0.17 +16 ~ (+0. From equation (i), we can say that 2 = 4, Hs) = x The value of o, is given as o, = > Now, we shall find out the value of sampling time (T) using the following relation: 2 tan[2 ae ply 2 or 4 pian(? T 4 2 7 or Ts ren 7] cw or 05 see k Using bilinear transformation, H(z) can be obtained by putting s = roa ast +0. Hq) = —05lz+1 ales [(ft}ooaf as [a(*")sor Pane. 0.5) mil alae [ef eo] +16 [ato] +16 2 +0. 4e-44 0.10401, or 12401 * 7 to = at eer eee ood |.12 - 3. a atl i [eae gig 18.812? = 31.982 +15.21) 516 ot) +1?. G.l2-A.941) * W681? —a1.987 ¥15.21 +162! Bde 16 AAD? 44.172 ~3.92-9.9 V2 40.27 -2.9 82.817" 40.0224 81.21 ~ A281? s027 0-41 91 This is the required transfor finetion for digital IIR filter fe Hi) Ans. EXAMPLET.19. The analog transfer function oflow- is 1 radisee. Design the digital filter using BLT method whose cut-off frequency is 20 x and sampling time is 0.0167 see by considering the warping effect, solution: Given analog transfer fanetion is. pass filter (LPR) is, H,(a)= —!— and its bandwidth 442 1 Wo= 5 eli) For the prewarping procedure, we have 2, = o 2) = 69.31 (i) +138.62 0.0167| 2+1 fe Hi) @+)) 69.3MerD att 119.762-119 882+18.86 3.732 +0.27 This is the required transfer function for digital ter. Ans. 7.40 BASIC ANALOG FILTER APPROXIMATIONS Many useful continuous-time IIR design methods have relatively simple ck Therefore, discrete-time IIR filter design methods bas formulas are rather simple to carryout. ed-form design formulas. «lon such standard continuous-time design As discussed earlier that the digital IIR filters are designed from the analog filters. Many times, itis necessary to approximate the characteristics of analog filter. This approximation is required because the practical characteristic ofa filter is not identical to the ideal characteristics. There are ¢ different types of approximation techniques as under () Butterworth filter approximation, ii) Chebyshev filter approximation. (ili) Elliptic filter approximation.SIGNAL PROCESSING, RWORTH AND CHEBYSHEV _ orworth and Chebyshev filters, R: BUTTERWORTH FILTER DIGITAL LOG FILTERS: BUTTE! cuss the design of digital Butt 7.41 ALL POLE ANAI In this article, we shall dise DESIGN OF DIGITAL BUTTERWORTH FILTE! APPROXIMATION Atypical characteristic of a bul TAZ is as shown in figure 7.14. torworth low-pass: filler (LPF) mart Toloranco Region frequency © OM 2 7 Oy J+ Passband [+ Transition +}+— Stopband —+] band loristics of analog low-pass filler (LPF). its main characteristic is ripples) in the passband. Fig. 7.14, Typical charact ponse is called as Butterworth response beca maximally fat, This means that there are no variation: squared response of low-pass Butterworth filter is given by: This type of that the passband is Now, the magnitude 2 1 | HQ)? = ———air AT53) Q It ( Q, } ‘This equation may also be expressed as fl | (754 |H(@)} —_ ww ne] 2 Qy Magnitude of analog low-pass filter (LPF). Cut-off frequency (~ 3 dB frequency). Passband edge frequency. Passband edge value. Stopband edge valu Parameter related to Here, |H(Q) , pples in passband, = Parameter related to ripples in stopband, i N= Order of the filter. le know that, i Pass Fi i in case of Low Pass Filter (LPF), the frequencies will pass upto the value of" cul “The wh 7 own go up and those who wore up go down RN. Tago |off requency (2). This is called as pase After that the frequencies are attenuated TH ‘ is called as stop band. Ideal charmace hi shown by dotted line in figure 7.15. Ides ihovalucofeut-off frequency Qa thee should be stopped. However, 1 Practical this is not happening, Now the order of filter Roughly we can say order of number of stages used in the design of filter. As the order of filler N incre, response of filter is more close response as shown in figure 7,15, ly, at Neivg is denoted hy N filter means, the log ASes, the to the ideal Frequency, © Fig. 7.15. Effect of N on frequency response . characteristics 7.424 Salient Features of Low Pass Butterworth Filter agnitude response is near that the passband is may (ii) There ‘onstant (equal to 1) at low mally flat are no ripples in the passband and stopband (iii) The maximum gain occurs at Q = 0 ‘or frequencies, This means and itis |H(O)| = 1 Gv) ‘The magnitude response is monotonically decreasing. 