Spectral Characteristics of Random Processes CIOINITIEINITIS = a Power density spectrum and its properties Relationship between power spectrum and autocorrelation function Cross-Power density spectrum and its properties Relationship between Cross-power spectrum and Cross-correlation function Scanned with CamScannerGATE ACADEMY PUBLICATIONS ™ Probability & Random Vs ‘ wetion is the time domain i.e. we have characterized oa Pree te onelaon. eosecorelation, and covariance functions Saat iy Gorelderation of spectral properties. Asis en ee eed domain and frequency domain analysis methods exist for analyzing a systems and deterministic waveforms. But what about random waveforms? Is there some way to describe random processes in the frequency domain? The answer is yes, and itis the purpose of this unit to introduce the most important concepts that apply to characterizing random processes in the frequency domain. 1._ Power Density Spectrum & Its Properties The spectral properties of a deterministic signal x(t) are contained in its Fourier transform X(a) given by ables 5-2 XO)=fis0eRds oe @ The function X(@), sometimes called simply the spectrum of x(t), has the unit of volts per hertz when x(t) is a voltage and describes the way in which relative signal voltage is distributed with frequency. The Fourier transform can, therefore, be considered to be a voltage density spectrum applicable to x(¢). Both the amplitudes and the phases of the frequencies present in x(/) are described by X(o). For this reason, if X(w) is known then x(t) can be recovered by means of the inverse Fourier transform 10-2] x@emao (ii) In other words, X(@) forms a complete description of x(t) and vice versa. In attempting to apply (i) to a random process, we immediately encounter problems. 2. Other problems arise if Laplace transforms are considered, On the other hand, if we turn. ow attention to the description of the power in the function does exist. This function is called th oes Seca ts tat mh Scanned with CamScannerACADEMY. PUBLICATIONS ™ 5-3 Spectral Characteristi & The Power Density Spectrum are Random Processes "For a random process X(@), let x,(t) be defined as that portion SRT x(t) that exists between ~T and 7; that is of a sample function 10={0 Teter 0 elsewhere + (iii) Now so long as Tis finite, we presume that x,(1) has bounded variation, will <, in a inded variation, will satisfy r |z-(|dt< eo J (Odean 8 AE SNR E Mees 8h sete (iv) and will have a Fourier transform which we denote X,(o), given by r r X,(@)= feat =f xea a J tr J (edt (vy) ‘The energy contained in x(¢) in the interval (-T,7) is r r ED) = | e@at= | ear ssseee (Vi) a : Since x,(t) is Fourier transformable, its energy must also be related to X,(o)by Parseval's theorem. Thus, from (vi) r Le Q)= | "Od Fog Jef do sseees (Vil) By dividing the expressions in (vii) by 27, we obtain the average power P(T) in x(t) over the interval (-7,T): _ P(T)= = j ¥Odt= x j Lecco do (viii) a : At this point we observe that |2,(@)|" /27 is a power density spectrum because power Tesults through its integration. However, it is not the function that we seek, for two reasons. One is the fact that (Vili) does not represent the power in an entire sample function. There remains the step of letting T become arbitrarily large so as to include all power