CONVOLUTION AND CORRELATION OF SIGNALS Seer Bet ecobites ght fort fw sp toe ses tirenosal termes oe rere, Seer ipeie octubre qptem tothe output signal, L } Correlation : Correlation Defwecn two signals is to Introduction 5.2 Concept of Convolution in Time Domain and Frequency Domain 5.3 Graphical Representation of Convolution measure the degree to which the two signals are similar ; ee it 15.4 Convolution sum } Gross Correlation : Cross corbelation is a measure of 15.5 Cross Correlation and Auto Be tes occa cra arn eli ar Pie acne oe Sr eign. 1 1 } ute Correlation : Auto correlgtion is w measure of of 5.6 Properties of Correlation Function 5.7 Energy Density Spectrum similarity between a signal and time delayed vers same signal. 1 Power Density Spectrum > Spectral Density : Spectral density defines the distribution Relation between Auto Correlation Function and Energy/Power af energy or power of the signal per whit bandwidth as Spectral Densify Function ——fetion of frequency. \ > Energy Spectral Density + Energy spectral density is Adfined as the distribution of energy of the signal per unit Aandi as function of frequency. \ Relation between Convolution and Correlation Detection of Periodic Signals in the Presence of Noise by Correlation e sors re + The average power (or sim en of averse pe dissipated by a oe +12 Extraction of a Signal from Noise by Filtering Solved Problems of University End Examsfethernatical operation used to combine two signals t0 form y 44. inal and Impulse response o 4. Convolution is an Important mé ¥ ‘al, Convolution gives the relationship between the input sigt © 1 Authear time trvariant (11) system to the output signal. ther important mathematical operation in the analysis of signaiy yyy ised to combine two Signals to top ™ Correlation is an ano’ systems, that 1s similar to convolution. Correlation Is also a few signal, Correlation between two signals 1s to measure the degree to which te (wo sign ; are similar. Correlation is widely used In practice, particularly In Radar, Sonar and digity Correlation 1s of two types, cross correlation and aut and time delayed version of anoth, ‘communications. 10 correlation, Crags correlation is a Measure of similarity between one signal Signal Auto correiaticn is @ measure of similarity between a signal and time delayed version oy same signal, 5.2 CONCEPT OF CONVOLUTION IN TIME DOMAIN AND FREQUENCY DOMAIN, operation which Is used to express the input-output relationship Convolution is @ mathematical an Linear Time Invariant (LT1) system. 5.2.1 Convolution Integral In continuous-time LTI system, the convolution integral, relates input and impulse response of the system to output. ‘Consider an LTI continuous-time system, with x(t) and y(t) as its input and output respectively as shown in Fig. 5.2.19). at, system 3) 0) xt) ORE ise the system is an impulse as shown in Fig, 5.2.1(b) then the output of the system called the impulse response of the system. be represented as,wi qn correlation of Signals (Unit 5] pi 8 78 Som in Fig, 5.2.1 is given by, . y(t) = HL Ex(t)] ¥ _ y(t) = 90 foo t-dee for a linear system, i y(t) = foo Halt - 2) de .. (5.2.1) If the response of the system due to impulse 8(t) is h(t), then the response of the system © gue to delayed impulse is, ie aaepeng ele a) (6.2.2) _ Heibstituting 4.) (5.2.2) In Eq. (5.2.1), we get, yit) = fo he ddr (623) system, the output due to input delayed by + sec is equal to the ‘output For a time invariant s delayed by + sec: That is, h(t, 1) = htt - 1) w» (5.2.4) substituting Eq. (5-2-4) in Eq. (5-2-3), Me get, yt) = fo hit - 2) dr 625) Eq. 5.2.5 indicates the convolution integral, oF simply convolution. The convolution of two signals a and h(t) can be represented 25 yt) = x(t) = 0) he convolution between tw denotes th 0 signals. Convolution Integral Gtion integral are as follows: {The commutative property of convolution integral states that, n(t) * x0) a property of convolution integral states that,0 screeners SUPINE and Correlation of Signals Seer 5.2.3 Time Convolution Theorem Associative Property : The associative property ‘of convolution integral states that, x(t) « [hy (t) # ha(t)] = Be) * HYCO * h(t) Shift Property : The shift property of convolution integral state that If, x(t) * h(t) = y(t) Then, x(t) # h(t = T) = vit = 1) ‘Similarly, x(t - T) * h(t) = y(t - T) And, x(t —T,) «Mt = T.) = vCt = Th - Te) @ Convolution with an Impulse : Convolution of a signal x(t) with a unit impulse is the sony Itself. That is, x(t) # a(t) = x(t) Width Property : Let the duration of x(t) and h(t) be T, and T, respectively. Then the duratan of the signal obtained by convolving x(t) and hit) is T; + Tz. StaTEMENT : The time convolution theorem states that convolution of two continuous-time signa in time domain is equivalent to multiplication of their respective spectra in frequency doman Mathematically, t, x(t) 1+X,(0) And xo(t) $L>x,(0) Then, x,(t) + x2(t) 7X, (0) X,(0) Proof + The Fourier transform of [x,(t)sx,(t)] is defined as,ef Signals (Unit - 5 {6-5} oa j = intl a fete Se See neta then we Wage ap Pe aol deme ¥ vipa sstort a - Foe) * x2(0) = Jul ramos as " _ «fel fate ahem ae ’ ; 5 Ce ——— : ee : x40) ~ Pasco xate) oom a: “fuer abu ae a Xo) = X,(@)X,(0) x(t) * x(t) EE X,(0)X2(0) w= (5.2.6) Eq. (5.2.6) represents the time convolution theorem. 5.2.4 Frequency Convolution Theorem “Statement : The frequency convolution theorem states that the multiplication of continuous time signals in time domain is equivalent to convolution of their respective spectra in frequency domain. Mathematically, y x(t) > x, (0) ind aft) FL x2(e) g(t) x2(0 Em) + X20)]6 $$ Convolution and Correlation of si, a Interchanging the order of Integration, FHethn(Ol= 2 fraiwo «| farern | Using the frequency-shifting property, the bracketed term is X,(@ ~ '). Substituting ty the above equation gives, Flv (x20 = 9 fle) Xyl0— ) Putting m - @’ = 2, in above equation, we get Fix,(t).x2(t)) = Z fue = 1X0) = EM) *X,(0)) : 2 X20 E+E fey(0) *x,00)] (or) 2xx,(t)x(t)F* +X, (0) * X2(0) In terms of frequency, we get, Fhe, (0-,(0) = X09 = (0) Eq. (5.2.7) and Eq. (5.2.8) raitehared the frequency convolution theorem. a) Find the ‘Seibel ‘the following signals: 2) 20 = 6 Vth) = 0 Ot bay = ot oH ato = t+) Given x,(t) = e* u(t), x(t) = € u(t) ls in time-domain is defined as, u(t = 1) dra; wo={ nyse Lo; toveOStar ere and ut-a)= i t a ese Misnl=\y eceusdmees Hence u(s).u(t= 5) = 1 only for 0 < + < t, For all other values of x, u(s).u(t - x) = 0. Therefore, ‘ x(t) * x2(t) = fe oe a t =e% fe Gr = eter]! ~ eB fet - 1) (et - UH) 3 b) Given x,(t) = e* u(t), x,(t) = u(t + 1) Convolution of signals in time-domain is defined as, x(t) * x2(t) = fac x,(t-1) dr = fer u(t) u(t + 1 = ) dr In this case, Bil go ipo ft. deed ult) { treo and uesi-g=f weed Hence u(r).u(t+1-x) = 1 only for 0 < + < t+1. For all other values of rt, u(x).u(t+1-1) = 0. Therefore, ea xy(t) * x(t) = j= d= [-e 3 sretteN 4 e% a1-etten Using time-convolution theorem. Find the convolution of two signals of Fourier transform. (Y= e* u(t) and x,(t) = e u(t) Given, x,(t) = e-? u(t) and x,(t) = e** u(t) We have,58 Conwolution and Correlation of Sigma ty, eg From the definition of time-convolution theorem, xy (t) *x2(t) AE Xj(0)-X2(0) = FLx,(t) * x,(t)] = X,(o) - X(0) = x(t) # x(t) = FX, (0) . X(0)] a1 1 1 4 i ‘GE raee = eae] 5.3 GRAPHICAL REPRESENTATION OF CONVOLUTION The convolution of two continuous-time signals x,(t) and x,(t) using the graphical method invojy trent u(t) the following steps: Step 1 (Change of Time Index) : Replace the independent variable t by a dummy variable + fy the given signals x,(t) and x,(t) and plot the graph for x,(z) and x(x). Step 2 (Folding) : Keep the function x,(x) fixed and fold (or invert) the function x,(x) about the vertical axis (t = 0) to get x,(-1). Step 3 (Shifting) : Shift the function x,(-x) by t units. The shifted x,(-r) now represents x,(t ~) Plot the graph for x,(x) and x,(t - t) on the same axis beginning with time shift t. Step 4 (Multiplication and Integration) : The signals x,(:) and the shifted signal x,(t - +) a multiplied to get a product signal x,(s)x,(t - x). For a particular value of t = k, integration of the product signal represents the area under the product curve (i.e., common area). This common area represents the convolution of x,(t) and x,(t) for a shift of t = k, That is, faitorace -)d = [uxt Repeat this procedure for different values of t by successively progressing the frame by different amounts and find the ‘of the convolution function x(t)«h(t) at those values of t. c shown in Fig. 5.3.1 with itself.jition of convolution, 5 . 