Sunteți pe pagina 1din 22

SOLUTIONS TO DIFFERENCE EQUATIONS

Introduction
One of the most important concepts of DSP is to be able to properly represent the
input/output relationship to a given LTI system. A linear constant-coefficient difference equation
(LCCDE) serves as a way to express just this relationship in a discrete-time system. Writing the
sequence of inputs and outputs, which represent the characteristics of the LTI system, as a
difference equation help in understanding and manipulating a system.

A difference equation is an equation that shows the relationship between consecutive


values of a sequence and the differences among them. They are often rearranged as a recursive
formula so that a systems output can be computed from the input signal and past outputs.

A digitized version of a differential equation results in a difference equation. In continuous-


time the relationships between different signals are expressed by differential equations. However,
in discreet time, difference equations are the counterpart.
We now show that a digitized version of a differential equation results in a difference
equation. Let us consider a simple first-order differential equation
𝑑𝑑𝑑𝑑
+ 𝑐𝑐𝑐𝑐(𝑡𝑡) = 𝑥𝑥(𝑡𝑡) (1.1a)
𝑑𝑑𝑑𝑑
Consider uniform samples of 𝑥𝑥(𝑡𝑡) at intervals of 𝑇𝑇 seconds. As usual, we use the notation
𝑥𝑥[𝑛𝑛] to denote 𝑥𝑥(𝑛𝑛𝑛𝑛), the nth sample of 𝑥𝑥(𝑡𝑡). Similarly, 𝑦𝑦[𝑛𝑛] denotes 𝑦𝑦[𝑛𝑛𝑛𝑛], the nth sample of
𝑦𝑦(𝑡𝑡). From the basic definition of a derivative, we can express Eq. (1.1a) at 𝑡𝑡 = 𝑛𝑛𝑛𝑛 as
𝑦𝑦[𝑛𝑛] − 𝑦𝑦[𝑛𝑛 − 1]
lim + 𝑐𝑐𝑐𝑐[𝑛𝑛] = 𝑥𝑥[𝑛𝑛]
𝑇𝑇→0 𝑇𝑇
Clearing the fractions and rearranging the terms yields (assuming nonzero, but very small
𝑇𝑇)

𝑦𝑦[𝑛𝑛] + 𝛼𝛼𝛼𝛼[𝑛𝑛 − 1] = 𝛽𝛽𝛽𝛽[𝑛𝑛] (1.1b)


Where
1 1
𝛼𝛼 = − and 𝛽𝛽 =
1+𝑐𝑐𝑐𝑐 1+𝑐𝑐𝑐𝑐

We can also express Eq. (1.1b) in advance operator form as

𝑦𝑦[𝑛𝑛 + 1] + 𝛼𝛼𝛼𝛼[𝑛𝑛] = 𝛽𝛽𝑥𝑥[𝑛𝑛 + 1] (1.1c)


It is clear that a differential equation can be approximated by a difference equation of the
same order. In this way, we can approximate an nth-order differential equation by a difference
equation of nth order. Indeed, a digital computer solves differential equations by using an
equivalent difference equation, which can be solved by means of simple operations of addition,
multiplication, and shifting. Recall that a computer can perform only these simple operations. It
must necessarily approximate complex operation like differentiation and integration in terms of
such simple operations. The approximation can be made as close to the exact answer as possible
by choosing sufficiently small value for T.

Forms of Difference Equations

In general, an Nth-order linear constant coefficient difference equation has the form
𝑁𝑁 𝑀𝑀

� 𝑎𝑎𝑘𝑘 𝑦𝑦[ 𝑛𝑛 − 𝑘𝑘 ] = � 𝑏𝑏𝑘𝑘 𝑥𝑥 [ 𝑛𝑛 − 𝑘𝑘 ] (2.1a)


𝑘𝑘=0 𝑘𝑘=0

where 𝑁𝑁 and 𝑀𝑀 are non-negative integers. Assuming the system is causal (and assuming
𝑎𝑎0 ≠ 0), solving the Eq (2.1a) for 𝑦𝑦[𝑛𝑛] results in the recursive definition
𝑀𝑀 𝑁𝑁
1
𝑦𝑦[𝑛𝑛] = �� 𝑏𝑏𝑘𝑘 𝑥𝑥 [ 𝑛𝑛 − 𝑘𝑘 ] − � 𝑎𝑎𝑘𝑘 𝑦𝑦[ 𝑛𝑛 − 𝑘𝑘 ]� (2.1b)
𝑎𝑎0
𝑘𝑘=0 𝑘𝑘=1

From this equation, note that 𝑦𝑦[ 𝑛𝑛 − 𝑘𝑘 ] represents the outputs and 𝑥𝑥 [ 𝑛𝑛 − 𝑘𝑘 ] represents the
inputs. The value of 𝑁𝑁 represents the order of the difference equation and corresponds to the
memory of the system being represented. Because this equation relies on past values of the output,
in order to compute a numerical solution, certain past outputs, referred to as the initial conditions,
must be known.
Consider some of example of difference equation such as

𝑦𝑦[𝑛𝑛] = 𝑎𝑎𝑎𝑎[𝑛𝑛 − 1] + 𝑥𝑥 [𝑛𝑛]


1 1
𝑦𝑦[𝑛𝑛] + 𝑦𝑦[𝑛𝑛 − 1] + 𝑦𝑦[𝑛𝑛 − 2] = 𝑥𝑥 [𝑛𝑛]
2 4
1
𝑦𝑦[𝑛𝑛] = (𝑥𝑥 [𝑛𝑛] − 𝑥𝑥 [𝑛𝑛 − 1])
𝑇𝑇
These are all linear equations with constant (not-time varying) coefficients. Time index 𝑛𝑛
arbitrary (can shift as desired).
Example

Given 𝑦𝑦[𝑛𝑛] = 𝑎𝑎𝑎𝑎[𝑛𝑛 − 1] + 𝑥𝑥[𝑛𝑛], transform to second order difference equation.

Solution

Let 𝑛𝑛 = 𝑛𝑛 + 2

𝑦𝑦[𝑛𝑛 + 2] = 𝑎𝑎𝑎𝑎[𝑛𝑛 + 2 − 1] + 𝑥𝑥 [𝑛𝑛 + 2]

𝑦𝑦[𝑛𝑛 + 2] = 𝑎𝑎𝑎𝑎[𝑛𝑛 + 1] + 𝑥𝑥[𝑛𝑛 + 2]

The general Advance Operator form is expressed as

𝑦𝑦[𝑛𝑛 + 𝑁𝑁] + 𝑎𝑎𝑁𝑁−1 𝑦𝑦[𝑛𝑛 + 𝑁𝑁 − 1] + ⋯ + 𝑎𝑎1 𝑦𝑦[𝑛𝑛 + 1] + 𝑎𝑎0 𝑦𝑦[𝑛𝑛] (2.2)


= 𝑏𝑏𝑀𝑀 𝑥𝑥 [𝑛𝑛 + 𝑀𝑀] + 𝑏𝑏𝑀𝑀−1 𝑥𝑥 [𝑛𝑛 + 𝑀𝑀 − 1] + ⋯ + 𝑏𝑏0 𝑥𝑥[𝑛𝑛]

The order of the systems is (𝑁𝑁,𝑀𝑀), indicated number of storage devices in the system.
Usually 𝑁𝑁≥ 𝑀𝑀 (for causality) and order is 𝑁𝑁.

