EENG-3151 DSP

CHAPTER-FOUR
DISCRETE TIME FOURIER TRANSFORM (DTFT) &
DISCRETE FOURIER TRANSFORM (DFT)

Discrete Time Fourier Transform (DTFT)

INTRODUCTION
The Fourier representation of signals plays an important role in both
continuous time and discrete-time signal processing.

It provides a method for mapping signals into another domain in which to

manipulate them.

This transform is useful to map the time-domain sequence into a continuous

function of a frequency variable ω.

Definition

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 1
EENG-3151 DSP

TWO IMPORTANT PROPERTIES

(or)

X[n] X(ejω)

Real & Even Real & Even

Real & Odd Imaginary & Odd
Imaginary & Even Imaginary & Even
Imaginary & Odd Real & Odd

DTFT OF SOME COMMON SEQUENCES

SEQUENCE DTFT

[] 1

[ −  ]  

1 2
()
 2
( −  )
 [], || < 1 1
1 −  

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 2
EENG-3151 DSP

− [− − 1], || > 1 1

1 −  
cos ( ) 
( −  ) + 
( +  )

Summary of DTFT Theorems & Properties

!"!
Let  () #$$$$$$$$$$$$% &' (
!"!
)() #$$$$$$$$$$$$$% *( )
PROPERTY SEQUENCE DTFT
Linearity  () + +)() aX'e/0 ( + bY(e/0 )
Time Shift ( −  ) e/034 X(e/0 )
Time –Reversal (−) X'e/0 (
Modulation  5 () X'e/(004 (
Convolution () ∗ )() X'e/0 (. Y(e/0 )
Conjugation  ∗ () X ∗ 'e/0 (
d
() j X'e/0 (
dω
Derivative
>
1
 ())() < X'e/= ( ∗ Y(e(/0=) )dθ
2π
Multiplication
>
Example:- If &' ( = 
( − A ) + 
( + A ) Compute the inverse
DTFT.
B B
Sol:-  () =  +  = cos ( )
C C

Relation Between DTFT & DFT

DFT is same as DTFT, but it is defined only for finite duration sequences.The DFT is itself a
sequence rather than a function of continuous variable, and it corresponds to samples, equally spaced
in frequency of the DTFT of the signal.

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 3
EENG-3151 DSP

Discrete Fourier Transform (DFT)

INTRODUCTION
The Discrete Fourier Transform (DFT) is a powerful computation tool which allows us to evaluate
the Fourier Transform &( ) on a digital computer or especially designed hardware.

DFT plays an important role in the implementation of many signal processing algorithms.

Definition
If  () = 0 EAF  < 0 G  ≥ I, then N point DFT is given by

--------------------- (1)
Then Inverse DFT is given by

--------------------- (2)

Let us define a term

which is known as a twiddle factor, and substituting in equations (1) & (2)

we get

----------------------------------- (3)

------------------------------ (4)

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 4
EENG-3151 DSP

DFT using matrix method

If we
assume that the inverse of WN exists, then the above equation can be inverted by
premultiplying both sides by JKB . Thus we obtain

------------------- ---- **

In fact, the IDFT as given by (**), can be expressed in matrix form as

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 5
EENG-3151 DSP

EXAMPLE: Find the DFT of a sequence  @ M1,1,0,0N and the

he IDFT of a
sequence *(L) = M1,0,1,0NN

Sol:-

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 6
EENG-3151 DSP

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 7
EENG-3151 DSP

Relationship of the DFT to other Transforms

1. The DFT and the Fourier Transform of x(n)
The Fourier Transform &  of a finite duration sequence  having length N is
given by
KB

&' ( @ O  
P
----------------------
------------ (5)

Where &  is a continuous function of ω.

The discrete Fourier Transform of  is given by

----------------
-------------- (6)

Comparing equations (5) & (6) we find that the DFT of  is a sampled version of
the Fourier Transform of the sequence and is given by

&L @ Q&'
' (R CST L @ 0,1,2,     ,
 I  1.
P
K

2. The DFT and the z--transform

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 8
EENG-3151 DSP

We obtain the desired expression for U  in the form

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 9
EENG-3151 DSP

------ (7)
The expression in the above equation has the form of a convolution sum. However it is not the
ordinary linear convolution, because the convolution sum involves the index 'V  (K and is
called circular convolution. Thus we conclude that multiplications of the DFTs of two sequences are
equivalent to the circular convolution of two sequences in the time-domain.

