Chapter 2.

Linear Time-Invariant Systems

2.1 Introduction
Two most important attributes of systems are linearity and time-invariance. An important subsets of
continuous-time and discrete-time systems are systems that satisfy the linearity and time-invariance
properties. Such systems are referred to as linear, time-invariant (LTI) systems. It will be shown that
the input-output relationship for LTI systems is described in terms of a convolution operation.

The importance of the convolution operation in LTI systems stems from the fact that knowledge of the
response of an LTI system to an unit impulse input allows us to find its output to any input signals. In
the time domain, continuous-time LTI systems can be analyzed using the convolution integral or a
linear constant-coefficient differential equation. Where as, discrete-time LTI systems can analyzed
using the convolution sum or a linear constant-coefficient difference equation.

2.2 Response of Continuous-Time LTI System and The Convolution Integral

2.2.1 Impulse Response

The impulse response of an LTI system is the response of the system to an impulse input signal. The
impulse response of a continuous-time LTI system, denoted by h(t), is defined as:

δ(t) → h(t) ⇒ h(t) = T [δ(t)]

If the system is time-invariant, then we have:

h(t) = T [δ(t)] = T [δ(t − τ )]

2.2.2 The Convolution Integral

The output y(t) of a continuous-time LTI system with impulse response h(t) and input x(t) can be
obtained using the convolution integral.

The convolution integral is defined as:

∞
y(t) = x(t) ∗ h(t) = ∫ ​
x(τ )h(t − τ )dτ
−∞
Given input x(t) and impulse response h(t) of a continuous-time LTI system, the convolution integral
can be evaluated graphically by following steps.

1. Sketch the waveform for input x(t) by changing the independent variable from t to τ and keep
the waveform for x(τ ) fixed during convolution.
2. Sketch the waveform for the impulse response h(t) by changing the independent variable from t
to τ .
3. Reflect h(τ ) about the vertical axis to obtain the time-reflected impulse response h(−τ ).
4. Shift the time-reflected impulse function h(−τ ) by a selected value of t. The resulting function
represents h(t − τ ).
5. Multiply function x(τ ) by h(t − τ ) and plot the product function x(τ )h(t − τ ).
6. Calculate the total area under the product function x(τ )h(t − τ ) by integrating it over τ =
[−∞, ∞].
7. Repeat steps 4 to 6 for different values of t to obtain y(t) for all time, −∞ ≤ t ≤ ∞.

The convolution integral has the following properties.

i. Commutative property

x(t) ∗ h(t) = h(t) ∗ x(t)

ii. Associative property

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

​ ​ ​ ​

iii. Distributive property

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

​ ​ ​ ​

Example
Graphically determine the continuous-time convolution of h(t) and x(t) for the following:
 x(t) = {
1, 0≤t≤4
0, otherwise
​ ​

and

h(t) = {
1, 0≤t≤4
0, otherwise
​ ​

Solution

To compute y(t) = x(t) ∗ h(t), first h(τ ) is to be obtained by inverting h(τ ) about the vertical axis.
Then, the product of x(τ ) and h(t − τ ) is formed, point by point, and this product is integrated to
compute y(t). Thus, the overlap area between the rectangles forming x(τ ) and h(t − τ ) is y(t).
Clearly, y(0) = 0 because there is no overlap between the rectangles forming x(τ ) and h(t − τ ) at
t = 0. For 0 < t < 8, there is overlap between the rectangles forming x(τ ) and h(t − τ ). For t ≥
8 there is no overlap, and hence, y(8) = 0.

These are illustrated below with the final result for y(t). The shaded portion represents the overlap
area of the product x(τ ) and h(t − τ ).
2.3 Properties of Continuous Time LTI Systems
2.3.1 Memoryless LTI Systems
The output of a memoryless system depends only on the present input. Applying the commutative
property of convolution, the output of a continuous-time LTI system may be expressed as:

y(t) = h(t) ∗ x(t)

∞
=∫ ​ h(τ )x(t − τ )dτ
−∞

For this system to be memoryless, y(t) must depend only on x(t) and cannot depend on x(t − τ )
for τ  0. This condition implies that h(τ ) = 0 for τ =
=  0. Thus, a continuous-time LTI system is
memoryless if and only if:

h(t) = Aδ(t) where: A is an arbitrary constant

All memoryless continuous-time LTI systems perform scalar multiplication on the input.

