FOURIER TRANSFORM

Subset of -transform As -plane is a complex number plane, = = + where is the real value, is the imaginary value. Thus & are as follows; = = 2 +
1 2 2

tan1

Figure 5.1: Radius and angle of a complex value

In Fourier Transform, = 1. Thus, Fourier Transform is actually -transform evaluated on the unit circle. If the formulation of -transform is

=
=

 Formulation for Fourier transform is similar to -transform but with =

=
=

Its inverse transformation is then 1 = 2

Note that the inverse FT is evaluated on = to = or with the range of 2. This is because value will be repeated after each 2.

3.1 Why do we need FT? Basically, signal has three information, Amplitude (), Frequency () and Phase ().

Figure 5.2:

In signal filtering system, filtering is based on the frequency of the signal, where the desired frequency component will be preserved while the unwanted frequency component will be removed from the signal.

 In this case, FT is used to sort the signal based on its frequency. When the signal has been sorted based on its frequency, the process of preserving and removing frequency component in the signal will be possible. In continuous signal, frequency is labeled with and the unit is Hz. In discrete signal, frequency is labeled with where = frequency of continuous signal Hz Sampling frequency Hz

 As is the ratio between and , it has no unit Besides Hertz (Hz), frequency is also normally presented in radian as given below, where the frequency is multiplied with 2 Continuous signal = 2 Discrete signal = 2 ( 1 ) ()

Based on the FT formulation, it can be seen that the FT is sorting the frequency component of the signal based on As the important component of the FT is , sometimes the formulation is written as () instead of ( ). If we want to present the FT in terms of , the formulation becomes

=
=

3.2 Magnitude and Phase Spectrum Similar to -transform, results of FT is also a complex value where the real and imaginary value is separated. Thus, to obtain the behavior of the signal, magnitude and phase spectrum are used for analysis

= + ()

Magnitude

Phase = tan1

In signal analysis, few other spectrums are also used. There are; Energy
2

= () (To obtain clearer plots as most of

Magnitude dB 20 log10 | | | | values are small)

Obviously, -axis of all of the spectrums , , 2 , is the frequency component of signal, and -axis of the spectrums is the value of the spectrum.

 Normally, the spectrums will be plotted from = to = . As the spectrum value will be repeated every 2, an alternative way of plotting the spectrum is from = 0 to = 2

Figure 5.3:

Example 1

Figure 5.4: Time-domain signal

Plot , , Solution:

= + 1 + + [ 1]

=
=

+ 1 + + 1

= + 1 + = 1 + + = 1 + 2 cos = 1 + 2cos() = 1 + 2 cos = 0

In trigonometry, the relationship between cosine and sine to exponential is as below + = 2 = 2 = +

= 1 + 2 cos = tan1

tan1

0 1 + 2 cos

0 0 < 0 = < 0
2

= 1 + 2 cos

0.5
Hreal () Himag()

-0.5

-1 -1

-0.8 -0.67

-0.4

-0.2

0 0.2 ( rad)

0.4

0.67 0.8

-1 -1

-0.8 -0.67

-0.4

-0.2

0.2

0.4

0.67 0.8

( rad)

3 2.5 2 1.5 1 0.5

4 3.142

H()

|H()|

-3.142 0 -1 -0.8 -0.67 -0.4 -0.2 0 0.2 ( rad) 0.4 0.67 0.8 1 -4 -1 -0.8 -0.67 -0.4 -0.2 0 0.2 0.4 0.67 0.8 1

( rad)

Figure 5.5: Frequency spectrums

Example 2 = cos (0.1) for 0 < < 40

1

0.5

x[n]

-0.5

-1

10

15

20 n

25

30

35

40

Figure 5.6: Time-domain input signal

From the equation above, it is known that = 0.05 where = 2 = 0.1. The frequency value can also be seen from Figure 5.6 above where = = 20 = 0.05. Now, lets plot the

signal in the frequency domain using Fourier Transform.

