Program : B.

Tech
Subject Name: Digital Communication
Subject Code: EC-502
Semester: 5th
Downloaded from www.rgpvnotes.in

Department of Electronics and Communication Engineering

Sub. Code: EC 5002 Sub. Name: Digital Communication

Unit 1
Syllabus: Random variables Cumulative distribution function, Probability density function, Mean, Variance
and standard deviations of random variable, Gaussian distribution, Error function, Correlation and
autocorrelation, Central-limit theorem, Error probability, Power Spectral density of digital data.

1.1 Random Variable:

The outcome of an event may always need not be numbers. It may be the name of the horse in an event
etc. An experiment whose outcome cannot be predicted exactly, and hence is random, is called a random
experiment.
The collective outcomes of a random experiment form a sample space. A particular outcome is called a
sample point or sample. Collection of outcomes is called an event. Thus an event is a subset of sample
space.
A Random Variable is a real valued function defined over the sample space of random experiments. The
random variables are denoted by uppercase letters such as X, Y etc and the values assumed by them are
denoted by lower case letters with subscript such as x 1, x2, y1, y2 etc.

Random variables are classified as discrete or continuous.

If in any finite interval X(λ) assumes only a finite number of distinct values then the random variable is
discrete. Ex. Tossing a die.

If in any finite interval X(λ) assumes continuous values then the random variable is continuous. For
Example, shift in the magnitude of miss of a bullet due to wind.

A random variable is a mapping from sample space Ω to a set of real numbers. What does this mean? Let’s
take the usual evergreen example of “flipping a coin”.
In a “coin-flipping” experiment, the outcome is not known prior to the experiment, that is we cannot
predict it with certainty. But we know the all possible outcomes – Head or Tail. Assign real numbers to the
all possible events, say “0” to “Head” and “1” to “Tail”, and associate a variable “could take these two
values. This variable “X” is called a random variable, since it can randomly take any value ‘0’ or ‘1’ before
performing the actual experiment.
Obviously, we do not want to wait till the coin-flipping experiment is done. Because the outcome will lose
its significance, we want to associate some probability to each of the possible event. In the coin-flipping
experiment, all outcomes are equally probable. This means that we can say that the probability of getting
Head as well that of getting Tail is 0.5.
This can be written as,
𝑃(𝑋 = 0) = 0.5 𝑎𝑛𝑑 𝑃(𝑋 = 1) = 0.5

1.2 Cumulative Distribution Function:

The cumulative distributive function or distribution function for a discrete random variable is defined as

𝐹 (𝑋 ) = 𝑃 (𝑋 ≤ 𝑥 ) = ∑ 𝑓 (𝑢 ) − ∞ < 𝑥 < ∞
𝑢≤𝑥

Page 1 of 9

Page no: 1 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

If X can take on the values x1, x2, x3, x4 … xn, then the distributive function is given by

0 − ∞ ≤ 𝑥 < 𝑥1
𝑓(𝑥1 ) 𝑥1 ≤ 𝑥 < 𝑥2
𝐹(𝑥) = 𝑓 (𝑥1 ) + 𝑓(𝑥2 ) 𝑥2 ≤ 𝑥 < 𝑥3
⋮
{ 𝑓(𝑥1 ) + 𝑓 (𝑥2 ) + ⋯ + 𝑓(𝑥𝑛 ) 𝑥𝑛 ≤ 𝑥 < ∞

Mathematically, a complete description of a random variable is given be “Cumulative Distribution

Function“- FX(x). Here the bold faced “X” is a random variable and “x” is a dummy variable which is a place
holder for all possible outcomes. The Cumulative Distribution Function is defined as,

𝑓𝑋 (𝑥) = 𝑃(𝑋 ≤ 𝑥)

If we plot the CDF for our coin-flipping experiment, it would look like the one shown in the figure

Figure 1.01 Cumulative Distribution Function

The example provided above is of discrete nature, as the values taken by the random variable are discrete
and therefore the random variable is called Discrete Random Variable.
If the values taken by the random variables are of continuous nature, then the random variable is called
Continuous Random Variable and the corresponding cumulative distribution function will be smoother
without discontinuities.

