H OMEWORK 2.1 I) Show that for the harmonic process H(t) = A cos (ωt + T )with A and T are independent RVs, and T uniformly distributed in [−π, π] the following is true: 1 Rπ g(T ) dT ; (1) The expected value of a function g(T ) is E {g(T )} = 2π −π (2) Mean: µH = 0; 1 2 (3) Variance: σH = E |A|2 ; 2 1 2 (4) Auto-correlation: RHH (t1 , t2 ) = RHH (∆t) = E |A|2 cos ω∆t = σH cos ω∆t; 2 Proof. (1) By definition Z ∞ 1 E {g(T )} = g(u) fT (u) du; 2π −∞ Z π g(T ) dT, −π but since T uniformly distributed in [−π, π], also by definition, its PDF should be 1 −π ≤ u ≤ π ; fT (u) = 2π 0 otherwise Substituting, the expected value of g(T ) becomes Z π 1 E {g(T )} = g(T ) dT. 2π −π (2) Since RVs A and T are independent (uncorrelated would also work) E {A g(T )} = E {A} E {g (T )}, the mean is Z π 1 µH = E {A cos (ωt + T )} = E {A} E {cos (ωt + T )} = E {A} cos(ωt + T )dT, 2π −π which, with the variable change θ = T + ωt, dθ = dT , T = ±π⇒ θ = ±π + ωt, therefore Z E {A} π+ωt E {A} E {A} µH = cos θdθ = [sin θ]π+ωt [sin (π + ωt) − sin (−π + ωt)] , −π+ωt = 2π 2π 2π −π+ωt and using the trigonometric formula for the difference of sines E {A} 2 sin π cos (ωt) = 0. 2π (3) Using the same change of variables as before: Z 2 2 E {A2 } π 2 2 σH (t) = E H (t) = E A cos (ωt + T ) = cos2 (ωt + T )dT, 2π −π 2 Z π+ωt E {A } 2 σH (t) = cos2 θdθ. 2π −π+ωt µH = 1 An easy way to calculate the above integral is as follows: denote Z π+a 1 cos2 θdθ, I = 2π −π+a Z π+a 1 J = sin2 θdθ; 2π −π+a then Z π+a Z π+a 1 1 2 2 cos θ + sin θ dθ = dθ = 1, I +J = 2π −π+a 2π −π+a Z π+a Z π+a 1 1 2 2 cos θ − sin θ dθ = cos 2θdθ = 0; I −J = 2π −π+a 2π −π+a 1 Therefore I = J = , hence the result. 2 (4) Auto-correlation: RHH (t1 , t2 ) = E {H(t1 )H(t2 )} = E A2 cos (ωt1 + T ) cos (ωt2 + T ) = Z π 2 1 = E A cos(ωt1 + T ) cos(ωt2 + T )dT, 2π −π where 1 2π Z π Z 1 cos(ωt1 + T ) cos(ωt2 + T )dT = 2π −π π −π 1 [cos(ω(t1 + t2 ) + 2T ) + cos ω(t1 − t2 )] dT ; 2 the integral of the firsty term is zero (see the calculation of the mean), and the second term yields Z π Z π 1 1 1 1 1 cos ω(t1 − t2 )dT = cos ω(t1 − t2 ) dT = cos ω∆t, 2π −π 2 2 2π −π 2 with ∆t = t2 − t1 , thus 1 2 RHH (t1 , t2 ) = E A2 cos ω∆t = σH cos ω∆t. 2 P II) Let H(t) = N n=1 Hn (t), where Hn (t) are harmonic processes, Hn (t) = An cos (ωn t + Tn ) with An and Tn mutually independent RVs, and Tn uniformly distributed in [−π, π] the following is true. Show that: (1) Processes Hn (t) are orthogonal, i.e. hHn , Hm i = Rnm (t1 , t2 ) = Rnn (t1 , t2 )δnm (2) The auto-correlation RHH (t1 , t2 ) = of process Hn . PN n=1 2 ( Rnn (t1 , t2 ) for m = n = ; 0 for m 6= n σn2 cos ωn (t1 − t2 ), where σn2 is the variance Proof. (1) Orthogonality: from the definitions of the inner product and process H hHn , Hm i = Rnm (t1 , t2 ) = E {H(t1 )H(t2 )} = = E {An Am cos (ωn t1 + Tn ) cos (ωm t2 + Tm )} = = E {An Am } E {cos (ωn t1 + Tn ) cos (ωm t2 + Tm )} . Since for m 6= n Tn and Tm are independent, the mean factorizes into the means of the factors E {cos (ωn t1 + Tn ) cos (ωm t2 + Tm )} = E {cos (ωn t1 + Tn )} E {cos (ωm t2 + Tm )} and the product cancels (see I.1), so in fact E {cos (ωn t1 + Tn ) cos (ωm t2 + Tm )} = E {cos (ωn t1 + Tn )} E {cos (ωn t2 + Tn )} δnm , that is hHn , Hm i = E {An Am } E {cos (ωn t1 + Tn )} E {cos (ωn t2 + Tn )} δnm , = E A2n E {cos (ωn t1 + Tn )} E {cos (ωn t2 + Tn )} δnm = = Rnn (t1 , t2 )δnm . (2) Auto-correlation: using the orthogonality of the processes RHH (t1 , t2 ) = N X N X Rnn (t1 , t2 )δnm = n=1 m=1 = N X N X Rnn (t1 , t2 ) n=1 Rnn (t1 , t2 ) = n=1 N X N X δnm m=1 σn2 cos ωn ∆t. n=1 III) Show that if a process X(t) is 2-order stationary and X 0 (t) is its derivative with respect to time, for any given time t RVs X(t) and X 0 (t) are orthogonal and uncorrelated. Hint: assume that you can exchange the order of the time differentiation and the expected value operator. Proof. Using the linearity of the mean 1 d 2 1 d 2 dX dX 1d 2 X = E X = σ =0 X, =E X =E dt dt 2 dt 2 dt 2 dt X 3