43

Sections 4.4.5 and 4.2.6

Relationship of Z -Transform and Fourier Transform
( ) ( )
n
n
z n x z X

+
=

= ROC: r
2
< |z| < r
1

let z = re
j
then ( ) ( ) ( ) ( ) ( )


X e n x z X e r n x z X
n j
n
e z
n j n
n
re z
j j
= =

+
=
=

+
=
=

and
Therefore, if z = e
j
is not within ROC, then X() doesnt exist. There are cases that X(z) exists
such as a
n
u(n), |a| > 1, because we can find an r such that ( ) <

+

n
r n x , but its X() doesnt
exist because |x(n)| is not finite when |a| >1. Note that in this example, ROC doesnt include z
= 1 since ROC will be |z| > r > |a| >1.
On the other hand, there are cases that X() exists on a weaker condition that the signals energy
is finite like ( )
n
n
n x
c

sin
= but it doesnt have a X(z). Therefore, if X(z) exists and if its ROC
includes unit circle, then X() exists too, while the other side around is not true.

If the system function, H(z), converges on the unit circle, we can obtain the frequency response
of the system by evaluating H(z) on the unit circle.
( ) ( ) ( )
n j
n
e z
e n h z H H j



+
=
=

= =
If H(z) can be written as ( )
( )
( ) z A
z B
z H = then
( )
( )
( )
( )
( )

=
+
= =
N
k
j
k
M
k
j
k
o N
k
k j
k
M
o k
k j
k
e P
e z
b
e a
e b
A
B
H
1
1
1
1
1
1

a
k
and b
k
are real, but z
k
and p
k
can be
complex.
Power Spectrum is defined as: |H()|
2
= H() H
*
()
44
( )
( )
( )
( )
( )

=
=
=
=

=
|
.
|

\
|

=
N
k
k
M
k
k
o N
k
j
k
M
k
j
k
o
z p
z z
b
z
H
e p
e z
b H
1
*
1
*
*
*
1
*
1
*
*
1
1
1
1
1

when h(n) is real, then complex z

k
and p
k

occur in complex-conjugate pairs. Then ( )
1
*
*
1

=
|
.
|

\
|
z H
z
H or equivalently H*() = H(-).
LTI Systems as Frequency Selective Filters
H() acts as a weighting function or a spectral shaping function as Y() = H()X(). From this
point of view, H() is a filter. An ideal filter has a constant gain in pass-band and is zero in stop-
band and also has a linear phase response. ( )
0
n = -linear characteristics within pass-band.
( )

d
d
g
= is called the group delay of the filter. ( )
g
is the time delay that a signal
component of frequency , undergoes as it passes from the input to the output of the system.
Obviously when ( )
g
= constant, then all the components have the same delay.
An example of ideal low-pass filter is:





Therefore, ( )
n
n
n h
c

sin
= but is not a casual signal and is not absolutely summable and is
also unstable. However, its frequency response can be very closely approximated by some
realizable filter. You have seen how the location of poles and zeros changes the frequency
response. Now lets have a graphical view of their location and type of filter.
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )

=
=
=
=
=

=
N
k
j
k
M
k
j
K
M N j
o N
k
k
j
M
k
k
j
M N j
o N
k
j
k
M
k
j
k
o
k
k
e U
e V
e b
p e
z e
e b
e p
e z
b H
1
1 _
1
1
1
1
1
1

|H()|

c

45













1) If a zero is on the unit circle at
k o
z p = , then =
=
0
o
k
V

|H(
o
)| = 0.
2) If a pole is on the unit circle at
k o
p p = , then ( ) = =
= o k
H U
o


0 .
From 1 and 2 it is clear that the presence of a zero close to the unit circle, makes |H()| to be
small at the frequencies close to that point and on the other hand, the presence of a pole close to
the unit circle, causes |H()| to be large at the frequencies close to that point.
Lets look at ( ) 0
1
1
1 1
1
= =

=

z a p
az
a
z H . So if a is real ( ) 0
1
= p p and close to unit
circle, then |H()| will be maximum at zero frequencies. Now if you add a zero on the unit
circle, but with -, it also attenuates the |H()| more at high frequencies. Therefore, we can
say what kind of filter it is by just looking at zero-poles locations.