The general Delay Operator form is expressed as

𝑦𝑦[𝑛𝑛] + 𝑎𝑎𝑁𝑁−1 𝑦𝑦[𝑛𝑛 − 1] + ⋯ + 𝑎𝑎1 𝑦𝑦[𝑛𝑛 − 𝑁𝑁 + 1] + 𝑎𝑎0 𝑦𝑦[𝑛𝑛 − 𝑁𝑁] (2.3)


= 𝑏𝑏𝑁𝑁 𝑥𝑥 [𝑛𝑛] + 𝑏𝑏𝑁𝑁−1 𝑥𝑥 [𝑛𝑛 − 1] + ⋯ + 𝑏𝑏0 𝑥𝑥[𝑛𝑛 − 𝑁𝑁]

Assumed causal 𝑀𝑀=𝑁𝑁 in above equation. If non-unity coefficient on 𝑦𝑦[𝑛𝑛], can divide to
force unity coefficient. It requires initial conditions, up to maximum system memory to perform
computations. Initial conditions 𝑦𝑦[−𝑁𝑁], 𝑦𝑦[−𝑛𝑛+1], …, 𝑦𝑦[−1].

Example:

Put into delay operator form the difference equation given by 3𝑦𝑦[𝑛𝑛 + 3] + 4y[n + 2] +
y[n] = 9x[n + 2] − 3x[n + 1].

Solution:

Let 𝑛𝑛 = 𝑛𝑛 − 3

3𝑦𝑦[𝑛𝑛] + 4𝑦𝑦[𝑛𝑛 − 1] + 𝑦𝑦[𝑛𝑛 − 3] = 9𝑥𝑥 [𝑛𝑛 − 1] − 3𝑥𝑥[𝑛𝑛 − 2]


𝟒𝟒 𝟏𝟏
𝒚𝒚[𝒏𝒏] + 𝒚𝒚[𝒏𝒏 − 𝟏𝟏] + 𝒚𝒚[𝒏𝒏 − 𝟑𝟑] = 𝟑𝟑𝟑𝟑[𝒏𝒏 − 𝟏𝟏] − 𝒙𝒙[𝒏𝒏 − 𝟐𝟐]
𝟑𝟑 𝟑𝟑
Iterative (Recursive) Solution of Difference Equation

The general delay operator form can be expressed as

𝑦𝑦[𝑛𝑛] = −𝑎𝑎𝑁𝑁−1 𝑦𝑦[𝑛𝑛 − 1] − 𝑎𝑎2 𝑦𝑦[𝑛𝑛 − 2] − ⋯ − 𝑎𝑎𝑁𝑁 𝑦𝑦[𝑛𝑛 − 𝑁𝑁]


(3.1)
+𝑏𝑏0 𝑥𝑥 [𝑛𝑛] + 𝑏𝑏1 𝑥𝑥 [𝑛𝑛 − 1] + ⋯ + 𝑏𝑏𝑁𝑁 𝑥𝑥[𝑛𝑛 − 𝑁𝑁]

In Eq. (3.1) 𝑦𝑦[𝑛𝑛] is computed from 2𝑁𝑁+1 pieces of information; the preceding 𝑁𝑁 values of
the output: 𝑦𝑦[𝑛𝑛−1], 𝑦𝑦[𝑛𝑛−2], . . . , 𝑦𝑦[𝑛𝑛−𝑁𝑁], and the preceding 𝑁𝑁 values of the input: x[𝑛𝑛−1] , x[𝑛𝑛−2],
. . . , x[𝑛𝑛−𝑁𝑁], and the present value of the input x[𝑛𝑛].

Suppose that the system is initially at rest and that 𝑥𝑥[𝑛𝑛] has nonzero values starting at 𝑛𝑛=0.
Then, the condition of initial rest means that 𝑦𝑦[−1]=𝑦𝑦[−1]=𝑦𝑦[−2]=…=𝑦𝑦[−𝑁𝑁]=0, and one can start
computing 𝑦𝑦[𝑛𝑛] recursively.

This is often how digital filters are implemented on a computer or a digital signal processor
board. This is also how a simulation of the response of a differential equation is typically
computed: first, the differential equation is discretized at a given sampling rate to obtain a
difference equation, and then the response of the difference equation is computed recursively.

Initially, to compute y[0], the 𝑁𝑁 initial conditions y[−1] , y[−2], . . . , y[−𝑁𝑁] serve as the
preceding 𝑁𝑁 output values.

Hence, knowing the 𝑁𝑁 initial conditions and the input, we can determine the entire output
y[0], y[1], y[2], y[3], . . . recursively, one value at a time
Example:

Solve iteratively the given difference equation 𝑦𝑦[𝑛𝑛 + 2] − 𝑦𝑦[𝑛𝑛 + 1] + 0.3𝑦𝑦[𝑛𝑛] =


𝑥𝑥 [𝑛𝑛 + 2] − 3𝑥𝑥 [𝑛𝑛 + 1] with initial conditions 𝑦𝑦[−1] = 1 and 𝑦𝑦[−2] = 2 and causal input 𝑥𝑥 [𝑛𝑛] =
𝑛𝑛 𝑢𝑢[𝑛𝑛]

Solution:

The given difference equation can be expressed as 𝑦𝑦[𝑛𝑛 + 2] = 𝑦𝑦[𝑛𝑛 + 1] − 0.3𝑦𝑦[𝑛𝑛] +


𝑥𝑥 [𝑛𝑛 + 2] − 3𝑥𝑥 [𝑛𝑛 + 1].

If we set 𝑛𝑛 = −2, we obtain

𝑦𝑦[0] = 𝑦𝑦[−1] − 0.3𝑦𝑦[−2] + 𝑥𝑥 [0] − 3𝑥𝑥[−1]

𝑦𝑦[0] = 1 − 0.3(2) + 1(0) − 3(0)

𝑦𝑦[0] = 1 − 0.6
𝒚𝒚[𝟎𝟎] = 𝟎𝟎. 𝟒𝟒

Next, setting 𝑛𝑛 = −1

𝑦𝑦[1] = 𝑦𝑦[0] − 0.3𝑦𝑦[−1] + 𝑥𝑥 [1] − 3𝑥𝑥 [0]

𝑦𝑦[1] = 0.4 − 0.3(1) + 1 − 3(0)

𝑦𝑦[1] = 0.4 − 0.3 + 1

𝒚𝒚[𝟏𝟏] = 𝟏𝟏. 𝟏𝟏

Continuing this iteratively we can obtain the following:

𝑛𝑛 = 0 : 𝑦𝑦[2] = 𝑦𝑦[1] − 0.3𝑦𝑦[0] + 𝑥𝑥 [2] − 3𝑥𝑥 [1] = −𝟎𝟎. 𝟎𝟎𝟎𝟎


𝑛𝑛 = 1 : 𝑦𝑦[3] = 𝑦𝑦[2] − 0.3𝑦𝑦[1] + 𝑥𝑥 [3] − 3𝑥𝑥 [2] = 𝟓𝟓. 𝟔𝟔𝟔𝟔

Note carefully the recursive nature of the computations. From the N initial conditions (and
the input) we obtained y[O] first. Then, using this value of y [O] and the preceding N - 1 initial
conditions (along with the input), we find y [ l ] . Next, using y [O], y [ l ] along with the past N -
2 initial conditions and input, we obtained y[2], and so on. This method is general and can be
applied to a recursive difference equation of any order.

We shall see in the future that the solution of a difference equation obtained in this direct
(iterative) way is useful in many situations. Despite the many uses of this method, a closed-form
solution of a difference equation is far more useful in study of the system behavior and its
dependence on the input and the various system parameters. For this reason we shall develop a
systematic procedure to analyze discrete-time systems along lines similar to those used for
continuous-time systems.

Operational Notation
In difference equations it is convenient to use operational notation similar to that used in
differential equations for the sake of compactness. In continuous-time systems we used the
operator D to denote the operation of differentiation. For discrete-time systems we shall use the
operator E to denote the operation for advancing a sequence by one time unit. Thus

𝐸𝐸𝐸𝐸 [𝑛𝑛] ≡ 𝑥𝑥 [𝑛𝑛 + 1]

𝐸𝐸2 𝑥𝑥 [𝑛𝑛] ≡ 𝑥𝑥 [𝑛𝑛 + 2]

𝐸𝐸3 𝑥𝑥 [𝑛𝑛] ≡ 𝑥𝑥 [𝑛𝑛 + 3]

𝐸𝐸𝑁𝑁 𝑥𝑥 [𝑛𝑛] ≡ 𝑥𝑥[𝑛𝑛 + 𝑁𝑁] (4.1)


Example
Express the following difference equation to operator notation.