1.3 Probability Distribution function:

Consider an experiment in which the probability of events is as follows. The probabilities of getting the
numbers 1,2,3,4 individually are 1/10, 2/10,3/10,4/10 respectively. It will be more convenient for us if we
have an equation for this experiment which will give these values based on the events. For example, the
equation for this experiment can be given by f(x)=x/10 where x=1,2,3,4. This equation is called probability
distribution function.

1.4 Probability Density function (PDF) and Probability Mass Function (PMF):
It’s more common deal with Probability Density Function (PDF)/Probability Mass Function (PMF) than CDF.
The PDF is given by taking the first derivate of CDF.

𝑑F𝑋 (𝑥)
𝑓𝑋 (𝑥 ) =
𝑑𝑥

For discrete random variable that takes on discrete values, is it common to defined Probability Mass
Function.
𝑓𝑋 (𝑥) = 𝑃(𝑋 = 𝑥)

Page 2 of 9

Page no: 2 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The previous example was simple. The problem becomes slightly complex if we are asked to find the
probability of getting a value less than or equal to 3. Now the straight forward approach will be to add the
probabilities of getting the values x=1,2,3 which comes out to be 1/10+2/10+3/10 =6/10. This can be easily
modeled as a probability density function which will be the integral of probability distribution function with
limits 1 to 3.
Based on the probability density function or how the PDF graph looks, PDF fall into different categories like
binomial distribution, Uniform distribution, Gaussian distribution, Chi-square distribution, Rayleigh
distribution, Rician distribution etc. Out of these distributions, you will encounter Gaussian distribution or
Gaussian Random variable in digital communication very often.

The PDF has the following properties:

a) 𝑓𝑥 (𝑥 ) ≥ 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑥

This results from the fact that F(x) increases monotonically as x increases, more outcomes are
included in the probability of occurrence represented by F(x).
∞
b) ∫−∞ 𝑓𝑥 (𝑥)𝑑𝑥 = 1

This result can be seen from the fact that

∞
∫ 𝑓𝑥 (𝑥)𝑑𝑥 = 1 𝐹(∞) − 𝐹(−∞) = 1 − 0 = 1
−∞

𝑥
c) 𝐹(𝑥 ) = ∫−∞ 𝑓𝑥 (𝑥)𝑑𝑥

This result from the integration of the definition of PDF.

1.5 Mean:
The mean of a random variable is defined as the weighted average of all possible values the random
variable can take. Probability of each outcome is used to weight each value when calculating the mean.
Mean is also called expectation (E[X])
For continues random variable X and probability density function fX(x)
∞
𝐸[𝑋] = ∫ 𝑥𝑓𝑋 (𝑥 )𝑑𝑥
−∞
For discrete random variable X, the mean is calculated as weighted average of all possible values (x i)
weighted with individual probability (pi)

1.6 Variance:
Variance measures the spread of a distribution. For a continuous random variable X, the variance is defined
as
∞
𝑣𝑎𝑟[𝑋] = ∫ (𝑥 − 𝐸[𝑋])2 𝑓𝑋 (𝑥 )𝑑𝑥
−∞

For discrete case, the variance is defined as

∞

𝑣𝑎𝑟[𝑋] = 𝜎 X = ∑(𝑥 − µ𝑋)2 𝑝𝑖

2

−∞

Standard Deviation (σ) is defined as the square root of variance σ2X

Page 3 of 9

Page no: 3 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Properties of Mean and Variance:

For a constant – “c” following properties will hold true for mean

𝐸 [𝑐𝑋] = 𝑐𝐸[𝑋]
𝐸 𝑋 + 𝑐 ] = 𝐸 [𝑋 ] + 𝑐
[
𝐸 [𝑐 ] = 𝑐

For a constant – “c” following properties will hold true for variance

𝑣𝑎𝑟[𝑐𝑋] = 𝑐 2 var [X]

𝑣𝑎𝑟[𝑋 + 𝑐 ] = var [X]
𝑣𝑎𝑟[𝑐 ] = 0

PDF and CDF define a random variable completely. For example: If two random variables X and Y have the
same PDF, then they will have the same CDF and therefore their mean and variance will be same.
On the other hand, mean and variance describes a random variable only partially. If two random variables
X and Y have the same mean and variance, they may or may not have the same PDF or CDF.

1.7 Binomial, Poisson and Normal (Gaussian) Distributions

The most important probability distributions are Binomial, Poisson and Normal (Gaussian). Binomial
and Poisson distributions are for discrete random variables, whereas the Normal distribution is for
continuous random variables.

