Ch3.
4: Z-Transform
Information source
and input transducer
Source Coding
Ch3.4:
z-Transform
Channel Coding
Modulator
Questions to be answered:
z-Transform: A general version of DTFT
Region of Convergence: Where the
transform exists.
Channel
Inverse z-Transform: Back to timedomain
LTI System in Transform Domain: A
general version of the frequency
response function
Information sink
and output transducer
Elec3100 Chapter 3.4
Source Decoding
Channel Decoding
Demodulator
(Matched Filter)
1
Ch3.4: z-Transform
Definition of z-Transform
Region of Convergence
Inverse z-Transform
Z-Transform Properties
LTI System in Transform Domain
Elec3100 Chapter 3.4
Generalizing DTFT
The DTFT provides a frequency-domain representation of discretetime signals and LTI discrete-time systems.
However, because of the convergence condition, in many cases,
the DTFT of a sequence may not exist.
As a result, it is not possible to make use of such frequency-domain
characterization in these cases.
Question: Can we generalize DTFT?
Elec3100 Chapter 3.4
z-Transform
The z-transform X(z) of a sequence x[n] is defined as
where z is a continuous complex variable. Generally, we can
where r =
express the complex variable z in polar form as
|z| > 0 is the magnitude and is the angle of z.
z-transform gives
In particular, when r = 1, then
becomes
Elec3100 Chapter 3.4
and the above expression
z-Transform
The relationship between
and
can also be
illustrated in the z-plane (a real-imaginary plane for complex
number).
The z-transform evaluated on the
unit circle (as varies from 0 to
2) corresponds to the DTFT of
x[n].
Elec3100 Chapter 3.4
z-Transform
Main advantages of using z-transform over DTFT:
z-transform can encompass a broader class of signals than
DTFT
Recall that a sufficient condition for convergence of the DTFT is:
|
Example: The DTFT of
0.5
1 does not exist
since it is unbounded in the negative direction:
-5 -4 -3 -2 -1 0
-8
-16
-4 -2
-32
Elec3100 Chapter 3.4
z-Transform
On the other hand, the z-transform exists if
We can choose region of convergence (ROC) for z such that the
z-transform converges.
Example: z-transform of
Proof:
0.5
0.5
0.5
0.5
1 exists if 0.5
.
.
Elec3100 Chapter 3.4
z-Transform
Other advantages of using z-transform:
More convenient notation than DTFT when dealing with
analytical problems
Convenient block diagram representation for
implementation of practical systems, Eg., a unit delay
system is expressed as:
x[n]
z-1
y[ n] = x[n 1]
Useful for determining and solving difference
equations for discrete-time systems
Elec3100 Chapter 3.4
Application: Filter Design
The convolution sum description of an LTI discrete-time system with
.
an impulse response
is given by
The output to an input
For any given input sequence , the output will be
indicating that
is an eigen-function for any LTI system.
How can we utilize this property?
Elec3100 Chapter 3.4
is given by
Use of z-Transform: Filter Design
Can you figure out what kind of filter has the following ztransform and magnitude spectrum?
Elec3100 Chapter 3.4
10
Ch3.4: z-Transform
Definition of z-Transform
Region of Convergence
Inverse z-Transform
Z-Transform Properties
LTI System in Transform Domain
Elec3100 Chapter 3.4
11
Region of Convergence (ROC)
Since z can be any point on the z-plane, generally there exists some
z which makes X(z) not converge
|
The set of z values for which X(z) converges
is called the region of convergence (ROC).
The ROC must be specified along with X(z) in order for the ztransform to be completely defined.
Elec3100 Chapter 3.4
12
Poles and Zeros in z-Transform
In many applications, X(z) can be expressed as a rational function:
M
P( z )
X ( z) =
=
Q( z )
b z
a z
k =0
N
b ( z c1 )(z c2 )L (z cM ) b0
= 0
=
a0 ( z d1 )( z d 2 )L (z d N ) a0
k =0
(z c )
k
k =1
N
(z d )
a0 0, b0 0
k =1
P(z) and Q(z) are numerator and denominator polynomials in z
The values of z for which X(z) = 0, i.e. roots
of P(z), are called zeros of X(z)
The values of z for which X(z) = , i.e. roots
of Q(z), are called poles of X(z)
Elec3100 Chapter 3.4
13
Properties of the ROC
The ROC is ring or disk in the z-plane centered at the origin, i.e.