1. 𝑦𝑦[𝑛𝑛] = 𝑎𝑎𝑎𝑎[𝑛𝑛 − 1] + 𝑥𝑥 [𝑛𝑛]


1 1
2. 𝑦𝑦[𝑛𝑛 + 2] + 𝑦𝑦[𝑛𝑛 + 1] + 𝑦𝑦[𝑛𝑛] = 𝑥𝑥[𝑛𝑛 + 2]
2 4

Solution 1
Given 𝑦𝑦[𝑛𝑛] = 𝑎𝑎𝑎𝑎[𝑛𝑛 − 1] + 𝑥𝑥[𝑛𝑛]

𝑦𝑦[𝑛𝑛] − 𝑎𝑎𝑎𝑎[𝑛𝑛 − 1] = 𝑥𝑥 [𝑛𝑛]

𝑦𝑦[𝑛𝑛 + 1] − 𝑎𝑎𝑎𝑎[𝑛𝑛] = 𝑥𝑥 [𝑛𝑛 + 1]

𝐸𝐸𝐸𝐸[𝑛𝑛] − 𝑎𝑎𝑎𝑎[𝑛𝑛] = 𝐸𝐸𝐸𝐸 [𝑛𝑛]


(𝑬𝑬 − 𝒂𝒂)𝒚𝒚[𝒏𝒏] = 𝑬𝑬𝑬𝑬[𝒏𝒏]

Solution 2
1 1
Given 𝑦𝑦[𝑛𝑛 + 2] + 2 𝑦𝑦[𝑛𝑛 + 1] + 4 𝑦𝑦[𝑛𝑛] = 𝑥𝑥[𝑛𝑛 + 2]

1 1
𝐸𝐸2 𝑦𝑦[𝑛𝑛] + 𝐸𝐸𝐸𝐸[𝑛𝑛] + 𝑦𝑦[𝑛𝑛] = 𝐸𝐸2 𝑥𝑥 [𝑛𝑛]
2 4
𝟏𝟏 𝟏𝟏
�𝑬𝑬𝟐𝟐 + 𝑬𝑬 + � 𝒚𝒚[𝒏𝒏] = 𝑬𝑬𝟐𝟐 𝒙𝒙[𝒏𝒏]
𝟐𝟐 𝟒𝟒

If we expressed Eq. (2.2) using operator notation, we obtain


(𝐸𝐸𝑁𝑁 + 𝑎𝑎𝑁𝑁−1 𝐸𝐸𝑁𝑁−1 + ⋯ + 𝑎𝑎1 𝐸𝐸 + 𝑎𝑎0 )𝑦𝑦[𝑛𝑛]
= (𝑏𝑏𝑁𝑁 𝐸𝐸𝑁𝑁 + 𝑏𝑏𝑁𝑁−1 𝐸𝐸𝑁𝑁−1 + ⋯ + 𝑏𝑏1 𝐸𝐸 + 𝑏𝑏0 )𝑥𝑥 [𝑛𝑛] (4.2)

or

𝑸𝑸[𝑬𝑬]𝒚𝒚[𝒏𝒏] = 𝑃𝑃[𝐸𝐸]𝒙𝒙[𝒏𝒏]

Where 𝑄𝑄[𝐸𝐸] and 𝑃𝑃[𝐸𝐸] are Nth-order polynomial operators

(4.3)
(4.4)
Zero-Input Response

The zero-input response 𝑦𝑦0 [𝑛𝑛] is the solution to the difference equation if the input 𝑥𝑥 [𝑛𝑛] =
0. That is,

𝑄𝑄 [𝐸𝐸]𝑦𝑦[𝑛𝑛] = 𝑃𝑃[𝐸𝐸]𝑥𝑥 [𝑛𝑛] → 𝑄𝑄[𝐸𝐸]𝑦𝑦[𝑛𝑛] = 0 (5.1)


The previous equation can be expanded as

𝑦𝑦0 [𝑛𝑛 + 𝑁𝑁] + 𝑎𝑎𝑁𝑁−1 𝑦𝑦[𝑛𝑛 + 𝑁𝑁 − 1] + ⋯ + 𝑎𝑎1 𝑦𝑦0 [𝑛𝑛 + 1] + 𝑎𝑎0 𝑦𝑦0 [𝑛𝑛] = 0 (5.2)
We can solve this equation systematically. But even a cursory examination of this equation
points to its solution. This equation states that a linear combination of 𝑦𝑦0 [𝑛𝑛] and advanced 𝑦𝑦0 [𝑛𝑛]
is zero, not for some values of 𝑛𝑛, but for all 𝑛𝑛. Such situation is possible if and only if 𝑦𝑦0 [𝑛𝑛] and
advanced 𝑦𝑦0 [𝑛𝑛] have the same form. Only an exponential function 𝛾𝛾 𝑛𝑛 has this property

𝛾𝛾 𝑛𝑛+𝑚𝑚 = 𝛾𝛾 𝑛𝑛 𝛾𝛾 𝑚𝑚 (5.3)

Equation (5.3) shows that 𝛾𝛾 𝑛𝑛 advanced by 𝑚𝑚units is a constant (𝛾𝛾 𝑚𝑚) time 𝛾𝛾 𝑛𝑛 . Therefore,
the solution must be of the form

𝑦𝑦0 [𝑛𝑛] = 𝑐𝑐𝛾𝛾 𝑛𝑛 (5.4)

To determine 𝑐𝑐 and 𝑦𝑦, we substitute this solution in Eq. (5.1). Equation (5.4) yields

Furthermore, it yields to

𝑐𝑐𝛾𝛾 𝑛𝑛 (𝑐𝑐𝛾𝛾 𝑁𝑁 + 𝑎𝑎𝑁𝑁−1 𝛾𝛾 𝑁𝑁−1 + 𝑎𝑎𝑁𝑁−2 𝛾𝛾 𝑁𝑁−2 + ⋯ + 𝑎𝑎1 𝛾𝛾 + 𝑎𝑎0 ) = 0 (5.5)


For a nontrivial solution of this equation

𝑐𝑐𝛾𝛾 𝑁𝑁 + 𝑎𝑎𝑁𝑁−1 𝛾𝛾 𝑁𝑁−1 + 𝑎𝑎𝑁𝑁−2 𝛾𝛾 𝑁𝑁−2 + ⋯ + 𝑎𝑎1 𝛾𝛾 + 𝑎𝑎0 = 0 (5.6)

Equation (5.6) is an Nth-order polynomial in 𝛾𝛾. This polynomial is the same as the 𝐸𝐸 operators
Q[𝐸𝐸] . Thus, we can express this polynomial as Q[𝛾𝛾 ]. This will give N-roots to the polynomial
which gives N solution to the equation 𝑄𝑄[𝐸𝐸]𝑦𝑦0 [𝑛𝑛] = 0

CASE I: DISTINCT ROOTS

𝑄𝑄[𝛾𝛾 ] is an nth-order polynomial and can be expressed. in the factored form (assuming all
distinct roots):

𝑄𝑄[𝛾𝛾 ] = (𝛾𝛾 − 𝛾𝛾1 )(𝛾𝛾 − 𝛾𝛾2 ) … (𝛾𝛾 − 𝛾𝛾𝑁𝑁 ) = 0 (5.7)


