Scribd
Upload
31 views

Module 4 DSP Iir Book

Uploaded by

ANU
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
31 views

Module 4 DSP Iir Book

Uploaded by

ANU
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 50
44 z — systems can be FIR or IR ‘The difference equations by Digital Signal Processing iscrete time EGE Structures for IIR systems wie doe «The digital filters are pasically discrete ume LTT systems aeseen ease we hav ‘These systems are described by’ ‘itference equation® actually implemented in hardware OF software. cet resizaton st «The various ways to implement difference equation = basis of COMP! digital filter structures. ‘These st are derived oF the implementation finite wordlength effects ef Sen ae Ss : The can be of two types. If om sed it i d ma inte . : it is called Can, nacture is equal to order of the difference eai® soy then ite called non-anay fom wealization. Tf the number of delays in the tonic realization. Describing Equation of IIR Syste! We know that the LTI systems are describ yr) wutational complexity, ease g a by the general diference equation a Bayes & meO-® imo ea = (4 _Taking z-transform of above equation, Ye $ az Y@)+ § by z* Xl) et Kao ref $ a | = § yz * X@) it ko We know that system function H(2) “43: thence from above equation we get, Sark z (412) He) = =o —__ we az it 1 « This is the rational form of system function which is expressed as the ratio of two polynomials in 2” 0A dis time is i st si Aan ren ied ying onion 411 cr eal 12h ions can be lence ste ways. We know that discrete time system is imy Jemented wi sa ot cl elements, multipliers and adders as elementary blocks. A brief summ: cases xe eee " nce its realization given in Fig. a system a this igure, u(r) is the input and Pela en (0) 8 the inpt snd yf) ae 2hO -H,@), the realization for H(z) ' ) Ua by casading Hy(2) and H(2) as shown in nears sq. a Cesc connection of Hy (2) and (2) spent direct form realization for H1(2) can be obtained by He “he realization of H;(2) [Fig. 42.5] and realization Fig. 42.6 Dict foim implomentagon of conten tig) fig. 42.6) This connection is shown in Fig. 428. ects raizaton of S70) inal! pol arly it is delayed and used in all Je delay element can be used fo Je system. Simil fs, Observe that sing zero system. Xx tiem But we have used 670 (01), This can be used by a! ‘This means the and zero system element. The Fes Fig. 429. In the alization ‘assumed N= M- ‘This realization of IR system duced memory Fequirenment comP: form] structure. maximum of {MN} direct form-II structure reduc it is called canonic form. structure requires same number (ie. M+N+1) ‘and additions (1. of direct foravI structure. can be me aking rolizat shown In is called ces Solution : /) To check whether FIR or HR filtor : Consider the given system function. It can be written as, 2 az 2 1722 -525241375 at H@ = 3+ ee ©) = 345-05" 2-005 ~ 207524 0125 mare that the system function has numerator polynomial of order 2 as well as denominator polynomial o ler2 ‘The system function has poles as well as zeros. Hence it il e Fl represents IIR filter. The FIR filter has all zero syst il) Direct form- realization : ‘We can write equation 4.2.13 as follows : 7-5.25271 4137522 a 1-075 271 + 0.12522 = os os ¥ expand summations of equation 42.1 for M=N =2 ie, HG) = thet byt V4m 27 + ap 22 ‘The direct form-1 realization of above equation for’ bove equation for gener ; seers. with equation 42.14 we get, ‘alized M and N is shown in Fig. 4.2.4. On compari" ee = a HG) = Seow Pt in Fitor Design an my pit formal canonic realization : cinema won wite equation 42.13 as follows, 752521 a HH) = 413752 loos. Gan 075274 010527 w= 4215) Y@ eJnow that H(2) ae Hence above equation will be, X@ © 1-07527 4010527 stus rearrange the above equation as, Y@_W@) _ 7-525244137527 W@) X@) — 1-07527+ 012527 We) Y@ 3 XG) WE) : t rt w= (42:16) ee FAQ) HO = apest ages CBT ) wo (4247), Wey ; Q Fae) “x(q ~ To07se7 + OMe? ww (42.18) @. 4137527 = 7-52529+ ad nye ross multiplying in equation 4.2.17 (all pole function) we gets W@)n-o7527 + 0125271 = XE) @)t-0.75 2-1 + oe ate we = x0 +075 (42.19) uation. Fe 42.11 shows the direct form realization nr eq : ‘ss multiply the terms in equation Loe sca) me7 NE) Sea abe 5257 ya = (7 o oe b ws (4.2.20) "king inverse z-transform of above €4 4M Digital Signal Processing equation. Fig. 4212 shows the direct form implementation of above 69 z x) rm) | anes ee } > Direct form reailzation of equati. L 5s Fig. 42.12 Direct form realization of equation 45, Fig, 42.11 Dirvct form realization of equation 4.2.19 realization of H2(z). The realiza Here note that Fig. 4.2.11 is realization of #;(2) and above figure shows see ae oO. ee te . (2) = Hy(2) - Hz(2) [Le.