LESSION -4 Date 28/4/2021

Topic :-- DFT Algorithm

N-POINT DTFT ALGORITHM

❖ For evaluating the spectrum of simple signals the Fourier transform formula is quite
useful. For more complex signals, using the Fourier transform formula becomes
difficult.
❖ The Fourier transform of any discrete time signal results in a specrum which is a
continuous function of ω. This function is “bounded” within a range of 2 π as, - π<
ω < + π.
❖ Graph of any continuous function may be approximated using a finite number of
“sample points”. N-point discrete time Fourier transform (dtft) algorithm employs
this very principle.
❖ For this purpose, any number of points can be chosen. In the following example, we
chose n = number of sample values of the given signal.
❖ The entire range of frequencies -π<ω<+π is equally divided into “n” discrete
values “ωk” and the Fourier transform formula is calculated at these frequencies.
❖ The approximate spectrum of the signal is the plot of these discrete points of the
actual spectrum.
❖ A more accurate approximation can be had if we chose more member of discrete
frequency components. In other words, we could calculate the Fourier transform at
more points, but, as it is , the calculation is quite long as we find out in the following
exercise. An exercise with more points can be carried out by adding some 0s behind
the given points of the signal to artificially increase “n” , so that the frequency range
of 2π can be divided into more number of discrete frequency values. Such an
exercise would involve an increased number of calculation steps. Hence such
exercise is tackled by “fft algorithms”.
❖ In this exercise we explain the simpler “dft algorithm”.

EXAMPLE : ---
CONSIDER THE x (n) AS --

BY FOURIER TRANSFORM FORMULA

 𝐧=(𝐍−𝟏)
Ƒ [𝐱 (𝐧) ] = 𝐗 (𝛚) = ∑𝐧=𝟎 𝐱 (𝐧). 𝐞−𝐣 𝛚 𝐧 ………….(A)

Eq. (A) IS CALCULATED AT SOME “N” DISCRETE VALUES OF ω = ωK , SUCH AS

𝟐𝝅𝒌
𝛚𝒌 = For k = 0,1 2,3 ,4 ,5
𝑵
WHEN WE SUBSTITUTE ωk IN Eq (A) WE HAVE ω = ωk n AS
𝟐𝝅𝒌𝒏
𝛚𝒌𝒏 = For k = 0,1 2,3 ,4 ,5 & n = 0,1 2,3 ,4 ,5
𝑵

CALCULATION STEPS

GIVEN,

1. TABULATE THE SAMPLE VALUES OF THE SIGNAL x (n)….. WHERE N = 6

{ x (0) = -2 : x (1) = 1 : x (2) = 2 : x (3) =2 : x (4) = -3 : x (5) = 1 }

2. CALCULATE THE FOURIER TRANSFORM X (k) AT EACH VALUE OF k AS FOLLOWS

3. AT k = 0 ALL THE FREQUENCY TERMS DEFINED BY ω kn IS 0.

𝟐𝝅𝒌𝒏
𝛚𝒌𝒏 = (For k = 0) 𝛚𝒌𝒏 = 𝟎
𝑵
WHILE SUBSTITUTING IN Eq (A) ALL THE TERMS 𝐞−𝐣 𝝎𝒌 𝒏 = 𝟏
USING THE SAMPLE VALUES OF x (n) WE GET

