ELEN E4810: Digital Signal Processing

Topic 5:
Transform-Domain Systems
1. Frequency Response (FR)
2. Transfer Function (TF)
3. Phase Delay and Group Delay

1
1. Frequency Response (FR)
 Fourier analysis expresses any
signal as the sum of sinusoids
 

e.g. IDTFT: x n    X e e d
j jn

 Sinusoids are the eigenfunctions of LSI

systems (only scaled, not ‘changed’)
 Knowing
 the scaling for every sinusoid
fully describes system behavior
describes how a
 frequency response system affects each
pure frequency
2
 Sinusoids as Eigenfunctions
 IR h[n] completely describes LSI system:
x[n] h[n] y[n] = x[n] * h[n]  k hk x n  k 
j 0 n
Complex sinusoid input i.e. xn  e
 yn  k hk e 
j 0 nk 

  hk e  j 0 k
e j 0 n H(ej)
k
 = |H(ej)|ej()
 yn  H e   j 0
 
 x n  H e j 0  e j  0 n  0 

 Output is sinusoid scaled by FT at 0


3
System Response from H(e )
j 

 If x[n] is a complex sinusoid at 0

then the output of a system with IR h[n]
is the same sinusoid scaled by |H(ej)|
and phase-shifted by arg{H(ej)} = ()
where H(ej) = DTFT{h[n]}
(Any signal can be expressed as sines...)
 |H(ej)| “magnitude response”  gain

 arg{H(ej)} “phase resp.”  phase shift

4
 Real Sinusoids
 In practice signals are real e.g.
xn  A cos0 n   
X(ej)
 A
2
e 
j 0 n  
 e  j 0 n   
 A2 e j e j0 n  A2 e  j e  j0 n

- 0 0

 yn  A e j H
2 e e j 0 j 0 n
 A e j H
2 e  j 0
e  j 0 n

 Real h[n]  H(e-j) = H*(ej) = |H(ej)|e-j()

 yn  A H e  cos0n     0 
j 0

5
 Real Sinusoids
Acos(0n+) h[n] |H(ej0)|Acos(0n++(0))

 A real sinusoid of frequency 0

passed through an LSI system
with a real impulse response h[n]
has its gain modified by |H(ej0)|
and its phase shifted by (0).

6
 Transient / Steady State
 Most signals start at a finite time e.g.
j 0 n
x n  e n What is the effect?
yn  hn * x n  
n j 0 nk 
 hk e
k
 
 hk e j 0 nk 
 hk e j 0 nk 

 k kn


H e  e
j 0 j 0 n
  
kn
hk e  j 0 k
e j 0 n

Steady state Transient response

- same as with pure sine input - consequence of gating

 7
 Transient / Steady State
j 0 n
 x n  e n transient
 yn  H e  
j 0
e j 0 n
 
kn
hk e  j 0 k
e j 0 n

  FT of IR h[n]’s tail from time n onwards

 zero for FIR h[n] for n ≥ N
 tends to zero with large n for any ‘stable’ IR

8
FR example
M 1
 MA filter yn   0 x n 
1 
M
 x n * hn
-1 1 2 3 4 5 6 n

 H e   DTFT hn
j

 M 1  jn
  hne  jn
 1
 e
n M n0
1 1 e 1  j M 1 / 2 sin M /2
 jM
  e
M 1 e  j
M sin  /2

9
 FR example
 MA filter: H e  
j 1
M
e  j M 1 /2 sinM /2 
sin /2 

H e

 
j 1 sin M / 2 
M sin  / 2 

 M 1
    2
 r
(jumps at sign changes:
r=M/2)

j 0 n j1n
  Response to xn  e e ...
10
 FR example
 MA filter
j 0 n j1n
 input xn  e e
0 = 0.1  H(ej0) ~ 4/5 ej0
1 = 0.5  H(ej1) ~ -1/5 ej1


output yn  H e  
j 0
e j 0 n
 
H e j 0
e j1n


11
2. Transfer Function (TF)
Linking LCCDE, ZT & Freq. Resp...

k0 dk yn  k   k0 pk xn  k 

N M
 LCCDE:

 Take ZT: k dk z Y z  k pk z X z

k k

 
Y z   k
pk z k
 Hence: X z 
k d k z k
Transfer
 function
 or: Y z   H z X z  H(z)

12
Transfer Function (TF)
 Alternatively, yn  hn * x n
 ZT  Y z   H z X z 
  pk z k ... if system

 Note:same H(z)    d k z k
has DE form

 

 n
h n  z n
... from IR

 e.g. FIR filter, h[n] = {h0, h1,... hM-1}

M 1
 pk=hk, d0=1, DE is 1 yn   hk x n  k 
 k0

13
 Transfer Function (TF)
 Hence, MA filter:
1 1 0n M
yn  1  x n    hn  M
M
M 0
0 o.w.
M 1 n
H z   1
M  n0
z Im{z}
M
1 z zM=1 i.e.