“equation 4.2.16] can be obtained by cascading 2 ‘Such realization is shown in Fig. 42.13. ‘Azer esiem| Fig. 42.13 Realization of H (2) = H, (2); Hz (2) In the above figure observe that the delays can be' combined into two. Then the realization becomes as shown ia Fig 42.14, In the above figure observe that there are.two : 7} delay elements. The order of H(z) is also two: Hence it is canonic realization. It is also called i as direct form II realization of IIR filters. | The direct form-II canonic realization for generalized N and M is given in Fig 429. Earlier we . obtained the values of | ay by bara and ap. By putting these values in SIRE Fig. 42.9 we can directly obtain ‘the Fig. 4.2.14 Direct forml canonic realization realization of Fig, 4.2.14, tun Fitse Boston pe ven erence eQILON CAN Be eno, YO) = Byer at C AYO~2) + xq! goon the direet forml structure ofthis ar(rl) difference equation, 32 (224 ) - O17 5H y Sager U8 >) ols i Fig. 4.2.16 Direct forma structure of given system Cascade Form Structure for IIR Systems Consider the rational system function of equation 41.2 ie, tt Eas ww (42.21) et aN ~ 422) Xone HAC) kel Qo K w= (42.23) bbe connected in cascade to obtain We know that the system functions Hi) Hal "alization of H(2). Thi is shown in Fi: 42.17. Now each (0), H2(2 can also be written as, wi), 4 AO = 8O.43 = Hy()Ha® MQ. or : Let Hal) = x@) Tra? + 407 “A This is all pole second order subsystem and, ae Hat) = Babes hae" * ha? wt We have inzuoed the peocedire for ebaning diet fom Ii st SUBSEON Proceeding 6 oe ne ng aoe ymin the direct fore structure for H,(-), which is spite into 190 functions given by equation 4224, ‘chown below in Fig. 42.18. equation 4.2.25. This direct form-II structure is Fig. 4.2.18 Direct form-t realization of second order subsystems used In cascade connection of IR systems ‘This cascade structure is described by following equations. amg(n) = — ayn wy(0~1) ~A42 e(-2) + 94-1 (7) ~@ yp) = eo y(t) + Bia (1-1) + Dia 48-2) ( ‘This equations represent the second order subsystem of Fig. 42.18. And cascading is represented by follow ‘equations : Y-a(n) = x4(n) H(t) = see1(*) « soli) = 26) , yr) = (0) ‘The above equations represent the cascading of second order subsystems as shown in Fig. 42.17. PR er esas) oS 45 y UR Fitor Design the ee ‘ ww (82.29) we ot ese ro enon ar nthe form of uatn 4225, They ae writen ao H@) = Tea 1 4230) anel4 : atta \) Hy(2) = —mthaz = 2 (4231) AG) * Tanrleag?t 1 Or me 12 4231) 2 ay eclia 3 aaiuaton of equation 4.2.23 is given in Fi Jo.18. The above two equations canbe realized BY Wn Fig. 42.18, pet9 shows the cascade realization of H¢ te) Sa ao! =F \ Scar o gei-4e tiie? 3 Here He pub ay ogo Fort ane bby 4, by = Uy dy 2 53 1 and gj =~ Ba = pasa — gy Fig. 4.2.29 shows direct form-II realization. Fig. 4.2.29 (a) Direct format realization Fig. 4.2.29 (b) Direct form4 realization Fig. 4231 shows direct form structure. It is obtained based on structure of Fig. 42.4. Here by = 6,6, =7,b,=1 and ~a, = 075,~4, =~ 0425 li) Direct form - it Fig. 42.31 Direct form realization Fig. 4232 shows the direct formII, realization of given difference equation. 1) Direct form4 The given system function can be written in polynomial form as, zh -2e3 22 +22 23 + 0.56252" - 0.06252 + 0.0313 H(z) Here bg = 1, by =-2,b =-1, bg =2 ned ay = 1, a= 05625, ay =-0.0625, a4 = 0.0813 Direct form-I realization is shown below. rect form structure of f(x) of example 4. oe Fig. 4.2.36 DI [Zaria PUBLICATIONS” An up tua for knowl MW) Direct format Pig. 4.2.36 shows the direct form-I Fig, 42:36 Direct form roazation of M() of example 4.2.8 “mosmwit Ur nz) OF oxample 4.2.8 gece On ar tet ‘elton: ) Cascade form realization ‘The given system function is, Ga? +5249-3)_ (2 +6245) @-624+8) Ht) = H,@)-Fa(@) Fig. 42.39 shows the realization of above equation. (Please see Fig. 4.2.39 on next page.) 2) l) Direct form-tl realization ‘The given system function is, (e-1)(@? +52+ 92-3) zh 423-122 -% +18, = GI +524 0-9) FE HO = (Fy 62+ -E+9) zee? 232" + 18+ 40 7 Se ares ple 4.2.9 Fig. 4.2.38 Cascade form realization of H(z) of exam +184 Fig. 4.2.40 shows the direct form-II realization of system given by above equation + Solution : |) To obtain difference equation The given system can be expressed in pol ee He) = UR Fiiter Design nee X@) ee 46 +0,04692"4 (z) = YG) ~ 1.14688 2°? ¥(2) 405913 2° Ye2) es 4 8.0469 2S Ya) + 10 X(a) + 833 2°! XC) = 20 2°? Xlz) + 6667 2-* C2) jqverse #transform, yin) = 1.875 y(n ~ 1) - 1.14688 y(n - 2) + 0.5313 yin - 3) ~ 0.0469 y(n ~ 4) +10 x(n) + 8.33 x (n— 1) ~ 20 x(n — 2) + 6.667 x (- 3) required difference equation. in direct form-| and ll pss he to realize wwehave, by 10, by = 8.33, by =~ 20, bs = 6.667 ay = — 1875, ay = 1.1468, as = - 05313, ag = 0.0469 Fg 4241 (a) shows the direct form-I realization of the system. Fig, 42.41 (b) shows the direct form cfthe system. realization roallzation of H(z) of example 4.2410 sae enalvation .. 