Beruflich Dokumente
Kultur Dokumente
ca/~dkundur/course/discrete-time-systems/
7.1
x n is a real valued sequence. The first five points of its 8-point DFT are:
{0.25, 0.125 - j0.3018, 0, 0.125 - j0.0518, 0}
To compute the 3 remaining points, we can use the following property for real valued
sequences:
X N k X * k X k (page 468 in the book)
In our case N=8 and therefore we have the equations for X 5 , X 6 and X 7 :
X 5 X 8 3 X * 3
0.125 j 0.0518
0.125 j 0.0518
X 6 X 8 2 X * 2
0
0
X 7 X 8 1 X * 1
0.125 j 0.3018
0.125 j 0.3018
7.3
1, 0 k kc , N kc k N 1,
H k
0, kc k N kc
Hence this leads to the conclusion that x n , the inverse N-point DFT of
X k , is a
lowpass version of x n .
7.7
exp j
N
N
n 0
exp j
exp j
N
N
N
n 0
2
2 k k0 n 1 N 1
2 k k0 n
1 N 1
x n exp j
x n exp j
2 n 0
N
N
2 n 0
1
1
X k k0 mod N X k k0 mod N
2
2
Operating the same way for the sequence xs n we get the corresponding DFT X s k :
X s k
1
1
X k k0 mod N X k k0 mod N
2j
2j
7.13 a)
N 1
X 1 k x n WNkn
n 0
X3 k
3 N 1
x n W
kn
3N
n0
X 3 k x n W3knN
n 0
N 1
2 N 1
x n W
n N
kn
3N
3 N 1
x n W
kn
3N
n2 N
N 1
N 1
n0
n0
n0
N 1
N 1
N 1
n0
N 1
n 0
n0
N 1
n 0
7.23
We have to compute the N-point DFT of 4 signals:
a) x n n
b) x n n n0 , 0 n0 N
c) x n a n , 0 n N 1
1, n even
h) x n
0 n N 1
0, n even
a) x n n
The N-point DFT of x n n is defined as:
N 1
2 kn
X k n exp j
n 0
2 k (0)
0 exp j
Therefore:
DFT
x n n , 0 n N 1
X k 1, 0 k N 1
N
b) x n n n0 , 0 n0 N
The N-point DFT of x n n n0 , 0 n0 N is defined as:
N 1
2 kn
X k n n0 exp j
n 0
2 k (n0 )
n0 exp j
2 k (n0 )
exp j
Therefore:
2 kn0
DFT
x n n n0 , 0 n0 N
X k exp j
, 0 k N 1
N
N
c) x n a n , 0 n N 1
The N-point DFT of x n a n , 0 n N 1 is defined as:
N 1
2 kn
X k a n exp j
n 0
2 k
a exp j
n0
N 1
2 k
1 a exp j
2 k
1 a exp j
1 aN
2 k
1 a exp j
Therefore:
DFT
x n a n , 0 n N 1
X k
N
1 aN
, 0 k N 1
2 k
1 a exp j
1, n even
h) x n
0 n N 1
0, n even
1, n even
The N-point DFT of x n
0 n N 1 is defined as:
0, n even
N 1
2 kn
X k x n exp j
n 0
2 k (2n)
2 k (4n)
2 k ( N 1)
X k 1 exp j
exp j
... exp j
N
N
N
N 1
terms
2
i.e.,
N 1
2 (2k ) 2
1 exp j
X k
2 (2 k )
1 exp j
2 k
1 exp j
4 k
1 exp j
2 k
1 exp j
2 k
2 k
1 exp j N 1 exp j N
1
2 k
1 exp j
Therefore:
1, n even
1
DFT
x n
0 n N 1
X k
, 0 k N 1
N
2 k
0, n even
1 exp j
7.28
n
a , n L
We are given a discrete-time signal x n
where a 0.95 and L 10 .
0, n L
(a) Here we need to compute and plot x n . Obviously from the given values of a and L ,
we have:
0.95 n , n 10
x n
n 10
0,
0.8
x(n)
0.6
0.4
0.2
0
-15
-10
-5
0
n
10
15
exp j n
exp j n
n L
n L
n 1
1 a n exp j n exp j n
n 1
L
1 a n 2 cos n
n 1
x 0 2 x n cos n
n 1
k
, k 0,1,..., N 1, can be found below.
100
20
15
X()
10
-5
0.5
1.5
2.5
1 2
X
N N
k , k 0,1,..., N 1
3.5
For N 30 , ck becomes:
ck
1 2
X
30 30
k , k 0,1,..., 29
ck
L
1
2
x
0
2
x n cos
kn , k 0,1,..., 29
30
30
n 1
1
2
n
1
2
0.95 cos kn , k 0,1,..., 29
30
30
n 1
ck
0.3
0.2
0.1
0
-0.1
10
15
k
20
25
k 0
30
29
1 2
2 kn
x n X
k exp j
N
30
k 0 30
1 29 2
2 kn
X
k exp j
30 k 0 30
N
1 29
X w exp jwn
30 k 0
x ~(n)
0.6
0.5
0.4
0.3
0.2
0.1
0
-15
-10
-5
0
n
x n lN ,
Ln L
x n 30l ,
Ln L
10
15
From the corresponding plot below, we can see that x1 n is a periodic/repeated version
of x n .
