CHAPTER 4:

FREQUENCY ANALYSIS OF
SIGNALS AND SYSTEMS

Lesson #13: Review of CTFT and Sampling

Lesson #14: Discrete-Time Fourier Transform (DTFT) & Inverse
DTFT
Lesson #15: DTFT properties
Lesson #16: Frequency spectrum of DT signals
Lesson #17: Frequency-domain characteristics of LTI systems
Lesson #18: Digital filters
 Lecture #13
Review of CTFT and Sampling

1. The Fourier series for CT periodic signals

2. The Fourier Transform for CT aperiodic signals (CTFT)
3. Sampling of CT signals
4. Aliasing
 The Fourier series of CT periodic signal

x(t) is periodic and satisfies the Dirichlet conditions

T
x(t)

 2

a e
j nt
Fourier series: x (t )  n
T
t
n  
T /2 2
1 j nt
Fourier coefficients: an 
T 
T / 2
x (t )e T
dt
 Power density spectrum of periodic signal
A periodic signal has infinite energy and finite average power, which is
T /2 T /2
1 1
Px   | x (t ) | dt   x(t ). x * (t )dt
2

T T / 2
T T / 2
2
1
T /2
   nt 
x (t )  an e
j
  dt
* T
T T / 2  n   
 
* 1
T /2
j
2
 
  an  dt    | an |2
nt

n   T

T / 2
x (t )e T

 n  
T /2 
1
Parseval’s relation: Px 
T  | x ( t ) |2
dt   n
| a
n  
|2

T / 2
 Lecture #13
Review of CTFT and Sampling

1. The Fourier series for CT periodic signals

2. The Fourier Transform for CT aperiodic signals
(CTFT)
3. Sampling of CT signals
4. Aliasing
The Fourier transform of CT aperiodic signal

x(t) is aperiodic and satisfies the Dirichlet conditions

Fourier transform pair:


X( )   x (t )e
 jt
dt  FT x (t )


x (t ) 
1
 X ( )e
jt
d  FT 1
X ( )
2 
 Fourier transform properties

F
Linearity: ax(t)  by(t)  aX ( )  bY ( )
F
- j 
Time shift: x (t   )  e X ( )
F
Time reverse: x ( t )  X (  )
F
Convolution in
time domain:
x (t ) * y (t )  X ( ).Y ( )
F 1
Multiplication in x (t ). y (t )  X ( ) * Y ( )
time domain: 2
 Energy density spectrum of aperiodic signal
Let x(t) be finite energy signal, which is
 
Ex   dt   x(t ). x * (t )dt
2
| x ( t ) |
 

 1 

  x (t )  X * ( )e
 jt
d dt
  2  
1

 
 1


 X * ( )  x (t )e dt d   | X ( ) |
 jt
 2
d
2     2 

 
1
Parseval’s relation: E x   | x(t ) | dt 
2
 | X ( ) |2
d

2 
 Lecture #13
Review of CTFT and Sampling

1. The Fourier series for CT periodic signals

2. The Fourier Transform for CT aperiodic signals (CTFT)
3. Sampling of CT signals
4. Aliasing
 Sampling of CT signals
Define the CT impulse train as:

x(t) is CT signal we want to sample:

To sample x(t), we will multiply x(t) by p(t)

 Sampling of CT signals

Let xs(t) be the sampled signal. Then,

x(t )  
 xs (t )
sampling

 
xs (t )   x(t ) (t  nT )   x(nT ) (t  nT )
n   n  

Signal xs(t) consists

of a train of CT
impulses – take off
the arrow heads to
get x(n) – a DT
signal
 Spectrum of sampled signals

Consider y(t) = xs(t) = x(t).p(t) in the frequency domain

Take Fourier transform of x(t)

X(ω) = FT{x(t)} and P(ω) = FT{p(t)}

 Spectrum of sampled signals

Finding FT{p(t)}, using the continuous-time FT of periodic signals

 
p(t )    (t
n  
 nT ) 
FT
P( )   2an (  ns )
n  

1 T /2 1
an    (t )e dt 
T T / 2
 jt

T
 P( )     (  n )
n  
s s

A CT impulse train has a FT that is an impulse train in frequency

Spacing between pulses in time is T
Spacing between pulses in frequency is 2π/T
Increasing period in time domain decreases it in frequency domain
 Spectrum of sampled signals
Back to Xs(ω); with ωs: the sampling frequency