0
| |
The DTFT of
converges absolutely iff the ROC of the ztransform includes the unit circle.
The ROC cannot contain any poles.
Example: The z-transform
is given by
.
of the sequence
0.6
0.6
Here, the ROC is just outside the circle going
through the point z
0.6.
Elec3100 Chapter 3.4
14
ROC of Finite-Length Sequence
If
is a finite-duration sequence, i.e. a sequence that is
, then
zero except over a finite interval
the ROC is the entire z-plane, except z = 0.
Example: Consider a finite-length sequence
defined for
M n N, where M and N are non-negative integers and
|
|
Its z-transform is given by
Note:
has M poles at
and N poles at
0.
Thus, the ROC is the entire z-plane except possibly at
and/or at
Elec3100 Chapter 3.4
N1
N2
15
ROC of Right-Sided Sequence
A right-sided sequence is a sequence with nonzero sample values
for
. A right-sided sequence with nonzero sample values for
0 is called a causal sequence.
Consider a causal sequence
converges exterior to a circle
,
including the point
.
On the other hand, a right-sided sequence
with nonzero
sample values only for
with M nonnegative has a ztransform
with M poles at
.
is exterior to a circle
, excluding the point
The ROC of
.
Elec3100 Chapter 3.4
, whose z-transform is given by
16
Example 5.1 Right-Sided Sequence
Determine the z-transform of
converges if
1
. This requires
| |, and gives
Note: If |a| < 1, the ROC contains the
unit circle and hence the DTFT of this
sequence x[n] exists.
Shaded area = ROC
Remark: u[n] has z-transform, but not DTFT
Elec3100 Chapter 3.4
x denotes as the pole of X(z)
o denotes as the zero of X(z)
17
ROC of Left-Sided Sequence
A left-sided sequence is a sequence with nonzero sample values for
. A left-sided sequence with nonzero sample values for
0
is called an anticausal sequence.
Consider an anticausal sequence
, whose z-transform is given
by
It can be shown that
converges interior to a
circle
, including the point
0.
On the other hand, a left-sided sequence
with nonzero sample
values only for
with N nonnegative has a z-transform
with N poles at
0.
is interior to a circle
, excluding the point
The ROC of
0.
Elec3100 Chapter 3.4
18
Example 5.2 Left-Sided Sequence
Determine the z-transform of another signal
converges if
1
. This requires
| |, and gives
Note: If |a| < 1, the ROC does not
contain the unit circle and the DTFT
of this sequence x[n] does not exist.
x denotes as the pole of X(z)
o denotes as the zero of X(z)
Elec3100 Chapter 3.4
19
Observation from Example 5.1 and 5.2
Different signals can give same z-transform, although the ROCs will
differ
x[n] = a n u[n]
x[n] = a nu[n 1]
1
X ( z) =
1 if z > a
1 az
X ( z) =
( ) exists if a < 1
( ) does not exist if a > 1
X e j
X e j
Elec3100 Chapter 3.4
1
if z < a
1
1 az
( ) exists if a > 1
( ) does not exist if
X e j
X e j
a <1
20
ROC of Two-Sided Sequence
The z-transform of a two-sided sequence
The first term on the RHS can be interpreted as the z-transform of a
right-sided sequence and it thus converges exterior to the circle
.
The second term on the RHS can be interpreted as the z-transform
of a left-sided sequence and it thus converges interior to the circle
.
, there is an overlapping ROC
If
given by
| |
. Otherwise, the
z-transform does not exist.
Elec3100 Chapter 3.4
can be expressed as
21
Example 5.3 Two-Sided Sequence
Consider the two-sided sequence
transform is given by
The first term on the RHS converges for
.
second term converges for
, whose z-
whereas the
There is no overlap between two regions.
Hence, the z-transform does not exist.
Elec3100 Chapter 3.4
22
Example 5.4 Sum of Two Sequences
Consider a signal that is the sum of two real sequences:
n
1
1
x[n] = u[n] + u[n]
2
3
Using the result in Example 5.1,
n
1
1
Z
x1 [n] = u[ n]
X 1 ( z) =
1 12 z 1
2
z >
if
1
2
1
1
Z
x2 [n] = u[n]
X 2 ( z) =
1 + 13 z 1
3
X ( z) =
1 12 z 1 1 + 13 z 1
1
2 z z 12
=
z 12 z + 13
Elec3100 Chapter 3.4
)(
if
if
z >
1
3
1
1
1
z > I z > z >
3
2
2
The overall ROC is the region
of overlap of the ROC for X1(z)
and the ROC for X2(z); unless
there is pole-zero cancellation.