1.7.1 Binomial Distribution:

Let us consider an event with only two possible outcomes. Such an experiment is known as
Bernoulli trial.) One outcome is called success and other is called failure. Let the experiment is repeated a
number of times. Let us consider that the probability of success 𝒑 in each trial is same, i.e. the trials are
independent. Then the probability of failure in each trial is 𝒒 = (𝟏 − 𝒑). The probability of x successes in n
trials is given by a probability function known as Binomial distribution:
𝒏 …1.7.1.1
𝒇(𝒙) = 𝑷(𝑿 = 𝒙) = ( ) 𝒑𝒙 𝒒𝒏−𝒙
𝒙

The properties of Binomial distribution are

𝑴𝒆𝒂𝒏 = 𝒏𝒑, 𝒗𝒂𝒓𝒊𝒂𝒏𝒄𝒆 = 𝒏𝒑𝒒

Example 1.7.1.1
A fair die is tossed 5 times. A toss is called a success if face 1 or 6 appears. Find (a) the probability of two
successes, (b) the mean and the standard deviation for the number of successes.
(a)
𝟐 𝟏 𝟏 𝟐
𝑵 = 𝟓, 𝒑 = = , 𝒒 = 𝟏 − 𝒑 = 𝟏 − =
𝟔 𝟑 𝟑 𝟑
𝟓 𝟏 𝟐 𝟐 𝟓−𝟐 𝟖𝟎 𝒏 𝒏!
𝑷 ( 𝑿 = 𝟐) = ( ) [ ] [ ] = ( )=
𝟐 𝟑 𝟑 𝟐𝟒𝟑 𝒙 𝒙! (𝒏 − 𝒙)!
(b)
𝟏
𝑴𝒆𝒂𝒏 = 𝒏𝒑 = 𝟓 × = 𝟏. 𝟔𝟔𝟕
𝟑
𝟏 𝟐
𝑺𝒕𝒂𝒏𝒅𝒂𝒓𝒅 𝒅𝒆𝒗𝒊𝒂𝒕𝒊𝒐𝒏 = √𝒏𝒑𝒒 = √𝟓 × × = 𝟏. 𝟎𝟓𝟒
𝟑 𝟑

Page 4 of 9

Page no: 4 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

1.7.2 Poisson Distribution:

Let X be a discrete random variable that can assume values 0, 1, 2… Then the probability function of X
is given by Poisson distribution:
𝝀𝒙 𝒆−𝝀 …1.7.2.1
𝒇(𝒙) = 𝑷(𝑿 = 𝒙) =
𝒙!

𝒙 = 𝟎, 𝟏, 𝟐, …
Where, 𝝀 is a positive constant. The properties of Poisson distribution are
𝑴𝒆𝒂𝒏 = 𝝀, 𝒗𝒂𝒓𝒊𝒂𝒏𝒄𝒆 = 𝝀

1.7.3 Poisson Approximation to Binomial Distribution:

In Binomial distribution if n is large and p is close to zero, then it can be approximated by Poisson
distribution with 𝝁 = 𝒏𝒑. In practice 𝒏 ≥ 𝟓𝟎 and 𝒏𝒑 ≤ 𝟓 give satisfactory approximation.
It can be seen from the equation of binomial distribution that if n is large, then calculation of a desired
probability considering Binomial distribution is tedious. On the other hand, calculation of a desired
probability considering Poisson distribution is fairly simple as seen from equation 1.7.2.1.

1.7.4 Normal (or Gaussian) distribution

Gaussian PDF looks like a bell. It is used most widely in communication engineering. This is the most
important continuous probability distribution as most of the natural phenomenon are characterized by
random variables with normal distribution. The importance of normal distribution is further enhanced
because of the central limit theorem. The density function (PDF) for normal (Gaussian) distribution is given
by
2
−(𝑥−µ) …1.7.4.1
1 2
𝑓 (𝑥 ) = 𝑒 2𝜎 −∞ <𝑥 < ∞
√2𝜋𝜎2

Where µ and 𝜎 are mean and standard deviation, respectively.

