Procesado Digital de Seales. 4 de Ingeniera Electrnica.

Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.

Transformada Discreta
de Fourier.
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Hasta ahora se ha visto......
Importancia de la respuesta
en frecuencia de un sistema
Tenemos otra forma de caracterizar
los sistemas L.T.I
Herramienta muy potente para
determinar salidas cuando las entradas
son sinusoides o combinacin de stas.
Permite realizar aplicaciones que en el
dominio temporal son difciles de
comprender (por ejemplo ltrado)
Transformada de Fourier de
una seal discreta
Tenemos una forma de expresar la
seal como una suma innita de
sinusoides
Utilizando esta descomposicin de la
seal junto con la respuesta en
frecuencia tenemos una forma sencilla
de determinar la salida de un sistema en
el estacionario.
PROBLEMA: TENEMOS UN SUMATORIO CON LMITES
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Desarrollo en serie de Fourier discreto (DFS)
Supongamos una seal discreta x(n) peridica (periodo N); esto es
x(n+N)=x(n). Sabemos que a nivel temporal esta seal se puede
expresar, en el primer periodo y usando deltas desplazadas, como:
!
x(n) = x(k) "# n $ k
( )
k= 0
N
%


!
x(n) =" # X(k) # e
$ j# 2# % # k# n
N
k= 0
N$1
&


!
x(n) " e
j" 2" # " s" n
N
=$ " X(k) " e
% j" 2" # " k" n
N
k= 0
N%1
&
'
(
)
)
*
+
,
,
" e
j" 2" # " s" n
N



!
x(n) " e
j" 2" # " s" n
N
n= 0
N$1
%
=& " X(k) " e
$ j" 2" # " k$s ( )" n
N
n= 0
N$1
%
k= 0
N$1
%



!
e
" j# 2# $ # k"s ( )# n
N
n= 0
N"1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
j" 2" # " s" n
N
n= 0
N$1
%
=& " N " X(s)




!
X(k) = x(n) " e
j" 2" # " k" n
N
n= 0
N$1
%
x(n) =
1
N
" X(k) " e
$ j" 2" # " k" n
N
k= 0
N$1
%









Podemos plantear otra representacin alternativa usando
exponenciales complejas que, como ya hemos visto, son la base
en el dominio frecuencial, esto es
Evidentemente el objetivo ahora es determinar X(k) y ; para ello algunas operaciones.....
Aplicando la ortogonalidad de las
exponenciales complejas, esto es,
!
x(n) = x(k) "# n $ k
( )
k= 0
N
%


!
x(n) =" # X(k) # e
$ j# 2# % # k# n
N
k= 0
N$1
&


!
x(n) " e
j" 2" # " s" n
N
=$ " X(k) " e
% j" 2" # " k" n
N
k= 0
N%1
&
'
(
)
)
*
+
,
,
" e
j" 2" # " s" n
N



!
x(n) " e
j" 2" # " s" n
N
n= 0
N$1
%
=& " X(k) " e
$ j" 2" # " k$s ( )" n
N
n= 0
N$1
%
k= 0
N$1
%



!
e
" j# 2# $ # k"s ( )# n
N
n= 0
N"1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
j" 2" # " s" n
N
n= 0
N$1
%
=& " N " X(s)




!
X(k) = x(n) " e
j" 2" # " k" n
N
n= 0
N$1
%
x(n) =
1
N
" X(k) " e
$ j" 2" # " k" n
N
k= 0
N$1
%











!
k(n) =
P
n"1
# x
n
1+
r
x
n
t
# P
n"1
#
r
x
n





!

e (n) = d(n) "
r
x
n
t
# w
n"1



!
w
n
= w
n"1
+ k(n) #

e (n)




!
P
n
= P
n"1
" k(n) #
r
x
n
t
# P
n"1





!
J = "
n#l
e
2
(l)
l= 0
n
$




!
k(n) =
P
n"1
# x
n
$ +
r
x
n
t
# P
n"1
#
r
x
n



!
P
n
=
P
n"1
" k(n) #
r
x
n
t
# P
n"1
$




!
x(n) = x(k) "# n $ k
( )
k= 0
N
%


!
x(n) =" # X(k) # e
j# 2# $ # k# n
N
k= 0
N%1
&


!
x(n) " e
# j" 2" $ " s" n
N
=% " X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
&
'
(
)
)
*
+
,
,
" e
# j" 2" $ " s" n
N