7.12.2 Designing Expressions and Designing Steps Let A, = Attenuation in passband. A, = Attenuation in stopband, Q, = Passband edge frequency. Q, = Cut-off frequency Q, = Stopband edge frequency. In numerical problems, the specifications of required digit n and it is asked to design a particular discrete time Butterworth filter. ‘Then the following steps must be used : al filter is usually gi 1. From the given specifications of digital filter, we obtain equivalent analog filter as under: (a) For impulse invariance method, we have o . (b) For bilinear transformation method, we have Q= Qe Pp tan Here, Q = Frequency of analog filter Frequency of digital filter T 2. We evaluate pling, time he order N of filter using the following expression:DIGITAL SIGNAL PROCESSING Ne oye (Q, 2.) te “top a ay Hore, 5, = Attennation in stopband ivon in decibels (AB) then, we make use of the following expression wl 7 F weed | i i If the specifieations are re og| Oe PS gh | 2 Q, e log 2 i “ | | 3. After that we determine cut-off frequency (2). The cut-off frequency (Q,) of analog filter is calculated as under: invariance method, we have | T (b) For bilinear transformation method, we have 4. Next, we determine the poles using the following expression: pie £0, oN DN, 5 50,1,2,..N—1 | IF the plas nee coumeex oonttugtia than orga / ee pals ae complex conjugate then organize the poles (p,) as eomplox conjugate pats tht s, and s,", 8, and 53 ete. f 5. Next, we calculate the system transfer function of analog filter using following expressio” | | | I o Hy(s)= ——8e _ (8~p,)(6=p,).. fT cg Toss wa cil Shag Toe da St TaToANTTN — G.B, ShawIS Sy)(s ~ 8) 6. Lastly, we design the digital iter using impuge ' method. £7.20 A digital filter has following frequene passband frequency = 0, = 0.2 5 Stopband frequency = @, ® What are the corresponding specific domain if: i) Impulse invariance technique is used for designing. i) Bilinear transformation is used for designing, solution: Here, let us assume sampling time T @ For impulse invariance method, we have invariance method or bilinear transformation 'Y specifientio ‘ations for passhand and stopband frequencies in analog. see ®, 2, 0.2 x = 0.63 rad/see, @, _ 038 and O,= FP = PS =0.3n=0.94 radisec, (i) For bilinear transformation, we have tan o2e1e) and 21an( 95182) = 019 ads Ans. EXAMPLE 7.21. Design a second order discrete-time Butterworth filter with cut-off frequency of L Hii and sampling frequency of 10‘ samples/see by bilinear transformation. Solution: In this problem, the specifications of digital filter are not given directly. So, first we have to obtain Tequired design specifications for digital filter. Then, for Butterworth approximation, we have to convert the Spetifications into specifications of equivalent analog filter. Finally, using bilinear transformation we have to obtain H(z), Given that Order of filter, N = 2 Cutoff frequency of analog filter, F, = 1 kHz = 1000 Haz. Sampling frequency, F, = 10' samples/sec = 10,000 Hz. First, let us determine the required design specifications of digital filter We have the equation to convert continuous frequency (F) into diserete frequency (0. Leis oe K Thus, Fe _ 1000 _ 6.1 eycles/sample © FR, 7 10,000 Now, the angular frequency (frequency of digital filter) ©, = nf, = 2nx 0.1 = 0.20 radians/sampleDIGITAL SI Now, let ux determin ations of a For bilinear (ransformation, we have the ape Qe A tan 1 Thus, 02 He 1 P= sampling time = = eee F, * 10,000 0.2 2 ry to calculate it. Thus, the specifications of analog filter are ay The value under: (2) Cutoff frequer ii) Order of filter = N Let us calculate the poles usi po , iN 425 + tnt | = 6198.39 radians/see. of N is given so it is not nece nfsec, =O, = 6498.99 radi Here N = 2, Thus, i = 0 to N ~ 1 means i = 0 and i = (i) For i=0,wehave 6498.39 ei”! = + 6498.39 |eas( + isin( = us) 4] Pg = + £2, 08 + Has = £ 6498.39 |- 0.707 + j 0.707] or Py = 6498.39 [~ 0.707 ~ j 0.707] and ~ 6498.39 [- 0.707 + j 0.707] or Py = ~ 4595.05 + j 4595.05 and 4595.05 ~ j 459: Gi) For i we have Pye £2, cl 42+ bana ++ 6498.39 ei) Po = (-4595.05 + j4595.05) * i or py = #6498.39 oon isin{ + 6498.39 [- 0.707 ~ j 0.7071 4595.05 orp, = 6498.39 |- 0.707 ~ 0.707) and ~ 6498.39 [- 0,707 ~ j 0.7071 or py = ~ 4595.05 ~ 4595.05 and 4595.05 + j 4595.05 Now, let us plot these poles as shown in figure 7.16. x a Py = (-4595.05 — 4595.05) (4595.05 — 4595.05) Fig. 7.16. Plot of poles, p, 2a Important Point: We know that analog filter is stable if the poles lie on the LHS. of s:plane. So, for the filter to be stable, we have to select the poles which are on the L.HS. of s-plane. Observe figure 7.16, We shall denote these poles by s, and s;; because these poles &* complex conjugate of each other. ‘Thus, and (05 + j 4695.05, 4595.05, Now, the transfer function of analog filter is obtained by using the following expression: __ (6198.39) 4595.05) (s + 4595.05 + j Simplifying, we gotDESIGN AND ANALYs}: SOF 1 ENTE IPULSE Response (UR) DIGITAL FILTERS Hs) = ——__(6498.39)? 