The properties of the normal distribution are
𝑴𝒆𝒂𝒏 = 𝝁, 𝒗𝒂𝒓𝒊𝒂𝒏𝒄𝒆 = 𝝈𝟐
The corresponding distribution function is
1 𝑥 2 2
𝐹 (𝑥 ) = 𝑃 (𝑋 ≤ 𝑥 ) = ∫−∞ 𝑒 −(𝑣−𝑢) /2𝜎 𝑑𝑣
𝜎√2π
1 𝑥 −(𝑣−𝑢)2 /2𝜎 2
1
=2+ ∫ 𝑒 𝑑𝑣 …1.7.4.2
𝜎√2π 0
𝑥−𝜇
Let Z be the standardized random variable corresponding to X. Thus if 𝑍 = , then the mean of Z is zero
𝜎
and its variance is 1. Hence
1 2
𝐹(𝑧) = √ 𝑒 −𝑧 /2 …1.7.4.3
2π
𝐹(𝑧) is known as standard normal density function. The corresponding distribution function is
1 𝑧 2 1 1 𝑧 2
𝐹(𝑧) = 𝑃(𝑍 ≤ 𝑧) = √ ∫−∞ 𝑒 −𝑢 /2 𝑑𝑢 = 2 + √ ∫0 𝑒 −𝑢 /2 𝑑𝑢 …1.7.4.4
2π 2π
The integral of equation 1.7.4.4 is not easily evaluated. However, it is related to the error function, whose
tabulated values are available in mathematical tables.

Error Function
The error function of z is defined as
2 𝑧 2
𝑒𝑟𝑓 𝑧 = π ∫0 𝑒 −𝑢 𝑑𝑢 …1.7.3.5
√
The error function has the values between 0 and 1.
𝑒𝑟𝑓 (0) = 0 𝑎𝑛𝑑 𝑒𝑟𝑓 (∞) = 1
The Complementary error function of z is defined as
2 ∞ 2
𝑒𝑟𝑓𝑐 𝑧 = 1 − 𝑒𝑟𝑓 𝑧 = π ∫𝑧 𝑒 −𝑢 𝑑𝑢 …1.7.3.6
√

Page 5 of 9

Page no: 5 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The relationship between 𝑓 (𝑥 ), 𝑒𝑟𝑓 𝑧 and 𝑒𝑟𝑓𝑐 𝑧 is as follows

1 𝑧 1 𝑧 …1.7.3.7
𝑓 (𝑥 ) = [1 + 𝑒𝑟𝑓 ( )] = 1 − 𝑒𝑟𝑓𝑐 ( )
2 √2 2 √2

1.7.5 Normal approximation to Binomial Distribution

If n is large and if neither p nor q is close to zero, then the binomial distribution can be closely
approximated by a normal distribution with standardized random variable given by
𝑋 − 𝑛𝑝
𝑍=
√𝑛𝑝𝑞
In practice 𝒏𝒑 ≥ 𝟓 and 𝒏𝒒 ≥ 𝟓 give the satisfactory approximation.

1.7.6 Normal approximation to Poisson Distribution

As Binomial distribution has relationship with both Poisson and Normal distributions, one would expect
that there should be some relationship between Poisson and Normal distributions. In fact it is found to be
so. It has been seen that the Poisson distribution approaches Normal distribution as 𝝀 → ∞.

1.8 Error Function:

In mathematics, the error function is a special function of sigmoid shape that occurs in probability,
statistics, and partial differential equations describing diffusion. It is defined as:
𝑥
1 2
erf(𝑥 ) = ∫ 𝑒 −𝑡 𝑑𝑡
√𝜋 −𝑥

𝑥
2 2
= ∫ 𝑒 −𝑡 𝑑𝑡
√𝜋 0

In statistics, for nonnegative values of x, the error function has the following interpretation: for a random
variable X that is normally distributed with mean 0 and variance , erf(x) describes the probability of X
falling in the range [−x, x].

1.9 Central Limit Theorem

Central Limit Theorem (CLT) states that irrespective of the underlying distribution of a, if you take a
number of samples of size N from the population, then the “sample mean” follow a normal distribution
with a mean of μ and a standard deviation of σ/ sqrt(N).The normality gets better as your sample size N
increases. Here the underlined word signifies that irrespective of the base distribution the probability
distribution curve will approach Gaussian or Normal distribution as the number of sample increases. In
other words, CLT states that the sum of independent and identically distributed random variables
approaches Normal distribution as 𝑁 ≥ ∞.