!
k(n) =
P
n"1
# x
n
1+
r
x
n
t
# P
n"1
#
r
x
n





!

e (n) = d(n) "
r
x
n
t
# w
n"1



!
w
n
= w
n"1
+ k(n) #

e (n)




!
P
n
= P
n"1
" k(n) #
r
x
n
t
# P
n"1





!
J = "
n#l
e
2
(l)
l= 0
n
$




!
k(n) =
P
n"1
# x
n
$ +
r
x
n
t
# P
n"1
#
r
x
n



!
P
n
=
P
n"1
" k(n) #
r
x
n
t
# P
n"1
$




!
x(n) = x(k) "# n $ k
( )
k= 0
N
%


!
x(n) =" # X(k) # e
j# 2# $ # k# n
N
k= 0
N%1
&


!
x(n) " e
# j" 2" $ " s" n
N
=% " X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
&
'
(
)
)
*
+
,
,
" e
# j" 2" $ " s" n
N



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%












!
W
N
k" n
= e
# j" 2" $ " k" n
N










tomando =1/N


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%












!
W
N
k" n
= e
# j" 2" $ " k" n
N










Ecuacin de sntesis
Ecuacin de anlisis


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%












!
W
N
k" n
= e
# j" 2" $ " k" n
N










Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Ejemplo de DFS.

Consideramos la siguiente seal
peridica aqu N=10


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5 " n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
k" n
= e
# j" 2" $ " k" n
N



!
X(k) = x(n) "
n= 0
N#1
$
e
# j" 2" % " k" n
N




!
X e
jw
( )
= x(n) " e
# j" w" n
n=#$
$
%

Aplicando la denicin de DFS se tiene
!
X(k) =
U(k) 0 " k " N #1
0 otro caso

$
%
&
















!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= e
# j" 2" $ " k" n
10
n= 0
4
%
=
1#e
# j" 2" $ " k" 5
10
1#e
# j" 2" $ " k
10



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*











Agrupando exponenciales (como siempre!!!).
!
X(k) =
U(k) 0 " k " N #1
0 otro caso

$
%
&
















!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= e
# j" 2" $ " k" n
10
n= 0
4
%
=
1#e
# j" 2" $ " k" 5
10
1#e
# j" 2" $ " k
10



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*











Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Propiedades del DFS.
!. E. Angoletta - DISP2003 - !"#$%&$'()(*+,%,'- .($/ 01''!"#$%&$'2&$%&,''''' 22 !"24
DFS properties DFS properties
!"#$%%%%%%%%%%%%%%%%%%%%%%&'$()$*+, !"#$%%%%%%%%%%%%%%%%%%%%%%&'$()$*+,
-.#./$*$"0, as[n] aS(k)
122"0"3"0, s[n] + u[n] S(k)+U(k)
4"*$5'"0, as[n] + bu[n] aS(k)+bU(k)
6)70"87"+50".* 9 s[n] u[n]
:.*3.7)0".* 9 S(k)U(k)
!"#$%;<"=0"*/ s[n - m]
&'$()$*+,%;<"=0"*/ S(k - h)
!
"
#
$
1 N
0 h
h) - S(h)U(k
N
1
!" !"#$%&'()*+"()")*$,"-**./0"&*1,23* #$%&'()*+"()")*$,"-**./0"&*1,23*
!
"
#
" $
1 N
0 m
m] u[n s[m]
S(k) e
T
m k 2!
j
$
$
"
s[n]
T
t h 2!
j
e $
%
Linealidad
Producto
Convolucin peridica
Desplazamiento temporal
Desplazamiento frecuencial.
N N 0
m
x
2
[m]
~
N N 0
m
x
2
[ m]
~
N N 0
m
x
1
[m]
~
N N 0
m
x
2
[1 m] = x
2
[(m 1)]
~ ~
x
2
[2 m] = x
2
[(m 2)]
~ ~
N N 0
m
Figure 8.3 Procedure for forming the periodic convolution of two periodic
sequences.
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
Ejemplo de convolucion peridica
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Muestreo de la Transformada de
Fourier de una secuencia discreta (I)
Sabemos que la Transformada de Fourier
de una secuencia discreta viene denida
por la siguiente expresin:


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5" n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
k" n
= e
# j" 2" $ " k" n
N



!
X(k) = x(n) "
n= 0
N#1
$
e
# j" 2" % " k" n
N




!
X e
jw
( )
= x(n) " e
# j" w" n
n=#$
$
%

!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



Muestreamos (EN EL DOMINIO FRECUENCIAL)
considerando N muestras; recordemos que X(jw) es
peridica con periodo
U(k) es peridica (periodo N) se puede
corresponder a los coecientes de un DFS.
Estoy interesado en la seal temporal que se
obtiene de esos coecientes
!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



Aplicando
!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Muestreo de la Transformada de
Fourier de una secuencia discreta (II)
!
U(k) = X e
" j# 2# $ # k
N
%
&
'
'
(
)
*
*
= x(n) # e
" j# 2# $ # k# n
N
n="+
+
,



!
u(n) =
1
N
" U(k) " e
j" 2" # " k" n
N
k= 0
N$1
%


!
u(n) =
1
N
" x(m) "
m=#$
$
%
e
# j" 2" & " k" m
N
'
(
)
)
*
+
,
,
"
k= 0
N#1
%
e
j" 2" & " k" n
N



!
u(n) = x(m) "
m=#$
$
%
1
N
" e
j" 2" & " k" n#m ( )
N
k= 0
N#1
%
'
(
)
)
*
+
,
,


!
e
j" 2" # " k" n$m ( )
N
k= 0
N$1
%
=
N n $ m = r " N
0 n $ m & r " N
r entero
'
(
)



!
u(n) = x(m) "#(n $ m$ r " N)
m=$%
%
&
'u(n) = x(n $ r " N)
r=$%
%
&



!
u(n) = x(n " r # N)
r="$
$
%



!
x(n) =
u(n) 0 " n " N #1
0 otro caso

$
%
&



Se llega nalmente a
... ...
12 0 8
N = 12
x[n] =
~
x[n r12]
r =

0 8
x[n]
(a)
(b)
n
n
Figure 8.8 (a) Finite-length sequence x[n]. (b) Periodic sequence x[n] corre-
sponding to sampling the Fourier transform of x[n] with N = 12.
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
... ...
12 0 8
N = 12
x[n] =
~
x[n r12]
r =

0 8
x[n]
(a)
(b)
n
n
Figure 8.8 (a) Finite-length sequence x[n]. (b) Periodic sequence x[n] corre-
sponding to sampling the Fourier transform of x[n] with N = 12.
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
... ...
14 7 0
N = 7
14
x[n] =
~
x[n r7]
r =

n
Figure 8.9 Periodic sequence x[n] corresponding to sampling the Fourier trans-
form of x[n] in Figure 8.8(a) with N = 7.
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Transformada Discreta de Fourier (DFT).
Conclusiones a tener en cuenta.
Si x(n) tiene longitud nita se puede recuperar
dicha seal a partir de muestras de su
Transformada de Fourier; no es necesario
conocer dicha Transformada en todas las
frecuencias.
La seal temporal obtenida usando dicho
muestreo de la Transformada de Fourier se
supone peridica por construccin; aunque a
nosotros slo nos interesa el primer periodo.
Se dene la Transformada Discreta de
Fourier (DFT) de una seal x(n) como .


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5" n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
= e
" j# 2# $
N



!
X(k) = x(n) "W
N
k" n
n= 0
N#1
$
x(n) =
1
N
" X(k) "W
N
#k" n
k= 0
N#1
$






La Transformada inversa queda
denida como (ecuacin de sintesis)


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5" n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
= e
" j# 2# $
N



!
X(k) = x(n) "W
N
k" n
n= 0
N#1
$
x(n) =
1
N
" X(k) "W
N
#k" n
k= 0
N#1
$






Si se dene


!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5 " n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
= e
" j# 2# $
N



!
X(k) = x(n) "W
N
k" n
n= 0
N#1
$
x(n) =
1
N
" X(k) "W
N
#k" n
k= 0
N#1
$








!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5 " n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
= e
" j# 2# $
N



