Indian Institute of Science

E9−207: Basics of Signal Processing

Instructor: Shayan G. Srinivasa
Homework #2 Solutions, Spring 2018
Solutions prepared by Chaitanya and Ankur
Late submission policy: Points scored = Correct points scored × e−d , d = # days late
Assigned date: Feb. 20th 2018 Due date: Mar. 1st 2018 by end of the day

PROBLEM 1:
(a)Let S1 and S2 be two vector spaces. Then show that S1 ∩ S2 is also a vector space.
(b)If A ∈ Cn×n and u, v ∈ Cn are non-zero vectors such that Au = 2u and Av = 3v, show that
u, v are linearly independent.
Solution:
(a) Consider two vectors x, y such that x, y ∈ S1 and x, y ∈ S2 . This implies x, y ∈ S1 ∩ S2 .
Similarly αx + βy ∈ S1 and αx + βy ∈ S2 for scalars α, β. Therefore, αx + βy ∈ S1 ∩ S2 . The
zero vector, 0 ∈ S1 and 0 ∈ S2 . Therefore, 0 ∈ S1 ∩ S2 . Hence, S1 ∩ S2 is a vector space.
(b)Suppose u, v are linearly dependent. Then we can write u = kv for some scalar k 6= 0. Then,
Au = A(kv) = kAv = 3kv. But Au = 2u. This means 2u = 3kv or u = 3k 2 v.
3k
Now, u = kv and u = 2 v both hold if only if k = 0. This is a contradiction as k 6= 0 as per
assumption. Therefore, u and v are linearly independent.

PROBLEM 2:
(a)Let A ∈ Cm×m be a matrix acting on vectors in the vector space Cm . We define a new
product between vectors x, y ∈ Cm as (x, y)A as x† Ay. Under what conditions is this a valid
inner product?
(b)Consider the matrix
 
1 a a
A = a 1 a
a a 1
√
For what values of a ∈ C is x†Ax a norm defined on C3 ?
Note: a† is the transpose conjugate of a. For example
 
1

v=  i  ⇒ v † = 1 −i i
−i
Solution:
(a) To be a valid inner product we need hx, xiA ≥ 0 i.e., x† Ax ≥ 0. This should hold for all x.
This holds if and only if A is positive definite matrix.
(b) A is positive semi-definite implies A† = A ⇒ a ∈ R. Secondly all its eigen values must be
positive. Therefore λ = 1 − a, 1 − a, 1 + 2a must be positive. Therefore a ∈ (− 21 , 1).

PROBLEM 3:
What is the minimum value of x − y − z subject to the constraint x2 + y 2 + z 2 = 1?
Solution:
For twopvectors u, v we havep |hu, vi| ≤ ||u|| ||v||. Fix u = (x, y, z)T , v = (1, −1, −1)T . Then
||u|| = x2 + y 2 + z 2 , ||v||p= 12 + (−1)2p + (−1)2 .
√
∴ |hu,√vi| = |x − y − z| ≤ 2 2 2 2 2 2
√ x + y + z · 1 + (−1) + (−1) = 1 · 3. √
⇒ − 3 ≤ x − y − z ≤ 3. This implies that the minimum value of x − y − z is − 3.
PROBLEM 4:
Consider the functions ϕk (t) = A sinc(π(t − k)) where k is an integer and A ∈ C.
For integers k, l evaluate
Z
ϕk (t)ϕ∗l (t)dt
R

Conclude that ϕ(t) ∈ L2 (R) and that {ϕk : k ∈ Z} forms an orthonormal set of functions in
L2 (R).
Solution:
Denote

ψk,l (t) = ϕk (t)ϕ∗l (t)dt

Z
ψ̂k,l (ω) = ψk,l (t)e−jωt dt
R

We need to evaluate
Z Z π
1
ψ̂k,l (ω)|ω=0 = ϕk (t)ϕ∗l (t)dt = ϕ̂k (ω)ϕ̂∗l (−ω)dω|ω=0
R 2π −π

We know
 
ω
φk (t) = A sinc(π(t − k)) ←→ A rect e−jωk
2π
 
ω
φ∗l (t) = A∗ sinc(π(t − l)) ←→ A∗ rect ejωl
2π

Now,
Z π
1
ϕ̂k (ω)ϕ̂∗l (−ω)dω|ω=0
2π −π
Z π
1
= |A|2 e−jω(k−l) dω|ω=0
2π −π
(
sin π(k − l) 0 for l 6= k
= |A|2 =
π(k − l) |A|2 for l = k

Therefore,
Z Z
ϕk (t)ϕ∗k (t)dt = |ϕk (t)|2 dt = |A|2 < ∞.
R R

|A|2 = 1 then {ϕk : k ∈ Z} form an orthonormal set of functions.