X (0) = (-2) + (1) + (2) + (2) + (-3) +(1) = 1

1. AT k = 1 , THE FREQUENCIES OF ALL THE ‘n’ TERMS AND THE CORRESPONDING

𝝎𝒌 𝒏 & 𝐞−𝐣 𝝎𝒌 𝒏
, ARE CALCULATED AND MULTILPIED WITH THE
CORRESPONDING SIGNAL SAMPLE x (n). THIS DEFINES THE SPECTRAL COMPONENT
𝟐 𝝅 .𝟏
X(1) , WHICH OCCURS AT ω1 = = 𝝅⁄𝟑
𝟔
−𝐣 𝝎𝒌 𝒏
N = 0 𝝎𝒌 𝒏 = 𝟎 & 𝐞 = 1 : ∴ x(0) 𝒆−𝐣 𝝎𝒌 𝟎 = ( -2).1 = -2 + j 0
𝟐 𝝅 𝟏.𝟏 𝝅
N = 1 𝝎𝒌 𝒏 = = 𝝅⁄𝟑 & 𝐱(𝟏). 𝐞 −𝐣 ⁄𝟑 =(1).( 0.5 – j 0.866)
𝟔
= 0.5 – j 0.866
𝟐 𝝅 𝟏.𝟐 𝝅
N = 2 𝝎𝒌 𝒏 = = 𝟐𝝅⁄𝟑 &𝐱(𝟐). 𝐞 −𝐣 𝟐 ⁄𝟑 =(2).( -0.5 – j 0.866)
𝟔
 = -1 – j 1.73
𝟐 𝝅 𝟏.𝟑
N=3 𝝎𝒌 𝒏 = = 𝛑 & 𝐱(𝟑) 𝐞 −𝐣 𝛑 = (2). ( -1 + j0) = -2 + j 0
𝟔
𝟐 𝝅 𝟏.𝟒 𝝅
N=4 𝝎𝒌 𝒏 = = 𝟒𝝅⁄𝟑 & 𝐱(𝟒) 𝐞 −𝐣 𝟒 ⁄𝟑 =(-3).( -0.5 + j 0.866)
𝟔
= 1.5 - j 2.6
𝟐 𝝅 𝟏.𝟓 𝝅
N=5 𝝎𝒌 𝒏 = = 𝟓𝝅⁄𝟑 & 𝐱(𝟓) 𝐞 −𝐣 𝟓 ⁄𝟑 = (1).(0.5 + j 0.866)
𝟔
= 0.5 + j 0.866
ADDING UP THESE COMPLEX NUMBERS WE GET THE SPECTRAL COMOPNENT
X (1) =(-2 + j 0 )+( 0.5 – j 0.866 )+( -1 – j 1.73 )+( -2 + j 0 ) +
(1.5 - j 2.6 )+( 0.5 + j 0.866 ) = - 2.5 – j 4.33
IN POLAR FORM
X (1) = 5 ∠60 0

THIS SPECTRAL COMPONENT HAS AN “AMPLITUDE” X(1) = 5 V AND IS

ASSOCOATED WITH A PHASE ANGLE OF + 60 0 AND THIS OCCURS AT THE
𝟐 𝝅 .𝟏
FREQUENCY OF ω 1 = = 𝝅⁄𝟑
𝟔

𝟐 𝝅 .𝟐
2. AT k = 2 ,WE DEFINE THE SPECTRAL COMPONENT X(2) AT ω 2 = = 𝟐𝝅⁄𝟑
𝟔
N = 0 𝝎𝒌 𝒏 = 𝟎 & 𝐞 −𝐣 𝝎𝒌 𝒏 = 1 : ∴ x(0) 𝒆−𝐣 𝝎𝒌 𝟎 = ( -2).1 = -2 + j 0
𝟐 𝝅 𝟐.𝟏 𝝅
N = 1 𝝎𝒌 𝒏 = 𝟔 = 𝟐𝝅⁄𝟑 & 𝐱(𝟏). 𝐞 −𝐣𝟐 ⁄𝟑 =(1).(-0.5 – j 0.866)
= -0.5 – j 0.866
𝟐 𝝅 𝟐.𝟐 𝟒𝝅 −𝐣 𝟒𝝅⁄𝟑
N = 2 𝝎𝒌 𝒏 = 𝟔 = ⁄𝟑 & 𝐱(𝟐). 𝐞 =(2).(-0.5 + j 0.866 )
= -1 + j 1.73
𝟐 𝝅 𝟐.𝟑 −𝐣 𝟐𝛑
N = 3 𝝎𝒌 𝒏 = 𝟔 = 𝟐𝛑 & 𝐱(𝟑). 𝐞 =(2). 1 + j 0 = 2 + j 0
𝟐 𝝅 𝟐.𝟒 𝝅
N = 4 𝝎𝒌 𝒏 = 𝟔 = 𝟖𝝅⁄𝟑 & 𝐱(𝟒). 𝐞 −𝐣 𝟖 ⁄𝟑 =(-3).( -0.5 - j 0.866 )
= 1.5 + j 2.6
𝟐 𝝅 𝟐.𝟓 𝟏𝟎𝝅 −𝐣 𝟏𝟎𝝅⁄𝟑
N = 5 𝝎𝒌 𝒏 = 𝟔 = ⁄𝟑 & 𝐱(𝟓). 𝐞 =(1).(- 0.5 + j 0.866)
= - 0.5 + j 0.866

X (2) =(-2 + j 0 )+( -0.5 – j 0.866)+( -1 + j 1.73)+( 2 + j 0) +

(1.5 + j 2.6 )+( - 0.5 + j 0.866 ) = - 0.5 + j 4.33
IN POLAR FORM
X (2) = 4.6 ∠ -83.40
 THIS SPECTRAL COMPONENT HAS AN “AMPLITUDE” X(2) = 4.6 V AND IS
ASSOCOATED WITH A PHASE ANGLE OF - 83.4 0 AND THIS OCCURS AT
THE FREQUENCY OF ω 2 = 𝟐 𝝅⁄𝟑