190.184 42.22x108 stl. let ws design digital filter using bilinear transformati ation, ce of bilinear transformation, H(z) is obtained ye 2fr-1] Inca ained by putting « = Hy ze} In equation of Hs) 1 Here. T= Too ‘Thus, we have to put, c= 2eioe [2 | in the equation of Hs) Therefore, z-1 : , . zai || *183-802%109/ a | aeazeao® This is the required transfer function for digital filter. Ans, EXAMPLE 7.22 Using bilinear transformation, design a Butterworth filter which satisfies the following conditions: 0.8 < |H(e%)| <1 O+(2"} +i 2) or Po = £0.75 |- 0.707 + j 0.707) or Po= t1-0.53 +j 0.53] Po= ~ 0.53 + j 0.53 and 0.53 ~ j 0.53 Fori=1, we have or Py = £0.75 oh +204 « 4 9.75 gion sno) 4 4 £0.75 (-0.707 ~}j 0.707) = ~ 0.53 ~}0.53 and 0.59 + oe) 1 53 +5 0.53 ‘ ‘The location of poles are as shown in figure 7.18, Fig. 7.18, Location of poles | +iDESIGN AND ANALYSIS OF INFINITE IMPULSE RESPONSE (IIR) DIGITAL FILTERS: eee Note: We kn 4 : : Tgagortant Re ne Se ats che stability the poles whieh are dt iw SL oF plane hoo aes complex conjugate stead chee Weett 4 DSS and - 080510 NEU ‘ : We wil ee oF 5,= ~053 +059 andst=-068-jonh nt A Jet us determine H,(s). (Now Fer function of analog filter isi oN ransfer funtion of analog eri given hy N Hf. —— (0.75 (s-s,)(s=s (840.53 -j0.63)(8 4 0.53 + 40.58) - Ys) oe +0.538 + 30.538 + 0.538 +0.28 + 0.28 — j0.538 — 0.28 + 0.28 (0.75) 0.56 Hs) = or S41.068+0.56 8° 41.06540.56 pax B: Determination of Transfer Function for Digital Filter We have to use bilinear transformation method. Then, H(z) is obtained by putting s = > pation of Hs) Tet us assume sampling time T = 1. Therefore, we have He) = ——,—*8 ___ 422) +e@x1.00(21)+056 *@ x41 zl ay 0.56(2+1)* o H@)= — e+ + A(z — 22-41) + (2.122-2.12)(2+1) +0.56(2 +1)" Be) 0.56(+0" a @) = =—_—_$__—— — ooo a2 80444 2.122? + 2.122—2.122-2.12+ 0.562" +1.122+ 0.56 0.56(2+1)* o Hee) = fe) 6.682" - 6.882 +2.44 This is the required transfer function to design the digital filter. Ans. EXAMPLE 7.23 (a) Find the order and cut-off frequency of a digital filter with the following ‘specifications: 0.89 < | Hie) | <1, O} _ S| vis (o.889)? +1 B= € ~ 0.882 Next, we shall calculate values ofr and R. Wehave Min fF eli =o, Bd or ani of ellipse = r=, . “3p (2.64) or r= 20. 22.6 2(2.64) ‘The major axis of an ellipse is expressed as: 2 2.64)? + BPH 99 2.68)" +1 Re 0, = Gee * 8049 Now, we shall calculate values of 6, as under: Gitte oto N—1 that means for i= 0. 2* oN x, O+I)n 8 ot oa o™ Using equation (i), we can write pole positions as under: 5) = ros Oy ~ JR sin 0, 8) = 22.6 cos x + j(30.19) sin x = Lastly, let us obtain the system function H,(s). ‘The system function is given by: or 22.6 H®)= (ogy 7 ov 226 Ai) i) (iii) div)DIGITAL SIGNAL PROCESSING Here, for N = 1, we have, i= by= 226 22.6 8+22.6 This is the required transfer function. Ans. Hence, we have H,(s) = EXAMPLE. 7.27. Design a digital low pass filter using Chebyshev filter seeing procedure ¢! t the following specifications: Passhand magnitude characteristics that is constant to within gre recurrences below w= 0.2 x and stopband attenuation of atleast 15 dB for frequencies bev @= 0.3 xand x. Use bilinear transformation. Ween Solution: In this problem, the specifications of digital filter are given. The given specifications are : Passband ripple = A,=16B Passband edge frequency = @, = 0.2 Stopband attenuation A,=15 dB Stopband edge frequency 0,=0.3 n First, we will calculate the values of edge frequencies for the analog filter. First, let us determine analog filter's edge frequencies. For the bilinear transformation, we have Important Point: Sampling time is not given. Therefore, we will assume sampling time T=1sec. Thus, 0.6498 and Again Hence, ‘ = 1.019 Next, we determine the parameter € 1 It is given by, Ae [ao va {io 10) ‘Now, let us determine the order (N) of a filter. Itis calculated using the equation, |HGQ)| in 4B = - 20 logyy € ~ 6(N ~ 1) - 20 N logy (2) Given attenuation is 15 dB. or = 15 dB = ~ 20 logy 0.5088 ~ 6(N — 1) ~ 20 N log,(1.019) or ~15 = 5.87-6N +6-0.1634 or 6.1634 N = 26.87 or N= 4.35 Thus, order of filter = N=5 ‘The pole position are given by: 5, = 1 c05 0, + jR sin 0; Now, we will calculate values of r, R and 6; li) Minor axis of ellipse =vivo ORR)? « 0.5088 | = 133 2 G oga98, 13-1 2xran * 0.186 “i And major axis ellipse = R= q, M41 ar R= 06499 039" +1 or E : 2xiag 7067 ain) Now, we shall calculate values of 8, z Qisne ve have, z,& Le 2 : 1=0,1,2,..N—1 thatmeans i= 0,1,2,8,4 z 6 Fori= 0: 3 S . x Qin 8x Fori=1: z ar 2* 40 10 zon For o'a ge Ey tL 1 3° 2" 10 ~ 10 ae ltn 2* 10 ~ 10 Now putting these values in equation (i), we get Fori=0, = 0.186 cos 0) +j0.67 sin 0 én or 5) = 0.186 cos al + 50.67 sin ry =~ 0.057 + j0.637 $1 = 0.186 cos 6, + 0.67 sin 0, Bn 8x , or 8 = 0.186 cos uo} +5067 sin 3] == 0.15 + j0.39 5) = 