Professional Documents
Culture Documents
{1} transmitted at a rate of 1/T bits/sec. He also determines the optimum pulse shape that was band limited to W Hz
and maximized the bit rate under the constraint that the pulse cause no Inter Symbol Interference at the sampling time.
His studies led him to conclude that the maximum pulse rate is 2W pulses/s. The rate is called Nyquist rate. Moreover
this pulse rate can be achieved by using the pulses g(t)=sinc(t). Nyquist result is equivalent to a version of the
sampling theorem of band limited signals. The sampling theorem states that a signal of bandwidth W can be
reconstructed from samples taken at Nyquist rate of 2W samples/s using interpolation formula.
In light of Nyquist work, Hartley considered the issue of the amount of the data that can be transmitted when multiple
amplitude levels are used.
Another significant advancement of communication was the work of Wiener who considered the problem of estimating
a desired signal waveform s(t) in the presence of additive noise n(t), based on the observation of the received signal
r(t)=s(t)+n(t). Wiener determined the linear filter whose output is the best mean-square approximation of the desired
signal. The resulting filter is called Optimum filter.
Shanon (1948) formulated the basic problem of reliable transmission of information in statistical terms.
The fundamental limitations of information transmission by electrical means are bandwidth and noise.
Taking both limitations into account, Shannon (1948)i stated that the rate of information transmission cannot
exceed the channel capacity. C = B log (1 + S/N)
This relationship, known as the Hartley-Shannon law, sets an upper limit on the performance of a communication
system with a given bandwidth and signal-to- noise ratio.
If the information rate is less than C, then it is theoretically possible to achieve reliable(error free) transmission through
the channel by appropriate coding. On the other hand regardless of signal processing used at the transmitter and
receiver if information rate > c then reliable transmission is not possible.
Hamming in his classic work devised error-detecting and error-correcting codes to combat channel noise.
The increase in demand for data transmission coupled with development of VLSI , has led to development of very
efficient and more reliable digital communication system.
Distortion and Interference
Distortion is waveform perturbation caused by imperfect response of the sys-tem to the desired signal itself. Unlike
noise and interference, distortion disappears when the signal is turned off. If the channel has a linear but distorting
response, then distortion may be corrected, or at least reduced, with the help of special filters called equalizers.
Page 1 A.Sarkar,ECE,JGEC,
Interference is contamination by extraneous signals from human sources- other transmitters, power lines and
machinery, switching circuits, and so on. Interference occurs most often in radio systems whose receiving antennas
usually intercept several signals at the same time. Radio-frequency interference (WI) also appears in cable systems if
the transmission wires or receiver circuitry pick up signals radiated from nearby sources. Appropriate filtering removes
interference to the extent that the interfering signals occupy different frequency bands than the desired signal.
One-way or simplex (SX) transmission. Two-way communication, of course, requires a transmitter and receiver at
each end. A full-duplex (FDX) system has a channel that allows simultaneous transmission in both directions. A halfduplex (HDX) system allows trans- mission in either direction but not at the same time.
Why is noise unavoidable? Rather curiously, the answer comes from kinetic theory. At any temperature above absolute
zero, thermal energy causes microscopic particles to exhibit random motion. The random motion of charged particles
such as electrons generates random currents or voltages called thermal noise. There are also other types of noise, but
thermal noise appears in every communication system.
Modulation Benefits
The primary purpose of modulation in a communication system is to generate a modulated signal suited to the
characteristics of the transmission channel. Actually, there are several practical benefits and applications of modulation
briefly discussed below.
Modulation for Efficient Transmission: Signal transmission over appreciable distance always involves a traveling
electromagnetic wave, with or without a guiding medium. The efficiency of any particular transmission method
depends upon the frequency of the signal being transmitted. By exploiting the frequency-translation property of CW
Modulation, message information can be impressed on a carrier whose frequency has been selected for the desired
transmission method.
As a case in point, efficient Line-of-sight ratio propagation requires antennas whose physical dimensions are at
least 1/10 of the signal's wavelength. Un-modulated transmission of an audio signal containing frequency components
down to 100 Hz would thus call for antennas some 300 km long. Modulated transmission at 100 MHz,
as in FM broadcasting, allows a practical antenna size of about one meter. It Likewise follows that signals with large
bandwidth should be modulated on high-frequency carriers. Since information rate is proportional to bandwidth,
according to the Hartley-Shannon law, we conclude that a high information rate requires a high carrier
frequency. For instance, a 5 GHz microwave system can accommodate 10,000 times as much information in a given
time interval as a 500 Hz radio channel. Going even higher in the electromagnetic spectrum, one optical laser beam has
a bandwidth potential equivalent to 10 million TV channels.