4 Yt) = x(t) * nity = fo Wt- 4 Given convolution of a signal with itself, so h(t) = x(t) only. The mathematical expression for x(t) and hit) Is, xm=no={F } O 3) ti CMe5.14 —___ conyatuttan and Carretation of Signals tugy —— The convolved signal y(t) = x(t)+h(t) is as shown in Fig. 5.3.11. vit) iy 5.4 CONVOLUTION SUM In discrete-time systems, the convolution sum, relates input and impulse response of the System ‘to output. It is mathematically expressed as, y(n) = (k) h(n -k 2 : + (5.4.1) +x The relationship expressed in Eq. (5.4.1) is called the convolution sum of input xn) and une impulse response h(n). This operation is represented symbolically as, y(n) = x(n)*h(n). Determine the convolution of the two sequences, xe) =(2)' voy ame ny =(2)" oy Sol: By definition, x(n)* hin) = Yaw hin) = YE . ay unk) ral In this case, a; k20 A; n-k205ksn w= fh | Keo I UN-K) fo p n-k Step 2 (Folding) : Invert hk) about the vertical axis (k = 0) to obtain h(-k). Step 3 (Shifting) : Shift the inverted hk), i.e,, h(-k) by n units to obtain h(n ~ k). Step 4 (Multiplication and Addition) : Finally multiply x(k) and h(n - k) for particular-values of Nand add all the products to obtain x(n)*h(n). The procedure is repeated for each value of n over the range -= to ©, Comment : If x(n) starts at n= 1, and h(u) starts at 1 = ny then choose n = nytiy as starting tine for evaluating the output sequence y(u) = x(n)*h(n). So Find the convolution of two signals x(n) = {1, 1,0, 1, 1} and h(n) = {1, -2,-3, 4} and represent then graphicelly. [Dec. - 2011] Sol: Step 182 (Change of Time Index and Folding) : The graphical plots of x(k) and h(k) shifted @round vertical axis (k = 0) i.e., h(-k) is as shown in Fig. 5.4.1. xk 4 bk Sree 9, 4 an 5 5 the starting time for evaluat reine ote equet ting y(n) for various _ values of n a-Y0 Mak) e& = x(0)M0) =1.1=1 YEA) = x09) ran =e) Po : =1.-2)+11=-1 «yin = Dixon = = WC) 416-2) +0=-8 re “Lrw-o PAGeM-D+Ovr1=2 mEt ho) + x“5.18 — Convolution and Correlation of Signals ‘um jealefpeg bbenaticlemne sae SN. Nig, viel = xt Ht ee 041.4404 1(-2)41.123 veo} = 7 xk) ho) = =0+0+0+(1.(-3)+(1.(-2)=-5 ie} = 7 x4) nk) = =0+0+0+14+1.(-3)=1 yin) = Y xan) a> 7 we get = for n Get Y¥(n) = 0. Therefore the convolution of given sequences x(n) and hin) Is, y(n) = x(n)*h(n) = {1, <1, <5, 2, 3, +5, 1, 4} n=0 The graphical plot of convolution of given two signals is as shown in Fig. 5.4.3. yin} S_ -5 5.5 CROSS CORRELATION AND AUTO CORRELATION FUNCTIONS ion that closely resembles convolution. Correlation is a measure namely cross-carrelation and auto- Correlation is a mathematical operati of similarity between the two signals. Correlation is of two types, correlation. 5.5.1 Cross Correlation of Energy Signals of similarity between one signal and time-delayed version of Cross-correlation is the measure lated to the another signal; it means that the cross-correlation explains how much one signal Is rel time delayed version of another signal. Let x(t) and y(t) be two different complex these two energy signals Is defined a5, valued energy signals, The cross-correlation between‘and Correlation of Signals (yp of meter, searching parameter or scarp, lation between two signals. The subscript, signals being correlated. The second possible cr, the order of subscripts xy, that is R,,(t) iS defined as 4 Rylt) = fro xa(t- 1) dt= fue 1) xt(t) dt ‘valued signals : R,x(*) = fro x(t -1) dt = fe +1) x(t) dt .o(*) will be finite over some range of « if the signals x(t) ay The cross-correlation function , larity. The energy signals x(t) and y(t) are said to be orthogonal i. (0) Is zero. Substituting x = 0 in Eq. (3.5.1) te ignals can be written as, y(t) have some simil 1 (having no similarity) over the entire time interval if R, condition for orthogonality of the two energy si Ryy(0) = fore dt =0 (653) Ry,(0) = froxre dt=0 on (654) or, 5.5.2 Cross Correlation of Power Signals ‘The cross correlation function between two different complex valued power signals x(t) and v() is mathematically defined by, 7 2 Ryglo)= lim + Jxore-nat- jim + Jtrovme 655 a tn If x(t) and y(t) are real, then, 0} bsi6li, , Feeiple sig: tap. so". 27 i sebhecdbais: 1 =s)dt= tim + fur v(t) dt (658) . gy! TR *-e interval if, ver the ¢, cross Correlation of Periodic Signals : An important special case of correlation of power ea _. js the correlation between two periodic signals whose fundamental periods are such that the e product of the two signals is also periodic. This will happen any time the ratio of their fundamental periods is a rational number, For two signals whose product has a period 7, the general form of the correlation function (for real power signals) is given by, T , 2 = J x(t) y(t = 2) ated J x(t +-xy(t) at ws (5.5.7) -in tp 53 Auto Correlation of Energy Signals auto correlation is the measure of similarity between signal and time delayed version of same signal. 1 Ruy) = ifx(t) Is @ complex-valued energy signal then the auto correlation function is defined as, Rag (2) = feo x(t- 2) dt= fue +2) x"(t) dt ~. (5.5.8) Ifx{t) is a real-valued signal then, Rald= front -oat= ft roan ae 554 Auto Correlation of Power Signals (5.5.99 x(t) is a complex-valued power signal, then its auto correlation function is defined as, 2 2 fh * _ . Raat = im 2 fx xt= 1) at = tim & J x(t +9) x") dt on (5.5.10) tn “in It x(t) is a gi valued signal then, om ve ‘b ni, 1 Rig(2) = gm £ Joo He-9 at = tim 2 Jxcro wy at eset) in a * Avto-Correlation of Periodic Signals : When the power signal x(t) Is a periodic signal, the integrand in Eq. (5.5.11) is also periodic and the time average can be taken over one period (T, say). Thus we may express the auto correlation function of a periodic signal X,(t) of period T as, ™ Pe RRB 1h. g Raplt)= = J xs .. (5.5.12) Time Signals ‘the graphical method St0ss correlation of a real valued si the following sequence of op 1 (Shifting) : shift the signal Same axis beginning with the‘Ste 2 (Multiplication and Integration) : This signals x(t) and y(t -'s) are multiplied tp - Product signal x(t) y(t - t). For particular value of x = k, the integration of the produc 2 I fepresents the area under the product curve. This common area represents the €r0ss-cont ‘Of x(t) and y(t) for a shift of « = k. That is, ) Repeat this procedure for different values of t and find the values of cross-correation fing, Ry(*) at those values of x. Convolution and Correlation of Signal, ee Ray ea fo Wet~ lew Find the cross-correlation of the signals x(t) and y(!) are shown in Fig. 5.5.1. x(t) eo) ‘The definition of correlation, Royle fo Vet- Dhow Given correlation of two signals x(t) and y(t), the mathematical expressions of these signals at, 2; O Real-Valued Signals : R,,(m) = jim — nm) y(n) # Ryy(m) = ti x(n) y(n 144 i2 (n=) = lim ) ¥(n) For t'vo signals whose product has 2 period Ny, th (for real power signals) is defined by, 'e general form of the correlation function Raylim) = Sr x0) nm) : re aN 562 Autocorrelation Function of Discrete-time Signals IFx(n) Is an energy signal, then its autocorrelation is, Ral) = Yum x(n) = Y'x(0-4m) x(0) f= to If x(n) is @ power signal, then its autocorrelation is, 4 y x(t) = lim wil” x{o—m) = lim wien) x(n) 5543 Graphical Representation of Correlation of Discrete-time Signals Cross correlation of a discrete-time signals x(n) and y(n) can be computed using graphical method by the following sequence of operations. Step 1 (Shifting) : Shift the sequence y(n) by m units to obtain y(n - m). Step 2 (Multiplication and Adding) : Multiply the shifted sequence y(n - m) with x(n) and add all the values to obtain R,,(m). Repeat this procedure for each value of m over the interval ~~ to =, [Bec. - 2011] in) = {4, 3, 2, 1) mt ;Convolution and Correlation of Signals (yj =—ee—_ Unity By definition, cross-correlation is given by, Renton) = x(n) inn) The graphical plots of x(n) and shifted signal h(n ~ m) for various values of m are shown , in Fig. 5.5.5, x{n] 7 1 Eyeegee Cr 5 4 : : h{n + 3] Ral] = Vx hin-m) a SE =) ot ee eee b[n + 2] [rem= =] rae E elowpi2 > fiat >| nl] 7 h(n—m) =2.1+1.0=2 Doditem les'rige z R,alim) = x(n) n=) =3.142.0+1.0=3 Rlem) = x(n) Kn—m) i e4.14+1195. 3 Realm) = Yu tin-m) 2 31-3 Te | ae 8 ay sae h(n -3] 4 Reals =) x00) Wom) Form=3 |" 2) ey [eas slog 1 2-1%912345 6 oT | Fig. 5.5.5 raphical Plots of Cross Correlation of Given Sequences for Various Shift Values 7 Therefore, from” Fig. 5.5.5 the cross-correlation of given sequences is, Ral) ={1,2,3,5,2,3,4) R : n=0 4 PROPERTIES OF CORRELATION FUNCTION Section, we shall define the properties of auto correlation and cross correlation function, Webi so 2: 5: 1 : erties of Cross Correlation of Energy Signals x valued eneray slanals x(t) and y(t) has the following properties, ‘ pie?