Example 1

Solution: Each sequence consists of four nonzero points. For the purposes of illustrating the
operations involving in circular convolution, it is desirable to graph sequence as points on a circle.
We note that the sequences are graphed in a counter clockwise direction on a circle. This establishes
the reference direction in rotating one of the sequences relative to the other.

Now, U V is obtained by circularly convolving B  G C  as specified by equation (7).

Beginning with m=0 we have

U

U 0) = O B ()C ((−))K

P

C ((−))W is simply the sequence C () folded and graphed on a circle as illustrated in fig. 3.1b. In
other words, the folded sequence is simply C () graphed in a clockwise direction.

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 10
EENG-3151 DSP

The product sequence is obtained by multiplying B  with C W , point by point. The
sequence is also illustrated in fig. 3.1b. Finally, we sum the values in the product sequence to obtain

U 0) = 14

For m=1 we have

U

U (1) = O B ()C ((1 − ))W

P

It is easily verified that C ((1 − ))W is simply the sequence C ((−))W rotated counter clockwise by
one unit in time as illustrated in fig 3.1c. This rotated sequence multiplies B () to yield the product
sequence, also illustrated in fig. 3.1c. Finally, we sum the values in the product sequence to obtain
U (1). Thus

U (1) = 16

For m=2 we have

U

U (2) = O B ()C ((2 − ))W

P

Now C ((2 − ))W is the folded sequence in fig.3.1b rotated two units of time in the counter
clockwise direction. The resultant sequence is illustrated in fig 3.1d along with the product sequence
B ()C ((2 − ))W. By summing the four terms in the product sequence, we obtain

U (2) = 14

For m= 3 we have
U

U (3) = O B ()C ((3 − ))W

P

The folded sequence C ((−))W is now rotated by three units in time to yield C ((3 − ))W and the
resultant sequence is multiplied by B () to yield the product sequence as illustrated in fig. 3.1e. The
sum of the values in the product sequence is

U (3) = 16

We observe that if the computation above is continued beyond m=3, we simply repeat the sequence
of four values obtained above. Therefore, the circular convolution of the two sequences B () and
C () yields the sequence

U () = M14, 16, 14, 16}

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 11
EENG-3151 DSP

Figure 3.1

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 12
EENG-3151 DSP

Circular Convolution Using Matrix Method

If ) @ ⨂\,, Using matrix approach we can write h(n) as N*N matrix form
f and x(n) as
column matrix.

Note that each row of the matrix in the above equation is a circular right-shift
right of the previous
row.

EFFICIENT COMPUTATION OF THE DFT : FAST

FOURIER TRANSFORM
INTRODUCTION
Fast Fourier transform (FFT) is a numerical algorithm
algorithm that has been developed to improve the
computational efficiency by an enormous amount and is the most popular method used in spectral
analysis in digital signal processing, specifically, to find the DFT of the signal and also the inverse
DFT of the frequency response to get the discrete-time
discrete time signal. This is only a computational algorithm
and not another transform.
In this FFT algorithm, when the value for the radix N is chosen as 2R, where R is a positive integer
(index of stage),, the number of complex multiplications is of the order (N/2)
( 2) log2(N) and the number
of complex additions is of the order N log2 N.
The basic strategy that is used in the FFT algorithm is one of "divide and conquer." which involves
invo
decomposing an N-point
point DFT into successively smaller DFTs.
Recall the definition of DFT:

DFT:

IDFT:

Where,

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 13
EENG-3151 DSP

RADIX-2 FFT ALGORITHMS

1. Decimation-in-Time FFT
The decimation-in-time FFT algorithm is based on splitting (decimating) x(n) into smaller sequences
and finding X(k) from the DFTs of these decimated sequences. This section describes how this
decimation leads to an efficient algorithm when the sequence length is a power of 2.
Let x(n) be a sequence of length N = 2R, and suppose that x(n) is split (decimated) into two sub
sequences, each of length N/2.

Separating x(n) into its even-and odd-numbered points.