1   
X s ( )  X ( ) *   s (  ns )
2 n   
s 
1 

2

n  
X (  ns )   X (  ns )
T n  
The effect of sampling is an infinite sum of scaled, shifted
copies of the continuous time signal's Fourier Transform
 Lecture #13
Review of CTFT and Sampling

1. The Fourier series for CT periodic signals

2. The Fourier Transform for CT aperiodic signals (CTFT)
3. Sampling of CT signals
4. Aliasing
 Aliasing

Note that the triangles don’t overlap – so with an ideal

low pass filter with cut-off frequency ω=π/T, we could
filter xs(t) to perfectly recover x(t)
 Aliasing (cont)

- The triangles overlap  can’t recover x(t)

- Happens when ωs = 2π/T < 2ωb

0 ωb ωs
Aliasing (cont)

Avoid
aliasing:
sampling faster
than twice the
highest
frequency
component –
The Nyquist-
Shannon
sampling
theorem
Examples
Examples
Examples
Examples
 Lecture #14
Discrete-Time Fourier Transform (DTFT)

1. From CTFT to DTFT

2. Convergence of the DTFT
3. The relation between DTFT and ZT
4. DTFT of signals with poles on the unit circle
5. Inverse DTFT
 From CTFT to DTFT

Take a CT signal x(t) and sample it:


x(t ) sampling
  xs (t )   x(nT ) (t  nT )
n  
The CTFT of the sampled signal is:

FTx s ( t )   x(nT)FT(t  nT)
n  
 
 
n  
x (nT)  ( t  nT)e  jt dt

 T 1 
  x (
n  
nT ) e  jnT

T  
 x (
n  
n ) e  jn
 X ()
 DTFT formula


X()  DTFTx[n ]   x[n] e  j n

n  

• Discrete in time, but continuous in frequency and periodic

with period of 2
• Gives the complex frequency spectrum of DT signal
• Not all DTFT is converge
 Lecture #14
Discrete-Time Fourier Transform (DTFT)

1. From CTFT to DTFT

2. Convergence of the DTFT
3. The relation between DTFT and ZT
4. DTFT of signals with poles on the unit circle
5. Inverse DTFT
 Convergence of the DTFT
We always have:  

 x[n ]e
n  
 jn
  x[n ]e
n  
 jn


  x[n ] e
n  
 jn

 DTFT exists when:

  x[n ]
n  


 x[n ]  
absolutely
summable
n  
 Examples

1) Find DTFT of x(n) where x[n]  a u[n]

n

  j
1 e
X()   a e n  jn
  (ae ) 
 j n
 j
 DTFT nexists
0 when: n 0 1  ae  j
e a
if |a| < 1
 Examples

2) Find DTFT of x(n) where y[n]  a u[n]

n

0 
1 a
Y()  a e
n  
n  jn
  (a e ) 
n 0
1 j n

1 a e
1 j
 j
e a
if |a| > 1
 Examples

3) Find DTFT of p(n) where p[n]  u[n]  u[n  N ]

Show that this DTFT has a linear phase term
N 1
1 e  jN

P()  1.e  jn


n 0 1 e  j

Phase: -Ω(N-1)/2  linear in phase

 Examples

4) Find DTFT of h(n) where h[n]   [n]  2 [n 1]  2 [n  2]   [n  3]

Show that this DTFT has a linear phase term
3
H()   h[n ]e  jn
 1  2e  j
 2e  j2 
e  j3 

n 0

Phase: -3Ω/2  linear in phase

 Lecture #14
Discrete-Time Fourier Transform (DTFT)

1. From CTFT to DTFT

2. Convergence of the DTFT
3. The relation between DTFT and ZT
4. DTFT of signals with poles on the unit circle
5. Inverse DTFT
 From ZT to DTFT

Recall ZT of x(t):

X( z)   x[n]z
n  
n

Evaluating X(z) on the unit circle, if the unit circle is in the

ROC of X(z)


X( z )
z e j 
  x[n]e
n  
 j n
 X()
 From ZT to DTFT

DTFT is the Z-transform of x(n) evaluated on the unit circle

X ()  X ( z )
z e j 

If the ROC of the ZT contains the unit circle, we can get the
DTFT from the ZT by substitution z = ejΩ
 Example

