ELEC442: Magnitude-Phase Representation

ELEC442: System analysis:
Magnitude-Phase Representation of systems

o Introduction
o Magnitude and Phase Representation of signals
o Amplitude-Phase Representation of signals
o System analysis:
o System function
o Group Delay of LTI Systems
o IIR and FIR Systems
oFrequency Response of Rational Systems
o Relation between Magnitude/Phase & poles/zeros
o All-pass systems
Dr. Aishy Amer o Minimum-phase systems
Concordia University
Electrical and Computer Engineering o Linear systems with generalized linear phase
o Appendix: Log-Magnitude and Bode plots
Content and figures are based on/from
•A. Oppenheim, A.S. Willsky and S.H. Nawab, Signals and Systems, 2nd Edition, Prentice-Hall, 1997
•Oppenheim, Schafer, DSP, 3ed.
•http://metalab.uniten.edu.my/~zainul/images/Signals&Systems
•Slides by Robert Akl, http://www.cse.unt.edu/~rakl/
•Dr. Güner Arslan, 351M Digital Signal Processing, http://signal.ece.utexas.edu/~arslan/courses/dsp
1 •Dr. Zheng-Hua Tan, Digital Signal Processing III, 2009, http://kom.aau.dk/~zt/cources/DSP/

Introduction: Complex numbers
* Cartesian representation : z  x  jy
M agnitude of z r | z | x 2  y 2
It is the distance of a point z from the origin
y
Phase (argument) of z   z  tan 1
x
 is the angle to the real positive axis
 can change by any multiple of 2 and still give the same angle
(radians not degrees are being used)
* Polar representation : z  z e j  z cos   j z sin 
 Complex Conjugate :
z *  x  jy ; ( z  z * ) and ( zz* ) are real
2
Introduction: Complex exponential
x[n]   n e j ( no  )
j j o
x[n]  C , where C | C | e ,  |  | e
n
j jno j ( no  )
x[n] | C ||  | e e n
| C ||  | e n
| C ||  |n cos(o n   )  j | C ||  |n sin(o n   )
|  | 1, real & imaginary parts are sinusoidal
|  | 1, sinusoidal decaying exponentially
|  | 1, sinusoidal growing exponentially
e j (o  2 ) n  e j 2n e jo n  is the phase shift (initial phase)

 e j o n e
j (  2  k ) j
 e , any integer k
3
 Periodicit y at 2  In general, phase is ambiguous
Introduction A cos(o n   )
 Both waves are periodic, i.e., after a certain time, called the period, they look
the same again
 Both waves look alike, but the cosine wave appears to start at its maximum,
while the sine wave starts at zero.
 An important property of sinusoids is “frequency”:

 tells us how many peaks and valleys we can count in a given period of time
 High frequency means many peaks and valleys
 Low frequency means few peaks and valleys
 "frequency" can mean Hz (cycles per second), or radians per second

4
Introduction: Phase A cos(o n   )
 The cosine wave starts out 1/4th later than the sine
wave in its period It has an offset
 Common to measure this offset in degree or radians
 One complete period equals 360° or 2π radian
 The cosine wave thus has an offset of 90° or π/2
 This offset is called the phase of a sinusoid
We cannot restrict a signal x[n] to start out at zero

phase or 90° phase all the time
Must determine signal frequency, amplitude, and
phase to uniquely describe that signal at any time
instant
5
Introduction:
Phase shift A cos(o n   )
 The phase θ of the signal is the offset in the
displacement from a specified reference point at
n=0
 θ represents a "shift" from zero phase
 A change in θ is also referred to as a phase-
shift
 For infinitely long sinusoids, a change in θ is the
same as a shift in time.
Ex: If x(t) is delayed (time-shifted) by 1/4 of its
cycle Signals with different
phase shift
x(t  1 / 4T )  A sin( (t  1 / 4T )   ) compared to the
 A sin(t   / 2   ) bottom signal
whose "phase" is now θ-π/2
it has been shifted by π/2
Source: Wikipedia
6
Introduction: Importance of Phase
7
Phase causes signal shift and/or

signal shape distortion
8 Source: A. Oppenheim, A.S. Willsky and S.H. Nawab, Signals and Systems, 2nd Edition, Prentice-Hall, 1997
 In images, as in sound signals, phase carries

considerable information
 The phase data dominates our perception in images
 A random distortion of the phases can dramatically
distort the reconstructed signal
 Signal delay corresponds to phase change:
Signal delay, in the frequency domain, alters only the

phase of the signal
No difference will be observed when viewing the FT
9
magnitude
 The value of each point determines the phase of the corresponding frequency
 The inverse FT to the above magnitude image while ignoring the phase, gives
o Magnitude tells "how much" of a certain frequency

component is present in x[n]
o Phase tells "where" the frequency component is in
the signal
10
Outline
o Introduction
o System analysis:
oSystem function
oGroup Delay of LTI Systems
oIIR and FIR Systems
o All-pass systems
o Minimum-phase systems
o Linear systems with generalized linear phase
oAppendix: Log-Magnitude and Bode plots
11
Magnitude and Phase Representation of signals
 LTI system is completely characterized by …
 Three domains
 Time domain: impulse response, convolution sum

y[n]  x[n] * h[n]   x[k ]h[n  k ]
k  
 Frequency domain: frequency response

Y (e j )  X (e j ) H (e j )
 z-transform: system function
Y ( z)  X ( z) H ( z)
12
Magnitude-Phase Representation of signals
 DT-FT
 Often x[n] is real but its FT is, in general, complex

