You are on page 1of 52

ECTE301: Digital Signal Processing

Lecture 4
The z-transform: Inverse Transform & Applications
Part A's contents
Textbook
Lecture Topic Practice
Chapter
Lab 1
L1 Introduction to DSP 1, 2.1, 2.2
Tutorial 1
Lab 2
L2 Convolution and LTI Systems 2.3, 2.4
Tutorial 2
The z-transform: Tutorial 3
L3 3.1, 3.2, 3.3
Forward Transform and Properties Lab 3
The z-transform:
L4 3.4, 3.5 Tutorial 4
Inverse Transform and Applications
2.4, 2.5,
Random signals and Tutorial 4
L5 2.6, 12.1
Cross-correlation Sequences Report

L6 Revision of Weeks 1 to 6 - -

Digital Signal Processing S. L. Phung & A. Bouzerdoum 2/52


Lecture 4's objectives

 Learn techniques to compute the inverse z-transform.

 Use the z-transform to analyse LTI systems.

 Learn the concept of complex spectral density.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 3/52


Lecture 4's sequence

4.1 The inverse z-transform

4.2 The z-transform and LTI systems

4.3 Complex spectral density

Digital Signal Processing S. L. Phung & A. Bouzerdoum 4/52


4.1 The inverse z-transform
 The inverse z-transform of X(z) is defined as

 Here, C is a counter-clockwise closed contour in the ROC of X(z).

 The inverse z-transform can be derived from the Cauchy integral


theorem:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 5/52


The inverse z-transform

There are three common methods for finding the inverse z-transform.

 Inspection method [Section 4.1.1]

 Power series method [Section 4.1.2]

 Partial-fraction expansion method [Section 4.1.3]

Digital Signal Processing S. L. Phung & A. Bouzerdoum 6/52


4.1.1 Inspection method

 Use Table 2 of common z-transform pairs to find directly signal x(n) for
the standard z-domain functions.

 Use Table 1 of z-transform properties to relate a given X(z) to the


standard z-domain functions.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 7/52


Table 2: Common z-transform pairs

Digital Signal Processing S. L. Phung & A. Bouzerdoum 8/52


Table 1: Properties of the z-transform

Digital Signal Processing S. L. Phung & A. Bouzerdoum 9/52


Inspection method

 Frequently used properties

● Linearity:

● Time-reversal:

● Time-shifting:

● Scaling:

● Differentiation:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 10/52


Inspection method
 Example 4.1: Find the inverse z-transform of

 Answer:
Using Table 2, we know that

Therefore, x(n) can be obtained directly as

Digital Signal Processing S. L. Phung & A. Bouzerdoum 11/52


Inspection method
 Example 4.2: Find the inverse z-transform of

 Answer:
According to the time-shifting property (Table 1):

Using Table 2, we know that (Example 4.1):

Therefore, we have:
Shifting by k = 1 sample!

Digital Signal Processing S. L. Phung & A. Bouzerdoum 12/52


4.1.2 Power series method
 If x(n) is causal sequence, then

 If we can express X(z) as a power series

then, by matching coefficients, we get

 This method is suitable if the power series of X(z)


● has only a few terms, or
● can be found easily using Taylor series expansion, or
● can be found easily using long-division.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 13/52


Power series method: Direct terms
 Example 4.3: Find the inverse z-transform of

 Answer: We have
X(z) has only
a few terms

Therefore,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 14/52


Power series method: Taylor series
 Example 4.4: Find the inverse z-transform of

 Answer: Note the Taylor series expansion:


ECTE301-Formula-
List-Part-A.pdf
We can write
X(z) has an easy-to-find
Taylor series

Hence,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 15/52


Power series method: Taylor series
 Example 4.5: Find the inverse z-transform of

 Answer: Note the Taylor series expansion:

We can write

Hence,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 16/52


Power series method: Taylor series
 Example 4.6: Find the inverse z-transform of

 Answer: We use the Taylor series expansion, ECTE301-Formula-


List-Part-A.pdf

Then,

Hence,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 17/52


Power series method: Long division
 Example 4.7: Find a causal sequence x(n) so that

 Answer:

● This method is applicable when X(z) is the ratio of two polynomials of z-1.

● Arrange the polynomials in the increasing power of z-1.

● Find the quotient polynomial, using long division.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 18/52


Power series method: Long division
Performing ECTE301-Formula-
long division List-Part-A.pdf


Hence,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 19/52


4.1.3 Partial fraction expansion method
 This method is applicable when X(z)/z can be expressed as the ratio of
two polynomials in z:

 There are two cases that we should consider.