Clearly, 𝛾𝛾 has 𝑁𝑁 solutions 𝛾𝛾1 , 𝛾𝛾2 , … 𝛾𝛾𝑁𝑁 and, therefore, Eq. (5.1) also has 𝑁𝑁 solutions
𝑐𝑐1𝛾𝛾 , 𝑐𝑐2 𝛾𝛾2𝑛𝑛 , … 𝑐𝑐𝑁𝑁 𝛾𝛾𝑁𝑁𝑛𝑛 . In such a case, we have shown that the general solution is a linear combination
𝑛𝑛

of the 𝑁𝑁 solutions. Thus

𝒚𝒚𝒐𝒐 [𝒏𝒏] = 𝒄𝒄𝟏𝟏 𝜸𝜸𝒏𝒏 + 𝒄𝒄𝟐𝟐 𝜸𝜸𝒏𝒏𝟐𝟐 + ⋯ + 𝒄𝒄𝑵𝑵 𝜸𝜸𝒏𝒏𝑵𝑵 (5.8)

where 𝛾𝛾1, 𝛾𝛾2 , … 𝛾𝛾𝑁𝑁 are the roots of Eq. (5.7) and 𝑐𝑐1, 𝑐𝑐2 , … 𝑐𝑐𝑁𝑁 are arbitrary constants determined
from 𝑁𝑁 auxiliary conditions, generally given in the form of initial conditions. The polynomial 𝑄𝑄[𝛾𝛾 ]
is called the characteristic polynomial of the system, and

𝑄𝑄[𝛾𝛾 ] = 0 (5.9)

is the characteristic equation of the system. Moreover, 𝛾𝛾1 , 𝛾𝛾2 , … 𝛾𝛾𝑁𝑁 , the roots of the characteristic
equation, are called characteristic roots or characteristic values (also eigenvalues) of the system.
The exponentials 𝛾𝛾𝑖𝑖𝑛𝑛 (𝑖𝑖 = 1, 2, … , 𝑁𝑁) are the characteristic modes or natural modes of the
system. A characteristic mode corresponds to each characteristic roots of the system, and the zero-
input response is a linear combination of the characteristic modes of the system.
Example:

Consider the DT LTI system described by 𝑦𝑦[𝑛𝑛 + 2] − 0.8𝑦𝑦[𝑛𝑛 + 1] − 0.09𝑦𝑦[𝑛𝑛] = 3 𝑥𝑥[𝑛𝑛 +


10
2]. Find the zero-input response if 𝑦𝑦[−1] = 0 and 𝑦𝑦[−2] = .
3

Solution:

Operator notation: (𝐸𝐸2 − 0.8𝐸𝐸 − 0.09)𝑦𝑦[𝑛𝑛] = 3𝐸𝐸2 𝑥𝑥[𝑛𝑛]

Characteristic polynomial: 𝛾𝛾 2 − 0.8𝛾𝛾 − 0.09

Characteristic equation:

𝛾𝛾 2 − 0.8𝛾𝛾 − 0.09 = 0(𝛾𝛾 − 0.9)(𝛾𝛾 + 0.1) = 0


Therefore,

𝑦𝑦0 [𝑛𝑛] = 𝑐𝑐1 (0.9)𝑛𝑛 + 𝑐𝑐2 (−0.1)𝑛𝑛

Solve for constants:

𝑛𝑛 = −1
0 = 𝑐𝑐1 (0.9)−1 + 𝑐𝑐2 (−0.1)−1
10
𝑐𝑐1 � � + 𝑐𝑐2 (10) = 0
9
𝑛𝑛 = −2
10
= 𝑐𝑐1 (0.9)−2 + 𝑐𝑐2(−0.1)−2
3
100 10
𝑐𝑐1 � � + 𝑐𝑐2 (100) =
81 3
Solving for systems of linear equations for 2 unknowns, we obtain 𝑐𝑐1 = 27/100 and 𝑐𝑐2 =
3/100.
Therefore,
𝟐𝟐𝟐𝟐 𝟑𝟑
𝒚𝒚𝟎𝟎 [𝒏𝒏] = (𝟎𝟎. 𝟗𝟗)𝒏𝒏 + (−𝟎𝟎. 𝟏𝟏)𝒏𝒏
𝟏𝟏𝟏𝟏𝟏𝟏 𝟏𝟏𝟏𝟏𝟏𝟏

CASE II: REPEATED ROOTS

So far we have assumed the system to have 𝑁𝑁 distinct characteristic roots 𝛾𝛾1, 𝛾𝛾2 , … 𝛾𝛾𝑁𝑁 with
corresponding characteristic modes 𝛾𝛾 𝑛𝑛 , 𝛾𝛾2𝑛𝑛 , … 𝛾𝛾𝑁𝑁𝑛𝑛 . If two or more roots coincide (repeated roots),
the form of characteristic modes is modified. Direct substitution shows that if a root 𝛾𝛾 repeats 𝑟𝑟
times (root of multiplicity 𝑟𝑟), the corresponding characteristic modes for this root are
𝛾𝛾1𝑛𝑛 , 𝑛𝑛𝛾𝛾1𝑛𝑛 , 𝑛𝑛2 𝛾𝛾1𝑛𝑛 , … , 𝑛𝑛𝑟𝑟−1 𝛾𝛾1𝑛𝑛 . Thus, if the characteristic equation of a system is
Q[𝛾𝛾 ] = (𝛾𝛾 − 𝛾𝛾1 )𝑟𝑟 (𝛾𝛾 − 𝛾𝛾𝑟𝑟+1 ) … (𝛾𝛾 − 𝛾𝛾𝑁𝑁 ) = 0 (5.10)
the zero-input response of the system is

𝒚𝒚𝒐𝒐 [𝒏𝒏] = (𝒄𝒄𝟏𝟏 + 𝒄𝒄𝟐𝟐 𝒏𝒏 + 𝒄𝒄𝟑𝟑 𝒏𝒏𝟐𝟐 + ⋯ + 𝒄𝒄𝒓𝒓 𝒏𝒏𝒓𝒓−𝟏𝟏 )𝜸𝜸𝒏𝒏𝟏𝟏 + 𝒄𝒄𝒓𝒓+𝟏𝟏 𝜸𝜸𝒏𝒏𝒓𝒓+𝟏𝟏 + ⋯ + 𝒄𝒄𝑵𝑵 𝜸𝜸𝒏𝒏𝑵𝑵 (5.11)

Example:

Consider the DT LTI system described by (𝐸𝐸2 + 6𝐸𝐸 + 9)𝑦𝑦[𝑛𝑛] = (2𝐸𝐸2 + 6𝐸𝐸)𝑥𝑥[𝑛𝑛]. Find
2 1
the zero-input response if 𝑦𝑦[−1] = − and 𝑦𝑦[−2] = .
3 9

Solution
𝛾𝛾 2 + 6𝛾𝛾 + 9 = 0

(𝛾𝛾 + 3)2 = 0

Characteristic modes are (−3)𝑛𝑛 and 𝑛𝑛(−3)𝑛𝑛

𝑦𝑦0 [𝑛𝑛] = 𝑐𝑐1 (−3)𝑛𝑛 + 𝑐𝑐2 𝑛𝑛(−3)𝑛𝑛


𝑦𝑦0 [𝑛𝑛] = (𝑐𝑐1 + 𝑐𝑐2𝑛𝑛)(−3)𝑛𝑛
Solving for the constants:

𝑛𝑛 = −1
2 1
− = (𝑐𝑐1 − 𝑐𝑐2 ) �− �
3 3

𝑛𝑛 = −2
1 1
= (𝑐𝑐1 − 2𝑐𝑐2 ) � �
9 9

Solving for systems of linear equations with 2 unknowns, we obtain 𝑐𝑐1 = 3 and 𝑐𝑐2 = 1

Therefore,

𝒚𝒚𝟎𝟎 [𝒏𝒏] = 𝟑𝟑(−𝟑𝟑)𝒏𝒏 + 𝒏𝒏(−𝟑𝟑)𝒏𝒏

CASE III: COMPLEX ROOTS


As in the case of continuous-time systems, the complex roots of a discrete time system will
occur in pairs of conjugates if the system equation coefficients are real. Complex roots can be
treated exactly as we would treat real roots. However, just as in the case of continuous-time
systems, we can also use the real form of solution as an alternative.