42.44 Diract form Bi av} VO Het VOe We =HOe 52H et gs heixe np feraye Fora ya 220) 5209 2 o- egy veto sek we 1)- Sy(e-d)- ype" x(i)+ £ x01) 14 realization Fig. 4.244 Direct form wey = “1247 shows the realization as Pet above equations Fig, 4.2.47 Implementation of network with h(n) = e/°0” peice: beostayr(n gdcion: yn) = 2bcoseny(n—1) — bPy(n—2) + x(n) - beostns(— 1) ‘king z-transform, | Y@) = 2beoswge 1Y(2) ~b*2*Y (2) + X(z) — beosuyz”1X(2) sity Sanptety suiuat Boe gaye Y@ 2 —_Aabeosag? eae yt HO) = 50) [ahconayg) Pe? Here by =1 heb and; ay =-2b and a, =? Direct Form:|.; ed ¥ - a ATONE tse an iota Sinal Processing co Ait Direct Form By = 4b =3,-m = 0, -uy #2 ‘The difference equation is given as, Mon) = ~ayyl-1) ayn 2) + byx(n) + 52(0-) Putting values in above equation, Yn) = —2y (n-2) 44 x(n) +3 x(n-1) (1 lt ciple)” 3 eye) her rs) (+z) (e)otrectfom -elenation (0) Cascade relegation ep eee a Fig. 42.56 Realization of H(z) of example 4.2.1 GAL PUBLICATIONS" An up tnt or nowede pares 4-37 IR Fitter Dosign TY sgl IR filters are usually designed from analog fiers, Fist een de system function fs converted from analog to digit analog filters are designed with the help ee s Analog Filter Design using Butterworth Approximation VID Ee MarelvO Fab s05 ZAUICOB) analog filter is designed for given specifications. lomain to obtain the digital filter. Of standard approximations such as butterworth, chebyshev, elliptic ese {Why filter approximations ? ‘The ideal low we Fesponse shown in Fig. 43.1 is not physically realizable. Hence the response is approximated with the help of standard functions such as Butterworth, Chebyshev, elliptic ete. «The magnitude squared frequency response of the Butterworth filter is given 25, |H@P = —1 uf 2 . wey fe, Here N is the order and Q, is - 3 dB cutoff frequency of the filter. «s Monotonically reducing magnitude response : «Fig. 432 shows the plot of equation (4.3.1) Its plotted for different values of N (order). (43) Fig, 4.3.4 Ideal lowpass filter CREED cette tet ie rene hh dri nr. 2 Response is monotonically feducing. SHAQ) = 05 for =, for-allN. |) KER Poles of Ha(s), ; sd el | H,(9 6: the aystem function of ‘spalog, iter. The poles of Butterworth approximation can be falculated by ~~ (43.2) (oh The poe of Hf) ee hse Sch en Be ag SY ofthe spon Poles of Butterworth filter lie on the circle in slane- FE) ordar oF tha Buttorworth Fitors Fig. 433 shows the different attenuations and fe filter specifications are mentioned as follows + Ay S\Hq(Q)|s1 for 052 2p TZLTnNGAL PUBLEATIONS™: ap Ort fr roweape .quencies represented in Butterworth filter approximation, The IRAQ) S A, for , 50 Here Ay is passband attenstation, A, is stopband attenuation, 2, is passband edge frequency: 1D, is stopband edge frequency 1 ek we (43.3) Not Ap ® and A, (OEE i oe ution Ay nd Ay ae they ae nea. “Teele ‘No the buttroorth ler i given a, ig. 4.3.3 Frequencies and attenuations Fla. 4 Butterworth approximation for As/Ap linear for Ay, A, in dB “Above equation can be expressed in terms of '¢ and ‘8 are as follows : [EEE cutoff Frequency of the Butterworth Filter ‘The cutoff frequency is required to calculate poles of Butterworth filter. It is given mers et prone | Qy 2 aor it < ABN aoMas yy BN Apanda sind or ¥ aes prey ~@ of ~& 3 utterworth oraz’ B Filter (Prototypo Fitter) putterworth filter 2, = 1 rade. we a Sr dcte Normalised las Gn bevel we wae one oe of potes or systemt plied the frequency transformation to get the desired filter. o aya Function Ha (s) of Butterworth Filter pepe obne By estion (43.2 are aranged incom iplex te tem funtion of Butterworth iter canbe expres soe pa a toe poe ei OF as, a then the syste ™ © eae a a | SEAMS PM ehy a cebarhy 4 Tacos orm fer | 439) pox romaine Duterwot fils, 2, = 1 in the above eqn, The plyomial values for normalized pafterworth filters ae obtained further in example 492. im (|S 02. for F 2 5000H= eu and F, = 1000 Hz A,=02 and F = 5000 Hz and Goce 9, = 2H, =m 000H2= 20008 124/see and 0, =2n F, =2nS000H2 = 10,0008 24/80 Step 4: To obtain order of the filter a el ae] 1 ae) iG ee iar Wot) ose Ne tog Oe “20007 ee in Step 2: To obtain cutoff frequency Cott frequency is given by equation (437) a5, Ne 9, | 200%, 10,0008 a ot 2-5 5 7 Ala yee (4) ee aA ae? oF i) isi 4 3 x rad/see 2 107245 rad/see = 34037 Sip 3: To obtain potes of Buttewort filter Poles are given by equation (432) 8) seenucal PUBLICATIONS: AH oat fr knowlege Diattal Sianat Prooassing ab | ye Ang dONNDN, fe Oo | ee 0, = 107245 and N=2 above equation becomes, : Me / py = e107 daHEADA IO, f= 01 ‘ : = £10745 A 8/4, k= 0/1 ef — yp = #107288 o9%/4 =—7588-+ {7588 and 7583 ~ j 7589 Frk py = £ 107245el@4¥9)"/4 with k=1 in equation (4.3.10) = #107245 d9*/4 = -7583 7583 and 7583 + 7583 Here observe that we obtained four poles. The poles lying in left half of s-plane are as follows. ‘9 = ~7583 + j7583. and sj = -7583 - 7563 Stop 4: To obtain system function H,(s) ‘The system function of the butterworth filter is given by equation (4.3.9) as, a aN 10724.5? ~ G=aye=a)(6 =e) Ee) 6 ¥75ES ~77588) 6 +7583 + 77583) 5 HA) 107245? 