Problem Set 2 Solution

1. Find the convolution of x(t) with h(t), where


t + 1,
 0≤t≤1
x(t) = 2 − t, 1≤t≤2 h(t) = δ(t + 2) + 2δ(t + 1),

0, elsewhere.


where δ(t) represents Dirac-delta impulse distribution.

Solution: The output y(t) is given by:
Z ∞
y(t) = h(t) ∗ x(t) = {δ(τ + 2) + 2δ(τ + 1)} x(t − τ )dτ = x(t + 2) + 2x(t + 1).
−∞

x(t)

−3 −2 −1 1 2 1 t

x(t + 2) 2x(t + 1)

4 4

+
3 3

2 2

1 1

−3 −2 −1 1 2 1 t −3 −2 −1 1 2 1 t

1
 The output therefore is


 0 t < −2,

t + 3 − 2 < t < −1,



y(t) = 4 + t − 1 < t < 0,

2(1 − t) 0 < t < 1,





0 t > 1.

2. For the signal x(t) given below, plot x(1 − 2t ).

x(t)

−1 1 2 t

Solution:

x(1 + t) x(1 − t)

−2 −1 0 1 t −1 0 1 2 t

x(1 − 12 t)

−2 −1 0 1 2 4 t

3. (a) Find the impulse response h(t) of a RC circuit given in Figure 1, considering zero
initial condition.
(b) For a given input x(t) = u(t − 2) find the system output y(t) using the convolution
integral.

1
Figure 1: RC circuit: R = 1Ω, C = F .
3

2
 Solution:
(a) Using KVL we have

x(t) = vr (t) + y(t) = Ri(t) + y(t).

Since i(t) = C dtd y(t), we have

d
x(t) = RC y(t) + y(t). (1)
dt
Corresponding to input x(t) = δ(t), the output is the impulse response h(t). Using this
in equation (1), we get

d
δ(t) = RC h(t) + h(t). (2)
dt
We propose a solution h(t) = Aept u(t) to be a solution. Using this in equation (2)

δ(t) = RCpAept u(t) + RCAδ(t) + Aept u(t). (3)

Therefore, we have
1 1
RCA = 1 ⇒ A = RCp + 1 = 0 ⇒ p = − .
RC RC
Thus, the impulse response is
1 − 1 t
h(t) = e RC u(t) = 3e−3t u(t).
RC

(b) The output is given by

Z ∞
y(t) = x(t) ∗ h(t) = 3 e−3τ u(τ )u(t − τ − 2)dτ.
−∞

Figure 2: RC circuit: R=1, c= 13

To evaluate the integral, we first need to determine the limits of the integration as illus-
trated in Figure 2.
• For t − 2 < 0 or t < 2 there is no overlap, so y(t) = 0.
• for t − 2 > 0 or t > 2 there is overlap so,
Z t−2
e−3τ t−2
y(t) = 3 e−3τ u(τ )u(t − τ − 2)dτ = 3 |0 = [1 − e−3(t−2) ]u(t − 2).
0 −3

4. Let h(t) and x(t) be the impulse response of a system and input to the system, respectively
as defined below:
(
0 ≤ t ≤ 1,
 
1 t
x(t) = and h(t) = x for 0 ≤ α ≤ 1.
0 elsewhere. α

3
 (a) Using convolution integral find y(t).
d
(b) What will be the value of α so that y(t) has three discontinuities only?
dt
Solution:
(a) Given
(
1 0≤t≤α
 
t
h(t) = x for 0 < α ≤ 1 =
α 0 elsewhere.

The corresponding integral limits are described in Figure 3.

• For t ≥ 0 and t − α ≤ 0 and 0 ≤ t ≤ α
Z t
y(t) = 1dτ = [τ ]t0 = t.
0

• For t − α ≥ 0 and t ≤ 1 or α ≤ t ≤ 1
Z t
y(t) = 1dτ = [τ ]tt−α = α.
t−α

• For t ≥ 1 and t − α ≤ 1 or 1 ≤ t ≤ 1 + α
Z 1
y(t) = 1dτ = [τ ]1t−α = 1 − t + α.
t−α

Therefore,


 t 0≤t≤α

α α≤t≤1
y(t) =


 1 − t + α 1 ≤ t ≤ (1 + α)

0 elsewhere.

x(t) h(t) x(t) x(t) x(t)

1 α t 1 1 1 t

h(−τ )

−α τ h(−τ + t) h(−τ + t)
h(−τ + t)
h(−τ + t)

t−α t τ
t−α t τ t−α t t−α t

Figure 3: Illustrations of integration limits for different cases.

