Anda di halaman 1dari 14

13.

Power Spectrum
For a deterministic signal x(t), the spectrum is well defined: If X (ω )
represents its Fourier transform, i.e., if
+∞
X (ω ) = ∫ −∞ x(t )e − jω t dt , (13-1)
then | X (ω ) |2 represents its energy spectrum. This follows from
Parseval’s theorem since the signal energy is given by
+∞ +∞
∫ −∞ x (t )dt = 1
∫ −∞ ω dω = E.
2 2
| X ( ) | (13-2)

Thus | X (ω ) |2 ∆ω represents the signal energy in the band (ω , ω + ∆ω )
(see Fig 13.1).
| X (ω )|2
X (t ) Energy in (ω ,ω +∆ω )

0 t 0 ω
ω ω + ∆ω
Fig 13.1
1
However for stochastic processes, a direct application of (13-1)
generates a sequence of random variables for every ω . Moreover,
for a stochastic process, E{| X(t) |2} represents the ensemble average
power (instantaneous energy) at the instant t.

To obtain the spectral distribution of power versus frequency for


stochastic processes, it is best to avoid infinite intervals to begin with,
and start with a finite interval (– T, T ) in (13-1). Formally, partial
Fourier transform of a process X(t) based on (– T, T ) is given by
T
X T (ω ) = ∫ −T X (t )e − jω t dt (13-3)
so that
| X T (ω ) |2 1 T 2


− jω t
= −T
X (t )e dt (13-4)
2T 2T
represents the power distribution associated with that realization based
on (– T, T ). Notice that (13-4) represents a random variable for every
ω , and its ensemble average gives, the average power distribution
based on (– T, T ). Thus 2
⎧ | X T (ω ) |2 ⎫ 1 T T
∫− T ∫− T
− jω ( t1 − t2 )
PT (ω ) = E ⎨ ⎬ = E { X ( t1 ) X *
( t 2 )}e dt1dt2
⎩ 2T ⎭ 2T
1 T T
∫ − T ∫− T XX 1 2
− jω ( t1 − t2 )
= R ( t , t ) e dt1dt2 (13-5)
2T
represents the power distribution of X(t) based on (– T, T ). For wide
sense stationary (w.s.s) processes, it is possible to further simplify
(13-5). Thus if X(t) is assumed to be w.s.s, then RXX (t1 , t2 ) = RXX (t1 − t2 )
and (13-5) simplifies to
1 T T
∫ −T ∫ −T XX 1 2
− jω ( t1 − t2 )
PT (ω ) = R ( t − t ) e dt1dt2 .
2T
Let τ = t1 − t2 and proceeding as in (14-24), we get
1 2T

− jωτ
PT (ω ) = − 2T
R XX
(τ ) e (2T − | τ |)dτ
2T
2T
= ∫ − 2T RXX (τ )e − jωτ (1 − 2|τT| )dτ ≥ 0 (13-6)
to be the power distribution of the w.s.s. process X(t) based on
(– T, T ). Finally letting T → ∞ in (13-6), we obtain 3
+∞
S XX (ω ) = lim PT (ω ) = ∫−∞ RXX (τ )e − jωτ dτ ≥ 0 (13-7)
T →∞

to be the power spectral density of the w.s.s process X(t). Notice that
RXX (ω ) ←⎯→
F⋅T
S XX (ω ) ≥ 0. (13-8)
i.e., the autocorrelation function and the power spectrum of a w.s.s
Process form a Fourier transform pair, a relation known as the
Wiener-Khinchin Theorem. From (13-8), the inverse formula gives
+∞
∫−∞ XX
jωτ
RXX (τ ) = 1
2π S (ω ) e dω (13-9)

and in particular for τ = 0, we get


+∞
1
∫−∞ XX ω ω = = } = P,
2
2π S ( ) d R XX
(0) E {| X ( t ) | the total power.
(13-10)
From (13-10), the area under S XX (ω ) represents the total power of the
process X(t), and hence S XX (ω ) truly represents the power
spectrum. (Fig 13.2). 4
S XX (ω ) S XX (ω ) ∆ω represents the power
in the band (ω , ω + ∆ω )