Find DTFT of x(n) where x[n]  a u[n] n

z
X ( z)  ROC :| z || a |
za
 DTFT exists when ROC includes the unit circle, which
means |a| < 1:
e j
X ( )  X ( z )  j
z  e j e a
 Lecture #14
Discrete-Time Fourier Transform (DTFT)

1. From CTFT to DTFT

2. Convergence of the DTFT
3. The relation between DTFT and ZT
4. DTFT of signals with poles on the unit circle
5. Inverse DTFT
DTFT of signal with poles on the unit circle

DTFT of x[n] can be determined by evaluating its ZT

X(z) on the unit circle, provided that the unit circle lies
within the ROC, or none of poles is on the unit circle

Rescue: just extend the DTFT!!!

Allowing the DTFT to contain impulses at certain
frequencies corresponding to the location of the poles
 Examples
Ex1: Consider this signal x[n] = u[n]
z
X ( z)  ROC :| z | 1  DTFT does not exist
z 1
Evaluate X(z) on the unit circle, except at z = 1 = ejk2π:
j j / 2 j / 2 j / 2
e e .e e
X ()  j  j / 2 j / 2  j / 2    k 2
e  1 e (e e ) 2 j sin( / 2)

1
X ( )    k 2
2 | sin( / 2) |

At Ω = k2π, X(Ω) contains impulses

 Examples
Ex2: Consider this signal x[n] = (-1)nu[n]
z
X ( z)  ROC :| z | 1  DTFT does not exist
z 1
Evaluate X(z) on the unit circle, except at z = -1 = ej(2k+1)π

e j e j / 2 .e j / 2 e j / 2
X ()  j  j / 2 j / 2  j / 2    (2k  1)
e  1 e (e e ) 2 cos( / 2)

1
X ( )    (2k  1)
2 | cos( / 2) |

At Ω = (2k+1)π, X(Ω) contains impulses

 Examples
Ex3: Consider this signal x[n] = (cosΩ0n)u[n]
z  cos 0
X ( z)  2 ROC :| z | 1  DTFT does not exist
z  2 z cos 0  1

Evaluate X(z) on the unit circle, except at

 Lecture #14
Discrete-Time Fourier Transform (DTFT)

1. From CTFT to DTFT

2. Convergence of the DTFT
3. The relation between DTFT and ZT
4. DTFT of signals with poles on the unit circle
5. Inverse DTFT
 The inverse of DTFT

X()   x[ n
n  
] e  j n

 
1 1  
 j n  j l

2  
j l
X()e d    
2   n  
x[n ]e 

e d

  1  j ( l n ) 
  x[n ]  e d  x[l]
n    2   

1

j n
x[n ]  X() e d
2  
 Examples

1,    c
Ex1. Find x(n) from its DTFT X(Ω): X ( )  
0,  c    


1 c
1 jn c sin c n
x[n ]   1.e d  
jn c
.e c
2   c
2jn  c n
0.7

0.6

0.5

0.4

0.3

0.2

0.1

-0.1
-π -Ωc 0 Ωc π Ω -0.2
-20 -15 -10 -5 0 5 10 15 20
 Examples

Ex2. Find x(n) from its DTFT X(Ω): X()  cos 2 

2
 e  e  1 j2  1 1  j2 
j  j

X ()     e   e
 2  4 2 4
1 1 1
 x[n ]  [n  2]  [n ]  [n  2]
4 2 4
1 0.5

0.9 0.45

0.8 0.4

0.7 0.35

0.6 0.3

0.5 0.25

0.4 0.2

0.3 0.15

0.2 0.1

0.1 0.05

0 0
-2 0 2 -5 0 5
 Examples

e j
Ex3. Find x(n) from its DTFT X(Ω): X() 
e j  2

z
X(z)  | z | 2
z2
 x[n ]  2 u[n  1]
n
SUMMARY
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Linearity

DTFT
ax[n ]  by[n]  aX()  bY()

The DTFT of a linear combination of two or more signals is

equal to the same linear combination of the DTFT of the
individual signals.
 Example
n
Determine the DTFT of the signal x[n]  a if |a| < 1

a n  0
n
 1 
n

x[n ]   n  a u[n ]     u[n  1]

n

a n0   a  
z z 1
X ( z)   ROC :| a || z |
z  a z  1/ a |a |
j j j j
e e e ae
 X ()  j  j  j  j
e  a e  1 / a e  a ae  1
1  a2 1  a2
 j  j