Modulation to Reduce Noise and Interference: A brute-force method for combating noise and interference is to increase
the signal power until it overwhelms the contaminations. But increasing power is costly and may damage equipment.
This property is called wideband noise reduction because it requires the trans-mission bandwidth to be much greater
than the bandwidth of the modulating signal. Wideband modulation thus allows the designer to exchange increased
bandwidth for decreased signal power, a trade-off implied by the Hartley-Shannon law. Note that a higher carrier
frequency may be needed to accommodate wideband modulation.
Modulation for Frequency Assignment: When you tune a radio or television set to a particular station, you are selecting
one of the many signals being received at that time. Since each station has a different assigned carrier frequency, the
desired signal can be separated from the others by filtering. Were it not for modulation, only one station could
broadcast in a given area; otherwise, two or more broadcasting stations would create a hopeless jumble of interference.
Modulation for Multiplexing: Multiplexing is the process of combining several signals for simultaneous transmission
on one channel. Frequency-division multiplexing (FDM) uses CW modulation to put each signal on a different carrier
frequency, and a bank of filters separates the signals at the destination. Time-division multiplexing (TDM) uses pulse
modulation to put samples of different signals in non-overlapping time slots.
The gaps between pulses could be filled with samples from other signals. A switching circuit at the destination then
separates the samples for signal reconstruction.
Page 2 A.Sarkar,ECE,JGEC,
Fourier analysis
Abstract: The Fourier Series and its applications to the Communication are discussed. The tutorial is written in a colloquial
style to avoid intimidating readers who are of a lesser level of intelligence
History:
Jean Baptiste Joseph Fourier (1768-1830) studied the mathematical theory of heat conduction in his major work, The Analytic
Theory of Heat. He established the partial differential equation governing heat diffusion and solved it using an infinite series of
trigonometric functions. The description of a signal in terms of elementary trigonometric functions had a profound effect on the
way signals are analyzed. The Fourier method is the most extensively applied signal-processing tool. The Fourier transform of a
signal lends itself to easy interpretation and manipulation, and leads to the concept of frequency analysis.
Fourier was interested in heat propagation, and presented a paper in 1807 to the Institute de France on the use of sinusoids to
represent temperature distributions. The paper contained the controversial claim that any continuous periodic signal could be
represented as the sum of properly chosen sinusoidal waves. Among the reviewers were two of history's most famous
mathematicians, Joseph Louis Lagrange (1736-1813), and Pierre Simon de Laplace (1749-1827).
While Laplace and the other reviewers voted to publish the paper, Lagrange adamantly protested. For nearly 50 years, Lagrange
had insisted that such an approach could not be used to represent signals with corners, i.e., discontinuous slopes, such as in square
waves. The Institute de France bowed to the prestige of Lagrange, and rejected Fourier's work. It was only after Lagrange died
that the paper was finally published, some 15 years later.
Who was right? It's a split decision. Lagrange was correct in his assertion that a summation of sinusoids cannot form a signal with
a corner. However, you can get very close. So close that the difference between the two has zero energy. In this sense, Fourier was
right, although 18th century science knew little about the concept of energy. This phenomenon now goes by the name: Gibbs
Effect,
Page 3 A.Sarkar,ECE,JGEC,
Page 4 A.Sarkar,ECE,JGEC,
For harmonically related sinusoids the integration interval can be taken over one period. Similar equations can be derived for the
product of cosines, or sine and cosine, of different frequencies. Orthogonality implies that the sinusoidal basis functions are
independent and can be processed independently. For example in a graphic equalizer we can change the relative amplitudes of one
set of frequencies, such as the bass, without affecting other frequencies
e jwot
Why are sinusoids used instead of, for instance, square or triangular waves?
Remember, there are an infinite number of ways that a signal can be decomposed. The goal of decomposition is to end up with
something easier to deal with than the original signal. For example, impulse decomposition allows signals to be examined one
point at a time, leading to the powerful technique of convolution. The component sine and cosine waves are simpler than the
original signal because they have a property that the original signal does not have: sinusoidal fidelity. A sinusoidal input to a
system is guaranteed to produce a sinusoidal output. Only the amplitude and phase of the signal can change; the frequency and
wave shape must remain the same. Sinusoids are the only waveforms that have this useful property. While square and triangular
decompositions are possible, there is no general reason for them to be useful.
Page 5 A.Sarkar,ECE,JGEC,
The integrals of certain trigonometric functions (evaluated from zero to twice pi, the period of sinusoidal functions) turn out to be
zero. (their inner products are zero).
