3F4 Power and Energy Spectral

Density
Dr. I. J. Wassell

Power and Energy Spectral

Density
The power spectral density (PSD) Sx() for
a signal is a measure of its power
distribution as a function of frequency
It is a useful concept which allows us to
determine the bandwidth required of a
transmission system
We will now present some basic results
which will be employed later on

PSD
Consider a signal x(t) with Fourier
Transform (FT) X()
X ( )

j t
x
(
t
)
e
dt

We wish to find the energy and power

distribution of x(t) as a function of
frequency

Deterministic Signals
If x(t) is the voltage across a R=1 resistor,
the instantaneous power is,
( x (t )) 2
( x(t )) 2
R

Thus the total energy in x(t) is,

Energy

2
x
(
t
)
dt

From Parsevals Theorem,

Energy

X ( )

df

Deterministic Signals
So,
Energy

X ( )

df

X (2f )

df

E (2f )df

Where E(2f) is termed the Energy Density Spectrum

(EDS), since the energy of x(t) in the range fo to fo+fo is,
E ( 2f o )f o

Deterministic Signals
For communications signals, the energy is
effectively infinite (the signals are of unlimited
duration), so we usually work with Power quantities
We find the average power by averaging over time
lim 1 T 2
2
Average power
(
x
(
t
))
dt
T

T T T 2
Where xT(t) is the same as x(t), but truncated to zero outside
the time window -T/2 to T/2

Using Parseval as before we obtain,

Deterministic Signals
lim 1 T 2
2
Average power
(
x
(
t
))
dt
T

T T T 2

lim 1
2

X T (2f ) df

T T
2

lim X T (2f )

df
T
T

S x (2f )df

Where Sx() is the Power Spectral Density (PSD)

PSD
S x ( )

lim

X T ( )

T
T
The power dissipated in the range fo to fo+fo is,

S x ( 2f o )f o
And Sx(.) has units Watts/Hz

Wiener-Khintchine Theorem
It can be shown that the PSD is also given by
the FT of the autocorrelation function (ACF),
rxx(),

S x ( ) rxx ( )e j d

Where,

lim 1 2
rxx ( )
x(t ) x(t )dt

T T T
2

Random Signals
The previous results apply to deterministic signals
In general, we deal with random signals, eg the
transmitted PAM signal is random because the
symbols (ak) take values at random
Fortunately, our earlier results can be extended to
cover random signals by the inclusion of an extra
averaging or expectation step, over all possible
values of the random signal x(t)

PSD, random signals

2

lim E[ X T ( ) ]
S x ( )
T
T
Where E[.] is the expectation operator

The W-K result holds for random signals,

choosing for x(t) any randomly selected
realisation of the signal
Note: Only applies for ergodic signals where the time
averages are the same as the corresponding ensemble
averages

Linear Systems and Power Spectra

Passing xT(t) through a linear filter H() gives
the output spectrum,
YT ( ) H ( ) X T ( )
Hence, the output PSD is,
2

lim E[ YT ( ) ]
lim E[ H ( ) X T ( ) ]
S y ( )

T
T
T
T
2
E[ X T ( ) ]
2 lim
S y ( ) H ( )
T
T
2

S y ( ) H ( ) S x ( )