Applications of Central Limit Theorem:

CLT is being applied in vast range of applications including Signal processing, channel modeling, random
process, population statistics, engineering research, predicting the confidence intervals, hypothesis testing,
even in Casino and gambling , etc…,
One such application is deriving the response of a cascaded series of Low pass filters by applying Central
limit theorem .The author has illustrated how the response of a cascaded series of low pass filters
approaches Gaussian shape as the number of filters in the series increases.

Page 6 of 9

Page no: 6 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

In digital communication, channel noise is often modeled as normally distributed. Modeling a channel as
normally distributed when the noise components in that channel are sufficiently large is justified by Central
limit theorem.

1.10 Correlation
The Correlation (more precisely Cross Correlation) between two waveforms is the measure of similarity
between one waveform and time delayed version of the other waveform. This express how much one
waveform is related to the time delayed version of the other waveform.
The expression for correlation is very close to convolution. Consider two general complex functions 𝑓1 (𝑡)
and 𝑓2 (𝑡), which may or may not be periodic, and not restricted, to finite interval. Then cross correlation or
simply correlation 𝑅1,2 (𝜏) between two function is defined as follows:
𝑇
2
…1.10.1a
𝑅1,2 (𝜏) = lim ∫ 𝑓1 (𝑡)𝑓2 ∗ (𝑡 + 𝜏)𝑑𝑡
𝜏→∞ −𝑇
2
(The conjugate symbol*, is removed if the function is real.)
This represents the shift of function 𝑓2 (𝑡) by an amount – 𝜏(i.e. towards left). A similar effect can be
obtained by shifting 𝑓1 (𝑡) by an amount +𝜏 (i.e. towards right). Therefore correlation may also be defined
as
𝑇
2
…1.10.1b
𝑅1,2 (𝜏) = lim ∫ 𝑓1 (𝑡 − 𝜏)𝑓2 ∗ (𝑡)𝑑𝑡
𝜏→∞ −𝑇
2
Let us define the correlation for two cases, (i) Energy (non periodic) signal and (ii) Power (Periodic) Signals.
In the definition of correlation limit of integration may be taken as infinite for energy signals,
∞ ∞
…1.10.2a
𝑅1,2 (𝜏) = ∫ 𝑓1 (𝑡)𝑓2 ∗ (𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑓1 (𝑡 − 𝜏)𝑓2 ∗ (𝑡)𝑑𝑡
−∞ −∞
For power signals of period T0, the definition in above equation may not converge. Therefore the average
correlation over a period T0 is defined as
𝑇0 𝑇0
1 2 1 2
…1.10.2b
𝑅1,2 (𝜏) = ∫ 𝑇 𝑓1 (𝑡)𝑓2 ∗ (𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑇 𝑓1 (𝑡 − 𝜏)𝑓2 ∗ (𝑡)𝑑𝑡
T0 − 0 T0 − 0
2 2
The correlation definition represents the overlapping area between the two functions.

1.11 Auto Correlation Function

Auto correlation is a special form of cross correlation. It is defined as correlation of a function with itself.
Auto correlation function is a measure of similarity between a signal & its time delayed version. It is
represented with 𝑅(𝜏).
Consider a signal 𝑓(𝑡). The auto correlation function of 𝑓(𝑡) with its time delayed version is given by

𝑇
1
𝑅11 (𝜏) = 𝑅(𝜏) = lim T ∫ 2𝑇 𝑓 (𝑡)𝑓 ∗ (𝑡 + 𝜏)𝑑𝑡 (+ve Shift)
𝑇→∞ −
2

𝑇
1
= lim T ∫ 2𝑇 𝑓(𝑡 − 𝜏)𝑓 ∗ (𝑡)𝑑𝑡 (-ve Shift)
𝑇→∞ −
2

Properties of auto correlation

Auto correlation exhibits conjugate symmetry i.e. (𝑅(𝜏) = 𝑅 ∗ (−𝜏)

Auto correlation function of energy signal at origin i.e. at 𝜏 = 0 is equal to total energy of that signal, which
is given as:

Page 7 of 9

Page no: 7 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

∞
𝑅(0) = ∫ |𝑥(𝑡)|2 𝑑𝑡
−∞
1
Auto correlation function ∞ 𝑇 ,
Auto correlation function is maximum at 𝜏 = 0 𝑖. 𝑒. |𝑅(𝜏)| ≤ 𝑅(0)∀𝜏
Auto correlation function and energy spectral densities are Fourier transform pairs. i.e.