Thus by the omnipotent laws of Linear Algebra, the function vectors are orthogonal and can span a space of periodic functions.
This means that any element in the function space can be written as a linear combination of these basis vector functions.
Now that we have established the concept of a set of functions that are orthogonal to one another we are ready to approach the
Fourier Series. It is a fundamental concept in linear algebra that a given set of orthogonal vectors span a space that has a
dimension equal to the number of vectors in the set. For instance, if we have three mutually orthogonal vectors we can think of
them as spanning Cartesian 3-space. From this follows the concept that any object in Cartesian 3-space can be described as a
linear combination of these basis vectors. The coefficients of this linear combination are regarded as that objects coordinates. The
set of orthogonal functions that compose the Fourier series can be thought of as spanning the space of periodic functions. Thus,
analogous to the vectors that span 3-space, this means that a periodic function can be described as a linear combination of the
Fourier basis functions.
It should be noted that not all periodic functions are included in this span. A periodic
function must satisfy three criteria known as the Dirichlet conditions. These conditions are as follows:
1. f(t) is piecewise continuous.
2. f(t) has isolated maxima and minima.
3. f(t) is absolutely integrable over a period.
Summary:Fourier Series
It is a representation of a function f(t) by the linear combination of elements of infinite mutual orthogonal functions.
Mutual orthogonal function:Two real time functions are said to be mutual orthogonal over an interval t1 and t2, if the integral of their product over
this integral is zero .i.e. f(t) is orthogonal to h(t) if
t2
t1
f (t )h(t ) = 0
Page 6 A.Sarkar,ECE,JGEC,
a0
f ( x) =
+ a n cos(nx) + bn sin( nx)
2 n =1
n =1
a0 =
an =
bn =
f ( x)dx
f ( x) cos(nx)dx
f ( x) sin(nx)dx
Fourier Series:
Any periodic function can be expressed as the sum of a series of sines and cosines (of varying amplitudes)
f ( x) =
a0 =
an =
bn =
a0
+ a n cos(nx) + bn sin( nx)
2 n =1
n =1
f ( x)dx
f ( x) cos(nx)dx
f ( x) sin(nx)dx
Page 7 A.Sarkar,ECE,JGEC,
e jwot
The set of exponential signals in above equation are periodic with a fundamental frequency
is the fundamental frequency. These signals form the set of basis functions for the
Fourier analysis. Any linear combination of these signals of the form
is also periodic with a period of T0. Conversely any periodic signal x(t) can be synthesized from a linear combination of
harmonically related exponentials.
The Fourier series representation of a periodic signal are given by the following synthesis and analysis equations:
Page 8 A.Sarkar,ECE,JGEC,
Page 9 A.Sarkar,ECE,JGEC,
The Fourier synthesis and analysis equations for non-periodic signals, known as the Fourier transform pair, are given by
Page 10 A.Sarkar,ECE,JGEC,
Page 11 A.Sarkar,ECE,JGEC,
Page 12 A.Sarkar,ECE,JGEC,
The Laplace transform is particularly useful in solving linear ordinary differential equations as it can transform relatively difficult
differential equations into relatively simple algebraic equations. The Laplace transform of x(t) is given by the integral
Page 13 A.Sarkar,ECE,JGEC,
Convolution:- The convolution f(t) of two time functions f1(t) and f2(t), is defined by the following integral
f (t ) =
f 1( ) f 2(t )d
It is a mathematical operation and is useful for describing the input/output relationship in a linear time invariant
system.
Steps to perform convolution graphically
1> f1() is the first function , where an independent variable (t) is replaced by a dummy variable .
2> f2(-) is the mirror image of f2().
3> f2(t-) represents the function f2(-) shifted to the right by t seconds.
4> For a particular value of t=b , integration of the product f1()f2(b-) represents the area under the product
curve(common area). This common area represents the convolution of f1(t) and f2(t) for a shift of t=b
5> The procedure is repeated for different values of t to evaluate the convolution.
6> The value of convolution obtained at different values of t may be plotted on a graph.,
Page 14 A.Sarkar,ECE,JGEC,
Page 15 A.Sarkar,ECE,JGEC,
Page 16 A.Sarkar,ECE,JGEC,
and
Proof:- F[f1(t)*f2(t)]=
hence
f 1(t ) f 2(t )
f 1(t )e jw d F 2( w) = F1( w) F 2( w)
and
Page 17 A.Sarkar,ECE,JGEC,
a1f1(t)+a2f2(t)a1r1(t)+a2r2(t)
Time Invariant system:-if a time shift in the input causes a equal time shift in the output then the system is time
invariant.
