The width of a signal is proportional to the
reciprocal of the width of its CTFT.
Recall Figure 1.
As
TT increases,
ft
f
t
widens and
Fiω
F
ω
narrows. In fact CTFT applied to
fat
f
a
t
yields
1aFiωa
1
a
F
ω
a
(and vice versa). This means that
ft
f
t
and
Fiω
F
ω
cannot be narrow simultaneously.
Let's make this mathematical:
- Definition 1: width
Define the width of a function as its
normalized standard deviation.
σ
f
2=1∥f∥2∫−∞∞t2|ft|2d
t
σ
f
2
1
f
2
t
t
2
f
t
2
(1)
σ
F
2=12π∥f∥2∫−∞∞ω2|Fiω|2d
ω
σ
F
2
1
2
f
2
ω
ω
2
F
ω
2
(2)
Figure 2.
σ
f
σ
F
≥12
σ
f
σ
F
1
2
(3)
In other words, the time width times the frequency width is
greater than or equal to cst.
Proof: use C.S.T.
The signal that has equality in bound (which means
that it is the most concentrated signal in both time and frequency
under this measure) is the Gaussian
(Figure 3 and Figure 4):
ft=
A
1
e−(αt2)
f
t
A
1
α
t
2
(4)
Fiω=
A
2
e−ω22α
F
ω
A
2
ω
2
2
α
(5)