45

Chapter 3

Continuous-Time Linear Time-Invariant

Systems

3.1 Introduction
In the previous chapter, we identified a number of properties that a system may possess. Two of these properties were
the linearity and time invariance properties. In this chapter, we focus our attention exclusively on systems with both
of these properties. Such systems are referred to as linear time-invariant (LTI) systems.

3.2 Continuous-Time Convolution

In the context of LTI systems, we often find an operation known as convolution to be particularly useful. The convo-
lution of the functions x(t) and h(t) is denoted as x(t) ∗ h(t) and defined as
Z ∞
x(t) ∗ h(t) = x(τ )h(t − τ )d τ . (3.1)
−∞

Throughout the remainder of these notes, the asterisk (or star) symbol (i.e., “∗”) will be used to denote convolution,
not multiplication. It is important to make a distinction between convolution and multiplication, since these two
operations are quite different and do not generally yield the same result.
One must also be careful when using the abbreviated (i.e., star notation) for convolution, since this notation can
sometimes behave in counter-intuitive ways. For example, the expressions x(t − t0 ) ∗ h(t − t0 ) and x(τ ) ∗ h(τ )|τ =t−t0 ,
where t0 is a nonzero constant, have very different meanings and are almost never equal.
Since the convolution operation is used extensively in system theory, we need some practical means for evaluating
a convolution integral. Suppose that, for the given functions x(t) and h(t), we wish to compute
Z ∞
y(t) = x(t) ∗ h(t) = x(τ )h(t − τ )d τ .
−∞

Of course, we could naively attempt to compute y(t) by evaluating a separate convolution integral for each possible
value of t. This approach, however, is not feasible, as t can assume an infinite number of values, and therefore, an
infinite number of integrals would need to be evaluated. Instead, we consider a slightly different approach. Let us
redefine the integrand in terms of the intermediate function wt (τ ) where
wt (τ ) = x(τ )h(t − τ ).
(Note that wt (τ ) is implicitly a function of t.) This means that we need to compute
Z ∞
y(t) = wt (τ )d τ .
−∞

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

46 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

Now, we observe that, for most functions x(t) and h(t) of practical interest, the form of wt (τ ) typically remains fixed
over particular ranges of t. Thus, we can compute the integral y(t) by first identifying each of the distinct expressions
for wt (τ ) and the range over which each expression is valid. Then, for each range, we evaluate a convolution integral.
In this way, we typically only need to compute a small number of integrals instead of the infinite number required
with the naive approach suggested above.
The above discussion leads us to propose the following general approach for performing the convolution operation:
1. Plot x(τ ) and h(t − τ ) as a function of τ .
2. Initially, consider an arbitrarily large negative value for t. This will result in h(t − τ ) being shifted very far to
the left on the time axis.
3. Write the mathematical expression for wt (τ ).
4. Increase t gradually until the expression for wt (τ ) changes form. Record the interval over which the expression
for wt (τ ) was valid.
5. Repeat steps 3 and 4 until t is an arbitrarily large positive value. This corresponds to h(t − τ ) being shifted very
far to the right on the time axis.
6. For each of the intervals identified above, integrate wt (τ ) in order to find an expression for y(t). This will yield
an expression for y(t) for each interval.
7. The results for the various intervals can be combined in order to obtain an expression for y(t) that is valid for
all t.
Example 3.1. Compute the convolution y(t) = x(t) ∗ h(t) where

−1 for −1 ≤ t < 0

x(t) = 1 for 0 ≤ t < 1


0 otherwise, and
h(t) = e−t u(t).

Solution. We begin by plotting the signals x(τ ) and h(τ ) as shown in Figures 3.1(a) and (b), respectively. Next, we
proceed to determine the time-reversed and time-shifted version of h(τ ). We can accomplish this in two steps. First,
we time-reverse h(τ ) to obtain h(−τ ) as shown in Figure 3.1(c). Second, we time-shift the resulting signal by t to
obtain h(t − τ ) as shown in Figure 3.1(d).
At this point, we are ready to begin considering the computation of the convolution integral. For each possible
value of t, we must multiply x(τ ) by h(t − τ ) and integrate the resulting product with respect to τ . Due to the form of
x(τ ) and h(τ ), we can break this process into a small number of cases. These cases are represented by the scenarios
illustrated in Figures 3.1(e) to (h).
First, we consider the case of t < −1. From Figure 3.1(e), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.2)
−∞

Second, we consider the case of −1 ≤ t < 0. From Figure 3.1(f), we can see that
Z ∞ Z t
x(τ )h(t − τ )d τ = −eτ −t d τ
−∞ −1
Z t
= −e−t eτ d τ
−1
= −e−t [eτ ]|t−1
−t t −1
= −e [e − e ]
= e−t−1 − 1. (3.3)

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.2. CONTINUOUS-TIME CONVOLUTION 47

Third, we consider the case of 0 ≤ t < 1. From Figure 3.1(g), we can see that
Z ∞ Z 0 Z t
x(τ )h(t − τ )d τ = −eτ −t d τ + eτ −t d τ
−∞ −1 0
Z 0 Z t
= −e−t eτ d τ + e−t eτ d τ
−1 0
= −e−t [eτ ]|0−1 + e−t [eτ ]|t0
−t −1 −t t
= −e [1 − e ] + e [e − 1]
= e−t [e−1 − 1 + et − 1]
= 1 + (e−1 − 2)e−t . (3.4)

Fourth, we consider the case of t ≥ 1. From Figure 3.1(h), we can see that
Z ∞ Z 0 Z 1
x(τ )h(t − τ )d τ = −eτ −t d τ + eτ −t d τ
−∞ −1 0
Z 0 Z 1
τ
= −e −t
e dτ + e −t
eτ d τ
−1 0
= −e−t [eτ ]|0−1 + e−t [eτ ]|10
−t −1
= e [e − 1 + e − 1]
= (e − 2 + e−1 )e−t . (3.5)

Combining the results of (3.2), (3.3), (3.4), and (3.5), we have that


 0 for t < −1

e−t−1 − 1 for −1 ≤ t < 0
x(t) ∗ h(t) =


 (e−1 − 2)e−t + 1 for 0 ≤ t < 1

(e − 2 + e−1 )e−t for 1 ≤ t.

The convolution result x(t) ∗ h(t) is plotted in Figure 3.1(i).

Example 3.2. Compute the convolution y(t) = x(t) ∗ h(t) where

(
1 for 0 ≤ t < 1
x(t) = and
0 otherwise,
(
t for 0 ≤ t < 1
h(t) =
0 otherwise.