23
Table 3.4.1:
Common
z-transform
pairs
Elec3100 Chapter 3.4
24
Ch3.4: z-Transform
Definition of z-Transform
Region of Convergence
Inverse z-Transform
Z-Transform Properties
LTI System in Transform Domain
Elec3100 Chapter 3.4
25
The Inverse z-Transform
There are 4 commonly used techniques to evaluate the inverse ztransform. They are
1. Inspection Method
2. Partial Fraction Expansion
3. Power Series Expansion
4. Cauchy Integral Theorem
This is the formal inverse z-transform expression based on
Contour integral (an integration techniques together with
right-hand rule) onto z-plane.
We will not cover this method in this course.
Elec3100 Chapter 3.4
26
Inverse z-transform: Inspection Method
Inspection Method: Familiar with, or recognizing by inspection,
certain transform pairs.
Example:
Find the inverse z-transform of
This gives
X ( z) =
z
X ( z) =
z ( 12 )
z>
1
2
1
z>
2
1 (12 )z 1
By making use of the transform pair in Table 3.4.1,
n
1
x[n] = u[n]
2
Elec3100 Chapter 3.4
27
Partial Fraction Expansion Method
A rational
If
can be expressed as
, then
can be re-expressed as
where the degree of
is less than N. The rational function
is
called a proper fraction.
Elec3100 Chapter 3.4
28
Inverse Transform: Partial-Fraction Expansion
Simple Poles: In most cases, the rational z-transform of interest
is a proper fraction with simple poles. Let the poles of
be
at
,1
. A partial-fraction expansion of
is then of
the form
1
The constants
Each term of the sum in partial-fraction expansion has an ROC
given by
and, thus has an inverse transform of the form
. Therefore, the inverse transform of G(z) is given by
Reminder: There are cases with multiple poles.
Elec3100 Chapter 3.4
are called the residues and given by
1
|
29
Inverse Transform: Partial-Fraction Expansion
Example 5.5: Determine
whose z-transform is given by
2
1 2
1 0.2
1 0.6
0.2
0.6
Solution: A partial-fraction expansion of
1
0.2
is of the form
0.6
Given
we have
Elec3100 Chapter 3.4
.
.
.
.
. Thus,
30
Inverse Transform by Long Division
The z-transform G(z) of a sequence
can be expanded in a
power series in
. In the series expansion, the coefficient
multiplying the term
is then the n-th sample of
.
For a rational z-transform expressed as a ratio of polynomials in
the power series expansion can be obtained by long division.
Example Consider
1
2
0.12
Long division of the numerator by the denominator yields
1
1
0.4
1.6
0.52
0.4
1 1.6 0.52 0.4
As a result:
result with that of Example 5.5.
Elec3100 Chapter 3.4
0.2224
0.2224 . Compare this
31
Ch3.4: z-Transform
Definition of z-Transform
Region of Convergence
Inverse z-Transform
Z-Transform Properties
LTI System in Transform Domain
Elec3100 Chapter 3.4
32
z-Transform Properties
Linearity
Z
Z
If x1[n]
X 1 ( z ), ROC = Rx1 and x2 [n ]
X 2 ( z ),
ROC = Rx
Z
then ax1 [n] + bx2 [n]
aX 1 ( z ) + bX 2 ( z ),
ROC contains Rx I Rx
1
Example: Consider the two-sided sequence
1.
and
Two sequences
,
| | and
transforms
With linearity property, we arrives at
1
ROC is given by the overlap regions of
| |?
Question: What happens if
Elec3100 Chapter 3.4
1 have z,
| |
1
1
| | and
| |.
33
z-Transform Properties
Time Shifting
Z
x[n n0 ]
z n0 X ( z ),
ROC = Rx
ROC is unchanged, except for the possible addition or deletion of
n
the points z = 0 or z = . It is because the factor z 0 can alter the
number of poles at z = 0 or z =.
