EE1101 Signals and Systems JAN—MAY 2019

Tutorial 4 Solutions

1) For causal Discrete time LTI systems, h[n]=0 for n<0

P∞
For stable Discrete time LTI systems, |h[n]| < ∞.
n=−∞

For Causal Continuous time LTI systems, h(t)=0 for t<0

R∞
For Stable Continuous time LTI systems, τ =−∞ |h(τ )|dτ < ∞.

(a) h[n] = (− 21 )n u[n] + (1.01)n u[1 − n]

Since u[1-n]6= 0 for n< 0 , substituting n < 0 gives h[n] 6= 0
Therefore ,the system is NON CAUSAL
∞ ∞
|(− 12 )n u[n] + (1.01)n u[1 − n]|
P P
|h[n]| =
n=−∞ n=−∞
∞ ∞
|(− 21 )n u[n]| |(− 12 )n | < ∞
P P
=
n=−∞ n=0
∞ 1
|(1.01)n u[1 − n]| = |(1.01)n | < ∞
P P
n=−∞ n=−∞
∞
P
Therefore, |h[n]| < ∞. So the system is STABLE
n=−∞

(b) h[n] = n( 13 )n u[n − 1]

Since u[n-1]= 0 for n< 1, h[n] = 0 for n<0
Therefore ,the system is CAUSAL
∞
Intuitively, n( 13 )n u[n − 1] has a converging sum.Therefore,
P
|h[n]| < ∞.
n=−∞
To prove mathematically,

∞ ∞
X X 1
|h[n]| = |n( )n u[n − 1]|
n−∞ n=−∞
3
∞
X 1
= |n( )n u[n − 1]|
n=1
3
1 2 3
= + + 3 + ....(sum of inf inite AGP )
3 32 3
= 0.75 < ∞

Therefore,the system is STABLE

(c) h(t) = e2t u(−1 − t)

Since u(-1-t)6= 0 for t< 0, h(t) )6= 0 for t <0;
Therefore,system is NON CAUSAL

1
 Z ∞ Z ∞
|h(τ )|dτ = |e2τ u(−1 − τ )|dτ
τ =−∞ τ =−∞
Z −1
= |e2τ |dτ
τ =−∞
−2
e
= <∞
2

Therefore,the system is STABLE

(d) h(t) = e−6|t|

Since h(t) has values at all instants of time, h(t) )6= 0 for t <0;
Therefore,system is NON CAUSAL

Z ∞ Z ∞
|h(τ )|dτ = |e−6|τ | |dτ
τ =−∞ τ =−∞
Z 0 Z ∞
= |e6τ |dτ + |e−6τ |dτ
τ =−∞ τ =0
1 1 1
=| |+| |= <∞
6 6 3

Therefore,the system is STABLE

(e) h(t) = (2e−t − e(t−100)/100 )u(t)

Since h(t) is some function multiplied by u(t), h(t) )= 0 for t <0;
Therefore,system is CAUSAL

Z ∞ Z ∞
|h(τ )|dτ = |(2e−τ − e(τ −100)/100 )u(τ )|dτ
τ =−∞ τ =−∞
Z ∞
= |(2e−τ − e(τ −100)/100 )|dτ
τ =0

In this, We have an exponentially decaying part ((2e−τ ),and

Exponentially increasing part (e(τ −100)/100 )
The exponentially increasing part makes the integral sum to infinte. Therefore ,system is UNSTABLE

2)
(a) (a) h[n] = (−1/2)n u[n]
for n< 0

s[n] = 0

for n ≥ 0

2
 Pn −1 k
s[n] = k=0 ( 2 )

Using G.P. formula for finite series

s[n]= 13 (2 + ( −1 n
2 ) )u[n]

(b) h[n] = nu[n]

for n < 0

s[n] = 0

for n≥ 0
Pn
s[n] = k=0 k

n(n+1)
s[n]= 2 u[n]

(c) h(t) = e−|t|

for t < 0
R0
s(t)= −∞
eτ dτ = et

for t≥ 0
R0 Rt
s(t) = −∞
eτ dτ + 0
e−τ dτ = 2- e−τ
(
et t<0
s(t)=
2 − e−t t≥0

(d) h(t)=(1/4)(u(t)-u(t-4))
for t < 0

s(t) = 0

for t < 4
1 t
dτ = 14 t
R
s(t) = 4 0

for t≥ 4
R4
s(t) = 14 0
dτ = 1

0
 t<0
1
s(t) = t 0≤t<4
4
1 t≥4


(e) h(t) = u(t)

for t < 0

s(t) = 0

for t≥ 0
Rt
s(t)= 0
dτ = t

3
 s(t) = tu(t)