ω ω + ∆ω ω
0

Fig 13.2
The nonnegative-definiteness property of the autocorrelation function
in (14-8) translates into the “nonnegative” property for its Fourier
transform (power spectrum), since from (14-8) and (13-9)
n n n n +∞
∑∑ ai a RXX (ti − t j ) = ∑∑ ai a 21π
jω ( ti − t j )
∫−∞ S (ω )e dω
* *
j j XX
i =1 j =1 i =1 j =1

+∞ 2
∫−∞ S (ω ) ∑ i =1 ai e jω ti
d ω ≥ 0.
n
= 21π XX
(13-11)
From (13-11), it follows that
RXX (τ ) nonnegative - definite ⇔ S XX (ω ) ≥ 0. (13-12)
5
If X(t) is a real w.s.s process, then RXX (τ ) = RXX (−τ ) so that
+∞
S XX (ω ) = ∫ −∞ RXX (τ )e − jωτ dτ
+∞
= ∫ −∞ RXX (τ ) cos ωτ dτ

= 2 ∫ 0 RXX (τ ) cos ωτ dτ = S XX ( −ω ) ≥ 0 (13-13)
so that the power spectrum is an even function, (in addition to being
real and nonnegative).

6
Power Spectra and Linear Systems

If a w.s.s process X(t) with autocorrelation


function RXX (τ ) ↔ S XX (τ ) ≥ 0 is X(t) h(t) Y(t)
applied to a linear system with impulse
response h(t), then the cross correlation Fig 13.3
function RXY (τ ) and the output autocorrelation function RYY (τ ) are
given by (14-40)-(14-41). From there
RXY (τ ) = RXX (τ ) ∗ h* (−τ ), RYY (τ ) = RXX (τ ) ∗ h* (−τ ) ∗ h(τ ). (13-14)
But if
f (t ) ↔ F (ω ), g (t ) ↔ G (ω ) (13-15)
Then
f (t ) ∗ g (t ) ↔ F (ω )G (ω ) (13-16)
since
+∞
F { f (t ) ∗ g (t )} = ∫ −∞ f (t ) ∗ g (t )e − jω t dt
7
+∞
F { f (t ) ∗ g (t )}= ∫ −∞ {∫ +∞
−∞ }
f (τ ) g (t − τ )dτ e − jω t dt
+∞ +∞
= ∫ −∞ f (τ )e − jωτ
dτ ∫ −∞ g (t − τ )e − jω ( t −τ ) d (t − τ )
=F (ω )G (ω ). (13-17)
Using (13-15)-(13-17) in (13-14) we get
S XY (ω ) = F {RXX (ω ) ∗ h* ( −τ )} = S XX (ω ) H * (ω ) (13-13)

since

( )
+∞ +∞ *

∫ −∞ h (−τ )e ∫ −∞ h(t )e
− jωτ − jω t
*
dτ = dt = H * (ω ),
where
+∞
H (ω ) = ∫ −∞ h(t )e − jω t dt (13-19)

represents the transfer function of the system, and


SYY (ω ) = F {RYY (τ )} = S XY (ω ) H (ω )
= S XX (ω ) | H (ω ) |2 . (13-20)
8
From (13-13), the cross spectrum need not be real or nonnegative;
However the output power spectrum is real and nonnegative and is
related to the input spectrum and the system transfer function as in
(13-20). Eq. (13-20) can be used for system identification as well.