(b) Taking derivative of y we can write,



 1 0<t<α

dy(t)  0 α<t<1
=
dt 

 −1 1 < t < 1 + α

0 elsewhere.

This function has 4 discountinuities. When α = 1, it will have three discontinuities.

4
5. Consider an LTI system S and a signal x(t) = 2e−3t u(t − 1). If the output from S due to
input x(t) is y(t) and the output due to input dtd x(t) is −3y(t) + e−2t u(t), then determine
the impulse response h(t) of S.
Solution: Since x(t) = 2e−3t u(t − 1), we have

d
x(t) = −6e−3t u(t − 1) + 2e−3t δ(t − 1) = −3x(t) + 2e−3t δ(t − 1).
dt
d
Let h(t) be the impulse response of the system. Hence, output due to input dt
x(t) is
given by
d
x(t) ∗ h(t) = −3x(t) + 2e−3t δ(t − 1) ∗ h(t)


y1 (t) :=
dt
= −3(x(t) ∗ h(t)) + 2(e−3t δ(t − 1)) ∗ h(t). (4)

Using the fact that y(t) = x(t) ∗ h(t) and y1 (t) = −3y(t) + e−2t u(t) in equation (4), we
have

−3y(t) + e−2t u(t) = −3y(t) + 2(e−3t δ(t − 1)) ∗ h(t) ⇒ e−2t u(t) = 2(e−3t δ(t − 1)) ∗ h(t).
(5)

Note that
Z ∞
−3t
(e δ(t − 1)) ∗ h(t) = e−3τ δ(τ − 1)h(t − τ )dτ = e−3 h(t − 1). (6)
−∞

Using equation (6) in equation (5), we have

1
e−2t u(t) = 2e−3 h(t − 1) ⇒ h(t − 1) = e3−2t u(t).
2
Therefore, the impulse response is
1 1
h(t) = e3−2(t+1) u(t + 1) = e1−2t u(t + 1).
2 2

6. Consider convolution of two causal sequences u[n] and h[n] (both with finite support) to
give y[n].

(a) Show that one can construct a ‘Toeplitz Matrix T ’ using entries of h such that
   
y(0) u(0)
y(1)
 = T u(1) with entries of T coming from h.
 
.. ..

. .

(b) Relate support of h, u with y.

(c) Suppose the support is fixed, but h and u are changed for this fixed support. Can
support of y change for different h, u, or does the support of y depend only on the
supports of h and u?
(d) Now consider the signals to be non-causal and having finite support. How is the
support of h related to the supports of u and y?

5
 Solution : (a) LetPthe support of u[n] and h[n] be Nu and Nh , respectively. Since
y[n] = u[n] ∗ h[n] = ∞
k=0 u[k]h[n − k], we have

···
 
h[0] 0 0
 h[1] h[0] ··· 0 
..
 
 .. .. .. 
 . . . . 
..
 
 .. .. 
 . . . h[0] 
.. ..
     
y[0]  ..  u[0]

y[1]  . . . h[1]  

= ..  u[1] (7)

.. .. ..
.. ..

 . . . . 
. 
 .. .. 
 .
h[Nh ] h[Nh − 1] . . 

 .. .. 
 0
 h[Nh ] . .  
 . .. .. .. 
 .. . . . 
0 0 ··· h[Nh ] T
| {z }

(b) Let the support of y[n] be Ny . From equation (7) it is clear that T ∈ C(Nu +Nh +1)×(Nu +1) .
Therefore, Ny = Nh + Nu + 1.
(c) From equation (7) it is clear that the support of y[n] depends on the number of rows
of T , i.e., (Nu + Nh + 1). Therefore, the support of y[n] depends only on the support of
u[n] and h[n].
(d) Note that a noncausal signal can always be uniquely into an anticausal and causal
part. Hence, both u[n] and h[n] being noncausal can be divided into an anticausal and
causal part. Further, since the signals h[n] and u[n] have finite support, the respective
anticausal and causal parts will also have finite support. Hence, we define the following:

• Support of the anticausal and causal part of h[n] be Nhanti and Nhcaus , respectively.
• Support of the anticausal and causal part of u[n] be Nuanti and Nucaus , respectively.

Hence the support of y[n] is (Nhanti + Nhcaus ) + (Nuanti + Nucaus ) + 1.

7. Prove or disprove the following:

(a) x(t) = 2 cos(t) + 3 cos(t/3) is a periodic signal of fundamental period 2π.