 The FT is generally represented by its magnitude and phase
(rather than its real and imaginary parts)
 Magnitude tells "how much" of a certain frequency component is

present in x[n]
 Phase tells "where" the frequency component is in the signal
 Two signals may have the same magnitudes function but the phase
differs
13
Magnitude and Phase Representation of systems
j j j
 Relationship between FT of input and output Y (e )  X (e ) H (e )
 In polar form
| Y (e j ) || X (e j ) |  | H (e j ) |
 Magnitude
 magnitude response, gain, attenuation, distortion
 Phase Y (e j )  X (e j )  H (e j )
phase response, phase shift, attenuation , distortion
Magnitude does not uniquely characterize the system
14
Magnitude and phase of systems from H(z)
15
 We will use this result to find the frequency response of any discrete-
time LTI system with rational transfer function given by
 At any frequency ω , find the magnitude and phase of the vectors

drawn from the poles and zeros to the point e jω
(a point on the unit circle with angle ω )
 The magnitude of H (ejω) at ω is equal to the product of the
magnitudes of all vectors associated with the zeros divided by the
product of the magnitudes of all vectors associated with the poles
 Similarly, the phase of H (ejω) ω is equal to the summation of the
angles of all vectors associated with the zeros minus the summation
of the angles of all vectors associated with the poles
16
One zero, two poles

 Consider an LTI system with
an impulse response h[n]
 Assume that H(z) is a rational
function of z whose pole-zero
configuration is given in the
plot
 From this plot, it can be
concluded that the z-transform
of the impulse response is:
17
 The amplitude and angle of the vectors v1, v2

and v3 depend on the frequency ω
 The magnitude of the frequency response of
this system is proportional to
 The phase of the frequency response of this

system is equal to
 The magnitude of the frequency response is

large at those frequencies that correspond to
the points on the unit circle which are close to
the poles and far from the zeros
 Similarly, the magnitude of the frequency
response is small at those frequencies that
correspond to the points on the unit circle
which are close to the zeros and far from the
poles
18
Magnitude-Phase Representation of systems
from H(z)
H (e j )  H ( z ) | z  e j
j j j j j 1 (  )
H (e )  Hr (e )  Hi(e ) | H (e ) | e
 1( )  H (e j )
Magnitude  | H (e j ) |
Phase ( )  H (e j )
•In magnitude and phase plots when H(z) has a zero on the unit
j
circle, z  e :
• as ω goes through that zero on the unit circle
• the magnitude will go to zero and
• the phase will flip by π (phase reversal)
19
Example 1:
h[n]   [n]   n  
| H (e j ) | 1, H (e j )  
j j
H (e )  1  e
In the magnitude-phase representation, a real-valued
frequency response does not necessarily mean that the
system is zero-phase
20
Example 2:
ideal lowpass filter
 Frequency response H (e j )   1, |  | c ,
0, c |  | 
 Impulse response
sin c n ?
hlp [n]  ,   n   h[n]  0, n  0
n
Non-causal, cannot be implemented!
21
Example 3:
ideal delay system
 Ideal delay system hid [n]   [n  nd ]
H id (e j )  e  jnd
| H id (e j ) | 1 Delay distortion
H id (e j )  nd , |  |  Linear phase distortion
 Ideal lowpass filter with linear phase

 jnd
j
 e , |  | c ,
H lp (e )   Ideal lowpass filter is
0, c |  |  always non-causal !
sin c (n  nd )
hlp [n]  ,   n  
22  ( n  nd )
Example 4: First-order Recursive DT Filter
+
x[n] y[n]
ay[n-1] D
a y[n-1]
y[n]  ay[n  1]  x[n]
1) Taking the DTFT : Y(e j ) - ae -jY (e j )  X (e j )
Y(e j ) j 1
 j
 H(e ) 
X (e ) 1  ae - j
2) Eigenfunct ion property: x[n]  e j n , then y[n]  H(e j )e j n
Substituti ng in the diff. eq. : H(e j )e j n  aH(e j )e j ( n -1)  e j n
j 1
H(e )  - j
23
1  ae
Example 4: first order recursive DT filter
F .T
j
h[n]  H(e )
F .T 1
a u[n] 
n
- j
1  ae
h[n]  a u[n], 0  a  1
n
0 n
24
Example 4: First order recursive filter
j 1
H (e )   j
1  ae
1
(1  a)
1
(1  a)
- 0 
tan 1 (a / 1  a 2 )
 2
jumps by π
-  2
 tan 1 (a / 1  a 2 )
•Typically, it is hard to infer much from a phase plot

( a group delay plot is more useful; see later)
25
Outline
o Introduction

o System analysis:
oSystem function
o All-pass systems
26
Amplitude-Phase Representation of signals
 Recall: In magnitude and phase plots:

 as ω goes through a zero on the unit circle of H(z)
• the magnitude will go to zero and
• the phase will flip by π (phase reversal)
j
 If H ( e ) is real but bipolar, an alternative representation
is to remove these jumps of π in a phase plot
 A bipolar signal may assume two polarities and is
usually symmetrical with respect to zero amplitude, i.e.,
the absolute values of the positive and negative signal
states are nominally equal
 An audio waveform is an example of a bipolar signal
27
 A signal can be expressed in terms of complex polar

coordinates or as Amplitude/Phase
Representation
H (e j )  A(e j )e j 2 ( ) H (e j ) | H (e j ) | e j 1( )
j  1( )  H (e j )
 2( )  H (e )
 A(e j ) is real but not restricted to be positive,

thus θ2(ω) is different from θ1(ω)
j
 The sign change is contained in A(e ) , and
the jumps of π do not exist in θ2(ω)
28
 Consider h(n)
j  j ( N 1) / 2
H (e )  Ae sin(N / 2) / sin( / 2)
 In the magnitude-phase representation: θ1(ω) has jumps of π at

the sign change
 In the amplitude-phase representation: θ2(ω)= -ω(N-1)/2, a straight
line with the slope –(N-1)/2
 Moreover, here phase would be the same whether A is positive or
negative
29
Outline
o Introduction
oSystem analysis:
o System function
o Frequency Response of Rational Systems
o All-pass systems
30
System function
 Linear constant-coefficient difference equation