2.3.2 Causal LTI Systems

The output of a causal LTI system depends only on past or present values of the input. Again, write the
convolution integral as:

y(t) = h(t) ∗ x(t)

∞
=∫ ​
h(τ )x(t − τ )dτ
−∞

Past and present values of the input, x(t), x(t − 1), x(t − 2), …, are associated with indices τ ≥
0 in the convolution integral. Future values of the input are associated with indices τ < 0. In order for
y(t) to depend only on past or present values of the input, we require h(τ ) = 0 for τ < 0.

Thus, a continuous-time LTI system is causal if and only if:

h(t) = 0 for t<0

For a causal continuous-time LTI system, the convolution integral can be expressed as:
∞
y(t) = ∫ ​ h(τ )x(t − τ )dτ
0
2.3.3 Stable LTI Systems
A system is bounded input-bounded output (BIBO) stable if the output is guaranteed to be bounded
for every bounded input. Consider the convolution integral of a continuous-time LTI system:
∞
y(t) = h(t) ∗ x(t) = ∫ ​ h(τ )x(t − τ )dτ
−∞

Taking the magnitude of both sides, we get:

∞
∣y(t)∣ = ∣h(t) ∗ x(t)∣ = ∫ ​ ​ h(τ )x(t − τ )dτ ​

−∞

Using the Schwartz inequality, the above equation reduces to:

∞
∣y(t)∣ ≤ ∫ ​
∣h(τ )∣ ∣x(t − τ )∣ dτ
−∞

Assume that the input is bounded, i.e.,

∣x(t)∣ ≤ Bx < ∞ ​

Thus, the above inequality reduces to:

∞
∣y(t)∣ ≤ Bx ∫ ​ ​ ∣h(τ )∣dτ
−∞

This implies that the output is bounded, ∣y(t)∣ < ∞, provided that the impulse response is absolutely
integrable.
Thus, a continuous-time LTI system is BIBO stable if and only if:
∞
∫ ​ ∣h(τ )∣dτ < ∞
−∞

2.4 Step Response of LTI Systems

The step response of LTI systems can be easily expressed in terms of the impulse response as
follows. Consider a continuous-time LTI system with impulse response h(t) and step response
denoted by s(t).

s(t) = h(t) ∗ u(t)

 ∞
=∫ ​ h(τ )u(t − τ )dτ
−∞

But, we know that

u(t − τ ) = {
0 for τ > t
1 for τ ≤ t
​ ​

Thus, the step response can be expressed as:

t
s(t) = ∫ ​ h(τ )dτ
−∞

That is, the step response of a continuous-time LTI system is expressed as the running integral of the
impulse response. We can also express the impulse response of a continuous-time LTI system in
terms of the step response as follows:

d
h(t) = s(t) ​

dt

Exercise 2(a)
Given:
Impulse response:

h(t) = e−4t u(t − 2)

determine whether the LTI system is memoryless, causal, and BIBO stable.

1. Memoryless?

From Section 2.3.1, a continuous-time LTI system is memoryless iff

h(t) = Aδ(t)

for some constant A.

Here,

h(t) = e−4t u(t − 2)

is not a scaled impulse—it is a delayed exponential starting at t = 2.

Not memoryless

2. Causal?

From Section 2.3.2, an LTI system is causal iff

h(t) = 0 for all t < 0

But note:
u(t − 2) = 0 for t < 2, and nonzero for t ≥ 2.
So h(t) = 0 for t < 2, which includes all t < 0.
Thus, h(t) = 0 for t < 0 → Causal

Important: Causality only requires no response before t = 0. A delayed response (e.g., starting
at t = 2) is still causal.

3. BIBO Stable?

From Section 2.3.3, the system is BIBO stable iff

∞
∫ ​ ∣h(t)∣ dt < ∞
−∞

Compute:
∞ ∞
∫ ​ ∣e −4t
u(t − 2)∣ dt = ∫ ​ e−4t dt
−∞ 2

(since e−4t > 0 for all real t)

∞
−1 −4t −1 −8 1
=[ e ] =0−( e ) = e−8 < ∞
4 4 4
​ ​ ​ ​

BIBO stable

Exercise 3(b)
Given:
Impulse response:
 h(t) = u(t) − u(t − 4)

find the unit step response s(t).

