Homework 6 SOLUTION EE235, Summer 2013 Due August 13th in class Reference material: • Riskin Interactive Notes Lectures 15-20. • Oppenheim and Willsky: – Filtering 3.9-3.10, – Fourier Transform Ch 4, – Frequency Response Ch 6 • Schaum’s: Fourier Transform, Filtering, Freq Response Ch 5 HW6 Topics: • Fourier Transform and LTI systems • Frequency Response • Filtering Problems: 1. The Fourier transform of a causal signal x(t) is 1 . 1 + jω X(jω) = (a) Find the total energy of this signal. x(t) = e−t u(t) Z ∞ energy of x(t) = |x(t)|2 dt −∞ Z ∞ = |e−t u(t)|2 dt −∞ Z ∞ = e−2t dt 0 1 = 2 (b) Find the time t0 (> 0) such that the signal x(t) on the interval [0, t0 ] contains 90% of its total energy. Z t0 1 0.9 = |x(t)|2 dt 2 −∞ Z t0 0.45 = |e−t u(t)|2 dt −∞ t0 Z = e−2t dt 0 1 1 = − e−2t0 + 2 2 0.1 = e−2t0 ln(0.1) = −2t0 t0 = − ln(0.1)/2 ≈ 1.15 1 (c) Find the frequency ω0 (> 0) such that the signal x(t) within the frequency band [−ω0 , ω0 ] contains 90% of its total energy. Hint: Use Parseval’s Theorem By Parseval’s theorem, Z Z ∞ 1 |X(w)|2 dw 2π −∞ −∞ 1 energy of x(t) = energy of X(w) 2π ∞ |x(t)|2 dt = so the energy of X(w) is π. Z w0 |X(w)|2 dw 0.9π = −w0 w0 Z 1 |2 dw 1 + jw 1 dw 1 + w2 | = −w0 w0 Z = −w0 0 = tan−1 (w)|w −w0 = tan−1 (w0 ) − tan−1 (−w0 ) since tan−1 (z) is an odd function 0.9π = 2 tan−1 (w0 ) 0.9π w0 = tan( ) ≈ 6.31 rad/s 2 x(t) 2 1 t −1 1 2 3 Figure 1: x(t) for problem 2 2. More Fourier transform properties. Let X(jω) denote the Fourier transform of the signal x(t) shown in figure 1. (a) Find X(0). Z ∞ x(t)e−jwt dt Z ∞ X(0) = X(w) w=0 = x(t)dt X(w) = −∞ −∞ So X(0) is the area under x(t) X(0) = 7 2 (b) Find R∞ −∞ X(jω)dω. 1 2π ∞ Z X(w)ejwt dw −∞ Z ∞ 1 X(w)dw x(0) = x(t)|t=0 = 2π −∞ x(t) = Z ∞ X(w)dw = 2πx(0) = 4π −∞ (c) Evaluate R∞ −∞ |X(jω)|2 dω. By Parseval’s theorem, Z ∞ 1 2π |x(t)|2 dt = −∞ so, this integral is Z ∞ Z 2 |X(w)| dw = 2π −∞ Z ∞ |X(w)|2 dw −∞ ∞ |x(t)|2 dt −∞ Z 0 = 2π[ Z 4dt + −1 1 0 Z 1 Z 2 (t − 4t + 4)dt + = 2π[4 + 0 3 2 Z (2 − t)2 dt + (t)2 dt + Z 3 4dt] 1 2 2 (t)2 dt + 4] 1 t3 t − 2t2 + 4t)|1t=0 + ( )|2t=1 + 4] 3 3 8 1 1 = 2π[4 + ( − 2 + 4) + ( − ) + 4] 3 3 3 2 38 76π = 2π[12 + ] = 2π[ ] = 3 3 3 = 2π[4 + ( Note: You should perform all these calculations without explicitly evaluating X(jω). 