N M
a
k 0
k y[n  k ]   bm x[n  m]
m 0
 z-transform format N M
 k Y ( z) 
a z k
 m X ( z)
b z m
M k 0 m0
Y ( z)
 bm z m
m 0
H ( z)   N
X ( z)
 k
a z k
(1  cm z 1 ) in the numerator
k 0
M a zero at z  cm ; a pole at z  0
b0 m 1
 m )
(1  c z 1
(1  d k z 1 ) in the denominato r
( ) N
a0
 (1  d k z 1 ) a zero at z  0 ; a pole at z  d k
31
k 1
System function: Frequency
Response
 DT-FT of a LTI rational system function
1  c e 
M M
b e k
 jk
 b0 k
 j
 
He j k 0
   k 1
 1  d e 
N N
 a0 
a e
k 0
k
 jk
k 1
k
 j
 Magnitude Response M
1c e k
 j
 
H e j 
b0
a0
k 1
N
 1
k 1
 dk e  j
32
System function: Log-Magnitude Response
 Log Magnitude in decibels (dB)
 
M N
b0
20 log10 H e j
 20 log10   20 log10 1  ck e  j
  20 log10 1  d k e  j
a0 k 1 k 1
 Frequency response is a sum of the contributions from each pole and
zero
 System gain in dB Gain in dB  20 log10 H e j  
 System attenuation in dB  
 - 20 log10 H e j  Gain in dB
 Example:
 |H(ej)|=0.001 translates into –60dB gain or 60dB attenuation
 |H(ej)|=0.5 translates into -6dB gain
 |H(ej)|=1 translates into 0dB gain
 |H(ej)| > 1 translates into positive gain
33
System function
 Stable
 h[n] absolutely summable
 H(z) has a ROC including the unit circle
 Causal
 h[n] right side sequence
 H(z) has a ROC being outside the outermost
pole
 Stable & causal:
 ROC must be outside the outermost pole and
include the unit circle
34
System function: Inverse systems
 Many (not all) systems have inverses

 specially systems with rational system functions
M
G( z)  H ( z) H i ( z)  1  (1  cm z 1 )
b0 m 1
H ( z)  ( ) N
1 a0
H i ( z)   k )
(1  d z 1
H ( z) k 1
N
g[n]  h[n] * hi [n]   [n]  (1  d k z 1 )
a0 k 1
H i ( z)  ( ) M
b0
 (1  c m z 1
)
m 1
 Poles become zeros and vice versa

 ROC: must have overlap between H(z) and
35
Hi(z) for the sake of G(z)
System function: Inverse systems
Example
1  0.5 z 1 1  0.9 z 1
H ( z)  1
, | z | 0.9 H i ( z ) 
1  0.9 z 1  0.5 z 1
1 0.9 z 1
 1

1  0.5 z 1  0.5 z 1
The only choice for ROC of Hi(z) that
If | z | 0.9, right - sided h[n] overlap with |z|>0.9 is |z|>0.5
hi [n]  (0.5) n u[n]  0.9(0.5) n 1 u[n  1]; stable & causal

 h[n] * hi[n]   [n]
If | z | 0.9, left - sided h[n]

hi [n]   [n]  0.8(0.5) n u[n]; unstable & noncausal
36
BUT h[n] * hi[n] !  [n]  not inverse
37  Minimum phase systems
Outline
o Introduction
o System analysis:
oSystem function
o All-pass systems
38
Group Delay of LTI Systems
 Typically, it is hard to infer much from a phase plot

 A group delay plot gives more useful information
d
Group delay is defined as  ( )   { H (e j ) }

d
 A group delay is about the delay of the system

 It measures the linearity of the phase
 It is about delay of samples on the signal
 Group delay is a measure of the transit time of a signal
through a system/device versus frequency
 Group delay is a useful measure of phase distortion
 The variations in group delay cause signal distortion
 In Matlab, group delay is calculated using the FT rather than

39
differentiation
Group Delay of LTI Systems
j j j
Y (e )  X (e ) H (e )
j j j
| Y (e ) || X (e ) || H (e ) | Y(e j )  H(e j )  X(e j )
It is a measure of the linearity of the phase

40
Group delay  ( )
s[n]
w
0
 Effect of phase distortion on a narrowband signal
x[n]  s[n] cos(0 n)
0 w0
 For this input x[n] with spectrum only around w0
 phase effect can be approximated around w0 as linear
(though in reality maybe nonlinear)
H (e j )  nd  0
 The output y[n] = x[n]*h[n] is approximately
y[n] | H (e j0 ) | s[n  nd ] cos(0 (n  nd )  0 )
 Group delay d
  grd[ H (e )]   {arg[ H (e j )]}
j
d
41
Example: consider the following freq. response and group delay of
a system (Fig. 5.4)
 At ±0.2π, we get 150 sample delay
 At ±0.4π, we get 10 sample delay
Larger gain at ± 0.4π than gain at ±0.2π

(both have |H|>1 ,so amplification of input x[n] occurs)
 At ± 0.8π |H|<1 ,so attenuation of input x[n] occurs
42
Example: Fig 5.5 shows an input signal + its FT to the system in Fig. 5.4
0.8π 0.2π 0.4π x[n] consist of 3 narrowband
pulses at w= 0.2π ,0.4π ,0.8π
43
j
Example to Fig 5.4 | Y (e j ) || X (e j ) || H (e ) |
44
Outline
o Introduction
o System analysis:
o System function
o IIR and FIR Systems: impulse response of rational systems

o All-pass systems
45
Infinite Impulse Response (IIR) Systems
 Rational system function

1c z 
M
b   k
1
Hz    0  k 1
 1  d z 
N
 a0  1
k
k 1
 If at least one nonzero pole does not cancel with a zero, in h[n] there will at
least one term of the form
a nun OR - a nu n  1
 Therefore, h[n], the impulse response, will be infinite length IIR system
 Example: Causal system of the form yn  ayn  1  xn

1
Hz  ROC : z  a from causality
1  az1
hn  anun
46
Finite Impulse Response (FIR) Systems
M
Y z  k
b z k
 The rational system function H z    k 0