 X(z) has simple poles: e.g. X(z) has two simple poles z = 1 and z = 0.5.

 X(z) has repeated poles: e.g. X(z) has a pole z = 1 with multiplicity m = 2.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 20/52


Partial fraction expansion method
 Steps when X(z) has only simple poles
● Express X(z)/z as a ratio of two polynomials in z.

● Perform partial fraction expansion to write X(z)/z as

where ai can be found as

● Write X(z) as a sum of simple terms:

● Finally, obtain x(n) using Table 2:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 21/52


Partial fraction expansion method
 Example 4.8: For causal x(n), find the inverse z-transform of

 Answer A:
● Express X(z)/z as a ratio of two polynomials in z:

● Perform the partial fraction expansion of X(z)/z:


ECTE301-Formula-
List-Part-A.pdf

Digital Signal Processing S. L. Phung & A. Bouzerdoum 22/52


Partial fraction expansion method
● Write X(z) as the sum of simple terms.

● Use Table 2 to obtain the final form of x(n)

● Verify x(n) by checking the value of x(0) and x(∞).


 According to (*), x(0) = 1 and x(∞) = 1.4.
 According to the initial value theorem:

 According to the final value theorem:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 23/52


Partial fraction expansion method
 Answer B:
● This fast technique is applicable when X(z) is given in terms of z-1.

● We can expand X(z) directly as ECTE301-Formula-


List-Part-A.pdf

● Then, find a1 and a2 by

● Therefore,

Digital Signal Processing S. L. Phung & A. Bouzerdoum 24/52


Partial fraction expansion method
 Steps when X(z) has repeated poles
● Suppose that D(z) has a root p with multiplicity of m, where m > 1.

● Expand X(z)/z as terms contributed by pole p

● Calculate the coefficients am, am-1, …, a1 as

● Write solution x(n).

Digital Signal Processing S. L. Phung & A. Bouzerdoum 25/52


Partial fraction expansion method
 Example 4.9: Find a causal sequence x(n) with the following z-transform

 Answer:
● We expand X(z)/z as

● Calculate the coefficients:

Calculus

Digital Signal Processing S. L. Phung & A. Bouzerdoum 26/52


Partial fraction expansion method
 Answer (continued):
● Therefore,

● Comparing with Table 2 of z-transform pairs, we get

Digital Signal Processing S. L. Phung & A. Bouzerdoum 27/52


Lecture 4's sequence

4.1 The inverse z-transform

4.2 The z-transform and LTI systems

4.3 Complex spectral density

Digital Signal Processing S. L. Phung & A. Bouzerdoum 28/52


4.2 The z-transform and LTI systems
 For a discrete-time LTI system, let H(z) be the z-transform of the
impulse response h(n):

 H(z) is called the pulse transfer function or system function. It completely


characterizes the system.

 H(z) is the ratio of Y(z) and X(z).

h(n)

H(z)

Digital Signal Processing S. L. Phung & A. Bouzerdoum 29/52


The z-transform and LTI systems

4.2.1 Finding unknown signals x(n), h(n), y(n)

4.2.2 Poles, zeros, ROC, stability

4.2.3 z-transform & difference equation

4.2.4 z-transform & frequency response

Digital Signal Processing S. L. Phung & A. Bouzerdoum 30/52


4.2.1 Finding unknown signals x, h, y

h(n)

H(z)

 Knowing two of the three signals {X(z), H(z), and Y(z)},


we can calculate the remaining signal.

● Finding output y(n):

● Finding impulse response h(n):

● Finding input x(n):

Digital Signal Processing S. L. Phung & A. Bouzerdoum 31/52


The z-transform and LTI systems
 Example 4.10: Compute the output y(n) of the system (a ≠ b):
LTI

 Answer:
● First, we find:

● Next, we perform the partial fraction expansion on Y(z):

● Finally, we can write:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 32/52


4.2.2 Poles, zeros, ROC, stability
 Poles: The values of z where H(z) goes to ∞.

 Zeros: The values of z where H(z) = 0.

 ROC: Region on the z-plane where H(z) is valid.

 Example 4.11: Consider a causal system with

● Poles: p1 = 1, p2 = 0.5.

● Zeros: z1 = 0.3, z2 = 0.

● ROC: |z| > 1.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 33/52


Stability
 Theorem 1: A LTI system is stable if its ROC contains the unit circle.

 Theorem 2: A causal LTI system is stable if all its poles are inside the
unit circle.