!
X(k) = x(n) "W
N
k" n
n= 0
N#1
$
x(n) =
1
N
" X(k) "W
N
#k" n
k= 0
N#1
$






Ec. Anlisis
Ec. Sntesis
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Ejemplo de DFT (I).
5
| X(e
j
)|
0
0
5 10 1 1
... ...
2
1
1
3 4 6 7 8 9 11
x[n]
~
x[n]
(c)
(b)
(a)
2 4
5
0 1 2 5 10 1 2 3 4 6 7 8 9 11
X[k]
(d)
... ...
0 5 10 15 20
0 4
n
n
X[k]
~
k
k
Figure 8.10 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 5. (c) Fourier series coefcients

X [k] for x[n]. To emphasize that the Fourier series coefcients are samples of the
Fourier transform, |X (e
j
)| is also shown. (d) DFT of x[n].
F
r
o
m

D
i
s
c
r
e
t
e
-
T
i
m
e

S
i
g
n
a
l

P
r
o
c
e
s
s
i
n
g
,

2
e

b
y

O
p
p
e
n
h
e
i
m
,

S
c
h
a
f
e
r
,

a
n
d

B
u
c
k

1
9
9
9
-
2
0
0
0

P
r
e
n
t
i
c
e

H
a
l
l
,

I
n
c
.

Sea la seal x(n); vamos a
calcular su DFT con N=5.
Implcitamente suponemos
que se tiene la seal
5
| X(e
j
)|
0
0
5 10 1 1
... ...
2
1
1
3 4 6 7 8 9 11
x[n]
~
x[n]
(c)
(b)
(a)
2 4
5
0 1 2 5 10 1 2 3 4 6 7 8 9 11
X[k]
(d)
... ...
0 5 10 15 20
0 4
n
n
X[k]
~
k
k
Figure 8.10 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 5. (c) Fourier series coefcients

X [k] for x[n]. To emphasize that the Fourier series coefcients are samples of the
Fourier transform, |X (e
j
)| is also shown. (d) DFT of x[n].
F
r
o
m
D
i
s
c
r
e
t
e
-
T
i
m
e
S
i
g
n
a
l
P
r
o
c
e
s
s
i
n
g
, 2
e
b
y
O
p
p
e
n
h
e
i
m
, S
c
h
a
f
e
r
, a
n
d
B
u
c
k

1
9
9
9
-
2
0
0
0
P
r
e
n
t
i
c
e
H
a
l
l
, I
n
c
.
Aplicando la denicin y
haciendo algunos clculos





!
X(k) = x(n) " e
# j" 2" $ " k" n
5
n= 0
4
%
=
1#e
# j" 2" $ " k
1#e
#2" j" $ " k
5
=
0 k & 5" r
5 k = 5" r
r entero
'
(
)



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*


5
| X(e
j
)|
0
0
5 10 1 1
... ...
2
1
1
3 4 6 7 8 9 11
x[n]
~
x[n]
(c)
(b)
(a)
2 4
5
0 1 2 5 10 1 2 3 4 6 7 8 9 11
X[k]
(d)
... ...
0 5 10 15 20
0 4
n
n
X[k]
~
k
k
Figure 8.10 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 5. (c) Fourier series coefcients

X [k] for x[n]. To emphasize that the Fourier series coefcients are samples of the
Fourier transform, |X (e
j
)| is also shown. (d) DFT of x[n].
F
r
o
m
D
is
c
r
e
te
-
T
im
e
S
ig
n
a
l P
r
o
c
e
s
s
in
g
, 2
e
b
y
O
p
p
e
n
h
e
im
, S
c
h
a
f
e
r
, a
n
d
B
u
c
k

1
9
9
9
-
2
0
0
0
P
r
e
n
tic
e
H
a
ll, I
n
c
.
5
| X(e
j
)|
0
0
5 10 1 1
... ...
2
1
1
3 4 6 7 8 9 11
x[n]
~
x[n]
(c)
(b)
(a)
2 4
5
0 1 2 5 10 1 2 3 4 6 7 8 9 11
X[k]
(d)
... ...
0 5 10 15 20
0 4
n
n
X[k]
~
k
k
Figure 8.10 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 5. (c) Fourier series coefcients