1
K
Substitution of n = 2r for n even and n = 2r+1 for n odd, where r=0, 1,- - -,
C

Consequently, the above equation can be rewritten as

` `
B B
Where, ](L) = ∑ a
P _[] J` G b(c) = ∑
T a
P ℎ[] J`T
a a
K
_[] = [2] G ℎ[] = [2 + 1],  = 0, 1, − − − − 1
C
&(L) = ](L) + JKT b(L)
`
K Tf
& dL + e = ](L) + JK a
b(L)
C
K CS K CS CS
Tf
JK C =  K
(Tf )
C =  K .   CS
T
= −  K
T
= −JKT

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 14
EENG-3151 DSP

K
& dL + e = ]L  JKT bL
C
Properties of JK :
`
Tf
Symmetry property: JK @ JKT a

Periodicity >> : JKTfK @ JKT

Recursion >> : JKC @ JK/C

Computation of N = 8 point DFT

Below figure depicts the Computation of N = 8 point DFT. We observe that computation is
performed in three stages, beginning with the computations of four two-point DFTs, then two four-
point DFTs and finally one eight-point DFT.
The combination of the smaller DFTs to form the larger DFT is illustrated in figure for N=8.

The below figure shows the basic butterfly computation in the decimation-in-time FFT algorithm.

The basic computation is performed at every stage as illustrated in the below figure.

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 15
EENG-3151 DSP

BUTTERFLY DIAGRAM

2. Decimation-in-Frequency FFT: It is based on decimating (splitting) the output

sequence X(k) into smaller and smaller subsequence’s.

Computation of N = 8 point DFT

Below figure depicts the Computation of N = 8 point DFT. We observe that computation is
performed in three stages, beginning with the computation of one eight-point DFT, then two four-
point DFTs and finally four two-point DFTs.

BUTTERFLY DIAGRAM

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 16
EENG-3151 DSP

NOTE:
In case of DIT-FFT algorithm, output sequence is in order and input sequence is in bit-reversal order.
In case of DIF-FFT algorithm, input sequence is in order and output sequence is in bit-reversal order.
Generally the twiddle factor hi JKT for both DIT and DIF algorithms, where k value is given by
Ij
JKT , c= ; j = 0,1,2, − − − − −, (2mB − 1).
2k
R is the index of stage (R=1 for stage 1, R=2 for stage 2,R=3 for stage 3 and so on).

Standard values of twiddle factors are given below

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 17
EENG-3151 DSP

npo @ q; nqo @ r; nps @ q; nqs @ p. tpt − rp. tpt; nus = −r; nvs = −p. tpt − rp. tpt

EXAMPLE
Compute the DFT of a sequence w(x) = Mq, −q, q, −q} using DIT algorithm.
Sol:- Let w() = M1, −1,1, −1} ; Length N=4.
Twiddle factors are npo = q; nqo = −r;
Butterfly diagram

The output sequence &(L) = M0,0,4,0}

EXAMPLE
Compute the DFT of a sequence () = M0,1,2,3} using DIT and DIF algorithms.
Sol:- Let w() = M0,1,2,3} ; Length N=4.
Twiddle factors are npo = q; nqo = −r;

DIT Butterfly Diagram

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 18
EENG-3151 DSP

DIF Butterfly Diagram

EXAMPLE
Compute the DFT of a sequence  @ M1,1,1,1,1,1,1,1} using DIT and DIF
algorithms.
Sol:- Let w() = M1,1,1,1,1,1,1,1} ; Length N=8
Twiddle factors are
Jy = 1; JyB = 0.707 − {0.707; JyC = −{; JyU = −0.707 − {0.707

DIT Butterfly Diagram

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 19
EENG-3151 DSP

DIF Butterfly Diagram

IDFT Using FFT Algorithms

FFT algorithms can be used to compute an inverse DFT without any change in the algorithm.
The inverse DFT of an N-point sequence

where .
Taking complex conjugate and multiply by N, we obtain,
KB
∗ 
I @ O & ∗ LJKT
TP

The right hand side of the equation is DFT of the sequence & ∗ L and may be computed using any
FFT algorithm.
The desired output sequence x(n) can then be found by complex conjugating the DFT and dividing by
N to give

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 20
EENG-3151 DSP

KB ∗
1
 @ |O & ∗ (L)JKT }
I
TP

Example
Find the IDFT of the sequence &(L ) = M10, −2 + 2{, −2, −2 − 2{} using DIT & DIF
algorithms.

Sol: - Let &(L) = M10, −2 + 2{, −2, −2 − 2{} ; N=4; JW = 1; JWB = −{;

I ∗ () = M4,8,12,16} G () = M1,2,3,4}

Try DIF algorithm, you get the same answer.

ELECTRICAL & COMPUTER ENG. DEPT., SCHOOL OF ENGINEERING, ASTU, ADAMA Page 21