X z  N
 k
a
k 0
z k
 If H(z) does not have any poles except at z=0

 In this case N=0
M
H  z    bk z  k
 The rational system function
k 0
 No partial fraction expansion possible (or needed)

M
 The impulse response can be seen to be hn   b n  k 

k
k 0
 Impulse response is of finite length  FIR system

47
Example5.5: FIR System
an 0  n  M
hn  
 finite length
 The system function 0 else
 M
1 - a M 1 z  M 1
H z    hnz
n  
n
 a z
n 0
n n

1  az 1
 Assuming a is both real and positive, the zeros can be written as
zk  ae j 2k / M 1 for k  0,1,..., M
 For k=0 we have a zero at z0=a

 The zero cancels the pole at z=a
M
 From convolution  yn    ak xn  k 
k 0
 From 1 - a M 1 z  M 1
 H z  
1  az 1
yn  ayn  1  xn  aM1xn  M  1 Need memory
 The more the zero cancels the pole at a ,the more equal the two equations become.
48
Outline
o Introduction
o System analysis:
o System function

o Frequency response
o Phase response
o Group delay
o Squared magnitude
o All-pass systems
49 oAppendix: Log-Magnitude and Bode plots
Frequency Response of Rational System
Functions
 DTFT of a stable and LTI rational system function
1  c e 
M M
b e k
 jk
 b0 k
 j
He  j k 0
   k 1
 1  d e 
N N
 a0 
a e
k 0
k
 jk
k 1
k
 j
M
 Magnitude Response 1c e k
 j
 
H e j 
b0
a0
k 1
N
 1  dk
k 1
e  j
 Magnitude Squared
 1  c e 1  c e 
M
 j * j
2
 b0  k
    He 
2 k
j 
He j
 He j
   k 1
 1  d e 1  d e 
N
 a0   j * j
k k
50 k 1
Log-Magnitude Response of
Rational System Functions
 Log Magnitude in decibels (dB)
 
M N
b0
20 log10 H e j
 20 log10   20 log10 1  ck e  j
  20 log10 1  d k e  j
a0 k 1 k 1
Frequency response is a sum of the contributions from each pole and zero
Gain in dB  20 log10 H e j 

 
Attenuation in dB  - 20 log10 H e j  Gain in dB
 Output of system
   
20 log10 Y e j  20 log10 H e j  20 log10 X e j  
51
Phase Response of Rational System Functions
 Phase response of a rational system function

 b0  M
     
N
H e j
      1  ck e  j
   1  dk e j
 a0  k 1 k 1
Phase response is a sum of the contributions from each pole and zero
 Corresponding group delay
     
M N
d d
grd H e j
  arg 1  ck e  j
 arg 1  d k e  j
k 1 d k 1 d
 Here arg[.] represents the continuous (unwrapped) phase
 Equivalently,we get
 
grdH e j
M dk
2
 Re dk e j  M

ck
2

 Re ck e j 
1 d k 1 k
2
 2 Re dk e   j
 k 1 1  ck
2

 2 Re ck e j 
52
Instantaneous, wrapped,
Unwrapped Phase
 Phase shift is in general ambiguous:
j (  2  k ) j
e  e , any integer k
 The instantaneous (or local) phase of a complex-
valued function is the real-valued function
 ( )  H (e j )
 When the phase is constrained to an interval, for
example, [- π,π], it is called the wrapped phase
it does not distinguish between multiples of 2π
 An unwrapped phase is a continuous function of ω
The unwrapped phase is not restricted to the wrapped
range, and has no jumps
53
Unwrapped (Continuous) Phase
 Phase is ambiguous: when calculating the

arctan(.) function on a computer
 Values between - and +
 Denoted in the book as ARG(.)
  
   ARG H e j  
 Any multiple of 2 would give the same

result
    
H e j  ARG H e j  2r
 Here r() is an integer for any given
value of 
 Group delay is the derivative of the
unwrapped phase
  
grd H e j

d
d
   
argH e j
54
Examples: Frequency and phase response, and group delay
 
M N
b0
20 log10 H e j
 20 log10   20 log10 1  ck e  j
  20 log10 1  d k e  j
a0 k 1 k 1
Frequency/phase/delay responses:
55 a sum of the contributions from each pole and zero
Frequency Response of a Single Zero or Pole
zero at re j
 Let’s analyze the effect of a single term
1  ck e  j 2
 1  re ej  j 2
 1  r 2  2r cos   
 If we represent it in dB
20 log10 1  re je j  10 log10 1  r2  2r cos   

The phase term is written as
 r sin   


ARG 1  re je j  arctan  
1  r cos  
 And the group delay obtained by differentiating the phase

r 2  r cos   r 2  r cos  

grd1  re e j  j
 
1  r  2r cos  
2

j  j
2
1  re e
 Minimum and maximum values of magnitude
  
10 log10 1  r2  2r cos    10 log10 1  r2  2r  20 log10 1  r  
56
10 log10 1  r 2
 2r cos    10 log 1  r 10
2
 2r   20 log 10 1r
The zero at θ
dominates H
Phase(w=θ) =0
57
Pole at θ dominates H
Pole at –θ dominates H
58
Outline
o Introduction
o System analysis
o System function
o All-pass systems
59
Relation between Magnitude and Phase
 For general LTI system
 Knowledge about magnitude/phase doesn’t provide any information
about phase/magnitude
 For linear constant-coefficient difference equations, however, there is

some constraint between magnitude and phase
 If magnitude and the number of pole-zeros are known
• Only a finite number of choices for phase
 If phase and number of pole-zeros are known
• Only a finite number of choices for magnitude (ignoring scale)
 A class of systems called minimum-phase

 Magnitude specifies phase uniquely
 Phase specifies magnitude uniquely
60
Relation between Magnitude and Phase
 Stable and causal system:

 poles inside unit circle, no restriction on zeros
 If zeros are also inside the unit circle
inverse system is also stable and causal
(in many situations, we need inverse systems!)
 (Minimum-phase systems: stable and causal and have

stable and causal inverses)
61
Magnitude-Squared System Function
 Explore possible choices of system function of the form
He   j
2
   