Causal System:- A system where the response does not begin before the input function is applied is known as
causal system. In other words , the value of the output ,r(t) at any instant t=t0 depends only on the values on the
input f(t) for tto.
The unit impulse response h(t) of a causal system is also causal i.e. h(t)=0 for t<0.
Causal system is physically realizable and they are operating in the real world. A physically realizable system
cannot have a response before the driving function or excitation applied. Pauley-wiener criterion gives the physical
realizable properties in frequency domain.
System function or transfer function:-The response r(t) of a linear system to given input f(t) can be determined by taking
advantage of superposition theorem.
1> Decomposition :- resolve the input function f(t) in terms of simpler function such as exponential, impulse for which the
response can easily be evaluated.
2> Determine individually the response of a linear system for the simple input function..
3> Synthesis: - Find the sum of the individual responses which become the output response.
Representation of a function f(t) as continuous sum of impulse function:Let us consider an arbitrary excitation f(t) a shown in figure. In the limit t0, nth element area may be constructed as
a rectangle of width t and height f(nt). This delta function is symbolically represented as
f(nt)( t)(t-nt).
The function is continuous sum of such impulse functions
f (t ) =
lim
f(nt )( t ) (t nt )
t 0 n =
Page 18 A.Sarkar,ECE,JGEC,
R ( w) = F ( w) H ( w)
H(w) is known as transfer function or system function.
Energy Signal & Power signal:A useful parameter of a signal f(t) is its normalized energy. We define the normalized energy ( or simply the energy) E
of a signal as the energy dissipated by a voltage f(t) applied to a 1 ohm resistor( or by passing a current f(t) through 1
ohm resistor). Thus
E=
f 2 (t )dt
The energy of a signal exists only if the integral is finite. The signals for which E is finite is called energy signal.
Non-periodic signals are examples of Energy signals.
The energy may be infinite for many signals and for such signals w used to define another quantity called average
power. Average power may exist if its energy is infinite. Such signals with finite average power is called power signal.
Periodic signals are examples of power signals.
Parsevals theorem for energy signal:- Energy can be calculated in terms of Fourier transformation. This theorem is
very useful as it helps for evaluating the energy of signal without knowing its time domain nature.
1
1
1
2
2
2
1
2
2
E = f 2 (t )dt =
F ( w) dw = F ( w) df
2
Energy Spectral Density:-Consider a signal f(t) is applied to an ideal narrowband filter. The transfer function of the
filter is H(w). The response of the system is given by R(w)=H(w)F(w).
The energy E of the output signal is given by
1
1
2
2
E=
R( w) dw =
R( w) F ( w) dw
2
2
Now it is obvious that H(w)=0 except for narrow band wm to wm for which it is unity. F(w) is constant with frequency
for a narrowband filter(w0). Hence energy over this narrowband w=2wm is given as
Page 19 A.Sarkar,ECE,JGEC,
1 wm
1
2
2
(
)
=
(
)
(2 wm )
E=
F
w
dw
F
w
2 wm
2
let 2 wm = w
E=
1
2
2
F ( w) (w) = F ( w) (f )
2
E
2
= F ( w) = Energy Spectral Density = ( w)
f
Thus E represents the contribution of energy due to bandwidth w of the signal. Hence energy contribution per unit
bandwidth is given as energy spectral density.The total energy E is given by
1
E=
2
1
F ( w) dw =
2
2
( w)dw
2
Energy contribution includes both negative and positive frequency and the contribution is equal. F ( w) = F ( w)
For real functions , the spectrum is symmetrical .Hence
E=
( w)dw
The relationship between energy densities of input and response can be derived as follows
R(w)=H(w)F(w)
Hence
2
R( w) = H ( w) F ( w) = H ( w) F ( w)
r ( w) = H ( w) f ( w)
Page 20 A.Sarkar,ECE,JGEC,
Average Power:- The average power dissipated by a voltage f(t) applied across 1-ohm resistor is defined as the average
power or simply power of the signal f(t). This is same as the power dissipated by a current f(t) passing through a 1 ohm
resistor.
2
lim 1 T / 2
f
(
t
)
dt = f 2 (t )
P=
T T T / 2
The above equation represents the mean square value of the signal f(t) and hence the average power is same as the
mean square value.
The power Density Spectrum:We will drive an expression for the power density spectrum assuming power signal as a limiting case of energy signal.