0.186 cos 8, + j0.67 sin 0, S_ = 0.186 cos m + j 0.67 sin x =~ 0.186 Fori=3, 5, = 0.186 cos 6, + j0.67 sin lan 12m) oo 1 49 e 85 = 0.186 cos (2) +5 0.67 | io | = 0.15 ~j0.39 Fori=4, 8, = 0.186 cos 0, + j0.67 sin 0, Man x . Le sy 0186 os (22) «oar sin [Ip] = 0087 sno Next, we determine the system transfer function of analog filter Itis given by, i H,(s) = Gens =H = BE 85)E-*) i HYe= GA TET WFO) + 0.186) +015 + 0.39) +0057 + 50.637) (6 +0.057 - j0.637)(7496 7 DIGITAL SIGNAL PROCI ‘SSING a or HY)= ae — —_— (g+0,816)(8° +0.38 + 0.1746)(s" +0.1148 4 0.409) Now, N=5 thus, j= by = Constant term in the denominator 186) (0.1746) (0.409) = 0.0135 or is by= 0.0132 | or HYs)= —— 379301746)" 40. o132______ | (6 +0.186)(62 + 0.88 + 0.1746)(s* + 0.1148 + 0.409) | Lastly, we determine the transfer function for digital filter. By using bilinear transformation, the transfer funetion for digital filter H(z) is obtained from Hs) 8) by | putting, 2fz-1 s= 2/2" Tlz+1 Here, T= Sampling time = 1 sec. (assumed) 0.0132 el eal al tal at8.8 DESIGN TECHNIQUES OF FIR FILTER ARIOUS KINDS OF WINDOWS =o —— nj eaeaeerrereeeerrrerj——=<=s i iple of the window design method is to truncate the ideal impulse response with eee ‘The corresponding effect in the frequency domain is that the ideal frequene; response is convolved with the Fourier transform of ee DO YOU KNOW? window. Ifthe idea filter isa low passfilter,thediscontinuity J oY ee. in its frequency response is smeared as the main lobe of the eiusation Inj Fourier transform of the window moves across the | Response ( : ) digital filter is discontinuity in the convolution process. To a first ]based on a frequency respone approximation, the width of the resulting transition band is eee aria ; es ‘ determined by the width of the main lobe of the Fourier | implementation is accomplishes transform of the window, and the passband and stopband_| t#king the oper camera oe ripples are determined by the side lobs of the Fourier gesined q cToduency-responsy transform of the window. Because the passband and stopband _[$Pecification. ripples are produced by integration of the symmetric window side lobs, the ripples in the passband and the stopband are approximately the same. Furthermor. toa very good approximation, the maximum passband and stopband deviations are not depentest on M and can be changed only by changing the shape of the window used. Design techniques for FIR filters without linear phase have been given by Chen and Paris (1987), Parks and Burrus (1987), Schussler and Steffen (1988), and Karam and McClellan (1995)- Asa matter of fact, various window functions are used for filter design. First, let us considertt® time and frequency domain representation of different types of windows. Basically, following are the types of window functions: (@ Rectangular window (i) Bartlett (1., Triangular window) (iv) Hamming window (vi) Kaiser window Gii) Blackmann window (¥) Hanning window | * What are different window functions? Write mathemati are these windows used in the design of FIR filter, expression for three window functio 5 Hot | | 1Not ‘The Bartlett, Hann Kes. ator. ‘The Hanning window is ofifgsometimes referred to as the ty (1958) to describe the operation ¢ a post widely used NaMe for the w A n Hann, an Austrian Meteoralogiat, oF applyinye thie ett hannine’ was used by Blackman ane ‘indo, With tae nee Minlow toa signal andl has since become sannine YE preferences for the choice of Hanning’ or 38d Rectangular Window wat) ca matter of fret, different typos of windows are nod to ft Nw petangular window: Figure 8.6 shows thereetanguly window,» Irisdenoted by wy. Hts magnitude is for the rane M~ 1. Since, the shape of window funetion ix 1 i aaled as rectangular window, The rectangular window of length TT oe a5 6 Mt Fig, 8.6, Rectangular window ‘Mis represented by w, if) and it is expressed as un {1 forn=0,1,2, sy60 = | »M-1 0 elsewhere (8.36) Now, let us consider the Fourier tr sform of rectangular window. It may be expressed as: Mat Wa) = x wa(nyem (8.37) Substituting for w_(n) from equation (8.36), equation (8.37) becomes, Mat Wa) = Yee (8.38) = Here, let us use the standard series relation as under: anger yak = 2 (8.39) Ina th a = €, equation (8.38) will take the following form uM W,(o) = 2 (8.40) R Te The last equation can be rearranged as under: MM joy 35 © Wy() = om We know that oi! e-i# = 2j sin 6. Hence, above expression becomes Sol! on 7 asin{ 0) 2 Wal) = o 2 2sin| ain( 3][50] DIGITAL SIGNAL ——— o(] 3 “64, we(oll Main lobe IWp@le 0| Side lobe we get magnit as under: in( OM a 60 (8.42) IW! = To saan -80 (3 =“ ds trum of s Fig. 8.7. Magnitude spectrum of rectangut Figure 8.7illustrates the response