First, we express the complex conjugate roots 𝛾𝛾 and 𝛾𝛾 ∗in polar form. If |𝛾𝛾 |is the magnitude
and 𝛽𝛽 is the angle of 𝛾𝛾, then

𝛾𝛾 = |𝛾𝛾 |𝑒𝑒 𝑗𝑗𝑗𝑗 and 𝛾𝛾 ∗ = |𝛾𝛾 |𝑒𝑒 −𝑗𝑗𝑗𝑗


Therefore, the solution is

𝑦𝑦𝑜𝑜 [𝑛𝑛] = 𝑐𝑐1 𝛾𝛾 𝑛𝑛 + 𝑐𝑐2 (𝛾𝛾 ∗)𝑛𝑛 𝑦𝑦𝑜𝑜 [𝑛𝑛] = 𝑐𝑐1 |𝛾𝛾 |𝑛𝑛 𝑒𝑒 𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑐𝑐2 |𝛾𝛾 |𝑛𝑛 𝑒𝑒 −𝑗𝑗𝑗𝑗𝑗𝑗

Constants 𝑐𝑐1 and 𝑐𝑐2 must be conjugates for a real system.


𝑐𝑐 𝑐𝑐
𝑐𝑐1 = 𝑒𝑒 𝑗𝑗𝑗𝑗 and 𝑐𝑐2 = 𝑒𝑒 −𝑗𝑗𝑗𝑗
2 2

Then
𝑐𝑐
𝑦𝑦𝑜𝑜 [𝑛𝑛] = |𝛾𝛾 |𝑛𝑛 �𝑒𝑒 𝑗𝑗(𝛽𝛽𝛽𝛽+𝜃𝜃) + 𝑒𝑒 −𝑗𝑗(𝛽𝛽𝛽𝛽+𝜃𝜃) �
2
𝒚𝒚𝒐𝒐 [𝒏𝒏] = 𝒄𝒄|𝜸𝜸|𝒏𝒏 𝐜𝐜𝐜𝐜𝐜𝐜 (𝜷𝜷𝜷𝜷 + 𝜽𝜽)
Example

Consider the DT LTI system described by (𝐸𝐸2 − 1.3𝐸𝐸 + 0.663)𝑦𝑦[𝑛𝑛] = 𝐸𝐸2 𝑥𝑥[𝑛𝑛]. Find the
zero-input response if 𝑦𝑦[−1] = 2 and 𝑦𝑦[−2] = 1.

Solution
Characteristic equation:

𝛾𝛾 2 − 1.3𝛾𝛾 + 0.663 = 0
(𝛾𝛾 − 0.64 + j0.49)(𝛾𝛾 − 0.65 − j0.49) = 0

Roots

𝛾𝛾 = 0.65 ± 𝑗𝑗0.49 = 0.814𝑒𝑒 ±𝑗𝑗0.646

|𝛾𝛾 | = �0.814𝑒𝑒 ±𝑗𝑗0.646� = 0.814

𝛽𝛽 = ∠�0.814𝑒𝑒 ±𝑗𝑗0.646 � = 0.646

Thus,

𝑦𝑦0 [𝑛𝑛] = 𝑐𝑐|0.814|𝑛𝑛 cos(0.646𝑛𝑛 + 𝜃𝜃)


Using the initial conditions to solve for constant:
𝑐𝑐
𝑛𝑛 = −1 : 2 = cos(−0.646 + 𝜃𝜃 )
0.814
𝑐𝑐
𝑛𝑛 = −2 : 1 = cos(−1.292 + 𝜃𝜃)
0.663

Simplifying further,
𝑐𝑐 𝑐𝑐
cos(−0.646 + 𝜃𝜃 ) = (cos(0.646) cos 𝜃𝜃 + sin(0.646) sin 𝜃𝜃)
0.814 0.814
𝑐𝑐 [0.981 cos 𝜃𝜃 + 0.740 sin 𝜃𝜃 ] = 2

0.981𝑐𝑐 cos 𝜃𝜃 + 0.740𝑐𝑐 sin 𝜃𝜃 = 2

𝑐𝑐 𝑐𝑐
cos(−1.292 + 𝜃𝜃 ) = (cos(1.292) cos 𝜃𝜃 + sin(1.292) sin 𝜃𝜃 )
0.663 0.663
𝑐𝑐 [0.415 cos 𝜃𝜃 + 1.450 sin 𝜃𝜃 ] = 1

0.415𝑐𝑐 cos 𝜃𝜃 + 1.450𝑐𝑐 sin 𝜃𝜃 = 1


Let 𝑥𝑥 = 𝑐𝑐 cos 𝜃𝜃 and y = 𝑐𝑐 sin 𝜃𝜃

Thus,

0.981𝑐𝑐 cos 𝜃𝜃 + 0.740𝑐𝑐 sin 𝜃𝜃 = 2 → 0.981 𝑥𝑥 + 0.740𝑦𝑦 = 2

0.415𝑐𝑐 cos 𝜃𝜃 + 1.450𝑐𝑐 sin 𝜃𝜃 = 1 → 0.415𝑥𝑥 + 1.450𝑦𝑦 = 1

Solving systems of linear equations with 2 unknowns, we obtain 𝑥𝑥 = 1.937 and y =


0.135

Since 𝑥𝑥 = 𝑐𝑐 cos 𝜃𝜃 and y = 𝑐𝑐 sin 𝜃𝜃,


𝑐𝑐 sin 𝜃𝜃 0.135
= tan 𝜃𝜃 = 𝜃𝜃 = 0.0696 𝑟𝑟𝑟𝑟𝑟𝑟
𝑐𝑐 cos 𝜃𝜃 1.937
Thus, 𝑐𝑐 = 1.942

Therefore,

𝒚𝒚𝟎𝟎 [𝒏𝒏] = 𝟏𝟏. 𝟗𝟗𝟗𝟗𝟗𝟗(𝟎𝟎. 𝟖𝟖𝟖𝟖𝟖𝟖)𝒏𝒏 𝐜𝐜𝐜𝐜𝐜𝐜(𝟎𝟎. 𝟔𝟔𝟔𝟔𝟔𝟔𝟔𝟔 + 𝟎𝟎. 𝟎𝟎𝟎𝟎𝟎𝟎𝟎𝟎)

Unit Impulse Response


The impulse response of the system is the output of the system at rest when a single unit
impulse is applied. This is a case where the system output is zero for all instance of time except
for a single input of 1 at t=0.
Consider an nth-order system specified by the equation
(𝐸𝐸𝑁𝑁 + 𝑎𝑎1 𝐸𝐸𝑁𝑁−1 + ⋯ + 𝑎𝑎𝑁𝑁−1 𝐸𝐸 + 𝑎𝑎𝑁𝑁 )𝑦𝑦[𝑛𝑛] = (𝑏𝑏0 𝐸𝐸𝑁𝑁 + 𝑏𝑏1 𝐸𝐸𝑁𝑁−1 + ⋯ + 𝑏𝑏𝑁𝑁−1 𝐸𝐸 + 𝑏𝑏𝑁𝑁 ) 𝑥𝑥[𝑛𝑛]

Or it can be represented as 𝑄𝑄[𝐸𝐸]𝑦𝑦[𝑛𝑛] = 𝑃𝑃[𝐸𝐸]𝑥𝑥[𝑛𝑛]

The unit impulse response ℎ[𝑛𝑛] is the solution of this equation for the input 𝛿𝛿[𝑛𝑛] with all
the initial conditions zero; that is,

𝑄𝑄[𝐸𝐸]ℎ[𝑛𝑛] = 𝑃𝑃[𝐸𝐸]𝛿𝛿[𝑛𝑛]

subject to initial conditions: ℎ[−1] = ℎ[−2] = ⋯ = ℎ[−𝑁𝑁] = 0. The equation


𝑄𝑄[𝐸𝐸]ℎ[𝑛𝑛] = 𝑃𝑃[𝐸𝐸]𝛿𝛿[𝑛𝑛] can be solved to determine ℎ[𝑛𝑛] iteratively or in a closed form.

