EECS 20N: Structure and Interpretation of Signals and Systems Problem Set 5

Department of Electrical Engineering and Computer Sciences SOLUTIONS

U NIVERSITY OF C ALIFORNIA B ERKELEY

HW 5.1 Solution

(a)

∞
X(ω) = x(n)e−iωn
n=−∞

d 
∞
d
i X(ω) = i x(n)e−iωn
dω dω n=−∞

∞
d −iωn
= i x(n) e
n=−∞
dω
∞
= i x(n)(−in)e−iωn
n=−∞
∞
= (nx(n))e−iωn
n=−∞
(n) = nx(n)
x

(b) Upsampling


∞
X(ω) = x(n)e−iωn
n=−∞
∞

X(ω) = (n)e−iωn
x
n=−∞
(n) is non-zero only at n = kN, k = . . . − 2, −1, 0, 1, 2, . . .
but x

∞

X(ω) = (kN)e−iωkN
x
k=−∞
∞
= x(k)e−i(ωN )k
k=−∞
= X(ωN)

Echo system
F
Remember that δ(n − n0 ) ↔ e−iωn0 , and that for LTI system Y (ω) =
H(ω)X(ω)
From the difference equation y(n) = x
(n) + α
y (n − N)

1
 Take Fourier transform of both sides

Y (ω) = X(ω)
 + αe−iωN Y (ω)
(1 − αe−iωN )Y (ω) = X(ω)


 Y (ω)
H(ω) =

X(ω)
1
=
1 − αe−iωN
For single sample echo system y(n) = x(n) + αy(n − 1)
We have frequency response H(ω) = 1−αe1−iωN by letting N = 1 in previ-
ous part.

So, H(ω) = H(ωN)
Find impulse response: h(n) is, by definition, the output when the input
is an impulse, so let x(n) = δ(n) and find y(n)

y(n) = δ(n) + αy(n − 1)

y(n) = 0, n < 0
y(0) = 1
y(1) = α
..
.
y(n) = αn , n ≥ 0
∴ h(n) = αn u(n) where u(n) is the unit step function.

By upsampling property,
 n
 αN if n ≥ 0 and n mod N = 0
h(n) =
0 otherwise

HW 5.2 Solution
(a)
X(ω) + X ∗ (ω) 1 − cos ω
Re{X(ω)} = =
2 2
iω −iω
1 e +e
= −
2 4
Sincexis a real-valued signal, by the properties of DTFT for real-valued signals,

2
 X(ω) = X ∗ (−ω) ⇒ X ∗ (ω) = X(−ω)

X(ω) + X(−ω) 1 eiω + e−iω

⇒ = −
2 2 4
f˜
Using the time-reversal property of DTFT (x(−n) ←→ X(−ω)) and tak-
ing inverse DTFT of both sides, we get
x(n) + x(−n) δ(n) δ(n − 1) + δ(n + 1)
= −
2 2 4
f˜ f˜
[Using δ(n) ←→ 1, x(n − n0 ) ←→ X(ω)e−iωn0 ]

δ(n − 1) + δ(n + 1)
∴ x(n) + x(−n) = δ(n) −
2
x being causal, x(−n) = 0, ∀n > 0
δ(n − 1) + δ(n + 1)
⇒ ∀n > 0, x(n) + 0 = δ(n) −
2
−δ(n − 1)
=
2
δ(−1) + δ(1)
n = 0 ⇒ 2x(0) = δ(0) −
2
= 1
= δ(0)
⇒ x(0) = 1/2
⎧
⎨ 1/2 n=0
∴ x(n) = −1/2 n=1
⎩
0 otherwise
δ(n) − δ(n − 1)
equivalently, x(n) =
2
Since we made no extra assumptions on x apart from what is given, the
solution above is the unique solution.
(b)
5 1 1
Note that − cos ω = ( − eiω )( − e−iω )
4 2 2
and
|X(ω)|2 = X(ω)X ∗(ω)
1 1
Thus, X(ω)could be − eiω or − e−iω .
2 2
3
 It was given in the problem that the signal was causal. From this as-
sumption, X(ω) cannot be 12 − eiω .

