Concept of Fourier Analysis:: K N W JWN
Concept of Fourier Analysis:: K N W JWN
Concept of Fourier Analysis:: K N W JWN
Introduction
Fourier methods are commonly used for signal analysis and system design in modern telecommunications,
radar, and image processing systems. Classical Fourier methods such as the Fourier series
and the Fourier integral are used for continuous time (CT) signals and systems, i.e., systems in which
a characteristic signal, s.t/, is defined at all values of t on the continuum −1 < t < 1 . Amore
recently developed set of Fouriermethods, including the discrete time Fourier transform(DTFT) and
the discrete Fourier transform (DFT), are extensions of basic Fourier concepts that apply to discrete
time (DT) signals.
We have already seen that any arbitrary sequence is weighted linear combination of unit
sample sequences ( wδ (n − k ) ). We consider in this chapter an alternate description of a
st
Let x(t) =e is input to a CT LTI system with impulse response h(t). We can determine
the output by convolution integral.
∞
y (t ) = ∫ h(t ) x(t − τ )dτ
−∞
∞
∫ h (t ) e
s ( t −τ )
y (t ) = dτ
−∞
∞
∫ h(t )e
− sτ
y (t ) = e st
dτ
−∞
y (t ) = e .H ( s)
st
∫ h(t )e
− sτ
where..H ( s) = eigen..value = dτ
−∞
e st = eigen.... function
{x[n]}={ e jw0n
}
o/p=convolution of i/p and unit sample response of the system
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 3of 31
∞
y[n] = ∑ h[k ]x[n − k ]
k = −∞
∞
y[n] = ∑ h[k ]e
k = −∞
jw0 ( n − k )
∞
y[n] = e jw0 n ∑ h[k ]e
k = −∞
− jw0 k
y[n] = e jw0 n
H (e jw0 )
∞
where...H (e jw0 ) = ∑ h[k ]e
k = −∞
− jw0 k
in..general..
∞
H (e jw ) = ∑ h[n]e
n = −∞
− jw0 n
.....DTFT
The o/p sequence is also a complex exponential at the same frequency, but has been
jw
multiplied by a complex valued time coefficient (H(e )). This coefficient has
introduced gain and phase factors.
We will observe that the DTFT of a sequence x[n] is a continuous function of w. It is also
periodic in w with period 2 π .
Proof of periodicity
∞
X (e j ( w1+ 2πk )
)= ∑ x[n]e
n = −∞
− j ( w1+ 2πk ) n
∞
= ∑ x[n]e
n = −∞
− jw1n
e − j 2πkn
− j 2πkn
e = 1...if ..k , n..are.. int eger
= X (e jw1 )
Thus if DT signal is unique over the frequency interval (0,2 π ) or (- π , π ).then
X(w)is periodic with period 2 π
IDTFT:-
π
1
x[ n] = ∫
jwn
X ( w) e dw
2π −π
Discrete Fourier Transformation:-
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 4of 31
DFT:-
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 5of 31
X [k ] = X ( w) 2πk
w=
N
where...0 ≤ k ≤ N − 1
N −1 j 2πnk
−
X [ K ] = ∑ x[n]e N
.....DFT
n =0
N −1 j 2πnk
1
x[n] =
N
∑ X [ K ]e
k =0
N
.....IDFT
Now when a CT signal is sampled with sampling time T, the spectrum of the
resulting DT sequence becomes a periodic function of frequency 2 π /T.
Similarly when X(w) is sampled with sampling points 2 π /N, the corresponding DT
sequence becomes periodic in time with period N.
Thus the periodic sequence Xp[n], is formed by adding an infinite number of shifted
replicas.
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 6of 31
Example:-
Find the DFT of a 3 point signal x(n) as shown in fig
solution:-
2 j 2πNk
−
X (K ) = ∑ x ( n )e
N =0
N
j 2π j 4π
− k − k
X ( K ) = 3 + 2e + 3e 3 3
X ( 0) = 3 + 2 + 3 = 8
j 2π j 4π
− −
X (1) = 3 + 2e 3
+ 3e 3
X (1) = 1 / 2 + j 3 / 2 = e 3
jπ
−
X ( 2) = e 3
K 0 1 2
|X(K)| 8 1 1
∠ X(K) 0 PI/3 -PI/3
observe that because x(n) is a3 point sequence, its DFT is also a 3 point sequence.