Example:

Find the impulse response for the DT system 𝑦𝑦[𝑛𝑛] − 0.5𝑦𝑦[𝑛𝑛 − 1] − 0.2𝑦𝑦[𝑛𝑛 − 2] =
3𝑥𝑥[𝑛𝑛] iteratively.
Solution

Let 𝑥𝑥 [𝑛𝑛] = 𝛿𝛿[𝑛𝑛] and y[𝑛𝑛] = ℎ[𝑛𝑛], then

ℎ[𝑛𝑛] − 0.5ℎ[𝑛𝑛 − 1] − 0.2ℎ[𝑛𝑛 − 2] = 3𝛿𝛿[𝑛𝑛]

Solving for ℎ[𝑛𝑛]

ℎ[𝑛𝑛] = 0.5ℎ[𝑛𝑛 − 1] + 0.2ℎ[𝑛𝑛 − 2] + 3𝛿𝛿[𝑛𝑛]

Take note that the initial conditions is always equal to zero at 𝑡𝑡 ≤ 0, thus

ℎ[−1] = ℎ[−2] = ⋯ = 0

𝑛𝑛 = 0

ℎ[0] = 0.5ℎ[−1] + 0.2ℎ[−2] + 3𝛿𝛿 [0]

ℎ[0] = 0 + 0 + 3(1)

𝒉𝒉[𝟎𝟎] = 𝟑𝟑

𝑛𝑛 = 1
ℎ[1] = 0.5ℎ[0] + 0.2ℎ[−1] + 3𝛿𝛿 [1]

ℎ[1] = 0.5(3) + 0 + 3(0)

𝒉𝒉[𝟏𝟏] = 𝟏𝟏. 𝟓𝟓

𝑛𝑛 = 2

ℎ[2] = 0.5ℎ[1] + 0.2ℎ[0] + 3𝛿𝛿 [2]

ℎ[2] = 0.5(1.5) + 0.2(3) + 3(0)

𝒉𝒉[𝟐𝟐] = 𝟏𝟏. 𝟑𝟑𝟑𝟑

Continuing this way, we can determine any number of terms of ℎ[𝑛𝑛]. Unfortunately, such
a solution does not yield a closed-form expression for ℎ[𝑛𝑛]. Nevertheless, determining a few values
of ℎ[𝑛𝑛] can be useful in determining the closed-form solution.

CLOSED FORM SOLUTION OF IMPULSE RESPONSE

Recall that ℎ[𝑛𝑛] is the system response to input 𝛿𝛿 [𝑛𝑛], which is zero for n > 0. We know
that when the input is zero, only the characteristic modes can be sustained by the system.
Therefore, ℎ[𝑛𝑛] must be made up of characteristic modes for n > 0. At n = 0, it may have some
nonzero value 𝐴𝐴0 , so that a general form of ℎ[𝑛𝑛] can be expressed as

where 𝑦𝑦𝑐𝑐 [𝑛𝑛] is a linear combination of the characteristic modes. We now substitute the
above equation in 𝑄𝑄[𝐸𝐸]ℎ[𝑛𝑛] = 𝑃𝑃[𝐸𝐸]𝛿𝛿[𝑛𝑛] to obtain 𝑄𝑄[𝐸𝐸](𝐴𝐴0 𝛿𝛿 [𝑛𝑛] + 𝑦𝑦𝑐𝑐 [𝑛𝑛]𝛿𝛿[𝑛𝑛]) . Because 𝑦𝑦𝑐𝑐 [𝑛𝑛]
is made up of characteristic modes, 𝑄𝑄 [𝐸𝐸]𝑦𝑦𝑐𝑐 [𝑛𝑛]𝑢𝑢[𝑛𝑛] = 0 and we obtain 𝐴𝐴0 𝑄𝑄[𝐸𝐸]𝛿𝛿 [𝑛𝑛] =
𝑃𝑃[𝐸𝐸] 𝛿𝛿[𝑛𝑛], that is,

Setting n = 0 in this equation and using the fact that 𝛿𝛿[𝑚𝑚] = 0 for all 𝑚𝑚 ≠ 0, and 𝛿𝛿 [0] = 1 , we
obtain

Therefore,
𝑏𝑏𝑁𝑁
ℎ[𝑛𝑛] = 𝛿𝛿 [𝑛𝑛] + 𝑦𝑦𝑐𝑐 [𝑛𝑛]𝑢𝑢[𝑛𝑛]
𝑎𝑎𝑁𝑁
Example

Find the closed form impulse response for the DT system 𝑦𝑦[𝑛𝑛] − 0.5𝑦𝑦[𝑛𝑛 − 1] −
0.2𝑦𝑦[𝑛𝑛 − 2] = 3𝑥𝑥[𝑛𝑛]

Solution

Let 𝑛𝑛 = 𝑛𝑛 + 2

𝑦𝑦[𝑛𝑛 + 2] − 0.5𝑦𝑦[𝑛𝑛 + 1] − 0.2𝑦𝑦[𝑛𝑛] = 3𝑥𝑥 [𝑛𝑛 + 2]


(𝐸𝐸2 − 0.5𝐸𝐸 − 0.2)𝑦𝑦[𝑛𝑛] = 3𝐸𝐸2 𝑥𝑥 [𝑛𝑛]

(𝛾𝛾 2 − 0.5𝛾𝛾 − 0.2) = 0

(𝛾𝛾 + 0.262)(𝛾𝛾 − 0.762) = 0

Thus, 𝑦𝑦𝐶𝐶 [𝑛𝑛] = 𝑐𝑐1 (−0.262)𝑛𝑛 + 𝑐𝑐2(0.762)𝑛𝑛

From the operator notation, we obtain 𝑎𝑎𝑁𝑁 = −0.2 and 𝑏𝑏𝑁𝑁 = 0

ℎ[𝑛𝑛] = [𝑐𝑐1(−0.262)𝑛𝑛 + 𝑐𝑐2 (0.762)𝑛𝑛 ]𝑢𝑢[𝑛𝑛]


Solving for 𝑐𝑐1 and 𝑐𝑐2, recall from the previous example

ℎ[0] = 3 and ℎ[1] = 1.5

Substituting this we obtain,


[𝑐𝑐1(−0.262)0 + 𝑐𝑐2 (0.762)0 ]𝑢𝑢[0] = 3

𝑐𝑐1 + 𝑐𝑐2 = 3

[𝑐𝑐1(−0.262)1 + 𝑐𝑐2 (0.762)1 ]𝑢𝑢[1] = 1.5

𝑐𝑐1(−0.262) + 𝑐𝑐2 (0.762) = 1.5

Solving systems of linear equation with two unknowns,


𝑐𝑐1 + 𝑐𝑐2 = 3𝑐𝑐1(−0.262) + 𝑐𝑐2 (0.762) = 1.5

We obtain 𝑐𝑐1 = 0.768 and 𝑐𝑐2 = 2.232


Therefore,

𝒉𝒉[𝒏𝒏] = [𝟎𝟎. 𝟕𝟕𝟕𝟕𝟕𝟕(−𝟎𝟎. 𝟐𝟐𝟐𝟐𝟐𝟐)𝒏𝒏 + 𝟐𝟐. 𝟐𝟐𝟐𝟐𝟐𝟐(𝟎𝟎. 𝟕𝟕𝟕𝟕𝟕𝟕)𝒏𝒏 ]𝒖𝒖[𝒏𝒏]

Zero-state Response

The zero-state response 𝑦𝑦[𝑛𝑛] is the system response to an input 𝑥𝑥[𝑛𝑛] when the system is in
the zero state. A signal 𝑥𝑥[𝑛𝑛] can be expressed as a sum of impulse components. The component of
𝑥𝑥[𝑛𝑛] at 𝑛𝑛 = 𝑚𝑚 is 𝑥𝑥[𝑚𝑚]𝛿𝛿[𝑛𝑛 − 𝑚𝑚], and 𝑥𝑥[𝑛𝑛] is the sum of all these components summed from 𝑚𝑚 =
−∞ 𝑡𝑡𝑡𝑡 ∞.