1
X(ω) = − e−iω
2

and taking inverse DT F T , we get

δ(n)
x(n) = − δ(n − 1)
2
which is clearly causal and real-valued. Letting
δ(n)
x(n) = δ(n − 1) −
2
which is also real-valued and causal gives us the same solution for |X(ω)2|.
So, we have ≥ 2 solutions
δ(n)
2
− δ(n − 1), δ(n − 1) − δ(n)
2
⇒Solutions not unique.
In this problem, only the magnitude of the frequency response was con-
strained leaving the phase ambiguous. So shifting the signal by any
number of samples in which the phase changes but the magnitude re-
mains the same would also produce a valid answer making the solution
not unique.
∗
(c) Using Im{X(ω)} = X(ω)−X2i
(ω)
,and that
∗
X (−ω) = X(ω) for real-valued signals, we get
X(ω) − X(−ω) ei2ω − e−i2ω
= − sin 2ω = −
2i 2i
⇒ X(ω) − X(−ω) = −e2iω + e−2iω
Takeing inverse DT F T , we get
x(n) − x(−n) = −δ(n + 2) + δ(n − 2)
For n > 0, ⇒ x(n) = −δ(n+2)+δ(n−2) = δ(n−2) [x(−n) = 0 since x is causal]
Using definition of DT F T ,
⇒ X(ω) = x(0)ei0ω + e−2iω = x(0) + e−2iω
∴ X(ω)|ω=0 = 1 ⇒ x(0) = 0
∴ x(n) = δ(n − 2) which is unique (no extra assumptions made)

4
HW 5.3 Solution
(a) (I)
⇒ h(n) = 0 when n<0
(II) (i)
A(ω) = H(ω)ei2ω ⇒ a(n) = h(n + 2) or h(n) = a(n − 2)
(ii) 
1
a(n) = A(ω)eiωn dω
2π 2π

1 1
a(0) = A(ω)dω = (12π) = 6
2π 2π 2π
(iii)
⇒ a(n) is symmetric. so a(n) and h(n)are both length 5 .
(iv)

2
A(ω) = a(n)e−iωn
n=−2


2
A(0) = a(n)
n=−2
= a(−2) + a(−1) + a(0) + a(1) + a(2)
= 2a(2) + 2a(1) + 6
= 8 (from the figure)
A(π) = A(−π)
 2
= a(n)e−iπn
n=−2

2
= a(n)eiπn
n=−2

2
= a(n)(−1)n
n=−2
= a(−2) − a(−1) + a(0) − a(1) + a(2)
= 2a(2) − 2a(1) + 6
= 12

5
 We have 2 equations for 2 unknowns...solving them gives
a(2) = a(−2) = 2, a(1) = a(−1) = −1

6 6
a(n) h(n)

2 2 2 2

n n
-2 -1 0 1 2 0 1 2 3 4
-1 -1 -1 -1

(b)

y(n) = (x ∗ h)(n)
= 2x(n) − x(n − 1) + 6x(n − 2) − x(n − 3) + 2x(n − 4)

from the sifting property of the delta funtion

x(n) 2 + y(n)

z −1
q1(n) −1 +

z −1
q2(n) 6 +

z −1
q3(n) −1 +

z −1
q4(n) 2

(c)

6
(d)


4
H(ω) = h(n)e−iωn
n=0
= 2 − e−iω + 6e−2iω − e−3iω + 2e−4iω
= e−2iω (2e2iω − eiω + 6 − e−iω + 2e−2iω )
= ei(−2ω) (6 − 2 cos(ω) + 4 cos(2ω))
∠H(ω) = −2ω
|H(ω)| = (6 − 2 cos(ω) + 4 cos(2ω)) = |A(ω)|
So the given figure can be used for (f)

(e) (i)
y(n) = 8, because |H(0)| = 8, ∠H(0) = 0
(ii)

x(n) = eiπn
y(n) = ei(−2π) (6 − 2 cos(π) + 4 cos(2π))eiπn
= 12eiπn
= 12(−1)n

(iii)
π
π

π π

y(n) = 6 − 2 cos + 4 cos cos n−

4
2 4 2
√ π
= (6 − 2) sin n
4