𝟐 𝝅 .𝟑
3. AT k = 3 ,WE DEFINE THE SPECTRAL COMPONENT X(3) AT ω 3 = 𝟔 = 𝝅
N = 0 𝝎𝒌 𝒏 = 𝟎 & 𝐞 −𝐣 𝝎𝒌 𝒏 = 1 : ∴ x(0) 𝒆−𝐣 𝝎𝒌 𝟎 =( -2).1 +j 0 = -2 + j 0

𝟐 𝝅 𝟑.𝟏
n=1 𝝎𝒌 𝒏 = = 𝛑 & 𝐱 (𝟏). 𝐞 −𝐣𝛑 = (1). (- 1 –j 0) = - 1 – j 0
𝟔
𝟐 𝝅 𝟑.𝟐
n=2 𝝎𝒌 𝒏 = = 𝟐𝛑 & 𝐱(𝟐). 𝐞 −𝐣 𝟐𝛑 =(2).( 1 + j0) = 2 + j 0
𝟔
𝟐 𝝅 𝟑.𝟑
n=3 𝝎𝒌 𝒏 = = 𝟑𝛑 & 𝐱(𝟑). 𝐞 −𝐣 𝟑𝛑 =(2).(-1 –j 0 ) = -2 – j 0
𝟔
𝟐 𝝅 𝟑.𝟒
n=4 𝝎𝒌 𝒏 = = 𝟒𝛑 & 𝐱(𝟒). 𝐞 −𝐣 𝟒𝛑 = (-3).(1 + j 0) = -3 + j 0
𝟔
𝟐 𝝅 𝟑.𝟓
n = 5 𝝎𝒌 𝒏 = = 𝟓𝛑 &𝐱(𝟓). 𝐞 −𝐣 𝟓𝛑 =(1). (- 1 – j 0) = -1 - j 0
𝟔

X (3) =(-2 + j 0 )+( - 1 – j 0)+( 2 + j 0)+( -2 – j 0) +

(-3 + j 0)+( -1 - j 0) = - 7 – j 0
IN POLAR FORM
X (3) = 7 ∠00

THIS SPECTRAL COMPONENT HAS AN “AMPLITUDE” X(3) = 7 V AND IS ASSOCOATED

WITH A PHASE ANGLE OF 00 AND THIS OCCURS AT THE FREQUENCY OF ω 2 = 𝝅

𝟐 𝝅 .𝟒
4. AT k = 4 ,WE DEFINE THE SPECTRAL COMPONENT X(2) AT ω 4 = 𝟔 = 𝟒𝝅⁄𝟑
n = 0 𝝎𝒌 𝒏 = 𝟎 & 𝐞 −𝐣 𝝎𝒌 𝒏 = 1 : ∴ x(0) 𝒆−𝐣 𝝎𝒌 𝟎 =( -2).1 +j 0 = -2 + j 0
𝟐 𝝅 𝟒.𝟏 𝝅
n = 1 𝝎𝒌 𝒏 = 𝟔 = 𝟒𝝅⁄𝟑 & 𝐱(𝟏). 𝐞 −𝐣𝟒 ⁄𝟑 =(1).( -0.5 + j 0.866)
= - 0.5 + j 0.866
𝟐 𝝅 𝟒.𝟐 𝝅
n = 2 𝝎𝒌 𝒏 = 𝟔 = 𝟖𝝅⁄𝟑 &𝐱(𝟐). 𝐞 −𝐣 𝟖 ⁄𝟑
=(2).( -0.5 - j 0.866)
= -1 – j 1.73
𝟐 𝝅 𝟒.𝟑 −𝐣 𝟒𝛑
n = 3 𝝎𝒌 𝒏 = 𝟔 = 𝟒𝛑 & 𝐱(𝟑). 𝐞 =(2).(1 + j 0 ) = 2 + j 0
𝟐 𝝅 𝟒.𝟒 𝝅
n = 4 𝝎𝒌 𝒏 = 𝟔 = 𝟏𝟔𝝅⁄𝟑 &𝐱(𝟒). 𝐞 −𝐣 𝟏𝟔 ⁄𝟑 =(-3).( -0.5 + j0.866)
= 1.5 – j 2.6
 𝟐 𝝅 𝟒.𝟓 𝝅⁄
n = 5 𝝎𝒌 𝒏 = = 𝟐𝟎𝝅⁄𝟑 &𝐱(𝟓). 𝐞 −𝐣 𝟐𝟎 𝟑=(1).(- 0.5 - j 0.866)
𝟔
= - 0.5 - j 0.866
X (4) =(-2 + j 0 )+( - 0.5 + j 0.866)+( -1 – j 1.73)+( 2 + j 0) +
(1.5 – j 2.6)+( - 0.5 - j 0.866 ) = -0.5 – j 4.33
IN POLAR FORM
X (4) = 4.6 ∠+83.4 0
THIS SPECTRAL COMPONENT HAS AN “AMPLITUDE” X(3) = 4.6 V AND IS
ASSOCOATED WITH A PHASE ANGLE OF +83.4 0 AND THIS OCCURS AT THE
FREQUENCY OF ω 4 =𝟒𝝅⁄𝟑