3)
(a) True. If h(t) periodic and nonzero, then
Z ∞ +∞ Z
X nT
|h(t)|dt = |h(t)|dt.
−∞ n=−∞ (n−1)T

Since each summand is same , the infinite sum is unbounded. Thus h(t) is unstable.
(b) False. For instance, suppose that the inverse of h[n] = δ[n − n0 ] is g[n]. Then,

⇒ h[n] ∗ g[n] = δ[n]

X∞
⇒ δ[k − n0 ]g[n − k] = g[n − n0 ] = δ[n]
k=−∞

⇒ g[n] = δ[n + n0 ]
which is noncausal.
(c) False. For example h[n] = u[n] implies that
∞
X
|h[n]| = ∞.
−∞

This is an unstable system.

(d) True. Assuming that h[n] is bounded in the range n1 ≤ n ≤ n2 ,
n2
X
|h[k]| < ∞.
k=n1

This implies that the system is stable.

(e) False. For example, h(t) = tu(t) is causal but not stable.
(f) False. For example, the cascade of a causal system with impulse response h1 [n] = δ[n − 1] and a non-causal
system with impulse response h2 [n] = δ[n + 1] leads to a system with overall impluse response given by
h[n] = h1 [n] ∗ h2 [n] = δ[n].
R∞
(g) False. For example, if h(t) = e−t u(t), then s(t) = e−t u(t) ∗ u(t) = τ =−∞ e−τ u(τ )u(t − τ )dτ
Rt
= τ =0 e−τ u(τ )dτ = (1 − e−t )u(t) and
Z ∞
|1 − e−t |dt = t + e−t |∞
0 = ∞.
0

Although the system is stable, the step response is not absolutely integrable.
(h) True. We may write u[n] = Σ∞
k=0 δ[n − k]. Therefore,

s[n] = Σ∞ ∞
k=−∞ u[k]h[n − k] = Σk=0 h[n − k].

If s[n] = 0 for n < 0, then h[n] = 0 for n < 0 and the system is causal.
4) Given: System A is LTI and system B is inverse of A. Let y1 (t) and y2 (t) be outputs of system A for inputs
x1 (t) and x2 (t) respectively. Combining these informations, we get,
A B
x1 (t) −
→ y1 (t) −
→ x1 (t). (1)
And,
A B
x2 (t) −
→ y2 (t) −
→ x2 (t). (2)

4
 (a) To prove system B is linear. Assume that system B is not linear. From equations 1 and 2, we observe
that an input ax1 (t) + bx2 (t) to system A can generate ay1 (t) + by2 (t), i.e.,
A
ax1 (t) + bx2 (t) −
→ ay1 (t) + by2 (t).

This is due to linearity property of the system A. Our assumption that system B is not linear implies that
the output of B for the input ay1 (t) + by2 (t) is not ax1 (t) + bx2 (t), i.e. the outputs of system B does not
add up linearly even if the inputs combine linearly. Therefore, we arrive at a situation which is as follows:
A B
ax1 (t) + bx2 (t) −
→ ay1 (t) + by2 (t) −
6 → ax1 (t) + bx2 (t).

This contradicts the fact that B is inverse of A. Hence, our assumption is incorrect, and so, system B is
linear.
(b) To prove system B is time-invariant. Assume system B is time variant. By time-invariant property
of system A, we have,
A
x1 (t − τ ) −
→ y1 (t − τ ).

When this output of system A, is fed to system B, we must not expect its response to be x1 (t − τ ) because
of our assumption. So, we land up in a situation where,
A B
x1 (t − τ ) −
→ y1 (t − τ ) −
6 → x1 (t − τ ).

This contradicts the fact that B is the inverse of A. So, the assumption about system B is incorrect.
Therefore, system B is also time-invariant.

