Scribd
Upload
69 views8 pages

Lect NB Noise

This document discusses narrowband noise representation. It begins by explaining how wideband noise is shaped into narrowband noise through bandpass filtering. It then presents equations to represent narrowband noise using quadrature components. These components have properties like being uncorrelated and having identical power spectral densities. The document uses this representation to analyze the performance of binary frequency-shift keying modulation in the presence of additive white Gaussian noise. It derives an expression for the average probability of error for this system.

Uploaded by

Ibra Nazla
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
69 views8 pages

Lect NB Noise

This document discusses narrowband noise representation. It begins by explaining how wideband noise is shaped into narrowband noise through bandpass filtering. It then presents equations to represent narrowband noise using quadrature components. These components have properties like being uncorrelated and having identical power spectral densities. The document uses this representation to analyze the performance of binary frequency-shift keying modulation in the presence of additive white Gaussian noise. It derives an expression for the average probability of error for this system.

Uploaded by

Ibra Nazla
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

Narrowband Noise Representation on Mac

28. Narrowband Noise Representation


In most communication systems, we are often dealing with band-pass filtering of signals.
Wideband noise will be shaped into bandlimited noise. If the bandwidth of the bandlimited
noise is relatively small compared to the carrier frequency, we refer to this as narrowband
noise. Figure 28.1 shows how to generate narrowband noise from wideband noise.
Figure 28.1 Generation of narrowband noise.
We can derive the power spectral density G n (f) and the auto-correlation function
R nn ( ) of the narrowband noise and use them to analyse the performance of linear
systems. In practice, we often deal with mixing (multiplication), which is a non-linear
operation, and the system analysis becomes difficult. In such a case, it is useful to express
the narrowband noise as
n(t) = x(t) cos 2fct - y(t) sin 2fct

(28.1)

where fc is the carrier frequency within the band occupied by the noise. x(t) and y(t)
are known as the quadrature components of the noise n(t). The Hibert transform of
n(t) is
n^ (t) = H[n(t)] = x(t) sin 2fct + y(t) cos 2fct

(28.2)

Proof.
The Fourier transform of n(t) is
1
1
1
1
N(f) = 2 X(f - fc) + 2 X(f+ fc) + 2 jY(f- fc) - 2 jY(f+ fc)
Let N^ ( f) be the Fourier transform of n^ ( t).
In the frequency domain,
N^ (f) = N(f)[-j sgn(f)]. We simply multiply all positive frequency components of
N(f) by -j and all negative frequency components of N(f) by j. Thus,
1
1
1
N^ (f) = -j 2 X(f-fc)+ j 2 X(f+ fc) - j 2 jY(f- fc) - j
1
1
1
N^ (f) = -j 2 X(f - fc) + j 2 X(f+ fc) + 2 Y(f- fc) +

1
2 jY(f+ fc)
1
2 Y(f+ fc)

and the inverse Fourier transform of N^ (f) is


n^ (t) = x(t) sin 2fct + y(t) cos 2fct
28.1

Q.E.D.

Narrowband Noise Representation on Mac

The quadrature components x(t) and y(t) can now be derived from equations (28.1)
and (28.2).
x(t) = n(t)cos 2fct + n^ (t)sin 2fct

(28.3)

y(t) = n(t)cos 2fct - n^ (t)sin 2fct

(28.4)

and

Given n(t), the quadrature components x(t) and y(t) can be obtained by using the
arrangement shown in Figure 28.2.
Figure 28.2 Generation of quadrature components of n(t).
x(t) and y(t) have the following properties:
1.
2.
3.
4.

E[x(t) y(t)] = 0. x(t) and y(t) are uncorrelated with each other.
x(t) and y(t) have the same means and variances as n(t).
If n(t) is Gaussian, then x(t) and y(t) are also Gaussian.
x(t) and y(t) have identical power spectral densities, related to the power
spectral density of n(t) by
Gx(f) = Gy(f) = Gn(f- fc) + Gn(f+ fc)

(28.5)

for fc - 0.5B < | f | < fc + 0.5B and B is the bandwidth of n(t).


Proof. (Under Construction)
Equation (28.5) is the key that will enable us to calculate the effect of noise on AM and FM
systems. It implies that the power spectral density of x(t) and y(t) can be found by
shifting the positive portion and negative portion of Gn(f) to zero frequency and adding
to give Gx(f) and Gy(f). This is shown in Figure 28.3.
Figure 28.3 (a) Power spectral density of n(t), (b) Power spectral density of x(t)
and y(t).
In the special case where G n (f) is symmetrical about the carrier frequency f c , the
positive- and negative-frequency contributions are shifted to zero frequency and added to
give

28.2

Narrowband Noise Representation on Mac

Gx(f) = Gy(f) = 2Gn(f- fc) = 2Gn(f+ fc)

(28.6)

Example 28.1
Given that the power spectral density of a narrowband Gaussian noise of variance 2 and
power N is

Gn(f) = N
2

1
2 2

(f fc )

2
2
e

2
+N
2

1
2 2

(f +fc )

2
2
e

where fc is the carrier frequency within the band occupied by the noise, then the power
spectral densities of the quadrature components of the noise are

Gx(f) = Gy(f) = 2Gn(f+ fc) = N

1
2 2

2
f
2
e 2

This is shown in Figure 28.4.