Solution. We begin by plotting the signals x(τ ) and h(τ ) as shown in Figures 3.2(a) and (b), respectively. Next, we
proceed to determine the time-reversed and time-shifted version of h(τ ). We can accomplish this in two steps. First,
we time-reverse h(τ ) to obtain h(−τ ) as shown in Figure 3.2(c). Second, we time-shift the resulting signal by t to
obtain h(t − τ ) as shown in Figure 3.2(d).
At this point, we are ready to begin considering the computation of the convolution integral. For each possible
value of t, we must multiply x(τ ) by h(t − τ ) and integrate the resulting product with respect to τ . Due to the form of
x(τ ) and h(τ ), we can break this process into a small number of cases. These cases are represented by the scenarios
illustrated in Figures 3.2(e) to (h).
First, we consider the case of t < 0. From Figure 3.2(e), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.6)
−∞

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

48 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(τ )
1
h(t − τ )

τ
x(τ ) t −1 1

1
−1
(e)
τ x(τ )
−1 1
1
h(t − τ )

−1
τ
−1 t 1
(a)

h(τ ) −1
(f)
1
x(τ )
1

h(t − τ )
0 τ
(b) τ
−1 t 1

h(−τ )
−1
1 (g)
eτ
x(τ )
1
0 τ h(t − τ )

(c)
τ
−1 1 t
h(t − τ )
1
−1
eτ −t
0 (h)

t τ 0.4

0.2
(d) 0
−2 −1 0 1 2 3 4
−0.2

−0.4

−0.6

−0.8

(i)

Figure 3.1: Evaluation of the convolution integral. The (a) input signal x(τ ), (b) impulse response h(τ ), (c) time-
reversed impulse response h(−τ ), and (d) impulse response after time-reversal and time-shifting h(t − τ ). The func-
tions associated with the product in the convolution integral for (e) t < −1, (f) −1 ≤ t < 0, (g) 0 ≤ t < 1, and (h) t ≥ 1,
(i) The convolution result x(t) ∗ h(t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.2. CONTINUOUS-TIME CONVOLUTION 49

Second, we consider the case of 0 ≤ t < 1. From Figure 3.2(f), we can see that
Z ∞ Z t
x(τ )h(t − τ )d τ = (t − τ )d τ
−∞ 0
= [t τ − 21 τ 2 ]|t0
= t 2 − 12 t 2
= 21 t 2 . (3.7)

Third, we consider the case of 1 ≤ t < 2. From Figure 3.2(g), we can see that
Z ∞ Z 1
x(τ )h(t − τ )d τ = (t − τ )d τ
−∞ t−1
= [t τ − 12 τ 2 ]t−1
1

= t − 21 (1)2 − [t(t − 1) − 12 (t − 1)2 ]

= t − 12 − [t 2 − t − 12 (t 2 − 2t + 1)]
= − 21 t 2 + t. (3.8)

Fourth, we consider the case of t ≥ 2. From Figure 3.2(h), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.9)
−∞

Combining the results of (3.6), (3.7), (3.8), and (3.9), we have that


0 for t < 0

 1 t2
2 for 0 ≤ t < 1
x(t) ∗ h(t) =
− 1 t2 + t for 1 ≤ t < 2
 2


0 for t ≥ 2.

The convolution result x(t) ∗ h(t) is plotted in Figure 3.2(i).

Example 3.3. Compute the quantity y(t) = x(t) ∗ h(t), where



0 for t < 0

t for 0 ≤ t < 1
x(t) = and


−t + 2 for 1 ≤ t < 2

0 for t > 2,
h(t) = u(t) − u(t − 1).

Solution. Due to the somewhat ugly nature of the expressions for x(t) and h(t), this problem can be more easily solved
if we use the graphical interpretation of convolution to guide us. We begin by plotting the signals x(τ ) and h(τ ), as
shown in Figures 3.3(a) and (b), respectively.
Next, we need to determine h(t − τ ), the time-reversed and time-shifted version of h(τ ). We can accomplish this in
two steps. First, we time-reverse h(τ ) to obtain h(−τ ) as shown in Figure 3.3(c). Second, we time-shift the resulting
signal by t to obtain h(t − τ ) as shown in Figure 3.3(d).
At this point, we are ready to begin considering the computation of the convolution integral. For each possible
value of t, we must multiply x(τ ) by h(t − τ ) and integrate the resulting product with respect to τ . Due to the form of
x(τ ) and h(τ ), we can break this process into a small number of cases. These cases are represented by the scenarios
illustrated in Figures 3.3(e) to (i).

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

50 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(τ )
1
h(t − τ )

x(τ )
τ
t −1 t 0 1
1 (e)

x(τ )
τ 1
0 1
h(t − τ )
(a)

h(τ ) τ
t −1 0 t 1
(f)
1

x(τ )
1
τ
0 1
h(t − τ )
(b)
τ
0 t −1 1 t
h(−τ )
(g)

1
x(τ )
1
h(t − τ )
τ
−1 0
(c)
τ
0 1 t −1 t

1 (h)
h(t − τ )
x(t) ∗ h(t)

τ
t −1 t
1
(d)
1
2

t
0 1 2
(i)

Figure 3.2: Evaluation of the convolution integral. The (a) input signal x(τ ), (b) impulse response h(τ ), (c) time-
reversed impulse response h(−τ ), and (d) impulse response after time-reversal and time-shifting h(t − τ ). The func-
tions associated with the product in the convolution integral for (e) t < 0, (f) 0 ≤ t < 1, (g) 1 ≤ t < 2, and (h) t ≥ 2,
(i) The convolution result x(t) ∗ h(t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.2. CONTINUOUS-TIME CONVOLUTION 51

First, we consider the case of t < 0. From Figure 3.3(e), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.10)
−∞

Second, we consider the case of 0 ≤ t < 1. From Figure 3.3(f), we can see that
Z ∞ Z t
x(τ )h(t − τ )d τ = τ dτ
−∞ 0
1 2 t
= [ 2 τ ]0
= 21 t 2 . (3.11)

Third, we consider the case of 1 ≤ t < 2. From Figure 3.3(g), we can see that
Z ∞ Z 1 Z t
x(τ )h(t − τ )d τ = τ dτ + (−τ + 2)d τ
−∞ t−1 1
= [ 21 τ 2 ]t−1
1
+ [− 12 τ 2 + 2τ ]t1
= 21 − [ 12 (t − 1)2 ] − 21 t 2 + 2t − [− 12 + 2]
= −t 2 + 3t − 23 . (3.12)

Fourth, we consider the case of 2 ≤ t < 3. From Figure 3.3(h), we can see that
Z ∞ Z 2
x(τ )h(t − τ )d τ = (−τ + 2)d τ
−∞ t−1
2
= [− 12 τ 2 + 2τ ]t−1
= 2 − [− 12 t 2 + 3t − 25 ]
= 21 t 2 − 3t + 92 . (3.13)

Lastly, we consider the case of t ≥ 3. From Figure 3.3(i), we can see that
Z ∞
x(τ )h(t − τ ) = 0. (3.14)
−∞

Combining the results of (3.10), (3.11), (3.12), (3.13), and (3.14) together, we have that



 0 for t < 0
 1 2
2t

 for 0 ≤ t < 1
x(t) ∗ h(t) = −t 2 + 3t − 32 for 1 ≤ t < 2

 1 2 9
2 t − 3t + 2 for 2 ≤ t < 3




0 for t ≥ 3.

The convolution result (i.e., x(t) ∗ h(t)) is plotted in Figure 3.3(j).

Example 3.4. Compute the convolution y(t) = x(t) ∗ h(t) where

x(t) = e−at u(t), and

h(t) = u(t),

and a is a positive real constant.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

52 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

h(t − τ )
1
x(τ )

τ
t −1 t 0 1 2
(e)
x(τ )
h(t − τ )
1
1 x(τ )

τ τ
0 1 2 t −1 0 t 1 2

(a) (f)

h(τ ) h(t − τ )
1
1
x(τ )

τ
τ 0 t −1 1 t 2
0 1
(g)
(b)
h(t − τ )
h(−τ ) x(τ )
1
1

τ
0 1 t −1 2 t
τ
−1 0 (h)
(c)
h(t − τ )
h(t − τ ) 1
x(τ )
1

τ
0 1 2 t −1 t
τ
t −1 t (i)
(d)
x(t) ∗ h(t)

1
3
4

3
t
0 1 2 2 3
(j)