ofrectangular i. doreorer etd window given by above expression for M = 50. i frectangular window described above, it mayte Important Point: In the magnitude spectrum o} ar a reo hat there is one main lobe and many side lobes. As ‘MW’ increas®®s the main lobe becomes narrower. The area under the side lobes remain same irrespective of changes in the value of I, 8.8.2 Bartlett (ie., Triangular) Window Bartlett window is also called as triangular window. It is expressed mathematically as: wr(n) = 41- for n=0, 1,...M-1 (B43) M-1 0 elsewhere Fourier transform of this window can be obtained in a similar manner as discussed for rectangular window. Figure 8.8 (a) shows the sketch of this window. Figure 8.8 (b) shows the magnitude response of this window. wn) W(o ° -20 -40 60 -80 Met 9-100 | (@) (b) | 1 23 4 Fig. 8.8. Bartlett window (a) Time domain graph (b) Magnitude response. | Koy Point: | Key Point: Asa matter of atthe Fourier transform af th Bartlett window canbe expressed | are can be er eansfoms of rectangular windows, andthe Fourier transforms ofthe other windor sums of frequeney-shifted Fourier transforms of the rectangul™ | | 8.8.3 Blackmann Window ‘The Blackman window h its ti : nk as bell like shape of its time-domain-samples. It is expressed iDESIGN A} ND ANALYSIS OF Fy INIT E IMPULSE ‘SPONSE (FIR) FILTER \ 551 0.42 2 wy(n) = M for n=0,1,..M-1 8,44) ) ab oF Blackman eee Figure 8.9 (a) shows the a » tear ellie n window, Figure: é i! si Me aponse. Here, it may be observed that the Width af ee Fitare 8.9 0) shows its magnitude re small side lobes, ) lnhe is increased, However, it has very (0) wal Wolea)l 0. 20, 0, ~60: -00 1-100. . (a) ae (o) Fig. 88. Blackman window (a) Time-domain graph (b) magnitude response 8.8.4 Hamming Window Hamming window is most commonly used window in speech proce: ing. It is given as 0.54 0.46 cos for n=0,1,.. M1 wy(n) = M-1 (8.45) 0 elsewhere This windows also has bell like shape. Its first and last samp shows the graph of Hamming window. Figure 8.10 (b) shows the magniti Ithas reduced side lobes but slightly increased main lobe. The are not zero. Figure 8.10 (a) ude response of this window. side lobes are higher than Blackmann window. wn) Wyo) o 20 40 | * 60 | i, + 100 wo aad (@) oot I (0) Fig. 8.10. Hamming window (a) Time-domain graph (b) Magnitude response 288 Hanning Window Hanning window has shape similar to those of Blackmann and Hamming, Les firs Samples are zero, It is given as, as ( 0s Wynn) = 42 } for n=0, 1,..M-1 (8.46) 0 elsewhereDIGITAL SIGNAL PROCESSING dow is commonly used for spectrum 3.11 @ shows the graph of his window. The Win eee nse of Ha —~ window. Her ye observed that ; ‘as narrow main lobe, but first few side lobes are signir’ analysis, speech and music proces sing i 1 ee, ul ; " indow. Here, it may be observe 8 cant. Then sidelobes reduce rapidly. f | Ma) Myo)! | | 0 -20 40 I 60: } 0: t 2 I I finn ep uals | op 12 3 @ M-1 (b) t Fig, 8.11. Hanning window (a) Time-domain graph (b) Magnitude response 8.8.6 Kaiser Window a Fadmadiay i indows. The mples i b-sections, we discussed four different types of wil a ‘ the eee ‘depend only upon ‘M’ ic,, length of the window. In other words, ee a of the main lobe and attenuation of sidelobes depends only upon length ‘M’ of the window. They cannot be controlled independently. ; - l However, Kaiser window allows separate control of width of the main lobe and attenuation of sidelobes. The Kaiser window is defined as under: 1 Iy ols-(22*)T 60 B= 19.5842 (A 21) 4 0.07886(A-21) for 2150 f= {0.5842 (A ~21)"! +0.07886(A-21) for 21< A <50 0 for A<21 alue of M using the following expression: Here, Aw = Transition width and is expr Aw = 0, = 0, Where «, = Stopband edge frequeney and @, = Passband edge frequency. sed as under (v) Now, we select the desired impulse response depending on type of filter e.g., low pass, high pass ote. (vi) Finally, we determine the FIR filter coefficients using the relation h(n) = hy (n) . We: 8.15 SIGN OF FIR FILTERS USING OTHER WINDOW FUNCTIONS indow method does not pert it individual control over the approximation errors in different bands. For muny applications, better filters result from a minimax strategy (minimization oft? maximum errors) or a frequency-weighted error criterion, Such designs can be achieved usiné algorithmic techniques, Different types of window functions are window is selected depending upon the different type of windows, available which reduce ringing effect. The particule application. Table 8.3 lists the time-domain equations fo"DESIGN AND ANALYSIS OF FiNtre IMPULSE RESPONSE (FIR) FILTER 's Window functions and their Cotresponding Shapes ‘Time-domatn Sequan : “a wOOsnsmere Te Blane bt Winey FeaeeR re] Name of No.