Signal Processing Fundamentals – Part I

Spectrum Analysis and Filtering

3. Spectrum Representation

Magnitude Spectrum for square wave

1.4

1.2

1
Amplitude

0.8

0.6

0.4

0.2

0
-200 -150 -100 -50 0 50 100 150 200
Frequency f (Hz)

3. Spectrum Representation

1
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

What is Spectrum?
A graphical representation of the
frequency content of a signal

• Allow us to see interrelationships among the different

frequency components and their relative amplitudes
• Applications:
• Equalizer of Hi-Fi
• Allow key changing for Karaoke
• Enable speech recognition
• Determine a star is flying away or towards us (Doppler’s
effect)
2
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Spectrum of a Sum of Sinusoids

• Many signals are generated by adding together a constant
and N sinusoids, each with different frequency, amplitude
and phase
N
x(t ) = A0 + ∑ Ak cos(2πf k t + φk )
k =1

If N = 1,
x(t ) = A0 + A1 cos(2πf1t + φ1 )
If N = 2,
x(t ) = A0 + A1 cos(2πf1t + φ1 ) + A2 cos(2πf 2t + φ2 )
3
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

N
x(t ) = A0 + ∑ Ak cos(2πf k t + φk )
k =1

• Express in phasor form

{ }
N
x(t ) = X 0 + ∑ Re X k e j 2πf k t

k =1

Where
jφ k
X 0 = A0 , X k = Ak e
4
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

• For any complex number A= a + jb

Re{ A } = Re{ a + jb } = a = (a + jb + a – jb)/2

= (A + A*)/2
Where A* is the complex conjugate of A

Hence

{ }=
j 2πf k t * − j 2πf k t
j 2πf k t X ke + X ke
Re X k e
2
5
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

{
Re X k e j 2πf k t
} 
=  X ke

j 2πf k t
(
+ X ke )
j 2πf k t * 
/2


(X ek
j 2πf k t *
) = (A e k )
j (2πf k t +φk ) *

= Ak cos(2πf k t + φk )
− jAk sin (2πf k t + φk )
− j (2πf k t +φk ) − jφk − j 2πf k t
= Ak e = Ak e e
* − j 2πf k t
6
= X ke
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

{ }
N
x(t ) = X 0 + ∑ Re X k e j 2πf k t
k =1
 X k j 2πf t X k* − j 2πf t 
N
⇒ x(t ) = X 0 + ∑  e k + e k

k =1  2 2 
• We define the two-side spectrum (or spectrum) of
a signal to be the set of 2N+1 complex phasors X
and 2N+1 frequencies f
  X1   X 1
*   X2   X 2*   XN   X N* 
( X 0 ,0 ),  , f1 , ,− f1 ,  , f 2 ,  ,− f 2  L  , f N ,  ,− f N  
  2   2  2
   2   2
   2 

7
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Example
x(t ) = 10 + 14 cos(2π 100t − π / 3) + 8 cos(2π 250t + π / 2 )
− jπ / 3 j 2π 100t jπ / 3 − j 2π 100t
x(t ) = 10 + 7e e + 7e e
+ 4e jπ / 2e j 2π 250t + 4e − jπ / 2e − j 2π 250t

( )(
(10,0 ), 7e − jπ / 3 ,100 , 7e jπ / 3 ,−100 ,  )

 4e (
jπ / 2
,250 , 4e )(
− jπ / 2
,−250 

)
8
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Spectrum - A Graphical Plot of

Frequencies
10
7e jπ/3 7e -jπ/3

4e -jπ/2 4e jπ/2

-250 -100 0 100 250 f (in Hz)

DC
9
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Product of Signals
• In many practical systems, we intentionally
multiply two sinusoids to generate a signal
• Resulted audio signal is sometimes called beat note
• Examples:
• Musical Instruments
• Radio broadcasting – such as Amplitude
Modulation (AM)
Demo: Beatconb
10
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

How to Multiply Two Signals?

• It is easier to understand for multiplying a sinusoid
with a square wave

Switch
multiply
0
1
0 control

11
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Sinusoid A
transistor

multiply

control

Sinusoid B

• The amount of sinusoid A that can pass thru the

transistor is controlled by sinusoid B
12
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Spectrum of the Product of

Sinusoids
• To display the spectrum of the product of two
sinusoids, we need to convert it to a sum of
sinusoids first
• Because spectrum is defined in the form of sum of
sinusoids
• Inverse Euler’s formula
jθ − jθ jθ − jθ
e +e e −e
cos(θ ) = sin (θ ) =
13
2 2j
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

• For example
x(t ) = 2 cos(2π 20t ) cos(2π 200t )

x(t ) = 2 cos(2π 20t ) cos(2π 200t )

= cos(2π 220t ) + cos(2π 180t )
= 0.5 e( j 2π 220t
+e − j 2π 220t
)+ 0.5(e j 2π 180t
+e − j 2π 180t
)
0.5 0.5 0.5 0.5

14 -220 -180 0 180 220 f (in Hz)

 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

• Spectrum above shows that we need not really

multiple the two 200Hz and 20Hz sinusoids,
• Can be done by a summation of 180Hz and 220Hz
sinusoids
• While the spectrum of every signal can be found, it
means that every signal can be constructed by
summing the required sinusoids

What are the required sinusoids of a signal?