Multiplication by an Exponential Sequence
Z
z 0n x[n]
X ( z z0 ),
ROC = z 0 Rx
where ROC | | denotes that the ROC
is scaled by | |; ie.
if
is the set of values of z such that rR < |z| < rL, then the new
ROC is the set of values of z such that |z0|rR < |z| < |z0| rL
Elec3100 Chapter 3.4
34
z-Transform Properties
Differentiation
dX ( z )
,
nx[n] z
dz
Z
ROC = Rx
ROC is unchanged except for the possible addition or deletion of z =
0.
Conjugation of a Complex Sequence
Z
x [n ]
X ( z ),
Z
Time Reversal x[ n ]
X (1 z ),
ROC = Rx
ROC = 1 Rx
where ROC 1/ denotes that the ROC
is inverted; i.e. if is
the set of values of z such that rR < |z| < rL, then the ROC is the set
of values of z such that 1/rL < |z| < 1/rR
Elec3100 Chapter 3.4
35
Example 5.6: Time-Reversal
Find the z-transform of x[ n] = a nu[ n]
Solution:
1
,
a u[ n]
z>a
1
1 az
1
a 1 z 1
Z
n
=
a u[ n]
,
1 1
1 az 1 a z
n
Elec3100 Chapter 3.4
z < a 1
36
z-Transform Properties
Convolution of Sequences
Z
Z
If x1 [n]
X 1 ( z ),
ROC = Rx1 and x2 [n ]
X 2 ( z ),
Z
then x1 [n] x2 [n] X 1 ( z ) X 2 ( z ),
ROC = Rx2
ROC contains Rx1 I Rx2
The resulting ROC contains at least the intersection of Rx1 and Rx2
If there is no pole-zero cancellation in the linear
combination, the ROC is exactly equal to Rx1 I Rx2 .
If there is pole-zero cancellation in the linear combination,
the ROC may be larger.
Initial-Value Theorem
If x[n] is zero for n < 0 (i.e. if x[n] is causal), then x[0] = lim X ( z )
z
Elec3100 Chapter 3.4
37
Example 5.7: Convolution
Find the convolution sum of
1.
and
, where
Solution:
Method 1: direct convolution
Z 1
Method 2: Y ( z ) = X 1 ( z ) X 2 ( z ) y[ n] = x1 [n ] x2 [n ]
1
1
X 1 ( z) =
,
z
>
a
and
X
(
z
)
=
,
z >1
2
1
1
1 az
1 z
This gives
Y ( z) =
1
1
z >1
1 az 1 z
1 1
a
=
1
1 a 1 z
1 az
y[n] =
Elec3100 Chapter 3.4
1
u[n] a n+1u[n]
1 a
38
Table 3.4.2:
z-transform
properties
Elec3100 Chapter 3.4
39
Ch3.4: z-Transform
Definition of z-Transform
Region of Convergence
Inverse z-Transform
Z-Transform Properties
LTI System in Transform Domain
Elec3100 Chapter 3.4
40
LTI Systems in the Transform Domain
An LTI discrete-time system is completely characterized in the timedomain by its impulse response sequence
.
Each LTI discrete-time
representation.
Such transform-domain representations provide additional insight
into the behavior of such systems. It is easier to design and
implement these systems in the transform-domain for certain
applications.
We consider now the use of the DTFT and the z-transform in
developing the transform domain representation of an LTI system.
Elec3100 Chapter 3.4
system
has
transform-domain
41
The Transfer Function
Transfer Function: A generalization of the frequency response
function.
The convolution sum description of an LTI discrete-time system with
.
an impulse response
is given by
Taking the z-transform of both sides, we get
/
is the z-transform of
Here,
transfer function or the system function.
Elec3100 Chapter 3.4
and is called the
42
FIR vs. IIR Filter
A finite-duration impulse response (FIR) filter has a impulse
response
bn 0 n M 1
h ( n) =
0
else
The difference equation representation is
y[ n] = b0 x[ n] + b1x[ n 1] + L + bM 1 x[ n M + 1]
which is a linear convolution of finite support.
IIR filters are characterized by infinite-duration impulse response.
The difference equation representation of an IIR filter is expressed as
y[n] =
Elec3100 Chapter 3.4
M 1
N 1
m=0
m =1
bm x[n m] am y[n m]
43
LTI Systems in the Transform Domain
Consider an LTI discrete-time system characterized by linear
constant coefficient difference equations of the form:
.
Applying the z-transform to both sides of the difference
equation, we arrive at
where
and
denote the z-transforms of
and
with associated ROCs, respectively.