𝐹. 𝑇. |𝑅(𝜏)| = Ψ(𝜔)
∞
Ψ(𝜔) = ∫ 𝑅(𝜏)𝑒 −𝑗𝜔𝑡 𝑑𝜏
−∞
𝑅(𝜏) = 𝑥(𝑡) ∗ 𝑥(−𝑡)

1.12 Power Spectral Density

The function which describes how the power of a signal got distributed at various frequencies, in the
frequency domain is called as Power Spectral Density (PSD).
PSD is the Fourier Transform of Auto-Correlation. It is in the form of a rectangular pulse.

𝑇 sin(𝜋𝑇𝑓)
𝑠 (𝑡 ) = 𝐴 |𝑡 | < 𝑠(𝑓) = 𝐴𝑇
2 𝜋𝑇𝑓
𝑇
𝑠 (𝑡 ) = 0 |𝑡 | >
2
Figure 1.02 Power Spectral Density

1.13 Power Spectral Density Derivation

According to the Einstein-Wiener-Khintchine theorem, if the auto correlation function or power spectral
density of a random process is known, the other can be found exactly.
Hence, to derive the power spectral density, we shall use the time auto-correlation (Rx(τ)) of a power
signal x(t) as shown below.
𝑇𝑃
1 2
𝑅𝑥 (𝜏) = lim ∫ 𝑥 (𝑡)𝑥(𝑡 + 𝜏) 𝑑𝑡
𝑇𝑃 →∞ 𝑇𝑃 −𝑇𝑃
2

Since x(t) consists of impulses, Rx(τ) can be written as

∞
1
𝑅𝑥 [𝜏] = ∑ 𝑅𝑛 𝛿(𝜏 − 𝑛𝑇)
𝑇
𝑛=−∞
1
Where 𝑅𝑛 = lim ∑𝑘 𝑎𝑘 𝑎𝑘+𝑛
𝑁→∞ 𝑁
Getting to know that Rn=R−n for real signals, we have

Page 8 of 9

Page no: 8 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

∞
1
𝑆𝑥 (𝜔) = (𝑅0 + 2 ∑ 𝑅𝑛 𝑐𝑜𝑠𝑛𝜔𝑇)
𝑇
𝑛=1
Since the pulse filter has the spectrum of (w)↔f(t), we have
𝑆𝑦 (𝜔) = |𝐹(𝜔)|2 𝑆𝑥 (𝜔)
∞
|𝐹 (𝜔)|2
= ( ∑ 𝑅𝑛 𝑒 −𝑗𝑛𝑢𝑇𝑏 )
𝑇
𝑛=∞
∞
|𝐹 (𝜔)|2
= (𝑅0 + 2 ∑ 𝑅𝑛 𝑐𝑜𝑠𝑛𝜔𝑇)
𝑇
𝑛=1
Hence, we get the equation for Power Spectral Density. Using this, we can find the PSD of various line
codes.

Page 9 of 9

Page no: 9 Get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]
Program : B.Tech
Subject Name: Digital Communication
Subject Code: EC-502
Semester: 5th
Downloaded from www.rgpvnotes.in

Department of Electronics and Communication Engineering

Sub. Code: EC 5002 Sub. Name: Digital Communication

Unit 2
Digital conversion of Analog Signals: Sampling theorem, sampling of band pass signals, Pulse Amplitude
Modulation (PAM), types of sampling (natural, flat-top), equalization, signal reconstruction and
reconstruction filters, aliasing and anti-aliasing filter, Pulse Width Modulation (PWM), Pulse Position
Modulation (PPM)
Digital transmission of Analog Signals: Quantization, quantization error, Pulse Code Modulation (PCM),
companding, scrambling, TDM-PCM, Differential PCM, Delta modulation, Adaptive Delta modulation,
vocoder.

PART I DIGITAL CONVERSION OF ANALOG SIGNALS

2.1 Sampling of Analog Signals

Sampling is defined as, “The process of measuring the instantaneous values of continuous-time signal in a
discrete form.” In the process of sampling an analog signal is converted into a corresponding sequences of
samples, that are uniformly spaced in time.
This discretization of analog signal is called as Sampling. The following figure indicates a continuous-time
signal x (t) and a sampled signal xs (t). When x (t) is multiplied by a periodic impulse train, the sampled
signal xs (t) is obtained.