—_ 3.28 evolution and Correlation of Si Wai. Proof : We have the cross-correlation between two complex valued signals x(t) and v(t)'Be Ral) = frto yrtt— 2) . ~ G6, Letting, p = t - « in Eq. (5.6.1), we get, ; ey Ragl= fxs +) yd oa Also, we know that, Ry lt) = fro xt 9) at a Substituting p = t in Eq. 5.5.3, we get, Ryald= fy) x 0-9 ap ‘a s . = RC) | froo-» | * fro xp - 1) dp (563 ‘Substituting + = -r in Eq. (5.6.5) we get, —_Raea= fy'@xo+9 op ( (B683: Fourier tra we INsform of cro: SS correls dor lation fu renency comain. ction is the product of two energy signals it a (668) the Ea. (5.6.8) is known as the correlation theorem. Fa Properties of Cross Correlation of Power Signals ai : i. i Cross-correlation function of two power signals x(t) and y(t) exhibits conjugate Ray) = Ry) Properry 2: The complex valued power signals x(t) and y(t) are said to be orthogonal over the entire time interval if R,,(r) is zero. That is, 1 Ry =tim + [x y"=0; 120 ts <1 3 Propeaty 3 : The cross-correlation does not satisfies commutative property, |e Ry C0) # Ral) 5.6.3 Properties of Autocorrelation of Energy Signals i Propenry 1 : The autocorrelation exhibits conjugate symmetry, 1+. Ryx(t) = Rial") = (5.6.9) This means that the real part of R,,(+) 's an even function of ¢ and the Imaginary part of R,,(*) |S ‘an odd function of «. correlation of an eneray signal x(t) is given by, (n) = fan xr(t-)dt= fiero x (5.6.10) ugate, we have, . fro x(t - 9 dt ww (5.6.11)en _— Convolution and Correlation of Signals i, Replacing x with =r, we get, Ri) © fue +a at ‘Comparing Eq. (5.6.10) with Eq. (5.6.12), we get, & Raylt) =Ria(-*) Propeery 2 : The value of autocorrelation of an energy signal at the origin (+ = 0) gives the ‘energy of that signal. i.e., R(0) = fi x()P dt =, Proof : We have, Raxlt) = foowee =a)dt Putting + = 0 gives, R,,(0) = fo x"( dt = fi xP dt =e, Property 3 : Auto correlation is maximum at + = 0, for all values of t. That is, WRyx(*)! < 1R,,(0)] for all + Proof : Consider the integral, ~ Joc ext cop at20 = 7~ rom property 2 defined by Eq, (5.6.14) an 6.14) and bj ty £4. (5.5.8), We get, YY definition of auto correlation function defined PRu(O) # WR 20 > : | a) = Ry(0) 2+ Ry lr) J considering magnitude, we get, 7 : IR, (Ol > IR, Le = proreerr 4 : The auto correlation function and the energy spectral density form @ Fourier transform pair, that is, — : Rls) EL wo) or WO) r 56.4 Properties of Autocorrelation of Power Signals Proreery 1: The autocorrelation function of power signals exhibits conjugate symmetry, that Is, R(t) = Rice) Proof : We have, a7 mm Rees) = im zt Jxore-oet- ims Jxroroa me mR Tr « ™ RE jim = fro x(t = 1) dt tn 2 a Rata fim Sf cates det -Ralo tn ‘ Rex) = Rix) ; Property 2 : The value of the autocorrelation function of @ power signal at the origin is equal to signal, that is, ww (5.6.16)ig a Conve Putting x = 0, we et, te ” “ ia Ral@)= + [x 9°(0 =F firme ar, an aa Propenry 3 : The autocorrelation R,,(+) and PSD S(o) of 2 power signal forms a Fourier transtan alr, 1.e., Reg(t) $2 5(0) or SEF) Propsary 4 : The autocorrelation of a periodic signal with period T is periodic with the same pera: 7 x(t) = x(t #1), Then, Ralt) = R(t #7) Proof : Let x(t) be a periodic signal with period T. Then, x(t) = x(t +1) x(t = 2) = x(t = + 1) and x(t + 1) = x(t ++ 1) We have auto-correlation function of (real valued) periodic signal as, 2 Realt) -t J x(t) x(t = 1) dt an Replacing « with x - T, we get, sal Using Ea.correlation of Signals (Un -3) : aie 5.33 g x with t + T we get, ™! Role T) = 4 if th X(t + 2+ T) x(t) at 2 1 =F [xteoaey, ¥ YMO.c fy Usingieg. ety) = R(t) Rexlt) = Ret T) 47 ENERGY DENSITY SPECTRUM Spectral density plays an important role in determining the power and energy of a signal. This parameter is a function of frequency and is directly related to the amplitude spectrum of the concerned signal. In simple words, spectral density defines the distribution of energy or power of the signal per unit bandwidth.as function of frequency. : In general we encounter two types of signals based on energy, some signals are finite energy signals (for example, pulse signals) and others are infinite energy (for example, periodic signals). Signals having finite energy are called energy signals while the signals having infinite energy are called power signals. When you deal with energy signals it is customary to use energy spectral density as the parameter for computation of the signal. Likewise when dealing with power signals we use power spectral density of the signal as characterizing the parameter. | * Normalized Energy : The normalized energy (or simply energy) of a signal x(t) is defined as ‘energy dissipated by a voltage signal applied across a 1 ohm resistor (or alternatively by a signal flowing through a1 ohm resistor). The signal x(t) may be complex or real valued. lized energy of the signal Is defined as,ution and Correlation of Signals [up ee ea eee 57 1 Parseval’s Theorem for Energy Signals ¢ Parsevals theorem foF energy signals states that eneray of @ signal is determined the area ving curve [X(o)|2 or [X(A)I2 as a function of frequency, That is, L a 2 ex fovataes fant at Proof : Consider a function xit) such that x(t)! +X(o). Let x*(t) be is conjugate, then, x*(t) SL x"(-0) ‘The eneray of a signal x(t) is given by, - = ee E= [jae at = [xctyxe(t) dt = J xr(t) at p (3.72) replacing x(t) in Eq, (5.7.2) using the definition of inverse Fourier transform, we have, E- f vl fame “| at Interchanging the order of integration, eek sf free ol to io) fateh at = x*(o)= frooerrate frroerear * % (874) signals, aso called Rayleigh" signal is determined by te #and correlation of Signals [Unit - 5) ‘6.35 fe ergy Spectral Density (ESD) 4 gy spectral density is defined as the distribution of energy of the signal per unit bandwidth sanction of fFeaUENCY. ESD Is denoted by y(t, wih'= Shey oe Unit bandwidth ~ aF as seen from the Rayleighs theorem |x(F)|? inthe frequency domain. Thus, gives the distribution of energy of the signal, x(t) WA) = 1X1? = 1X(o) |? Let us consider @ signal x(t) applied at the input of an ideal bandpass filter whose transfer function is shown in Fig. 5.7.1(a). This filter Suppresses all frequencies except a narrow band af(sf 4.0) centered at frequency fy. If Y(f) is the Fourier transform of the response y(t) of the filter, then, Y(f) = X(F) HA) ~ (6.7.5) the energy E, of the output y(t) is given by, = fi Y(f) I? of : (5.7.6) Using Eq. (5.7.5) the energy of the bandpass filter output can be written as, E, = fi Xx(F) H(F) I? df ~ (5.6.7) Since |H(f)| = © everywhere except over a narrow band Af where it is unity, we have, E, = 21X(f,)I? af on (5.7.8) 1H Jerr Sites 1 =fo 0 fo 4 ° fo t (2) Transter Function of Bandpass (b) Spectral Distribution of the Input Signal Showing the Contribution to the Output Energy (Shaded Portion) Signals ond Systems:$8 rir rrr The energy of the output signal is thus 2|X(fp)I74f. This factor 2|X(fo)I7AF represents i, contribution to the energy of x(t) by the frequency components of x(t) lying within a narrow bad ‘Af Centered at f,. Hence, 21X(f,)I? ts the energy per positive frequency components contributes ong, half of the eneray’2IX(f,)I? while the remaining half is contributed by the negative frequency components centered at f,. Therefore, factor 21X(fo)I? can be interpreted as the energy per uni bandwidth of the input signal which Is contributed by the frequency components (positive o negative) around the frequency f. That is, si 2 € pin = E=v 679) ‘Therefore the total energy E of a signal can be written in terms of energy density spectrum as, Nerd ror a Ju oo (5.7.10) Since |X(NI? = IX(-f)I2 , v,(f) is an even function of f and we may write Eq. (5.7.10) as, = 1 ? = = 1 fe Joc df 2 front df Jue df = 6 3. Energy Spectral Densities of Input and Output If x(t) and y(t) are the input and response respectively of a linear system (shown in Fig. 9.72) then, : Yen) = xf A) (6.7.11) Where, HH(f) + The transfer function of the system X(f), Y(f) » The Fourier transforms of the input and the response respectively. “The energy spectral density of the input (excitation) is [X(F)|? and that of the ‘output (response) is [¥(f)|2. Therefore, w= XO? Af) = BHCOIE = THK XC = THEO? COE (5.7.2) of the output (response) of a linear syste" input and square of the magnitude of thena correlation of Slonals [Unit 5) a a properties of Energy Spectral Density 11: The total area und ler the energy spectral density function is equal to the total energY pat signal He e c= fune 2 If x(t) Is input to a linear time invariant (LT1) system with transfer function H(o), then spit and output enersy spectral density functions are related as, yo) = IH(o)? yo) | were, : yola) > Output energy spectral density function vfe) > Input energy spectral density function |H(o)|?_ > Energy gain at frequency o. ‘hora 3 + The auto correlation function Ry(s) and energy spectral densty funtion vy,(@) form __sfourier transform pair i.e. ! Ryx(t) AL v,(0) Sn) Determine the energy of the sinc pulse defined by, xt) = A sine(2W1) The energy of the given signal can be written as, ee fi x(t)? dt = A? Jrnccewn at mie The above integral which Is difficult to eval igh energy theorem. rt with the following Fourler transform pairs, can be calculated indirectly by using purpose we star i “ane 5) fh at sine (2) ‘equation becomes, i539 ——___ Convolution and Correlation of Signals tuay, ae Applying the duality property, we get, ansine() «PL, 2eh Rect (3) = 2A Rect (3) ‘Applying the time-scaling property (choose time-scaling factor equal to 2nW), we get, ~ dAsinet2Wt) EE» ra Ret (35%) (* ot) 9h (2 3) Using linear property, j.e., 2 x(t)+—*+a X(a), we get, ° Asinc(2wt) 2+ — - Rect (- fm P do The frequency spectrum of given function x(t) is shown in Fig. 5.6.3. (0) A2W -2aW 230 ‘0 MEDIREEEED Freauency Spectrum of x(t) = A sinc(2we) ae 1) ee miaewt ee +200 ~ sagt * ur 5.8 POWER DENSITY SPECTRUM Samat wae ‘energy (for example periodic signals) are called power signals. The meaningfil eae ere The average power (or simply pore? sipated by a voltage x(t) applied across a1 ole 11 oltm resistor). ‘Signals and System™Gee eeestetetynh oy ee average power P, is given by, 2 Pe = tim = | ic? at an tF ve (5.BAY ' Power P, defined by Eq. (5.8.1) Is actually the mean square value or the time average Of yred signal. Thus we may write, 2 32] = tim 4 [rer at on (58.2) is Theorem for Power Signals power theorem defines the power of a signal in terms of its Fourier series coefficients, terms of the harmonic components present in the signal. Mathematically, it is given by, = (5.8.3) sider a function x(t). The average power of x(t) for one cycle is, 1! 