Scribd
Upload
32 views

DSP Test 2

Uploaded by

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

DSP Test 2

Uploaded by

Surya
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 22
OGHFormAGIb - Her. Neen SJcIT Pere 1 ai [ Internal Test Question paper format Fame of he staf: Ravi M V/ Manjula K Date:_02/12/2020 Cc Reviewer's Signature: NOTE: Only the following information's to be given tothe students. S.J.C. Institute of Technology Department:-Electronics & Communication Engg. Test: 10 . Semester: V Section: A Subject Name & Code: Digital Signal Processing (18EC52) Duration:1.5 Hour Max Mark Note: Answer 5 Questions choosing Ifull question each part cate Marks | co umber Mlustrate & prove the following properties of DFT i)Symmetry property ii)Parseval's theorem. 10M | COI OR TFx(n)=(1,2,0,3,-2,4,7,5). Interpret the following 19K (0) i1)XG) ii)E" Co CK) iv) Deo [XG |? tomo! a. Compare the differences and similarities between DIT and DIF FFT algorithms? am | coz 3 | b. Summarize and prove the two properties of phase factor Wr that are exploited in Fast Fourier Transform algorithms? om | coz 1 OR 3 Explicate with necessary diagrams and equations the concept] of overlap-save method for linear filtering. A long sequence x(n) is filtered with a filter with impulse response h(n) to produce output y(n). If 5 x(n)= (1,4,3,0,7,4,-7,-7,-1,3,4,3} and h(n)={1,2}.Infer y(n) | 10M | coz using overlap add method by using only 5 point circular convolution. 10M | coz OR Interpret 8-point DFT of the sequence, x(n)=(I,1,1,1,0,0,0,0) 6 |using Redix-2 DIT-FFT algorithm by showing’ the intermediate results. 10M co2 SJCIT 064Form#02b - Rev.? Page: 12 Devélop DIT-FFT algorithm and write Siggal Flow graph for N=8. 10M cc OR Solve for 4-point Circular Convolution of x[nJ={1,1,1,1} and_h{n]=(1,0,1,0) using radix -2 DIF-FFT algorithm, 10M _L co A filter is to be designed with the following desired frequency response w . ei, as" Xk) = Yoxrywyi wi" (since wi" = 1) aad N-1 yxy" aad ~ Hence, X*(k) = X(N-k) ‘The above equation conveys the message that the DFT of a real sequence possesses conjugate symmetry about the midpoint. = X*(k) i] IN is odd, the conjugate symmetry is about ¥. The index, k = $ is called the folding inde, This aspect is illustrated in Fig. 3.7 Conjugate symmetry: X*(k) = AWW = A) or A(R) = PUN ~ Conjugate symmetry: X(k) = X(N ~ k) we (LEE _ - # 0 2 3 5 eee) Not . 120. CHAPTERS © Digital Signal Processing Fig. 2.7 Symmetry of Xk) for x(n) being real. Conjugate symmetry implies that we need to compute only half of the DET values to find the entire DFT sequences ~ a great labor saving help! A similar result holds good for IDFT also. — siete Yori Serced ale q 3.7.10 Inner product (Parsoval) Proof: From the definition of IDFT, we have W-1 Lermym => P earay nel) ind N-I | -t rma Dy xaowy' def ‘Taking conjugates on both the sides, we get Hence, Corollary: If y(n) = x(n), we get W-I Yr'mrn nw Wet atin) = 0 Na dad NeI YL ayn = n=l Not yy Not > (3 Ds awe) y(n) 120 nw 1 “2! N-1 ww (= voy) imo ned NI I 7 LX ra) dad w-1 DietmpP n=0 The DFT andFFT =—-145, N-I 1 = VL wxe k=O 