in the ensemble member. The second reason is that(viii) is only the power in one sample function and does not Tepresent the process. In other words, P(T) is actually Process. By taking the expected value in (viii), the random process. te Hence we must still form the limit as 7 Obtain a suitable power density spectrum °Perations are performed, (vii) can be written a random variable with respect to the random we can obtain an average power Py, for co and take the expected value of (viii) for the random process. After these Scanned with CamScanner —Probability & Random Variables 5-4 GATE ACADEMY PUBLICATIONS «Fx, (oy'] 1 intl a fafa] tea nt facts. First, average power Py in a randoy dO seesss (ix) Equation (ix) establishes two importai i process X(t) is given by the time average of its second moment: Pre singer] =4{e[X70} sees (8) ; FA exe For a process that is at least wide-sense stationary, E[X*()]=2", a constant, a Py. =X". Second, Pa can be obtained by a frequency domain integration. If wi define the power density spectrum for the random process by z[x,@f] 2T the applicable integral, which we call the power formula, is ++ (a) Sxx(@) = lim Pry = 7] 5aoae saves (ii) from (ix). Properties of the Power Density Spectrum : The power density spectrum possesses a number of important properties 1 Sq(o)20 2. Sy(-0)=Sy(a) X(P real 3. Spx(0) is real. 4. Fp] Sado afer} Property 1 follows from the definition (xi) and the fact that the expected value of a nonnegative function is nonnegative. Similarly, property 3 is true from (xi) since |x, is real. 5. Seo) = 0S px () sss (Xvi) It says that the power density spectrum of the derivative X(t) =dX(1)/ dt is a’ times| the power, spectrum of X(t). 6 ¢ I Sq (o)edo= A[Ry(t,1+2)] Sonn Sq(o)=fAlRatttgleMdr (xix) It states that the power density spectrum and the time average of the auto-correlation function form a Fourier transform pair, Property 6 indicates that the power spectrum and the autocorrelation function form a Fourier transform pair, Thus a Scanned with CamScannerAcaDEMY PUBLICATIONS 74 a8 Sye (0 1h 1 Red = 5 f Sulorerde for a wide-sense stationary process, _ pandwidth of the Power Density Spectrum: Assume that X(0) is a low-pass Process; that is, its spectral clustered near Gy and have decreasing magnitudes at highs testa ic ies. processes are ag called baseband, Except for the fact that the area a (w)is ra necessarily unity, Sy(@) has characteristics similar to a probability density function (it is nonnegative and real). Indeed, by dividing S,,(w) by its area, a new function is formed with area of unity that is analogous to a density function, The analogous quantity for the normalized power spectrum is a measure of its spread that we call rms bandwidth, which we denote W,, (rad/s). Now since Sye(.) is an even function for a real process, its “mean value” is zero and its “standard deviation” is the square root of its second moment. Thus, upon normalization, the rms bandwidth is given by J O'S, (0) do WZ, =—___ J Sx(@)do The above concept is readily extended to a process that has a bandpass form of power spectrum; that is, its significant spectral components cluster near some frequencies G,and -&,. If we assume that the process X(l) is real, Sy-() will be real and gave even symmetry about =0. With this assumption we define a mean frequency &, by Jose (o)do Bett ty Sintes BN oe (oii) =e JSec(o)do 0 and rms bandwidth by sfo-B' Su (odo ee f Sy (o)do : Wore Scanned with CamScannerON Probability & Random Variables Example 1.1 Consider the random process X(0)= A, cos(iogt +8) 5 where. 