CHAPTER II

Time-Domain Representations of
LTI Systems
Lecture 2: Discrete-Time Systems

University of Engineering and Technology - VNU Hanoi

Signals and Systems 2010 1 / 19

Difference Equations of Discrete-Time LTI Systems Difference Equation Representation of Discrete-Time Systems

The model of a discrete-time system can be

obtained by discretizing its corresponding
continuous-time system.
The discrete version of a differential equation is
called difference equation.
Example: a continuous-time system is described by
the following equation

dy(t)/dt + ay(t) = bx(t)

dy(nT ) y(nT )−y(nT −T )
By using the approximation dt ≈ T ,
we obtain the following difference equation of the
discrete-time system with the sampling period of T :

(1 + aT )y(n) − y(n − 1) = bTx(n)

Signals and Systems 2010 2 / 19

Difference Equations of Discrete-Time LTI Systems Constant-coefficient linear difference equations

Discrete-time LTI systems can be represented

by constant-coefficient linear difference
equations.
General form of constant-coefficient linear
difference equations:
N
X M
X
ai y(n − i) = bj x(n − j)
i=0 j=0

in which, x(n) is the input signal and y(n) is the

output signal of a system.
By solving the above difference equation, the
output signal y(n) is determined when the input
signal x(n) has been known.
Signals and Systems 2010 3 / 19
Difference Equations of Discrete-Time LTI Systems Solutions of constant-coefficient linear difference equations

The general solution of a constant-coefficient

linear difference equation has the following form:

y(n) = y0 (n) + ys (n)

y0 (n): the initial response, which is the solution

of the following homogeneous equation:
N
X
ai y(n − i) = 0 (1)
i=0

ys (n): the zero-state response, which is the

particular solution of the equation for an input
signal x(n).
Signals and Systems 2010 4 / 19
Difference Equations of Discrete-Time LTI Systems Determining the initial response

y0 (n) is the response of the system to the

system conditions at the initial point (n = 0),
without input signal x(n).
The homogeneous equation (1) has a solution
in the form z n in which z is a complex variable,
replacing y(n) in the equation by it yields:
N
X
ai z N−i = 0 (2)
i=0

which is the characteristic equation of the

system.

Signals and Systems 2010 5 / 19

Difference Equations of Discrete-Time LTI Systems Determining the initial response

Let the roots of the equation (2) be

{zk |k = 1..N}, the general solution of the
homogeneous equation (1) then has the
following form if all {zk } are distinct roots:
N
X
y0 (n) = ck zkn
k=1

Values of the coefficients {ck } will be

determined from initial conditions.

Signals and Systems 2010 6 / 19

Difference Equations of Discrete-Time LTI Systems Determining the initial response

In case the equation (2) has repeated roots, the

general form of the homogeneous solution will
be:
pk −1
!
X X
y0 (n) = ck zkn ni
k i=0
in which each root zk is repeated pk times.

Signals and Systems 2010 7 / 19

Difference Equations of Discrete-Time LTI Systems Determining the zero-state response

ys (n) is the system response to the input signal

x(n) when all initial conditions are zeros.
ys (n) is also called the particular solution of the
constant-coefficient linear difference equation
representing the system.
To determine ys (n), it is usually assumed that
ys (n) has a form similar to the input signal x(n)
with some unknown coefficients which will be
identified later from the equation.

Signals and Systems 2010 8 / 19

Difference Equations of Discrete-Time LTI Systems Determining the zero-state response

Note when suggesting the form of ys (n): ys (n)

must be independent of all terms of the
homogeneous solution y0 (n).
For example, if x(n) = αn , we may have to
consider the following cases:
If αn is not a component of y0 (n), then we can
suggest that ys (n) is in the form of cαn .
If α is a distinct root of the characteristic equation (2)
→ αn is a component of y0 (n) → ys (n) is in the form
of cnαn .
If α is a repeated root with the order of p of the
characteristic equation (2) → αn , nαn ,...,np−1 αn are
components of y0 (n) → ys (n) has the form of cnp αn .

Signals and Systems 2010 9 / 19

 Representing Systems by Impulse Response Convolution of two discrete-time signals

The convolution of two discrete-time signals f (n)

and g(n), denoted f (n) ∗ g(n), is defined as:
+∞
X
f (n) ∗ g(n) = f (k)g(n − k)
k=−∞

Convolution sum evaluation: see Proakis,

Section 2.3 (pages 102-114).