107245? © G7583)? +756)? | 57+ 151665 + 2047583)" ‘This is the system function of required butterworth filter. Factors stl s+ V2 +1 (2 +541) +1) (2+ 076536 s+ 1) (6? + 1847768 +0) (+1) (7 + 061805 + 1) (2 +16180s + U) (240517684) (2 = VB) (s? +1.93185+1) (+1) + 048505 + 1) (G2 +12456s+1) (6? +1802 5+1) py @eun9s+D e+} Let 9) = oe Pas ea MPM the tamara fa ssh wha passband of 1 radjece. Use transfortation to ji the tran ber gir fests of the a analog filters.) A Towpass fier with passband of 10 rat. 1) To obtain lowpass filter with 1p = 10 radisec Lowpass to lowpass transformation is given as , 100 1 = HQ) = H.(), pe ay A HOES +100 ‘10 ll) To obtain highpass filter with cut-off frequency 10 radi/sec. Here we have to apply lowpass to highpass transformation with Qjzp = 10 rad / seci.e. ww 8o5 1 2 H,() = Hohe! * yoo, 10/8, ~ F +10 fFs+ 100 5 +1 $ i“ op upFilter Design by Bilinear Transformation ae pinch: ‘The bilinear {tanstorm maps complete : j © axis of s-plane on the unit circle only once. Hence there is qo aliasing of frequencies. Bilinear transform is a confirmal mapping. (Oa s 77 Transformation ‘he bilinear transformation is given by following equation, (tet Tlie) i proof : Step 1: System function of analog filter Consider the system function of following analog filter b (4.10.1) HO = 3a sos (410.2) ye). b XG) #8 6) [s+] = # X00) avGe) +a Y06) = BX(6) (4.103) Stop 2 : Differential equation of anal08 filter Taking inverse Laplace transform of equation (4103), 410. gayi = BO» ree 2 SP =s¥0) oy tt ie, +g) = ta Bete yo : ncn PLEATINS 9 tb nee Distal Signal Processing 4280. Rr, Let us evaluate above equation at t = nT. Then, VOD+ay(nT) = b xlnT) or vloT) = ~ay(aT+bx(07) ~ (ty ‘Step 3: Approximation of an integral by trapezoidal formula ee Let y'(t) be derivative of y(t). Then y(t) can be integrated to obtain y(t). ie, ' w) = Jy) der ylto) ‘o x Ce at mtg integral can be approxi Here derivative y‘(x) is integrated from its initial value at f = to Above: integ PPPOximated 4 trapezoidal formula, at os take two pots of tapezoidal formula along time aXs at f= nT-—T andy 9 ‘Then trapezoidal rule approximation of above equation will be, wot) = Diy + yor) + yinr-1) oo (dag ay Here ta) = (nT - 7) Stop 4: To obtain difference equation for analog filter From equation (4104), y(n? ~ 1) = ~ a y(n ~ 1) + b x (nT ~ 1). Putting for y(nT) from equation (4.104) ang ¥(oT~T) from above equation in equation (4.10.6) we get, MOD = F ayer + bait) ~ay(nT—T)sbx(at-T]o Mnt=1) Here let y(n) s y(niT) and y(n —1) = MnT— 1). Similarly x(n) = x(n7) and x (n - 1) = x(n ~ 7) as standard shot ‘hand notation. Then above equation will be, yin) = Fray + bx(n) ~ay(n-1) + bx (n-1)}+ y(n—1) H+ Tyd—voy + Tynan = Fain sina) (1+Z}s- -F) v0 = Fete stnny Step 5 : To obtain Ht) by taking z-transform of difference equation ‘Taking the z-transform of difference equation obtained by equation (4.10.7), (+F}0-( -F)me = Frese aw (100 ar [+F-(-F) ro = Zaseyxy ye Fay 3) * at a T, 7 Hq) = 27042) =¥@) air Here tie) = 38. e Fass (ese Faye TECHNICAL PUBLICATIONS” ‘0 09 trust or kromodge ey up Fitter Dostgn o e = 7) een . lat é wos (410.9) ee ccc sneurnan by fo re psi of nega lim ue Plane to z-Plane Mapping ee know that s'= + j& and z ='r od”, Putting We JQ and z =r e®. Putting these values of s and z in equation (4.10.9) we get, Vere B(rehat) _ 2 rleosorjsina)-1 _ 2 reosu=1+ jrsine ré41)] Tr(cos@+jsina)+1 ~ T reos@s1 + jrsino 2 2, Teosw-1+ jrsinw , reosw+l~jrsinw T reosw+1 + jrsino “rcoso+1-jrsino (7? cos? w+rcosi- jr? cose sinw-reosw—1 2 +jrsinot jr? coswsinw+ jrsinatr? sin? «) (reos@+1)? + (rsina)® Poot w+? sin? @-1+ jarsino _-2 lees? w+ sino) = =1+ jasino r Zeogho+ Drcosoti+r? sin? 72 (cos? @ +sin? a) +1+2r cos " sis 2 Potsjrsine ye od orsnt oot T +r? +2rcoso 2 Soret es =F [yrPa2reoso ot s+ (4.10.10) Praise tal eed oe aos ak O = Tiptree 2 _xsine aa : soe (4.10.11) and. O = FZ jp Fe treoso . . biel fer ce jane : From equation (4.10.0) Hf = 4, then @ = 0. In other words, j Q axis in s:plane * Mapping of j @ axis of s-plane * jane (72D (6 = 0) is mapped to unit circle i plane ation (4.1020) Hf > 1, then o will be positive (o>0. In other words, qual Mapping of RHS of e-plane + From et ‘unit circle im zplane (r > 2) od outside, f hh ad of pe (> 0 up a rn RIGS * oa coe enh ng naan yan eas ee a’ = IR Fe ‘Digna 4-92 ey tion (410.10), If x < 1, then o will be negative (© <0). In othe , in zplane (7 < D ‘ * Mapping of LHS of s-plane : From equat Ugh hand side of s-plane (0 < 0) is mapped inside unit circle CUED ThE espping BF sinitar ta that shown in Fig. 4.9.1: But ‘iinzar transform frequency. relationship i given by equation (4.1 U + As per the mapping didcussed above, bilinear transform maps all poles in left hand side in s-plane inigey, unt circle in z-plane, Thus stable analog filter is converted to a stable digital filter. ww 5 ERDEW Frequency Warping Consider equation (4.10.1), hoe 2 2rsinw aoe traing T 141? +2rcosw To obtain the relationship between continuous time frequency (2) and discrete. time frequency (a), we have ty execute above equation on j & axis. For this r = 1. Putting r = 1 in above equation, oe in cos a= 2. 2sino 2, sino _ 2 2sing cosy “Ta Sei, By trignometrc relation (41032) oo 41013 s TAP one Gtuation gives the relationship between discrete time frequency (a) and continuous time frequency (0), ; 4.10.1 shows the plot of this relationship, Sea an nonlinear raorhip beeen continuous time fequeny (2) and discrete time frequency (0) is called i weney frequency warping. In Fig. 4.10.1 (a) observe that relationship between @ and w is highly if ELT laseaets et on Romney tans | iP Fier Design of warping can be compensated yj : th the help ie from digital ler trequency (0) withthe hen gg on PMMA: The analog fer frequency (0) 6 ae 2,0 1 of equation (410.12) te, ve at ow equeney Bands of dia fiter wit he oe pd bgh frequency bands will be expanded. oe nen de bilinear transformation is sppied on prevarped analog {eos me ies, the higher frequency bands get compressed, which were panded due to prewarping. Thus overall effect is nullified, And the |-—{—i/-- f equency mapping becomes nearly linear, 2 es = = “Fig, 440.2 Prowa fear Transformation Advantages 1, The mapping of frequency axis is one to‘one on the unit circle in z-plane. 2 Aliasing effect is absent, since the mapping is one to one. advantages and Disadvantages of | 4. Bilinear transformation maps poles as well as zeros. | 4 Stable analog filter is converted to stable digital filter. | Disadvantages: “1. The mapping of frquency axis is highly nonlinear. It s warping, 1. Additional processing such as prewarping is required to reduce effect of warping. TEE comparison between Impulse Invariance and Bilinear Transformation Methods Sf olan tiger fet of algal er wing aie Tse ace : Uy 4.04 it Digte! Signa! Processing Solution : Here) = by [Biineat transformation Is given 98, 2-1 20-27) with T= 1 see Tit ae) ‘The jpnsfer function of digital filter wil be, 1 ao HG) = Ho), 1-21 =z 20- 20-7) Ie) = Hel, Pi -eD her ston” + Obtain the system ESSER ten fino he a fle on ws 1 = 136 fon Aigital fer using bilinear transformation which is resonant at, = me | Solution : From the denominator of H, (6) we can write the poles of analog filter as, (+ 0D? +16 2+ 01-j4) (6+ 01+ j s = -Q1+j4 and s ‘There are the two complex conjugate poles. We know that Poles are, a1 ja + Thus the values of 'o and 01 and Q=44 A function is said tobe resonant at its poles. Hence H, (s) will be resonant at, s = -014j4 In other words we can state that H, (3) will be resonant at Q= 4 It is resonant at @, = 3. required that the digital filter should 2 This means the bilinear transformation should map 0 = From equation (41033 the relationship between ‘Q and ‘Wis given as, 2.8 SSNs ‘ a Fs . TH dtm Putting for Q=4 and o, == 5, of analog filter will map into w, = of dist! fiven by equation (4.10.9) as, Putting for this value of ‘s! in H, (9) we get H (2) Le, H® = #6) ern 4-95 UR Fitter Design | cxsmpliying the above equation we get, : Hq) = U28+ 00062" -o1z2 _ 01282? + 00062-0122 1+ 000062" + 09752 ~~? + 00062 + 0975) ‘te mots of denominator of H(z) are poles of H(z). They are located at z=~ 0.0003 + j 09874208 and (90008 - j 09874208 converting these poles to their polar values, 2 = 09874298 )1S7H100 elnow that z=rd®, hence, r + 09874208 and w= 15711001 = %, Thus the two complex conjugate poles are baad t=. Hence H (2) will be resonant at o=5, SRE De «wt ith’ 3-B digital Butterworth filter of first order by ising bilieat transformation. juton: Here @,=1 for 3dB, N=1 1)To obtain 2, by prewarping 2. we _ 2, 1 1093 O, = plang = peng. ii) System function of analog filter of 1st order Qe .093/7) Hy) = 570, = 5+ 003/D iis tic iii) To obtain H(@) by using bilinear transformation He) = Hol =) tay Diatal: 98 : Signal Processing 4 % oes sp yd iHerworth filter, into digit Hf Fk CEST con ai er cin’ of he 2 “rer Bil a 4 bilinear transformation: iven 05, 4 Solution : The system function of normalized ond ‘order butterworth filter is gi ; ; HO) * Bev : Bilinear transformation is given as, = 071 + 2278 42-2) -1 © 1-0.7662T+ 0.27722 ] ee Bilinear transforination is given as, H(z) = et ] Applying this transformation to given filter with T = 05, A422) Ho 3 = 0571+ 2271 +272) 2 (1-24 1-0.4.T 0.2632? : [ilies ES] +f )-3] GEREN Using the bitin 3 ed DP is the 2p cecal ee te wsliget EAE CELaSanTe The s-plane value is, s = /2 Bot tet = cos + join = j Putting s = j in equation for z, pails yd = Lene? a lepet tee 5 ty "2". tases ice ev escae tee, Tndsnc of a=; it z=j in zplane, mis 8c f O=1 in the s-plane is mapped TECHNICAL PUBLICATIONS * Ant trast for kroutedige processing so 4-99 nn Fito Dosign e Transform the analog filter H(s) w wate? a aoe Cantey M08 Hata ier sing tincar transformation. The digital ee wae. ROE Step 1: Given data et Hg = —tL_ + @+0I) +9 step 2: To obtain pole frequency 2 : +01)" +9 = (G+ 01-j3) (+0143) pales are at 3 = -014j3 Here's = - 0.1 and Q=+3 Step 3: To obtain T __ 81629 1 ~005372"" -0946@"%) 1-036%71 + 09875"? my ty 2 g into gl fier whose ee futon is () = Teeneag) HN MNO EEEEERIED cncrt the analog Ate : . TTT ‘ansformation. with T = 0.1 sec- : Solution; HG) = HO). 2 © Grier Thad Examples for Practice asin aan: Apply bine range” HO exer nT jossras 4%, “d= 0sse 1) : ceatha eeeernmmen SS tal Signal Processing 1000 Digital Signe fil that has 3 dB banat of ‘Hz A on at, VS een the aero he ples of BT aig Too ‘of 20 dB et 2000 Hz. Find the system function H (6) by bili z Solution : Given data : } A, = 34Br F, =1000Hz } a Ay = 2008, %=2000H2 1 Rep = = 1, Hence. Fer T= 00" c Here Fr is sampling frequency. Defining specifications of digital filter : : Let us convert the analog frequencies to:their discrete time values. Le. | ™ J _ 1000 > Sp * Fag ~ 30,000 S o op, = In fy = O1X2R= O27 F _ 2000 _ and fo = Fee 770,000 97 & ay = Inf, =02X2n=04n ‘Thus we have the specifications of equivalent digital filter as, Ay =3 4B, Z ‘Ag =20B, } a ‘To obtain specifications of equivalent analog filter for bilinear transformation : Here we are using bilinear transformation. Hence frequency. specifications of equation (4.12.5) should t converted to their equivalent arialog values according to bilinear transformation. We know that the frequeny relationship in bilinear transformation is given as, Thus we have the specifications of equivalent analog filter according to bilinear transformation are : a : ‘Ay =3dB, 9, =649839 } 4 ww As=20aB, ” Q,'=1453085 To obtain order of the filter : Here attenuations are gi in dB. attenuations are given in dB. Hence we can use equation (4.3.5) to determine order. ofthe filter TECHNICAL PUBLICATIONS". An up tsa fora IR Fito 4-108 oy ter tat Has 3 4. band if 1000 Ha on tony Sf (GENTRY Determine the onder and the poles of a bulterworth es ranean 8 20-48 ef 2000s ind the syste funtion 11) by ne transformation 12198 * 7 T0000 Diptal simal Processing Solution : Given data : 1 1000 Hz - Ay = 3D, Fy #100 | Ang Ap = 20dB, = 200012 1 se. Fop = 2 = 10,000 T= qghgp’ Hence Brey Here Ky is sampling frequency. Dofining specifications of digital fite E Let us convert the analog frequencies to their discrete time valties. p= Bm fb = Fp" 0,008 (wor =H)’ op = 2nf, =01x2n= 02% and = F220, Fo = Fe “70,000 "9? 0, = Inf, =02x2n=O5n ‘Thus we have the specifications of equivalent digital filter a, A,=3dB, - o, =02n (42g As=200B, 0, = 047 To obtain specifications of equivatent analog filter for bilinear transformation Here we are using bilinear transformation. Hence frequency. specifications of equation (44125) should be converted to their equivalent arialog values according to bilinear transformation. We know that the frequency relationship in bilinear transformation is given as, Saw a= Ftang 3 Zane : of RTOs We have 4:=10,000 and’ @, = 027 hence Q, = 210,000 tan SS = 6498.39 simialy, 2, = Bian «220000100 4 = ass085 Thus-we have the specifications of equivalent analog filter according to bilinear transformation are A, =3.4B, 2, 649839 } ‘A,=204B, 9, 1453085 1 (4100 To obtain order of the filter : Here attenuations are given in dB. Hence we can use equation (4.3.5) to determi rie fe r TECHNICAL PUBLICATIONS”. An up thus or knowiodoe 4-407 A Filor Dostan bel ae ) “, a, Pp pang vavous values from equation (127) in above equa ; © equation we get, 19x: 1o w= 3, Sho og (Ses) z CRIT N = 2858 the order of the filter cannot be frac iter should be N = 3. ion, we have to round ‘N’ to the next higher integer. Hence order of to determine cutoff frequency 2, : We know that Q, is 3 dB frequency. From specificat ; : : = . be verified from equation (4.3.7) in as, jons of equation (4.12.7), 2, = 6498.39. This can From equation (4.3.4) in we know that (s a} ge fe .) hence above equation becomes, 7 Dy wae a, = Puting for 2, and Ay from equation (412.7) in above equation we 6, 2, = 0822. = 6498.39 fe hoa ‘Thus Q, is same as Q, since it is 3 dB frequency’ To determine poles of Ha(S) + We have seen that poles