Figure 3.3: Evaluation of the convolution integral. The (a) input signal x(τ ), (b) impulse response h(τ ), (c) time-
reversed impulse response h(−τ ), and (d) impulse response after time-reversal and time-shifting h(t − τ ). The func-
tions associated with the product in the convolution integral for (e) t < 0, (f) 0 ≤ t < 1, (g) 1 ≤ t < 2, (h) 2 ≤ t < 3,
and (i) t ≥ 3. (j) The convolution result x(t) ∗ h(t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.3. PROPERTIES OF CONVOLUTION 53

Solution. Since x(t) and h(t) are relatively simple functions, we will solve this problem without the aid of graphs.
Using the definition of the convolution operation, we can write:
Z ∞
y(t) = x(t) ∗ h(t) = x(τ )h(t − τ )d τ
−∞
Z ∞
= e−aτ u(τ )u(t − τ )d τ . (3.15)
−∞
The integrand is zero for τ < 0 or τ > t. Conversely, the integrand can only be nonzero for τ ≥ 0 and τ ≤ t. So, if
t < 0, the integrand will be zero, and y(t) = 0. Now, let us consider the case of t > 0. From (3.15), we can write:
Z t
y(t) = e−aτ d τ
0
= [− a1 e−aτ ]t0
= a1 (1 − e−at ).
Thus, we have
(
1 −at )
y(t) = a (1 − e for t > 0
0 otherwise
= 1a (1 − e−at )u(t).
(If some steps in the above solution are unclear, it would probably be helpful to sketch the corresponding graphs. This
will yield the graphs shown in Figure 3.4.)

3.3 Properties of Convolution

Since convolution is frequently employed in the study of LTI systems, it is important for us to know some of its basic
properties. In what follows, we examine some of these properties. We will later use these properties in the context of
LTI system.

3.3.1 Commutative Property

The convolution operation is commutative. That is, for any two signals x(t) and h(t), we have
x(t) ∗ h(t) = h(t) ∗ x(t). (3.16)
Thus, the result of the convolution operation is not affected by the order of the operands.
We now provide a proof of the commutative property stated above. To begin, we expand the left-hand side of (3.16)
as follows:
Z ∞
x(t) ∗ h(t) = x(τ )h(t − τ )d τ .
−∞

Next, we perform a change of variable. Let v = t − τ which implies that τ = t − v and d τ = −dv. Using this change
of variable, we can rewrite the previous equation as
Z −∞
x(t) ∗ h(t) = x(t − v)h(v)(−dv)
Z∞∞
= x(t − v)h(v)dv
−∞
Z ∞
= h(v)x(t − v)dv
−∞
= h(t) ∗ x(t).
Thus, we have proven that convolution is commutative.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

54 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(τ ) h(τ )

1 1
e−at

τ
0 τ 0
(a) (b)
h(−τ ) h(t − τ )
1
1

τ τ
0 t
(c) (d)
h(t − τ ) h(t − τ )
1
1
x(τ ) x(τ )

τ τ
t 0 0 t

(e) (f)

Figure 3.4: Evaluation of the convolution integral. The (a) input signal x(τ ), (b) impulse response h(τ ), (c) time-
reversed impulse response h(−τ ), and (d) impulse response after time-reversal and time-shifting h(t − τ ). The func-
tions associated with the product in the convolution integral for (e) t < 0 and (f) t > 0.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.3. PROPERTIES OF CONVOLUTION 55

3.3.2 Associative Property

The convolution operation is associative. That is, for any signals x(t), h1 (t), and h2 (t), we have

[x(t) ∗ h1 (t)] ∗ h2 (t) = x(t) ∗ [h1 (t) ∗ h2 (t)]. (3.17)

In other words, the final convolution result does not depend on how the intermediate operations are grouped.
The proof of this property is relatively straightforward, although somewhat confusing notationally. To begin, we
use the definition of the convolution operation to expand the left-hand side of (3.17) as follows:
Z ∞
[x(t) ∗ h1 (t)] ∗ h2 (t) = [x(v) ∗ h1 (v)]h2 (t − v)dv
−∞
Z ∞ Z ∞ 
= x(τ )h1 (v − τ )d τ h2 (t − v)dv.
−∞ −∞

Now, we change the order of integration to obtain

Z ∞Z ∞
[x(t) ∗ h1 (t)] ∗ h2 (t) = x(τ )h1 (v − τ )h2 (t − v)dvd τ .
−∞ −∞

Pulling the factor of x(τ ) out of the inner integral yields

Z ∞ Z ∞
[x(t) ∗ h1 (t)] ∗ h2 (t) = x(τ ) h1 (v − τ )h2 (t − v)dvd τ .
−∞ −∞

Next, we perform a change of variable. Let λ = v − τ which implies that v = λ + τ and d λ = dv. Using this change
of variable, we can write
Z ∞ Z ∞
[x(t) ∗ h1 (t)] ∗ h2 (t) = x(τ ) h1 (λ )h2 (t − λ − τ )d λ d τ
−∞ −∞
Z ∞ Z ∞

= x(τ ) h1 (λ )h2 ([t − τ ] − λ )d λ d τ
−∞ −∞
Z ∞
= x(τ ) [h1 (ρ ) ∗ h2 (ρ )]|ρ =t−τ d τ
−∞
= x(t) ∗ [h1 (t) ∗ h2 (t)].

Thus, we have proven that convolution is associative.

3.3.3 Distributive Property

The convolution operation is distributive. That is, for any signals x(t), h1 (t), and h2 (t), we have

x(t) ∗ [h1 (t) + h2 (t)] = x(t) ∗ h1 (t) + x(t) ∗ h2 (t). (3.18)

The proof of this property is relatively simple. Expanding the left-hand side of (3.18), we have:
Z ∞
x(t) ∗ [h1 (t) + h2 (t)] = x(τ )[h1 (t − τ ) + h2 (t − τ )]d τ
−∞
Z ∞ Z ∞
= x(τ )h1 (t − τ )d τ + x(τ )h2 (t − τ )d τ
−∞ −∞
= x(t) ∗ h1 (t) + x(t) ∗ h2 (t).

Thus, we have shown that convolution is distributive.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

56 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(t) y(t)
H

Figure 3.5: System.

3.4 Representation of Continuous-Time Signals Using Impulses

The unit-impulse function is of fundamental importance in many aspects of system theory. For this reason, we
sometimes have the need to express an arbitrary signal in terms of impulse functions. In what follows, we develop a
means for accomplishing just this.
Suppose that we have an arbitrary signal x(t). From the equivalence property of the impulse function given
in (2.30), we can write
x(t)δ (t − τ ) = x(τ )δ (t − τ ).
Now, let us integrate both sides of the preceding equation with respect to τ to obtain
Z ∞ Z ∞
x(t)δ (t − τ )d τ = x(τ )δ (t − τ )d τ . (3.19)
−∞ −∞

The left-hand side of this equation can be further simplified as follows:

Z ∞ Z ∞
x(t)δ (t − τ )d τ = x(t) δ (t − τ )d τ
−∞ −∞
Z −∞
= x(t) δ (λ )(−d λ )
Z∞∞
= x(t) δ (λ )d λ
−∞
= x(t). (3.20)

(In the above simplification, we used the change of variable λ = t − τ , d λ = −d τ .) Combining (3.19) and (3.20), we
obtain Z ∞
x(t) = x(τ )δ (t − τ )d τ = x(t) ∗ δ (t). (3.21)
−∞
Thus, we can represent any signal x(t) using an expression containing the impulse function. Furthermore, we have
also just shown that the unit impulse function is the convolutional identity. That is, we have that, for any x(t),

x(t) ∗ δ (t) = x(t)

(i.e., convolving a function x(t) with the unit-impulse function δ (t) simply yields x(t)).