From Section 2.4, the step response is:

t
s(t) = h(t) ∗ u(t) = ∫ ​ h(τ ) dτ
−∞

(because convolution with unit step = running integral of h(t))

So compute:
t
s(t) = ∫ [u(τ ) − u(τ − 4)] dτ
​

−∞

Note that:

u(τ ) − u(τ − 4) = 1 for 0 ≤ τ < 4, and 0 elsewhere.

So the integrand is a rectangular pulse from 0 to 4.

Now evaluate the integral as a function of t:

Case 1: t <0

No overlap with pulse → integral = 0

s(t) = 0

Case 2: 0 ≤t<4

Integrate from 0 to t:
t
s(t) = ∫ 1 dτ = t ​

Case 3: t ≥4

Integrate over entire pulse:

4
s(t) = ∫ ​ 1 dτ = 4
0

Final step response:

 ⎧0, t<0
s(t) = ⎨t, 0≤t<4
⎩
​ ​ ​

4, t≥4

Or using unit step functions:

s(t) = t[u(t) − u(t − 4)] + 4u(t − 4)

2.5 Differential and Difference Equations Representation

The general form of an N th order linear constant-coefficient differential equation is given by:

N M
dk dk
∑ ak k y(t) = ∑ bk k x(t)
​ ​ ​ ​ ​ ​

dt dt
k=0 k=0

N M
dk dk
⇒ y(t) + ∑ ak k y(t) = ∑ bk k x(t),
​ ​ ​ ​ ​ a0 = 1
​

dt dt
k=1 k=0

N M
dk dk
⇒ y(t) = − ∑ ak k y(t) + ∑ bk k x(t) ​ ​ ​ ​ ​ ​

dt dt
k=1 k=0

where: ak 's and bk 's are constant coefficients.

​ ​

The total solution of the above linear constant-coefficient differential equation can be expressed as a
sum of two parts:

y(t) = yh (t) + yp (t) ​ ​

yh (t): Homogeneous Solution

​

yp (t): Particular Solution

​

Homogeneous Solution
The homogeneous solution yh (t) is the solution of the differential equation with zero input.
​

The homogeneous differential equation with zero input is given by:

N
dk
∑ ak k y(t) = 0
​ ​ ​

dt
k=0
The solution to the above zero-input homogeneous equation is characterized by exponential
responses of the form:

yh (t) = Aert​

where A is an arbitrary constant.

Plugging in yh (t) into the zero-input equation, we get:

​

N
dk
∑ ak k (Aert ) = 0
​ ​ ​

dt
k=0

⇒ Aert (a0 + a1 r + … + aN −1 rN −1 + aN rN ) = 0
​ ​ ​ ​

⇒ aN rN + aN −1 rN −1 + … + a1 r + a0 = 0
​ ​ ​ ​

This equation is known as the characteristic equation of the linear constant coefficient differential
equation.
Because the characteristic equation is of degree N , it will have N roots which may be either real or
complex.
The N roots can also be distinct or repeated.

Case I: Distinct roots

If the N roots ri are distinct, ri
​ ​ = rk for k =
  i, the general solution to the homogeneous
​

differential equation is:

N
yh (t) = ∑ Ai eri t = A1 er1 t + A2 er2 t + A3 er3 t + … + AN erN t
​ ​ ​
​

​
​

​
​

​
​

​
​

i=1

where the constants Ai are chosen to satisfy the initial conditions.

​

Case II: Repeated roots

For repeated roots, the solution must be modified as follows:

If r1 is a repeated root of multiplicity m with the remaining N

​
− m roots distinct, the
homogeneous solution becomes:

N
yh (t) = (A1 + A2 t + … + Am t
​ ​ ​ ​
m−1
)e r1 t ​

+ ∑ Ai er i t ​
​

i=m+1

where the constants Ai are chosen to satisfy the initial conditions.

​
Particular Solution
The particular solution is the response of the system to the input x(t) assuming zero initial conditions.
Thus, the particular solution yp (t) depends on the form of the input x(t).
​

Because the input x(t) can take different forms, there is no single specific form for a particular
solution.

Steps to solve linear constant-coefficient differential equations:

i. Find the homogeneous solution, yh (t), by solving the roots of the characteristic equation.
​

ii. Find the particular solution, yp (t), by assuming that it is of the same form as the input signal x(t)
​

iii. Find the total solution by adding yh (t) and yp (t), i.e.,
​ ​

y(t) = yh (t) + yp (t)

​ ​

iv. Determine the constant coefficients in the homogeneous solution by plugging initial conditions in
the total solution.