Solution:
Solution: Fourier
Fourier Analysis
Analysis

15
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Fourier Series
• Fourier first shows that any periodic waveform
can be synthesized by a sum of sinusoids that are
harmonically related

Periodic
Periodic signal:
signal:
Signals
Signalsthat
thatrepeat
repeat
itself
itselfwithin
withinaafixed
fixed
period
period

16
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Harmonically
Harmonically related
related sinusoids:
sinusoids:
Sinusoids
Sinusoidswith
withfrequencies
frequenciesthat
thatare
aremultiple
multipleof
ofaafundamental
fundamental
frequency
frequency

fo = 100Hz (fundamental f = 300Hz = 3*fo (3rd harmonic)

frequency or 1st harmonic)

f = 200Hz = 2*fo (2nd harmonic) f = 500Hz = 5*fo (5th harmonic)

17
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

• Any continuous-time periodic waveform can be

represented by the Fourier series as follows:
∞
x(t ) = A0 + ∑ Ak cos(2πkf 0t + φk )
k =1 An infinite series

x(t ) = A0 + A1 cos(2πf 0t + φ1 ) + K + A∞ cos(2π∞f 0t + φ∞ )

{ }
∞
x(t ) = X 0 + ∑ Re X k e j 2πkfot

Where k =1
jφ k
X 0 = A0 , X k = Ak e
18
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

• {Xo, X1, X2, …, X∞} are the Fourier series

coefficients
• Each coefficient controls the amplitude of the
corresponding harmonic required to construct a
signal
• Some signals need less harmonics, some signals
need more
• For example, to 100% synthesize a square wave,
basically need ∞ number of harmonics
The fact is even if you have ∞ number of harmonics,
you still cannot obtain a 100% square wave. It is due to
the Gibbs phenomenon
19
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Gibbs
phenomenon

20
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Fourier Analysis
• For a given periodic signal, how do we know its
Fourier series coefficients?
• By Fourier series analysis
2 T0 − j 2πkt / T0
Xk =
T0 ∫0 x(t )e dt

Where
1
T0 = = fundamental period
f0
21
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Example
• Let’s determine the Fourier Series Coefficients of a
square wave
 1 0 ≤ t < T0 / 2
x(t ) = 
− 1 T0 / 2 ≤ t < T0
1
0

-1 t
T0

22
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

2 T0 − j 2πkt / T0
Xk =
T0 ∫0 x(t )e dt

2 T0 / 2 − j 2πkt / T0 2 T0
= ∫0 (1)e dt + ∫T / 2 (−1)e − j 2πkt / T0 dt
T0 T0 0

− j 2πkt / T0 T0 / 2  − j 2πk (T0 / 2 ) / T0 − j 2πk ( 0) / T0 

2 e 2 e −e 
=
T0 − j 2πk / T0 T0  − j 2πk / T0 

0

e − jπk − 1
=
23 − jπk
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

− jπk − jπk − j 2πk

e −1 e −e
Xk = +
− jπk − jπk

=
2 − 2e − jπk
=
(
2 1 − (−1) k
)
jπk jπk
Since e-jπk = 1 if k
 4 is even and = -1 if
 jπk k = ±1,±3,±5,... k is odd
Xk = 

 0 k = 0,±2,±4,±6,...
24
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

 4 j2π 25t 4 j2π 3*25t 4 j2π 5*25t 4 j2π 7*25t 

x(t) = Re e + e + e + e 
 jπ jπ 3 jπ 5 jπ 7 
1 − jπ / 2
Since = e , hence
j
 4 − jπ / 2 j2π 25t 4 − jπ / 2 j2π 3*25t 
π e e + e e 
x(t) = Re
π 3

 4 − jπ / 2 j2π 5*25t 4 − jπ / 2 j2π 7*25t 
+ e e + e e
 π5 π7 
25
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

If we are only interested in the magnitude of each coefficient,

we have
4 − jπ / 2 4 4
e = (cos(π / 2)) + (− sin (π / 2)) =
2 2
πk πk πk
Magnitude Spectrum for square wave

1.4

1.2

1
Amplitude

0.8

0.6

0.4

0.2

0
-200 -150 -100 -50 0 50 100 150 200

26 Frequency f (Hz)
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Exercise 2
• Let’s determine the Fourier Series Coefficients of a
sawtooth wave
Hint:
T0
x(t ) =  − t 0 ≤ t < T0 te at
e at
2 ∫ dt = a − a 2
te at

T0/2

-T0/2
T0
27
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

Solution to Exercise 2
T0
(
X 0 = ∫ (T0 / 2 − t )dt = (T0 / 2)t − t / 2
0
2
)T0

0
=0
T0
X k = ∫ (T0 / 2 − t )e − j 2πkt / T0 dt
0
T0
− j 2πkt / T0 − j 2πkt / T0
2  te e 

=− +
T0  − j 2πk / T0 (2πk / T0 ) 2 
0
T0 T0 − jπ / 2
= = e
28 jπk πk
 Signal Processing Fundamentals – Part I
Spectrum Analysis and Filtering
3. Spectrum Representation

 T0 j2πt /T0 T0 j2π 2t /T0 T0 j2π 3t /T0 

x(t) = Re e + e + e +...
 jπ jπ 2 jπ3 
Assume f = 1/T0 = 25
Magnitude Spectrum for square wave

0.014

0.012

0.01
Amplitude

0.008

0.006

0.004

0.002

0
-150 -100 -50 0 50 100 150

29 Frequency f (Hz)