3.5 Continuous-Time Unit-Impulse Response and Convolution Integral Rep-

resentation of LTI Systems
Suppose that we have an arbitrary system with input x(t) and output y(t). Let us represent the processing performed by
the system with the operator H , so that y(t) = H {x(t)}. In other words, we have the system shown in Figure 3.5. As
a matter of terminology, the system response to the unit-impulse function input is referred to as the impulse response.
Let us denote the impulse response of the system as h(t). Mathematically, we can state the definition of the impulse
response as
h(t) = H {δ (t)}. (3.22)
Now, let us assume that the system is LTI (i.e., the operator H is both linear and time invariant). As we shall
demonstrate below, the behavior of a LTI system is completely characterized by its impulse response. That is, if the
impulse response of a system is known, we can determine the response of the system to any input.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.5. CONTINUOUS-TIME UNIT-IMPULSE RESPONSE AND CONVOLUTION INTEGRAL
REPRESENTATION OF LTI SYSTEMS 57

From the earlier result, we know that we can represent any signal in the form of (3.21). So, let us express the input
to the system in this form as follows: Z ∞
x(t) = x(τ )δ (t − τ )d τ . (3.23)
−∞
Now, let us consider the form of the output y(t). To begin, we know that
y(t) = H {x(t)}.

From (3.23), we can rewrite this equation as

Z ∞

y(t) = H x(τ )δ (t − τ )d τ .
−∞

Since H is a linear operator, we can move H inside the integral, and simplify the result to obtain
Z ∞
y(t) = H {x(τ )δ (t − τ )}d τ
−∞
Z ∞
= x(τ )H {δ (t − τ )}d τ . (3.24)
−∞

Since the system is time invariant (by assumption), we know that

h(t − τ ) = H {δ (t − τ )}. (3.25)
Thus, we can rewrite (3.24) as
Z ∞
y(t) = x(τ )h(t − τ )d τ
−∞
= x(t) ∗ h(t).
In other words, the output y(t) is simply the convolution of the input x(t) and the impulse response h(t).
Clearly, the impulse response provides a very powerful tool for the study of LTI systems. If we know the impulse
response of a system, we can determine the response of the system to any input.
Example 3.5. Suppose that we are given a LTI system with input x(t), output y(t), and impulse response h(t) where
h(t) = u(t). (3.26)
Show that this system is the integrator characterized by the equation
Z t
y(t) = x(τ )d τ . (3.27)
−∞

Solution. Since the system is LTI, we have that

y(t) = x(t) ∗ h(t).
Substituting (3.26) into the preceding equation, and simplifying we obtain
y(t) = x(t) ∗ h(t)
= x(t) ∗ u(t)
Z ∞
= x(τ )u(t − τ )d τ
−∞
Z t Z ∞
= x(τ )u(t − τ )d τ + x(τ )u(t − τ )d τ
−∞ t
Z t
= x(τ )d τ .
−∞

Therefore, the system with the impulse response h(t) given by (3.26) is, in fact, the integrator given by (3.27).

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

58 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

Example 3.6. Suppose that we have a LTI system with input x(t), output y(t), and impulse response h(t) where
(
1 for 0 ≤ t < 1
h(t) =
0 otherwise.

Find and plot the response of the system to the particular input x(t) given by
(
1 for 0 ≤ t < 1
x(t) =
0 otherwise.
Plots of x(t) and h(t) are given in Figures 3.6(a) and (b), respectively.
Solution. Since the system is LTI, we know that
y(t) = x(t) ∗ h(t).
Thus, in order to find the response of the system to the input x(t), we simply need to compute the convolution
x(t) ∗ h(t).
We begin by plotting the signals x(τ ) and h(τ ) as shown in Figures 3.6(a) and (b), respectively. Next, we proceed
to determine the time-reversed and time-shifted version of h(τ ). We can accomplish this in two steps. First, we
time-reverse h(τ ) to obtain h(−τ ) as shown in Figure 3.6(c). Second, we time-shift the resulting signal by t to obtain
h(t − τ ) as shown in Figure 3.6(d).
At this point, we are ready to begin considering the computation of the convolution integral. For each possible
value of t, we must multiply x(τ ) by h(t − τ ) and integrate the resulting product with respect to τ . Due to the form of
x(τ ) and h(τ ), we can break this process into a small number of cases. These cases are represented by the scenarios
illustrated in Figures 3.6(e) to (h).
First, we consider the case of t < 0. From Figure 3.6(e), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.28)
−∞

Second, we consider the case of 0 ≤ t < 1. From Figure 3.6(f), we can see that
Z ∞ Z t
x(τ )h(t − τ )d τ = dτ
−∞ 0
= [τ ]|t0
= t. (3.29)
Third, we consider the case of 1 ≤ t < 2. From Figure 3.6(g), we can see that
Z ∞ Z 1
x(τ )h(t − τ )d τ = dτ
−∞ t−1
1
= [τ ]|t−1
= 1 − (t − 1)
= 2 − t. (3.30)
Fourth, we consider the case of t ≥ 2. From Figure 3.6(h), we can see that
Z ∞
x(τ )h(t − τ )d τ = 0. (3.31)
−∞

Combining the results of (3.28), (3.29), (3.30), and (3.31), we have that


0 for t < 0

t for 0 ≤ t < 1
x(t) ∗ h(t) =


2 − t for 1 ≤ t < 2

0 for t ≥ 2.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.6. UNIT-STEP RESPONSE OF LTI SYSTEMS 59

The convolution result x(t) ∗ h(t) is plotted in Figure 3.6(i). The response of the system to the specified input is simply
the result of the convolution (i.e., x(t) ∗ h(t)).

3.6 Unit-Step Response of LTI Systems

Suppose that we have a system with input x(t), output y(t), and impulse response h(t). Consider the response s(t) of
the system to the unit-step function input u(t). We call s(t) the step response of the system. This response s(t) is
given by

s(t) = u(t) ∗ h(t)

= h(t) ∗ u(t)
Z ∞
= h(τ )u(t − τ )d τ
−∞
Z t
= h(τ )d τ .
−∞

Taking the derivative of s(t) with respect to t, we obtain

ds(t) s(t + ∆t) − s(t)

= lim
dt ∆t→0 ∆t
Z t+∆t Z t 
1
= lim h(τ )d τ − h(τ )d τ
∆t→0 ∆t −∞ −∞
Z t+∆t
1
= lim h(τ )d τ
∆t→0 ∆t t
1
= lim (h(t)∆t)
∆t→0 ∆t
= h(t).

Thus, we have shown that

ds(t)
= h(t).
dt

That is, the impulse response h(t) of a system is equal to the derivative of its step response s(t). Therefore, the impulse
response of a system can be determined from its step response simply through differentiation.
The step response is often of great practical interest, since it can be used to determine the impulse response of
a LTI system. From a practical point of view, the step response is more useful for characterizing a system based on
experimental measurements. Obviously, we cannot directly measure the impulse response of a system because we
cannot (in the real world) produce a unit-impulse signal. We can, however, produce a reasonably good approximation
of the unit-step function in the real world. Thus, we can measure the step response and from it determine the impulse
response.