A more convenient form of the z-domain representation of
the difference equation is given by
Elec3100 Chapter 3.4
44
LTI Systems in the Transform Domain
Thus,
, ,,
are finite zeros, and , , ,
are finite poles.
If
, there are additional
zeros at
0.
If
, there are additional
poles at
0.
For a causal digital filter, the impulse response is a causal
sequence. The ROC of the causal transfer function is exterior to a
circle going through the pole furthest from the origin with
max | |
Elec3100 Chapter 3.4
45
Example 5.8: Moving Average Filter
Consider the M-point moving-average FIR filter with an impulse
,0
1
response,
0, otherwise
Its transfer function is given by
1
1
1
The transfer function has M zeros on the unit circle at
/ ,0
1.
There is an
1)th order pole at
0
and a single pole at
1.
The ROC is the entire z-plane except
0
Elec3100 Chapter 3.4
46
Example 5.8: Moving Average Filter
Elec3100 Chapter 3.4
47
Frequency Response from Transfer Function
If the ROC of the transfer function H(z) includes the unit circle, then
the frequency response of the LTI digital filter can be obtained
|
simply as follows:
For a stable rational transfer function in the form
the frequency response is given by
It is convenient to visualize the contributions of the zero factor and
the pole factor from the factored form of the frequency response.
Elec3100 Chapter 3.4
48
Frequency Response from Transfer Function
The magnitude response function is given by:
|
|
| | ||
|
|
|
| |
|
|
|
|
The phase response is of the form
arg
arg
arg
How can we use this?
Elec3100 Chapter 3.4
49
Geometric Interpretation
The factored form of frequency response
is convenient to develop a geometric interpretation of the
frequency response computation from the pole-zero plot as varies
from 0 to 2 on the unit circle.
A typical factor in the factored form of the frequency response is
where
is a zero or pole.
given by
It can be observed from the right figure that
represents a vector starting at
z
and ending on the unit circle
.
Elec3100 Chapter 3.4
50
Geometric Interpretation
As indicated by |
|
|
|
|
, the magnitude
response at a specific frequency
is given by the product of the
magnitudes of all zero vectors divided by the product of the
magnitudes of all pole vectors.
Question: When can we obtain the smallest magnitude?
A zero (pole) vector has the smallest magnitude when
To attenuate signal components in a specified frequency range, we
need to place zeros very close to or on the unit circle in this range.
To emphasize signal components in a
specified frequency range, we need to
place poles very close to or on the unit
circle in this range.
Elec3100 Chapter 3.4
51
Can we build an ideal bandpass filter?
Elec3100 Chapter 3.4
52
Can we build an ideal bandpass filter?
The z-transform of an FIR impulse response can be expressed as a
simple polynomial P(z) of degree L - 1 where L is the number of
nonzero taps of the filter.
The fundamental theorem of algebra states that such a polynomial
has at most L - 1 roots. Thus, the frequency response of an FIR filter
can never be identically zero over a frequency interval since, if it
were, its z-transform would have an infinite number of roots.
The argument can be easily extended to rational transfer functions,
confirming the impossibility of a realizable filter whose characteristic
is piecewise perfectly flat.
Then, how can we design a bandpass filter? (To be continued)
Elec3100 Chapter 3.4
53
Stability Condition
A causal LTI digital filter is BIBO stable if and only if its impulse
response h[n] is absolutely summable, i.e., S
|
| .
Thus, an FIR digital filter with bounded impulse response is always
stable.
Question: What is the condition in terms of pole locations of
?
The ROC of the
for the impulse response sequence
is
defined by values of
where
|
| . Thus, if the
digital filter is stable, the ROC includes the unit circle.
Elec3100 Chapter 3.4
54
Stability Condition in Terms of Pole Position
Consider the causal IIR digital filter with a rational transfer function
H(z) given by
. Its impulse response is right-sided
sequence.
The ROC of the
is exterior to a circle going through the pole
furthest from z 0. But stability requires ROC includes the unit
cycle.
Conclusion: All poles of a causal stable transfer function H(z) must
be strictly inside the unit circle.
The stability region (shown shaded) in the z-plane is shown below
Elec3100 Chapter 3.4
z-Transform
55
Example 5.9: Stability Condition IIR Filter
Elec3100 Chapter 3.4
56