Figure 2.01 Sampling

2.1.1 Sampling Rate
To discretize the signals, the gap between the samples should be fixed. That gap can be termed as a
sampling period 𝑇𝑠.
𝑆𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 = 1/𝑇𝑠 = 𝑓𝑠
Where,
𝑇𝑠 = 𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑡𝑖𝑚𝑒
𝑓𝑠 = 𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑟 𝑡ℎ𝑒 𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑟𝑎𝑡𝑒

Sampling frequency is the reciprocal of the sampling period. This sampling frequency can be simply called
as Sampling rate. The sampling rate denotes the number of samples taken per second, or for a finite set of
values.

Page 1 of 24

Page no: 1 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

For an analog signal to be reconstructed from the digitized signal, the sampling rate should be highly
considered. The rate of sampling should be such that the data in the message signal should neither be lost
nor it should get over-lapped. Hence, a rate was fixed for this, called as Nyquist rate.

2.1.2 Signals Sampling Techniques

There are three types of sampling techniques:
a. Impulse sampling.
b. Natural sampling.
c. Flat Top sampling.

(a) Impulse Sampling

Impulse sampling can be performed by multiplying input signal x(t) with impulse train of period 'T'. Here,
the amplitude of impulse changes with respect to amplitude of input signal x(t). The output of sampler is
given by

Figure 2.1.2.1 Impulse Sampling

𝑦(𝑡) = 𝑥(𝑡) × 𝑖𝑚𝑝𝑢𝑙𝑠𝑒 𝑡𝑟𝑎𝑖𝑛

∞

= 𝑥(𝑡) × ∑ 𝑥(𝑛𝑡)𝛿 (𝑡 − 𝑛𝑡)

𝑛=−∞
∞

𝑦(𝑡) = 𝑦𝑛 (𝑡) = ∑ 𝑥(𝑛𝑡)𝛿 (𝑡 − 𝑛𝑡) … …(2.1.2.1)

𝑛=−∞
To get the spectrum of sampled signal, consider Fourier transform of equation 1 on both sides
∞
1
𝑌(𝜔) = ∑ 𝑋(𝜔 − 𝑛𝜔𝑠 )
𝑇0
𝑛=−∞
This is called ideal sampling or impulse sampling. You cannot use this practically because pulse width
cannot be zero and the generation of impulse train is not possible practically.
(b) Natural Sampling
Natural sampling is similar to impulse sampling, except the impulse train is replaced by pulse train of
period T. i.e. you multiply input signal x(t) to pulse train ∑∞
𝑛=−∞ 𝑃(𝑡 − 𝑛𝑇 ) as shown below

Figure 2.1.2.2 Natural Sampling

Page 2 of 24

Page no: 2 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝑦(𝑡) = 𝑥(𝑡) × 𝑝𝑢𝑙𝑠𝑒 𝑡𝑟𝑎𝑖𝑛

= 𝑥(𝑡) × 𝑝(𝑡)
∞

= 𝑥(𝑡) × ∑ 𝑃(𝑡 − 𝑛𝑇) …(2.1.2.2)

𝑛=−∞

The exponential Fourier series representation of p(t) can be given as

∞

𝑝(𝑡) = ∑ 𝐹𝑛 𝑒 𝑗𝑛𝜔0 𝑡 …(2.1.2.3)

𝑛=−∞

= ∑ 𝐹𝑛 𝑒 𝑗2𝜋𝑛𝑓0 𝑡
𝑛=−∞

𝑇
1
Where 𝐹𝑛 = 𝑇 ∫2𝑇 𝑝(𝑡)𝑒 𝑗𝑛𝜔0 𝑡 𝑑𝑡
−
2

1
= (𝑛𝜔𝑠 )
𝑇𝑃
Substitute Fn value in equation 2.1.2.2
∞
1
∴ p(t) = ∑ P(nωs )ejnω0 t
T
n=−∞
∞
1
= ∑ P(nωs )ejnω0 t
T
n=−∞