1  ae  ae  a 2
1  2a cos   a 2
 Example (cont)
1 For a = 0.8

0.5

0
-20 -15 -10 -5 0 5 10 15 20

10

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Time shifting

DTFT
x[n  n 0 ]  e  j n 0
X()
Proof: infer from the shifting property of ZT, then
evaluate ZT on the unit circle
ZT
x[n  n 0 ]  z X(z) n 0

 A shift in time causes a linear phase shift in

frequency – no change in DTFT magnitude
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Frequency shifting and modulation
DTFT
e j 0 n
x[n ]  X (   0 )
DTFT
1 1
cos(0 n ) x[n ]  X (   0 )  X (   0 )
2 2
Proof:
DTFT  
e j 0 n
x[n]   (e
n  
j 0 n
x[n])e  jn
  x[ n
n  
]e  j (   0 ) n
 X (   0 )

1 j 0 n 1  j  0 n
cos( 0 n)  e  e  ...
2 2

Modulation causes a shift in frequency

 Example

y1[n]  x[n] cos 0.5n and y2 [n]  x[n] cos n

X(Ω)

-2π -π -π/2 0 π/2 π 2π Ω

Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Differentiation in the frequency domain
DTFT dX ()
nx[n]  j
d

Proof: infer from the differentiation-in-the-z-domain property

of ZT, then evaluate ZT on the unit circle

Z dX ( z )
nx[n ]   z
dz
j dX () j dX () dX ()
DTFT
nx[n ]   e j
 e j
 j
d (e ) je d d
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Convolution in time domain

DTFT
x1[n ]  x 2 [n]  X1 ().X2 ()

Proof:

Infer from the convolution-in-time-domain property

of ZT, then evaluate ZT on the unit circle

Convolution in time  Multiplication in frequency

 Example
Given h[n] = anu[n], |a| < 1

Find its inverse system hi[n] but don’t use ZT

j
e
H ()  j
e a
j
e a  j
 H i()  j
 1  ae
e
 hi [n ]   [n ]  a [n  1]
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Convolution in frequency domain
DTFT
1 1
x1[n ].x 2 [n ]  
2 2 
X1 ()X 2 (  )d  X1 ()  X 2 ()
2
DTFT 
x1[n ].x 2 [n ]   1
( x
n  
[ n ].x 2
[ n ])e  jn

Multiplication 
 1 
in time    X (  ) e j n
d  2
x [ n ]e  jn

n    2 2  
1

1  x [n ] e  j(  ) n d

  X (  ) 2 
2 2   n  
1

1

2 2 
 X1
(  ) X 2
(    ) d Convolution in
frequency
Lecture #15
DTFT properties

1. Linearity
2. Time shifting
3. Frequency shifting and modulation
4. Differentiation in the frequency domain
5. Convolution in time domain
6. Convolution in frequency domain
7. Symmetry
 Symmetry properties
Consider x[n] and X(Ω) are complex-valued functions

X ()   x[
n  
n ]e  jn


 (x
n  
R [n ]  jx I [n ])(cos n  j sin n )

 X R ()  jX I ()

 X R()  (x
n  
R [n ] cos n  xI [n ] sin n )

 X I()    ( xR [n ] sin n  x I [n ] cos n )
n  
 Symmetry properties
Consider x[n] and X(Ω) are complex-valued functions

1

jn
x[n ]  X (  ) e d
2 

1

2 
[ X R ()  jX I ()](cos n  j sin n )d

 xR [n ]  jx I [n ]

1
 xR [n ] 
2 [ X

R () cos n  X I () sin n ]d


1
 x I [n ] 
2 [ X

R () sin n  X I () cos n ]d
 Real signal
x[n ]  xR [n ] and x I [n ]  0

 X R()   x[n] cos n  X
n  
( ) : even
R


 X I()    x[n ] sin n   X I( ) : odd
n  

 X * ()  X ( )
| X () | X 2 ()  X 2 () | X ( ) |: even
 R i
 X I ()
X ()  arctg  X ( ) : odd
 X R ()
 Real signal (cont)

x[n ] even :

X R ()  x[0]  2 x[n ] cos n; X I ()  0
n 1

1
x[n ] 
  X
0
R () cos n d

x[n ] odd :