1 N-1 = FL ikwer We Thus, we have proved that N-I 1 We 2 2 Leo =y Lor n=O k=O slays eal sole 3 Re Ree BALERS TS 4,7,5),N=8, 1) To obtain x0) 1 DFT, XW) = Sawin Fn} With k= 0 and N = 8, x00) = Fxq =1424+043-24+44+745=20 : 0 : Above equation shows that X(Q is summation of x(x). Hence if x(n) is real, X(0 is also real. Il) To obtain x{4) = DFT, XW) = Sane m0 N With k= =4 and N=8, X(4) = Secs ) n=O = . = Fejy ction an = Baeye we Sere - Sayer ned = Seay since €F** = (4)" 0 = #(0)~*(0) +2(2)-2(3) +2(4) -2(5) + 2(6) 0) = 1-24+0-3-2-447-52-8 Ill) To obtain Fxn) : 0 141 IDFT, ata) = Ole 1 5x With n = 0 and N=8, (0) Digital Signe! Processing Properties of DFT and its Ap Sw = 8x(0) =8x1=8 iv) To obtain Sx? ke Parseval's theorem, be Nin ' = Deer N= 8, and real x(n) , = 83240) =9[2+20)+2 Ort?) vat B[l+4+0+9+4+16+49+25] = 864. EE ue ne Finesse eteneinnmncnenen te caesarean Differences between FFT algorithms _ Ae DET at the output jain bit reversed order. Decimates the sequence step by step to 2-point ce nnd cae DER pes i Properties of Phase Factor (Ww) © Now let us consider the properties of Wy. These properties are used by FFT algorithms to redu of calculations. * We know the N-point DFT of sequence x(n) is given as, = X= D xQwe, k=0,1,..N-1 n=0 © Here Wy is called twiddle factor and it is defined as, * This twiddle factor exhibit symmetry and periodicity properties, Periodicity Property of Wy x TECHNICAL PUBLICATIONS” An up thrust for knowledge We know that Ww = 2 Ween we TRON tite w= (254) ~j2%% eet NE ein Here eF2% = cos 2n-jsin2n =1-j0= 1 always Hence equation (2.5.4) becomes, 2 Win = ‘This shows that Wi, is periodic with period ‘N’. This property was illustrated earlier in Fig, 24.1. Observe that for N= 8 wpe = Wh ww (255) Proof : We know that ~ (256) Here” cit a cosn~jsin ts ajon tna Hence equation (2.5.6) becomes, ah Wy ? 2 ~ 257) (258) yeh Is fo fuel aa NM UL hea N - 2 um — le ES Bi | 0 for clock gl ca ofa fo be divded Orr ton) — fucerker> MMe. —a_ —* jae vin wl eect > -b > a L= Ntaiel c [sl Toe tk? iM. sy ae ee: > a —-> Petes io zee | een Giep 3 tome Ma Pret i Ge Vid Shp on = BG). sate E Ken. -- so Sod mor CAD) Semyes from JOH), Feed np 5 Tee. dak Ceawtivhan Gip Is Chien cl bee Contaiting je “Cy ter ce) > br ceoe nel nepared by Bh Deeps a ve Saot Hn) = (1, ie. M=2 and x(t) = {1,4,3,0,7,4,-7,-7,-1,3,4,3} Since 5 - point circular convolution is to be used, N = 5. Since N= M+L-1, Now, And x ~ Now let us calculate, Last M1 Le. 1 sample of eich yi(1) should be added to frst M~1 Le, 1 sample of succeeding iC)" 5 =2+L-1 Hn) = {1,2,0,0,0} x(n) = £,4,3,0,0} xo(n) = {7,4,-7,-7,0} xy(n) = £1,3,4,3,0} vie) = 1) He) vale) = x20) 0) ws) = 3@®© +0) n ae vi) ey wQ2)| =}3 4 (| |o 3 n(4) 90 v()| [7 o77 4 y)| [4 7 07 7 nQ)=\7 4 7 0 -7Vol= =L=4 w(3)| |7 7 4 7 wi] loz7s ys(0)] [20 3 4 y(l)| |3 -1 0 3 w(2)|=|4 3 20 w(| [0 3 43 4 so ou ne y(3)| |3 4 3-1 TECHNICAL PUBLICATIONS" An up trust for knowledge a Processing pose ; 2107 Proportss of DFT and its Applications ‘This is shown below ‘Thus the output sequence is, 0 {1,6,11,6,7,18,1,-21,-15,1,10,11,6 } uel (EIIEER) Develop an’ 8-point