Recall Maths
General Form of a First-Order LCCDE

A linear constant-coefficient differential equation (LCCDE) of first order looks like:

dy(t)
+ a y(t) = b x(t)
​

dt
a, b: real constants
x(t): input (forcing function)
y(t): output (system response)
Initial condition (IC): usually given as y(0) or y(t0 ) ​

Solution Strategy: Total Response = Homogeneous + Particular

the total solution is:

y(t) = yh (t) + yp (t)

​ ​

Where:
 yh (t): homogeneous (natural) solution → response due to system dynamics (with x(t) = 0)
​

yp (t): particular (forced) solution → steady-state response due to input x(t)

​

Then, apply initial conditions to solve for unknown constants.

🔹 Step 1: Homogeneous Solution yh (t) ​

Set input x(t) = 0:

dyh
+ ayh = 0
​

​ ​

dt
This is a separable ODE. Solution:

yh (t) = Ae−at
​

A: constant determined by IC
The exponent −a comes from the characteristic equation: r + a = 0 ⇒ r = −a

🔹 Step 2: Particular Solution yp (t) ​

Guess a form similar to the input x(t). Common cases:

Input x(t) Guess for yp (t)

​

Constant (e.g., u(t)) Constant K

est Kest (unless s = −a)

sin(ωt) or cos(ωt) K1 sin(ωt) + K2 cos(ωt)

​ ​

🔹 Step 3: Total Solution

y(t) = Ae−at + yp (t) ​

🔹 Step 4: Apply Initial Condition

Plug in t = 0 (or given time) into y(t) to solve for A.

Important: The IC applies to the total response, not just the homogeneous part.
Exercise 4(a)
Determine the output of the system described by the following differential equation with input and initial
condition as specified:

dy(t)
+ 4y(t) = 2x(t),
​ x(t) = u(t), y(0) = 1
dt
This is a first-order linear constant-coefficient differential equation (LCCDE).

Section 2.5 outlines a 4-step procedure:

1. Find the homogeneous solution yh (t) ​

2. Find the particular solution yp (t) ​

3. Form the total solution: y(t) = yh (t) + yp (t)

​ ​

4. Apply initial conditions to solve for unknown constants

Step 1: Homogeneous Solution yh (t) ​

Set the input x(t) = 0 (zero-input response):

dyh (t)
+ 4yh (t) = 0
​

​ ​

dt
dyh
+ 4yh = 0 ⇒ yh (t) = Ae−4t
​

​ ​ ​

dt
Step 2: Particular Solution yp (t) ​

Input: x(t) = u(t) → a unit step, which is constant for t ≥ 0

For a constant input, assume a constant particular solution: yp (t) ​ = K (for t ≥ 0)

dyp
Then dt = 0, so plug into the original DE:
​

1
0 + 4K = 2 ⋅ 1 (since x(t) = u(t) = 1 for t ≥ 0) ⇒ 4K = 2 ⇒ K =
2
​

So,

1
yp (t) = , t≥0
2
​ ​

( write yp (t) ​ = 12 u(t), but since we’re solving for t ≥ 0 with IC at t = 0, it's fine to treat it as
​

constant.)
Step 3: Total Solution

1
y(t) = yh (t) + yp (t) = Ae−4t + , t≥0
2
​ ​ ​

Step 4: Apply Initial Condition

Given: y(0) =1
1 1 1 1
y(0) = Ae0 + =A+ =1⇒A=1− =
2 2 2 2
​ ​ ​ ​

Answer

1 1
y(t) = ( e−4t + ) u(t)
2 2
​ ​ ​

Note: multiply by u(t) to emphasize that the solution is valid for t ≥ 0 (causal system with step
input). If the context assumes t ≥ 0 implicitly, you may write simply:
1
y(t) = (1 + e−4t ) , t≥0
2
​

2.6. Convolution Sum

The convolution sum is a fundamental operation in discrete-time Linear Time-Invariant (LTI) systems.
It describes how to calculate the output signal y[n] based on the input x[n] and the system’s impulse
response h[n].

The convolution of two discrete-time sequences x[n] and h[n] is defined as:
∞
y[n] = x[n] ∗ h[n] = ∑ x[k] h[n − k] ​ (Eq. 1)
k=−∞

Where:

∗ denotes the convolution operation.

x[k] is the input signal.
h[n − k] is the impulse response flipped and shifted.