Now we want to know about the DTFT from the DFT. The DTFT of X(K) is given by
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 7of 31
2
X (e jw ) = ∑ x(n)e − jwn
k =0
− jw
= 3 + 2e + 3e − j 2 w
= e − jw (2 + 3e jw + 3e − jw )
= e − jw (2 + 6 cos w)
F ( w) =| 2 + 6 cos w |
⎧ − w for (2 + 6 cos w) > 0
∠F ( w) = ⎨
⎩π − w for (2 + 6 cos w) < 0
Observe that DFT values are equal to DTFT values at sampling frequencies. But DFT in
this example has too few points to give a reasonable picture of DTFT. The peaks of the
DTFT appearing between the 2nd and 3rd sample is missing by the DFT. The goal can be
accomplished by zero padding.
Use of zero padding:-Seeing the DTFT through DFT is like a viewing F(w) through a
picket fence. Only the spectral components at the sampled frequencies will be visible. If
DFT has too few points, major peaks and valleys of F(w) existing DFT points will not be
visible. Actually it is possible to compute DTFT from DFT by using interpolation
formulae. But in that case we seek to reduce the distance at which we sampled. ( i.e.
equal to 2.PI/N).So N must be increased. So length of the sample must be increased. It
can be done by adding sufficient number of zero points to input sample sequence.
BY appending sufficient number of zeros to input sequence, the period N will increase
and thereby number of points in the DFT increases. Thus zero valued dummy elements
contribute nothing to the sum of DFT but act to decrease the frequency spacing.
jπ j 2π
− k − k
X ( K ) = 3 + 2e 3
+ 3e 3
X ( 0) = 3 + 2 + 3 = 8
− jπ
X (1) = 5e 3
jπ
X (2) = e 3
X (3) = 4
− jπ
X (4) = e 3
jπ
X (5) = 5e 3
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 8of 31
we can now view DFT and IDFT as linear combination on sequences {x[n]} and {X[k]}.
Let us define a N point vector x of the signal sequence.
therefore
X = DN x
x = XD N−1 ....(1)
agian.. from...IDFT ...Eqn
1
x = D N* X
N
1
D N−1 = D N*
N
D N D N = N .I N
*
1 1 1 1
1 2
1 W 4 W 4 W43
2 4
1 W 4 W 4 W46
3 6
1 W 4 W 4 W49
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 10of 31
− j 2π
W41 = e 4
= cos π / 2 − j sin π / 2 = − j
− j 2π .2
W42 = e 4
= cos π − j sin π = −1
W42 = W40+ 4 / 2 = W NK + N / 2 = −W NK = −W40 = −1
W44 = W42+ 4 / 2 = −W42 = 1
1 1 1 1
1 -j -1 j
1 -1 1 -1
1 j -1 -j
⎡0⎤ ⎡1 1 1 1 ⎤
⎢1 ⎥ ⎢1 − j − 1 j ⎥
X 4 = ⎢ ⎥⎢ ⎥
⎢2⎥ ⎢1 − 1 1 − 1 ⎥
⎢ ⎥⎢ ⎥
⎣3⎦ ⎣1 j − 1 − j ⎦
⎡ 6 ⎤
⎢− 2 + 2 j ⎥
X4 = ⎢ ⎥
⎢ −2 ⎥
⎢ ⎥
⎣− 2 − 2 j ⎦
example:-
compute the IDFT of the 4 point sequencex(n)={6,-2+2j,-2,-2-2j}.