| Window Rectangular 1 Bartlett | (criangular) Blackmann | 0.42-0.5c0s2™ s9.g win) | M-1 1 | | ann cal : M-1 o4 0 sil Hilt, 7 | a Mot | 2mm win). | | cos, 4. | Hamming 0.54 - 0.46 cos. M-1 i ob atl | I. 5. | Hanning a cos 0 Not 22 mn 6. | Kaiser 1648) 1-( *) 08 . 0.6 | El li pi el M ln Table 83, 1,(4) in Kaiser window function is the modified Bessel function. There are two more "v8 which are not listed in the above table. ‘They are Lanczos window and Tukey window. Figure 8.15 illustrates the frequency response of low-pass filter with cut-off frequency ©, designed “Sig Various types of windows. The magnitude response is shown in dB scaleDIGITAL SIGNAL PROCESSING np i R filter designed usit k : 8.16, it may be observed that I'l id using Hama SEE eT Nbal yeduced sidelobes a8 compared to gaetaneulet ‘window, Blac nd Fa eine 8.16(@) has sinallest sidelobes but width of main lobe i increased, Hee, tt SE caneal sidelobes and width of main lobe. Kaiser yin e between attenuation of bes K Ea fri lity to the designer. Therefore, it is more commonly used window form filter design. IH(o) BB [H(e) BB i | “20 -20 = 40 4 60 -80| -80) a F @ 7100) 4: af x i Magnitude response of FIR low pass fiter creat trae’ ete traces [H(o) BB [H(o)|48 0 ) -20 -20/. 0 —40 60) 60 -80) -80 100 oe 100 2 @ O% (c) Magnitude response of FIR low-pass filter (d) Magnitude response of FIR low-pass filter dese! designed using Blackmann window. It has very using Kaiser window. It has reduced side lobes ‘small side lobes but increased width of main lobe. & transition band is also narrow. Fig. 8.15. Magnitude response of low pass FIR filter designed by various windows EXAMPLE 8.9 Design the symmetric FIR low pass filter for which desired frequency respo™** expressed as I forlwl— = 27. Similarly, from equation (i), @, = @, = 0.8 rad sample.DESIGN AND ANALYSIS OF FINITE IMPULSE RESPONSE (FIR) FILTER ‘substituting these values in above equa We obtain sin[0.3n(n - 27) m — a =27) [oss 0.46. o({ 3) for n #27 0.3} 0.54 - = 5 [oss oabon( | for n=27 for n=0to 54 Important Point: The values of unit samplo response may be obtained above equation to get desired linear hace FIR filter. EXAMPLE 8.12. Design a normalised linear east 40 dB attenuation in the stopband, Als ‘solution: First, et us obtain @, phase FIR filter having the phase delay of t= 4 and at 0, obtain the magnitude/frequency response of the filter. ‘The linear phase FIR filter is normalised. This means cut off frequency will be: ©, = 1 rad/ sample Now, let us obtain length M of the filter Phase delay, t= 4. We know that for a linear phase FIR filter, we have Now, t= 4 gives, 45 or M=9 ‘Therefore, the length of the filter is M= 9. ‘Now, let us choose the type of the window. At least, 40 dB attenuation is required in stopband. From Table 8.2, it is obvious that Hanning window provides minimum of 44 dB attenuation. Therefore, Window used: Hanning window Now, let us obtain desired unit sample response hy(n). In example 8.11, we have observed that hg (n) has been obtained for an ideal low-pass filter of cut-off frequency o,. Such h, (n) is given by equation (iv) in previous example as under : sin aa 2 MM for n# hyo) = fn . oe for n= ® M-1 In above expression, we have to substitute w, = 1 rad/sample and —— =+ = 4. This means that sin(0-49) ge ned x(n) Ltr ned ® ham) = sali) Now, let us obtain h(n) by windowing. In this example, we have selected Hanning window.DIGITAL SIGNAL PROCESSING From Table 8.3, Hanning window is given as unde a rae = time) Substituting for M, above expression will take the following form Vy aoe ™ an) © i! ova mt) Also, hin) may be obtained by windowing as under: hin) = han).wn) mn sin(n— 4) () cos } 1) for nea —_| or ha = ax(n-0) ep for n=4 forn=0to8 This is the unit sample response of the desired linear phase FIR filter. Now, let us obtain the magnitude (frequency) response. Here. order of the filter is, M = 9 ie., odd, For odd M, the magnitude/frequency response of FIR filters x given by: M-3 MS) s ayeose{n x *) n=O 1H)! Substituting for M = 9, above expression becomes, 3 |H(@)| = h(4)+2 D) h(n) cos a(n - 4) neo Values of h (n) can be substituted in above expression from equation (ii). Then | H (w)| may be obtained fr various values of EXAMPLE €.13 Design the bandpass linear phase FIR filter having cut-off frequencies of @,,=1 rad/ sample and ,,=2 rad/ sample. Obtain the unit sample response through following window: 1 for 0 3 aie and B = 3.395 Substituting the values in equation (8.47), the window will be defined as under: 1 1p}3. a Ca} P w(n) = 1b T)@.395) for 0 si Bis integer In other words, filter length is M+ ‘Therefore, ha(n) will be given 293 + 1= 224, ie., even. for0

