2/17/2020

The Z-Transform

Quote of the Day

Such is the advantage of a well-constructed
language that its simplified notation often
becomes the source of profound theories.

Laplace

The z-Transform
• Counterpart of the Laplace transform for discrete-time signals
• Generalization of the Fourier Transform
– Fourier Transform does not exist for all signals
• The z-Transform is often time more convenient to use
• Definition:

Xz    xn z
n  
n

• Compare to DTFT definition:

   xn e

X e j   jn

n  

• z is a complex variable that can be represented as z=r ej

• Substituting z=ej will reduce the z-transform to DTFT

1
 2/17/2020

The z-transform and the DTFT

• The z-transform is a function of the complex z variable
• Convenient to describe on the complex z-plane
• If we plot z=ej for =0 to 2 we get the unit circle

Im  
X e j

Unit Circle

r=1
 0
Re 
2 0 2

Convergence of the z-Transform

• DTFT does not always converge
   xn e

X e j   jn

n  
– Infinite sum not always finite if x[n] no absolute summable
– Example: x[n] = anu[n] for |a|>1 does not have a DTFT

• Complex variable z can be written as r ej so the z-transform

   xn re   
 
X re j   j  n
  xn r n e jn
n   n  

• DTFT of x[n] multiplied with exponential sequence r -n

– For certain choices of r the sum maybe made finite



 xn r
n  
-n


2
 2/17/2020

Region of Convergence
• The set of values of z for which the z-transform converges
• Each value of r represents a circle of radius r
• The region of convergence is made of circles
• Example: z-transform converges for
values of 0.5<r<2
Im – ROC is shown on the left
– In this example the ROC includes
the unit circle, so DTFT exists
• Not all sequence have a z-transform
Re • Example: xn  coson
– Does not converge for any r
– No ROC, No z-transform
– But DTFT exists?!
– Sequence has finite energy
– DTFT converges in the mean-
squared sense
5

Right-Sided Exponential Sequence Example

 az 
 
xn  anun  Xz    anunz n  1 n

n   n0

• For Convergence we require Im

 n
 az
n0
1


• Hence the ROC is defined as

n a 1
az 1 1 z  a o x Re

• Inside the ROC series converges to

 az 

1 z
Xz   1 n
1
 
n0 1  az z a
• Geometric series formula • Region outside the circle of
radius a is the ROC
N2
aN1  aN2 1
a  1  a
n
• Right-sided sequence ROCs
n  N1 extend outside a circle
6

3
 2/17/2020

Same Example Alternative Way

 az 
 
xn  anun  Xz    a unz
n n
 1 n

n   n0
N2 N1 N2 1
 

n  N1
n

1

az   az 
1 0 1 

 az 

1 n

n0 1  az 1
|z|>2
• For the term with infinite exponential to vanish we need
az 1  1  a  z
– Determines the ROC (same as the previous approach)
• In the ROC the sum converges to

 az 

1
Xz   1 n

n0 1  az 1
7

Two-Sided Exponential Sequence Example

n n
 1 1
xn     un -   u- n - 1
 3 2
0  1 1
 1 1   1 1  ROC :  z 1
n  z    z  3

 1 1   3   3  1
  z  
n0  3  1 1

1 1
 z
1 z 1  z 1 3
3 3
 1 1 

 1 1 
0 1 1
 z   z  ROC : z 1
 1 1 
n
1 2
2 2
1
  
  z  
n   2  1 1 1
1 z  1
1  z 1  z Im
2 2 2

 1 
2z z   1 1
1 1  12  
Xz     3x oo
2
x Re
1 1 1 1  1  1 1
1 z 1 z  z   z   12
3 2  3  2
8