Solution:-
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 11of 31
1 *
xN = WN X N
N
1
W N−1 = W N*
N
x N = W N−1 X N
⎡1 1 1 1 ⎤
⎢1 j − 1 − j ⎥
1
W4−1 = ⎢ ⎥
4 ⎢1 − 1 1 − 1 ⎥
⎢ ⎥
⎣1 − j − 1 j ⎦
⎡1 1 1 1 ⎤⎡ 6 ⎤
⎢1 j − 1 − j ⎥ ⎢− 2 + 2 j ⎥
1
x4 = ⎢ ⎥⎢ ⎥
4 ⎢1 − 1 1 − 1 ⎥ ⎢ − 2 ⎥
⎢ ⎥⎢ ⎥
⎣1 − j − 1 j ⎦ ⎣ − 2 − 2 j ⎦
⎡0 ⎤
⎢1 ⎥
x4 = ⎢ ⎥
⎢ 2⎥
⎢ ⎥
⎣ 3⎦
Circular shift of a sequence:-
∞
The periodic extension of sequence x[n] can be written as x p (n) = ∑ x(n − lN )
l = −∞
One way of visualizing the periodic sequence x p (n) is wrapping the finite duration
sequence x(n) around a circle in counterclockwise direction which is selected as +ve
direction.
As we repeatedly traverse the circumference of the circle , we see the finite length
sequence ,periodically repeated on a circular(modulo N) time axis. Thus we can write
x p (n) =x[(n modulo N)].
If the argument n is between 0 to N-1, leave it as it is , otherwise add/subtract multiples
of N, until the result is between 0 to N-1.
Example:-
x(-3 modulo 4)=x(1)
x(10 modulo 8)=x(2)
x(-11 modulo 4)=x(1)
x 'p = x p (n − k )
∞
= ∑ x(n − lN − k )
l = −∞
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 12of 31
In In
In last figure we find x(n) and its periodic extension x p (n) . The sequence x 'p (n) is
obtained by shifting x p (n) by 2 units right.
From the figure we find that the sequence x’(n) is not a linear shift of x(n) and x’(n) can
be formed by shifting x(n) so that as a sequence value leaves the interval 0 to N-1 at one
end it enters at the other end.
We know that a periodic sequence x p (n) can be interpreted
from the finite duration sequence x(n) by displaying x(n) around the circumference of a
circle with a circumference of exactly N points.
The linear shift of the periodic sequence
x p (n) is equal shifting the sequence x(n) circularly in anticlockwise direction.
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 13of 31
Summary:-If you consider a length-N sequence defined for 0<=n<=N-1. such a sequence
have sample values equal to zero for n<0 and n>=N.
If x[n] is such a sequence , then for any arbitrary integer n0, the shifted sequence
x1[n]=x[n-n0] is no longer defined for the range 0<=n<=N-1. we therefore need to define
another type of shift that will always keep the shifted sequence in the range 0<=n<=N-1.
This is known as circular shift.
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 14of 31
The last expression has the form of a convolution sum. However it is not the ordinary
linear convolution. Instead the convolution sum involves the index and is
called circular convolution. Thus we conclude that multiplication of the DFT of 2
sequence is equivalent to the circular convolution of the to sequences in time domain.
example:-
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 15of 31
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 16of 31
similarly
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 17of 31
Example:-
x1(n)={2,1,2,1} x2(n)={1,2,3,4}
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 18of 31
⎡ x 2 ( 0) x 2 ( N − 1) x 2 ( N − 2)....... x 2 ( 2) x 2 (1) ⎤
⎢ x (1) x 2 ( N − 1)....... x 2 ( 2) ⎥⎦
⎣ 2 x 2 ( 0) x 2 (3)
n -3 -2 -1 0 1 2 3
x1(n) 2 1 2 1
x2(n) 1 2 3 4
When m=0
x2(-n) 4 3 2 1 4 3 2 x3(0)=2.1+1.4+2.4+1.2=14
When m=1
x2(1-n) 4 3 2 1 4 3 x3(1)=2.2+1.1+2.4+1.3=16
When m-2
x2(2-n) 4 3 2 1 4 x3(2)=2.3+1.2+2.1+1.4=14
When m=3
x2(3-n) 4 3 2 1 x3(3)=2.4+1.3+2.2+1.1=16
x3(m)={14,16,14,16}
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 20of 31
In Linear filter the basic building block is a LTI system. Let us consider a LTI system of
frequency response H(w)
H(w)
X(w) Y(w)=H(w)X(w)
h(n)/H[k]
x(n)/X[k] Y(k)=H(K)X(k)
But IDFT of Y[k] is the circular convolution. But the desired output is linear convolution. So we
must do something to get linear convolution output.
