MASSACHUSETTS INSTITUTE OF TECHNOLOGY Department of Electrical Engineering and Computer Science 6.003: Signals and Systems — Spring 2007 Tutorial 9 Solutions April 16, 2007 Problem 9.1 (O&W Problem 8.36) (a) Let us assume that we are working with real signals, so each of the transforms depicted in this problem is symmetric about the ω = 0 axis. We are given Y (jω) and H1 (jω): Y (jω) 2ωM- C C C C C C 2ωM- 1 CC −ωc C C C C C C CC ω ωc H1 (jω) 2ωM- C 2ωT 2ωM- K C C C C C CC −ωc C 2ωT C C C C C CC ωc Let’s the call output of the coarse tunable filter s(t) so that S(jω) = Y (jω)H1 (jω): 1 ω S(jω) 2ωM- C C CC C C C 2ωM- K C C C C 2ωT C -C CC C C C CC C C C C C C C 2ωT C -C CC C −ωc ωc ω As we’ve seen in previous problems, multiplication by cos((ωc + ωf )t) shifts S(jω) by ±(ωc + ωf ) and scales it down by a factor of 2 to obtain Z(jω) (only ω > 0 shown here): Z(jω) 2ωM- K 2 C C CC C C C 2ωMC C C C C CC C C C C C C C 2ωT C -C CC C C 2ωT C -C CC C ωf 2ωc + ωf ω (b) We know that X1 (jω) is: X1 (jω) 2 C −ωM C C C C C CC ωM ω The height of the triangle is 2 because after modulating x1 (t) by cos ωf t, the spectrum scaled down by a factor of 2 to 1. So, R(jω) is: 2 R(jω) 2ωM- C C C C C C 1 CC 2ωM- −ωf C C C C C C CC ωf ω ωT must be such that the “non-Y1 ” portion of Y (jω) does not enter the non-zero region of the spectrum above. Looking at Z(jω), we see that the “tail” cannot interfere with the spectrum we want, which requires that: −ωf + ωT =⇒ ωT < < ωf − ωM 2ωf − ωM (c) We need the cutoffs α and β to be the same as the boundaries of the information we want and a compensating gain to bring the height of the triangle to 1, as required by the diagram of R(jω). So, α = ωf − ωM , = ωf + ωM , 2 G = . K β Problem 9.2 (O&W Problem 8.37) First, we rewrite z(t) using only the first four terms of the power series, 1 1 z(t) = 1 + y(t) + y(t)2 + y(t)3 2 6 1 1 = 1 + (x(t) + cos ωc t) + (x(t) + cos ωc t)2 + (x(t) + cos ωc t)3 2 6 5 5 1 1 1 9 2 3 = + x(t) + x(t) + x(t) + + x(t) + x(t)2 cos ωc t 4 4 2 6 8 2 1 1 cos 3ωc t + (1 + x(t)) cos 2ωc t + 4 24 (a) Using the expression for z(t) above we can sketch the spectrum of z(t). Note that the bandwidth at baseband is 6ω1 , at ωc is 4ω1 , and at 2ωc is 2ω1 . At 3ωc there is only a tone. 3 Z ( j ω ) ω 3 ω 1 ω c 2 ω 1 ω c ω c + 2 ω 1 2 ω c 3 ω c 0 2 ω c ω 1 2 ω c + ω 1 (b) In order to get an amplitude modulated version of x(t) we must filter out everything except for the components centered around 2ωc . This means that ωc + 2ω1 <α < 2ωc − ω1 2ωc + ω1 <β < 3ωc . Problem 9.3 (O&W Problem 8.48) (a) Since p[n] is periodic, we can use the DT FourierPSeries analysis equation to find the general form for ak 2πk and then use the relationship that P (ejω ) = 2π +∞ k=−∞ ak δ(ω − N ): ak = = 1 N X 2π p[n]e−jk N n n=<N > M 1 X −jk 2π n N e N n=0 2π ak = = = = 1 1 − e−jk N (M+1) 2π N 1 − e−jk N 1 e N M +1 −jk 2π N ( 2 ) (e since: al = l=0 M +1 jk 2π N ( 2 ) 2π 1 2π 1 e−jk N ( 2 ) (ejk N ( 2 ) " B X −e − M +1 −jk 2π N ( 2 ) 1 − aB+1 1−a ) 2π 1 e−jk N ( 2 ) ) sin( kπ N (M + 2j sin( kπ N ) 1 −jk 2π ( M ) 2j N 2 e N # " (M + 1)) sin( kπ π −j kM N N e N sin( kπ N ) 1)) # Note: You can also use Tables 5.1 and 5.2 on pp391 - 392 to find this result. Therefore: P (ejω ) = 2π +∞ X ak δ(ω − k=−∞ ak = ( e−j kM π N h 2πk ) N sin( kπ N (M+1)) N sin( kπ N ) M+1 N For N = 4, M = 1, P (ejω ) will look like: 4 where i , k 6= N i for i ∈ Z , k = N i for i ∈ Z P (ejω ) 6 π 6 j π4 −j π4 .71πe .71πe 6 6 π π .71πej 4 .71πe−j 4 6 6 - −π −3π 2 (b) Let ωM = π 2N . 