3.7 Block Diagram Representation of Continuous-Time LTI Systems

Frequently, it is convenient to represent continuous-time LTI systems in block diagram form. Since such systems are
completely characterized by their impulse response, we often label the system with its impulse response. That is, we
represent a system with input x(t), output y(t), and impulse response h(t), as shown in Figure 3.7.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

60 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

h(t − τ ) x(τ )
1

x(τ )
τ
t −1 t 0 1
1
(e)

τ x(τ )
0 1 1
(a) h(t − τ )

h(τ )
τ
t −1 0 t 1

1 (f)

x(τ )
τ 1
0 1 h(t − τ )
(b)

τ
h(−τ ) 0 t −1 1 t
(g)
1

x(τ ) h(t − τ )
1
τ
−1 0
(c)
τ
0 1 t −1 t
h(t − τ )
(h)
1

x(t) ∗ h(t)

τ 1
t −1 t
(d)

t
0 1 2
(i)

Figure 3.6: Evaluation of the convolution integral. The (a) input signal x(τ ), (b) impulse response h(τ ), (c) time-
reversed impulse response h(−τ ), and (d) impulse response after time-reversal and time-shifting h(t − τ ). The func-
tions associated with the product in the convolution integral for (e) t < 0, (f) 0 ≤ t < 1, (g) 1 ≤ t < 2, and (h) t ≥ 2,
(i) The convolution result x(t) ∗ h(t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.8. INTERCONNECTION OF CONTINUOUS-TIME LTI SYSTEMS 61

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

Figure 3.7: Block diagram representation of continuous-time LTI system.

x(t) y(t) x(t) y(t)

h1 (t) h2 (t) ≡ h1 (t) ∗ h2 (t)

(a)

x(t) y(t) x(t) y(t)

h1 (t) h2 (t) ≡ h2 (t) h1 (t)

(b)

Figure 3.8: Series interconnection of continuous-time LTI systems.

3.8 Interconnection of Continuous-Time LTI Systems

Suppose that we have an LTI system with input x(t), output y(t), and impulse response h(t). We know that x(t) and
y(t) are related as y(t) = x(t) ∗ h(t). In other words, the system can be viewed as performing a convolution operation.
From the properties of convolution introduced earlier, we can derive a number of equivalences involving the impulse
responses of series- and parallel-interconnected systems.
Suppose that we have two LTI systems with impulse responses h1 (t) and h2 (t) that are connected in a series
configuration, as shown on the left-side of Figure 3.8(a). For convenience, let us define v(t) = x(t) ∗ h1 (t). Using the
associative property of convolution, we can simplify the expression for the output y(t) as follows:

y(t) = v(t) ∗ h2 (t)

= [x(t) ∗ h1 (t)] ∗ h2 (t)
= x(t) ∗ [h1 (t) ∗ h2 (t)].

In other words, we have the equivalence shown in Figure 3.8(a).

Suppose that we have two LTI systems with impulse responses h1 (t) and h2 (t) that are connected in a series
configuration, as shown on the left-side of Figure 3.8(b). Using the commutative property of convolution, we can
simplify the expression for the output y(t) as follows:

y(t) = x(t) ∗ h1 (t) ∗ h2 (t)

= x(t) ∗ h2 (t) ∗ h1 (t).

In other words, we have the equivalence shown in Figure 3.8(b).

Suppose that we have two LTI systems with impulse responses h1 (t) and h2 (t) that are connected in a parallel
configuration, as shown on the left-side of Figure 3.9. Using the distributive property of convolution, we can simplify
the expression for the output y(t) as follows:

y(t) = x(t) ∗ h1 (t) + x(t) ∗ h2 (t)

= x(t) ∗ [h1 (t) + h2 (t)].

Thus, we have the equivalence shown in Figure 3.9.

Example 3.7. Consider the system shown in Figure 3.10 with input x(t), output y(t). Find the impulse response h(t)
of the system. Use the properties of convolution in order to accomplish this.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

62 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(t) y(t)
h1 (t) +
x(t) y(t)
≡ h1 (t) + h2 (t)
h2 (t)

Figure 3.9: Parallel interconnection of continuous-time LTI systems.

x(t) v(t) y(t)

h1 (t) + h3 (t)

h2 (t)

Figure 3.10: System interconnection example.

Solution. From the diagram, we can write:

v(t) = x(t) + x(t) ∗ h1 (t) + x(t) ∗ h2 (t)

= x(t) ∗ δ (t) + x(t) ∗ h1 (t) + x(t) ∗ h2 (t)
= x(t) ∗ [δ (t) + h1 (t) + h2 (t)].

Similarly, we can write:

y(t) = v(t) ∗ h3 (t).

Substituting the expression for v(t) into the preceding equation we obtain:

y(t) = v(t) ∗ h3 (t)

= x(t) ∗ [δ (t) + h1 (t) + h2 (t)] ∗ h3 (t)
= x(t) ∗ [h3 (t) + h1 (t) ∗ h3 (t) + h2 (t) ∗ h3 (t)].

Therefore, the impulse response h(t) of the overall system is

h(t) = h3 (t) + h1 (t) ∗ h3 (t) + h2 (t) ∗ h3 (t).

3.9 Properties of Continuous-Time LTI Systems

In the previous chapter, we introduced a number of properties that might be possessed by a system (e.g., memory,
causality, stability, invertibility). Since a LTI system is completely characterized by its impulse response, one might
wonder if there is a relationship between some of the properties introduced previously and the impulse response. In
what follows, we explore some of these relationships.

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.9. PROPERTIES OF CONTINUOUS-TIME LTI SYSTEMS 63

3.9.1 Memory
Recall that a system is memoryless if its output y(t) at any arbitrary time t0 depends only on the value of its input x(t)
at that same time. Suppose now that we have a LTI system with input x(t), output y(t), and impulse response h(t).
The output y(t) at some arbitrary time t = t0 is given by
y(t0 ) = x(t) ∗ h(t)|t=t0
= h(t) ∗ x(t)|t=t0
Z ∞
= h(τ )x(t − τ )d τ
−∞ t=t0
Z ∞
= h(τ )x(t0 − τ )d τ .
−∞
Consider the integral in the above equation. In order for the system to be memoryless, the result of the integration
must depend only on x(t) at t = t0 . This, however, is only possible if
h(t) = 0 for all t 6= 0. (3.32)
Consequently, h(t) must be of the form
h(t) = K δ (t) (3.33)
where K is a complex constant. Thus, we have that a LTI system is memoryless if and only if its impulse response
satisfies (3.32). This, in turn, implies that the impulse response is of the form of (3.33). As a consequence of this fact,
we also have that all memoryless LTI systems must have an input-output relation of the form
y(t) = x(t) ∗ K δ (t)
= Kx(t).
Example 3.8. Suppose that we have the LTI system with the impulse response h(t) given by
h(t) = e−at u(t)
where a is a real constant. Determine whether this system has memory.
Solution. The system has memory since h(t) 6= 0 for some t 6= 0 (e.g., h(1) = e−a 6= 0).
Example 3.9. Suppose that we have the LTI system with the impulse response h(t) given by
h(t) = δ (t).
Determine whether this system has memory.
Solution. Clearly, h(t) is only nonzero at t = 0. This follows immediately from the definition of the unit-impulse
function δ (t). Therefore, the system is memoryless (i.e., does not have memory).

3.9.2 Causality
Recall that a system is causal if its output y(t) at any arbitrary time t0 depends only on its input x(t) for t ≤ t0 . Suppose
that we have the LTI system with input x(t), output y(t), and impulse response h(t). The value of the output y(t) for
t = t0 is given by
y(t0 ) = [x(t) ∗ h(t)]|t=t0
Z ∞
= x(τ )h(t − τ )d τ
−∞ t=t0
Z ∞
= x(τ )h(t0 − τ )d τ
−∞
Z t0 Z ∞
= x(τ )h(t0 − τ )d τ + x(τ )h(t0 − τ )d τ . (3.34)
−∞ t0+

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

64 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

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

Figure 3.11: System.