=
= 40

=
=0 40

cos (0.1) 1 0.1 + 0.1 2 ( 0.1) + (+0.1)

=0 40

=
=0

1 = 2 1 = 2

40

(0.1)
=0

1 + 2

40

+ ( +0.1)
=0

1 1 41 0.1 1 41 +0.1 = + 0.1 2 1 1 +0.1

25 20 15 10 5 0

|X()|

0.2

0.4

0.6

0.8

1 (rad)

1.2

1.4

1.6

1.8

Figure 5.7: Magnitude spectrum where 0 2

25

20

15
|X()|

10

0 -1

-0.8

-0.6

-0.4

-0.1

0 0.1 (rad)

0.4

0.6

0.8

Figure 5.8: Magnitude spectrum where

X()

-3 0 0.1 0.4 0.8 1 1.2 (rad) 1.6 1.9 2

Figure 5.9: Phase spectrum where 0 2

Example 2 = cos (0.1 0.5) for 0 < < 40

Figure 5.10 shows [] where it is a 5 samples delayed of signal in Figure 5.6 ( = 0.5)
1 0.5
x[n]

0 -0.5 -1

10

15

20 n

25

30

35

40

Figure 5.10: Time-domain input signal of Figure 5.6 with = 0.5

25 20
|X()|

15 10 5 0 -1 -0.8 -0.6 -0.4 -0.1 0 0.1 ( rad) 0.4 0.6 0.8 1

Figure 5.11: Magnitude spectrum

H() = 0.5
0.5

H() x

-0.5

H() = -0.5

-1 -1

-0.8

-0.4

-0.1

0 0.1 ( rad)

0.4

0.8

Figure 5.12: Phase spectrum. The arrows show the phase value at =0.1

Quiz 1 Find the frequency response, () of the signals below, where all the signals are stable.
1 1

1. h n = 1 0 0 3 2. h n = an u n 3. h n = n 3 + an u n 1 4. h n = sin 0 n 5. h n = 2u n u n 3 6. H z = 1 + z 1 + 3z 3

7. H z = 10.5z 1 + 1+2z 1 8. y n = x n + 3x n 1 + 0.5x n 2 9. y n = x n + y n 1 10. y n = x n 2 y n 2 + 2y n 1 11. h n = a n , 1 < < 1

3.3 Fourier Transform Properties

Linear Time-shifting Frequency-shifting Time-reversal Differentiation 1 + 2 0 () 1 + 2 () () ( 0 ) ()

Convolution Multiplication

1 2 1 2

1 2 () 1 2 ()

3.4 Symmetry Properties of FT for LTI System By substituting = cos () into FT formulation, the real and imaginary part of the transform can be separated.

=
=

cos sin

=
=

=
=

cos

=
=

sin ()

 Since cos = cos() and sin = sin(), thus, it follows that = , = ,

Or, in single equation, it is written as = The magnitude and phase spectra also follows that = ,

= , ()

Figure 5.13: Correct shapes of frequency response

Figure 5.14: False shapes of frequency response

3.5 Frequency Response based on Poles and Zeros Because frequency response for LTI system is symmetry, poles and zeros will exist as a conjugate pair on the -plane. Thus, when there is a pole/zero at = , there will also be a pole/zero at = . When the poles/zeros are real ( = 0 , ), there will be no pair

Figure 5.15: Pairs of poles and zeros as conjugation

3.5.1

Frequency Response for a Pair of Poles

System function when only a pair of poles exist is =

Imaginary Part

1 1

1 1 1

1 2 = + 1 1 1 1

Real Part

Figure 5.16: A pair of poles

 As = 1 2 = + 1 1 1 2 = + ( ) 1 1 (+ ) Based on the equation, () is maximum when = and = , where

and

+ becomes 1. The shape of the frequency response is shown in Figure 5.17 For both causal and stable system, 0 < < 1. Within this range, max ] will increase [ when increases.