X I ()  2 x[n ] sin n; X R ()  0
n 1

1
x[n ]  
  X
0
I () sin n d
 x[ n ] X ()
Symmetry x * [n ] X * ( )
properties x * [  n ]
X * ()
x R [n ] X e ()
jx I [n ] X o ()
xe [ n ] X R ()
xo [n ] jX I ()
X ()  X * ( )
any real signal | X () || X ( ) |
X ()  X ( )
real and even real and even
real and odd imaginary and odd
 Lecture #16
Frequency spectrum of DT signals

1. Frequency spectrum
2. Amplitude spectrum and phase spectrum
3. Energy spectral density (ESD)
4. Bandwidth
 Frequency spectrum

 Representation of the signal in the frequency domain

 Being generated via frequency analysis tools:
- CTFT for aperiodic CT signal
- CT Fourier series for periodic CT signal
- DTFT for aperiodic DT signal
- DT Fourier series for periodic DT signal
 Frequency spectrum analysis

 The technical process of decomposing a complex signal into

simpler parts.
 The process of quantifying the various amounts (e.g. amplitudes,
powers, intensities, or phases), versus frequency.
 Summary:
- Aperiodic CT signals have aperiodic continuous-frequency spectra
- Periodic CT signals have aperiodic discrete-frequency spectra
- Aperiodic DT signals have periodic continuous-frequency spectra
- Periodic DT signals have periodic discrete-frequency spectra
Frequency range of some biological signals
Type of signal Frequency range (Hz)
Electroretinogram 0-20
Pneumogram 0-40
Electrocardiogram (ECG) 0-100
Electroencephalogram (EEG) 0-100
Electromyogram 10-200
Sphygmomanogram 0-200
100-4000
Frequency range of some biological signals

Type of signal Frequency range (Hz)

Wind noise 100-1000
Seismic exploration signals 10-100
Earthquake and nuclear explosion signals 0.01-10
Seismic signals 0.1-1
 Frequency range of some EM signals
Type of signal Frequency range (Hz)
Radio broadcast 3x10-4 – 3x106
Shortwave radio signals 3x106 – 3x1010
Radar, satellite communications… 3x108 – 3x1010
Infrared 3x1011 – 3x1014
Visible light 3.7x1014 – 7.7x1014
Ultraviolet 3x1015 – 3x1016
X-rays 3x1017 – 3x1018
 Frequency spectrum representation

 Real and imaginary parts:

X ()  X R ()  jX I ()

 Absolute and argument:

 X () 
jarctg I 

X () | X () | e j (  )  [ X R ()]2  [ X I ()]2 e  R 
X ( )

 Magnitude and phase:

X ()  A()e j (  )
A() can be positive, negative or zero
 ()  2k if A()  0
 ()  
 ()  ( 2k  1) if A()  0
 Example

 Consider this even signal:

1,  2  n  2
x[n ]  
0, elsewhere
 Its frequency spectrum:
2
X ()  X R ()  1  2 cos n
n 1

 1  2 cos n  2 cos 2n

 Magnitude and phase:  A()  1  2 cos n  2 cos 2n
 ()  0
 Absolute and argument:| X () || 1  2 cos n  2 cos 2n |
0 if X ()  0
 ()  
 if X ()  0
 Lecture #16
Frequency spectrum of DT signals

1. Frequency spectrum
2. Amplitude spectrum and phase spectrum
3. Energy spectral density (ESD)
4. Bandwidth
 Amplitude spectrum and phase spectrum

j (  )
X ()  X () e
Amplitude spectrum Phase spectrum
 
X ( )   x[n]e
n  
 jn
; X (  )   x[n]e
n  
jn

 X ( )  X (  )
*

 | X() || X() | and X()  X()

 Example

Find and plot amplitude spectrum and phase spectrum:

x[n] = u[n] - u[n - 4]

 Using Matlab to plot amplitude
spectrum and phase spectrum

w = -2*pi:pi/255:2*pi; % freq. -2π  2π, resolution of π/255

X =4*sinc(2*w/pi)./sinc(w/(2*pi)).*exp(-j*1.5*w);
subplot(2,1,1);
plot(w/pi,abs(X)); % plot amplitude spectrum
subplot(2,1,2);
plot(w/pi,phase(X)); % plot phase spectrum
 Amplitude spectrum
6

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

-2

-4
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Phase spectrum
 Lecture #16
Frequency spectrum of DT signals