W.S.S White Noise Process: If W(t) is a w.s.s white noise process,


then from (14-43)
RWW (τ ) = qδ (τ ) ⇒ SWW (ω ) = q. (13-21)
Thus the spectrum of a white noise process is flat, thus justifying its
name. Notice that a white noise process is unrealizable since its total
power is indeterminate.
From (13-20), if the input to an unknown system in Fig 13.3 is
a white noise process, then the output spectrum is given by
SYY (ω ) = q | H (ω ) |2 (13-22)
Notice that the output spectrum captures the system transfer function
characteristics entirely, and for rational systems Eq (13-22) may be
used to determine the pole/zero locations of the underlying system.
9
Example 13.1: A w.s.s white noise process W(t) is passed
through a low pass filter (LPF) with bandwidth B/2. Find the
autocorrelation function of the output process.
Solution: Let X(t) represent the output of the LPF. Then from (13-22)
⎧ q, | ω |≤ B / 2
S XX (ω ) = q | H (ω ) | = ⎨
2
. (13-23)
⎩0, | ω |> B / 2
Inverse transform of S XX (ω ) gives the output autocorrelation function
to be
RXX (τ ) = ∫ − B / 2 S XX (ω )e jωτ d ω = q ∫− B / 2 e jωτ d ω
B/2 B/2

sin( Bτ / 2)
= qB = qB sinc( Bτ / 2) (13-24)
( Bτ / 2) R (τ )
XX

| H (ω )|2 qB
1

ω τ
−B /2 B/2

(a) LPF (b)


Fig. 13.4 10
Eq (13-23) represents colored noise spectrum and (13-24) its
autocorrelation function (see Fig 13.4).
Example 13.2: Let
1 t +T
Y (t ) = ∫
2T t −T
X (τ )dτ (13-25)

represent a “smoothing” operation using a moving window on the input


process X(t). Find the spectrum of the output Y(t) in term of that of X(t).
h (t )
1 / 2T
Solution: If we define an LTI system
with impulse response h(t) as in Fig 13.5, t
−T T
then in term of h(t), Eq (13-25) reduces to Fig 13.5
+∞
Y (t ) = ∫ −∞ h(t − τ ) X (τ )dτ = h(t ) ∗ X (t ) (13-26)
so that
SYY (ω ) = S XX (ω ) | H (ω ) |2 . (13-27)
Here
+T
H (ω ) = ∫ −T 1
2T e − jωt dt = sinc(ωT ) (13-28)
11
so that
SYY (ω ) = S XX (ω ) sinc 2 (ω T ). (13-29)
S XX ( ω ) sinc 2 (ω T ) S YY ( ω )

ω ω ω
π
T

Fig 13.6
Notice that the effect of the smoothing operation in (13-25) is to
suppress the high frequency components in the input (beyond π / T ),
and the equivalent linear system acts as a low-pass filter (continuous-
time moving average) with bandwidth 2π / T in this case.

12
Discrete – Time Processes
For discrete-time w.s.s stochastic processes X(nT) with
autocorrelation sequence {rk }+∞
−∞ ,
(proceeding as above) or formally
defining a continuous time process X (t ) = ∑ n X (nT )δ (t − nT ), we get
the corresponding autocorrelation function to be
+∞
RXX (τ ) = ∑ rk δ (τ − kT ).
k =−∞
Its Fourier transform is given by
+∞
S XX (ω ) = ∑ rk e − jωT ≥ 0, (13-30)
k =−∞

and it defines the power spectrum of the discrete-time process X(nT).


From (13-30),
S XX (ω ) = S XX (ω + 2π / T ) (13-31)
so that S XX (ω ) is a periodic function with period

2B = . (13-32)
T 13
This gives the inverse relation
1 B

jkωT
rk = S
− B XX
(ω ) e dω (13-33)
2B
and
1 B
r0 = E{| X ( nT ) | } = ∫ S (ω )d ω
2
(13-34)
− B XX
2B
represents the total power of the discrete-time process X(nT). The
input-output relations for discrete-time system h(nT) in (14-65)-(14-67)
translate into
S XY (ω ) = S XX (ω ) H * (e jω ) (13-35)
and
SYY (ω ) = S XX (ω ) | H (e jω ) |2 (13-36)
where
+∞

H (e ) = ∑ h( nT ) e − jω nT (13-37)
n =−∞
represents the discrete-time system transfer function. 14