x
x
x
x
x
x
x
x
x
x
x
x
LP
HP
R
e
1

p
k

e
j

( )
( )
( ) ( )
k
j
k
k
j
k
j
K k
j
z e
z e V
e V z e
k
=
=
=

p
I
m

( )
( )
( )
( ) ( )
k
j
k
k
j
k
k
j j
k
p e
p e U
p e e U
k
=
=
=

p

( )
( )
( )

=
=
=
N
k
k
M
k
k
o
U
V
b H
1
1

z
k
U
k

V
k

46
Example of a Band-Pass Filter
Design a 2-pole band-pass filter with center frequency
2

= and 2-zeros at = o, , also

( )
2
1
9
4 =
=

H . Lets have the zeros on the unit circle: 1 , 1

2 1
= = = =
+ j jo
e z e z . If we want
the filter coefficients to be real, the complex poles must be complex conjugate. So let
jr
re p
jr
re p
j j
=
=
=
=

2
2
2
1


( )
( )( )
( )( ) jr z jr z
z z
G z H
+
+
=
1 1
will choose G such that
( )
2
1
1
1
2 1
1
2
2
2 2 2
2
2

= =
+

=
+

= =
|
.
|

\
|
=
=
r
G
r
G
r z
z
G z H H
j z
e z
j

To determine the value r, we should use the corner frequency or the 3db frequency.
( ) ( )
2
1
2
9
4
2
9
4 =
)
`

=
)
`

= =

j
e z
H H

{
2
1
9
8
r 2 r 1
9
8
2 2
2
1 r
9
8
j r
9
8
9
8
j 1
9
8
2
r
2 4
2
2
2
G
1 2
=
+ +

=
+
|
.
|

\
|
+
+
|
.
|

\
|

cos
cos
sin cos
sin cos
r
2
= 0.7
( )
2
2
7 . 0 1
1
15 . 0

=
z
z
z H

) (w H p
2
1

( ) H
dB
0
-3

9
4
47
Converting LP to HP Filters
( ) ( ) =
LP Hp
H H ( ) ( ) ( ) ( ) n h n h e n h
LP
n
LP
n j
HP
1 = =


All-Pass Filters
If |H()| = 1 for o < < H() is an all-pass system like a delay system. H(z) = z
-k
. In an
all-pass system, if z
0
is a pole, then
0
1
z
is a zero. Its main application is in phase equalizer to
compensate for poor phase characteristics to produce an overall linear phase response.
Read all of Section 4.5.

Invertibility of LIT Systems

if ( )
( ) z H
z H
1
2
1
= .
If ( )
( )
( ) z A
z B
z H =
1
, then ( )
( )
( ) z B
z A
z H =
2
meaning that the zeros of H
1
(z) are the poles of H
2
(z). Is
every system invertible?
If an invertible system cannot be expressed by z-transform, we may find the inverse at the
system by convolution of the two h
1
(n) and h
2
(n).
( ) ( ) ( ) n k n h k h
n
k
=

=
2
0
1
h
2
(n) = 0 for n < 0. For n = 0 h
2
(0) =
( ) 0
1
1
h
.
For n 1, ( )
( )
( ) ( ) k n h k h
h
n h
n
k

=

=
2
1
1
1
2
0
1
. You can prove these easily by expanding the
convolution sum.
This method does not work if h
1
(0) = 0 but this can be easily resolved by introducing a delay.

Minimum Phase Maximum Phase Systems
Consider these two systems:
( )
( )

|
.
|

\
|
+ = + =
+ =
|
.
|

\
|
+ =


1
2
1
2
1
2
1
1
2
1
1 1
2
1 1
1
z z z z H
z z z z H

Both are all-zeros systems.
H
1
(z) H
2
(z)
x(n) x(n) y(n)
48








( )

sin
2
cos
2
1
1
2
1
1
1
j
e H
j

|
.
|

\
|
+ = + =


( ) ( ) cos
4
5
sin
4
1
cos
2
1
1
2
2
2
1
+ = + + = H
( )

sin cos
2
1
2
1
2
j e H
j

|
.
|

\
|
+ = + =


( ) cos
4
5 2
2
+ = H

( ) ( )
2 1
H H = because
1
2
1
z
z = reciprocal but ( )