You might also like

• 6.0 Iir Filter Bee3213

  

  No ratings yet

  6.0 Iir Filter Bee3213

  54 pages
• Digital Filter Design

  

  No ratings yet

  Digital Filter Design

  31 pages
• Unit-2 - Kec503 DSP - 2023-24

  

  No ratings yet

  Unit-2 - Kec503 DSP - 2023-24

  22 pages
• Impulse & Bilinear

  

  No ratings yet

  Impulse & Bilinear

  16 pages
• Lecture9-IIR Filter

  

  100% (1)

  Lecture9-IIR Filter

  54 pages
• Dsp-Unit 5.2 Iir Filters

  

  No ratings yet

  Dsp-Unit 5.2 Iir Filters

  81 pages
• IIR Filter Design Techniques

  

  No ratings yet

  IIR Filter Design Techniques

  21 pages
• IIR Filter Design and Techniques

  

  100% (1)

  IIR Filter Design and Techniques

  50 pages
• Unit3 IIR Design

  

  No ratings yet

  Unit3 IIR Design

  103 pages
• Digital Filter Design Essentials

  

  No ratings yet

  Digital Filter Design Essentials

  103 pages
• IIR Filter Design: Impulse Invariant

  

  No ratings yet

  IIR Filter Design: Impulse Invariant

  4 pages
• Unit3 Iir Design Lecture Notes

  

  No ratings yet

  Unit3 Iir Design Lecture Notes

  24 pages
• DF Lesson 06

  

  No ratings yet

  DF Lesson 06

  71 pages
• IIR Filters

  

  No ratings yet

  IIR Filters

  46 pages
• Unit-4 - Design of Digital Filter

  

  No ratings yet

  Unit-4 - Design of Digital Filter

  107 pages
• LAB 7: IIR Filter Design Objectives

  

  0% (1)

  LAB 7: IIR Filter Design Objectives

  13 pages
• DSP Test Ii: Figure 2a An LCR Filter

  

  No ratings yet

  DSP Test Ii: Figure 2a An LCR Filter

  5 pages
• Chapter 6 Digital Filter Design For Teaching

  