4, and oy are real:constant and 0 is random variable uniformly distributed} on the interval (0, n/2), Find the average power Pyin X().__ [CSVTU May 2014) Given that, X(t) = 4, cos(w,t +0) where 4, and @, are real constant and @ is random variable uniformly distributed on the interval (0, n/2) GATE ACADEMY PUBLICATIONS ™ 5-6 Sol. ELX2@)]= ELA cos*(a+0)] = [40 +08 2(0t+ oy] 2 yas = [f+ 4 £08 2(ay oo) J 0s 2(eyt + 0)d0 4 sn (20,1) 2 | This process is not even ildeasse stationary, since the above function is time- dependent. The time average of above expression is AELO]}= = ins | i [| $-4meara] [xo] 4 on the interval (0, x i (@) Is X(@) wide sense stationary? "© Find the power spectrum of XO “and ee? i ‘Given that, — X() = A.cos(a,t +0) where 4, and a, are real constant and @ is a random on the interval (0, 7). (@) The mean value of random process is given by, E[x@]= fe Sado Power, Do your two powers | variable uniformly distributed Scanned with CamScanneryy ACADEMY PUBLICATIONS 14 i gat $27 _ Spectral Charvctortation of Random Procawson ALYO)s Heewiarodana: ‘tle oh ® cox(0j1 1. 0)10 *fsin(aye eoyp ~4 [-rinoyeainaye] ait : int \ 4 This mean value is. timed dependent so (0) ly not whde-uense stationary. Ans. of random process is given by, BLO] EL cosy Js “fs 1 Zeon, in | (b) The mean square value 7 4 +5 [Semcoy + 2» | * EF 20,40 2 -4 Lena 420)] Ae DAO 4 4 Thus, Py = fins j BLM Od fi * Vjded Ans. (©) Fourier transform of random process X() can be calculated as, X,(0)= i Ay cos((ogt + O)e"! dt = js [etn ne tv ertede * cveral eo fomena wna} feomea] Ayn [eT Tarn [em] Te maa) [2 | aera) Ae’ fel etl, Act (eo =) “eo = 2) (@, F9) ap sin(@-@,)T ” ina =a] (0-®) J+ [ (o+o,) Next we find |X, (a)|’ = X;(0) X7(@) : Ix} =[TAe"Sao- —c,)P + ATeSa(o+ 4)? ] [Tae PSa(o- O4)T + ATeSa(o+0,)T] T3Sa (co Wy T+ AT 2€7*" Sa co+ Og )Sa(C0— Oy) : *e AT Salw+Oy)TS,(0-Oy)T EM + AT*Sa2 (+ 04)T Scanned with CamScanner t 4 ! { | | I' ‘ Probability & Random Variables 5-8 GATE ACADEMY PUBLICATIONS 1% Neglecting cross terms each having factors at widely separate frequencies «, ~«, . Such that their product is small. We get, |x, Cf = 477Sa*(o-0,)T + AT*Sa°(@+ @,)T ELL L8fT sq ast Zsiove] 2T 2a « Thus, power spectrum is the limit at 70. We know that, lim =| [saat] =8(a)=[Sa(a7)]* rex| oF Sq(0)= 87 [e(0-«))+8(0+0,)] Average power can be calculated as, ra=t]s (do= 1] B2[a10-0,)+50+0,)]40=4 Ang fala a “Int 2 2 The to power auton do agree Example 1.3 We are given the random process X(0) = 4 sin(ajt+8) Where 4, and , are real constant and 0 is a random variable uniformly distribut on the interval (0, x). (a) Is X(®) wide sense stationary? (©) Find the power in X(). (© Find the power spectrum of X(t) and calculated power. Do your two es agree? : Sol. Giventhat, —_X()= Asin(a,t+8) Where 4, and @, are real constant and @ is a random variable uniformly distributed| on the interval (0, x ). (2) Themean value of random process is given by, z[x@]= { x f,(0)d0 = ja sin (on +)440 =A feostos +f, = eosuy X(t) is not wide-sense stationary because E[X(0)] is