Solution: False. At t = 0, x(0) = 5 and at t = 2π, x(2π) = 23 × 21 = 21 . Hence,
the period of x(t) cannot be 2π.
(b) If h(t) is the impulse response of an LTI system and h(t) is periodic and nonzero,
the system is unstable. R∞
Solution: True. For a system to be stable −∞ |h(t)|dt = M , where M ∈ R.
If h(t) is periodic with period T , then
Z ∞ Z T /2
|h(t)|dt = lim N |h(t)|dt.
−∞ N →∞ −T /2

R T /2
Since h(t) is non-zero and −T /2 |h(t)|dt is a finite number for a periodic impulse
R T /2
response, limN →∞ N −T /2 |h(t)|dt = ∞. Hence, the system is unstable.
(c) If an LTI system is causal, it is stable.
Solution: False. Consider the system P∞with impulse P
response h[n] = u[n]. The
∞
system is causal. However the sum k=−∞ |h[n]| = k=0 1 does not converge.
Therefore, the system is not stable.

6
8. If an auditorium has a perceptible echo, then an initial acoustic impulse will be followed
by attenuated versions of the sound at regularly spaced intervals. The model for this
phenomenon is an LTI system with an impulse response:
∞
X
h(t) = hk δ(t − kT ).
k=0

Here the echoes occur T seconds apart, and hk represents the gain factor on the k-th echo
resulting from an initial acoustic impulse.
Suppose x(t) represents the original acoustic signal and that y(t) = x(t)∗h(t) is the actual
signal that is heard if no processing is done to remove the echoes. In order to remove
the distortion due to echoes, assume a microphone is used to sense y(t) and the resulting
signal is transduced into an electrical signal – we call this electrical signal y(t) as well.
Can we find an LTI system with an impulse response g(t) of the form:
∞
X
g(t) = gk δ(t − kT ),
k=0

such that the effect of echo is removed. If yes what should be the values of g0 , g1 and g2
in terms of hk .
Solution: In order to remove the effect of echo the LTI system g(t) must be inverse of
the system h(t). In other words,
∞
X ∞
X
h(t) ∗ g(t) = δ(t) ⇒ hm δ(t − mT ) ∗ gk δ(t − kT ) = δ(t). (8)
m=0 k=0

Note that using the distributive property of convolution we have

∞
X ∞
X ∞ X
X ∞
hm δ(t − mT ) ∗ gk δ(t − kT ) = hm gk (δ(t − mT ) ∗ δ(t − kT )) (9)
m=0 k=0 m=0 k=0

Now we evaluate the term δ(t − mT ) ∗ δ(t − kT ). Let φ(t) be a test-function. Then,
Z ∞
φ(t) (δ(t − mT ) ∗ δ(t − kT )) dt
−∞
Z ∞ Z ∞ 
= φ(t) δ(t − mT − τ )δ(τ − kT )dτ dt
−∞ −∞
Z ∞ Z ∞ 
= (φ(t)δ(t − mT − τ )dt) δ(τ − kT )dτ
−∞ −∞
Z ∞
= φ(mT + τ )δ(τ − kT )dτ = φ(mT + kT ).
−∞

Therefore, the effect of (δ(t − mT ) ∗ δ(t − kT )) is equivalent to δ(t − (m + k)T ). Hence,

δ(t − mT ) ∗ δ(t − kT ) = δ(t − (m + k)T ). Using this in equation (9) and then invoking
(8) we have
∞ X
X ∞
h(t) ∗ g(t) = hm gk δ(t − (m + k)T ) = δ(t).
m=0 k=0

Therefore, on equating the coefficients of δ(t), δ(t−T ) and δ(t−2T ), we get the following:
1
h0 g0 = 1 ⇒ g0 =
.
h0
h1
h0 g1 + h1 g0 = 0 ⇒ g1 = − 2 .
h0
2
h h2
h0 g2 + h1 g1 + h2 g0 = 0 ⇒ g2 = 13 − .
h0 h1 h0

7
9. Consider a discrete-time LTI system with unit sample response

h[n] = (n + 1)αn u[n],

where |α| < 1. Show that the step response of this system is
 
1 α n α n
s[n] = − α + (n + 1)α u[n].
(α − 1)2 (α − 1)2 (α − 1)

Solution (9): The output due to an unit step input x[n] = u[n] is
n n+1
n
X
k d X k
y[n] = h[n] ∗ x[n] = ((n + 1)α u[n]) ∗ u[n] = (k + 1)α = α , n > 0.
k=0
dα k=0

Since |α| < 1, the sum above converges. Thus, we have

d 1 − αn+2
 
y[n] = , n>0
dα 1−α
 
1 α n α n
= − α + (n + 1)α u[n].
(α − 1)2 (α − 1)2 (α − 1)