In order for the expression for y(t0 ) in (3.34) not to depend on x(t) for t > t0 , we must have that

h(t) = 0 for t < 0 (3.35)

(i.e., h(t) is a causal signal). In this case, (3.34) simplifies to

Z t0
y(t0 ) = x(τ )h(t0 − τ )d τ .
−∞

Clearly, the result of this integration does not depend on x(t) for t > t0 (since τ varies from −∞ to t0 ). Therefore, a
LTI system is causal if its impulse response h(t) satisfies (3.35).

Example 3.10. Suppose that we have the LTI system with impulse response h(t) given by

h(t) = e−at u(t),

where a is a real constant. Determine whether this system is causal.

Solution. Clearly, h(t) = 0 for t < 0 (due to the u(t) factor in the expression for h(t)). Therefore, the system is
causal.

Example 3.11. Suppose that we have the LTI system with impulse response h(t) given by

h(t) = δ (t + t0 ),

where t0 is a strictly positive real constant. Determine whether this system is causal.

Solution. From the definition of δ (t), we can easily deduce that h(t) = 0 except at t = −t0 . Since −t0 < 0, the system
is not causal.

3.9.3 Invertibility
Recall that a system is invertible if we can always uniquely determine its input x(t) from its output y(t). An equivalent
way of stating this is that an inverse system must exist.
Suppose now that we have a LTI system with input x(t), output y(t), and impulse response h(t). Such a system is
illustrated in Figure 3.11.
One can readily show that the inverse of a LTI system, if it exists, must also be LTI. This follows from the fact that
the inverse of a linear system must be linear, and the inverse of a time-invariant system must be time invariant. Since
the inverse system, if it exists, must be LTI, the system can be completely characterized by its impulse response. Let
us denote the impulse response of the inverse system as hinv (t). Now, we want to find the relationship between h(t)
and hinv (t).
If the inverse system exists, then by definition, it must be such that

x(t) ∗ h(t) ∗ hinv (t) = x(t). (3.36)

This relationship is expressed diagrammatically in Figure 3.12. Since the unit-impulse function is the convolutional
identity, we can equivalently rewrite (3.36) as

x(t) ∗ h(t) ∗ hinv (t) = x(t) ∗ δ (t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.9. PROPERTIES OF CONTINUOUS-TIME LTI SYSTEMS 65

x(t) y(t) x(t)

h(t) hinv (t)

Figure 3.12: System in cascade with its inverse.

By comparing the left- and right-hand sides of the preceding equation, we have

h(t) ∗ hinv (t) = δ (t). (3.37)

Therefore, a system will have an inverse if and only if a solution for hinv (t) exists in (3.37). Thus, a system is invertible
if and only if a solution to this equation exists.

Example 3.12. Suppose that we have the LTI system with impulse response h(t) given by

h(t) = Aδ (t − t0 ) (3.38)

where A is a nonzero real constant and t0 is a real constant. Determine whether this system is invertible. If it is
invertible, find the impulse response hinv (t) of the inverse system.

Solution. The inverse system, if it exists, is given by the solution to the equation

h(t) ∗ hinv (t) = δ (t).

So, let us attempt to solve this equation for hinv (t). Substituting (3.38) into (3.37) and using straightforward algebraic
manipulation, we can write:

h(t) ∗ hinv (t) = δ (t)

Z ∞
⇒ h(τ )hinv (t − τ )d τ = δ (t)
−∞
Z ∞
⇒ Aδ (τ − t0 )hinv (t − τ )d τ = δ (t)
−∞
Z ∞
⇒ δ (τ − t0 )hinv (t − τ )d τ = A1 δ (t).
−∞

Using the sifting property of the unit-impulse function, we can simplify the integral expression in the preceding
equation to obtain

hinv (t − t0 ) = A1 δ (t). (3.39)

Now, we perform a change of variable. Let τ = t − t0 so that t = τ + t0 . Using this change of variable, we can
rewrite (3.39) as

hinv (τ ) = A1 δ (τ + t0 ).

Since A 6= 0, the function hinv (t) is always well defined. Therefore, the inverse system exists and has the impulse
response hinv (t) given by

hinv (t) = A1 δ (t + t0 ).

Example 3.13. Consider the system shown in Figure 3.13 with the input x(t) and output y(t). Use the notion of an
inverse system in order to express y(t) in terms of x(t).

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

66 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

x(t) v(t) y(t)

+ h1 (t)
−

h2 (t)

Figure 3.13: Feedback system.

Solution. From Figure 3.13, we can write:

v(t) = x(t) − y(t) ∗ h2 (t) and (3.40)

y(t) = v(t) ∗ h1 (t). (3.41)

Substituting (3.40) into (3.41), and simplifying we obtain:

y(t) = [x(t) − y(t) ∗ h2 (t)] ∗ h1 (t)

⇒ y(t) = x(t) ∗ h1 (t) − y(t) ∗ h2 (t) ∗ h1 (t)
⇒ y(t) + y(t) ∗ h2 (t) ∗ h1 (t) = x(t) ∗ h1 (t)
⇒ y(t) ∗ δ (t) + y(t) ∗ h2 (t) ∗ h1 (t) = x(t) ∗ h1 (t)
⇒ y(t) ∗ [δ (t) + h2 (t) ∗ h1 (t)] = x(t) ∗ h1 (t). (3.42)

For convenience, we now define the function g(t) as

g(t) = δ (t) + h2 (t) ∗ h1 (t). (3.43)

So, we can rewrite (3.42) as

y(t) ∗ g(t) = x(t) ∗ h1 (t). (3.44)

Thus, we have almost solved for y(t) in terms of x(t). To complete the solution, we need to eliminate g(t) from the
left-hand side of the equation. To do this, we use the notion of an inverse system. Consider the inverse of the system
with impulse response g(t). This inverse system has an impulse response ginv (t) given by

g(t) ∗ ginv (t) = δ (t). (3.45)

This relationship follows from the definition of an inverse system (associated with (3.37)). Now, we use ginv (t) in
order to simplify (3.44) as follows:

y(t) ∗ g(t) = x(t) ∗ h1 (t)

⇒ y(t) ∗ g(t) ∗ ginv (t) = x(t) ∗ h1 (t) ∗ ginv (t)
⇒ y(t) ∗ δ (t) = x(t) ∗ h1 (t) ∗ ginv (t)
⇒ y(t) = x(t) ∗ h1 (t) ∗ ginv (t).

Thus, we can express the output y(t) in terms of the input x(t) as

y(t) = x(t) ∗ h1 (t) ∗ ginv (t),

where ginv (t) is given by (3.45) and g(t) is given by (3.43).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.9. PROPERTIES OF CONTINUOUS-TIME LTI SYSTEMS 67

3.9.4 Stability
Recall that a system is BIBO stable if any arbitrary bounded input produces a bounded output. Suppose that we have
a LTI system with input x(t), output y(t), and impulse response h(t). Further, suppose that |x(t)| ≤ A < ∞ for all t
(i.e., x(t) is bounded). We can write

y(t) = x(t) ∗ h(t)

= h(t) ∗ x(t)
Z ∞
= h(τ )x(t − τ )d τ .
−∞

So, we have (by taking the magnitude of both sides of the preceding equation)
Z ∞
|y(t)| = h(τ )x(t − τ )d τ . (3.46)
−∞

One can show, for any two functions f1 (t) and f2 (t), that
Z ∞ Z ∞
f1 (t) f2 (t)dt ≤ | f1 (t) f2 (t)| dt.
−∞ −∞

