Documentos de Académico
Documentos de Profesional
Documentos de Cultura
October 2, 2023 1 / 20
Theoretical Fourier Analysis
operations required for their evaluation. The presence of an infinite sum or integral prevents
exact numerical computation of the corresponding transform
Capı́tulo 3.4 Transformada discreta de Fourier
Direct transform Inverse transform Exact
(spectral analysis) (signal reconstruction) computation
N−1 N−1
1 " 2π " 2π
DTFS c̃k = x̃[n]e− j N kn x̃[n] = c̃k e j N kn yes
N
n=0 k=0
finite summation finite summation
∞
" ! 2π
1
DTFT X̃(e j! ) = x[n]e− j!n x[n] = X̃(e j! )e j! dω no
2π 0
n=−∞
infinite summation integration
! T0 ∞
"
1
CTFS ck = x̃c (t)e− jk!0 t dt x̃c (t) = ck e jk!0 t no
T0 0
k=−∞
integration infinite summation
! ∞ ! ∞
1
CTFT Xc ( j!) = xc (t)e− j!t dt xc (t) = Xc ( j!)e j!t d! no
−∞ 2π −∞
integration integration
October 2, 2023 2 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
Cómo calculamos numéricamente las transformadas?
Para calcular la TFTD se requiere
−1
I usar un conjunto finito de valores {x[n]}N
0
N
X −1 ∞
X
X̃N (ejω ) = x[n]e−jωn = xN [n]e−jωn
n=0 n=−∞
October 2, 2023 3 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
Cada valor de la TFTD finita es una suma de N productos
N
X −1
X̃N (ejω ) = x[n]e−jωn
n=0
X̃N (e jωk
) = x[0]ejωk 0 + · · · + x[N − 1]ejωk (N −1) = X[k]
Propiedades
(WN−k )N = (ej2πk/N )N = ej2πk = 1, k= 0, 1, ...N − 1
m(N −1) N, si WNm = 1
1 + WNm + WN2m + · · · + WN =
0, si WNm 6= 1
Ortogonalidad
N −1 N −1
X
j2πk(m−n)/N
X k(m−n) N, si m − n = rN
e = WN =
0, para otro valor
k=0 k=0 October 2, 2023 5 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
N
X −1 N
X −1
X[k] = x[n]e−j2πkn/N = x[n]WNkn , k = 0, 1, ..., N − 1
n=0 n=0
October 2, 2023 6 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
k = 0, 1, ..., N − 1 n = 0, 1, ..., N − 1
October 2, 2023 7 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
El factor de giro es periódico con respecto a k y n
(k+N )n k(n+N )
WN = WNkn WN = WNkn
? ?
0 N−1 n
DFT
IDFT
... ...
0 N−1 n
October 2, 2023 9 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
October 2, 2023 10 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
Si x[n] = 0 para n ∈
/ [0, L − 1] y N ≥ L → x̃[n] = x̃N [n]
October 2, 2023 11 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
e 4
(a)
IDTFT 2
n 0
0 L 0
4
(b)
IDFT 2
N =16
n 0
0 16 0
(c)
4
IDFT 2
N =8
n 0
0 8 0
Figure 7.4 (a) Finite-length (L = 12) sequence x[n] and its DTFT. (b) Periodic
October replication
2, 2023 of 12 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
Qué pasa con una secuencia infinita?
TFTD 1
x[n] = an u[n], 0 < a < 1 ←→ X̃(ejω ) =
1 − ae−jω
Si se muestrea la TFTD se tiene una señal periódica en el tiempo
∞
TFTD
X
x̃[n] = x[n − lN ] ←→ X̃(ej2πk/N )
l=∞
Puesto que conocemos x[n] podemos calcular el valor de x̃[n] exacto ya
que las replicas de x[n − lN ] para l > 1 no afectan el valor de x̃[n] en el
rango 0 ≤ n ≤ N − 1
0 ∞
X X an x[n]
x̃[n] = an−lN = an (aN )l = N
= , 0≤n≤N −1
1−a 1 − aN
l=−∞ l=0
y reconstruir
N −1
ˆ X 1 1 − aN e−jωN
jω
X̃(e ) = x̃[n]e−jωn = −→ X̃(ejω )
1 − aN 1 − ae−jω N →∞
n=0
October 2, 2023 13 / 20
8 The Discrete Fourier Transform
8
Magnitude
Signal Amplitudes
1.2
1
Amplitude
0.8
0.6
0.4
0 5 10 15
Time index (n)
4
DTFT
(a)
0
0 3 0
4
DFT
(b) 4
0
0 3 0 1 2 3
4
DFT
(c) 8
0
0 3 7 0 1 2 3 4 5 6 7
4
DFT
(d) 16
0
0 3 15 0 4 8 12 15
Figure 7.6 (a) A finite-length (N = 4) rectangular pulse x[n] and the magnitude of its DTFT.
(b)–(d) Magnitude of the K-point DFT of x[n] for K = 4, 8, and 16, obtained by using October 2, 2023 15 / 20
The Discrete Fourier Transform
Capı́tulo 3.4 Transformada discreta de Fourier
8
6
Magnitude
4
64-point DFT 8-point DFT
2
0
0
2
8-point DFT
Phase in units
64-point DFT
−2
0
October 2, 2023 16 / 20
Since ω = !T, the N-point DFT X[k] is obtained by sampling the DTFT X(e ) at ω =
2πk/N or X̃(e j!T ) at ! = 2πk/NT for 0 ≤ k ≤ N − 1. The result is
Capı́tulo 3.4 Transformada
1 !
"
2πk
discreta
2π
# ∞de Fourier
X[k] = Relación j
Xc entre− TFTC
j m . yk = 0, 1, . . . , N − 1
TDF (7.74)
T m=−∞ NT T
c c
CTFT
0 t 0
Sampling Periodization
c e c
DTFT
0 nT 0
Periodization Sampling
DTFS
DFT
c c
N
DFT
0 N n 0 N k
October 2, 2023 17 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
376 The Discrete Fourier Transform
Representaciones alternativas
0 n
(a)
Blue
0 N n
(b)
0 1 2 n
(c) (d)
October 2, 2023 18 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
The N
La operación ”n módulo n ”Modulo N Operation
entrega una relación entre el indice temporal
n y el circular r
n = l + N r, 0 ≤ r ≤ N − 1 → hniN = ”n módulo N ” = r
Es decir dado un n y un N se escoge un l tal que r ∈ [0, N − 1].
October 2, 2023 19 / 20
Capı́tulo 3.4 Transformada discreta de Fourier
La operación z[n] = x[−n] no se puede realizar cuando x[n] no esta
definida fuera del rango 0 ≤ n ≤ N − 1 → periodicidad inherente de la
377 7.4 Properties of the Discrete Fourier Transform
TDF
z[n] = x[ n]
Figure 7.10 Circular folding using periodic extension, linear time-reversal, and extraction of
the primary period.
Doblamiento circular
MATLAB
z[n] = x[h−nifunction
N] y=circfold(x,N) function y=circfold(x,N)
% Circular time reversal (folding) % Circular time reversal (folding)
if length(x) > N; error(’N < length(x)’); end
x[0], n=0
if length(x) > N; error(’N < length(x)’); end
, x=[x zeros(1,N-length(x))]; x=[x zeros(1,N-length(x))];
− n], 1 ≤ n ≤ N − 1
x[Nn=(0:1:N-1); n=(0:1:N-1);
y=x(mod(-n,N)+1); y=x(mod(-n,N)+1);
October 2, 2023 20 / 20