1a pat J Ix) P dt = 4 J x(t) x7(t) dt 68.) ain “in But, we have the exponential Fourier series representation of x(t) as, x)=) Cet - (5.8.5) iS camel dt 7 1 ale 8 — a= foe “(ty edie af "(tera at c Fo J ms “2 "in Siew P : (5.8.6) I's power theorem and it defines that the power of a signal ~ Eq, (5.8.6) is called Parseva F tscess P oS to the sum of square of the magnitudes Of Various harmonics present in the discrete of power signal x(t):5.40 Convolution and Correlation of Signals (Yay Squyelitleyrand Gontaballante! Sanals tial 5.8.2) Power Spectral Density (PSD) Power spectral density is defined as distribution of power of the signal per unit bandwidth as fy In Of frequency, It is also called power density spectrum. It is denoted by S(f). 7 Consider an arbitrary power signal x(t) as shown in Fig. 5.8.1(a). x(t) x(t), @) (b) (EERE (0) Av Arbitrary Power Signat (6) Its Truncated Version To determine the PSD, assume the power signal as a limiting case of energy signal. For this purpas, ‘truncate the power signal x(t) shown in Fig. 5.8.1(a), so that it is Fourier transformable. The truncated function X,(t) is shown in Fig. 5.8.1(b). Mathematically truncated function is defined as, x(t) 5 Itl= 0 ; — Elsewhere xq(t)= | 5) Let, Xq(t)} $24 X4(F) oF X(0) The signal x,(t) is of finite duration T and hence it is an energy signal with energy E given bj, E+ ji xP dee ficcor do As x(t) over the interval, (~T/2, T/2) is same as x,(t) over the interval (-«, #), we have, : ” Jisor t= freer ke ain ™ ‘ Jor ated fisior do (588) mR ~ ‘ If T +, the left-hand side of Eq. (5.8.8) represents the average power P of the function x(t) That is, a Pouing | eat Patimste fixer de sm a z RE iranond Correlation of Signals (Unit - 5] 790, 1X(@)17/T approaches y 'S a finite value. Let this finite value is denoted S(o). That is, S(o) = Jim Xx(o) Fen tT . (5.8.10) the average power P of the function x(t) is given by, Ja rs P= x(t) = fan af Ushi Where x2(p) Is the mean square value of x(t). The average power is therefore given by, ! foo do = 2fane (5.8.12) 3 a ‘The PSD of a periodic function is given by, = ny ICn [2 8( ~ 1%) .. (5.8.13) of, (5.8.14) 5.8.3 Properties of PSD ‘The following are the properties of PSD, \ Properry 1: The area under the PSD function IS equal to the average power of that signal, 1.€. am fa - w» (5.8.15) Property 2 : The input and output PSPS of an LTT system are related a5, Sf) = HCI? S00) eat Where, sf) + Power spectral density of output y(t) sf) + Power spectral density of input x(t) WEG >. Transfer function of the system: Proptary 3 ; The autocorrelation function R(t) and PSD S(f) form a Fourier transform pair, | Riyals) PE Si()5.42 5.8.4 Comparison of ESD and PSD Comparison of ESD and PSD of a function x(t) are depicted in Table 5.8.1. FECES Comparison of ESD and PSD Convolution and Correlation of Signals (yy, ee i It defines the distribution of energy of | It defines the distribution of power of a sign) signal in frequency domain. in frequency domain, Itis given by, Itis given by, wef) = XP S(t) = tim LOE +757 The total energy is given by, The total power is given by, & frnaet [vem be feinged [am ao “ig ! 4 a ‘The autocorrelation for an energy signal | The autocorrelation fora power signal and its and its ESD forma Fourier transform pair, | PSD form a Fourier transform pair. Ryle y(") Rylt)L2 si Or Or Ry (1) 2+ Wo) Ryd) —L+ Ko), 5.9 RELATION BETWEEN AUTO CORRELATION FUNCTION OF ENERGY/ POWER SPECTRAL DENSITY FUNCTION 5.9.1 Relation between ESD and Autocorrelation Function The autocomelation function R,.,(x) for an energy signal and its energy spectral density function vi) form @ Fourier transform pair, |.¢., Ry) AL» y(o) Proof: ‘The autocorrelation of a function x(t) Is given as, Ral) fuo x(t at Replacing x*(t ~ x) by its cies Fourier transform, we have, x Reltde fo from 4 «| te afl from ” | at ee 2 . Borah ner | Toanging the order of integration, we have, 1] y ' 1 i * “hh h X°(@) X(a)e do = fi X(o) Pe do r +: (neeeane iain =F Afy(o)] ; 2 Yo) = FIR Q(1)] 1 ‘is proves that R(:) and y(o) form a Fourier transform pair, ; Ry lt) =D ,y(a) 492 Relation between PSD and Autocorrelation Function f ‘he autocorrelation function R,,(x) for a power signal and its power spectral density (PSD) function Se) of a power signal form a Fourier transform pait, ie, 1" Ryx(t) +450) 7 Proof: The autocorrelation function of a power (periodic) signal x(t) in terms of Fourier series tteffcients is given as, | Ral)= ) eer Where c, and C., are the exponential Fourier series coefficients, Ral= Yie Pemeo Fooling Fourier transform on both sides of above equation we get, FR (0) = ip IcaP oem MsHs Wit) Interchanging the order of integration and summation, we get, FAR = D7 [Cy fome-mer or But, tel, 280) and 1,4! «$5 2n6(u) wp). Thus, above ‘equation becomes, D = 2x y 1Gp I? 8( ~ 1199) = Dy IC, 8(f = nfo) ‘The RHS is the PSD S(w) or S(f) of the periodic function x(t), FAR, (1)) = S(0) for $(1) he, Rat) EL, S(a) for S(f)) 5.10°\ RELATION BETWEEN CONVOLUTION AND CORRELATION Both the convolution and correlation mathematical tools have a striking similarity, Ofcourse the two Integrals are closely related. To obtain the cross correlation of x(t) and y(t), according to the equation Ryn) = fun y(t = ) dt, we multiply x(t) with function y(t) displaced by + sec, The area under the product Curve is the cross correlation between x(t) and y(t) at t = 1, On the other hand, the convolution of x(t) and y(t) at t = + is obtained by folding y(t) backward about the vertical axis at the origin and taking the area under the product curve of x(t) and the folded function y(-t) displaced by +, It, therefore, follows that the cross correlation of x(t) and y(t) Is the same as the convolution of x(t) and y(~t). The same conclusion can be arrived at analytically as follows: The convolution of x(t) and y(-t) Is given by, wie yl 9 x00) lt x(t) » y(ct) = fo rt) de Replacing the dummy variable + in the above Integral by another variable k, we have, a0) 0) = fat yk —1) ak ‘Changing the variable from t to +, we get, 10) #164) fr) 0-1) dk = RyRyylt) = x(t) * y(t), Nea and Rl) = YOO * (4), be directly applied techni iof the techniques used to evaluate the convolution of two functions can red for convolution eter to find the correlation of two functions, si go PY to correlation. Similarly, all of the results deriv sf one of the function is an even function of t, let us say y{t) Is an even function of t, he, WO 2 y(t): Then the cross correlation and convolution are equivalent sl DETECTION OF PERIODIC SIGNALS IN THE PRESENCE OF NOISE BY CORRELATION tn detection theory, detecting the periodic signals in presence of random noise/AWG importance. Tt finds various applications in RADAR, SONAR detection, detection of periodic components in brain waves, the detection of cyclic components in ocean Wave analysis, in meteorology, etc. The correlation techniques discussed earlier finds to be @ powerful too! in the ation of the above problems. N is of utmost ‘the noise signal encountered in practice is a signal with random amplitude variations, Such ‘signal is uncorrelated with any periodic signal Let x(t) represent the periodic signal, n(t) represent the noise, » Then, 1 ma 1 Ren( = Nim Jo ft ~1) dt = 0 sin ‘ers SII Detection by Autocorrelation Let x(t) be the periodic signal mixed with noise n(t) so the received signal wil be, ada of pride signal and nase nt), eu yit) = x(t) + nit) Where the received signal y(t) is also periodic, ‘Autocorrelation function of periodic signal y(t) is siven by, 1 ® = im + foe syat = tim 1 ffx 4 n(ty) xt = 1) + nt =) at é 1 fio dt at x(t) rt x) at ' ' fo x(t apd foo ten j 1 ={muT eo (S.AL1)Since s(t) and n(t) are uncorrelated, we have, Rynlt) = Rae) = 0 RAO RG + RS We know that the autocorrelation function of a periodic signal is periodic and that of on-pernge function tends to zero for large values of +, AS x(t) Is periodic and n(t) is non-periodic, (=) is also periodic and RG) is arbitrary smay for large values of x. Therefore, for sufficiently large values of r, the autocorrelation of a given signal can be calculate by the numerical techniques used for convolution, even on digital computers. 