4.2. THE DISCRETE-TIME FOURIER TRANSFORM
To see this sampling effect, consider a finite-duration sequence x[n] and construct a periodic signal x ˜[n], for which x[n] is one period x ˜[n] 0 ≤ n ≤ N − 1 x[n] = 0 otherwise
Discrete-Time Fourier Transform The FT of an discrete-time signal (DTFT) is similarly defined as n=∞ X jω x[n]e−jωn (1) X(e ) =
The periodic signal x ˜[n] has FS coefficients, or the DFT, as 1 X x ˜[n]e−jk(2πn)/N ak = N
n=−∞
then the inverse discrete-time FT is Z 1 X(ejω )ejωn dω x[n] = 2π 2π
(2)
n=
Let ω0 = 2π/N , then these coefficients are scaled samples of the DTFT as
Notes: • The DTFT is periodic with the fundamental period of 2π
ak =
X(ejω ) = X(ej(ω+2π) )
1 X(ejkω0 ) N
• The DTFT is continuous in the frequency domain. This is different from the DFT (Discrete Fourier Transform, or discrete-time Fourier Series), which is discrete in both the time and frequency domains.
The FS representation of x ˜[n] becomes X 1 x ˜[n] = X(ejkω0 )ejkω0 n N k= X 1 X(ejkω0 )ejkω0 n ω0 = 2π
• The DFT is the sampled version of the DTFT in the frequency domain.
As N → ∞, ω0 → 0, and the above sum approaches the integral expression for x[n] in (2).
ES156 – Harvard SEAS
k=
1
ES156 – Harvard SEAS
2
Convergence Issues
Examples
Convergence of the infinite summation in the analysis equation is guaranteed if either x[n] is absolutely summable, or if the sequence has finite energy, i.e.,
• x[n] = an u[n], |a| < 1 has the DTFT jω
X(e ) = =
∞ X
an u[n]e−jωn
n=−∞ ∞ X
(ae−jω )n =
n=0
∞ X
1 . 1 − ae−jω
|x[n]| < ∞
n=−∞
or
1, |n| ≤ N 1 • The rectangular pulse s[n] = 0, |n| > N1
∞ X
|x[n]|2 < ∞
n=−∞
has the DTFT
jω
X(e ) =
N1 X
e−jωn = . . .
n=−N1
sin(ω(N1 + 12 )) = . sin(ω/2) This is the discrete-time counterpart to the sinc function, which appears as the Fourier Transform of the rectangular pulse. Unlike the continuous-time counterpart, the discrete-time sinc function is periodic with period 2π
ES156 – Harvard SEAS
3
ES156 – Harvard SEAS
4
• The periodic impulse train of period N, P∞ x[n] = k=−∞ δ[n − kN ] has Fourier coefficients ak = N1 for all k (check!) and so has the DTFT,
Fourier Transform for Periodic Signals The DTFT of a periodic signal is a periodic impulse train in the frequency domain. First, by substitution into the synthesis equation, we see that x[n] = e
jω0 n
jω
⇔ X(e ) =
∞ X
∞ X
¶ 2πk 2πak δ ω − X(e ) = N k=−∞ ¶ µ ∞ 2πk 2π X δ ω− = N N jω
2πδ(w − w0 − 2πl)
l=−∞
If we consider a periodic sequence x[n] with period N and Fourier series representation X ak ej(2π/N )n x[n] =
µ
k=−∞
k=
then its DTFT is given by, jω
X(e ) =
∞ X
k=−∞
2πak δ(ω −
2πk ). N
Examples: • The periodic signal x[n] = cos(ω0 n) = 12 ejω0 n + 12 e−jω0 n , with ω0 = 2π 5 has the DTFT, for −π ≤ ω < π, ¶ µ ¶ µ 2π 2π + πδ ω + , X(ejω ) = πδ ω − 5 5 which repeats periodically with period 2π. ES156 – Harvard SEAS
5
ES156 – Harvard SEAS
6
This is the duality between the discrete time Fourier transform and the continuous time Fourier series. Notice the similarities: Z 1 x[n] = X(ejω )ejωn dω 2π 2π ∞ X jω x[n]e−jωn X(e ) =
Properties of the DTFT The DTFT has properties analogous to the continuous-time FT. A few notable differences are as follows. ¡ ¢ ¡ ¢ • Periodicity: X ejω = X ej(ω+2π) • Differencing in time: F
n=−∞
¡
x[n] − x[n − 1] ←→ 1 − e • Accumulation: n X
x[k]
¢ −jω
¡
X e
compared to
¢ jω
x(t) =
←→
k=−∞
+πX(ej0 )
∞ X
1 ak = T
1 |x[n]| = 2π n=−∞
Z
T
2π
• Linearity:
¯ ¯ ¯X(ejω )¯2 dω
ax1 [n] + bx2 [n] ↔ aX1 (ejω ) + bX2 (ejω )
Note that all the energy is contained in one period of the DTFT. ¡ ¢ • Duality: Since X ejω is periodic, it has FS representation with coefficients given by x[−n]. ES156 – Harvard SEAS
x(t)e−jkω0 t dt
Other properties that are similar to the CTFT:
• Parseval’s relation: 2
Z
This property can be used to ease the computation of certain Fourier series coefficients.
δ(ω − 2πk)
k=−∞
∞ X
ak ejkω0 t
k=−∞
¡ jω ¢ 1 X e 1 − e−jω
F
∞ X
• Time and Frequency shifting: x[n − n0 ] ↔ e−jωn0 X(ejω ) ejω0 n x[n] ↔ X(ej(ω−ω0 ) ) 7
ES156 – Harvard SEAS
8
1 ↔ Y (e ) = 2π
• Conjugation and Conjugate Symmetry:
jω
x∗ [n] ↔ X ∗ (e−jω ) jω
Ev{x[n]} ↔
X1 (ejθ )X2 (ej(ω−θ) dθ 2π
Example: System input x[n] = (α)n u[n] and impulse response h[n] = (β)n u[n] lead to output y[n] = x[n] ∗ h[n], then by taking the DTFTs, we see that
jω
And if x[n] is real, this means X(e ) = X (e ). Furthermore, ∗
Z
Re{X(ejω )
Od{x[n]} ↔ jIm{X(ejω )
Y (ejω ) = X(ejω )H(ejω ) 1 1 · −jω 1 − βe 1 − αe−jω α/(α − β) −β/(α − β) = + 1 − αejω 1 − βejω α β (α)n u[n] − (β)n u[n] ⇒ y[n] = α−β α−β for α 6= β
• Time reversal: x[−n] ↔ X(e−jω ).
=
8