M 1 z 

1
 M roots of 1
@ z=ejr/M 
1
Re{z}

M
z 1
 ROC?
M  z z 1
M 1
pole @ z=1 z-plane
cancels
14
 TF example
y[n] = x[n-1] - 1.2 x[n-2] + x[n-3]
+ 1.3 y[n-1] - 1.04 y[n-2] + 0.222 y[n-3]
Y z  1 2
z 1.2z  z 3
 H z  
X z 1 1.3z 1.04z  0.222z
1 2 3

 factorize:

H z 
z 
1
1   0 z 1  
1 * 1
0z  0 = 0.6+j0.8
0 = 0.3
1  0 z 1  1z 1  1z 
1 1 * 1
1 = 0.5+j0.7


15
 TF example
H z 
z 1   0 z 1  
1 1 * 1
0z 
1  0 z 1  1z 1  1z 
1 1 * 1 Im{z}
0
1 
Re{z}
0 
1
*1 
 Poles i  ROC *0
 causal  ROC is |z| > max|i|

 includes u.circle  stable

16
 TF  FR
 DTFT H(ej) = ZT H(z)|z = ej
i.e. Frequency Response is
Transfer Function eval’d on Unit Circle
p0  
1   k z  p0 z  z   k 
factor: M 1 M M

H z  k1
 k1

d0  1   k z  d0 z  z   k 
N 1 N N
k1 k1

  k
M j
p e  
 H e  e
j 0 j  NM  k1

k1 e   k 
N
d0 j

17
 TF  FR
  k
M j
p e   i, i are
H e  e
j 0 j NM  k1 TF roots
k1 e   k 
N
d0 j on z-plane

p0 k1 e   k
M j
H e  
j

d0  e j  
N
Magnitude
response
k1 k
p0 
    arg   N  M  Phase
d0  response

 arge   k   arge  k 
M j N j
 k1 k1
18
 FR: Geometric Interpretation
  k
M j
p e  
 Have H e  d e
j 0 j NM  k1

k1 e   k 
N j
0
Constant/
linear part Product/ratio of terms
related to poles/zeros
 On z-plane:
 Im{z} Each (ej - n) term corresponds
to a vector from pole/zero n to

eji point ej on the unit circle
eji ej 
Overall FR is product/ratio of
 all these vectors
Re{z}

19

 FR: Geometric Interpretation
Im{z}  Magnitude |H(ej)| is product of
lengths of vectors from zeros
i
divided by product of lengths
i 
eji of vectors from poles
eji ej 
 Phase () is sum of angles of

Re{z}
vectors from zeros
minus sum of angles of
vectors from poles


20
FR: Geometric Interpretation
 Magnitude and phase of a single zero:

 Pole is reciprocal mag. / negated phase

21
FR: Geometric Interpretation
 Multiple
poles / H z 

z  0.8e j0.3 

z  0.8e  j0.3 

zeros: z 
 0.9e j0.3 
z
 0.9e  j0.3 




22
Geom. Interp. vs. 3D surface
 3D magnitude surface for same system

23
Geom. Interp: Observations
 Roots near unit circle
 rapid changes in magnitude & phase
 zeros cause mag. minima (= 0  on u.c.)
 poles cause mag. peaks ( 1÷0= at u.c.)
 rapid change in relative angle  phase
 Pole and zero ‘near’ each other
cancel out when seen from ‘afar’;
affect behavior when z = ej gets ‘close’

24
 Filtering
 Idea: Separate information in frequency
with constructed H(ejw)
 e.g. x n  A cos1n  B cos 2 n
interested don’t care about
in this part this part

 Construct a filter: X(ej) “filtered


|H(ej1)|~1 H(ej)
out”
 
|H(e 2)|~0
j
2 1 1 2
 Then yn  hn * x n  A cos1n   1 
25
 Filtering example b
a a
 3 pt FIR filter h[n] = aba -2 -1 1 2 3 4
n
 mix of 1 = 0.1 rad/samp  to remove
and 2 = 0.4 rad/samp i.e. make
j
H(e 1) = 0

  n hne  a  be  ae
H e j  jn  j 2 j

 e b  a e  e  e b  2a cos  
 j j  j  j

 H e   b  2a cos  want = 1 at  = 0.4

j

= 0 at  = 0.1 ...
 b 13.46,a 6.76...
 26
M
Filtering example
 Filter
IR

 Freq.
resp

 input/
output

27
 3. Phase- and group-delay
 For sinusoidal input x[n] = cos0n,
 
we saw yn  H e j 0 cos 0 n    0 
gain phase shift
    0  or time shift
 i.e. cos 0 n  
   0  subtraction so
 positive p means
or cos 0 n   p  0  delay (causal)

  
 where  p    is phase delay

 28
 Phase delay example
b
 For our 3pt filter: a a
n
  e
-2 -1 1 2 3 4
j  j
H e b  2a cos  
     
 
  p      1
   
 i.e. 1 sample delay (at all frequencies)
(as observed)

29
Group Delay
 Consider a modulated carrier
e.g. x[n] = A[n]·cos(cn)
with A[n] = Acos(mn) and m << c

 Delay of envelope and carrier may differ

30
 X(ej)
Group Delay
c c
-m +m
 So: x[n] = Acos(mn)·cos(cn) c 
 A2 cos c   m n  cos c   m n
 Now: yn  hn * x n


 
H e j  c  m  cos    n 
A  c m


 
2 H e j  c  m  cos c   m n 
 
 Assume |H(ej)|~1 around cm
but (c-m) = l ; (c+m) = u ...
 31
Group Delay
cos c   m n   l  
yn  A2  
 cos c   m n   u 
  u   l    u   l 
 A cos c n   cos m n  
 2   2 
phase shift phase shift
of carrier of envelope

32
Group Delay
 If (c) is locally linear i.e.
d  
(c+D) = (c) + SD, S 
d   c
l  u
 Then carrier phase shift    c 
  c  2
so carrier delay    p, phase delay
  c
  l
 Envelope phase shift u
 m  S
d  2
 
 delay  g  c    group delay
 d   c
33
Group Delay

n
c g, group 
p, phase delay
delay 
 If () is not linear around c, A[n]
suffers “phase distortion”  correction...
34
exercise

35