time dependent. (b) The mean square value of random process is given by, £[7@]= [4 sin*(oy+0)]= [40 ~cos2(wst+ al] 2 tip 4-4 gfcos2(0,1+0) Scanned with CamScanneree a8 ACADEMY PUBLICATIONS ™ ELX (]= (©) Fourier transform of random process X(t) can be calculated as, 7 a eee gL __ir 2j f@-) 27 I, Tay | P 1 A fellas estes X;(o) = i Asin (oy + 0)e™ar= Ale = ] ma aA we Keno) A nlf fr] fom va] Aen ; r Alen lL A” fa leone OT) a go] 2 H(,-0) 2 F(O, +0) =Aen( smell) A ea(seieteyr) i U@a) Ji (ora) = AT oPSa(o-o,)P -€*Sa(0+ 0, yr] @l= Fe {Sa?[(o-0,)T]+5a? [(o+0,)T]} Sn(0) = AZ [5(0-e)+K0+0)] Average power can be clesitied as, Py =e xl Sp(a)do= [a0 -0,) +5(0+0,)]d0=-4 Ans. The two power calculations do agree. Scanned with CamScannerGATE ACADEMY PUBLICATIONS m Probability & Random Variables 5-10 Sol. (a) Sy(a)= a oy +3 Since S.q (o) is even function and for all @ Sy 20. Hence Sy(w) is a valid, ©) Sy(@)=ee" Since S_y(«) not an even function of w. Hence Sy, (w) not a valid function. 7 © Sa()=7-8@) For @=0, Sy(@)=0-1=-1 Not a meee and hence not a valid function. @) Sy@)=— | Tea + jo" S,q(@) should be real. Hence not a valid function. as o ) © Sa(o)= For ©= 1, Sy-()=——; = negative I+: rd Not notregutive and hence not a valid power density spectrum. O Sa)= ey Since Sy-(w) is even function for all @and Sy(o)20. Hence S,,(w) is a valid power density spectrum. -— ® Su= oF Sx(@)# Sy(-0) Hence not a valid power density spectrum. 1 H) Syx(o)= ar For 20, Sy-(o) is imaginary. But Syy' (@) should always be real and hence not a valid power density spectrum Scanned with CamScanner\ PUBLICATIONS ™ qaTB ACADEMY NB: 5-11_ Spectral Charactoristica of Random Processes RMS bandwidth is given by, Jo'Sy(o)do W2, ==—___ J Sw(oydo 7 Ps [sacordo= | [rope 20s 4pw Wii, on fe Syx(@)do = t Po on(22)a0 -v WwW 7 i” ~oos( 38 2) ~sin( 32 2) =2e] 2W}) ,2|__\2v) =/2W (wy | *7|%erawy ‘Example 1.6 is Determine the rms bandwidth of the power spectrums given by : ie (spel? 2 lS vey satay altel bolsw A 0. ol>w 5 iel> where P and W are real sitive constants. lol< jof> 7 7 ” J Sp(w)do= f Pdo=2P¥ 4 iv . Ed Sol, (a) Given that, saio-{y : no all Din Jors.c(oda= J otrdo=? yoy forse(ordo 2 py? a Won, 2 pS J Sarto Ans, Scanned with CamScanner— 5-12 _ GATE ACADEMY PUBLICATIONS ny ‘eel me (b) Given that, Sy(o)= wv 0 [ol>w Jsatodo f off] ae “fp rfie@)eos]e(-§ ia =f, «(oT ere ble] oofon Jorn 20 =PW 7 Jo'sa(oro= joel +e} do= oI rw'-(2)] cof ERE af rie = Probability & Random Vari 4W 304 6 Jose(de eset pyrig_ PY 6 j Sqq(o)do Sol. Giventhat, — Ryy(t)=3+ 2exp(-4r) (@) We know that, Ry) Sy, (0) etn olin etn 1 1 Here >= oe 20 Bras 8 inet ME ne 2 Sex) = 320) 80) + Ves " ‘Ans. Scanned with CamScannersai cADEMY PUBLICATIONS ™ 5-13 Spectral Characteristics of Random Processes } (9) Average power can be calculated as, Ry (0)= Sout sw Ans. (9 Power in ye sey -f [6x8(0)+ Jee “"de] 7 0128) 2342 f ee aye aea| ef 2 Wz oa seal (U2)- (ae =3+a[(2)-1+e(J)]-3+2120598-4 = 3.3948 Ans, [Example 1.8 ‘ é ‘Assume a random process has a power spectrum. ce (@?/9) ise: 0 otherwise Find : i (a) The average power. (b) The rms bandwidth, (Q) The autocorrelation: function of the process. So Giventhat, — Syq(w) = {fe Co) pil S otherwise (@) The average power can be calculated as, 1? 