10 9 8 r = 1.1 7 6
|H()|

11 10 9 8 7
|H()|

r = 0.9

5 4 r = 1.3 3 2 1 0 -1 -0.8 -0.6 -0.3 0 ( rad) 0.3 0.6 0.8 1 r = 1.5

6 5 4 3 2 1 -1 r = 0.5 r = 0.7

-0.8 -0.6

-0.3

0 ( rad)

0.3

0.6

0.8

Figure 5.17: Frequency response for a pair of poles with =0.3 and various values of r

Example 3 Plot for = 2 cos [] Solution:

() =
=0

0 + 0

=
=0

1 1

+
=0

+ 0

=
=0

0 +
=0

+ 0

 When = 0 or 0 , . Thus, use Hospitals rule. Then, we obtain = 0 + + 0 .

1

0.5

Imaginary Part

o o

-0.5

-1 -1 -0.5 0 Real Part 0.5 1

(a)

(b)

Figure 5.18: (a) Magnitude response, (b) Pole-zero ploy

Example 4 Plot for = 0.5 Solution: 1 = 1 0.5 1 1 = 1 0.5 It has 1 pole at = 0.5 and 1 zero at = 0 () is maximum when = 0 ()

1 = =2 1 0.5

 Thus, the | | is
2

1
1.8

1.6

0.5
Imaginary Part

1.4

|H()|

1.2

-0.5

0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 ( rad) 0.4 0.6 0.8 1

-1

-0.5

0 Real Part

0.5

(a) Figure 5.20: (a) Magnitude response, (b) Pole-zero plot

(b)

3.5.2

Frequency Response for a Pair of Zeros

System function when only a pair of zeros exist is = 1 1 1 1 = 1 = 1

1 1 ( +)
Imaginary Part

is minimum when = and when = The shape of the frequency response is shown in Figure 5.22

Real Part

Figure 5.21: A pair of zeros

3.5 r = 1.3 3 r = 1.1 2.5 r=1

3 2.5 2 r = 0.7 r = 0.9 r=1 1.5 1 0.5 0 -1 -0.8

|H()|

1.5 1 0.5 0 -1 -0.8

-0.4

0 ( rad)

0.4

0.8 1

|H()|

-0.4

0 ( rad)

0.4

0.8 1

Figure 5.22: Frequency response for a pair of zeros with =0.4 and various values of =0.4

Example 5 = 1 0.5 0.25 Solution: = + 0.5[ 1 + 0.25[ 2] = 1 + 0.5 1 + 0.25 2 = 1 1 1 1 = 0.25 + 0.433 = 0.5 60
0 0

= 0.25 0.433 = 0.5 60 = 1

60 0 0.5

60 0 0.5

1 0.8 0.6 0.4

Imaginary Part

0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1 -0.5 0 Real Part 0.5 1 2

Figure 5.23: Pole-zero plot

= 1 0.5 60 = 1 0.5
60 0

1 0.5 60 1 0.5
+60 0

() is minimum (or close to minimum) when = 600 / or = 600 / 600 = 60

= 1 0.5 1.2500 + 0.4330 = 0.6250 + 0.2165

= 0.6614
2 1.8

|H()|

1.2

0.6614

-pi

-pi/2 -pi/3

0 (rad)

pi/3

pi/2

pi

Figure 5.24: Magnitude response

Example 6 = 2
1

Solution:
0.5

= 2 = + 1 = 1 + 1 = 1 + It has 1 zero at = 1 and 1 pole at = 0

Imaginary Part

-0.5

-1 -1 -0.5 0 Real Part 0.5 1

Figure 5.25: Pole-zero plot

 () is minimum at =

= = 1 +
1.5

= 1 + cos
|H()|

= 1 + 1 = 0 is maximum at = 0 | = 0 = 1 + 0 =1+1 =2

0.5

0 -pi

-pi/2

0 (rad)