Using this inequality, we can rewrite (3.46) as

Z ∞ Z ∞
|y(t)| ≤ |h(τ )x(t − τ )| d τ = |h(τ )| |x(t − τ )| d τ .
−∞ −∞

We know (by assumption) that |x(t)| ≤ A, so we can replace |x(t)| by its bound A in the above inequality to obtain
Z ∞ Z ∞ Z ∞
|y(t)| ≤ |h(τ )| |x(t − τ )| d τ ≤ A |h(τ )| d τ = A |h(τ )| d τ . (3.47)
−∞ −∞ −∞

Thus, we have Z ∞
|y(t)| ≤ A |h(τ )| d τ . (3.48)
−∞
Since A is finite, we can deduce from (3.48) that y(t) is bounded if
Z ∞
|h(t)| dt < ∞. (3.49)
−∞

(i.e., h(t) is absolutely integrable). Therefore, a LTI system is BIBO stable if its impulse response h(t) is absolutely
integrable (as in (3.49)). In other words, the absolute integrability of the impulse response h(t) is a sufficient condition
for BIBO stability.
As it turns out, condition (3.49) is also necessary for BIBO stability. Suppose that h(t) is not absolutely integrable.
That is, suppose that
Z ∞
|h(t)| dt = ∞.
−∞

If such is the case, we can show that the system is not BIBO stable. To begin, consider the particular input x(t) given
by the following:

x(t) = sgn(h(−t)),

where

−1 for α < 0

sgn α = 0 for α = 0


1 for α > 0.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

68 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

The signal x(t) can only assume the values −1, 0, and 1. So, x(t) is obviously bounded (i.e., |x(t)| ≤ 1). The output
y(t) is given by

y(t) = x(t) ∗ h(t)

Z ∞
= x(τ )h(t − τ )d τ
−∞
Z ∞
= (sgn h(−τ ))h(t − τ )d τ . (3.50)
−∞

Now, let us consider the output y(t) at t = 0. From (3.50), we have

Z ∞
y(0) = (sgn h(−τ ))h(−τ )d τ . (3.51)
−∞

Then, we observe that

α sgn α = |α | for any real α .

So (sgn h(−τ ))h(−τ ) = |h(−τ )|, and we can simplify (3.51) to obtain
Z ∞
y(0) = |h(−τ )| d τ
−∞
Z ∞
= |h(τ )| d τ
−∞
= ∞.

Thus, we have shown that the bounded input x(t) will result in an unbounded output y(t) (where y(t) is unbounded
at t = 0). Consequently, if the impulse response h(t) is not absolutely integrable (i.e., does not satisfy (3.49)), the
system is not BIBO stable. Consequently, the condition (3.49) is not only sufficient but also necessary for a system
to be BIBO stable. In other words, we have that a LTI system is BIBO stable if and only if its impulse response h(t)
satisfies
Z ∞
|h(t)| dt < ∞
−∞

(i.e., h(t) is absolutely integrable).

Example 3.14. Suppose that we have the LTI system with impulse response h(t) given by

h(t) = eat u(t),

where a is a real constant. Determine for what values of the constant a the system is BIBO stable.

Solution. We need to determine for what values of a the impulse response h(t) is absolutely integrable. Suppose that
a 6= 0. We can write
Z ∞ Z ∞
|h(t)| dt = eat u(t) dt
−∞ −∞
Z ∞
= eat dt
0
Z ∞
= eat dt
0
= [ 1a eat ]|∞
0
= a1 (ea∞ − 1) .

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.9. PROPERTIES OF CONTINUOUS-TIME LTI SYSTEMS 69

We can see that the result of the above integration is finite if a < 0 and infinite if a > 0. In particular, if a < 0, we have
Z ∞
|h(t)| dt = 0 − 1a
−∞
= − 1a .

Let us now consider the case of a = 0. In this case, we have

Z ∞ Z ∞
|h(t)| dt = |u(t)| dt
−∞ −∞
Z ∞
= dt
0
= [t]|∞
0
= ∞.

Thus, we have shown that

(
− a1
Z ∞
for a < 0
|h(t)| dt =
−∞ ∞ for a ≥ 0.

In other words, the impulse response h(t) is absolutely integrable if a < 0. Consequently, the system is BIBO stable
if a < 0.

Example 3.15. Suppose that we have the LTI system with input x(t) and output y(t) defined by
Z t
y(t) = x(τ )d τ
−∞

(i.e., an ideal integrator). Determine whether this system is BIBO stable.

Solution. First, we find the impulse response h(t) of the system.

Z t
h(t) = δ (τ )d τ
−∞
(
1 if t > 0
=
0 if t < 0
= u(t).

Using this expression for h(t), we now check to see if h(t) is absolutely integrable. We have
Z ∞ Z ∞
|h(t)| dt = |u(t)| dt
−∞ −∞
Z ∞
= dt
0
= ∞.

So, we have shown that h(t) is not absolutely integrable. Therefore, the system is not BIBO stable.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

70 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

3.10 Response of Continuous-Time LTI Systems to Complex Exponential

Signals
Suppose that we have a LTI system with input x(t), output y(t), and impulse response h(t). Consider now the response
of the system to the complex exponential x(t) = est . We can write:

y(t) = x(t) ∗ h(t)

= h(t) ∗ x(t)
Z ∞
= h(τ )x(t − τ )d τ
−∞
Z ∞
= h(τ )es(t−τ ) d τ
−∞
Z ∞
= est h(τ )e−sτ d τ
−∞
st
= H(s)e

where H(s) is the complex function given by

Z ∞
H(s) = h(τ )e−sτ d τ . (3.52)
−∞

(The function H(s) depends only on the complex quantity s.) From above, we can see that the response of a LTI
system to a complex exponential is the same complex exponential multiplied by the complex factor H(s). Thus, we
have that est and H(s) satisfy an eigenvalue problem of the form

H {Φ(t)} = λ Φ(t)

where Φ(t) = est and λ = H(s). As a matter of terminology, Φ(t) is said to be an eigenfunction and λ its correspond-
ing eigenvalue. Therefore, est is an eigenfunction of a LTI system and H(s) is the corresponding eigenvalue.
Suppose now that we can express some arbitrary input signal x(t) as a sum of complex exponentials as follows:

x(t) = ∑ ak esk t .
k

From the eigenvalue property, the response to the input ak esk t is ak H(sk )esk t . By using this knowledge and the super-
position property, we can write

y(t) = H {x(t)}
( )
sk t
=H ∑ ak e
k
= ∑ ak H {esk t }
k
= ∑ ak H(sk )esk t .
k

Thus, if an input to a LTI system can be represented as a linear combination of complex exponentials, the output can
also be represented as linear combination of the same complex exponentials. As a matter of terminology, we refer to
H(s) as the system function.
Example 3.16. Suppose that we have the LTI system with the impulse response h(t) given by

h(t) = δ (t − 1). (3.53)

Find the system function H(s).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.10. RESPONSE OF CONTINUOUS-TIME LTI SYSTEMS TO COMPLEX EXPONENTIAL SIGNALS 71

Solution. Substituting (3.53) into (3.52), we obtain

Z ∞
H(s) = h(t)e−st dt
−∞
Z ∞
= δ (t − 1)e−st dt
−∞
−s
=e .

Therefore, the system function H(s) is given by H(s) = e−s .

Example 3.17. Suppose that we have the LTI system from Example 3.16. Use the system function to determine the
response y(t) of the system to the particular input x(t) given by

x(t) = et cos(π t) = 21 e(1+ jπ )t + 21 e(1− jπ )t .