1t(, o& Per = 32 J ae aan Ans. ™ 27 Bd : &» 1 Fars wdo= thor a0 Scanned with CamScannerJsulodo (6) We know that, Ry (1)? Sex) Lt eh f(s 2 ema Ralt)= 52 | Saloe™de aC =e : if 1 to! sg = 7g] edo 3 | 7 ae sale], “nit 2 4 sin6t 7 t = 74 sa(6r) OUR € Lee teal, is fy os. =—| LW 56 ay 2 2 36 18n|" jt je it ie 36sin6r_sin6t , 12cos6r al S6sinGs " sin6r | 12cos6e Onl o 7 ~1[ SBS 2ainte Icons aL ot OP OOF isha tees seh + Scanned with CamScannerGate. ‘sole The average power is given by, oy Py = mei sli (Given the power peti 1Se(@)aeelOn S [+ enor} where the dB bandwidths 10 radians per second ind [CSVTU Dec 2013} : Az) leas we I= ‘2 ike We ten (2) : j oS, (0)do Sol, Rms bandwidth is given by, W72, =" J Sx@)do First calculate, _f__10do _igs f__do Ise Kets lietoy © laoreF 1 =10° 2. + tan {ean a |. 2000 t F__l0a*dw sf _ ade ees) [1+@rtor]" “lore J stimi(2y |. = saastoax| |.*20" (aso Wp, = [00% = 10rad’s Ans. Scanned with CamScannerProbability & Random Variables NS 16 GATE ACADEMY PUBLICATIONS 1 eee aes“ . Relationship Between Power Spectrum & Autocorrelation Function (csvru May 2014) It was stated that the inverse Fourier transform of the power density spectrum is the time average of the autocorrelation function; that is (i) Le ps Fy J Sa lode" deo = A[Re(t +9)] This expression will now be proved. t : We have Sx (0) = lim 8 ¥ Jxqpelar, J xieras] * ES Gi) im if yj) Hinge | | BEKO XG emda ‘The expectation in the integrand of (ii) is identified as the autocorrelation function of X(): E[XG)X(4)]= Reet 4) -T <(f, and t,) the condition -7'<1+t0 and A, are constants. Find Sy(@).\ 5 ss 1-(r|/7)] -TsesT | Sol. Giventhat, Ry (t)= ah 7] . elsewhere | Ry; (2) is sketched in figure (a) for 4, > 0. The power spectrum is found from the Fourier transform Sy (0) = ji Ry (etdt 4 1 jer =A f [eq/nlemar+Aff-G/ Me dt ay ‘ Scanned with CamScanneri ee Probability & Random Variables 5-18 _ GATE ACADEMY PUBLICATIONS n Syy(@) = AT Sa*(@T 12) which is sketched in figure (b). Rye(0) Al) =r 0 T =m oe 0 ee r T (b) T T Fig. The autocorrelation function (a), and power spectrum (b), for the wide-sense stationary process Ry()= Pet Power spectrum is given by, Sw(o)= J PeMe made ° ° 040) ory =Pf eomdes Pf erg, = p|_© a +A @ Se + jo) 3- jo}. Scanned with CamScannerye EMYPUBLICATIONS™ —g_ é is nee an 49 _Spectral Characteristics of Random Processes _ . Sm (0) P i 2PI3 7 t ‘Auto-correlation Function pamele 23 ‘The autocorrelation function of random variable X is R, bar vay! (6) = c0s(ayi) where 4, te i 3; and o, are constant. Find power spectrum and plot. ICSVTU May 2014] Jo. From Wiener-Khintchine relations, Reg) Se (0) ie. autocorrelation and power spectral density are Fourier transform pair 2 2 : Ry(t)= Scostaye) = 4 [e= tem] From frequency-shifting property of Fourier transforms given by e** € 2nB(w-0,) e™ € 2n5(0+0,) By using this result, the Fourier to transform of R,,(t) becomes a 3y(o) = 2% [5(0-0,)+8(0+0,) This function and Ry, (x) are illustrated in figure. Ralf 4 2 ~ Scanned with CamScannerarse ee cement i, | Probability & Random Variables 5-20 GATE ACADEMY PUBLICATIONS ™ Example 2.4 A random process in given by IP()= AX()+ YO) ny : : where A and Bare real constant and X(0) and Y(t) are jointly wide-sense stationary Processes, (a) Find the power spectrum Syy (1) of wo. (b) Find Syy(«) if X(#) and ¥() are uncorrelated. (©) Find cross power spectrum Sy» («) and Sr(@)- Sol. Given