3. An LTI system has an impulse response given by h(t) = δ(t) − 5 sinc(5t). π H(ω) = 1 − rect(ω/10). (a) Classify the above filter as lowpass/highpass/bandpass/bandstop/allpass. (It may help to draw the Fourier transform.) Highpass H(ω) 1 ω −8 −6 −4 −2 3 2 4 6 8 (b) Find the output y(t) corresponding to the input x(t) = sin(3t). The Fourier transform of the input is X(s) = πj [δ(w −3)−δ(w +3)], which is two impulses at w = 3 and w = −3. Since the cutoff of the highpass filter is wc = 5, these impulses get filtered out, so y(t) = 0. Mathematically, Y (w) = X(w)H(w) π = [δ(w − 3) − δ(w + 3)][1 − rect(ω/10)] j π = [δ(w − 3) − δ(w + 3) − δ(w − 3) + δ(w + 3)] j F−1 = 0 −→ y(t) = 0 (c) Find the output y(t) corresponding to the input x(t) = cos(2t) + sin(8t). The Fourier transform of the input is X(ω) = πδ(ω − 2) + πδ(ω + 2) + jπδ(ω − 8) − jπδ(ω + 8) The deltas at ±2 are filtered out (they become zero), so the output is. y(t) = sin(8t). 4. Write these filters by using rect(ω/2B) functions. (a) Write an ideal bandpass filter G(ω) that passes only frequencies 3 < |ω| < 7. G(ω) = rect( ω ω ) − rect( ) 14 6 (b) Write an ideal bandreject filter H(ω) that rejects only frequencies 3 < |ω| < 7, and passes all other frequencies. H(ω) = 1 − G(ω) = 1 − rect( ω ω ) + rect( ) 14 6 5. A causal LTI system is described by the following differential equation: dy(t) + 5y(t) = 2x(t). dt (a) Find the frequency response H(ω) of this system jωY (ω) + 5Y (ω) = 2X(ω) (jω + 5)Y (ω) = 2X(ω) H(ω) = 4 2 Y (ω) = X(ω) 5 + jω (b) Find the magnitude of the frequency response, |H(ω)|. 2 5 + jω H(ω) = 2 =√ ω 2 etan−1 (−ω/5) 25 + −1 2 =√ e− tan (−ω/5) 2 25 + ω | {z } |H(ω)| p H(ω)H ∗ (ω) p |H(ω)| = H(ω)H ∗ (ω) r 2 2 = · 5 + jω 5 − jω 2 =√ 25 + ω 2 Or by using the formula |H(ω)| = (c) Sketch the magnitude of the frequency response (for both positive and negative ω). |H(ω)| 2 1 6 5 4 3 2 1 0 1 2 3 4 5 6ω (d) Classify this system as lowpass/highpass/bandpass/bandstop/allpass. Lowpass (e) Find the impulse response h(t) of this system. h(t) = 2e−5t u(t) (f) We build another LTI system by making its impulse response h1 (t) = h(t) cos(50t). Classify this system as lowpass/highpass/bandpass/bandstop/allpass. Justify your answer by giving the approximate frequency range which the filter passes. (It may be useful to try to sketch |H1 (ω)|.) H1 (w) = 1 [H(w − 50) + H(w + 50)] 2 |H1 (ω)| 2 1 50 ω 50 Bandpass; It passes frequencies around 50 rad/s (approximately 48 rad/s to 52 rad/s). 6. Consider an LTI system with impulse response h(t) = e−4|t| . 5 (a) Show the frequency response of this LTI system. Z 0 Z 4t −jωt H(ω) = e e dt + −∞ Z 0 e (4−jω)t e(−4−jω)t dt = dt + −∞ e−4t e−jωt dt 0 ∞ Z ∞ 0 1 1 8 − = 2 4 − jω 4 + jω ω + 16 (b) We have an input x1 (t) = +∞ X δ(t − n). n=−∞ Is the output a periodic signal? Justify your answer. If “yes,” find its Fouries series coefficients. This is a periodic pulse train with period T=1. The following Fourier Series pair can be applied: ∞ 2kπ 2π X ) δ(ω − δ(t − nT ) = T T n=−∞ n=−∞ ∞ X Therfore the output is: ∞ 8 × 2π X δ(ω − 2kπ) ω 2 + 16 n=−∞ As a result, the Fourier Series Coefficients are (for integer values of k): 16π δ(ω − 2k) + 16 ω2 (c) Repeat the same questions for another input x2 (t) = +∞ X (−1)n δ(t − n). n=−∞ The input is a series of alternating positive and negative deltas that can be rewritten as: x2 (t) = +∞ X δ(t − 2n) − n=−∞ X(ω) = +∞ X δ(t − (2n + 1)) n=−∞ ∞ ∞ 2kπ 2π X 2kπ 2π X δ(ω − ) − e−jω δ(ω − ) 2 n=−∞ 2 2 n=−∞ 2 Therefore the fourier series coefficients are: 2πk 8π δ ω − (1 − e−jω ) ω 2 + 16 2 7. Let sin t . πt Assuming that x(t) is band-limited and X(jω) = 0 for |ω| ≥ 1, does there exist an LTI system T {·} such that T {x(t)} = y(t)? Justify your answer. If “yes,” give the impulse response of this LTI system. From trigonometric identities we have: 1 cos(2t) sin(t) y(t) = x(t) × + ∗ 2 2 πt y(t) = (x(t) cos2 t) ∗ 6 Next we take the Fourier transform: π π ω 1 δ(ω) + δ(ω − 2) + δ(ω + 2) × rect( ) Y (jω) = X(jω) ∗ 2 2 2 2 1 ω × {X(jω) + πX(j(ω − 2)) + πX(j(ω + 2))} × rect( ) 2 2 ω Since the signal is band-limited and rect( 2 ) function is zero for |ω| > 1 the X(j(ω + 2)) and X(j(ω − 2)) terms disappear and we are left with: Y (jω) = 1 X(jω) 2 Y (jω) = Y (jω) = X(jω)H(jω) H(jω) = 1 2 Taking the inverse Fourier transform yields: h(t) = 1 δ(t) 2 8. From previous lectures we know that the energy of a complex signal x(t) is, Z ∞ Wx = |x(t)|2 dt −∞ Noting that |x(t)|2 = x(t)x∗ (t) determine the relationship between finding the energy of a signal in the time domain and finding the energy of the signal in the frequency domain. This is an important relation called Parsivals Theorem. (Hint: determine what x∗ (t) is equal to in the frequency domain as your first step) Given X(ω) we can write x(t) and x∗ (t) as: Z ∞ 1 x(t) = X(ω)ejωt dω 2π −∞ 1 x (t) = 2π ∗ Z ∞ X ∗ (ω)e−jωt dω −∞ Next, we can substitute into the following equation replacing one of the ω’s with α: Z ∞ Z ∞ |x(t)|2 dt = x(t)x∗ (t)dt −∞ Z ∞ −∞ x(t) −∞ 1 2π Z ∞ X ∗ (jω)e−jωt dω dt −∞ Reordering: 1 2π Z ∞ −∞ X ∗ (jω) Z ∞ 1 x(t)e−jωt dt dω = X ∗ (jω)X(jω)dω 2π −∞ −∞ Z ∞ 9. Use Parsivals Theorem to solve the following problem, Z ∞ 2 χ= dω |jω + 2|2 −∞ 7 e−2t u(t) ↔ 1 jω + 2 Using Parseval’s Theorem: Z ∞ −∞ 2 dω = |jω + 2|2 Z 4π √ 2 Z ∞ 2 |e−2t u(t)|2 dt dω = 4π jω + 2 −∞ −∞ Z ∞ ∞ e−4t dt = − 0 4π (0 − 1) = π 4 BONUS LEARNING: If an image has some feature with a strong spatial frequency and you undersample the image spatially, then you can get an aliasing effect called moiré. Check out some examples and more discussion of moiré at http://en.wikipedia.org/wiki/Moire. 8