Vous êtes sur la page 1sur 2

EE341 Communications Systems

Assignment 5 7 March 2014



Estimating essential bandwidth of pulses by numerical integration

What? This tutorial is about using numerical integration to estimate the bandwidth required to transmit
certain kinds of pulses. Parsevals theorem states that the energy of a signal is given by the area under the
energy spectral density curve of the signal. Mathematically this is written as,
2
2
( ) ( )
g
E g t dt G f df


= =
} }

Why? Bandwidth comes at a premium because all channels are band-limited and the traffic is ever-
increasing. Every communications system must therefore estimate the bandwidth required to transmit
various types of pulses. The essential bandwidth depends not only on the shape of each individual pulse
but also on the specific attributes of the so-called line code (pattern of logic transitions) that is used to
represent the digital bits.

Problems
1. Estimate the essential bandwidth W (in rad/s) of the signal ( ) ( )
at
g t e u t

= if the essential bandwidth


is said to contain 95% of the signal energy. This problem should be done by hand first and then compared
with the result from numerical integration.

2. Use the same definition as above to estimate the essential bandwidth W (in rad/s) of the rectangular
pulse
( )
t
g t
T
| |
=
|
\ .
[
. Obtain an expression for the Plot the fractional energy passed as a function of the
bandwidth. You need to use numerical integration for this problem.

3. Three popular line codes are the polar NRZ, bipolarNRZ and the Manchester NRZ. Here RZ and NRZ
stand for return-to-zero and non-return-to-zero respectively. Plot the following three power spectral
density functions on a single pair of axes and estimate the essential bandwidth in each case. Think about
what the comparison tells you.

(a)
2
2
sin
( )
b
polarNRZ b
b
fT
P f A T
fT
t
t
| |
=
|
\ .
(b) ( )
2
2
2
sin
( ) sin
4
b b
bipolarRZ b
b
A T fT
P f fT
fT
t
t
t
| |
=
|
\ .


(c) ( )
2
2 2
sin / 2
( ) sin / 2
/ 2
Manches
b
b ter Z
b
N b R
fT
P f A T fT
fT
t
t
t
| |
=
|
\ .

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Time
A
m
p
l i t u
d
e
Numerical integration algorithms
Our objective is to find the approximate value of the integral ( ) ( )
b
a
I f f x dx =
}
, where f(x) is a continuous function
in [a, b]. One would resort to numerical integration if f(x) is either too complicated to integrate by hand or if only a
sample of discrete values of f(x) are known from experimental observations of the process that is described by f(x).

a. Midpoint formula
A simple procedure to approximate I(f) can be devised by partitioning the interval [a, b] into sub-intervals I
k
=
[x
k-1
, x
k
] , k = 1,, M, with x
k
= a + kh, = 0,, M and h = (b-a)/M.
Since,
1
( ) ( )
k
M
k
I
I f f x dx
=
=

}
, on each sub-interval we can approximate the exact integral of f by that of a
polynomial f approximating f on
k
I . The simplest solution consists in choosing f as the constant
polynomial interpolating f at the middle point of
k
I , that is, at
1
2
k k
k
x x
x

+
= . In such a way we obtain the
composite midpoint quadrature formula:

1
( ) ( )
M
midpt k
k
I f h f x
=
=

(1)

b. Trapezoidal formula
Another formula can be obtained by replacing f on
k
I by the linear polynomial interpolating f at the nodes
x
k-1
and x
k
. This yields the so-called composite trapezoidal formula given by,
| | | |
1
1
1 1
( ) ( ) ( ) ( ) ( ) ( )
2 2
M M
trap k k k
k k
h h
I f f x f x f a f b h f x

= =
= + = + +

(2)

Using Matlab
Look up the following Matlab commands - quad, quadl and trapz.
The function you want to integrate needs to be defined as an anonymous function which is then passed
as an argument to one of the three integration functions.

Illustrative example problems
1. (Simple functions) Consider ( ) 2 sin(2 ) f x x = + . Calculate the approximate integral of f(x) over the
interval [1, 6]. Use the two numerical integration methods given below with 11 sample points to compute the
approximate integral. How do they compare?

2. (Demography) Consider the population of a very large number M of individuals. The distribution of their
height can be represented by a bell function characterized by the mean value h0 of the height and standard
deviation ,
2 2
0
( ) /(2 )
( )
2
h h
M
N h e
o
o t

=


Calculate the number of individuals whose height is between 1.8m and 1.9m for a group of 200 people with a
mean height of 1.7m and a standard deviation of 0.1m.

Vous aimerez peut-être aussi