X [k] for x[n]. To emphasize that the Fourier series coefcients are samples of the
Fourier transform, |X (e
j
)| is also shown. (d) DFT of x[n].
F
r
o
m
D
is
c
r
e
te
-
T
im
e
S
ig
n
a
l P
r
o
c
e
s
s
in
g
, 2
e
b
y
O
p
p
e
n
h
e
im
, S
c
h
a
f
e
r
, a
n
d
B
u
c
k

1
9
9
9
-
2
0
0
0
P
r
e
n
tic
e
H
a
ll, I
n
c
.
Lo que obtenemos son innitas
muestras periodicas; slo nos
interesa el primer periodo
Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Ejemplo de DFT (II).
Ahora consideramos N=10.
Implcitamente suponemos
que se tiene la seal
x[n]
~
(b)
10 0 4 10
| X[k]|
(c)
10 0 10
x[n]
(a)
0
1
1
5
3.24 3.24
1.24 1.24
0.4
0.2
0.2
0.4
1
4
X[k]
(d)
10 0 10 k
k
n
nn
Figure 8.11 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 10. (c) DFT magnitude. (d) DFT
phase. (xs indicate indeterminate values.)
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
x[n]
~
(b)
10 0 4 10
| X[k]|
(c)
10 0 10
x[n]
(a)
0
1
1
5
3.24 3.24
1.24 1.24
0.4
0.2
0.2
0.4
1
4
X[k]
(d)
10 0 10 k
k
n
nn
Figure 8.11 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 10. (c) DFT magnitude. (d) DFT
phase. (xs indicate indeterminate values.)
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
x[n]
~
(b)
10 0 4 10
| X[k]|
(c)
10 0 10
x[n]
(a)
0
1
1
5
3.24 3.24
1.24 1.24
0.4
0.2
0.2
0.4
1
4
X[k]
(d)
10 0 10 k
k
n
nn
Figure 8.11 Illustration of the DFT. (a) Finite-length sequence x[n]. (b) Periodic
sequence x[n] formed from x[n] with period N = 10. (c) DFT magnitude. (d) DFT
phase. (xs indicate indeterminate values.)
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.
Operando





!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= x(n) " e
# j" $ " k" n
5
n= 0
4
%
=
1#e
# j" $ " k
1#e
# j" $ " k
5
= e
# j" $ " 0.4" k
"
sen
$ " k
2
&
'
(
)
*
+
sen
$ " k
10
&
'
(
)
*
+



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*







!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= x(n) " e
# j" $ " k" n
5
n= 0
4
%
=
1#e
# j" $ " k
1#e
# j" $ " k
5
= e
# j" $ " 0.4" k
"
sen
$ " k
2
&
'
(
)
*
+
sen
$ " k
10
&
'
(
)
*
+



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*


Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Propiedades de la DFT
TABLE 8.2
Finite-Length Sequence (Length N) N-point DFT (Length N)
1. x[n] X[k]
2. x
1
[n], x
2
[n] X
1
[k], X
2
[k]
3. ax
1
[n] +bx
2
[n] aX
1
[k] +bX
2
[k]
4. X[n] Nx[((k))
N
]
5. x[((n m))
N
] W
km
N
X[k]
6. W
n
N
x[n] X[((k ))
N
]
7.
N1

m=0
x
1
(m)x
2
[((n m))
N
] X
1
[k]X
2
[k]
8. x
1
[n]x
2
[n]
1
N
N1

=0
X
1
()X
2
[((k ))
N
]
9. x

[n] X

[((k))
N
]
10. x

[((n))
N
] X

[k]
11. Re{x[n]} X
ep
[k] =
1
2
{X[((k))
N
] + X

[((k))
N
]}
12. j Jm{x[n]} X
op
[k] =
1
2
{X[((k))
N
] X

[((k))
N
]}
13. x
ep
[n] =
1
2
{x[n] + x

[((n))
N
]} Re{X[k]}
14. x
op
[n] =
1
2
{x[n] x

[((n))
N
]} j Jm{X[k]}
Properties 1517 apply only when x[n] is real.
15. Symmetry properties

X[k] = X

[((k))
N
]
Re{X[k]} = Re{X[((k))
N
]}
Jm{X[k]} = Jm{X[((k))
N
]}
|X[k]| = |X[((k))
N
]|
{X[k]} = {X[((k))
N
]}
16. x
ep
[n] =
1
2
{x[n] + x[((n))
N
]} Re{X[k]}
17. x
op
[n] =
1
2
{x[n] x[((n))
N
]} j Jm{X[k]}
From Discrete-Time Signal Processing, 2e by Oppenheim, Schafer, and Buck 1999-2000 Prentice Hall, Inc.