0 −π 2 P (e ) = 2π = ( +∞ X ak = π N < 2π N kπ e−j N 2πk ) where N h i sin( 2πk N ) , k 6= N i for i ∈ Z kπ N sin( ) ak δ(ω − k=−∞ π 2N ω 3π 2 Since M = 1, then: jω Since 2 · π π 2 N 2 N , k = N i for i ∈ Z (no aliasing) and Y (ejω ) = Y (ejω ) = +∞ X 1 jω 2π X(e ) ak X(ej(ω− ∗ P (ejω ), then: 2πk N ) ) k=−∞ For N = 4, Y (ejω ) will look like: Y (ejω ) 6 2 1 π .35ej 4 π .35e−j 4 - −π −π 2 π 8 −π 8 π 2 π (c) For this problem, we are given: X(ejω ) = nonzero , |ω| < ωM 0 , ωM < |ω| ≤ π where X(ejω ) repeats every 2π. In order to make sure that there is no aliasing: 2ωM < ωs = Therefore, ωM < π N 2π N where N ∈ Z+ . Notice that this calculation does not depend on M. 5 ω (d) To get back the original signal, we need to low-pass filter y[n]: N M+1 y[n] - - −π −π N x[n] π π N Problem 9.4 Here we divide each system into sub components as shown in system 1 and determine the output at each sub system (see next page). 6 • System 1 1 a(t) b(t) . - J J 6 c(t) - - ω ? A s(t) A 6 A A A −ωm - J J 6 p(t) = cos(ωm t) q(t) ωm 1 - - ω d(t) −ωm ωm p(t) = cos(ωm t) From the diagram above, s(t) = a(t) + b(t) cos(ωm t) S(jw) is shown below Clearly, filtering this signal with H(jω) would not give us something proportional to a(t), thus c(t) is NOT proportional to a(t). S(jω) & % −2ωm 2ωm ω Modulating s(t) by a cosine gives us q(t). Q(jω) is shown below Q(jω) −3ωm Again, filtering this signal with H(jω) would not give us something proportional to b(t). Thus, d(t) is NOT proportional to b(t). % & 6 3ωm ω ωm 7 • System 2 Using the same technique in system 1, we divide the system into small parts again. This time from figure H and I, we can conclude that only d(t) = K2 b(t) where K2 = 2. a(t) J 6 G(jω) ? C s(t) 6 C C C r(t) = cos(2ωm t) b(t) - ? J G(jω) j f(t) j - ω -j - c’(t) J 6 1 ω −ωm ωm p(t) = sin(ωm t) ? - J ω 1 d’(t) -j ω −ωm ωm The real and imaginary parts of S(jω) are given below: Re{S(jω)} T T T T −2ωm T T T T Im{S(jω)} 2ωm ω ω 2ωm −2ωm Passing s(t) through G(jω) gives f(t): Re{F (jω)} Im{F (jω)} ω −2ωm 2ωm T T T −2ωm 8 c(t) T 2ωm T T T T ω d(t) A J a(t) D ? CC 6 C C C C 6 r(t) = cos(2ωm t) ? J B b(t) E F J 6 G(jω) G(jω) j ω -j j - ω -j H - 1 ωm −ωm ω p(t) = sin(ωm t) ? - J G I - d(t) 1 ωm −ωm c(t) ω Recall that P (jω) = −jπδ(ω − 2ωm ) + jπδ(ω + 2ωm ) C ′ (jω) is shown below: Im{C ′ (jω)} Re{C ′ (jω)} −4ωm @ @ @ A A A AA −4ωm 4ωm ω @ 2ωm @ −2ωm @ 4ωm −2ωm 2ωm ω Filtering C ′ (jω) with H(jω) would then give something proportional to a(t). Thus c(t) IS proportional to a(t). D ′ (jω) is shown below: Re{D′ (jω)} Im{D′ (jω)} −4ωm 4ωm −2ωm 2ωm ω A A A −4ωm AA 4ωm −2ωm Filtering D ′ (jω) with H(jω) would then give something proportional to b(t). Thus d(t) IS proportional to b(t) 9 2ωm AA A A A ω