Let x[n] is finite duration sequence of length L
And h[n] is finite duration sequence of length M
We can write
X[n]=0 for n<0 and n >= L
And h[n]=0 for n<0 and n>=M
Both x[n] and h[n] are causal sequence
Therefore]
M −1
y[n] = ∑ h[k ]x[n − k ] = x[n] * h[n]
k =0
Impulse response of the linear system is a finite duration sequence, hence called FIR filter or
finite duration Impulse Response Filter.
The output sequence is also a finite duration sequence of length (L+M-1).
Conversely we can say that if the output sequence contains at least (L+M-1) samples or more ,
then taking IDFT of output sequence we can get the time domain sequence y[n]. So the size of the
DFT must be greater than (L+M-1) or
N >= L+M-1
Note:- N can have any value greater than (L+M-1). But the value of N should be power of 2
otherwise DFT computation will have large amount of time.
The output signal spectrum
Y(K)=Y(w)|w=2.PI.K/N, k=0,1,2,…N-1
= X(w).H(w)| w=2.PI.K/N
=X(k).H(K)
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 21of 31
Example:=BY means of DFT and IDFT determine the response of FIR filter with impulse
response h[n]={1,2,3,} to the input sequence x[n]={1,2,2,1}
Solution:- In this case L=4 & M=3
N>=3+4-1 or N>=6
Choosing N as power of 2 ,let N=8
Linear convolution of x[n] & h[n]={1,4,9,11,8,3}
1 2 2 1
1 1 2 2 1
2 2 4 4 2
3 3 6 6 3
2+ 2 4+3 2 H ( 0) = 6
X (1) = − j( )
2 2 H (1) = 1 + 2 − j (3 + 2)
X ( 2) = −1 − j
H ( 2) = −2 − j 2
2− 2 4 −3 2
X ( 3) = + j( ) H ( 3) = 1 − 2 + j (3 − 2)
2 2
X ( 4) = 0 H ( 4) = 2
2− 2 4 −3 2 H ( 5) = 1 − 2 − j (3 − 2)
X ( 5) = − j( ) H ( 6) = −2 + j 2
2 2
X ( 6) = −1 + j H (7) = 1 + 2 + j (3 + 2)
2+ 2 4+3 2
X (7) = + j( )
2 2
Y(0)=36
Y(1)=-14.07-j17.48
Y(2)=j4
Y(3)=0.07+j0.515
Y(4)=0
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 22of 31
Y(5)=0.07-j0.515
Y(6)=-j4
Y(7)=-14.07+j17.48
1 7
y[n]= ∑ Y [ K ]e
j 2πkn / 8
; n = 0,1,2...7
8 k =0
y[n]={1,4,9,11,8,3,0,0}
Linear Filtering:=
x[n]={1,2,2,1}Æ L=4
h[n]={1,2,3}ÆM=3
N>=L+M-1=6
Summary:= since the DFT provides a discrete frequency representation of a finite duration
sequence in the frequency domain, It is interesting to explore its use as a computational tool for
linear system analysis, specially for linear filtering. Digital computers can only store and perform
computations on quantities at discrete frequencies.
So we can use linear filtering in freq. domain using DFT as an alternative to time domain
convolution.
In fact Freq. domain approach based on DFT is computationally more efficient than time domain
convolution due to existence of efficient algorithms for computing the DFT known a FFT.
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 23of 31
We have demonstrated that the product of 2 DFT is equivalent to circular convolution of the time-
domain sequences. But circular convolution is of no use to us if our objective is to determine the
output of a linear filter to a given input sequence.