1. Frequency spectrum
2. Amplitude spectrum and phase spectrum
3. Energy spectral density (ESD)
4. Bandwidth
 Energy spectral density (ESD)


    1 
E   | x[n] |   x[n]x [n]   x[n]
2 *
 X
*
()e  jn
d
n   n   n    2  

Changing the order of summation & intergral:

 
1  
 jn  1
E  X ()   x[n ]e d   X() d
* 2

2    2  
n  
Energy spectral density
 Example

Given x[n] = anu[n], -1<a<1

1
X ( ) 
1  ae  j
 Example
3

2.5
mat do pho nang luong

E
S 1.5

D
1

0.5

0
-3 -2 -1 0 1 2 3
tan so omega
 Lecture #16
Frequency spectrum of DT signals

1. Frequency spectrum
2. Amplitude spectrum and phase spectrum
3. Energy spectral density (ESD)
4. Bandwidth
 Bandwidth of a signal

 Bandwidth: the range of frequencies over which

the power or energy density spectrum is
concentrated.
 Ex: a signal has 95% of its ESD or PSD concentrated
in the range from F1 to F2 then the 95% bandwidth
of the signal is (F2-F1).
 Similarly, we may define 75% or 90% or 99%
bandwidth of the signal
 Frequency-domain classification of signals

 Low-frequency signal: PSD or ESD concentrates about zero

frequency
 High-frequency signal: PSD or ESD concentrates at high
frequencies
 Bandpass signal: PSD or ESD concentrates somewhere in the
broad frequency range between low frequency and high frequency
- Narrowband signal: the bandwidth (F2-F1) is much smaller (by a
factor of 10+) than (F1+F2)/2
- Wideband signal
 Frequency-domain classification of signals

Low-frequency High-frequency
signal signal

Ω Ω

Narrowband signal Wideband signal

Ω Ω
Lecture #17
Frequency-domain characteristics of LTI systems

1. Frequency response function

2. Response to complex exponential and sinusoidal
signals
3. Steady-state and transient response
4. Response to aperiodic input signals
 Frequency response

 For impulse response, h(n), its DTFT is often called frequency

response H(Ω)
 H(Ω) completely characterizes a LTI system in the frequency
domain
 H(Ω) allows us to determine the steady-state response of the
system to any arbitrary weighted linear combination of sinusoids or
complex exponential
 H(Ω) exists if the system is BIBO stable

H ()  H () e j ()

Amplitude response Phase response

 Determination of frequency response
1. From impulse response: Just take DTFT of h[n]

2. From difference equation:

- Take DTFT for both side
- Put the Y(Ω) on left side
- Divide both side by X(Ω)

3. From block diagram:

(1) Find the difference equation, then find H(Ω) from equation
(2) Put X(Ω)=1 as input, then directly find Y(Ω)=H(Ω)

4. From transfer function:

Evaluate H(z) on the unit circle
 Example of frequency response

 A LTI causal system is described by the following equation:

y[n]  0.1y[n 1]  0.85 y[n  2]  x[n]  0.3x[n 1]

 First, checking the stability of the system (by using Matlab):
b = [1 -0.3];
a = [1 0.1 0.85];
zplane(b,a) % plot zeros and poles to check if all poles are
inside the unit circle

 Second, take DTFT for two sides:

1  0.3e  j
H () 
1  0.1e  j  0.85e  j 2
Example of amplitude and phase responses
1
H() 
1  0.4e  j
Example of amplitude and phase responses
1
H() 
1  0.4e  j
2

1.5

0.5
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

0.5

-0.5
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
 Example of moving average filter
1
y[n ]  ( x[n  1}  x[n ]  x[n  1])
3
1 j  j 1
H ()  ( e  1  e )  (1  2 cos )
3 3
Hence
1
| H () | | 1  2 cos  |
3
0, 0    2 / 3
H ()  
 , 2 / 3    
Example of moving average filter
1

0.5

0
0 0.2 0.4 0.6 0.8 1

0
0 0.2 0.4 0.6 0.8 1
Lecture #17
Frequency-domain characteristics of LTI systems

1. Frequency response function

2. Response to complex exponential and
sinusoidal signals
3. Steady-state and transient response
4. Response to aperiodic input signals
Response to complex exponential signals


x[n]  Ae j0 n
,   n   y[n ]   h[k] x[n  k]
k  

 

y[n]   h[ k
k  
] Ae j 0 ( n  k )