Im{z}
× denotes a pole.
×
×
1 Re{z}
z-plane

Digital Signal Processing S. L. Phung & A. Bouzerdoum 34/52


Stability
 Example 4.12: Determine the stability of the following causal systems:

 Answer:
a) Find poles:

Because |p1| < 1 and |p2| < 1, system H1(z) is stable.

b) Find poles:

Because |p1| > 1, system H2(z) is unstable.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 35/52


4.2.3 The z-transform & difference equation
 Consider an LTI system given by the difference equation:

 Applying the z-transform to both sides of equation:

 The system function is

Digital Signal Processing S. L. Phung & A. Bouzerdoum 36/52


The z-transform and difference equation

 If ak = 0 for all k, we have an all-zero system:

 This is a FIR system:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 37/52


The z-transform and difference equation

 If bk = 0 for k = 1, 2, …, M, we have an all-pole system:

 This is called an autoregressive (AR) system:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 38/52


4.2.4 The z-transform & frequency response

 Consider an LTI system with an impulse response of h(n) and system


function H(z).

 The frequency response H(jω) is the system function H(z) evaluated on


the unit circle:

Im{z}
1

 Here, ω is an angle between [0, 2π]. !


Re{z}
z-plane

Digital Signal Processing S. L. Phung & A. Bouzerdoum 39/52


Magnitude and phase responses
 In general, H(jω) is a complex number.

real part imaginary part

 The magnitude response of the system is defined as

 The phase response of the system is defined as

Digital Signal Processing S. L. Phung & A. Bouzerdoum 40/52


Frequency response of a LTI system
 Example 4.13: Find the frequency response of the system

 Answer: In z-domain, the system can be expressed as

Therefore, the system function is

We substitute z = ejω to obtain the frequency response:

Digital Signal Processing S. L. Phung & A. Bouzerdoum 41/52


Frequency response of a LTI system

Magnitude Response
0.4

0.35

0.3

0.25

0.2
-6 -4 -2 0 2 4 6

Phase Response
2

-2

-4
-6 -4 -2 0 2 4 6
ω
Plot of magnitude and phase responses.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 42/52


Steady-state output
 Consider a sinusoidal input with a frequency ω0:

 The steady-state output of the system is

 Conclusion:

● output frequency = input frequency,


● output magnitude = input magnitude × |H(jω0)|,

● output phase = input phase + angle{Hjω0)}.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 43/52


Steady-state output
 Example 4.14: Find steady-state response of the system in Example 4.13
for the input signal

 Answer: We already find that

For input signal x(n), frequency ω0 = 0.1π. Therefore,

The steady-state output is

Digital Signal Processing S. L. Phung & A. Bouzerdoum 44/52


Steady-state output

x[n] 0

-1
transient 0 10 20 30 40 50 60
response steady-
0.5 state
y[n]

-0.5
0 10 20 30 40 50 60
n

Digital Signal Processing S. L. Phung & A. Bouzerdoum 45/52


Lecture 4's sequence

4.1 The inverse z-transform

4.2 The z-transform and LTI systems

4.3 Complex spectral density

Digital Signal Processing S. L. Phung & A. Bouzerdoum 46/52


4.3 Complex spectral density

 The complex spectral density (CSD) of a sequence x(n) is the z-transform


of the auto-correlation rxx(l):

 The complex cross-spectral density of sequences x(n) and y(n) is the


z-transform of the cross-correlation rxy(l):

Digital Signal Processing S. L. Phung & A. Bouzerdoum 47/52


Complex spectral density
 The CSD of a real signal x(n) is given by:
Because

 The cross CSD of x(n) and y(n) is given by:


Because

 It follows that

Digital Signal Processing S. L. Phung & A. Bouzerdoum 48/52


Complex spectral density
 Example 4.15: The CSD function of a signal is given by

Find autocorrelation rxx(l) and power spectral density Pxx(ω).

 Answer: Applying the power fraction expansion:

The ROC of Rxx(z) is 0.8 < |z| < 1.25. Therefore,

The power spectral density is

Digital Signal Processing S. L. Phung & A. Bouzerdoum 49/52


Input-output correlation sequences

h(n)

h(l)

h(l)

rhh(l)

Transformation of correlation sequences by a LTI system.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 50/52


Input-output correlation sequences

H(z)

H(z)

H(z)

H(z) H(z-1)

Transformation of spectral densities by a LTI system.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 51/52


Lecture 4's summary
To revise for this lecture:

 Learn three techniques to compute the inverse z-transform.

 Use the z-transform to analyse LTI systems.

 Compute unknown signals x(n), y(n) or h(n) given the two others.

 Calculate system zeros, poles, ROC, stability, frequency response.

 Practice: Tutorial 4 exercises.

Digital Signal Processing S. L. Phung & A. Bouzerdoum 52/52

You might also like