!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= x(n) " e
# j" $ " k" n
5
n= 0
4
%
=
1#e
# j" $ " k
1#e
# j" $ " k
5
= e
# j" $ " 0.4" k
"
sen
$ " k
2
&
'
(
)
*
+
sen
$ " k
10
&
'
(
)
*
+



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*


!
x n
( ) ( )
N
[ ]
= x n mdulo N
( ) [ ]



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " X(k) " e
j" 2" $ " k#s ( )" n
N
n= 0
N#1
%
k= 0
N#1
%



!
e
j" 2" # " k$s ( )" n
N
n= 0
N$1
%
=
0 k & s
N k = s
'
(
)



!
x(n) " e
# j" 2" $ " s" n
N
n= 0
N#1
%
=& " N " X(s)




!
x(n) =
1 0 " n " 4
0 5" n " 9
#
$
%


!
X(k) = x(n) " e
# j" 2" $ " k" n
N
n= 0
N#1
%
x(n) =
1
N
" X(k) " e
j" 2" $ " k" n
N
k= 0
N#1
%


!
W
N
= e
" j# 2# $
N



!
X(k) = x(n) "W
N
k" n
n= 0
N#1
$
x(n) =
1
N
" X(k) "W
N
#k" n
k= 0
N#1
$






Procesado Digital de Seales. 4 de Ingeniera Electrnica.
Profesor Emilio Soria. E.T.S.E. Universitat de Valncia.
Cuestiones a tener en cuenta con la DFT
Aqu tenamos dos problemas; por una parte tenemos un sumatorio innito; necesito conocer
TODA la secuencia x(n) desde para calcular dicha Transformada de Fourier.
Por otra parte, la Transformada de Fourier es funcin de una variable (w) que puede tomar
innitos valores (intervalo 0-2).
Las soluciones han sido por una parte restringir el tamao de x(n); este hecho tendr una
repercusin directa sobre la RESOLUCIN de la DFT. Se dene dicha resolucin como la
mnima frecuencia que la DFT puede discernir. De modo intuitivo si tengo una seal de
duracin 1 s la mnima frecuencia que podr diferenciar es 1 Hz (en ese intervalo temporal la
sinusoide completa con frecuencia mnima es la de 1 Hz). Otro efecto relacionado con el
anterior es el del enventanado que se comentar ms adelante.
Por otra parte, se ha muestreado la Transformada de Fourier lo que conduce a efectos
parecidos a los que se tena cuando se muestreaba a nivel temporal; por una parte la
repeticin de espetros (a nivel temporal) se traduce aqu en una repeticin de la seal
temporal. Esto adems conlleva a tener en cuenta los posible efectos del aliasing
temporal.
La Transformada de Fourier de una secuencia
discreta x(n) viene denida de la siguiente forma





!
X(k) = x(n) " e
# j" 2" $ " k" n
10
n= 0
9
%
= x(n) " e
# j" $ " k" n
5
n= 0
4
%
=
1#e
# j" $ " k
1#e
# j" $ " k
5
= e
# j" $ " 0.4" k
"
sen
$ " k
2
&
'
(
)
*
+
sen
$ " k
10
&
'
(
)
*
+



!
X(k) =
e
" j# $ # k
2
e
" j# $ # k
10
#
e
j# $ # k
2
"e
" j# $ # k
2
e
j# $ # k
10
"e
" j# $ # k
10
= e
" j# $ # 0.4# k
#
sen
$ # k
2
%
&
'
(
)
*
sen
$ # k
10
%
&
'
(
)
*


!
x n
( ) ( )
N
[ ]
= x n mdulo N
( ) [ ]


!
X e
j" w
( )
= x l
( )
" e
j" w" l
l=#$
$
%