+
=

cos
2
1
sin
tan
1
1

( )

+
=

cos 2
sin
tan
1
2










X X
H
1

H
2

-2
You can plot and check these with
[H, W] = freqz(b,a,1024,whole)
Then plot (w, angle(H))


-


( )
1

( )
2

2

-0.5
49
As goes from 0 , the phase change of ( )
1
is zero but for ( )
2
is . Therefore H1 has
the minimum phase while H2 has the aximum phase charge. So a Min-Phase system has zeros
all inside the unit circle. On the other hand, all zeros of a Max-Phase system are outside the unit
circle. Min-phase Minimum delay.
Showing Minimum Phase Graphically:
( ) ( )

j
j
e
a e
H
z
a z
az z H

=

= =
1
1 z
1
= a, P
1
= 0
Case 1: |a| < 1 p H() = () -

In this case, if changes from zero to ,
( ) H p changes from zero to zero.
At ( ) ( ) 0 0 0 0 = = = H p
At ( ) ( ) 0 = = = H p
Net change = o
Minimum Phase System
Case II: |a|>1

At ( ) = = = 0 0 H p
At ( ) 0 = = = H p
Therefore, in this case, net charge = and
the system is therefore, a Max-Phase system.


So in general, if a system has M zeros outside the unit circle, its net phase change will be
M. over the range (0, ), while the net phase change for the zeros or poles inside the unit circle
is zero.

Phase Compensation
Any casual and stable system can be decomposed to a Min-Phase system cascaded with an all-
pass system.

e
j

a
|H()|
()
e
j

|H()|
()
50
( ) ( ) ( ) z H z H z H
ap
=
min

Lets say ( )

=
=

+
+
=
+
+
=

2
1
3

2
1
3
2
1
1
3 1
1
1
1
1
p
z
z
z
z
z
z H

We need to choose the appropriate H
ap
to reflect the zero that is outside the unit circle to inside.
So H
ap
should have a zero at z
1
= -3 and since it is an all-pass system, it must have a pole at
3
1 1
1

=
z
. Therefore,
1
1
3
1
1
3 1
3
1
3

+
+
=
+
+
=
z
z
z
z
H
ap

( )
( )
1
1
min
2
1
1
3
1
1
2
1
3
1

+
+
=
+
+
= =
z
z
z
z
z H
z H
H
ap

Another Example
( )

= =

|
.
|

\
|

|
|
.
|

\
|
+
|
|
.
|

\
|
+
=

o p p
e z
z
z e z e
z H
j
j j
2 1
4
2 , 1
1
1
4
1
4
,
3
1
2
3
3
1
1
2
3
1
2
3
1



( )
|
.
|

\
|
+
|
.
|

\
|
+
|
.
|

\
|
+
|
.
|

\
|
+
=

4 4
4 4
3
2
3
2
2
3
2
3


j j
j j
ap
e z e z
e z e z
z H
( )
1
1
4
1
4 4 4
min
3
1
1
3
2
1
3
2
1
3
1
3
2
3
2

|
.
|

\
|
+
|
.
|

\
|
+
=
|
.
|

\
|

|
.
|

\
|
+
|
.
|

\
|
+
=
z
z e z e
z z
e z e z
z H
j j j j


Verify these with MATLAB

Min-Phase System Properties
1. Min-Phase System has the smallest group delay.
Proof: H(z) = H
min
(z) . H
ap
(z)

g
() =
gmin
() +
gap
()
Since ( ) o
gap
for ( ) ( )
min g g
o
2. The Min-Phase System has the largest partial energy. Partial Energy of a system is defined
as ( ) ( )
2

=
=
n
o k
k h n E
Note that H
ap
(z) in
these cases is also a
Max-Phase System
51
It can be shown that among all systems having the same |H()| and same total energy
( ) ( )
2

=
=
o k
k h E , the Min-Phase System has the largest partial energy. In particular, it can be
also concluded that ( ) ( ) 0 0 h h
MP
.
Use initial value theorem and prove this as assignment. Look also to following problems in
Chapter 4: 6(a,c), 8, 28, 30, 51, 59 and 100.