In this case we seek a frequency domain methodology equivalent to linear convolution.
The N point circular convolution is equivalent to linear convolution. So here we first find the N
point DFT of x[n] and h[n]. Then multiply them and then find IDFT. To make the x[n] and h[n] N
point sequence we use zero padding technique. Where N>=L+M-1.
So finally with the help of zero padding DFT can be used to perform linear filleting.
The length M is fixed but L can have any value. So the entire long duration sequence is divided
into blocks. For each block DFT is performed and output sequence is calculated. The output
sequence of each block is overlapped to get the final output sequence. This is called Overlap-Save
method.
A block contains (M-1) data points from previous block & L no. of points from current block. So
Total No of data points =L+M-1;which is the required condition for DFT/IDFT method.
For the first block, since there is no previous block , we must take
(M-1) zeros.
L+M-1
In each block output seq. , discard first (M-1) no. of data points a they will be computed due to
aliasing error.
In the h[n] seq. , we have to pad (L-1) no. of zeros , to make the length =L+M-1
Example:=
Ls=15 M=3 let us choose N=5
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 25of 31
5=L+M-1
5=L+3-1
therefore L=3
x1(n)={0,0,x(0),x(1),x(2)}
x2(n)={x(1),x(2),x(3),x(4),x(5)}
x3(n)={x(4),x(5),x(6),x(7),x(8)}
x4(n)={x(7),x(8),x(9),x(10),x(11)}
x5(n)={x(10),x(11),x(12),x(13),x(14)}
x6(n)={x(13),x(14),0,0,0}
h(n)={h(0),h(1),h(2),0,0}
x1(n)={x(0),x(1),……..x(L-1),0,0,0..0,0}
M-1 zeros
x2(n)={x(L),x(L+1),……x(2L-1),0,0,0..0,0}
x3(n)={x(2L),x(2L+1),……x(3L-1),0,0,0..0,0}
y1(n)={y1(0),y1(1),….y1(N-1)}
y2(n)={y2(0),y2(1),….y2(N-1)}
y3(n)={y3(0),y3(1),….y3(N-1)}
y(n)={ y1(0),y1(1),….y1(L-1),y1(L)+y2(0),….y1(N-1)+y2(M-2),y2(M-
1,…..y3(L)+y3(0),y2(L+1)+y3(1),…..}
Example:-
Find the output y[n] of a filter whose impulse response is h[n]={1,1,1} and the input signal is
x[n]={3,-1,0,1,3,2,0,1,2,1} using i)overlap-save ii)overlap-add method
x1[n]={0,0,3,-1,0}
x2[n]={-1,0,1,3,2}
x3[n]={3,2,0,1,2}
x4[n]={1,2,1,0,0}
h[n]={1,1,1,0,0}
⎡1 0 0 1 1⎤ ⎡ 0 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢ 0 ⎥⎥
⎢