  No ratings yet

  Chapter 6 Digital Filter Design For Teaching

  27 pages
• DSP Unit 4 Gate

  

  No ratings yet

  DSP Unit 4 Gate

  44 pages
• IIR Filters

  

  No ratings yet

  IIR Filters

  31 pages
• 18ee63 DSP Module-4

  

  No ratings yet

  18ee63 DSP Module-4

  18 pages
• 1.7.3 Iir Filter Design

  

  No ratings yet

  1.7.3 Iir Filter Design

  34 pages
• Discrete-Time IIR Filter Design Guide

  

  No ratings yet

  Discrete-Time IIR Filter Design Guide

  21 pages
• 16 IIR Filter Design

  

  No ratings yet

  16 IIR Filter Design

  23 pages
• IIR Filter Design Guide

  

  No ratings yet

  IIR Filter Design Guide

  81 pages
• Chapter 8

  

  No ratings yet

  Chapter 8

  42 pages
• Design of Infinite Impulse Response (IIR) Digital Filters

  

  No ratings yet

  Design of Infinite Impulse Response (IIR) Digital Filters

  16 pages
• Analysis and Design of Digital Filter: EE 422G Notes Instructor: Cheung

  

  No ratings yet

  Analysis and Design of Digital Filter: EE 422G Notes Instructor: Cheung

  13 pages
• Analysis and Design of Digital Filter: EE 422G Notes Instructor: Cheung

  

  No ratings yet

  Analysis and Design of Digital Filter: EE 422G Notes Instructor: Cheung

  13 pages
• Some FIR Filter Designs

  

  No ratings yet

  Some FIR Filter Designs

  10 pages
• Design and Implementation of IIR Filter

  

  67% (3)

  Design and Implementation of IIR Filter

  45 pages
• Digital Filter Design for Engineers

  

  No ratings yet

  Digital Filter Design for Engineers

  10 pages
• Digital FIR Filter Design

  

  No ratings yet

  Digital FIR Filter Design

  86 pages
• Iir Digital Filters: Unit-V

  

  No ratings yet

  Iir Digital Filters: Unit-V

  81 pages
• Digital Filter MBICT

  

  No ratings yet

  Digital Filter MBICT

  28 pages
• Digital Filters

  

  No ratings yet

  Digital Filters

  27 pages
• Impulse Invariant Method

  

  No ratings yet

  Impulse Invariant Method

  13 pages
• Impulse Invariant Method

  

  No ratings yet

  Impulse Invariant Method

  21 pages
• CPEN 304 L06 - IIR Filter Design 2

  

  No ratings yet

  CPEN 304 L06 - IIR Filter Design 2

  36 pages
• P S CK: Seshadri Rao Knowledge Village::Gudlavalleru

  

  No ratings yet

  P S CK: Seshadri Rao Knowledge Village::Gudlavalleru

  2 pages
• H (Ejw: Design of Iir Digital Filters

  

  No ratings yet

  H (Ejw: Design of Iir Digital Filters

  5 pages
• Dspa Word File

  

  No ratings yet

  Dspa Word File

  82 pages
• Digital Filter Design Guide

  

  No ratings yet

  Digital Filter Design Guide

  26 pages
• Summary Dig Filt

  

  No ratings yet

  Summary Dig Filt

  21 pages
• Digital Filter Design for Engineers

  

  No ratings yet

  Digital Filter Design for Engineers

  10 pages
• ECE 4316: Digital Signal Processing: Dr. Hany M. Zamel

  

  No ratings yet

  ECE 4316: Digital Signal Processing: Dr. Hany M. Zamel

  19 pages
• Unit-Iii Iir Digital Filters

  

  No ratings yet

  Unit-Iii Iir Digital Filters

  43 pages
• Unit 2 BEC 503

  

  No ratings yet

  Unit 2 BEC 503

  107 pages
• Lab.6&7. IIR Filters

  

  No ratings yet

  Lab.6&7. IIR Filters

  8 pages
• Digital Signal Processing: Subject Code - ECN-312

  

  No ratings yet

  Digital Signal Processing: Subject Code - ECN-312

  40 pages
• Digital Signal Processoer

  

  No ratings yet

  Digital Signal Processoer

  16 pages
• Design of IIR Filters

  

  No ratings yet

  Design of IIR Filters

  31 pages
• DSP CH-10f (IIR - Imp - Inv)

  

  No ratings yet

  DSP CH-10f (IIR - Imp - Inv)

  61 pages
• DSP Compiled All

  

  No ratings yet

  DSP Compiled All

  375 pages
• Design of IIR Filters 2

  

  No ratings yet

  Design of IIR Filters 2

  53 pages
• Digital IIR Filter Design Using Bilinear Transformation in MATLAB

  

  No ratings yet

  Digital IIR Filter Design Using Bilinear Transformation in MATLAB

  6 pages