4
 2/17/2020

Finite Length Sequence

an 0  n  N  1
xn  
0 otherwise


1  az 1 N
1 zN  aN
 az 
N 1 N 1
Xz    anz n  1 n
 
n0 n0 1  az 1 z N 1 z  a

Properties of The ROC of Z-Transform

• The ROC is a ring or disk centered at the origin
• DTFT exists if and only if the ROC includes the unit circle
• The ROC cannot contain any poles
• The ROC for finite-length sequence is the entire z-plane
– except possibly z=0 and z=
• The ROC for a right-handed sequence extends outward from
the outermost pole possibly including z= 
• The ROC for a left-handed sequence extends inward from the
innermost pole possibly including z=0
• The ROC of a two-sided sequence is a ring bounded by poles
• The ROC must be a connected region
• A z-transform does not uniquely determine a sequence without
specifying the ROC

10

5
 2/17/2020

Stability, Causality, and the ROC

• Consider a system with impulse response h[n]
• The z-transform H(z) and the pole-zero plot shown below
• Without any other information h[n] is not uniquely determined
– |z|>2 or |z|<½ or ½<|z|<2
• If system stable ROC must include unit-circle: ½<|z|<2
• If system is causal must be right sided: |z|>2

11

Z-transform and Linear Systems

 Z-transform of a causal FIR system

M
yn   bm xn  m
m 0
n n<0 0 1 2 3 … M N> M
h[n] 0 b0 b1 b2 b3 … bM 0
M
 The impulse response is hn   bm n  m
m0
 Take the z-transform on both sides

6
 2/17/2020

Z-transform of Causal FIR System (continue)

 M 
 M
Y z   Z yn  Z   bm xn  m    bm xn  mz  n
m  0  n   m  0
M  M   
  bm  xn  mz n
  bm z m 
 
xn  m z  n  m  

m 0 n   m 0  n   
M
  bm z m Z xn  Z hnX z   H z X z 
m 0

 Thus, the z-transform of the output of a FIR system is

the product of the z-transform of the input signal and
the z-transform of the impulse response.

Z-transform of Causal FIR System

(continue)

M
H z    bm z m
m0
 H(z) is called the system function (or transfer
function) of a (FIR) LTI system.

x[n] y[n]
h[n]

X(z) Y(z)
H(z)

7
 2/17/2020

Multiplication Rule of Cascading System

X(z) Y(z) Y(z) V(z)

H1(z) H2(z)

X(z) Y(z) V(z)

 H1(z) H2(z)

X(z) Y(z)
 H1(z)H2(z)

Example

 Consider the FIR system y[n] = 6x[n] -5x[n-1] + x[n-2]

 The z-transform system function is

H z   6  5 z 1  z 2
 1  1
 z   z  
 
 3  z 1 2  z 1  6
3 
z2
2

8
 2/17/2020

Delay of one Sample

 Consider the FIR system y[n] = x[n-1], i.e., the one-

sample-delay system.
 The z-transform system function is

H z   z 1

z-1

Delay of k Samples

 Similarly, the FIR system y[n] = x[n-k], i.e., the k-

sample-delay system, is the z-transform of the
impulse response h[n - k].

H z   z  k

Z-k

9
 2/17/2020

System Diagram of A Causal FIR System

 The signal-flow graph of a causal FIR system

can be re-represented by z-transforms.

x[n] b0 + x[n] b0 +
y[n] y[n]
TD z1
x[n-1] b1 + x[n-1] b1 +
TD z1
x[n-2] b2 + x[n-2] b2 +

TD z1
x[n-M] bM + x[n-M] bM +

Z-transform of General Difference Equation

 Remember that the general form of a linear

constant-coefficient difference equation is
N M
 ak yn  k    bm xn  m for all n
k 0 m 0

 When a0 is normalized to a0 = 1, the system

diagram can be shown as below

10
 2/17/2020

Review of
Linear Constant-coefficient Difference Equation

x[n] b0 + + y[n]

TD TD
b1 + + a1 y[n-1]
x[n-1]
TD TD
x[n-2] b2 + + a2 y[n-2]

TD
TD aN
+ y[n-N]
x[n-M] bM +

Z-transform of Linear Constant-coefficient

Difference Equation

 The signal-flow graph of difference equations

represented by z-transforms.