5) Pn
k
 k=0 (k + 1)α
 n≥0
s[n] = h[n] ∗ u[n] =

0 otherwise


Given that
n n+1
d 1 − αn+2
 
X
k d X k
(k + 1)α = α =
dα dα 1−α
k=0 k=0

1 − (n + 2)αn+1 1 − αn+2
 
s[n] = + u[n]
1−α 1−α
α2
 
1 n α n
s[n] = − α + (n + 2)α u[n]
(α − 1)2 (α − 1)2 (α − 1)
α2
 
1 n α(n + 1) n α n
= − α + α + α u[n]
(α − 1)2 (α − 1)2 (α − 1) (α − 1)
α2
 
1 n α(n + 1) n α(α − 1) n
= − α + α + α u[n]
(α − 1)2 (α − 1)2 (α − 1) (α − 1)2
 
1 α n α n
= − α + (n + 1)α u[n]
(α − 1)2 (α − 1)2 (α − 1)

6) (a) The autocorrelation function for x1(t):

case 1:t > 2 φx1 x1 (t) = 0

5
 3

2
x1(t + τ ) x1(τ )

−t 2−t 0 1 2 3 4 τ
R 2−t
case 2:0 ≤ t ≤ 2 φx1 x1 (t) = 0
(τ /2) ∗ ((t + τ )/2)dτ = (t3 /24) − (t/2) + (2/3)

2
x1(t + τ )x1(τ )
1

−t 0 2−t 2 3 4 τ
R2
case 3:−2 ≤ t ≤ 0 φx1 x1 (t) = −t
(τ /2) ∗ ((t + τ )/2)dτ = (−t3 /24) + (t/2) + (2/3)

2
x1(τ ) x1(t + τ )
1

0 −t 2 2−t 4 τ

case 4:t ≤ −2 φx1 x1 (t) = 0

2
x1(τ ) x1(t + τ )
1

0 1 2 −t 2−t τ

Consolidating all the cases,

(
(|t|3 /24) − (|t|/2) + (2/3), 0 ≤ |t| ≤ 2
φx1 x1 (t) =
0, |t| > 2

and φx1 x1 (t) = φx1 x1 (−t)

6
The autocorrelation function for x2(t):
case 1:t > 7 φx2 x2 (t) = 0

2
x2 (t + τ ) x2 (τ )
1

−t 1−t 2−t 3−t 4−t 5−t 6−t 7−t 0 1 2 3 4 5 6 7

−1
R 7−t
case 2:6 ≤ t ≤ 7 φx2 x2 (t) = 0
−1dτ = t − 7

2
x2 (t + τ ) x2 (τ )
1

−t 1−t 2−t 3−t 4−t 5−t 6−t 7−t 2 3 4 5 6 7

−1
R 6−t R 7−t
case 3:5 ≤ t ≤ 6 φx2 x2 (t) = 0
1dτ + 6−t
−1dτ = 5 − t

2
x2 (t + τ ) x2 (τ )
1

−t 1−t 2−t 3−t 4−t 5−t 6−t 7−t 3 4 5 6 7

−1

case 4:4 ≤ t ≤ 5 φx2 x2 (t) = t − 5

7
 3

2
x2 (t + τ ) x2 (τ )
1

−t 1−t 2−t 3−t 4−t 5−t 6−t 7−t 4 5 6 7

−1

By shifting the signal further we get the following cases

case 5:3 ≤ t ≤ 4 φx2 x2 (t) = 3 − t
case 6:2 ≤ t ≤ 3 φx2 x2 (t) = t − 3
case 7:1 ≤ t ≤ 2 φx2 x2 (t) = 1 − t
case 8:0 ≤ t ≤ 1 φx2 x2 (t) = 7(1 − t)
case 9:t < −7 φx2 x2 (t) = 0
case 10:−7 ≤ t ≤ −6 φx2 x2 (t) = −t − 7
case 11:−6 ≤ t ≤ −5 φx2 x2 (t) = 5 + t
case 12:−5 ≤ t ≤ −4 φx2 x2 (t) = −t − 5
case 13:−4 ≤ t ≤ −3 φx2 x2 (t) = 3 + t
case 14:−3 ≤ t ≤ −2 φx2 x2 (t) = −t − 3
case 15:−2 ≤ t ≤ −11 φx2 x2 (t) = 1 + t
case 16:−1 ≤ t ≤ 0 φx2 x2 (t) = 7(1 + t)
Consolidating all the cases, 

 7(1 − |t|), 0 ≤ |t| ≤ 1

1 − |t|, 1 ≤ |t| ≤ 2





|t| − 3, 2 ≤ |t| ≤ 3





3 − |t|, 3 ≤ |t| ≤ 4
φx2 x2 (t) =
|t| − 5,
 4 ≤ |t| ≤ 5

5 − |t|, 5 ≤ |t| ≤ 6





|t| − 7, 6 ≤ |t| ≤ 7





0, |t| > 7


and φx2 x2 (t) = φx2 x2 (−t)