that, — W(t)= AX(Q+BY(0) where A and B are real constant and X(¢) and ¥(‘) are jointly wide-sense stationary Processes. (a) The autocorrelation is given by, Rew (by 141) = E[{AX(9 + BY OHA +2)BY¢+0}] = APR yy (1) + B?Ry (1) + ABRyy (4) + ABR yy (2) We know that, Ry, (1) Sy, (a) Sy (@) = A°S yx (@) + B’Syy (0) + ABS py (@) + ABS (@) Ans, (b) For uncorrelated X() and Y(0), Ry (t)=X¥ Ry(=XY Sy (@) = AS yx (0) + B?S yy (w) + 42. X ¥ ABS(@) Ans. (0) The cross correlation is given by, Rey (t, +2) = ELX(@{AXG+0)+BY(t+0)}] = ARyy (1) + BRyy (1) Srp (©) = AS yx (@) + BS (0) Rey (t, +2) = ELYO{AX(+1)4 BY(+0}] = ARy (1) + BRy(1) Syp (3) = ASye(0)+ BS (0) Ans. f The cross-correlation of jointly wide-sense stationary processes X(t) and ¥(0) is assumed to be a Ryy (9) = Bu(s)exp(-W) where B>0O and W>0 are constants. (a) Find Ry (x). (b) Find S,,(o) and S,.(o), Sol. Given that, Ry (2) = Bu (t)exp(-W2) where B>0 and W>0 are constants, (a) We know that, | Rox (8) = Rey (—1) = Bu(-t)exp (Wz) ‘Ans. | Scanned with CamScannerr ACADEMY PUBLICATIONS 7 5:21 Spectral Characteristics of Random P () Power spectrum is given by, Sm(@)=F [Ra ()]= f Bulsyexp(-Wedze™ . feet -maee B Ee Wijo Ans. We know that, Srx(@) = Syy(-@) = 3 bial) W-jo Ans. 26 Find the cross-correlation function corresponding tothe cross power spectrum i 6 RALMOT CID: oe Sx (@)=G(@)G,(@) 6 1 G(oy=—,, Eis where A (@) oor OO FF a= and g,(t)=te* u(t) Ra = | a @se-de= | g@g(r—-HePae For r<0, Ry()= fe pe a= be ° : For t20, Re (O= fe ede + fe (e-Qe ae = (ee =i Thus, Ry (2) = 59 [od u(s)+u(—1)] il, Cross-Power Density Spectrum & Its Properties Consider a real random process I¥(t) given by the sum of two other real processes X (Nand ¥(): WO=XOHVOQ eae @ The autocorrelation function of 1(¢) is Rew (t+) = EP OW (+O) =E(XOFYO|XG+9+¥C4+0]}} Reg (tt #1) + Roy (t+ 1) + Rey (tt 0) + Rey (+0) «(iD Now if we take the time average of both sides of (ii) and Fourier transform the Tesulting expression by applying Wiener Khintchine relation, we have Spy (0) = Sipe (00) + Syy (@) + F{A[Ryy (t+ O)]} + F {A[Rre (tt +2}. Gli) Scanned with CamScanner2 GATE ACADEMY PUBLICATIONS ™ , Probability & Random Variables where F{} represents the Fourier t the power spectrum of I¥(0) Similarly, spectrums of A(t) and Ya), respectively. cross-power density spectrums. 1. The Cross-Power Density Spectrum + For two real random processes X(t) and ¥(#), we define x,() and y,(t) as truncated ensemble members; that is ransform. It is clear that the left side of (ii) is just the first two right-side terms are the power The second two right-side terms are called “fa -T 0: ila t e £[X;(@)¥, r=} | Aatin= 2] LAO Hie EO gy an 8) Finally, It is clear that the integrand involving w can be defined as a cross-power density it is a function of @ which we denote spectrum; Sp (= im EL eterno) a) Scanned with CamScannerof Random Proces sonDentd PUBLICATIONS™ 5-23 Spectral Character 8 what we call the cross-power formula A Py 3] Su(ordo xiii) y repeating the above procedure, we can also define another cross-power density spectrum by E[Y%()X} Sy(0) = Jim aioe) saies (xiv) Cross power is given by 1¢ . Px =, | Salo)do= Fy seseen (XV) Total cross power Pr, + Fy can be interpreted as the additional power two processes are capable of generating, over and above their individual powers, due to the fact that they are not orthogonal. 