of butte Pk ‘above equation becomes, é a42He/6, k= 0,1,2 wn (4128) worth fiter are given by equation (432) in a, ba, NAREADAAN, b= OD NT Since N = 3 and 2, = 6498.9, ! re = + 6198.29 With k= 0 in above equation we 6° poe +¢.6498.39 of 4/6 = ~ 924919 + 5627.77, and 324939 - j5627.7 With k=1 in equation (4.128) we get, = = 6498.39 and 6498.39 ast 498.394 * With k =2 in equation (4.128) we 6% S pp = £49899 Nae ncn PLENTONS a tal be A 8R16 w = 924919 ~)562777 and 324919 j 862777 IP Fer, left half of 2 axts in ey ‘pala tegen ts 4-108, tying in Here we oblained total 6 poles. Out of that we have to consider polet These poles will be poles of H,(9, then filter will be stable, £4 == 3240194 § 8627.77 } complex conjugate pair. 8} =~ S24919 = § 5627.77 82 == 6498.39 gate pairs to avoid imaginary coefficients, ‘There are the poles of H,(6). Always we combine complex conju To determine system function H4(s) : : Ee For butterworth approximation the system function for 3" order filter is given 2 HO = ) order filter. Putting values of poles and Q- in abory G5) @-4) 6 Here observe that the numerator is Q3 since it is 3% ‘equation we get, (6198.39? HAO) = SE SRBID= JST + IIIT FST TC + ED), (6498.39)? H,(s) = O” Rs ea 865 + 42229 108) + BE) To’ determine H (2) using bilinear transformation : We know that bilinear transformation is given as, 2(i-2t Tliezt We know that 41000 then above equation becomes, ayt Ee ance a ] ~ ns Teet Putting for this value of in equation (4.1239) we get H(2). te, (6498.39)° H@= (On simplifying above equation we get, el 0.0343 (1+ 27! H®@ = = > ) 48 -17695=71 + 0782) (15n9 -arsI eT = one (ie2e1422) (ise) ‘This is the required system function of digital filer, Design digital Nighpass buttertorth fer for citof frequency = 30 He Solution : Given data : Highpass filter with “Fyp = 90 Hz and B= 150 Hz Gnd sampling: frequency = 150 He TECHNICAL PUBLICATIONS". An Up tual or mnowindpo onder of Blter to be 1. Le. N= « 5 vn pecifcations of equivalent digital tor: aie equency of digital highpass iter wl be given as (f ab -F)ie, Sao = Hip 30 “7159702 cycles/sample 4 Opp = 2n fyp =2nx02=047 radians/sample of These 0 obtalt Hore WE relationship is given 85 are specifications of equivalent digital filter. isformation ¢ relationship: THIS feaency in tho specifications of equivalent analog fiter according to bilinear tran: vl redefine the frequencies of analog filter according to biineat frequency one 2 OQyp = Fen} 1 seg, and Opp = 04. Hence above equation gives 150 ow = 75 wan (5 ) 2 217963. radians/Se°- (os) sus we have the specifications of ansloS fate acordng to biiese transformation yp = 217968 and Nel ‘ . ove that the procedure accused Kee & sometimes called as prewarping: We Se obtaining the oe frequency of analog filter. When W° sppy bilinear tarformation © sete warping effects compensate for prototype first order IOWPSSS fiter function of normalized lowpass filter (e+ To obtain system function We want to design highpass also called prototype filter) and then frequency he normalized lowpass SE is given as, ‘Ths eystem function of the Le ahead +b, s+) Hal) * Wine eles ot py equation (43.19) sn (412.10) oo +3 i sowpass butterworth iter itor by frequency ‘transformation : : ation on aystem funetlon ‘of prototype lowpass filter to obtain system or gansformation is given by Table 45:1 sy get highpass filter system function. To obtain system ni Now we have to apply sequen ane joe function of highpass filter. The 1 pass to RIBRPAE O, Que tm Diet Signal Processing 4-110 Fae ong * ‘1’. Hence Q, Here Q, is passband edge frequency of lowpass filter, which is equal to bp =I. Hence ating ‘transformation will be, 13 Me : .. Hence, ‘We have obtained Quip as 217,963 ie. prewarped cutoff frequency of highpass filter. 217.963 2 Hence system function of highpass filter is given as, ‘ ~ (421 ‘This is the system function of the required digital highpass filter. 1 a 7 Hig ©) = Hon @h-sSH2 = 25,2965 ” HIBS, * FF HIS mie 27568, ‘This is the system function of analog highpass filter. ; ‘To obtain 'H (z) by bilinear transformation : ‘We know that bilinear transformation is given as, 2 2(teet S* Tlaiset Digital Signal Processing (GEEBTIRERID Desiga ar‘ aigital iter that hier g a Polloiving’ equinalent analog specifications : eo ‘AB or greater at 1000 7 rad/sec 5 (0 LPF with -1 dB cutoff at 100 x radfoce | (i) Stopband siecle Se ae < (G0 Monotonic stopband and passband (Go) Sampling rte of 2000 «ee eSPTRTTETS Solution : (i) Given data: oe is butterworth approximation, The passband and stopbands are monotonic. This characteristic #8 obtained by Hog we have to design butterworth filter. Sampling frequency Fer = 2000 Hz Passband attenuation, A, =1 dB for , =100 m rad/sec or Fy =50 Hz ‘Stopband attenuation, A, = 35 dB for &2, = 1000 x rad /sec or F; =500 He. Hence we will assume bilinear transformation, Here digital filter transformation is not given. (il) To define specifications of digital filter = F Let us convert the analog frequencies to their discrete time values i.e (vit =p]: 2n f, = 2nx0025 = 0.157 5 So Fe a0 0.025, Hence o, ~B _ 500 f= a ‘Thus the specifications of equivalent digital filter will be, Ap=14B, —@ =0157 } w= (42 A,=35 dB, @,=157 (li) To obtain specifications of equivalent analog filter for bilinear transformation (prowarping) : Here we are using bilinear transformation. Hence frequency specifications of, equation 4.12.16 should be converted to their equivalent analog values according to bilinear transformation. This is called, prewarping. The bilinear transformation frequency relationship is given as, ad \r ‘an 2 We have seen that 2 cancels out during calculations. Hence 2 can be considered 1. Hence above equation wil be, =0078 rad/sec 0.999 rad/sec ‘Thus the specifications of equivalent analog filter according to bilinear transformation are Ay=1dB, 2, =0078rad / sec Ag=354B, 9, =0999rad / sec (iv) To obtain order of the butterworth filter : Here attenuations are given in dB hence let us use equation 4.35 to obtain order of the filter. ie, TECHNICAL PUBLICATIONS”. An up tnt or node UR Fitar Design select N=2 esas oe frequency is given by equation 43; 7 1 7 (3 aoe a >" ion 43.4 1 4 be} gam equation 434 we know that 7 -1(y8i4y 8 e 1) hence above equiton canbe wate a, Oo, = en. Pearl 7 + (0 oe =F (Cr “iy patting values in above equation, nie 0 0.121 rad/sec ww (4.12.18) (v) To determine system function of the normalized (prototype) lowpass butterworth fier: he normalized butterwvorth system function is given By esuaton 439 as, nena Ga with b Han @) = Wy byas™ +b, 841 1 een Ha = Fanon 22 Ikis hy = V2, Hence above ‘equation becomes, from table 432 for N=? The value of by can be obtained ea re Han @) = Fa sis+2 nction of required analog filter by frequency transformation = junction (vil) To obtain the system f ‘order normalized: lowpass filter (equation 4.12.19), We want the We obtained the system on oa af Hee i we ow vo lowpass frequency lowpass filter having cur? ete transformation. It is given #9? er = Bee 6? Ov atid iter, 1 Op a1 rad/see and Qyp is passband edge rroqueney of TT aan. Henee save transformation will be, Here @, is passband °48° Y yp 2 08 Diatal ae Siena! Processing 4-120 tr 0146 © P+ 0171s + 00146 * Way This is the system function of equivalent analog filter. {vill To determine H (2) using bilinear transformation : ets i mation is given as, Next step is to apply bilinear transformation to H,(¢) to get H(¢). Bilinear transfo : We have considered 2 =1, since it cancels out Hence above equation will ey ae Applying this transformation to equation 4.1220 we get, 012 (1+2" {1662 + O77 He = Hu), 22 ter + 0146 ‘This is the system function of the required butterworth digital filter. VY Design’ a lowpass’ digital filter using’ bilinear’ transfortiation.. The. filter’is'to be monotonic in both stop ant passbands and has all of the following characteristics : 5 D An acceptable passband ripple of 1 dB, ii) A passhand edge of 0.3 x rad, and ‘ ti) Stopband attenuation of 40 dB or greater beyond 0.6 7 rad. ties Solution : Given data Ay = 14B Ay = 40.43 0, = 04 Step 1 : Prewarping 03: tan % w tan 93% = 0.509 rad/see O,= tan % = tan 9S% = 1.376 rad/see Step 2 : Order of butterworth filter (Monotonic stop and passbands) : 1.376 (10100 .4)28 ————— eo ein Fg TH aNeiON OF normale ana nro Doves 7 0g wbles of normalized by ce ea fot Ng ter. Ha () = china E+ 0.517; Grmye an FHSIee ep 52 FEGEREY transformation ; s+1) ops 10 lowpass transformation i given ag s Q, 7 aps 9, +} for nommalized filter and O45 =0, «og 5 06 #@*HaOkse = 1 : (a) sosur(ga)eal (ge) 2B] (se) 2 a (06° * 0315+ 038) (7+ 0848s+ 036) (FH I866+ 038 step 6: Applying bilinear transformation to get Hl) : (t= 20.58 271 + 0.231 27 ‘ansformation for. the following Specifications = CEE Dien a ee : : BH: : ‘ SEE _ |ite®)|<02 for a6 nsost ‘Sittin : Given data a A, = 08 = 1.93 4B Age 02. = 14 B Step 1: Prewappring * 4 en pe 2 Digital Signal Processing 4-122 Step 2: Order of filter: 10P14saB_y 1 °8|- orapw, a = 13=2 log = > Step 3 : Cut-off frequency Q.: Q, a, =} pg dt | 0808, _ 2 [tte (10m 4saeaye] “|e ayPe Ome BE Step 4 : System function of prototype lowpass filter : For N = 2, from normalized butterworth filter coefficient tables, 1 Hy) = =—>— © Fie Step 5 : To obtain H,(s) by frequency transformation : Q, ‘Lowpass to lowpass transformation : §—> ap? 5 HereQ, =1, yp =05 = 8 FE 1 a HQ) = Ho Oh § ° 7 7 5 2+ 0707s+ (ae) Step 6: To apply bilinear transformation : 0.25 027 (1+ 221 +27) He) = HO), lez? 1 ie) s ——, 98s ( ( =| 1-0.7662"1 + 0.27727 = 05 rad/see

You might also like