5.11.2 Detection by Cross-correlation The presence of noise in a periodic signal can also be carried out by cross correlating the recehed signal with another periodic signal of the same frequency. Detection by cross-correlation is much more effective than by the autocorrelation. However The drawback is that it Is necessary to know the frequency of the signal to be detected, earlier. In many cases such as RADAR, etc., the frequeng is known earlier. Let the received signal by, yet) = x(t) + n(t) Let the locally generated signal of same frequency as that of s(t) be o(t). Baja (4) ak 4 Ryle) = slim + foro +ngt-dr’ ft i faced tet ie, Ryg(*) = Rag(t) + Ring(t) If g(t) and nit) are uncorrelated, Ry(z) = 0 Ryolt) = Raglt) It Is obvious that if the cross-correlation of the corrupted signal y(t) with g(t) gives a periodt signal, y(t) must also contain a periodic component of the same frequency as that of a(t)- Ry) = RG) + Ra) vabe Hence, we can conclude that whether a periodic component is present or not, for any of s, we have, 5 R= Real Saints [o1ci3073 0.0 rr edena correlation of Signats (Uni 5) — Ere que is nothing by ing techn 19 but extract) Be rected ic the tine ee fF @ Given signal in the frequency domain. If the given n It Is known as co al relation technique. selationshiP between Correlation and Filtering t) and f(t) ar consiver (0) (t) are the signals, whose cross correlation function is R,,(t) y(t)< 1 (0) t f(t) =F, (0) ‘hen, fyz(t) $1 F (0) tiplication of the pation of te eco FJ) ad F(t gure domain eqn to cos ‘correlation of f,(t) and f,(t) in the time domain. The time domain relation of cross correlation function js obtained by evaluating the integral by a cross correlator. Cross correlation in time and frequency gomain is shown in following figures. hi tt) Rie) (a) ®) TEITIEEEREN css correiation in Time and Frequency Domain In the given figures, f) > _ Input signal Fm) > Transfer function Rial) > Output that when the signal f,(t) is applied to F,(-e) the output F,(-0), the cross correlation between signals f,(t) ly represents filtering. The impulse response of a From the Fig. 5.12.1(b), it is noted will be R,,(z). When the signal f,(«) is applied to and f,(t) may be effected. This operation basic system is given by, in(t) = FFAG 1 ( peeee | nab [Ane Reo), ts cross correlation function is the response of a system dt rte re the signals an f(t) and f,(t) al ne driving function is f,(t). With the impulse response f,(-t) when i Ram —stt~=t*S Signals andee Convolution and Correlation of Signals [Uni 4 noise componey If x(t) is the desired periodic signal component and n(t) is the random PONeNE they the received signal is given by, y(t) = x(t) + n(t) ‘The cross correlating signal f(t) with another periodic signal c(t) of pe a, aid aS that of x(t), detects the desired periodic signal components x(t) which is present in y(t). The above saig cross correlation function is performed by a system which has a unit impulse response c(-t) or . transfer function C(-a) le., C(t) 4 Clo) CL-t) Clo) Where, C(t) + Periodic signal with period Ty. Clo) Fourier transform of periodic signal c(t) Therefore, C(a) consist of impulses located at w = 0, + 20g, £ 30, «.. Nog. Similarly C(-a) also consist of impulses at these frequencies, The magnitudes of impulses located different frequencies is equal to 2 times the corresponding coefficients of the exponential Fourier series for C(-t). Therefore, the Fourier series representation of c(t) is given by, at)= y cent Wh oy #28 ere, oT FITC] = Clo) = ay Cio Step) And, Cla) = 28) C,*8o-no) fs C-0) = C*(o)] 5.13 SOLVED PROBLEMS OF UNIVERSITY END EXAMS XS Applying the convolution theorem, find Fourier Transforms of [Ac-Iit sine2W1]. [Nov. -08, 07] Sol. : The given function is, Ae- incawe, Let, x,(t) = sinc(2Wt) and x,(t) = Aerett The convolution theorem in frequency domain is given as, FTO) 00) = 2 D4 (0) * Xa(0)] erata Ftfe-bt] = 22g for 3? FIT{sincWE Ae-Hty . [rm sad Fro WwW) Peet enc « ea) ea (alent l(a] From the definition of convolution, we have Aals)*hal0) = Prseacte-a o Pe:al ise “We know that, t ress ® real! 1; ists? orec(A\a1 ; 2wsns2W (*) Fetst 2 nea( Zh) =0 elsewhere =O; elewhere raved [aca | Ee {tem Ce 2) ( feta) «(fn H5)-o( ] +0 Alen Cl ee) =o sige | ree “ae = = as =P on3( Ape acters) ai (acta) ‘(toa pt ginc2WE = a aro! - aw “ws | RATAeee Convolution a $80) Find and sketch the convolution of two signals ind Correlation of Signals (yn, ¥ 4 x)= anf) and h(t) nf) [May/Jun Sol. : By definition, yt) = x(t) * Nt) = fro ht -1) de (t=2)], the mathematical expressions of these functions Given, x(t) = an{ t=] and nt) =n] are defined by, 7 O 5.51 +g (Shifting) + Shift the signal ny. me graphical pots of x(x) and ht gunn nit 1) Is defined as, - nt-n= [ti O<(t-ye4 (0; Otherwise Sotet-rtca-tot-deret x) hit-9 2 1 Mei }2 93 4 $) 6 7 iF _ PUREE Groptical Plots of x(x) and h(t - ©) .n and integration to find x(t) ft are shown in Fig. 5.13-4- +) -y a Units. The shied h(-r) now represents M(t = own In Fig, 5.13.3. The mathematical expressions sre 4, 5 (Multiplication and Integration): Perform multplcatio ‘ h(t) for all values of t. The graphical plots for various intervals of (t-1) in the interval = < + Fig, 5:13.4(a) shows the plots of x(s) and I jap, hence the product of (a), we can see that the signals do not over! ost Vin salad Bas a mR : ™ e tim 2: { [cos100" - cos(200t = 200%)} at Bis “raat J, : 1 "A im 2--¢08100+ [t) | im ater (any Regs) = £ 605200" Ra) = Rysls) + Baal) * Ral) = Ar eos200 08300 + §Now, power, P= R(0) -* ale 2 Therefore, P= A’ Now, RMS value = WP = /4 PSD, S(o) = FIR(*)] 2 ‘= | €0s200r+ 3 cos300r + £ coi00 S(o) = © sie + 200) + &(w - 200)] + give + 300) + 8(o = 300)] + qi + 100) + 8 - 10 oan sie) mp =300 200-100 «= 0) 100 200-300 EFI PSD of Given Example Problem Fig. 5. ES Verify Parseval’s theorem for the energy signal x(!) =e“ u(t),a > 0. (Dec. - 2010] Sol.: The energy of a signal x(t) is given by, E- fxor dt = fiew u(t) |? dt fer a(S F = few rae @ 1 Therefore, E= 55 Now, according to Parseval’s theorem, we have, e= 2 fixer do Signals and Systems Samond nhXo) = Fodor = Pratye- 5 Juve ot ot a j aida ; = fet eran feormna ; } [ew* px], = i ario as ar) 5 amleaa = eat Steet 1X0) = Weare (GABA) of “zali-(4)-2 -o from Eq. (5.13.1) and Ea. (5.13.2), we see that the energy is same in both cases. Parseval’s theorem is verified. ) = 1/(1 + fo). Find the energy spectral se ts ere an input x(t) = €" u(t) and transfer function, He x(t) = e* ult) en transfer function, 1 Ho) = 75 jo 2 1 Ho) ? = 155 Fol 1 JH)? = 70? Tro dt) EH)een ee eet ‘Then, Nr 3+ fo laze Input ESD, velo) = 1X(0) Now, output ESD, wylo) = 1 Ho) P-vx(o) +0) Wyle) = (Ale Signals and Sy" .a + Sempling is the process of converting a “eentimuous-time signal into a discrete-time signal such that | the amplitudes of the continuousttime signals are defined eal ice instants of tinal I score aren nse a ‘any band-limited lowpass signal, x(t) bandlimited to W Hz ‘ie, X(o) = 0 for || > W. Then i is possible to reconstruct ‘back original message signal x(t) completely, without any | distortion from its samples, if tha sampling interval, T., is Introduction to Sampling, Sampling Theorem for Band Limited (Low Pass) Signals Nyquist Rate of Sampling Effect of Under Sam, Aliasing ek eat 7, lelsw ‘Sampling of the given input signal x (t) at the sampli ; xt by the impuse train 5,9. The impulse tra rie rate f,(Hz) can be obtained by multiplying : «{8 consists of unit impulse repeating periodically at the rate of 'T., T, == which is shown in Fig. 6.2.1(c) Multiplying the impulse train with the input si ignal x(t) using iItiplie ‘i i in the sampled signal x,(t) as shown in Fig. 6.2.1(d) Te x) Multiplier x) x0) 0 t Th al mee | 50 Cy @ CO} 350) t ST, 47, 37, 27, -T 0 OT TT TT © [EEEEERE Sempiing Operation X,(t) consists of a sequence of impulses located at regular intervals of T, sec and having amplitude equal to the amplitude of x(t) at the corresponding instant ([.e,t = nTy, Where M = ....2, 71,0, 1, .). Mathematically x,(t) is represented as, x(t) = x(t)-5y,(t) Where impulse train is mathematically defined as, 5,(t) = ») a(t - nT) : a 2 (=x) at-oT,) vv (6.211) Fe og cnaaenieeSe ed Applying Fourier transform on both sides of Eq, (6.2.1), we get, Xt) = an) ® - y {t si 3) (Using (9908) Ey 409 + ayy ~- (622) ‘Where X,(0) Is the spectrum of the sampled signal x,(t). From Eq (6.2.2) it can be concludes ‘that the spectrum of sampled signal, X,(F) (or X,(o)) Is an infinite sum of shifted replica of origina signal spectrum X(f) at regular intervals of f, = 1/T, (or o, 2n/T,). Fig. 6.2.2 shows the plots of message spectrum X(F) and sampled signal X,(F) under three cases: ‘Cast 1 (f, > 2W) : When f, > 2W, the spectral replicates have a larger separation between them, known as guard band (see Fig. 6.2.2(b)) which makes the process of filtering much easier and effective. Even @ non-ideal filter which does not have a sharp cutoff can also be used. ‘Case 2 (1, = 2W) + When f, > 2W, there is no separation between the replicates, so no guard band (see Fig. 6.2.2(c)) exists, and X(f) can be obtained from X,(f) by using only an ideal low pass fit (LPF) with sharp cutoff. Case 3 (1, < 2) : When f, < 2W, the low frequency components in XCF = f,) overlaps with igh frequency components in X,(f) (see Fig. 6.2.2(d)). Distortion in original signal x(t) occur on recovered from X,(f) by using any filter. This type of distortion Is called aliasing. Aliasing can be avoided whe? f, 2 2W or T, < 1/2W. xn[Unit - 6) =, W-few RWW () EEEEIIEET Frequency Spectrum of Sampling Operation From above discussion, we can say that for signal recovery without distortion, sampling rate must be equal to or greater than twice the highest frequency component in message signal x(t), that Is, - (6.2.3) (or) Tsou sos (6.2.4) Thus sampling theorem is proved. 