(b) The cross correlation function of x2(t) and x1(t):
R +∞
φx1 x2 (t) = −∞ x1(t + τ )x2(τ )dτ
case 1:t > 2 φx1 x2 (t) = 0

2
x1 (t + τ ) x2 (τ )
1

−t 2−t 2 3 4 5 6 7
−1
R 2−t
case 2:0 ≤ t ≤ 2 φx1 x2 (t) = 0
(t + τ )/2dτ = 1 − (t2 /4)

8
 3

2
x1 (t + τ ) x2 (τ )
1

−t 2−t 2 3 4 5 6 7
−1
R 2−t
case 3:−1 ≤ t ≤ 0 φx1 x2 (t) = −t
(t + τ )/2dτ = 1

2
x2 (τ ) x1 (t + τ )
1

−t 2 − t3 4 5 6 7
−1
R3 R 2−t
case 4:−3 ≤ t ≤ −1 φx1 x2 (t) = −t
(t + τ )/2dτ + 3
−(t + τ )/2dτ = 3t + 3.5

2
x2 (τ ) x1 (t + τ )
1

−t 3 2−t 5 6 7
−1
R5 R 2−t
case 5:−4 ≤ t ≤ −3 φx1 x2 (t) = −t
−(t + τ )/2dτ + 5
(t + τ )/2dτ = −(t2 /2) − 5t − (23/2)

9
 3

2
x2 (τ ) x1 (t + τ )
1

0 1 2 −t 2−t 6 7
−1
R5 R6 R 2−t
case 6:−5 ≤ t ≤ −4 φx1 x2 (t) = −t
−(t + τ )/2dτ + 5
(t + τ )/2dτ 6
−(t + τ )/2dτ = (3t/2) + (29/4)

2
x2 (τ ) x1 (t + τ )
1

0 1 2 −t 5 62 − t
−1
R6 R7
case 7:−6 ≤ t ≤ −5 φx1 x2 (t) = −t
(t + τ )/2dτ + 6
−(t + τ )/2dτ = (t2 /4) + (5t/2) + (49/4)

2
x2 (τ ) x1 (t + τ )
1

0 1 2 3 4 −t 6 2−
−1
R7
case 8:−7 ≤ t ≤ −6 φx1 x2 (t) = −t
−(t + τ )/2dτ = −(t2 /4) − (7t/2) − (49/4)

10
 3

2
x2 (τ ) x1 (t + τ )
1

0 1 2 3 4 5 6 −t 7
−1

case 9:−t < −7 φx1 x2 (t) = 0

(c) The crosscorrelation function:
R +∞
φxy (t) = −∞
x(t + τ )y(τ )dτ
Let τ =-u then
R +∞
φxy (t) = −∞
x(t − u)y(−u)du
So,
φxy (t) = x(t) ∗ y(−t)
7) Impulse response of the system is

h[n]=δ[n]+ aδ[n − k]

hinv [n] is causal as mentioned in the question

(
1 n=0
h[n] =
a n=k

Pl=∞
h[n]*hinv [n] = l=−∞ h[l]hinv [n − l]

= h[0]hinv [n − 0] + h[k]hinv [n − k]

We know that h[n]*hinv [n] = δ[n]

δ[n] =hinv [n] + ahinv [n − k]

since hinv [n] is causal so ahinv [n − k] exists only for positive values of k
for n < 0

hinv [n]= 0

for n = 0

1 = hinv [0] + 0

for n > 0

hinv [n] = −ahinv [n − k]

which means hinv [n]is nonzero only for postive multiples of k, this statement is verified by susbstituting some
values of n
Substituting n = 1

11
 hinv [1] = −ahinv [1 − k]
(
−a k = 1
hinv [1] =
0 k>1

Substituting n = 2

hinv [2] = −ahinv [2 − k]


2
(−a)
 k=1
hinv [2] = (−a) k=2

0 k>2


Substituting n = 3

hinv [3] = −ahinv [3 − k]

 3
(−a)


k =1
0 k =2
hinv [3] =
(−a)
 k =3

0 k >3


Hence hinv [n]can be written as

Pp=∞
hinv [n] = p=0 (−a)p δ[n − pk]

12