2. Properties of the Cross-Power Density Spectrum: Some properties of the cross-power spectrum of real random processes X(t) and Y(#) are listed below. 1. Sq(0)=Sx(-@) = Six(@) Re[Sxy(@)] and Re[ Sy, (w)] are even functions of o. Im[S,7(0)] and Im[S,,(0)] are odd functions of w. Sp(@) =Oand S,,(@)=0is X(*) and Y(t) are orthogonal. If X(t) and Y(#) are uncorrelated and have constant means X and Y Syy (©) = Spy (w) = 20. ¥ ¥ 8(@) 6. ARG t+D] & Sq) A[Ry (4t+9] O S,,(0) - Qodi) In the above properties, Re[-] and Im[-] represent the real and imaginary parts, respectively, and A[-] represents the time average. Property 6 states that the cross-power density spectrum and the time average of the cross-correlation function are a Fourier transform pair. For the case of jointly wide- sense stationary processes, (xxi) and (xxii) reduce to the especially useful forms Spa) = J Ry (Dede see =) - (ox) iss (ey Raia ae NE EOE (xxiv) Ry = 7 PSeteremda Pens ie ce. at metals (xxv) Ra O=Z | Su (werd Scanned with CamScanner7 - Probability & Random Variables 4 __ GATE ACADEMY PUBLICATIONS 74 | Example 3.4 : ae © Suppose we are given a cross-power spectrum defined by ‘a+ jbolW -Wa and bare real constants. Find the cross-correlation function. “W 0, a and b are real constants. The cross-correlation function is given by, 1 Ef boys 46 ot Feapecagrey DIE, ee bo) ast dey = [ dar j—— [wel Ry) LT (euRe do a ot Taw Lee ‘do This expression will reduce to od ” ale foe 1 (0a SL, awe Ee a, 1 |(aWx—b)sin(W)+bWrcos(W9)] Ans. 7s IV. Relationship Between Cross-Power Spectrum & Cross-Correlation Function In this we show that Sy(@)= iin | ant role ‘de cones The development consists of using the transforms of the truncated processes given by X,(@)= j Xe at es : ¥,(o)= J Year, o First, we use (ii) and (ii) to form A : X7@)¥,(@) = J X(e"dt f Y(h)e™*dt, EL% om oy We know that Syy(c)= lim. rf X(eldt 5 V(nye tah lim —— ar! =f fro telat at, a i Scanned with CamScannerre il = fins J J Re tnd@G, -1-1)dh, dt .- (vi) sme definition ofthe impulse allows the immediate solution for the integral over 4, 1t "dep = li rr J Px (oe**do = lim (vii) r ap Rater which is valid for ~7'@ in the limit. Equation (vii) indicates that the cross-power spectrum and the time-average cross-correlation function form a Fourier transform pair. The result proves (i) because itis the direct transform part of the pair. It should be noted from (vii) that given the cross-power spectrum, the cross- correlation function cannot in general be recovered; only its time average can. For jointly wide-sense stationary processes, however, the cross-correlation function R(t) canbe found from S,,(«) since its time average is just Ryy(t) - Example 4.1 Let the cross-correlation function of two processes (i) and Y(0) be Ry (t+) = A sinto,2) +-c0s[«,(21 +} where 4, Band @, are constants. Find the cross-power spectrum. Given that, Ry(ttt+d = 2 fsintays)+c0s[a,(2r+9)} where A, Band «, are constants. First, the time average is formed Cee | ae lf B fim ap | Bart Dat Zr sinlogt)+ > iar j cos[a,(21+1)]dt cf The average value of cosine is find to be zero over a cycle. Finally Fourier transform the time-averaged cross-correlation function we get Sqy(0)= F{ Beincao} - [5(o-a,)-8(@+0,)] Ans. 2 ‘Scanned with CamScanner