63 NYQUIST RATE OF SAMPLING ‘Nyquist rate of sampling is the theoretical minimum sampling rate at which @ signal can be sampled 2nd still be reconstructed from its samples without any distortion, Nyquist rate is always equal to twice the highest frequency component present in the signal, That is, Nyquist rate = 2(Highest frequency component) = 2f,, os (3a) Where f,, is highest frequency component present in the signal,Sa Sol: fa signal sampled sampling rate at less than its Nyquist rate is said to be under sampleq_ any two adjacent samples when a Semeliog (Un. A signal sampled at sampling rate greater than Nyquist rate is said to be over sampieg The reciprocal of Nyquist rate is Nyquist interval and is defined as the time interval pling rate iS Nyquist rate. Nyquist interval = Nyquist rate Determine the Nyquist rate and Nyquist interval for 1) x(t) = 1 + cos 2000xt + sin 4000x1 xq = sine(t00nt iit) x(t) = sin?e(100xt) . Wy) x(t) = cos(200xt) cos(100nt) ¥) x) = Reat(500n), ) Given, x(t) = 1 + cos(2000xt) + sin(400nt) Let, x(t) = 1, x,(t) = cos(2000nt) and x,(t) = sin(4000xt) The highest frequency component in 1 is %m, =0 The highest frequency component in 1 is cos(2000kt) = cos| t) Is om, = 2000 The highest frequency component in sin(4000xt) = SiN (i,t) is @m, = 4000x, So the maximum frequency component in x(t) iS Max [my.®qge ®ay = Max{O, 20008 4000n) = 4000r. Therefore, @q, = 4000" mp _ 4000n dn On & Nyquist rate = 2f,, = 4000 Hz = 2000 Hz 1 5 Sin(2O0nt) | sin(ogt) Given, x(t) = sinc (100nt) « “TO ~~ ‘And Nyquist interval = aha «0.25 ms 4000 100x rad/secNyquist rate = 2f, = 2 x 50 = 100 Hz ‘And Nyquist interval =! 1 ig ms 2,” 100 fi) Given, x(t) = sin c2(100xt) = x(t) = $9.(100xt)_ sin(toost) 100s "~ 100% 1 ~ Toons lsin?(100%)) ‘a 1 _[1-cos(200xty ) © 10% Here, @, = 200%, So fn * Nyquist rate = 2f,, = 2 x 100 = 200 Hz And Nyquist interval = 31 1 = 399755 WM) Given, x(t) = cos(200zt) cos(100xt) ss 5e o5(200nt) cos(100xt)] = § [cos(300xt) + cos (100nt)} (2 C08 cos® « cos(A + 8) + cos(a ~ 8) €05(300 nt) + 4c05(100 at)» Feoslog,t) + $08(09,0) 300r and wm, = 100x, Maximum frequency component in x(t) Is, © = MAX[0p,, Om, ] = 300 x, * 2. Nyquist rate = 2f, +2(3a) 1 . And Nyquist interval = me" 305 67ee OR yy 7 ¥) Given, x(t) = Rect(Soot) Sampling [Uni 5 t Comparing with standard rectangular pulse arect(*), We have, A= 1 and r= r ae We know that, Fouriet transform of rectangular pulse is a sinc function that is, 7 fv -ome (o ) (2sf 14 i. xf @)=sine(*) (5 ere ae oar at Meo)== sine) = (509) "(3 "s00) ~ 500 "| 595 Since a sinc function stays up to infinity, hence the nyquist rate is also infinite ang the nyquist interval is zero. _ 64 EFFECT OF UNDER SAMPLING : ALIASING A sianal sampled at less than jts Nyquist rate is sald to be under sampling (i.e, f, < 2¥, where wis highest frequency component in signal). Consider a signal x(t) whose spectrum Is bandlimited fo W Hz: Letithis-signalibe-sampled at a sampling rate less than twice the highest frequency Component present In x(t), that Is f, < 2W. The sampled signal obtained by sampling x(t) with a a Sampling frequency f, < 2W is as shown in Fig, 6.4.1 This is high frequency 'w’ of spectrun® of X(F+f,). But it overlaps cn X(F) and appears 2 low frequency of f-W ‘Overtapping Xi) ‘Spectrums These Frequencies are Aliased High Frequency Appearing as Low Frequency is Called Aliasing : WEEE effects cf incersamping or aaeng _ “AS seen in’Fig. 6.4.1, the spectrums located at X(f), X(f'- f,), etc., overlap on each other, Ths is called aliasing effect. ‘Aliasing effect is defined as the phenomenon in which the high frequency "+ components of the signal x(t) tal[unit 6) - esto “os Beiiasng can occur if either of the following condition exists: 1) The sampling rate is less than twice the highest frequency present in the signal 2) The signal is not bandlimited to a finite range. the effects of aliasing are: 1) Distortion in signal recovery is generated since the yey and low frequencies interfere with each other. 2) The data is lost and it cannot be recovered. Different methods are available to avoid aliasing: 1) To increase the sampling rate f, such that f, > 2W. 2) To put anti-aliasing filter before the signal x(t) is sampled. 64.1 Sampling Rate f, Higher than 2W f, > 2W, the frequency spectrum-of the sampled continuous signal is as shown in Fig. 6.4.2 and there is no overlapping between the samples and the original signal can be reconstructed without : | | _ |. If the sampling rate is higher than 2W (where W is highest frequency component in signal) i.e., | __ aliasing. $4.2 Anti-aliasing Filter ‘The low-pass filter used for band-limiting a signal before sampling Is generally referred to as anti- aliasing filter. The anti-aliasing filter H,,() put before the sampler is shown in Fig. 6.4.3. x(t) is the continuous signal and is passed through the anti-aliasing filter H,,(@) which gives the output X(t). The signal H(t) is passed through the sampler and recovered as x,(t). The continuous signal x(t) is passed through an anti-aliasing filter whose cut off frequency Is f,/2. All the frequency components of x(t) beyond f,/2 are eliminated before sampling of x(t) Is started. The anti-aliasing filter essentially band limits the signal to f,/2. By this, the components of x(t) beyond f,/2 are lost. However, these Suppressed components cannot corrupt the components of x(t) whose frequency is less than f,/ 2, Thus the spectrum below f,/2 remains Intact and completely recovered. The noise produced by the aliasing is very much reduced when anti-aliasing filter is used,6.10 Sampling [Uni Tt also suppresses the entire noise spectrum beyond the frequency f,/2. 5,0 65 SAMPLING TECHNIQUES Sampling process can be obtained in practise by the following three techniques 1) eal-or impulse sampling. 2) Natural sampling. 3) Flat-top sampling. 6.5.1 Ideal or Impulse Sampling ‘Ideal or instantaneous sampling uses the principle of multiplication. Fig. 6.5.1 shows the circuit to produce instantaneous sampled signal. This circuit is also known as switching. sampler. Let mit) be the input signal to be sampled as shown in Fig. 6.5.2(a). bot, mit) OFF: m(t) WEEN Forctionar Diagram ofa switching Sampler ‘The working operation of this circuit is very simple, When switch will be in ON position then ‘we will get input reproduced at output, For that ON time the amplitude value of m(t) will be taken. When switch is in OFF position, the output will be zero, Fig. 6.5.2 shows the working principle of ‘switching sampler using graphical representation of input and output signals. mitt fa Piao SA, mmwchan 729984 (a) Baseband Signal[unit - 6] 611 \ 8,00) ‘ T3TyaaTy Te OTe ate 3Tp 4s OTs (b) Impulse Train Pulse Width Heres 0 (c) Sampled Signal Graphical Representation of Input and Output Signals of Switching Sampler ¢ Mathematical Analysis : The train of impulses is defined as, 8,0 = > at - ATs) (65.1) o This is known as sampling function and its waveform is shown in Fig. 6.5.2(b). The sampled signal m,(t) expressed as the multiplication of m(t) and 5,,(t) is given by, m,(t) = m(t) - 5y,(t) . (6.5.2) = m(t)- wt - nT.) (6.5.3) = mt) =F mor.) ht -nTs) wn (6.5.4) The Fourier transform of the ideally sampled signal of Eq. (6.5.4) may be expressed as, Mc) =f, 3 X(f nf) sw (6.5.5) Comment : Eq. (6.5.5) gives the spectrum of ideally sampled signal. It shows that the spectrum | M,( is periodic in f, and weighted by f,, However, it may be noted that ideal or instantaneous sampling is possible only in theory since it is impossible to have a pulse whose width approaches zero. Practically flat-topSempling [Unt 6.12 6.5.2 Natural Sampling In natural sampling signal used is a pulse trains of some finite width (duration) +. In this sampling technique, the top of pulses follows the shape of the applied continuous-time signal. Fig. 6.5.3 shows the circuitry used to produce natural applied to a switching circuit is controlled by a sampling signal 9,(t), succession of pulses of width x and amplitude A. Let m,,(t) be the output of the sampling icy From Fig. 6.5.3, it is evident that when switch 's’ is ON, the output of circuit, ma,(t) will be produ of instantaneous value of m(t) and train of pulse. ily sampled signal. An analog signal my that consists of an infinite When switch ‘s’ is OFF, the output, m,,(t) will be zero. Mathematically, sampled output Is defineg as, rer e when g,(t) =A 0 ; when gp(t)=0 s 9.