𝟐 𝝅 .𝟓
5. AT k = 5 ,WE DEFINE THE SPECTRAL COMPONENT X(2) AT ω 5 = = 𝟓𝝅⁄𝟑
𝟔

n = 0 𝝎𝒌 𝒏 = 𝟎 & 𝐞 −𝐣 𝝎𝒌 𝒏 = 1 : ∴ x(0) 𝒆−𝐣 𝝎𝒌 𝟎 =( -2).1 +j 0 = -2 + j 0

𝟐 𝝅 𝟓.𝟏 𝝅⁄
n=1 𝝎𝒌 𝒏 = = 𝟓𝝅⁄𝟑 &𝐱(𝟏). 𝐞 −𝐣𝟓 =( 1).( 0.5 + j 0.866 )
𝟑
𝟔
= 0.5 + j 0.866
𝟐 𝝅 𝟓.𝟐 𝝅
n=2 𝝎𝒌 𝒏 = = 𝟏𝟎𝝅⁄𝟑 &𝐱(𝟐). 𝐞 −𝐣 𝟏𝟎 ⁄𝟑 =(2).(- 0.5 + j 0.866)
𝟔
= - 1 + j 1.73

𝟐 𝝅 𝟓.𝟑 𝟏𝟓𝝅⁄
n=3 𝝎𝒌 𝒏 = = 𝟏𝟓𝝅⁄𝟑 &𝐱(𝟑). 𝐞 −𝐣 𝟑 =(2).(- 1 +j 0 )
𝟔
= -2 + j 0
𝟐 𝝅 𝟓.𝟒 −𝐣 𝟐𝟎𝝅⁄𝟑
n = 4 𝝎𝒌 𝒏 = 𝟔 = 𝟐𝟎𝝅⁄𝟑 &𝐱(𝟒). 𝐞 =(-3).(-0.5 - j 0.866)
= 1.5 + j 2.6
𝟐 𝝅 𝟓.𝟓 𝟐𝟓𝝅 −𝐣 𝟐𝟓𝝅⁄𝟑
n = 5 𝝎𝒌 𝒏 = 𝟔 = ⁄𝟑 &𝐱(𝟓). 𝐞 =(1) .( 0.5 - j 0.866)
= 0.5 - j 0.866

X (5) =(-2 + j 0 )+( 0.5 + j 0.866 )+( -1 + j 1.73)+( -2 + j 0) +

(1.5 + j 2.6)+( 0.5 - j 0.866 ) = - 2.5 + j 4.33
IN POLAR FORM
X (5) = 5 ∠ - 60 0
THIS SPECTRAL COMPONENT HAS AN “AMPLITUDE” X(3) = 5 V AND IS
ASSOCOATED WITH A PHASE ANGLE OF -60 0 AND THIS OCCURS AT THE FREQUENCY
OF ω 5 =𝟓𝝅⁄𝟑
 THE CALCULATED SPECTRAL VALUES ARE TABULATED FOR CONVENIENCE, AGAINST
THE DISCRETE FREQUNCY VALUES AS FOLLOWS--

USING THESE SPECTRAL VALUES, THE APPROXIMATE SPECTRUM IS SKETCHED

FREQUENCY ωk 0 π/3 2π/3 π 4π/3 5π/3

(radian)
SPECTRAL 1 5 4.6 7 4.6 5
COMPONENTS (V)

WE HAD SEEN EARLIER, WHILE STUDYING FOURIER TRANSFORM OF DISCRETE TIME

NON-PERIODIC SIGNAL, THAT THE SPECCTRUM OF SUCH SIGNALS IS CONTINUOUS IN
FREQUENCY . THIS WAS VERIFIED IN EXAMPLE -4

THE SPECTRUM OBTAINED BY N- POINT DFT ALGORITHM AND SKETCHED ABOVE IS

AN APPROXIMATION OF THE ACTUAL CONTINUOUS SPECTRUM
THE APPROXIMATION BECOMES BETTER IF WE USE MORE SAMPLE POINTS, IN OTHE
RWORDS, WE USE MORE NUMBER OF POINTS OF “N”.
WHEN WE HAVE MORE POINTS, THE CALCULATION BECOMES MORE COMPLICATED.
THEREFORE SUCH CALCULATIONS CAN ONLY BE PERFORMED WITH COMPUTERS. FOR THIS
FFT ALGORITHM HAS TO BE USED.