Consider a power signal f(t) (extending to infinity) . Let us truncate this signal so that it is zero outside the interval
T/2.
f (t ) | t |< T / 2
Let us call FT (t ) =
0 otherwsie
Signal FT(t) is of finite duration T and hence it is an energy signal with energy E given by
ET=
f T (t ) dt =
2
FT ( w) df
Since f(t) over the integral (-T/2,T/2) is same as fT(t) over the interval(-,) we have
f T (t ) dt =
2
T /2
T / 2
f (t ) dt
2
1 T /2
1
f
(
t
)
dt
=
F
(
w
)
df
T
T T / 2
T
In the limit T , the left hand side represents the average power P of the function f(t).
2
lim FT ( w)
P=
df
T
T
2
S ( w) =
lim
FT ( w)
FT ( w)
T
may approach a finite value. Let us denote this finite value by S(w) i.e.
P = f 2 (t ) = S ( w)df =
1
2
S ( w)dw
According to above equation the total power is obtained by multiplying S(w) with bandwidth w and integrating
over the entire bandwidth. Therefore S(w) may be thought as average power per unit bandwidth and hence known as
power spectral density.
2
FT ( w) = FT ( w) the contribution of +ve and ve frequencies are equal. Hence average power may be written as
Page 21 A.Sarkar,ECE,JGEC,
1
P = 2 S ( w)df = S ( w)dw
0
0
PSD of input and the response:Let us apply a power signal f(t) at the input of a linear system with transfer function H(w) and let the output signal be
r(t).
The signals fT(t) and rT(t) represent signals f(t) and r(t) respectively truncated |t|=T/2.
If we apply this truncated signal at the input the response will not be rT(t), it will extend beyond t=T/2. However since
the input is zero for t>T/2, for a stable system the response for t>T/2 must decay with time. In the limit as T this
contribution (beyond t=T/2) will be of no significance.
Hence for T the response for fT(t) may be considered to be rT(t0 without much error.
lim
f T (t ) rT (t )
T
lim
RT ( w) = H ( w) FT ( w)
T
lim 1
2
S r ( w) =
RT ( w)
T T
lim 1
2
S r ( w) =
H ( w) FT ( w)
T T
lim 1
2
2
S r ( w) = H ( w)
FT ( w)
T T
2
S r ( w) = H ( w) S f ( w)
Page 22 A.Sarkar,ECE,JGEC,
Correlation:
the application of correlation to signal detection in radar, where a signal pulse is transmitted in order to detect a
suspected target. If a target is present, the pulse will be reflected by it. If there is no pulse, there will be no pulse, just a
noise. By detecting the presence or absence of the reflected pulse we confirm the presence or absence of a target. By
measuring the time delay between the transmitted and received pulses we determine the distance of the target.
For some value of there is a strong correlation, then we not only detect the target also can detect relative time shift of
transmitted signal with received signal. The cross correlation between two functions can be given by
Consider
R1, 2 ( ) =
f 1(t ) f 2(t + )d =
f 1(t ) f 2(t )d
The searching parameter introduced in the expressions of correlations is needed to find out the maximum possible
correlation between waveforms. It may happen when =0 two waveforms have no correlation but they may have
significant correlation with suitable value of . For example two non overlapping signals correlation is zero but
correlation increases as increases.
Difference between Convolution and correlation:Correlation is a function of the delay parameter , where as convolution is a function of t. In convolution delay plays
the role of a dummy variable and it disappears after solution of an integral. Where as in correlation physical time t
plays the role of dummy variable.
Convolution does not depend on which function is shifted and which direction it is shifted but correlation does. So
convolution is commutative.
Autocorrelation:It is a special form of cross-correlation.
R( ) =
f (t ) f (t + )dt =
f (t ) f (t )dt
Autocorrelation is a measure of similarity of a function with its delayed replica. An analogy comparison of your
photo with the photo five year back
Page 23 A.Sarkar,ECE,JGEC,
R( ) =
R( ) =
This shows the for a real f(t), the autocorrelation function is an even function of , that is R()=R(-)
We now show that ESD (w)=|F(w)|2 is the Fourier transform of the autocorrelation function R().
F [ R ( )] = f (t ) f (t + )dt e jw d = f (t ) f (t + )e jw d dt
The inner integral is the Fourier transform of f(t+) which is f() left shifted by t seconds and by time shifting property
it is in frequency domain F(w)ejwt . Therefore
F [ R ( )] = F ( w)
f (t )e jwt dt = F ( w) F ( w) =| F ( w) |2 = ( w)
)
+
=
T
T
T T
T T
2
now recall that R()|FT(w)| hence the Fourier transform of the preceding equation gives
2
lim FT ( w)
R( )
= S ( w)
T
T
Page 24 A.Sarkar,ECE,JGEC,
Page 25 A.Sarkar,ECE,JGEC,
Probability
we'll deal with random signals whose exact behavior cannot be described in advance. Random signals occur
in communication both as unwanted noise and as desired information-bearing waveforms. Lacking detailed
knowledge of the time variation of a random signal, we must speak instead in terms of probabilities and
statistical properties.