(0_ Fou mi(t) malt) [EDIE Fnctional Circuit Diagram of Natural sampling The waveforms in Fig. 6.5.4 shows the graphical representation of base band, carrier and naturally sampled output. ¢ mt) 0; t (2) Base band Signalao) 613, (©) Natural Sampled Output TEERERY craptical Representation of natural Samping Mathematical Analysis : Consider a message signal m(t) sampled by a sampling function g,(t) onsisting of a train of pulses of amplitude ‘A’ and duration x. Let T, be the period of the pulse train as shown in Fig. 6.5.4(b) The train of pulse (sampling function) is a periodic signal and can be expressed in the form of a complex Fourier series as, ACS of 9,(t) = FS sin | | evar The output of multiplier, m,,(t) shown in Fig. 6.5.4(c) is given by, Melt) = m(t) - 9,(t) At § len {2 everalney f ce 7M (0) ties Va 3 + (6.5.6) EG. (6.5.5) represents the time domain expression for natural sampled PAM signal. Applying Fourier transform to both sides of Eq. (6.5.6), We get the frequency-domain Fepresentation of a naturally sampled PAM signal as, ASG, Mf) rosin (nif) M(F - nf.) (65.7) * Frequency Spectrum of Naturally Sampled PAM Signal : Assuming the spectrum of message Signal to be bandlimited extending from (-W to W) Hz as shown in Fig. 6.5.5(a), then the frequency spectrum of naturally sampled PAM signal will be as shown in Fig. 6.5.5(b) mo) wow poe (a) Message Spectrum6.14 Sampling (Unit. 6) . 2 Frequency % T (b) Spectrum of Naturally Sampled PAM Signal [ERIE Fe cuency Spectrum of Naturally Sampled PAM Signal + In this illustration the sampling rate has been assumed to be equal to Nyquist rate 2W, so that there is no aliasing. The natural sampling, thus results in the mult ,Jlication of the Aran e( at ni lobe of the spectrum of the sampled signal by a factor 7-SINC| 7"). It can be clearly seen that the signal m(t) can be recovered from m,,(t) by passing m,,(t) through an ideal low-pass filter with cut-off frequency W. Thus we may conclude that the finite duration of the sampling pulses has no effect on the sampling process. 4 Limitation of Natural Sampling : In natural sampling, the top of the pulse at output is according to instantaneous value of applied signal m(t). So with in one output pulse, we have infinite number of amplitude variations. As power of signal is directly proportional to amplitude, so we have infinite values of power levels. Due to above said reason, the circuit should be capable to handle these variations. Practically this type of circuit design is very difficulty. 6.5.3 Flat Top Sampling In flat-top sampling, the top of the samples (pulses) is flat, i.e., the sample is having a constant ‘amplitude throughout the duration of pulse. Sample and hold; circuit is used to produce flat-top sampled signal as shown in Fig. 6.5.6(a). Value Holded by Capacitor ¢ Wer Veit Output, mA!) m(t) (2) Sample and Hold Circuit Generating Flat Top Sampled PAM (6) Flat-top Sampled Signal WGERERA Fat Top Samping Signals and Systems aSlL sample and Hol - ins PNG uns, ‘a Consists of two field effect transistor (FET) switches 2! BUR 0, bering sie losed for a short duration by short pulse applied to the the Instantaneous sample value ic the capacitor ‘C’ is quickly charged upto a voltage the capacitor °C’ holds the ne the incoming signal m(t). Now, the sampling switch 's We switch is then closed b ’arge until the pulse is applied at gate 2 of transistor Qa: YY pulse applied to gate G, of the other transistor Q,- OU® the capacitor °C’ is discharged to zero volts. The discharge switch is then opened and thus itor has no voltage. Hence, the output of the sample and hold circuit consists of a sequence of flat top samples” shown in.Fig. 6.5.6(b). ¢ Mathematical Analysis : The flat-top sampling can be better understood from ideal sampling: Consider the situation in which the signal m(t) is sampled instantaneously by ideal sampling 1 at 2 rate 7, but the duration of each sample is lengthened for a time x. Since the bandwidth — of transmission is inversely proportional to pulse duration, this type of sampling will reduce { the bandwidth requirement for transmission. The flat-top sampled signal m,(t) can be written as, mg(t) = mor,) At =n) (6.58) Where hit) is a rectangular pulse of unit amplitude, given by, fens ws (6.5.9) fa MDs fo j Otherwise The output of the ideal sampling technique, |-e» m,(t) is given by. mgt) ={mg(haeu = Dr Kt =0T) Att = nT, we have, mgt) =) rdor,) tn) convolving mt) with pulse M(t), We get, macnent= fmsAt-e = JOoeTD dentate a a Soret fac-orant-n ar =6. Di sting Using Using the shitting property of delta function, we get, matt) * h(t Sorwot ne-o7,) 654 Comparing Eqs. (6.5.9) and (6.5.10), we find that, my(t)=m,(t)*h(t) Thus from Eq, (6.5.11), we can say, “flat top sampled signal is equivalent to corootutcg of ideal sampled signal and rectangular pulse signal’, ‘Taking Fourier transform on both sides of Eq. (6.5.11) we get, Ma(f) = MA) H(A) ~» (6.5.12) Where M,(f), M,(f) and H(f) are the Fourier transforms of the flat-top sampled signal m, ideal sampled signal m,(t) and the pulse signal h(t) respectively. We have already seen that the Fourier transform of an ideal-sampled signal is given by, Substituting the value of M.(f) in Eq. (6.5.12), we get the frequency domain representation of a flat-top sampled PAM signal as, (6.5.13) @? Aperture Effect and Equalization : Aperture effect is defined as the distortion occurred in the reconstructed signal x(t) obtained from the flat-top sampled version of the signal, where in the amplitudes of the high-frequency components are reduced relative to the amplitudes of the low frequency components. For purpose of explantation of aperture effect, assume that m(t) has a spectrum as shown in Fig. 6.5.7(a). Consider the frequency domain representation of flat-top sampled signal 2&5 | defined by Eq. (6.5.13) as, a0-£ 9 H(-2] HN | (torn | Signals and systems | inea HT Where H({) Is a Fourler transform of rectangutar pulse h(t) of width + defined by Ea. (6.5.9) tically H(f) Is given by, P FLn(t)] = H(f) = (rsincfr)e2 sine(fr) ev" ‘Thus, H(f) is a sinc function as shown in Fig. 6.5.7(b), and will have its first zero crossings Wonly at f== and +1/x. Since + << T, these zero values of H(t) i.e, 41/s, wil be far away from f, and ~f,. Since us0-2yra(t-2] HAf) hence its plot will be as shown in Fig. 6.5.7(¢) TO) EAD RW. (2) Plot of M(A) IH(AL a (b) Plot of Hi), ut = 4) Mat) Characteristics of YA Reconstruction Fter“aan 6.18 Samaling (Uni. ey As can be seen in Fig. 6.5.7(c) amplitudes of the high-frequency components in flat-top Samp) ee signal are reduced relative to the amplitudes of the low-frequency components, because OF th Multiplication of m(f ~ nf,) by H(f). If we pass the sampled signal x4(t) through the reconstruc, tion filter, we find amplitude distortion of x(t). This distortion in x(t) caused by lengthening the wid of samples is referred to as the aperture effect. This distortion may be corrected by connecting an equalizer in cascade with the low.pa., 8 reconstruction filter as shown in Fig. 6.5.8. The equalizer has effect of decreasing the in-band ig of the reconstruction filter as the frequency increases in such a manner so as to compensate f, the aperture effect. Ideally, the amplitude response of the equalizer, H,(F) is given by, 1 a Hel = FraA]” esineh) aa rsine{fr) In practise the amount of equalization needed is usually small Equalizer [GEESE 8 ock Disoram of Reconstruction Circuit 6.5.4 Comparison of Sampling Techniques Reconstructed Message Signal Flat-top Sampled Signal x,At) Table 6.5.1 lists the comparison of three techniques. - EECIIERY Comparison of Tiree Sampling Techniques Heel ot, Natural Flat Top Instantaneous sala Sampling ‘Sampling High switching General switching | Sample and hold circuits used circuits used circuit is used f kK “Yann, ¥. mit) ies m9 | ace) m0) | my got cy ne ‘ { { ‘gy ‘Signals and Systems Peapeiscnm(t) mt) m,(t) Waveforms 1, Feasibility This method cannot | This method is used | This method is also be used practically | practically since | used practically since itis impossible | pulse train can be togenerateideal | generated impulse train of zero width Sampling rate Tends to infinity Nyquist criteria, Nyquist criteria. Noise Nominal Maximam Minimum interference Time = Ee m= 5 matt) =A£ my) = representation MAT.) a(t =nT,). |} m(t) sine{nf,) maT.) Xt —0T,) \ Frequency Mult domain Te representation xf =nf,) sine{nf,x) M(F ~ nf,) MMF = nf KF) 46 RECONSTRUCTION OF SIGNAL FROM ITS SAMPLES The original continuous-time signal x(t) can be reconstructed from the sampled signal x,(t) by a technique called interpolation. This reconstruction is accomplished by passing the sampled signal x,(t) through an ideal low- Pass filter of gain T, and having @ bandwidth of any value between W and (f, ~ W) none | PRctical view point, a good choice is the middle value f/2 = 1/2T, Hz or x/T, rad/s as shown in 6.6.1. With this choice of cutoff frequency and gain T,, the ideal low-pass filter required for ; € : reconstruction (or interpolarition) is, To6.20 Semplng [Unit 4) x(t) x MEER Frequency Response of a Practical low-Pass Filter The output of LPF shown in Fig. 6.6.1 in frequency-domain is given by, XC = XA) H(A) From the definition of inverse Fourier transform we have, 3 c wet, x(t) = Janene df= fro Hfje2** df = J T,X,(f)-e2** of (6.6.1) im ait, We know that the sampled signal x,(t) is obtained by multiplying analog signal x(t) with train of impulses, that is, s X,(t) = x(t) 6y,(0) = xt) > at-or,) a oy x(nT,) (t = nT.) Using the formula of the