Drr-FFT algoritn: Draw te’ signal oi graph Debora the DET ofthe Flat Beueee NY x) = {11,1,1,0,0,0,g a using the signal flow graph. Shoo all the intermediate results on the signal flow graph. : I We have already discussed the theory of DIT-FFT algorithm. The signal flow graph for 8-point DIT-FFT Soluti algorithm is given in Fig. 3.2.7. Let us use this signal flow graph to obtain the DFT of given sequence, The samples of x(n) are: x@=1 x(4)=0 x()=1 x @)=0 TECHNICAL PUBLICATIONS”. An up trust for knowedye x(Qal x(=0 x(3)"1 x@)=0 The phase factors are calculated in example 3.2.1 and given in Table 3.2.3. These factors are reproduce convenience. weed Ww} = 0707 ~/ 0707 We = -j We = -0707-j 0707 Computation of 2-point DFTs : Observe the signal flow graph of Fig. 327. The 2-point DFTs are first computed. Table 324 thon computation of these 2-point DFTs. nal flow graphs aa tee computaion f-pont DFTs Cokcisalions of 2 6ehk OEE 4x0) = x(0) + Wy" xc) 4 yee em | = x(0)—Wy'x(4) = 1-10 =a ° = x(2) + Wy x(6) = 1410 4 ie. = x(2)-W,"x6) = x(t) + Wy x5) = 1+ 4x0 = x(1) =W4"x5) pe te Fam=t] = (3) + Wear) t+ so Le. = x(3)—W4"x¢7) 1-10 1 Table 3.24 Computation of 2polnt DFTa Combining 2-point DFTs : ‘The next stage is to combine the above 2-point DFTs to 4 ll Fig. 327 and its calculations are shown in Table 3.2.5 = “Point DFTs, This part of the signal flow 47" inal Processing 3-10 I flow for toes tty Sy on tame} FO) = Vs(O)eWNg V4 210) ret | fer) oe} Fy Vy (neg Vat a! aa ) Wg t F\(2) = Vyy(OWNy Vj 210) ett oe ray F (8) = Vy tN Vat = lepxt et PT F20) = V2y(0}+\%y Vo2(0) F(t) = Vay 0g Vag 1) Combining 4-point DFTe : eat i Gy F2) = Vay(O-We V22(0) sion iy Ea) F (3) = V2y(1}-W¥g Vo9(1) sitet =f Table 3.2.6 Combining 2-point DFTs: The next stage is to combine the two 4-point DFTs to get &point DFT. This part of the signal flow graph of F327 and its calculations are shown below in Table 32.6. Refer Table 32.6 on next page) Tho complete signal flow graph : The complete signal flow graph with all the values calculated is reproduced below in Fig. 32.14. (Refer Fig. 32.14 on page 3-21.) Afar. tyferuite? eer=n-’ A dre te be he chetnet REM Ue fallerang det iret Lewned aoe Mt ef ods gh = Ufa Ha ener Bedesvedt sit the covepsctenss Fyn) pee hea) UNS Lea pecderpaliy By fagirad Be atone t 1 [quenced agen fe paaparee, ONS Sy aun, pe per ~ f aeaferes © Klace rete Ole aneted har eeeee cakenae meade) LP sanocen | a] The GH mn Bate SCI broke aga ot jaune 2 nxt een! ase wie off ts ae Freee, Pret acer Hfe) = Ferns) : Pare Gr fic fo eT] lave wen, TECHNICAL PUBLICATIONS”. An up thrust for knowedg gta! Signal Processing an . For n=3, fy(t) = =B fo de ‘Thus a(n) forns3 ) Step 2 : To perform windowing and obtain h(n), Hore hamming window Is given, y(n) = ost nbof 1) For M= 7, wy(n) = ost a46ens( 3 for n= 04, for M1 = 0,1 pornnenne 6 Following table shows uC toy (2) and hn) sin .75 n=O)" fore 3° iyi fore for es An) = {e006 00:929,017325,075,017325,004925 0006} Stop 3: To obtain frequency response M3 wagon) n=O peo EF 0) 22 SM)20(0-3) = 0.75 + 2[0005 cos 310-001929 cos2ea+ 017325 cos . nad

You might also like