PROBLEM 5:
(a)A baseband signal s(t) with 50 Hz bandwidth is sampled at a rate F s. The resultant signal
is downsampled by a factor 2 to obtain the discrete samples ŝ(n). What is the minimum value
of Fs in Hz to reconstruct back the signal s(t) from the samples ŝ(n)?
(b) Let s(n) be any discrete time signal with energy Es . The signal is downsampled by 3. What
is the evergy of the resultant signal if there is no aliasing after decimation?
Solution:

(a) F2s ≥ 100Hz. Therefore, minimum value of Fs = 200Hz

(b) Let X (ω) be the frequency response of the original signal. The frequency response after
 downsampling is
1  1
  1
  1

X̂ (z) = X 1 · z 3 + X Ω · z 3 + X Ω2 · z 3 where Ω3 = 1,
3    
1 ω  ω − 2π ω − 4π
=⇒ X̂ (ω) = X +X +X .
3 3 3 3
ω

ω−2π

ω−4π


Since
there is no

aliasing,

the responses X 3 , X 3 and X 3 do not overlap i.e.,
ω ω−2π ω−4π
X 3 X 3 X 2 = 0 ∀ω.
Therefore,
Z2π  2 Z6π Z6π    Z6π   2 Z6π   2
  1  2
 1  ω 2 1 
X ω − 2π 
 dω + 1  
X ω − 4π  dω
X̂ (ω) dω = X̂ (ω) dω = X  dω +    
3 27 3 27 3 27 3
0 0 0 0 0
Z2π Zπ Zπ
1 1 1
= |X (ω1 )|2 dω1 + |X (ω2 )|2 dω2 + + |X (ω3 )|2 dω2
9 9 9
0 −π −π
1 1 1 1
= Es + Es + Es = Es .
9 9 9 3
ω ω−2π ω−4π
We have substituted ω1 = 3, ω2 = 3 and ω3 = 3 in the above equations.
PROBLEM 6:
(a) A signal x(t) is obtained by convolving signals x1 (t) and x2 (t) with the following character-
istics:
|X1 (ω)| = 0 for |ω| > 500π,
|X2 (ω)| = 0 for |ω| > 250π.
∞
X
Impulse train sampling is performed on x(t) to get xs (t) = x(nT )δ(t − nT ). Specify the
−∞
range of values of T so that x(t) may be recovered from xs (t). (4 pts)

(
1 − |t| for − 1 ≤ t ≤ 1
(b) The signal s(t) = is passed through a system to obtain the out-
0 otherwise.
put ŝ(t). The system has a resonant frequency of 23 Hz and hence allows only frequencies of
2
R2 2
3 Hz and its harmonics along with d.c. component. What is the value of −2 |ŝ(t)| ? (8 pts)
Solution:
(a) Convolution in time domain is equivalent to multiplication it the frequency domain.
X(ω) = 0 for |ω| > 250π ⇒ X(f ) = 0 for |f | > 125 Hz. Sampling frequency should be
1
fs ≥ 2 · 125 Hz. That is, the sampling period should be T < 250 s = 4ms.

(b) This is same as sampling in the frequency domain. Here, the frequency response S (f )
P
∞

is multiplied by δ f − k 32 . Therefore, it results in time-domain signal convolved with its
k=−∞
P
∞ 2 P
∞

Fourier inverse e−j2πt 3 k = 3
2 δ t − k 32 . Therefore the output signal is
k=−∞ k=−∞
∞   ∞  
3 X 3 3 X 3
ŝ (t) = s (t) ∗ δ t−k = s t−k .
2 2 2 2
k=−∞ k=−∞

The input and output signals are shown in the following figure:
 s(t)

-1 0 1 t
s(t − 32 k)
k = −1
k=1
k = −2 k=2
k=0
1

-4 - 72 -3 - 52 -2 - 32 -1 - 12 0 1
2 1 3
2 2 5
2 3 7
2 4 t

3
P
∞

ŝ(t) = 2 s t − k 32
k=−∞

3
2

3
4

-4 - 72 -3 - 52 -2 - 32 -1 - 12 0 1
2 1 3
2 2 5
2 3 7
2 4 t

From the figure, the energy of ŝ (t) in the interval [−2, 2] is

Z2 Z0.5 Z1
2 2
|ŝ (t)| dt = 3 |ŝ (t)| dt +2 |ŝ (t)|2 dt
−2 −0.5 0.5
| {z } | {z }
3 triangle portions 2 flat portions
Z0.5 Z1
= 6 |ŝ (t)| dt + 2 |ŝ (t)|2 dt
2

0 0.5
Z0.5 2 Z1  2
3  3

= 6  (1 − t) dt + 2   dt

2 4
0 0.5
" #0.5  
27 (1 − t)3 9 27 1 1 9 63 9 9
= − + = × − + + = + = .
2 3 16 2 8×3 3 16 16 16 2
0

R2
Therefore, |ŝ (t)|2 dt = 18
4 = 4.5.
−2