Fourier transform, we obtain, X,()= fumere dt= {> x(nT,) at ~ ma] enBat gt Changing the order of summation and integration, we get, X,(N= Dy x(0T,) fac aT,) e-P" t eS 1s0-| Sn Bla [ ficto x(t = x(0, 2 XA =) a(at,) ePas 4 iel substituting Eq. (6.6.2) in Eq, (6.6.1), we get any, 0-1, { [Sarpenm u vim pa Me) oot "| aPe Ot aT ¥ x01) je mr of oe wat, ft, sory [e201 2 [zr AD vr Srey | [eee eam Gate iy Y «or, sinxt-nt,)/7, (sono 2se? oy \ a4 : >} nor, [eat =nT,)/T,] aoe xt-s,)/T, | x(t) = rT, hy Yr, sn 2 (t-ary} : 5 (6.6.3) . (6.6.3) Is call Eq. (6.6.3) is called the interpolation formula. Eq. (6.6.3) indicates that the original signal x(t) ‘an be reconstructed by weighing each sample by a sinc function centered at the sample time and summing. The signal reconstructed is shown in Fig. 6.6.2 The sampling theorem discussed in section 6.2 applies to bandlimited signals centered at origin low-pass signals) for which X(f) = 0 for f> f,, where f,, is highest frequency component present ‘signal. Let us now cons der a more general case of bandpass signals with f, as highest frequency and f, as lower cut-off frequency centered at f, as shown in Fig. 6.7.1(2). AS can from Fig. 6.7.1(a), in case of band pass signals X(f) = 0 for all frequencies outside ff < fy. Let the bandwidth of the signal shown in Fig. 6.7.1(a) be B = fy = ‘Signalsve Semetng tn, From Fig. 6.7.1(a), it is evident that maximum frequency component present in the signal is ¢ | f.+f, and hence, the minimum sampling rate is expected to be 2(f,+f,), but actually it is much j, les than this value. The minimum sampling rate is specified as follows, Cast 1 (f, is Not an Integer Multiple of Bandwidth) : In this case, sampling theorem for bandpa, ss signals states that a bandpass signal x(t) which has a spectrum of bandwidth B = fy-f, can 4, WF. Can be recovered without any distortion from the sampled signal x,(t) by bandpass filtering, if the samp ing rate is such that, “-(e)-C) Where m is the largest integer less than f,/B. Case 2 : (f, is not an Integer Multiple of Bandwidth) : If we assume that the highest frequency ‘component present in the band pass signal is multiple of bandwidth, i.e., fy = mB = m(2f,,), then the band pass sampling theorem states that the band pass signal x(t) whose maximum bandwidth se is 2f,, can be completely recovered from its samples if the minimum sampling rate (or Nyquist rate) is equal to twice that of bandwidth, i.e., If the spectrum of the band pass signal is X(F), then the spectrum of the sampled band pass signal & ee 2 zy X(F ~ 2nB) X(f) ® ‘Where X,(f) is the sum ‘of the original Fourier transform X(f) and shifted replicas of X(f) and "then scaled by 1/T. Fig. 6.7.1(b) shows the spectrum of the original signal and sampled signal From this figure, we can find that the original signal can be recovered by passing x(t) though ‘an ideal band pass filter witha pass band aiven, by. f, < Ifl, 2f,). The required Leip on fy/B. If f, > 2fy, there will not be 27 2B would suffice and will not produce gee civdihe ieribie. ita ty tbe in ae’ : zs Signals cna System vt aliax) ° (2) Spectrum of Band Pass Signal x0 2 “ F t ° te & (©) Spectrum ct Sampled Band Pass Signal HO T es, +, ° i ats (€) Band Pass Fiter Characteristics to Recover Original Signal PERERA Sampling operation Aband pass signal has a spectrum as depicted in Fig. 6.7.2. What is the minimum sampling frequency that can be used? Show that no aliasing takes place when this sampling frequency 6.23Sempling [Uni “6.24 Sol: Here, fy, the highest frequency component is 30 kHz, The minimum sampling frequency is given by, 2, 2h far fa Where m is the largest integer less than 5, that is, (B= 30 - 22 = 9 kHz) ‘Therefore the largest integer value of m which satisfiers above inequality Is m = 3. Therefore = 2x == 20kHz fa2xF Fig. 6.7.3 shows the spectrum of sampled signal x,(F) at sampling rate f, = 20 kHz, From Fig, 6.7.3, it can be noticed that no aliasing occurs with this sampling frequency. x) AMIM.MMM.N -30-22-20-18 -10 -2 lo2 10 18 20 22 30 38 40 42 50 58.6062 70 78.80 82 90 100 f,-30f,-22 ) £422 £430 ‘Spectrum of Sampled Version of x 6.8 SOLVED PROBLEMS OF UNIVERSITY END EXAMS University Problem [Gl Determine the Nyquist sampling rate and Nyquist sampling interval for the signals, i) sine(10nt) ii) sinc?(1074) fii) sinc(100xt) + 3sinc?(60n1). [Now./Dec. 2009, 2010] ¥ : sin (100xt) Sol.: i) Given x(t) sinc(100xt) = = (opst) The highest frequency component in sin (100xt) = sina,t iS @, = 100x. Therefore, 0 = 1008 1008 50 Hz so = ‘ omssl Sa 6.25 Nyquist rate = 2f,, = 100 Hz and Nyquist interval= 1. 1 9.10 ms 2, 100 i) Given, . marl el Highest frequency component in 1/2(100st)? is zero. Highest frequency component in (1/2(100nt)? cos200xt = C0S(_it) iS ©; = 200%, So the maximum frequency component in x(t) is, my = Om = 2008 200; 2x =100 Nyquist rate = 2f,, = 2 x 100 = 200 Hz And Nyquist interval = 1/2f, = 1/200 = 0.5 ms. ii) Let, x(t) = sinc{100xt) + 3 sinc?(60xt) _ sin100zt | 5 pete ~"100nt 6Ont oat, [t= eset] 1 100xt 2 (60nt)? _ sin 100nt , __3. cos 1601) “qOoxt ” 3600x7t 2 : 1 . Dnt0det TB Aafan3 120 | = 100zt || zaaae] lar aaa (0 +0) Highest frequency component in x,(t) IS eq * 1008, 0. Highest frequency component in ¥3(t) 'S mz Highest frequency component in x(t) Is yy = 120%So the maximum frequency component in x(t) = Maxlimqrs Omar Omg) = Max{100%, Op 120%) = 120n. Therefore, ae 2. My = 1208 we tat coe a as 2. Nyquist rate = 21, = 120 Hz ‘And Nyquist interval = 1/2f,, = 1/120 = 8.33 ms. ‘A low pass signal x(t) has @ spectrum X(f) given by, X(f)= I= 597 |f1< 200 ‘Assume thot x(t) is ideally sampled at f, = 300 Hz, sketch the spectrum of x1). [Nov. - 2007] ‘Sol. : The given spectrum X(f) = 1 - (If1/200), If] < 200 is a triangular pulse as shown in Fig. 6.8.1(2). ‘From the spectrum, we observe that f,, = 200 Hz. Nyquist rate = 2 x 200 eo = 400 Hz ‘But given sampling frequency, f, = 300 Hz < 2f, (= 400 Hz). Therefore. oe MiG brotien ‘arises, and the resulting overlapped sampled signal spectrum X,(f) is as6.27 A signal x(1) = 2 cos 40:1 + 6 cos 640z1 is ideally sampled at f, = 500 Hz. If the sompled signal Is passed through an ideal low poss filter with @ cutoff frequency 400 Hz, what frequency components will appear in the output? Sketch the output spectrum. Also find the output signal. [May/June - 09] Given x(t) = 2cos400nt + 6cos640nt = x(t) = 2cos[2n(200)t] + 6cos[2x(320)t] Applying Fourier transform on both sides of above equation, we get, X(F) = 2{ Sef + 200) + aff - 220i) ps che + 320) + a(f - 320)} 2 2 The spectrum of the given signal x(t) is shown in Fig. 6.8.2. 3 KO 3 -320 -200 «10-200 30 fF EEREET Srectum ore) Given signal is sampled at a sampling rate f, = 500 Hz. Since spectrum of sampled signal x,(t) is given by, X02 xt -nf) Hence the spectrum of sampled signal is a periodic repitation of x(f) at regular intervals Of nf, Le., £500 n = 1, 2, 3....... Fig. 5.8.3 depicts the spectrum of sampled signal at f, = S00 Hz xt an, aT, an,tt, If the sempled signal shown in Fig. 6.8.3 is passed through an ideal low pass Fiter wi, ‘cutoff frequency of 400 Hz [shown in Fig. 6.8.4(a)], the frequency spectrum Of the outpue yt) will be as shown in Fig. 6.8.4(b). 40 =4002=«0 400 f — -320-300-200-180 10 180 200300320 (2) (b) Spectrum of Sampled Signal when Passed through LPF with Cut-off Frequency of 400 Hz “Thus from Fig. 6.8.4(b) the frequency components present at the output are #180 Hz, +200 Hz, #300 Hz and £320 Hz. The output signal in frequency domain is given by, Y(f)= 2ut +180) + oar - 180) + Sat +200) + hut - 200) : 5 : Fi + Lage + 300) + Lat - 300) + 2aee + 320) + 2a¢t - 320) T, Ts, T, 1; (* +180) af - oe rn 2{# = 200) 4 + zo Ts 6 = W-F| 2 [Alf - 300) + aff + 300)] , 6 [alt - 320) + Hf + 320) 2 T, > Applying inverse Fourier transform, to above equation we get, Bie craie ye) = Scostantsoyy + 2 costanc2001 + 2 costan(s00y] + $ cost2(320)) = 1/f, = 1/500. Therefore, = 3000cos(360zt) + 1000cos(400xt) + 1000cos(600zt) + 3000cos(640x!) : a a. f, = 50 kHz [Nov. - 2008}, [Feb. - 2008)BSS Mics 5 0 a9 I 8 Spectrum of Band Pass Sigral] When f, = 25 kHz, the ee vs CorTesponding spectrum of the sampled signal X,(f) is shown in “Fig, 68.6. Tels a periodic repetition of D. ) at regular intervals of tnf,, ie, + 25n, 9 = Ss wo i oss as al oe 25 teas f felS fy25 EG Spectrum When f, = 25 kite From Fig. 6.8.6, we see that x(t) can be recovered from the sampled signal by using 2 bandpass filter of the frequency response. iz Ty, fe Sf] s 25kH2 with 10 kHz s fe = 15 kHz 0, Otherwise 9 { When f, = 45 kHz, the corresponding spectrum of sampled signal is shown in Fig. 6.8.7 from Fig. 6.8.7 we notice that x(t) cannot be recovered from the sampled signal x(t). x) Aliasing nasi | 15 ° 15 30 45 60 Spectrum When f= 45 kt orresponding spectrum of sampled signal § shown in Fig, 6.8.8 (at regular intervals of nf, Le, SOM where n= t, 2, 3, 60 45 -30 f, = 50 kHz, the c fa periodic repetition of Signals and Systems. —15 WERE stectim went = soe From Fig. 6.8.8, we see that x(t) can be recovered from the sa ideal low-pass filter of the frequency response. impled signal by using wo-f i fl s25kH2 0 5 Otherwise