Solution. The input x(t) has already been expressed in the form
1
x(t) = ∑ ak eskt
k=0

where a0 = a1 = 21 , s0 = 1 + jπ , and s1 = s∗0 = 1 − jπ . In Example 3.16, we found the system function H(s) to be
H(s) = e−s . So we can calculate y(t) by using the system function as follows:

y(t) = ∑ ak H(sk )esk t

k
= a0 H(s0 )es0 t + a1 H(s1 )es1 t
= 12 H(1 + jπ )e(1+ jπ )t + 21 H(1 − jπ )e(1− jπ )t
= 21 e−(1+ jπ ) e(1+ jπ )t + 12 e−(1− jπ ) e(1− jπ )t
= 21 et−1+ jπ t− jπ + 12 et−1− jπ t+ jπ
= 21 et−1 e jπ (t−1) + 21 et−1 e− jπ (t−1)
h  i
= et−1 21 e jπ (t−1) + e− jπ (t−1)
= et−1 cos π (t − 1).

In passing, we note that we should expect the above result for y(t), since the output of the system is nothing more than
the input time-shifted by one.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

72 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

3.11 Problems

3.1 Using graphical methods, for each pair of signals x(t) and h(t) given in the figures below, compute the convo-
lution y(t) = x(t) ∗ h(t).

x(t) h(t)
2 2

1
1 1

t t
−1 0 1 2 3 −1 0 1 2 3
(a)

h(t)
2
x(t)
1 1

− 21
t 1 t
−1 1 −1 2 1

−1
−1

−2

(b)

x(t) h(t)

2 1

t
−1 1
t −1
0 1
(c)

h(t)
x(t) 2

1
1
t
0 1 2 3
t
−1 0
(d)

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.11. PROBLEMS 73

h(t)

x(t)
1

1
t
−1 1
t
0 1 2
−1

(e)

x(t) h(t)
5 5
4 4
3 3
2 2
1 1
t t
−3 −2 −1 0 1 2 −2 −1 0
(f)

3.2 For each pair of signals x(t) and h(t) given below, compute the convolution y(t) = x(t) ∗ h(t).
(a) x(t) = eat u(t) and h(t) = e−at u(t) where a is a nonzero real constant;
(b) x(t) = e− jω0 t u(t) and h(t) = e jω0 t u(t) where ω0 is a strictly positive real constant;
(c) x(t) = u(t − 2) and h(t) = u(t + 3);
(d) x(t) = u(t) and h(t) = e−2t u(t − 1);
(e) x(t) = u(t − 1) − u(t − 2) and h(t) = et u(−t).

3.3 Let y(t) = x(t) ∗ h(t). Given that

Z ∞
v(t) = x(−τ − b)h(τ + at)d τ ,
−∞

where a and b are constants, express v(t) in terms of y(t).

3.4 Consider the convolution y(t) = x(t) ∗ h(t). Assuming that the convolution y(t) exists, prove that each of the
following assertions is true:
(a) If x(t) is periodic then y(t) is periodic.
(b) If x(t) is even and h(t) is odd, then y(t) is odd.
d d
3.5 From the definition of the convolution operation, show that if y(t) = x(t) ∗ h(t), then dt y(t) = x(t) ∗ [ dt h(t)].

3.6 Let x(t) and h(t) be signals satisfying

x(t) = 0 for t < A1 or t > A2 , and

h(t) = 0 for t < B1 or t > B2

(i.e., x(t) and h(t) are finite duration). Determine for which values of t the convolution y(t) = x(t) ∗ h(t) must
be zero.

3.7 Find the impulse response of the LTI system characterized by each of the equations below. In each case, the
input and output of the system are denoted as x(t) and y(t), respectively.
Z t+1
(a) y(t) = x(τ )d τ ;
−∞

Version: 2013-09-11 Copyright c 2013 Michael D. Adams

74 CHAPTER 3. CONTINUOUS-TIME LINEAR TIME-INVARIANT SYSTEMS

Z ∞
(b) y(t) = x(τ + 5)eτ −t+1 u(t − τ − 2)d τ ;
Z −∞
t
(c) y(t) = x(τ )v(t − τ )d τ and
Z−∞
t
(d) y(t) = x(τ )d τ .
t−1

3.8 Consider the system with input x(t) and output y(t) as shown in the figure below. Suppose that the systems H1 ,
H2 , and H3 are LTI systems with impulse responses h1 (t), h2 (t), and h3 (t), respectively.

x(t) y(t)
H1 H2 +

H3

(a) Find the impulse response h(t) of the overall system in terms of h1 (t), h2 (t), and h3 (t).
(b) Determine the impulse response h(t) in the specific case that

h1 (t) = δ (t + 1), h2 (t) = δ (t), and h3 (t) = δ (t).

3.9 Consider a LTI system whose response to the signal x1 (t) = u(t) − u(t − 1) is the signal y1 (t). Determine the
response y2 (t) of the system to the input x2 (t) shown in the figure below in terms of y1 (t).

x2 (t)

t
−3 −2 −1 1 2 3

−1

−2

−3

3.10 Suppose that we have the system shown in the figure below with input x(t) and output y(t). This system is
formed by the interconnection of two LTI systems with the impulse responses h1 (t) and h2 (t).

x(t) y(t)
h1 (t) h2 (t)

For each pair of h1 (t) and h2 (t) given below, find the output y(t) if the input x(t) = u(t).
(a) h1 (t) = δ (t) and h2 (t) = δ (t);
(b) h1 (t) = δ (t + 1) and h2 (t) = δ (t + 1);
(c) h1 (t) = e−3t u(t) and h2 (t) = δ (t).

Copyright c 2013 Michael D. Adams Version: 2013-09-11

3.11. PROBLEMS 75

3.11 Show that a linear system is invertible if and only if the only input x(t) that produces the output y(t) = 0 for all
t is x(t) = 0 for all t.

3.12 Consider the LTI systems with the impulse responses given below. Determine whether each of these systems is
causal and/or memoryless.
(a) h(t) = (t + 1)u(t − 1);
(b) h(t) = 2δ (t + 1);
(c) h(t) = ωπc sinc ωct;
(d) h(t) = e−4t u(t − 1);
(e) h(t) = et u(−1 − t);
(f) h(t) = e−3|t| ; and
(g) h(t) = 3δ (t).

3.13 Consider the LTI systems with the impulse responses given below. Determine whether each of these systems is
BIBO stable.
(a) h(t) = eat u(−t) where a is a strictly positive real constant;
(b) h(t) = (1/t)u(t − 1);
(c) h(t) = et u(t);
(d) h(t) = δ (t − 10);
(e) h(t) = rectt; and
(f) h(t) = e−|t| .

3.14 Suppose that we have two LTI systems with impulse responses

h1 (t) = 21 δ (t − 1) and h2 (t) = 2δ (t + 1).

Determine whether these systems are inverses of one another.

3.15 Consider the system shown in the figure below, where H is a LTI system and G is known to be the inverse sys-
tem of H . Let y1 (t) and y2 (t) denote the responses of the system H to the inputs x1 (t) and x2 (t), respectively.

x(t) y(t) x̂(t)

H G

(a) Determine the response of the system G to the input a1 y1 (t)+a2 y2 (t) where a1 and a2 are complex constants.
(b) Determine the response of the system G to the input y1 (t − t0 ) where t0 is a real constant.
(c) Using the results of the previous parts of this question, determine whether the system G is linear and/or time
invariant.

3.16 Suppose that we have the systems H1 , H2 , H3 , and H4 , whose responses to a complex exponential input e j2t
are given by
H
e j2t −→
1
2e j2t ,
H
e j2t −→
2
te j2t ,
H
e j2t −→
3
e j2t+π /3 , and
H
e j2t −→
4
cos 2t.

Indicate which of these systems cannot be LTI.

Version: 2013-09-11 Copyright c 2013 Michael D. Adams