y1[n]= ⎢1 1 1 0 0⎥ ⎢ 3 ⎥ = [− 1 0 3 2 2]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢− 1⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣ 0 ⎥⎦
⎡1 0 0 1 1⎤ ⎡− 1⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢ 0 ⎥⎥
⎢
y2[n]= ⎢1 1 1 0 0⎥ ⎢ 1 ⎥ = [4 1 0 4 6]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢ 3 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣ 2 ⎥⎦
⎡1 0 0 1 1 ⎤ ⎡ 3⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢2⎥⎥
⎢
y3[n]= ⎢1 1 1 0 0⎥ ⎢0⎥ = [6 7 5 3 3]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢1 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣2⎥⎦
⎡1 0 0 1 1 ⎤ ⎡1 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢2⎥⎥
⎢
y4[n]= ⎢1 1 1 0 0⎥ ⎢1⎥ = [1 3 4 3 1]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢0 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣0⎥⎦
y[n]={3,2,2,0,4,6,5,3,3,4,3,1}
⎡1 0 0 1⎤ ⎡ 0 ⎤
0 1
⎢1 1 0 1⎥⎥ ⎢⎢ 0 ⎥⎥
0 0
⎢
⎢1 1 1 0⎥ ⎢ 3 ⎥
0 0
y1[n] = ⎢ ⎥⎢ ⎥ = 1 1 3 2 2 0
⎢0 1 1 0⎥ ⎢− 1⎥
1 0
⎢0 0 1 0⎥ ⎢ 0 ⎥
1 1
⎢ ⎥⎢ ⎥
⎣⎢0 0 0 1 1 1⎦⎥ ⎣⎢ 1 ⎦⎥
⎡1 0 0 0 1 1 ⎤ ⎡0 ⎤
⎢1 1 0 0 0 1⎥⎥ ⎢⎢1 ⎥⎥
⎢
⎢1 1 1 0 0 0 ⎥ ⎢ 3⎥
y 2[n] = ⎢ ⎥⎢ ⎥ = 1 2 4 6 5 3
⎢0 1 1 1 0 0 ⎥ ⎢ 2⎥
⎢0 0 1 1 1 0 ⎥ ⎢0 ⎥
⎢ ⎥⎢ ⎥
⎢⎣0 0 0 1 1 1⎥⎦ ⎢⎣1 ⎥⎦
⎡1 0 0 0 1 1⎤ ⎡0 ⎤
⎢1 1 0 0 0 1⎥⎥ ⎢⎢1⎥⎥
⎢
⎢1 1 1 0 0 0⎥ ⎢ 2 ⎥
y3[n] = ⎢ ⎥⎢ ⎥ = 0 1 3 4 3 1
⎢0 1 1 1 0 0⎥ ⎢1 ⎥
⎢0 0 1 1 1 0⎥ ⎢0 ⎥
⎢ ⎥⎢ ⎥
⎣⎢0 0 0 1 1 1⎦⎥ ⎣⎢0⎦⎥
y[n]= y[n]={3,2,2,0,4,6,5,3,3,4,3,1}
ii)Overlap-Add method:-
X[n]= x[n]={3,-1,0,1,3,2,0,1,2,1}
x1(n)={3,-1,0,0,0}
x2(n)={1,3,2,0,0}
x3(n)={0,1,2,0,0}
x4(n)={1,0,0,0,0}
h[n]={1,1,1,0,0}
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 29of 31
⎡1 0 0 1 1⎤ ⎡ 3 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢− 1⎥⎥
⎢
y1[n]= ⎢1 1 1 0 0⎥ ⎢ 0 ⎥ = [3 2 2 − 1 0]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢ 0 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣ 0 ⎥⎦
⎡1 0 0 1 1 ⎤ ⎡1 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢3⎥⎥
⎢
y2[n]= ⎢1 1 1 0 0⎥ ⎢2⎥ = [1 4 6 5 2]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢0 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣0⎥⎦
⎡1 0 0 1 1⎤ ⎡0 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢1⎥⎥
⎢
y3[n]= ⎢1 1 1 0 0⎥ ⎢2⎥ = [0 1 3 3 2]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢0 ⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣0⎥⎦