Random Experiment:- Whose outcome cannot be predicted with certainty like flipping a coin, throwing a
die.
Outcome:- elementary result of experiment
Sample space:- set of all possible outcome
Event:- subset of sample space, collection of outcome. Example: for the experiment throwing a die event
outcome is odd {1,3,5}
Disjoint event:- if their intersection is empty.
For this purpose, let's identify a specific event A as something that might be observed on any trial of a chance
experiment. We repeat the experiment N times and record NA the number of times A occurs. The ratio NA / N then
equals the relative frequency of occurrence of the event A for that sequence of trials.
as N becomes very large and if every sequence of trials yields the same limiting value. Under these conditions we take
the probability of A to be P(A)=NA/N
N
The union event A + B (also symbolized by A U B) stands for the occurrence of A or B or both, so its subset consists
of all events either A or B.
The intersection event AB (also symbolized by A B) stands for the occurrence of A and B, so its subset consists
only of those events in both A and B.
P(A1+A2)=P(A1)+P(A2) IF A1A2=0
A,A, = 0 means that they are mutually exclusive.
Consider events A and B that are not mutually exclusive, P(A+B)=P(A)+P(B)-P(AB)
1.
0P(A)1
2.
Conditional probability:Conditional probabilities are introduced here to account for event dependence and also to define statistical
independence. Let us assume two events A and B have probabilities P(A) and P(B) . I an observer knows that the event
A has occurred , the the probability that event B will occur will not be P(B).
We measure the dependence of B on A in terms of the conditional probability. P(B|A)=P(AB)/P(A)
. If P(A/B)=P(A) then the knowledge of B does not change the probability of occurrence of A. In this case A and B are
said to be independent. for independent events
P(A B)=P(A) P(B)
The notation B/A stands for the event B given A, and P(B/A) represents the probability of B conditioned by the
knowledge that A has occurred. If the events happen to be mutually exclusive, then P(AB) = 0 and confirms that
P(B(A) = 0 as expected. With P(AB) 0, as N
P(B|A)=NAB/N|NA/N= NAB/ NA
and we thereby obtain two relations for the joint probability, namely
Or we could eliminate P(AB) to get Bayes' theorem
Total probability:-
Page 26 A.Sarkar,ECE,JGEC,
Random Variable:It is a mapping from sample space to a set of real numbers. I.e. assignment of real numbers to the outcome of a random
experiment. Despite the name, a random variable is neither random nor a variable. Instead, it's a function that generates
numbers from the outcomes of a chance experiment.
Although a mapping relationship underlies every RV, we usually care only about the resulting numbers. We'll therefore
adopt a more direct viewpoint and treat X itself as the general symbol for the experimental outcomes. This viewpoint
allows us to deal with numerical-valued events such as X = a or X a, where a is some point on the real line.
Furthermore, if we replace the constant a with the independent variable x, then we get probability functions that help us
calculate probabilities of numerical-valued events.
Page 27 A.Sarkar,ECE,JGEC,
A PDF is a non negative function whose total area is unity and whose area in the range a<x<b equals the probability of
observing x in that range.
let a = x - dx and b = x. The integral then reduces to the differential area px(x) dx and we see that
Page 28 A.Sarkar,ECE,JGEC,
Gaussian PDF
Page 29 A.Sarkar,ECE,JGEC,
STATISTICAL AVERAGES:For some purposes, a probability function provides more information about an RV than actually needed. Indeed, the
complete description of an RV may prove to be an embarrassment of riches, more confusing than illuminating. Thus, we
often find it more convenient to describe an RV by a few characteristic numbers. These numbers are the various statistical
averages presented here.
The mean of the random variable X is a constant m, that equals the sum of the values of X weighted by their probabilities.
This statistical average corresponds to an ordinary experimental average in the sense that the sum of the values observed
over
N >> 1 trials are expected to be about Nmx. For that reason, we also call mx, the
Expected value of X, and we write E[XI or X to stand for the expectation operation
that yields mx .
which expresses the mean of a discrete RV in terms of its frequency function Px(xi).
for continuous RV,
Page 30 A.Sarkar,ECE,JGEC,
If g(X) = Xn then E[Xn] is known as the nth moment of X. The first moment, of course, is just the mean value E[X] = mx. The
2
second moment E[X2] or <X2> is called the mean-square value, as distinguished from the mean squared mx2 = X with g(X) =
X2, we have
The mean-square value will be particularly significant when we get to random signals and noise.