 H e j H e j  H* 1 / z* Hz

  z  e j
 Restricting the system to be rational


1c z 
M
1  c z
M
b   k
1
 b0 
*
k
Hz    0  k 1 *
H 1/z *
    k 1
 1  d z 
N
 a0 
 1  d z
N
k
1
 a0  *
k
k 1 k 1
 1  c z 1  c z 
 The square system function M
1 *
2
 b0 
 
k
d k  1/ d k C z   H z H 1 / z   
k
* * * k 1
 1  d z 1  d z 
N
 a0  1 *
k
Given He 
k
2

j
we can get C(z) k 1
62
 What information on H(z) can we get from C(z)? Original pole Newly added
Poles and Zeros of Magnitude Square System
Function d *k  1 / d k ;
 1  c z 1  c z 
M
1
z  1 / d *k or z  d k
*
2
b 
 
k
C z   H z H * 1 / z *   0 
k
k 1
 1  d z 1  d z 
N
 a0  1 *
k k
k 1
 For every pole dk in H(z) there is a pole of C(z) at dk and (1/dk)*

 For every zero ck in H(z) there is a zero of C(z) at ck and (1/ck)*
 Poles and zeros of C(z) occur in conjugate reciprocal pairs
 If one of the pole/zero is inside the unit circle the reciprocal will be outside
 Unless there are both on the unit circle
 If H(z) is stable all poles have to be inside the unit circle
 We can infer which poles of C(z) belong to H(z)
 However, zeros cannot be uniquely determined
 Example to follow
63
Example5.9: Systems with same C(z)
H1  z  
  
2 1  z 1 1  0.5 z 1
 
1  0.8e j / 4 z 1 1  0.8e  j / 4 z 1 
H z  
1  z 1  2 z 
1 1 Both share the same
magnitude square
2
1  0.8e z 1  0.8e
j / 4 1  j / 4 1
z  system function
 Differ in zeros
H1  z  H 2 z 
If we are given this system,

we can change the zero position and C z   H1 z H1* 1 / z *  
64
still have the same magnitude

 H 2 z H 2* 1 / z * 
Example5.10: C(z)  H(z) C(z)
M M
Y ( z) b m z m
b0  (1  c m z 1 )
H ( z)   m 0
N
( ) m 1
N
a z  (1  d
X ( z) k a0
k k z 1 )
k 0 k 1
 Given is C(z)
1. Find P & Z of H(z)
- For every pole dk in H(z) there is a pole of C(z) at dk and (1/dk)*
- For every zero ck in H(z) there is a zero of C(z) at ck and (1/ck)*
2. How many stable & causal systems exist?
1. If we assume H(z) is stable & causal

 poles must be inside the unit circle (zeros can be anywhere) = 3 poles
2. If we assume ak and bk are real, the zeros and ploes either are real or
occur in complex conjugate pairs
 There are 4 different stable and causal systems with 3 poles and 3 zeros,
and for which the frequency-response magnitude is the same
3. If we assume ak and bk are NOT real, the number of choices of stable and
causal systems would be greater
65
66
Outline
o Introduction
o System analysis
o System function
o All-pass systems
67
All-Pass System
 A system with a constant frequency response magnitude

The system passes all of the frequency components of its input with constant
gain/attenuation 1
z  a*
H z  
 First-order all-pass system
Pole: z=a & zero: z=1/a*
1  az 1
ap
 Most general form with real impulse response: M=N=2Mc+Mr
 A: positive constant; dk: real poles; ck: complex poles
Hap z  A

z 1  dk Mc z 1  ck* z 1  ck
Mr
 
1 
k 1 1  d k z k 1 1  
ck z 1
1  c 
* 1
kz 
 M=N Poles & Zeros
 For Causal & Stable all-pass:
 Each pole of is paired with conjugate reciprocal zero
68
All-Pass System z 1  a*
Hap z 
1  az1
Num. & denom are complex

 Magnitude response constant conjugate of each other
 j * j
 have same magnitude
 
H ap e j  
*
e a  j 1 a e
 e
1  ae  j 1  ae  j
 
| H ap e j | 1
Typical P&Z plot for an all-pass system

69
Phase of All-Pass Systems
 j * j
 
H ap e  
*
j e a  j 1 a e
 j
e
1  ae 1  ae  j
 The phase represented in polar form
 j  e  j
 re  j
  r sin     
a  re ;  j  j 
   2 arctan  
 1  re e  1  r cos     
 The group delay of this system can be written as
 e j  re j  1  r2 1  r2
grd j  j 
 
1  re e  1  2r cos     r 2
j  j
1  re e
2
 Group delay of all-pass systems is always positive
 For stable and causal system |r|<1

 For causal systems:
 Phase between 0 and  is always non-positive
 Group delay is always positive
  
arg H ap e j  0
for 0    
70
Example 5.13: First-order all-pass system
Pole: z=0.9: θ=0, r=0.9

Pole: z=-0.9: θ=pi, r=0.9
To avoid the jumps in arg[.],

differentiate to get grd[.]
71
Example 5.13: Second-order all-pass system
72
All-Pass System: Summary
73
All-Pass System: Summary
74
Applications of all-pass systems
 Compensating for phase distortion or

group delay (Ch7)
 Transforming a frequency-selective low-
pass into other frequency-selective forms
(Ch7)
 Obtaining variable-cutoff frequency-
selective filters (Ch7)
 Theory of minimum-phase systems (Ch5.6)
 stable and causal systems that have stable and causal inverses
system
 Any H(z)=Hmin(z)Hap(z)
75
Outline
o Introduction
o System analysis
o All-pass systems
76
Minimum-phase systems
 Magnitude does not uniquely characterize the

system
 Stable and causal
 poles inside unit circle, no restriction on zeros
 Zeros are also inside unit circle
inverse system is also stable and causal
(in many situations, we need inverse systems!)
Such systems are called minimum-phase systems

are stable and causal and have stable and causal
inverses
77
Minimum-phase and all-pass
decomposition
 Any rational system function can be expressed as:

H ( z)  H min ( z) H ap ( z)
 Suppose H(z) has one zero outside the unit circle at
z  1 / c* , | c | 1
H ( z )  H 1 ( z )( z 1  c * )
z 1  c *
1
 H 1 ( z )(1  cz )
1  cz 1
minimum-phase all-pass
78
Properties of minimum-phase systems:
The Minimum-Phase-Lag Property
 From minimum-phase and all-pass decomposition
H ( z)  H min ( z) H ap ( z)
arg[ H (e j )]  arg[ H min (e j )]  arg[ H ap (e j )]
 From previous figure, the continuous-phase curve of

an all-pass system is negative for 0    
 So change from minimum-phase to non-minimum-
phase (+all-pass phase) always decreases the
continuous phase
 or increases the negative of the phase (called the
phase-lag function)
 Minimum-phase is more precisely called minimum phase-
79 lag system
Properties of minimum-phase systems
 The Minimum-Group-Delay Property

 From minimum-phase and all-pass decomposition
H ( z)  H min ( z) H ap ( z)
grd [ H (e j )]  grd [ H min (e j )]  grd [ H ap (e j )]
Group delay of minimum-phase system is less than
non-minimum-phase systems
 Minimum-Energy-Delay Property n n
 |
m 0
h[ n ] 2
|   | h
m 0
min[ n ] | 2
 The energy of Min. Phase systems is delayed the

least of all systems having same magnitude
80
Example : Minimum-Phase System
 Consider the following system
1  3z 1
H1 z  
1
1  z 1
2
 One pole inside the unit circle:
 Make part of minimum-phase system
 One zero outside the unit circle:
 Add an all-pass system to reflect this zero inside the unit
circle
1 1
1 1 z
1  3z 1  1 1  1  1 1
H1 z   3 z    3 z   3
1 1 3 1 3 1
1  z 1 1  z 1  1  z 1  1  z 1
2 2 2 3
 1 1  1 1 
 1 z  z  
H1 z    3 3  3   H z H z 
min ap
 1 1 1  1 1 
 z  1  z 
 2  3 
81
Frequency response compensation
 When the distortion system is not minimum-phase

system: 1
H d ( z)  H d min ( z) H ap ( z) H c ( z) 
H d min ( z )
G( z)  H d ( z) H c ( z)  H ap ( z)
 Frequency response magnitude is compensated

82
 Phase response is the phase of the all-pass
Frequency-Response Compensation
 In some applications a signal is distorted by an LTI system

 Could filter with inverse filter to recover input signal
 Would work only with minimum-phase systems
 Make use of minimum-phase all-pass decomposition
 Invert minimum phase part
 Assume a distorting system Hd(z)
 Decompose it into
Hd z  Hd,min zHd,ap z
 Define compensating system as
1
Hc z 
Hd,min z
 Cascade of the distorting system and compensating system
1
Gz  Hc zHd z  Hd,min zHd,ap z  Hd,ap z
Hd,min z
83
Outline
o Introduction
o System analysis
oIIRand FIR Systems
o All-pass systems
o Linear Phase & Generalized linear-phase systems
84
Linear Phase & Generalized linear-
phase systems
 Linear phase filters

H (e j ) | H (e j ) | e  j 
 Generalized linear phase filters

H (e j )  A(e j )e  j  j
A(e j ) is a real function of  ,
 and  are real constants
85
Linear phase systems
 Sometime we desire a system with

 Constant frequency response magnitude or
 zero phase (when |H|=constant not possible)
Accept phase distortion, in particular linear phase
since it only introduces time shift
 Linear phase is a property of a filter
 The phase response of the filter is a linear
function of frequency
 Linear phase filter has the property of the
86
true time delay
Non-linear phase systems
 Non-linear phase
 will change the shape of the input signal even if
the magnitude response is constant
 Non-linear phase shift delays some frequency
components more than others
o A filter with non-linear phase has a group
delay that varies with frequency, resulting
in considerable phase distortion
 Most real system functions have a non-linear
phase shift as a function of frequency
87
Linear phase systems: Examples
j  j 
Ideal delay system H id (e )e , |  |  ;  real
 j
j
H id (e )    linear phase
| H id (e ) | 1
grd[ H id (e j )]    Constant group delay
sin  (n   )
 hid [n] 
 (n   )
Time-shift: when   nd with nd integer  hid [n]   [n  nd ]
Ideal lowpass with linear phase sin c (n  nd )

hlp [n] 
 (n  nd )
88
A linear phase filter has constant group

delay
All frequency components of a signal
have equal delay times
There is no distortion due to selected
frequencies
89
 Systems with constant group delay are

referred to as linear phase systems
 They are desirable when we want to minimize
the distortion on the shape of a signal
 A linear phase shift as a function of frequency
corresponds to simple delay through the time-
shifting property of the FT
90
Group delay / Phase delay
j j
grd[ H (e )] H (e )
 The group delay and phase delay (phase
shift) are functions of ω
 In a linear phase system

 Both group and phase delay are equal to the
same constant delay of the system
 The phase shift of the system increases linearly
with frequency ω
91
Group delay / Phase delay:
Examples
92
Group delay / Phase delay :
Example 5.14
sin  (n   )
hid [n] 
 (n   )
93
Group delay / Phase delay
Examples: Sound reproduction
 Components of an audio system introduce
group delay in the input audio signal
 Components: loudspeakers, filters, ..
 Important to know: the threshold of audibility of
group delay with respect to frequency
o For example:
o 1 kHz  2 ms
o 4 kHz 1.5 ms
94
Linear Phase Systems: non-integer delay
 Ideal Delay System
 