X(z) b0 + + Y(z)

z1 z 1
b1 + + a1
z1 z 1
b2 + + a2

z1
z 1 aN
+
bM +

11
 2/17/2020

Z-transform of Difference Equation (continue)

 From the signal-flow graph,

M N
Y z    bm X z z m
  ak Y z z  k
m0 k 1
N M

 Thus,  ak Y z z k   bm X z z m
k 0 m 0
M

Y z  m 0
bm z  m
We have 
X z 

N
 ak z  k
k 0

Z-transform of Difference Equation (continue)

 Let M N
H z    bm z m
/  ak z k
m 0 k 0
 H(z) is called the system function of the LTI system defined by
the linear constant-coefficient difference equation.
 The multiplication rule still holds: Y(z) = H(z)X(z), i.e.,
Z{y[n]} = H(z)Z{x[n]}.
 The system function of a difference equation is a rational form
X(z) = P(z)/Q(z).
 Since LTI systems are often realized by difference equations, the
rational form is the most common and useful for z-transforms.

12
 2/17/2020

Z-transform of Difference Equation (continue)

 When ak = 0 for k = 1 … N, the difference equation

degenerates to a FIR system we have investigated
before. M
H z   
bm z  m
m0

 It can still be represented by a rational form of the

variable z as M

bm z M  m
H z   m 0
zM

System Function and Impulse Response

 When the input x[n] = ẟ[n], the z-transform of the

impulse response satisfies the following equation:
Z{h[n]} = H(z)Z{ẟ[n]}.
 Since the z-transform of the unit impulse ẟ [n] is
equal to one, we have
Z{h[n]} = H(z)
 That is, the system function H(z) is the z-
transform of the impulse response h[n].

13
 2/17/2020

System Function and Impulse Response

(continue)

 Generally, for a linear system,

y[n] = T{x[n]}
 it can be shown that
Y{z} = H(z)X(z).
where H(z), the system function, is the z-transform of the
impulse response of this system T{}.
 Also, cascading of systems becomes multiplication of
system function under z-transforms.

X(z) Y(z) (= H(z)X(z)) Z-transform

H(z)/H(ejw)
X(ejw) Y(ejw) (= H(ejw)X(ejw)) Fourier transform

Poles and Zeros

 Pole:
 The pole of a z-transform X(z) are the values of z for
which X(z)= ∞.
 Zero:
 The zero of a z-transform X(z) are the values of z for
which X(z)=0.
 When X(z) = P(z)/Q(z) is a rational form, and
both P(z) and Q(z) are polynomials of z, the
poles of are the roots of Q(z), and the zeros are
the roots of P(z), respectively.

14
 2/17/2020

Examples

 Zeros of a system function

 The system function of the FIR system
y[n] = 6x[n] - 5x[n-1] + x[n-2] has been shown as

 1  1
 z   z  
3  2  P z 
H z   6  
z 2 Q z 

 The zeros of this system are 1/3 and 1/2, and

the pole is 0.
 Since 0 and 0 are double roots of Q(z), the pole
is a second-order pole.

Example: Right-sided Exponential Sequence

 Right-sided sequence:
 A discrete-time signal is right-sided if it is nonzero
only for n≥0.
 Consider the signal x[n] = anu[n].

 
 
X z    a unz   az 1
n n n

 az 1 
n   n 0 
n
 For convergent X(z), we need 
n 0

 Thus, the ROC is the range of values of z for which

|az-1| < 1 or, equivalently, |z| > a.

15
 2/17/2020

Example: Right-sided Exponential Sequence

(continue)
 By sum of power series,

 az 

1 z
X z   1 n
  , z  a
n 0 1  az 1 za
 There is one zero, at z=0, and one pole, at z=a.