Therefore

𝑥𝑥[𝑛𝑛] = 𝑥𝑥 [0]𝛿𝛿 [𝑛𝑛] + 𝑥𝑥[1]𝛿𝛿[𝑛𝑛 − 1] + 𝑥𝑥 [2]𝛿𝛿[𝑛𝑛 − 2] + ⋯ + 𝑥𝑥[−1]𝛿𝛿 [𝑛𝑛 + 1] + 𝑥𝑥[−2]𝛿𝛿[𝑛𝑛 + 2]


+ ⋯ 𝑥𝑥 [𝑛𝑛] = � 𝑥𝑥 [𝑚𝑚]𝛿𝛿 [𝑛𝑛 − 𝑚𝑚]


𝑚𝑚=−∞

For a linear system, knowing the system response to impulse 𝛿𝛿[𝑛𝑛], the system response to
any arbitrary input could be obtain by summing the system response to various impulse
components.
Let ℎ[𝑛𝑛] be the system response to impulse input 𝛿𝛿[𝑛𝑛]. We shall use the notation

𝑥𝑥[𝑛𝑛] ⇒ 𝑦𝑦[𝑛𝑛]

to indicate the input and the corresponding response of the system. Thus, if

𝛿𝛿[𝑛𝑛] ⇒ ℎ[𝑛𝑛]

then because of time invariance

𝛿𝛿[𝑛𝑛 − 𝑚𝑚] ⇒ ℎ[𝑛𝑛 − 𝑚𝑚]

and because of linearity

𝑥𝑥[𝑚𝑚]𝛿𝛿[𝑛𝑛 − 𝑚𝑚] ⇒ 𝑥𝑥 [𝑚𝑚]ℎ[𝑛𝑛 − 𝑚𝑚]


and again because of linearity
∞ ∞

� 𝑥𝑥[𝑚𝑚]𝛿𝛿[𝑛𝑛 − 𝑚𝑚] ⇒ � 𝑥𝑥 [𝑚𝑚]ℎ[𝑛𝑛 − 𝑚𝑚]


𝑚𝑚=−∞ 𝑚𝑚=−∞

The left-hand side is 𝑥𝑥[𝑛𝑛] and the right-hand side is the system response 𝑦𝑦[𝑛𝑛] to input 𝑥𝑥 [𝑛𝑛].

Therefore,

𝑦𝑦[𝑛𝑛] = � 𝑥𝑥 [𝑚𝑚]ℎ[𝑛𝑛 − 𝑚𝑚]


𝑚𝑚=−∞

The summation on the right-hand side is known as the convolution sum of 𝑥𝑥[𝑛𝑛] and ℎ[𝑛𝑛], and is
represented symbolically by 𝑥𝑥 [𝑛𝑛] ∗ ℎ[𝑛𝑛]

𝒙𝒙[𝒏𝒏] ∗ 𝒉𝒉[𝒏𝒏] = � 𝒙𝒙[𝒎𝒎]𝒉𝒉[𝒏𝒏 − 𝒎𝒎]


𝒎𝒎=−∞

CONVOLUTION SUM FROM A TABLE


Just in the continuous-time case, convolution sums may be determined directly for a
variety of signal pairs using a convolution sum table.
Example:

Find the zero-state response 𝑦𝑦[𝑛𝑛] of an LTID system described by the equation
𝑦𝑦[𝑛𝑛 + 2] − 0.6𝑦𝑦[𝑛𝑛 + 1] − 0.16𝑦𝑦[𝑛𝑛] = 5𝑥𝑥[𝑛𝑛 + 2] if the input 𝑥𝑥 [𝑛𝑛] = 2−𝑛𝑛 𝑢𝑢[𝑛𝑛].
Solution:

Zero-state response is given by 𝑦𝑦[𝑛𝑛] = 𝑥𝑥[𝑛𝑛] ∗ ℎ[𝑛𝑛]

From the previous example: ℎ[𝑛𝑛] = [(−0.2)𝑛𝑛 + 4(0.8)𝑛𝑛 ]𝑢𝑢[𝑛𝑛]

𝑦𝑦[𝑛𝑛] = 𝑥𝑥 [𝑛𝑛] ∗ ℎ[𝑛𝑛]

𝑦𝑦[𝑛𝑛] = 2−𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ [(−0.2)𝑛𝑛 + 4(0.8)𝑛𝑛 ]𝑢𝑢[𝑛𝑛]𝑦𝑦[𝑛𝑛]


= 2−𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ (−0.2)𝑛𝑛 𝑢𝑢[𝑛𝑛] + 2−𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 4(0.8)𝑛𝑛 𝑢𝑢[𝑛𝑛]
From the convolution table (pair 4):
𝛾𝛾1𝑛𝑛+1 − 𝛾𝛾2𝑛𝑛+1
𝛾𝛾1𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 𝛾𝛾2𝑛𝑛 𝑢𝑢[𝑛𝑛] =� � 𝑢𝑢[𝑛𝑛]
𝛾𝛾1 − 𝛾𝛾2

Solving for the 1st term : 2−𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ (−0.2)𝑛𝑛 𝑢𝑢[𝑛𝑛]

2−𝑛𝑛 = (2−1 )𝑛𝑛 = (0.5)𝑛𝑛


(0.5)𝑛𝑛+1 − (−0.2)𝑛𝑛+1 0.5(0.5)𝑛𝑛 + 0.2(−0.2)𝑛𝑛
𝛾𝛾1𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 𝛾𝛾2𝑛𝑛 𝑢𝑢[𝑛𝑛] = � � 𝑢𝑢[𝑛𝑛] = � � 𝑢𝑢[𝑛𝑛]
(0.5) − (−0.2) 0.7
5 2
= (0.5)𝑛𝑛 𝑢𝑢[𝑛𝑛] + (−0.2)𝑛𝑛 𝑢𝑢[𝑛𝑛]
7 7

Solving for 2nd term :2−𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 4(0.8)𝑛𝑛 𝑢𝑢[𝑛𝑛]

2−𝑛𝑛 = (2−1 )𝑛𝑛 = (0.5)𝑛𝑛


(0.5)𝑛𝑛+1 − (0.8)𝑛𝑛+1 0.5(0.5)𝑛𝑛 + 0.8(0.8)𝑛𝑛
𝛾𝛾1𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 𝛾𝛾2𝑛𝑛 𝑢𝑢[𝑛𝑛] = 4 � � 𝑢𝑢[𝑛𝑛] = 4 � � 𝑢𝑢[𝑛𝑛]
(0.5) − (0.8) −0.3
3.2(0.8)𝑛𝑛 − 2(0.5)𝑛𝑛 32 20
=� � 𝑢𝑢[𝑛𝑛] = (0.8)𝑛𝑛 𝑢𝑢[𝑛𝑛] − (0.5)𝑛𝑛 𝑢𝑢[𝑛𝑛]
0.3 3 3

𝑦𝑦[𝑛𝑛] = 𝛾𝛾1𝑛𝑛 𝑢𝑢[𝑛𝑛] ∗ 𝛾𝛾2𝑛𝑛 𝑢𝑢[𝑛𝑛]


5 2 32 20
𝑦𝑦[𝑛𝑛] = (0.5)𝑛𝑛 𝑢𝑢[𝑛𝑛] + (−0.2)𝑛𝑛 𝑢𝑢[𝑛𝑛] + (0.8)𝑛𝑛 𝑢𝑢[𝑛𝑛] − (0.5)𝑛𝑛 𝑢𝑢[𝑛𝑛]
7 7 3 3
𝟐𝟐 𝟏𝟏𝟏𝟏𝟏𝟏 𝟑𝟑𝟑𝟑
𝒚𝒚[𝒏𝒏] = � (−𝟎𝟎. 𝟐𝟐)𝒏𝒏 − (𝟎𝟎. 𝟓𝟓)𝒏𝒏 + (𝟎𝟎. 𝟖𝟖)𝒏𝒏 � 𝒖𝒖[𝒏𝒏]
𝟕𝟕 𝟐𝟐𝟐𝟐 𝟑𝟑