The output y[n] of any discrete-time LTI system is the convolution of the input x[n] with the impulse
response h[n].
2.6.1. Properties of the Convolution Sum
The convolution sum has several key properties that make it easier to work with discrete-time
systems:

1. Commutative Property:
Convolution is commutative, meaning the order of signals does not matter:

x[n] ∗ h[n] = h[n] ∗ x[n] (Eq. 2)

2. Associative Property:
The grouping of convolutions does not affect the result:

(x[n] ∗ h1 [n]) ∗ h2 [n] = x[n] ∗ (h1 [n] ∗ h2 [n])

​ ​ ​ ​ (Eq. 3)

3. Distributive Property:
Convolution distributes over addition:

x[n] ∗ (h1 [n] + h2 [n]) = x[n] ∗ h1 [n] + x[n] ∗ h2 [n]

​ ​ ​ ​ (Eq. 4)

2.6.2. Convolution Sum Operation

Using the commutative property (Eq. 2), we can rewrite the convolution sum (Eq. 1) as:

∞
y[n] = ∑ h[k] x[n − k]
​ (Eq. 5)
k=−∞

This form is often easier to evaluate in practice. The process of performing convolution step-by-step is
as follows:

1. Time Reversal: Reverse the impulse response h[n] to obtain h[−k].

2. Shifting: Shift the reversed impulse response by n to get h[n − k].
3. Multiplication: Multiply x[k] and h[n − k] term by term for all k .
4. Summation: Sum all these products to get the final output y[n].

Example
Convolve the following two sequences x(n) and h(n) to find y(n):

x(n) = {1, 1, 1, 1}
h(n) = {2, 2}

We want to compute:
 ∞
y[n] = x[n] ∗ h[n] = ∑ x[k] h[n − k]
​ (Eq. 1)
k=−∞

step 1 Determine the Length of y[n]

The length of the convolution output is:

L = Lx + Lh − 1 = 4 + 2 − 1 = 5
​ ​

So, y[n] will have 5 samples: y[0], y[1], y[2], y[3], y[4].

Step 2: Compute Each Output Sample

We will use the convolution sum formula (Eq. 1) for each value of n:

y[0]
y[0] = x[0]h[0] = 1 × 2 = 2

y[1]
y[1] = x[0]h[1] + x[1]h[0] = (1)(2) + (1)(2) = 2 + 2 = 4

y[2]
y[2] = x[1]h[1] + x[2]h[0] = (1)(2) + (1)(2) = 2 + 2 = 4

y[3]
y[3] = x[2]h[1] + x[3]h[0] = (1)(2) + (1)(2) = 2 + 2 = 4

y[4]
y[4] = x[3]h[1] = (1)(2) = 2

Final Convolution Output

y[n] = {2, 4, 4, 4, 2} ​
 Each output value is computed as the sum of products between x[k] and a time-reversed,
shifted version of h[n].
Since h[n] = {2, 2}, every pair of overlapping values from x[n] contributes 2 + 2 = 4 to the
middle samples.
The convolution result is symmetric due to the uniform values in x[n] and h[n].

Exercise
1 Determine the output y(t) of the continuous-time LTI systems with the following pairs of input
signals and impulse responses

={ and h(t) = {
1, −2 ≤ t ≤ 2 1, −2 ≤ t ≤ 2
a. x(t)
0, otherwise 0, otherwise
​ ​ ​ ​

b. x(t) = u(t − 1) and h(t) = e−3t u(t)

2 Determine whether the LTI systems with the following impulse responses are memoryless, causal,
stable or not.

a. h(t) = e−4t u(t − 2)

b. h(t) = e2t u(3 − t)

c. h(t) = e−2t u(t + 10)

3 Determine the unit step response of the systems with the following impulse responses.

a. h(t) = e−∣t∣

b. h(t) = u(t) − u(t − 4)

c. h(t) = u(t + 1)

d. h(t) = e2t u(−t − 1)

4 Determine the output of the systems described by the following differential equations with input and
initial conditions as specified.

d
a. dt​y(t) + 4y(t) = 2x(t) and x(t) = u(t), y(0) = 1
d2
b. dt 2 y (t) + 6 dt y(t) + 8y(t)
​
d
​ = 2x(t) and x(t) = e−t u(t), y(0) = −1, y ′ (0) = 1