Hid e j  e j   
 Magnitude, phase, and group delay
 
Hid e j  1
H e   
id
j
grdH e   
id
j
 Impulse response
sinn   
hid n 
n   
 If =nd is integer
hid n  n  nd 
 For integer  linear phase system delays the input

yn  xn  hid n  xn  n  nd   xn  nd 
95
Linear Phase Systems: non-integer delay
 For non-integer  the output is an interpolation of samples
 Easiest way of representing is to think of it in continuous
hc t   t  T  and Hc j  e jT

 This representation can be used even if x[n] was not originally
derived from a continuous-time signal
 The output of the system is
yn  xnT  T 
 Samples of a time-shifted, band-limited interpolation of the input
sequence x[n]
 A linear phase system can be thought as
   
H e j  H e j e j 
 A zero-phase system output is delayed by 
96
Symmetry of Linear Phase Impulse
Responses
 Linear-phase =5
systems
   
H e j  H e j e j 
=4.5
 If 2 is integer
 Impulse response
symmetric
h2  n  hn
=4.3
97
Symmetry of Linear Phase Impulse Responses
98
Generalized Linear Phase System
 Generalized Linear Phase
 
   Ae e
He j j  j  j
A e j : Real function of 
 and  constants
 Additive constant in addition to linear term

 Has constant group delay
  
  grd H e j  
d
d
     
argH e j
 And linear phase of general form
  
argH e j     0
99
Condition for Generalized Linear Phase
 A generalized linear phase system response
       
H e j  A e j e j  j  A e j cos     jA e j sin   
We know
   hne
o   
H e j 
n  
 jn
  hncosn  j  hnsinn
n   n  
 The tangent of the phase angle of this system is
  hnsin n 
sin    
tan(   )   n 
cos   
 hncosn
n  
 Cross multiply to get necessary condition for generalized
linear

phase 
 hn cosn sin      hn sinn cos     0

n   n  

 hncosn sin     sinn cos     0

n  
 
100
 hn sin    n   hnsin  n     0
n   n  
Symmetry of Generalized Linear
Phase
 Necessary condition for generalized linear phase

 hnsin  n    0

n  
 For =0 or 

 hnsinn    0 
n  
 h2  n  hn
 For = /2 or 3/2


 hncosn    0 
n  
 h2  n  hn
101
Causal Generalized Linear-Phase
System
 If the system is causal and generalized linear-phase
 Since h[n]=0 for n<0 we get
hn  0 n  0 and n  M
 An FIR impulse response of length M+1 is

generalized linear phase if they are symmetric
 Here M is an even integer
102
Types of Generalized Linear-Phase FIR Filters
103
Type I FIR G. Linear-Phase
System
 Type I system is defined with

symmetric impulse response
hn  hM  n for 0  n  M

 M is an even integer
 The frequency response can be written

as
    hne
M
j  jn
He
n0
M / 2 
e  jM / 2
 an cosn
 n0 
 Where
a0  hM / 2
104 ak   2hM / 2  k  for k  1,2,..., M/2
Type II FIR G. Linear-Phase System
 Type II system is defined with

hn  hM  n for 0  n  M
 M is an odd integer

as
   hne
M
H e j   jn
n0
M1 / 2   1  
e  jM / 2
  b n cos 
  n   
 n1   2  
 Where
bk   2hM  1 / 2  k 
for k  1,2,..., M  1/2
105
Type III FIR G. Linear-Phase System
 Type III system is defined with

hn  hM  n for 0  n  M
 M is an even integer
   hne
as M
H e j   jn
n0
M / 2 
 je  jM / 2
 cn sinn
 n 1 
 Where
ck   2hM / 2  k 
for k  1,2,..., M/2
106
Type IV FIR G. Linear-Phase
System
 Type IV system is defined with

hn  hM  n for 0  n  M
 M is an odd integer
 The frequency response can be

written as M
 
H e j   hne  jn
n0
M1 / 2   1  
 je  jM / 2
  d n sin 
  n   
 n1   2  
 Where
dk   2hM  1 / 2  k 
107 for k  1,2,..., M  1/2
Location of Zeros for Symmetric Cases
 For type I and II we have

hn  hM  n 
z
  Hz  zMH z1  
 So if z0 is a zero 1/z0 is also a zero of the system
 If h[n] is real and z0 is a zero z0* is also a zero
 So for real and symmetric h[n] zeros come in sets of four
 Special cases where zeros come in pairs
 If a zero is on the unit circle reciprocal is equal to conjugate
 If a zero is real conjugate is equal to itself
 Special cases where a zero come by itself
 If z=1 both the reciprocal and conjugate is itself
 Particular importance of z=-1
H 1   1 H 1
M
 If M is odd implies that

H 1  0
 Cannot design high-pass filter with symmetric FIR filter and
108 M odd
Location of Zeros for Antisymmetric Cases
 For type III and IV we have
hn  hM  n 
z
  Hz  zMH z1  
 All properties of symmetric systems holds
 Particular importance of both z=+1 and z=-1
 If z=1
H1  H1  H1  0
• Independent from M: odd or even
 If z=-1
H 1   1 H 1
M1
• If M+1 is odd implies that
H 1  0
109
Typical Zero Locations
110
Outline
o Introduction
o System analysis
o IIR and FIR systems
oAll-pass systems
111
Log-Magnitude and Bode plots
|H(jw)| is normally converted into

decibels defined as 20 log10 | H ( ) |
Plots of 20 log10 | H ( j ) |
and H ( j ) versus log10 ( )
are known as Bode plots
112
Bode Diagram
 Although linear plots, e.g., H(jw) versus w, of frequency responses are
accurate, however, they do not always reveal important system behavior
 The plots of the two quite different-looking frequency responses look
identical
1 30
H1 ( f )  H2 ( f ) 
j 2 f  1 30  4 2 f 2  j 62 f
113
Bode diagram
o If we plot the logarithm of magnitude instead of
the magnitude, the difference can be seen more easily
Log-magnitude plots of the two frequency responses