The standard deviation of X, denoted by x provides a measure of the spread of observed values of X relative to m,. The square of
the standard deviation is called the variance, or second central moment, defined by
A small standard deviation therefore implies a small spread of likely values, RV concentrated around mean so less random and
vice versa.
Page 31 A.Sarkar,ECE,JGEC,
Page 32 A.Sarkar,ECE,JGEC,
Page 33 A.Sarkar,ECE,JGEC,
Page 34 A.Sarkar,ECE,JGEC,
Page 35 A.Sarkar,ECE,JGEC,
Page 36 A.Sarkar,ECE,JGEC,
Page 37 A.Sarkar,ECE,JGEC,
Page 38 A.Sarkar,ECE,JGEC,
Page 39 A.Sarkar,ECE,JGEC,
Introduction to sampling & reconstruction:Signals can be classified depending on the chars. of the time (independent) variable and the values they take.
Continuous Time (Analog) signals: They are defined for every value of time where they take values in the
continuous interval (a,b). a can be - and b can be . Mathematically they are described by function of a continuous
variable.
X(t)=cost -<t<.
Discrete Time signals:-They are defined only at certain specific values of time. These time instants may not be
equidistant but in usual practice, they are usually taken at equally spaced intervals for computational convenience and
mathematical tractability.
If we use n of the discrete time instants as the independent variable, the signal values become a function of integer
variable (i.e. a sequence of numbers). Thus a DT signal is represented by a sequence of numbers.
where =+ve angular freq. and - =-ve angular freq units are rad/sec .,
Page 40 A.Sarkar,ECE,JGEC,
On the other hand, the sequence at any two sinusoids with frequencies in the range - <w< or
- 1/2<f<1/2 are distinct.
Consequently ,DT sinusoidal signals with frequency |w| or |f|1/2 are unique.
Any sequence resulting from a sinusoid with a frequency |w| or |f|1/2 is identical to sequence obtained from a
sinusoid with frequency |w|<. Because of this similarity, we call the sinusoids with frequency |w| an alias of
corresponding sinusoid with frequency |w|<.
Thus we regard frequency in the range - <w< or - 1/2<f<1/2 are unique and all frequencies |w| or |f|1/2
as alias.
3>The highest rate of oscillation in a DT sinusoid is attained when w= ( or w=-) or f=1/2 or(f=-1/2)
Proof:- let x[n]=coswon To see what happens for w02. We consider sinusoid with frequency w1=w0
and w2=2- w0
as w1 varies from to 2
w2 varies from to 0
x1[n]=ACos w1n= ACos w0n
x2[n]=ACos w2n= ACos(2- w0)n= ACos (-w0n)=x1[n]
Hence w2 is an alias of w1. If you use a sine function, result will be same except for a 180 deg phase shift.
As we increase the relative frequency wo from to 2, rate of oscillation decreases.
For wo =2 , the result is a constant signal as in the case of wo =0. Obviously for wo =(f=1/2) we have
the highest rate of oscillation.
Page 41 A.Sarkar,ECE,JGEC,
xa(t)=Acos(2Ft+)
xa(nT)=Acos(2FnT+)
xa(nT)=Acos(2nF/ Fs +)
x[n]= Acos(2nf +)
Therefore f=F/Fs Therefore w=T
-<F<
-1/2<f<1/2
-<<
-<w<
Fundamental difference between CT and DT signal is their range of value for F,f,w,.
Periodic sampling of CT signal implies frequency range F into a finite frequency range f. Since the highest
frequency in a DT signal w= or f=1/2.
Fmax=Fs/2=1/2T
max=Fs=/T
Therefore sampling introduces ambiguity, since the highest frequency in a CT signal that can be uniquely
distinguished when such a signal is sampled at a rate Fs=1/T is Fmax=Fs/2. To see what happens to frequency above
Fs/2 consider the following example.
Page 42 A.Sarkar,ECE,JGEC,
-1/2fo1/2
-Fs/2FoFs/2
In This case the relationship between Fo and fo is one- to- one, and hence it is possible to identify or reconstruct the
analog signal xa(t) from samples of x[n].
On the other hand xa(t)= Acos(2Fkt+)
Where Fk=Fo+kFs,k=1, 2, 3.
Is sampled at a rate of Fs, it is clear that the frequency Fk is outside the fundamental frequency range
-Fs/2FFs/2
consequently x[n]=xa(nT)= Acos(2(Fo+kFs)/Fs.n+)= Acos(2nFo/Fs++2kn)= Acos(2fon+)
which is identical to DT signal obtained earlier.