pi/2

pi

Figure 5.26: Magnitude response

3.5.3 Frequency Response for Multiple Pairs of Poles and Zeros At each pole, | | goes towards infinity, at each zero, | | goes towards zero Example 7
3 1 2.5 0.5
Imaginary Part

r = 0.5 2
|H()|

60o 0

60o 45o r = 0.8

1.5 1

-0.5

0.5 -1 -1 -0.5 0 Real Part 0.5 1 0 -180 -120 -60-45 0 (degree) 45 60 120 180

Figure 5.27: Left - Pole-zero plot, Right Magnitude spectrum

Example 8 Poles: 0.8 30 , 0.8 30 , 0.8 60 , 0.8 60 Zeros:

1

120 ,

120 ,

150 ,

150
90 80 70

0.5
Imaginary Part
|H()|

60 50 40 30

-0.5
20 10

-1 -1 -0.5 0 Real Part 0.5 1

0 -180 -180 -120

-60

-30

0 30 (degree)

60

120

150

180

(a)

(b) Figure 5.28: (a) Pole-zero plot, (b) Magnitude response

40

20

|H()|db

-20

-40

-60

-80 -180

-150

-120

-60

-30

0 (degree)

30

60

120

150

180

Figure 5.29: Magnitude db

3.6 Phase Response of LTI System (()) There are 3 types of phase response o Zero phase system o Linear phase system o Non-linear phase system Phase response is important in order to determine whether the LTI system is having a phase distortion or not. Signal can be understood as the combination of sinusoids of different frequencies. If one of these frequency components is time delayed by any amount other than zero, then it can be said that the signal has a phase distortion.

 Frequency component of = 0.5 is delayed by = From the 3 phase response types, zero phase and linear phase do not have phase distortion while the non-linear phase has the phase distortion.

3.6.1

Zero Phase System

[] must be symmetry (even signal) or anti-symmetry (odd signal) Symmetry: = Anti-symmetry: =

 The problem with zero phase system is that it is not causal where 0 for < 0. Thus, the system is not practical.

3.6.2

Linear Phase System

is delayed symmetry or anti-symmetry signal

 To make sure that [] is causal, the signal is delayed so that it starts at = 0. Thus, the causality problem in zero phase system is solved. This system is only possible for FIR signal. It is not possible to have a delayed symmetry or anti-symmetry signal for IIR.

3.6.3

Non-Linear Phase System

[] can be both FIR and IIR [] can also be both stable and causal Other than zero-phase and linear phase systems, the system is considered as non-linear phase system. The problem is, it suffers phase distortion.

Quiz Determine whether the signal has zero phase, linear phase or non-linear phase. 1. = 2. = + 3. = + 1 4. = + 2 5. = 2 1 + 2 + 1 6. = , < 1 7. = , < 1 8. = 1 + 9. = 10. = 2 1 + 2 1 Plot the pole-zero magnitude response and frequency response of these signals

3.7 Sampling A process of converting analog signal to discrete signal = , = In frequency domain, = 2 =
2 ,

at = 2, =

Thus, frequency component preserved after the sampling is the frequency less than 2

Example 9 = cos 21000 If = 2000Hz 21000 = cos 2000 = cos = cos ) ( = Thus, when = 2000Hz, the signal frequency of = 21000 is mapped at = Number of sample in [] is calculated as =
10

for 0 < 10

= 0.5 = 20

10

1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 0 1 2 3 4 5 t (ms) 6 7 8 9 10

1 0.8 0.6 0.4 0.2

x[n]

x(t)

0 -0.2 -0.4 -0.6 -0.8 -1 0 2 4 6 8 10 n 12 14 16 18 20

(a)

(b)

Figure 5.35: (a)Continuous signal, (b) Discrete signal sampled at = 1000

20 18 16 14 12

|H()|

10 8 6 4 2 0 - or -Fs/2

0 ( rad)

or
Fs/2

Figure 5.36: Magnitude response - frequency = 21000 1 is mapped at =

 If = 6000Hz

21000 1 = cos = cos 6000 3

10

o Number of sample in [] is calculated as =

1 0.8 0.6 0.4 0.2

10 0.1667

= 60

0.5

x[n]

x(t)

