0% found this document useful (0 votes)
55 views8 pages

The Digital Fourier Transform

1. The document discusses the discrete Fourier transform (DFT) and its relation to the discrete-time Fourier transform (DTFT) of periodic sequences. 2. Code examples are provided to calculate the DFT of various test signals using the FFT algorithm and plot the resulting spectra. 3. The examples analyze the spectra of a sinusoid, amplitude modulated signal, sums of sinusoids, and a Gaussian function. Peaks are located correctly and phases match expected values.

Uploaded by

Aravind Vinas
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
55 views8 pages

The Digital Fourier Transform

1. The document discusses the discrete Fourier transform (DFT) and its relation to the discrete-time Fourier transform (DTFT) of periodic sequences. 2. Code examples are provided to calculate the DFT of various test signals using the FFT algorithm and plot the resulting spectra. 3. The examples analyze the spectra of a sinusoid, amplitude modulated signal, sums of sinusoids, and a Gaussian function. Peaks are located correctly and phases match expected values.

Uploaded by

Aravind Vinas
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

The Digital Fourier Transform

The Digital Fourier Transform

EE16B131
Dept. of Electrical Engineering

April 4, 2017

1 Introduction:
F(θ ) is continuous and periodic. f [n] is discrete and aperiodic. Suppose now f [n]
is itself periodic with a period N, i.e.,
f [n+N] = f [n] ∀n
Then, it should have samples for its DTFT. This is true, and leads to the Discrete
Fourier Transform or the DFT:
Suppose f [n] is a periodic sequence of samples, with a period N. Then
the DTFT of the sequence is also a periodic sequence F[k] with the same
period N. (You will prove this in the DSP course). So we have
P
F[k] = f [n]exp(−2πnkj/N )
1P −nk
f[n] =
N
F [k]W
In this assignment, we want to explore how to obtain the DFT.

2 Codes:
for problems 1 to 3,

from pylab import ∗


x=r a n d ( 1 0 0 )
X= f f t ( x )
y= i f f t (X)
c_ [ x , y ]
print a b s ( x−y ) . max ( )

N = 512
x=l i n s p a c e ( − 4 ∗ p i , 4 ∗ p i , N+1)
x= x [ : −1]

def fun ( x , y ,N) :


Y= f f t ( y )
Y = f f t s h i f t (Y ) ∗ ( 1 . 0 /N)
w = f f t f r e q ( l e n ( x ) , d=4) ∗ l e n ( x )
w = f f t s h i f t (w)
figure ()
subplot (2 ,1 ,1)
p l o t (w , a b s (Y) , l w =2)
t i t l e ( " Spectrum of y ")
y l a b e l ( " |Y| " )
xlim ( −40 ,40)
g r i d ( True )
subplot (2 ,1 ,2)
p l o t (w , a n g l e (Y) , " r o " , lw =2)
y l a b e l ( " Phase o f Y" )
xlim ( −40 ,40)
g r i d ( True )
show ( )

y1=s i n ( 5 ∗ x )
y2 = cos (10∗ x )+0.1∗ cos ( x )∗ cos (10∗ x )
y3 = ( 3 ∗ s i n ( x)− s i n ( 3 ∗ x ) ) ∗ ( 0 . 2 5 )
y4 = ( 3 ∗ c o s ( x)+ c o s ( 3 ∗ x ) ) ∗ ( 0 . 2 5 )
y5 = c o s ( 2 0 ∗ x +5∗ c o s ( x ) )
f u n ( x , y1 , N)
f u n ( x , y2 , N)
f u n ( x , y3 , N)
f u n ( x , y4 , N)
f u n ( x , y5 , N)

...
for problem 4:

from pylab import ∗

N = 1024
x=l i n s p a c e ( − 10 ∗ p i , 1 0 ∗ p i , N+1)
x= x [ : −1]
y = e x p ( −( x ∗ ∗ ( 2 ) ) / 2 . 0 )
Y= f f t ( y )
Y = f f t s h i f t (Y ) ∗ ( 1 . 0 /N)
w = f f t f r e q ( l e n ( x ) , d =10) ∗ l e n ( x )
w = f f t s h i f t (w)
figure ()
subplot (2 ,1 ,1)
p l o t (w , a b s (Y) ∗ 2 0 ∗ p i , lw =2)
t i t l e ( " Spectrum of y ")
y l a b e l ( " |Y| " )
xlim ( −10 ,10)
g r i d ( True )
subplot (2 ,1 ,2)
p l o t (w , a n g l e (Y) , " r o " , lw =2)
y l a b e l ( " Phase o f Y" )
xlim ( −10 ,10)
g r i d ( True )
show ( )

d = Y∗ 2 0 ∗ p i
s = ( ( 2 ∗ p i ) ∗ ∗ ( 0 . 5 ) ) ∗ e x p ( − (w ∗ ∗ ( 2 ) ) / 2 . 0 )
e r r = a b s ( a b s ( d)− s ) . max ( )
print " errror " , err

2.1 Problem 1:
e5jx −e−5jx
y = sin(5x) =
2j

2
So the expected spectrum is
1
Y(w) =
2j
[δ(ω − 5) − δ(ω + 5)]
plots are,

..
The peaks are at exactly -5 and 5.And the phase at -5 is Π/2 ,and at 5 is -Π/2.

2.2 problem 1.2

Suppose we now look at AM modulation. The function we want to analyse is


f(t) = (1+0.1 cos(t)) cos(10t)
And we get ...
0.1 cos(10t) cos(t) = 0.05(cos 11t +cos 9t)
and plots are,

3
..
We have three clear spikes. The phases at those spikes are zero to machine
precision. The location of the spikes are 9, 10 and 11 radians per sec.

2.3 problem 2:

3
sin x = (3*sin(x)-sin(3*x))*(0.25)
and
3
cos x = (3*cos(x)+cos(3*x))*(0.25)
plots are,
5
..
..
for (3*sin(x)-sin(3*x))*(0.25) the peaks are expected to be at -3,-1,1,3 as in we got in the plot.And
phase at -1 is π/2, at 1 is -π/2.And at -3 is -π/2 ,at 3 is π/2.
next for ,
3
(3*cos(x)+cos(3*x))*(0.25) the peaks are at the same values of sin x.But the phase becomes zero(0)
at all peaks.

2.4 problem 3:

spectrum of cos(20t +5 cos(t)),


..we have for cos(20t +5 cos(t)),
The sinsoidal peaks are located in the range of 10 to 30 and also at -10 to -30.

2.5 problem 4:

−ax2
The Fourier transform of a Gaussian function f(x)=e is given by,

..so for a = 2 we get


√ 2
2π e−w /2
plots are,
..
for range x=linspace(-10π ,10π ,1025)
√ 1
at w=0 the value is 2π . Fourier coecients have
T
factor so,we need to multiply with the interval
(4x = (20π )).
Since Gaussian is not bandlimited in frequency which is obtained by multipling the 4x = (20π )
,resulting in |Y| = abs(Y)(20π ) .And
−15
the calculated erroor is 2.77555756156e-15 (which is so small) order of 10 .

You might also like