⎡1 0 0 1 1 ⎤ ⎡1 ⎤
⎢1 1 0 0 1⎥⎥ ⎢⎢0⎥⎥
⎢
y4[n]= ⎢1 1 1 0 0⎥ ⎢0⎥ = [1 1 1 0 0]
⎢ ⎥⎢ ⎥
⎢0 1 1 1 0⎥ ⎢0⎥
⎢⎣0 0 1 1 1⎥⎦ ⎢⎣0⎥⎦
y[n]={3,2,2,0,4,6,5,3,3,4,3,1}
Exercise:-
i) overlap-add method:-
M=2
Ls=12 let L=4
N=L+M-1=4+2-1=5
x1[n]={1,2,-1,2,0}
x2[n]={3,-2,-3,1,0}
x3[n]={1,1,2,-1,0}
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 30of 31
⎡1 0 0 0 2 ⎤ ⎡ 1 ⎤ ⎡1 ⎤
⎢2 1 0 0 0⎥⎥ ⎢⎢ 2 ⎥⎥ ⎢⎢4⎥⎥
⎢
y1[n]= ⎢0 2 1 0 0⎥ ⎢− 1⎥ = ⎢3⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0 ⎥ ⎢ 2 ⎥ ⎢0 ⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣ 0 ⎥⎦ ⎢⎣4⎥⎦
⎡1 0 0 0 2⎤ ⎡ 3 ⎤ ⎡ 3 ⎤
⎢2 1 0 0 0⎥⎥ ⎢⎢− 2⎥⎥ ⎢⎢ 4 ⎥⎥
⎢
y2[n]= ⎢0 2 1 0 0 ⎥ ⎢ − 3⎥ = ⎢ − 7 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0⎥ ⎢ − 1 ⎥ ⎢− 7 ⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣ 0 ⎥⎦ ⎢⎣− 2⎥⎦
⎡1 0 0 0 2⎤ ⎡ 1 ⎤ ⎡ 1 ⎤
⎢2 1 0 0 0⎥⎥ ⎢⎢ 1 ⎥⎥ ⎢⎢ 3 ⎥⎥
⎢
y3[n]= ⎢0 2 1 0 0⎥ ⎢ 2 ⎥ = ⎢ 4 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0⎥ ⎢− 1⎥ ⎢ 3 ⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣ 0 ⎥⎦ ⎢⎣− 2⎥⎦
therefore y[n]={1,4,3,0,7,4,-7,-7,-1,3,4,3,-2}
ii)overlap-save method:-
Ls=12
Let N=5 5=L+M-1=L+1
L=4
x1(n)={0,1,2,-1,2}
x2(n)={2,3,-2,-3,-1}
x3(n)={-1,1,1,2,-1}
x4(n)={-1,0,0,0,0}
h[n]={1,2,0,0,0}
⎡1 0 0 0 2 ⎤ ⎡ 0 ⎤ ⎡ 4 ⎤
⎢ 2 1 0 0 0 ⎥ ⎢ 1 ⎥ ⎢1 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
y1[n]= ⎢0 2 1 0 0⎥ ⎢ 2 ⎥ = ⎢4⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0⎥ ⎢− 1⎥ ⎢3⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣ 2 ⎥⎦ ⎢⎣0⎥⎦
⎡1 0 0 0 2 ⎤ ⎡ 2 ⎤ ⎡ 0 ⎤
⎢2 1 0 0 0⎥ ⎢ 3 ⎥ ⎢ 7 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
y2[n]= ⎢0 2 1 0 0⎥ ⎢− 2⎥ = ⎢ 4 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0 ⎥ ⎢ − 3⎥ ⎢ − 7 ⎥
⎢⎣0 0 0 2 1 ⎥⎦ ⎢⎣ − 1⎥⎦ ⎢⎣− 7⎥⎦
©angsuman sarkar,ece,jgec,angsumansarkar@gmail.com page 31of 31
⎡1 0 0 0 2⎤ ⎡− 1⎤ ⎡− 3⎤
⎢2 1 0 0 0⎥⎥ ⎢⎢ 1 ⎥⎥ ⎢⎢ − 1⎥⎥
⎢
y3[n]= ⎢0 2 1 0 0⎥ ⎢ 1 ⎥ = ⎢ 3 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0⎥ ⎢ 2 ⎥ ⎢ 4 ⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣− 1⎥⎦ ⎢⎣ 3 ⎥⎦
⎡1 0 0 0 2⎤ ⎡− 1⎤ ⎡ − 1⎤
⎢2 1 0 0 0⎥⎥ ⎢⎢ 0 ⎥⎥ ⎢⎢− 2⎥⎥
⎢
y4[n]= ⎢0 2 1 0 0⎥ ⎢ 0 ⎥ = ⎢ 0 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 2 1 0⎥ ⎢ 0 ⎥ ⎢ 0 ⎥
⎢⎣0 0 0 2 1⎥⎦ ⎢⎣ 0 ⎥⎦ ⎢⎣ 0 ⎥⎦
y[n]={1,4,3,0,7,4,-7,-7,-1,3,4,3,-2}