Total Response
The total response of an LTID system can be express as a sum of the zero-input and zero-
state components:

𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑦𝑦[𝑛𝑛] = 𝑦𝑦0 [𝑛𝑛] + 𝑥𝑥 [𝑛𝑛] ∗ ℎ[𝑛𝑛]

Where

𝑦𝑦0 [𝑛𝑛] → zero-input response

𝑥𝑥 [𝑛𝑛] ∗ ℎ[𝑛𝑛] → zero-state response


Example

Find the total response of the DT LTI system described by 𝑦𝑦[𝑛𝑛 + 2] − 0.6𝑦𝑦[𝑛𝑛 + 1] −
25
0.16𝑦𝑦[𝑛𝑛] = 5𝑥𝑥[𝑛𝑛 + 2] given the initial conditions 𝑦𝑦[−1] = 0, 𝑦𝑦[−2] = , and input 𝑥𝑥 [𝑛𝑛] =
4
2𝑛𝑛 𝑢𝑢[𝑛𝑛]

Solution:
(𝐸𝐸2 − 0.6𝐸𝐸 − 0.16)𝑦𝑦[𝑛𝑛] = 5𝐸𝐸2 𝑥𝑥[𝑛𝑛]

(𝛾𝛾 2 − 0.6𝛾𝛾 − 0.16) = 0


(𝛾𝛾 + 0.2)(𝛾𝛾 − 0.8) = 0

Thus, 𝑦𝑦0 [𝑛𝑛] = 𝑐𝑐1 (−0.2)𝑛𝑛 + 𝑐𝑐2 (0.8)𝑛𝑛

Solving for 𝑐𝑐1 and 𝑐𝑐2


5
𝑛𝑛 = −1 : 0 = 𝑐𝑐1 (−0.2)−1 + 𝑐𝑐2 (0.8)−1 = 𝑐𝑐1(−5) + 𝑐𝑐2 � �
4
25 25
𝑛𝑛 = −2 : = 𝑐𝑐1(−0.2)−2 + 𝑐𝑐2(0.8)−2 = 𝑐𝑐1 (25) + 𝑐𝑐2 � �
4 16
1 4
Solving systems of linear equation with 2 unknowns, we obtain 𝑐𝑐1 = and 𝑐𝑐2 =
5 5

Therefore,
1 4
𝑦𝑦0 [𝑛𝑛] = (−0.2)𝑛𝑛 + (0.8)𝑛𝑛
5 5
From the previous examples,

ℎ[𝑛𝑛] = [(−0.2)𝑛𝑛 + 4(0.8)𝑛𝑛 ]𝑢𝑢[𝑛𝑛]


And
2 125 32
𝑥𝑥[𝑛𝑛] ∗ ℎ[𝑛𝑛] = � (−0.2)𝑛𝑛 − (0.5)𝑛𝑛 + (0.8)𝑛𝑛 � 𝑢𝑢[𝑛𝑛]
7 21 3
Therefore, the total response of the system is,
𝟏𝟏 𝟒𝟒 𝟐𝟐 𝟏𝟏𝟏𝟏𝟏𝟏 𝟑𝟑𝟑𝟑
𝒚𝒚[𝒏𝒏] = (−𝟎𝟎. 𝟐𝟐)𝒏𝒏 + (𝟎𝟎. 𝟖𝟖)𝒏𝒏 + � (−𝟎𝟎. 𝟐𝟐)𝒏𝒏 − (𝟎𝟎. 𝟓𝟓)𝒏𝒏 + (𝟎𝟎. 𝟖𝟖)𝒏𝒏 � 𝒖𝒖[𝒏𝒏]
𝟓𝟓 𝟓𝟓 𝟕𝟕 𝟐𝟐𝟐𝟐 𝟑𝟑
Natural and Forced Response

The characteristic modes of the previous system are (−0.2)𝑛𝑛 and (0.8)𝑛𝑛 The zero-input
component is made up of characteristic modes exclusively, as expected, but the characteristic
modes also appear in the zero-state response.
When all the characteristic mode terms in the total response are lumped together, the
resulting component is the natural response.

The remaining part of the total response that is· made up of non-characteristic modes is
the forced response

Example:
Simplifying further the previous equation, we obtain
𝟏𝟏𝟏𝟏 𝟏𝟏𝟏𝟏𝟏𝟏 𝟏𝟏𝟏𝟏𝟏𝟏
𝒚𝒚[𝒏𝒏] = (−𝟎𝟎. 𝟐𝟐)𝒏𝒏 + (𝟎𝟎. 𝟖𝟖)𝒏𝒏 − (𝟎𝟎. 𝟓𝟓)𝒏𝒏
𝟑𝟑𝟑𝟑 𝟏𝟏𝟏𝟏 𝟐𝟐𝟐𝟐
𝟏𝟏𝟏𝟏 𝟏𝟏𝟏𝟏𝟏𝟏
Natural response: (−𝟎𝟎. 𝟐𝟐)𝒏𝒏 + (𝟎𝟎. 𝟖𝟖)𝒏𝒏
𝟑𝟑𝟑𝟑 𝟏𝟏𝟏𝟏

𝟏𝟏𝟏𝟏𝟏𝟏
Forced response: − (𝟎𝟎. 𝟓𝟓)𝒏𝒏
𝟐𝟐𝟐𝟐

References:
 Principles of Linear Systems and Signals 2nd Edition by B.P. Lathi
 Fundamentals of Signals and Systems by Benoit Boulet
 Signals and Systems by Richard Baranuik
 https://www.youtube.com/user/agpanagos
CRITIQUE

Title: On a Fractional-order Backward Difference Equation Solution solved by a

Microcontroller.

The paper focuses on the analysis of the fractional-order backward difference and linear

time-invariant difference equations which was realized by implementing the solution in a

microcontroller and was compared using Matlab. These analysis proposed solutions to address the

“calculation tail” problems which was experienced by microcontrollers that causes errors in real-

time calculation due to the linearly growing number of multiplications and additions while

performing these operations.

According to Miller et. al (1993) and Podlubny (1999), as cited on the paper by Matusiak

et. al (2018), discrete fractional calculus has a significant role and become an effective

mathematical tool in the identification, analysis and synthesis of real dynamical systems in various

fields of science and engineering. However, sophisticated computational algorithms cause errors

in real-time calculations of microcontroller. Therefore, one should pay attention to these complex

algorithms for better mathematical models.

The concept of analyzing fractional-order backward difference equation using a

microcontroller is an excellent idea because it discusses on finding solutions on how to address

the “growing tail calculations” and on how to improve mathematical models for complex

calculation algorithms implemented in a microcontroller. The main problem found was that the

microprocessor adds additional operations in each step of calculation thus leading to a situation

where it will not be able to perform desired operations within the specified sampling period. As a
solution, the author suggests to apply some simplifying assumptions as remedy. The paper

proposes two remedies for these problems as discussed in the paper.

The paper develops a mathematical model t first to address these problems then analyzed

the results using a microcontroller. The limitations were identified and taken into consideration

especially the necessary precision for floating point values. Floating point values are important

because large number of calculations might not get a valid representation in the microprocessor

memory. However, decision of floating-point type leads to another limitation related to available

memory size. Finally, the last problem identified is related to the sampling time. The paper

analyzed these problems using STM32F746ZGT6 (by STMicroelectronics) microcontroller.

In conclusion, the paper determined that to avoid a problem of the fractional-order

difference/sum realization by a microprocessor caused by a lack of a memory and growing

calculation time one should apply “optimized” algorithms.

References
Matusiak, M., Sankowski, D., & Ostalczyk, P. (2018). On a Fractional-order Backward
Difference Equation Solution Solved by a Microcontroller. 2018 International
Interdisciplinary PhD Workshop (IIPhDW) (pp. 330-334). Swinoujście, Poland: IEEE.
doi:10.1109/IIPHDW.2018.8388384

S-ar putea să vă placă și