Thus an infinite number of CT sinusoid is represented by sampling the same DT signal. Consequently an ambiguity
exists as to which CT signals these values represent.
Page 43 A.Sarkar,ECE,JGEC,
Since Fs/2 , which corresponds to w=, the highest frequency that can be represented unique with a sampling rate
Fs; it is a simple matter to determine the mapping at any alias frequency above Fs/2(w=) into equivalent
frequency below Fs/2. We can use Fs/2 or w= as the pivotal point and reflect or fold the alias frequency to the
range 0w. The point of reflection Fs/2(w=) is called the folding frequency.
Sampling Theorem:We know that the highest frequency in an analog signal that can be unambiguously represented when the signal is
sampled at a rate Fs=1/T is Fs/2.
Any frequency above Fs/2 or below Fs/2 results in samples that are identical with corresponding frequency in the
range -Fs/2FFs/2
To avoid this ambiguity resulting from aliasing, we must select the sampling rate to be sufficiently high i.e we must
select
Fs/2>Fmax
Fs>2Fmax
With this sampling rate, any frequency component, say |Fi|<Fmax in the analog signal is mapped into a DT sinusoid
with frequency
-1/2fi=Fi/Fs1/2
-wi=2fi
|f|=1/2 or |w|= is highest frequency in a DT signal, choice of sampling rate avoids aliasing. Fs>2Fmax ensures that
all the sinusoidal components in the analog signal are mapped into corresponding DT frequency components with
frequencies in the fundamental interval.
Example:
Page 44 A.Sarkar,ECE,JGEC,
example:-
Page 45 A.Sarkar,ECE,JGEC,
Sampling process:-
xs(t)=x(t)g(t)
C e
g(t)=
jn 2f s t
1
Where Cn=
T
T /2
n =
T / 2
Page 46 A.Sarkar,ECE,JGEC,
The Ct signal x(t) must be sampled in such a way that the original signal can be reconstructed from these samples
Otherwise the sampling process is useless. Let us obtain the condition necessary to faithfully reconstruct the original
signals from the samples of the signal. The condition can be easily obtained if the signals are analyzed in the frequency
domain.
x s (t ) = x (t ) C n e
jn 2 f s t
n =
= C n x (t ) e jn 2f s t
n =
Xs( f ) = xs (t )e j 2ft dt =
C x(t )e
n =
jn 2f s t j 2ft
dt
Xs( f ) =
Cn x(t )e j 2 ( f nf s )dt
n =
Page 47 A.Sarkar,ECE,JGEC,
x(t )e j 2 ( f nf s ) tdt = X ( f nf s )
Therefore
Xs( f ) =
C X ( f nf )
n =
Spectrum of sampled CT signal =spectrum of x(t)+spectrum of x(t) translated to each harmonic of the sampling
frequency.
To avoid overlapping fs-fhfh therefore fs2fh fs=2fh is called Nyquist rate.
Sampling Theorem:- a band limited signal x(t) having no frequency components above fh Hz , is completely specified
by samples that are taken at a uniform rate greater than 2fh hz.
If X(f)=0 for |f|>fh is called band limited.
Page 48 A.Sarkar,ECE,JGEC,
(t nT )
g (t ) =
n =
By Fourier Series
g(t)=
C n e jn2f st
n =
1 T /2
jn 2f s t
dt where fs=1/T=sampling rate
Where Cn= T / 2 (t )e
T
(t)=1 at t=0
=0 otherwise
Cn =
1 0
e = 1/ T = f s
T
Xs ( f ) = f s
X ( f nf )
n =
Page 49 A.Sarkar,ECE,JGEC,
Signal Reconstruction:Spectrum of sampled signal has an amplitude equal to fs=1/T Therefore in order to remove this scalingh constant, the
LPF must have an amplitude response of 1/fs=T.
BW=fs/2=2fh(assuming sampling is done at fs=2fh)
h(t ) = T
fs / 2
fs / 2
e j 2ft df
T
(e jf s t e jf s t )
j 2t
sin f st
h(t ) = Tf s
= sin c( f st )
f st
h(t ) =
x(t ) =
n =
n =
Page 50 A.Sarkar,ECE,JGEC,
Page 51 A.Sarkar,ECE,JGEC,
Source
encoder
Channel
encoder
modulator
noise
channel
Inf
sink/destinatio
ns
Source
decoder
Channel
decoder
demodulator
Page 53 A.Sarkar,ECE,JGEC,