0 -0.2 -0.4 -0.6 -0.8 -1

-0.5

-1
0 1 2 3 4 5 t (ms) 6 7 8 9 10

10

20

30 n

40

50

60

(a)

(b)

Figure 5.37: (a)Continuous signal, (b) Discrete signal sampled at = 6000

 When = 5000Hz, the signal frequency = 21000 is mapped at =

30 25 20
|H()|

1 3

15 10 5 0 -pi

-pi/3

0 (rad)

pi/3 (1000Hz)

pi (3000Hz)
1 3

Figure 5.38: Magnitude response - frequency = 21000 1 is mapped at =

Example 10: Aliasing = cos 21000 + sin 24000 For = 3000Hz, Number of sample in [] is =
2

for 0 < 10 = 0.3333 = 30 as shown in Figure 5.40.

10

10

0.366

x(t)

-1 -1.366

-2

0.33 0.67

1.33 1.67

2.33 2.67

3.33 3.67

4.33 4.67

5 5.33 5.67 t (ms)

6.33 6.67

7.33 7.67

8.33 8.67

9.33 9.67

Figure 5.39: Continuous signal of ()

0.366 0
x[n]

-1.366 0 2 4 6 8 10 12 14 n 16 18 20 22 24 26 28

Figure 5.40: Discrete version of signal in Figure 5.39

In equation, the discrete signal is written as: 21000 24000 = cos + sin 3000 3000

2 8 [] = cos + sin 3 3 2 2 = cos + sin 2 3 3 2 2 = cos + sin 3 3

It is shown that when = 3000Hz, both frequency of the 1000Hz and the 4000Hz of the continuous signal are mapped to = 3 . In other words, although there are two frequencies exist in the continuous signal, only one frequency appears in its discrete form. Figure 5.41 shows the frequency response of the discrete signal.
2

25 20

|H()|

15 10 5 0 -pi

-2pi/3

0 (rad)

2pi/3

pi

pi

H()

pi/4 0 -pi/4

-pi -pi

-2pi/3

0 (rad)

2pi/3

pi

Figure 5.41: Magnitude response (upper figure), Phase response (lower figure)

 Thus, when [] is converted back to its continuous signal, the frequency component of 4000Hz is missing as shown in equation below and Figure 5.42. This phenomena is called aliasing. = cos 21000 + sin 21000
2 1

x(t)

0 -1 -2

5 t (ms)

10

Figure 5.42: Continuous signal reconstructed from discrete signal of []

 To obtain a good sampling output, 2 , where is the maximum frequency in the original continuous signal. This is called the Nyquist Theorem If we now choose = 10000Hz 2 8000Hz for Example 10, 21000 24000 = cos + sin 10000 10000 4 = cos + sin 5 5 Now, when = 10000Hz, frequency of 1000Hz is mapped to = 5 and frequency of 4000Hz is mapped to =
4 . 5

This shows that both of the frequencies in the continuous-

time signal are preserved where no aliasing occur.

 Figure 5.42 and figure 5.43 show the discrete time domain signal and its frequency spectrum respectively.
2 1

x[n]

0 -1 -2

10

20

30

40

50 n

60

70

80

90

100

Figure 5.42: Time-domain discrete signal of [] with = 10000

60

40
|H()|

20

0 -pi

-4pi/5

-pi/5

0 . (rad)

pi/5 (1000Hz)

4pi/5

pi

(4000Hz) (10000Hz)

pi pi/2

H()

0 -pi/2

-pi -pi

-4pi/5

-pi/5

0 (rad)

pi/5

4pi/5

pi

Figure 5.43: Frequency response of signal in Figure 5.42. (upper fig.) Magnitude, (lower fig) Phase.