114
Bode diagram
 A more common way of displaying frequency response is the
Bode diagram or Bode plot
 A log-magnitude plot is logarithmic in one dimension
 A Bode diagram is logarithmic in both dimensions
 A bode diagram is a plot of the logarithm of the magnitude of a

frequency response against a logarithmic frequency scale
 log |H(jw)| versus scaled w
 In a bode diagram, |H(jw)| is converted to a logarithmic scale

using the unit decibel (dB)
 The name of the unit decibel come from the original unit defined by
Bell Telephone engineers
 The bel (B), named in honor of Alexander Graham Bell, the inventor
of the telephone
115
Bode diagram
 The decibel is defined as the common logarithm (base 10) of a power ratio
Since the signal power is proportional to the square of the signal itself,
 Py   X2   X  2  X
10 log10    10 log10  2  =10 log10    = 20 log10  
 Px  Y   Y   Y 
 The plot of the previous two quite different system functions on the same
scale for better comparison:
116
Bode diagram
o LTI system are described by linear differential equations with
constant coefficients
D
dk N
dk

k 0
ak
dt k
y (t )   bk
k 0 dt k
x(t )
Y ( j )  k
b ( j ) k
 H ( j )  = k 0
X ( j ) D
 k
a
k 0
( j ) k
117
Bode diagram
o A factored form of H(jw)
  j   j    j 
1     1     ...... 1   
 z1   z2   zN 
H ( j )  A  
  j     j     j  
1     1     ......  1   
  p1     p1     pD  
 The kth pole of H(jw) occurs when jω = p

 The p’s are the values of jω at which the magnitude |H| goes to
infinity
 when referring to the p’s as poles, we mean the values of jω
where H(jw) goes to infinity
 The same holds true for the zeros of H(jw) which occur at jω = z
118
Bode diagram
 From the factored form, H(jw) can be considered as the cascade of
multiple subsystems, each having a frequency response with one pole
or one zero
 Each simple system will have a Bode diagram
 Because the magnitude Bode diagrams are plotted in decibels, which

is a logarithmic scale. the overall magnitude Bode diagram is the sum
of the individual magnitude Bode diagrams
 Phase is plotted linearly as before (against a logarithmic frequency

scale)
The overall phase Bode plot is the sum of all the phases contributed by
the subsystems.
119
Bode diagram
A system transfer function represented as a

cascade of simpler systems
120
Bode diagram
Frequency Independent Gain A:
o A is called a frequency-independent gain
o The gain constant A is assumed to be positive  the phase is
zero
o If A is negative, the phase is ±π rad.
121
Bode diagram: Component
Diagrams
1
H ( j ) 
j
One-Real-PoIe System: 1
pk
122
Bode diagram: Component
Diagrams
j
One-Real-Zero System: H ( j )  1 
zk
123
Bode diagram
1
Integrator: a system component with a single pole at zero H ( j ) 
j
124
Bode diagram
Differentiator (zero at zero): H ( j )  j
125

ELEC442: Magnitude-Phase Representation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ELEC442: Magnitude-Phase Representation

Uploaded by

Copyright:

Available Formats

ELEC442: System analysis:

Magnitude-Phase Representation of systems

1 •Dr. Zheng-Hua Tan, Digital Signal Processing III, 2009, http://kom.aau.dk/~zt/cources/DSP/

e j (o  2 ) n  e j 2n e jo n  is the phase shift (initial phase)

 An important property of sinusoids is “frequency”:

 "frequency" can mean Hz (cycles per second), or radians per second

We cannot restrict a signal x[n] to start out at zero

Phase causes signal shift and/or

 In images, as in sound signals, phase carries

 Signal delay corresponds to phase change:

Signal delay, in the frequency domain, alters only the

o Magnitude tells "how much" of a certain frequency

 LTI system is completely characterized by …

 Frequency domain: frequency response

 Often x[n] is real but its FT is, in general, complex

 Magnitude tells "how much" of a certain frequency component is

Magnitude does not uniquely characterize the system

 At any frequency ω , find the magnitude and phase of the vectors

One zero, two poles

 The amplitude and angle of the vectors v1, v2

 The phase of the frequency response of this

 The magnitude of the frequency response is

Non-causal, cannot be implemented!

H id (e j )  nd , |  |  Linear phase distortion

 Ideal lowpass filter with linear phase

•Typically, it is hard to infer much from a phase plot

o Amplitude-Phase Representation of signals

 Recall: In magnitude and phase plots:

 A signal can be expressed in terms of complex polar

 A(e j ) is real but not restricted to be positive,

 In the magnitude-phase representation: θ1(ω) has jumps of π at

 Linear constant-coefficient difference equation

 Log Magnitude in decibels (dB)

 Many (not all) systems have inverses

 Poles become zeros and vice versa

hi [n]  (0.5) n u[n]  0.9(0.5) n 1 u[n  1]; stable & causal

If | z | 0.9, left - sided h[n]

 Typically, it is hard to infer much from a phase plot

 A group delay is about the delay of the system

 In Matlab, group delay is calculated using the FT rather than

It is a measure of the linearity of the phase

y[n] | H (e j0 ) | s[n  nd ] cos(0 (n  nd )  0 )

 At ±0.2π, we get 150 sample delay

 At ±0.4π, we get 10 sample delay

Larger gain at ± 0.4π than gain at ±0.2π

 At ± 0.8π |H|<1 ,so attenuation of input x[n] occurs

o IIR and FIR Systems: impulse response of rational systems

 Example: Causal system of the form yn  ayn  1  xn

 The rational system function H z    k 0

 If H(z) does not have any poles except at z=0

 No partial fraction expansion possible (or needed)

 The impulse response can be seen to be hn   b n  k 

 Impulse response is of finite length  FIR system

 For k=0 we have a zero at z0=a

oFrequency Response of Rational Systems

 Log Magnitude in decibels (dB)

Gain in dB  20 log10 H e j 

 Phase response of a rational system function

 Here arg[.] represents the continuous (unwrapped) phase

 Phase is ambiguous: when calculating the

 Any multiple of 2 would give the same

20 log10 1  re je j  10 log10 1  r2  2r cos   

 And the group delay obtained by differentiating the phase