N=30
1
0.9
0.8
0.7
x ~1(n)
0.6
0.5
0.4
0.3
0.2
0.1
0
-80
-60
-40
-20
0
n
20
40
60
80
(f) Here we just have to replace N by 15 instead of 30 in the previously obtained equation.
This is trivial so just the new plots are being shown.
N=15
1.2
1
0.8
ck
0.6
0.4
0.2
0
-0.2
8
k
10
12
14
N=15
1.5
1.4
1.3
x ~(n)
1.2
1.1
0.9
0.8
-10
-8
-6
-4
-2
0
n
10
N=15
1.5
1.4
1.3
x ~1(n)
1.2
1.1
0.9
0.8
-40
-30
-20
-10
0
n
10
20
30
8.1
2 k
To show that exp j
, 0 k N 1 is an Nth root of unity we just have to show
N
2 k
that X N 1 for X exp j
, 0 k N 1 . This is fairly obvious since:
N
N
2 k
exp j N exp j 2 k 1 .
2 k
Hence, exp j
, 0 k N 1 is an Nth root of unity.
N
Now if we consider the sum used in the orthogonality property, we can rewrite that sum
as:
2 k l n
2 kn
2 ln N 1
exp j
exp j
exp j
N
N n0
N
n 0
N 1
If k l , the terms in the sum represent the N equally spaced unity roots on the unit circle
which add to zero.
Mathematical proof if k l :
N 1
n 0
exp
2 k l
1 exp j
N
2 k l n
N
2 k l
1 exp j
1 1
2 k l
1 exp j
N
N
n 0
n 0
N 1
N 1
1
n0
8.3
n 0
N
1
2
2 kn
X k x n W x n WNkn , where WNkn exp j
N
N
n 0
n
N 1
kn
N
N
1
2
N
1
2
N k n
x n W x n WN 2
2
n0
n 0
kn
N
X ' k x n WN(2 k 1) n x n WN 2
2
n 0
N
N 1
n0
2
2
N
2
N
1 :
N 1
n 0
2
2
n 0
2
8.4
We want to develop a method to compute a 24-point DFT from three 8-point DFTs.
Let Y k denote the 24-point DFT and Y1 k , Y2 k , Y3 k denote the three 8-point
DFTs. We then have:
N 1
23
n0
n 0
Y k y n WNkn y n WNkn
We can rewrite this sum as three sums that would take values n among the sets {0, 3,
6, , 21}, {1, 4, 7, , 22} and {2, 5, 8, , 23} respectively.
21
Y k
n 0,3,6,...
22
y n WNkn
n 1,4,7,...
y n WNkn
23
n 2,5,8,...
y n WNkn
n 0.
n 0.
7
7
8 pt DFT
8 pt DFT
8 pt DFT
Y1 k Y2 k WNk Y3 k WN2 k
With three 8-point DFTs, Y1 k , Y2 k , Y3 k , we can create a 24-point DFT Y k
using the following formulae:
Y k Y1 k Y2 k WNk Y3 k WN2 k
8.7
We want to derive the radix-2 decimation in time using the steps 8.1.16 to 8.1.18 in the
book.
Page 519 in the book already gives some guidelines on how to proceed such as selecting
N
M
and L 2 .
2
1) The first step to follow (8.1.16) makes us compute the M-point DFTs F l , q defined
as:
M 1
F l , q x l , m WMmq , 0 l L 1; 0 q M 1
m0
N
-point DFTs to compute for l 0 and l 1 .
2
M 1
N
1
2
m 0
m 0
F 0, q x 0, m WMmq x 0, m WNmq
2
N
1
2
F 1, q x 1, m W Nmq
m0
G l , q WNlq F l , q , 0 l L 1; 0 q M 1
Therefore we have two rectangular arrays to compute for l 0 and l 1 .
G 0, q WM(0) q F 0, q F 0, q
N
1
2
3) The third and last step (8.1.18) consists in computing the L-point DFT X p, q
defined as:
L 1
X p, q G l , q WLlp , 0 p L 1; 0 q M 1
l 0
L 1
l 0
l 0
X 0, q G l , q WL(0)l G l , q
G 0, q G 1, q
F 0, q WNq F 1, q
L 1
L 1
l 0
l 0
X 1, q G l , q WL(1)l G l , q WLl
G 0, q W20 G 1, q W21
G 0, q G 1, q as W21 1
F 0, q WNq F 1, q
X 0, q X k F1 k WNk F2 k
k
X 1, q X k 2 F1 k WN F2 k
,0 k
N
1
2