Figure 28.4 (a) Power spectral density of narrowband Gaussian noise n(t), (b) Power
spectral density of x(t) and y(t).
Performance of Binary FSK
Figure 28.5 Synchronous detection of binary FSK signals.
Consider the synchronous detector of binary FSK signals shown in Figure 28.5. In the
presence of additive white Gaussian noise (AWGN), w(t), the received signal is
r(t) = Acos 2fc1t + w(t)
where A is a constant and fc1 is the carrier frequency employed if a 1 has been sent. The
signals at the output of the band-pass filters of centre frequencies fc1 and fc2 are
r1(t) = Acos 2fc1t + n1(t)
and

28.3

Narrowband Noise Representation on Mac

r2(t) = n2(t)
where
n1(t) = x1(t) cos 2fc t - y1(t) sin 2fc t
1
1
and
n2(t) = x2(t) cos 2fc t - y2(t) sin 2fc t
2
2
are the narrowband noise. With appropriate design of low-pass filter and sampling period,
the sampled output signals are
vo1 = A + x1
vo2 = x2
and
v = A + [x1 - x2].
x1 and x2 are statistically independent Gaussian random variables with zero mean and
fixed variance 2 = N, where N is the power of the random variable. It can be seen
that one of the detectors has signal plus noise, the other detector has noise only.
When fc2 is the carrier frequency employed for sending a 0, the received signal is
r(t) = Acos 2fc2t + w(t).
It can be shown that
v = -A + [x1 - x2]
Since E [x 1 - x 2 ] 2 = E [x 1 ] 2 - 2E [x 1 x 2 ] 2 + E [x 2 ] 2 = E [x 1 ] 2 + E [x 2 ] 2 =
2 + 2 , the total variance t 2 = 2 2 . The two distributions of v are shown in
Figure 28.6.
Figure 28.6 Conditional probability density function.

28.4

Narrowband Noise Representation on Mac

The conditional probability density function of v assuming a 0 is sent is


(v + A)2
2
e 2 t

f(v/0) =

2 t2

and the probability of error given a 0 is sent is

= f(v/0)dv
0

P e0

(v + A)2
2
e 2 t dv

2
0 2 t

(28.7)

Similarly, the probability of error given a 1 is sent is

0
P e1

(v A)2
2
e 2 t dv

2
2 t

= P e0
Let p0 be the probability of sending a 0 and p1 be the probability of sending a 1. For
equally likely transmission of binary signals, we have p 0 = p 1 = 0.5. The average
probability of error is given by
Pe

= p0 Pe0 + p1 Pe1
= P e0

(v + A)2
and d u =
Let u = v + A . Then u 2 =
2 t
2 t2
equation (28.7), we get
Pe

A
u =
2t

1
2
= 2[

dv . Substituting u and du into


2 t

1 -u2
e
du

e-u2 du ]

(28.8)

28.5

Narrowband Noise Representation on Mac

Equation (28.8) becomes


Pe

1
A )
= 2 erfc(
2t
1
= 2 erfc( A )
2
1
= 2 erfc( A ).
2 N

Similarly, we can use this approach to derive the average probability of error for BASK
and BPSK systems. In the BASK system, the synchronous detector output consists of a
signal A plus noise or a noise alone. In the BPSK system, the synchronous detector
output consists of a polar signal + A plus noise. The results are summarised in Table 28.1
________________________________________________________________________
1
BPSK
P e = 2 erfc( A )
2N
1
BFSK
P e = 2 erfc( A )
2 N
1
BASK
P e = 2 erfc( A
)
2 2N
________________________________________________________________________
Table 28.1 Performance of various modulation systems.
References
[1]

J. G. Porakis and M. Salehi, Communication Systems Engineering, Prentice Hall,


1994. ISBN 0-13-158932-6.

[2]

M. S. Roden, Analog and Digital Communication Systems, 3/e, Prentice Hall,


1991, ISBN 0-13-033325-5.

[3]

H. Taub and D. L. Schilling, Principles of Communication Systems, 2/e,


MaGraw-Hill, 1996, ISBN 0-07-062955-2.

28.6

Narrowband Noise Representation on Mac

Wideband
Narrowband
noise
noise
Bandpass
G n (f )
filter

fc
-f c

0f

Frequency

Figure 28.1 Generation of narrowband noise.

x ( t )+ x ( t ) cos 4 f c t - y ( t ) sin 4 f c t

LPF

~
~
2n (t )

x (t )

cos 2 f c t
LPF

~
~

y (t )

-sin 2 f c t
Figure 28.2 Generation of quadrature components of n(t).

G x ( f ) = G y (f )
G x ( 0) = 2G n ( f c )

Gn ( f )

B
Frequency

Frequency

-f c

0
(a)

fc

(b)

Figure 28.3 (a) Power spectral density of n(t). (b) Power spectral density of x(t)
and y(t).

28.7

Narrowband Noise Representation on Mac

G n (f )

G x (f ) = G y (f )
G x ( 0) = 2G n ( f c )

Frequency

Frequency
0

-f c

fc

(b)

(a)

Figure 28.4 (a) Power spectral density of narrowband Gaussian noise n(t). (b) Power
spectral density of x(t) and y(t).

r ( t ) = A cos 2 f c t + n ( t )
1
1
1
v 01 ( t ) = A + x ( t )
1
BPF 1
LPF

~
~
~

fc
r ( t ) = A cos 2 f c t +
1
w (t )

~
~
Gain = 2

cos 2 f 1 t
LPF

BPF 2

~
~
~

fc

~
~

v 01 = A + x 1
+

v 02 = x 2

Gain = 2 v 02 ( t ) = x ( t )
2
cos 2 f 2 t

r ( t ) = n (t )
2
2

Figure 28.5 Synchronous detection of binary FSK signals.

1
2 t2

Probability density function


2
)

(v - A ) 2
1
f ( v /1) =
e
2 t2
2 t2

(v + A
2 t2 = f ( v /0)

-A

v
A

Figure 28.6 Conditional probability density function.

28.8

You might also like