 
 A  h[k ] e  
 j 0 k  j  0 n
e
k   
j 0 n
 ( Ae ) H ( 0 )  x[n]H ( 0 )

y[n]  x[n]H (0 )  Ae j0n H (0 )

eigenfunction eigenvalue
 Example
Determine the output signal of system h[n]  (1 / 2) n u[n]

j n
to this input signal x[n ]  Ae 2
,   n  

1
H ()   h[n ]e  jn

1  j
n   1 e
2
   1 2  j 26.60
At n  : H     e
2  2  1 j 2
1
5

    
2  j26.6 2A
j n j n  26.60 
y[n ]  x[n ]H   Ae . e 
0
2 
2
e
2 5 5
 Response to sinusoidal signals
A j n A  j n
x[n ]  A cos(0 n )  e  e 0
, n  0

2 2
A j0n A  j0n
y[ n ]  e H (  0 )  e H (  0 )
2 2
A j0n jH ( 0 ) A  j0n
 e | H (0 ) | e  e | H (0 ) | e  jH ( 0 )
2 2
A
 | H (0 ) | e e
2

j0n jH ( 0 )
e  j0n  jH ( 0 )
e 
y[n]  A | H (0 ) | cos0n  H (0 )
 Example

Determine the response of the system h[n]  (1 / 2) n u[n]

to the input signal

x[n ]  10  5 sin n  20 cos n,    n  
2
 One more example

 A LTI causal system is described by the following equation:

y[n] = ay[n-1] + bx[n], a real and 0<a<1
a) Determine the amplitude and phase responses
b) Choose b so that the maximum value of |H(Ω)| is unity,
and sketch |H(Ω)| and <H(Ω) for a = 0.9
c) Determine the output of the system to the input:

  
x[n]  5  12 sin n  20 cos  n  ,    n  
2  4
 One more example

a) The frequency response is:

b
H ( ) 
1  ae  j
The amplitude response and phase response are:
 One more example

Choose a = 0.9 and b = 1-a :

1 a 0.1
| H () | 
1  a 2  2a cos  1.81  1.8 cos 
0.9 sin 
H ()  0  arctg
1  0.9 cos 
 One more example
b) 1

0.5

0
0 0.2 0.4 0.6 0.8 1

-0.5

-1

-1.5
0 0.2 0.4 0.6 0.8 1
 One more example
c) The amplitude and phase response at Ω=0, π/2, and π
respectively:

The output:
Lecture #17
Frequency-domain characteristics of LTI systems

1. Frequency response function

2. Response to complex exponential and sinusoidal
signals
3. Steady-state and transient response
4. Response to aperiodic input signals
 Steady-state and transient response

The system response can be considered as a sum of 2 terms:

y[n]  yss[n]  ytr [n]

yss[n]: steady-state response
.

ytr[n]: transient response, decays toward zero as n  ∞

In many practical applications, the transient response is
unimportant and therefore, it is usually ignored
 Example

The LTI system described by first-order equation:

y[n]  ay[n  1]  x[n] |a| < 1

jn
Its response to the input: x[n]  Ae ,n0
is determined as
.

n 1  j ( n 1)
Aa e A
y[n ]  a y[1] 
n 1
e jn
 e jn
, n0
1  ae  j 1  ae  j

where y[-1] is the initial condition

 Example

n 1  j ( n 1)
Aa e A
y[n ]  a y[1] 
n 1
e jn
 e jn

1  ae  j
1  ae  j

Transient response Steady-state response

yss[n]  AH ()e jn

 AH (e )e j jn

M M
x[n ]   A k z n
k  yss[n ]   A k H(z k )z n
k
k 1 k 1
Lecture #17
Frequency-domain characteristics of LTI systems

1. Frequency response function

2. Response to complex exponential and sinusoidal
signals
3. Steady-state and transient response
4. Response to aperiodic input signals
 Response to aperiodic input signals
x[n] y[n]
LTI
X(Ω) Y(Ω)

 From the convolution property, we have:

Y ( )  X ( ) H ( ) For   0 :
| Y () || X (
. ) || H ( ) | | Y (0 ) || X (0 ) || H (0 ) |
Y ()  X ()  H () Y (0 )  X (0 )  H (0 )