: zeros
 : poles
Gray region: ROC

Example: Left-sided Exponential Sequence

 Left-sided sequence:
 A discrete-time signal is left-sided if it is nonzero
only for n <1.
 Consider the signal x[n] = -anu[-n-1].
 1
X z     a u n  1z
n n
  a n z n
n   n  

 a z 
 
 n n 1 n
 a z  1
n 1 n0

 If |az-1| < 1 or, equivalently, |z| < a, the sum converges.

16
 2/17/2020

Example: Left-sided Exponential Sequence

(continue)

 By sum of power series,

1  a 1 z z
X z   1    , z  a
1  a 1 z 1  a 1 z z  a
 There is one zero, at z=0, and one pole, at z=a.

The pole-zero plot and the

algebraic expression of the
system function are the same
as those in the previous
example, but the ROC is
different.

Example: Sum of Two Exponential Sequences

n n
1  1
Given xn     u n      u n 
2  3
 n  n
1  1
Then X z      u n z  n      u n z  n
n    2  n    3 
 n n
 1  n   1  n
   z    z
n 0  2  n0  3 

 1
2 z z  
1 1  12 
  
1 1  1  1
1  z 1 1  z 1  z   z  
2 3  2  3

17
 2/17/2020

Example: Sum of Two Exponential Sequences

(continue)

n z
1 1 1
  u n   , z 
2 1 1 2
1 z
n 2
z
 1 1 1
   u n   , z 
 3 1 1 3
1 z
Thus 3

n n z
1  1 1 1 1
  u n      u n    , z 
2  3 1 1 1 1 2
1 z 1 z
2 3

Example: Sum of Two Exponential Sequences

(continue)

18
 2/17/2020

Example: Two-sided Exponential Sequence

n n
 1 1
Given x n      u n     u  n  1
 3 2
n z
 1 1 1
Since    u n   , z 
 3 1 1 3
1 z
3
and by the left-sided sequence example

n z
1 1 1
   u  n  1  , z 
2 1 1 2
1 z
2

Example: Two-sided Exponential Sequence

(continue)
 1
2 z z  
1 1  12 
X z    
1 1  1  1
1  z 1 1  z 1  z   z  
3 2  3  2

Again, the poles

and zeros are the
same as the
previous example,
but the ROC is not.

19
 2/17/2020

Example: Finite-length Sequence (FIR System)

a n 0  n  N 1
Given xn   
 0 otherwise
Then
X z  
N 1
a n n
z   az 
N 1
1 n

 
1  az 1
N

n0 n0 1  az 1
1 zN  aN

z N 1 za
There are the N roots of zN = aN, zk = aej(2πk/N). The root of k = 0
cancels the pole at z=a. Thus there are N-1 zeros, zk = aej(2π
k/N), k = 1 …N, and a (N-1)th order pole at zero.

Pole-zero Plot

20
 2/17/2020

Some Common Z-transform Pairs

 n  1 ROC : all z.

1
un  ROC : z  1.
1  z 1
1
 u n  1  ROC : z  1.
1  z 1
 n  m  z  m ROC : all z except 0 (if m  0) or  (if m  0) .
1
a n un  ROC : z  a .
1  az 1
1
 a n un  ROC : z  a .
1  az 1
az 1
na n un   ROC : z  a .
1  az 
1 2

Some Common Z-transform Pairs

(continue)
az 1
 na n u n  1  ROC : z  a .
1  az 1 2

1  cos w0 z 1
cos w0 nun  ROC : z  1.
1  2 cos w0 z 1  z  2

sin w0 nun  sin w0 z 1 ROC : z  1.

1  2 cos w0 z 1  z  2

r n cos w0 nun  1  21rcosr cos w0 z 1

w0 z 1  r 2 z  2
ROC : z  r.

r n sin w0 nun  1  2r cos

r sin w0 z 1
w0 z 1  r 2 z  2
ROC : z  r.

a n 0  n  N 1 1 a N z N
  ROC : z  0.
 0 otherwise 1  az 1

21