Signals and Systems 2010 10 / 19

 Representing Systems by Impulse Response Properties of convolution

Commutativity:

f (n) ∗ g(n) = g(n) ∗ f (n)

Associativity:

[f (n) ∗ g(n)] ∗ h(n) = f (n) ∗ [g(n) ∗ h(n)]

Distributivity:

[f (n) + g(n)] ∗ h(n) = f (n) ∗ h(n) + g(n) ∗ h(n)

Signals and Systems 2010 11 / 19

 Representing Systems by Impulse Response Properties of convolution

Time shift: if x(n) = f (n) ∗ g(n), then

x(n − n0 ) = f (n − n0 ) ∗ g(n) = f (n) ∗ g(n − n0 )

Convolution of a signal with the unit impulse:

f (n) ∗ δ(n) = f (n)

Causality: if f (n) and g(n) are causal signals

then f (n) ∗ g(n) is also causal.

Signals and Systems 2010 12 / 19

 Representing Systems by Impulse Response Impulse response of a discrete-time LTI system

Given an LTI system y(n) = T[x(n)], we derive:

" +∞ #
X
y(n) = T[x(n) ∗ δ(n)] = T x(k)δ(n − k)
k=−∞
Z ∞
= x(τ )T[δ(t − τ )]dτ = x(t) ∗ h(t)
−∞

in which, h(n) = T[δ(n)] is called the impulse

response of the system represented by the
function T.

Signals and Systems 2010 13 / 19

 Representing Systems by Impulse Response Analyzing impulse response of a discrete-time LTI system

Memoryless: impulse response has only one

non-zero value at n = 0.
Causal systems: impulse response is a causal
signal.
Stable systems: if and only if the following
condition is satisfied
+∞
X
|h(n)| < ∞
k=−∞

Signals and Systems 2010 14 / 19

State-Variable Models for Discrete-Time Systems State equations of a discrete-time system

Let {u1 (n), u2 (n)...} be input signals,

{y1 (n), y2 (n)...} be output signals, and
{q1 (n), q2 (n)...} be state variables of a
discrete-time LTI system.
State equations of the system are then:
X X
qi (n + 1) = aij qj (n) + bik uk (n) (i = 1, 2, ...)
j k

Output signals are computed from state

variables and input signals as follows:
X X
yi (n) = cij qj (n) + dik uk (n) (i = 1, 2, ...)
j k

Signals and Systems 2010 15 / 19

State-Variable Models for Discrete-Time Systems State equations of a discrete-time system

State-variable model of a discrete-time LTI

system is often represented in the following
matrix form:

q(n + 1) = Aq(n) + Bu(n)

y(n) = Cq(n) + Du(n)

in which, u(n), y(n) and q(n) are column vectors

whose elements are respectively input signals,
output signals and state variables of the system;
A, B, C and D are matrices of coefficients.

Signals and Systems 2010 16 / 19

State-Variable Models for Discrete-Time Systems Establishing state equations

State-variable models of discrete-time systems

can be derived from state-variable models of
their continuous-time counterparts: Given the
following equations of a continuous-time system
dq(t)
= Aq(t) + Bu(t)
dt
y(t) = Cq(t) + Du(t)
Discretizing them with the sampling time of T
and approximating dq(nTdt
)
≈ q(nT +TT)−q(nT ) , we
obtain this discrete-time model:
q(n + 1) = (T A + I)q(n) + T Bu(n)
y(n) = Cq(n) + Du(n)
Signals and Systems 2010 17 / 19
State-Variable Models for Discrete-Time Systems Establishing state equations

State equations can also be formulated from the

difference equation of a discrete-time LTI
system:
N
X M
X
ai y(n − i) = bj x(n − j)
i=0 j=0

Denote uj (n) = x(n − j) (j = 0..M) input signals

of the system and rewrite the above equation in
the following form:
N
X M
X
ai y(n − i) = bj uj (n)
i=0 j=0

Signals and Systems 2010 18 / 19

State-Variable Models for Discrete-Time Systems Establishing state equations

Choose the following state variables:

q1 (n) = y(n − N), q2 (n) = y(n − N + 1), ...

qN (n) = y(n − 1)

We derive the following state equations:

q1 (n + 1) = q2 (n), q2 (n + 1) = q3 (n), ...

qN−1 (n + 1) = qN (n)

N M
1  X X
qN (n + 1) = − ai qN−i+1 (n) + bj uj (n)
a0
i=1 j=0

Signals and Systems 2010 19 / 19