 The LTI system attenuates some frequency components and

amplifies other frequency components of the input
 The output can’t contain frequency components that are not
contained in the input
 Example
 A LTI causal system is described by the following equation:
y[n] – 0.5y[n-1] = x[n]
Determine the spectrum and the ESD of the output when the
n
input signal is: 1
x[n ]    u[n ]
4
 The frequency response: H ()  1
1  12 e  j
The input spectrum: 1
 X () 
1  14 e  j
The output spectrum: 1
Y ()  X (). H () 
1  14 e j 1  12 e j 


1
ESD | Y () |  2

 45  cos  1716  12 cos  

 Example

1.6 2

1.4
X(Ω) 1.5 H(Ω)
1.2
1
1

0.8 0.5
0 0.5 1 0 0.5 1

3 80
Y(Ω) 60 ESD
2
40
1
20

0 0
0 0.5 1 0 0.5 1
Lecture #18
Digital filters

1. Digital filters
2. Ideal filters
 Filters

 Provide a convenient means to change the nature of a signal.

 Change the frequency characteristics of a signal in a specific way,
letting some frequencies in the signal pass while blocking others
 4 basic types:
.
- Low pass filter (LPF): lets low frequencies pass while blocking
high frequencies
- High pass filter (HPF): does the opposite
- Band pass filter (BPF): allows a band of frequencies to pass
- Band stop filter (BSF): block all frequencies inside a band
 Digital filters

 DT systems that perform mathematical operations on

a DT signal to reduce or enhance certain aspects of that
signal.
 Digital filters are difference equations defined by a list
of filter coefficients
.

 Ex: Moving average filter:

y[n] = 1/3{x[n+1]+x[n]+x[n-1]}
 Typical applications of digital filters

 Noise suppression: radio signal, biomedical signal, analog

media signal...
 Enhancement of selected frequency range: treble/bass
control, equalizers in audio, image edge enhancement…
 Bandwidth limiting:
.
aliasing prevention, interference
avoidance…
 Removal of specific frequencies: DC removal, 60 Hz signal
removal, notch filter…
 Special operations: differentiation, integration, phase shift…
Example of signal after lowpass filtering

.
Example of signal after highpass filtering

.
Lecture #18
Digital filters

1. Digital filters
2. Ideal filters
 Ideal digital filters

Cx[n  n 0 ], 1    2
y[n ]  
0 , 

Y()  CX()e  j n 0
 X()H(), 1    2

Ce  j n , 1    2
0

H ( )  
0,  
| H() | C, 1    2 - Constant gain
()   n 0 , 1    2 - Linear phase
 Amplitude responses for some
ideal digital filters
| H() | | H() |

 
c c

| H() | | H() |

 
a b a b
 Ideal digital lowpass filters

1,   c
Frequency response: H L ( )  
0, c    


c
1 c sin c n
 1.e
jn
hL [n]  d 
2 c
 c n
0.7

0.6

0.5

0.4

0.3

0.2

0.1

-0.1
-π -Ωc 0 Ωc π Ω -0.2
-20 -15 -10 -5 0 5 10 15 20
 Ideal digital highpass filters
Frequency response: 1, c    

H H ( )  
0,   c


sin  n c sin c n
H H ()  H L 2 ()  H L1 ()  h[n]  
n  c n
0.4

HL1
HL2
0.3

0.2

0.1

-0.1

-π -Ωc 0 Ωc π Ω -0.2

-0.3
-20 -15 -10 -5 0 5 10 15 20
 Ideal digital bandpass filters
Frequency response: 1, c1    c 2  
H B ( )  
0, elsewhere

c 2 sin c 2n c1 sin c1n

H H ()  H L 2 ()  H L1 ()  h[n]  
 c 2 n  c1n
0.1

HL1 0.08

HL2 0.06

0.04

0.02

-0.02

-0.04

-π -Ωc2 0 Ωc1 π Ω -0.06

-0.08
-20 -15 -10 -5 0 5 10 15 20
 Ideal digital bandstop filters
1,   c1
Frequency response: 
H B ()  0, c1    c 2

1, c 2    
c1 sin c1n c 2 sin c 2n
H H ()  H L ()  H H ()  h[n]    [n ] 
 c1n  c 2 n
0.6

HH HL 0.5

0.4

0.3

0.2

0.1

-π -Ωc2 0 Ωc1 π Ω 0

-0.1
-20 -15 -10 -5 0 5 10 15 20
Amplitude responses for some actual digital filters