Substitute p(t) in equation 2.1.2.1

𝑦(𝑡) = 𝑥(𝑡) × 𝑝(𝑡)
∞
1
= 𝑥(𝑡) × ∑ 𝑃(𝑛𝜔𝑠 )𝑒 𝑗𝑛𝜔0 𝑡
𝑇
𝑛=−∞

∞
1
= ∑ 𝑃(𝑛𝜔𝑠 )𝑥(𝑡)𝑒 𝑗𝑛𝜔0 𝑡
𝑇
𝑛=−∞

To get the spectrum of sampled signal, consider the Fourier transform on both sides.
∞
1
𝐹. 𝑇. [y(t)] = F. T. ∑ [P(nωs )x(t)ejnω0 t ]
T
n=−∞
∞
1
= ∑ P(nωs )F. T. [x(t)ejnω0 t ]
T
n=−∞

According to frequency shifting property

F. T. [x(t)ejnω0 t ] = X[𝜔 − 𝑛𝜔𝑠 ]
∞
1
∴ Y[𝜔] = ∑ P(nωs ) X[𝜔 − 𝑛𝜔𝑠 ]
T
n=−∞

Page 3 of 24

Page no: 3 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

(c) Flat Top Sampling

During transmission, noise is introduced at top of the transmission pulse which can be easily removed if
the pulse is in the form of flat top. Here, the top of the samples are flat i.e. they have constant amplitude.
Hence, it is called as flat top sampling or practical sampling. Flat top sampling makes use of sample and
hold circuit.

Figure 2.1.2.3 Flat Top Sampling

Theoretically, the sampled signal can be obtained by convolution of rectangular pulse p(t) with ideally
sampled signal say yδ(t) as shown in the diagram:
i.e. 𝑦(𝑡) = 𝑝(𝑡) × 𝑦𝛿 (𝑡) …(2.1.2.4)

Figure 2.11 Sampling

To get the sampled spectrum, consider Fourier transform on both sides for equation 1
𝑌[𝜔] = 𝐹. 𝑇. [𝑃(𝑡) × 𝑦𝛿 (𝑡)]
By the knowledge of convolution property,
𝑌[𝜔] = 𝑃 (𝜔)𝑌𝛿 (𝜔)
𝜔𝑇
Here 𝑃(𝜔) = 𝑇𝑆𝑎 ( ) = 2𝑠𝑖𝑛𝜔𝑇/𝜔
2

2.1.3 Sampling Theorem

The sampling theorem states that, “a signal whose spectrum is band limited to B Hz, [G(ω) = 0 for |𝜔| >
2𝜋𝐵] can be reproduced exactly from its samples if it is sampled at the rate 𝑓𝑠 which is greater than twice
the maximum frequency ω of the signal to be sampled.” Therefore minimum sampling frequency is
𝑓𝑠 = 2𝐵 𝐻𝑧
Proof: Consider a continuous time signal x(t). The spectrum of x(t) is a band limited to f m Hz i.e. the
spectrum of x(t) is zero for |ω|>ωm.
Sampling of input signal x(t) can be obtained by multiplying x(t) with an impulse train δ(t) of period Ts. The
output of multiplier is a discrete signal called sampled signal which is represented with y(t) in the following
diagrams:

Page 4 of 24

Page no: 4 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 2.1.3.1 Sampling of signal x(t)

Here, you can observe that the sampled signal takes the period of impulse. The process of sampling can be
understood as under.
The sampled signal is given by
𝑦(𝑡) = 𝑥(𝑡). 𝛿 (𝑡) = 𝑥(𝑡). 𝛿𝑇𝑠 (𝑡)
= ∑ 𝑥(𝑛𝑇𝑠 ). 𝛿(𝑡−𝑛𝑇𝑠 ) ….(2.1.3.1)
𝑛
The impulse train 𝛿𝑇𝑠 (𝑡) is a periodic signal of period 𝑇𝑠 , hence it can be expressed as a Fourier series
1
𝛿𝑇𝑠 (𝑡) = [1 + 2 𝑐𝑜𝑠𝜔𝑠 𝑡 + 2 𝑐𝑜𝑠2𝜔𝑠 𝑡 + 2 𝑐𝑜𝑠3𝜔𝑠 𝑡 + ⋯ ]
𝑇𝑠
2𝜋
Where 𝜔𝑠 = 𝑇 = 2𝜋𝑓𝑠 ….(2.1.3.2)
𝑠

Substitute 𝛿 (𝑡)in equation 1.

→ 𝑦 (𝑡 ) = 𝑥 (𝑡 ). 𝛿 (𝑡 )
1
𝑦(𝑡) = [𝑥(𝑡) + 2𝑥(𝑡) 𝑐𝑜𝑠𝜔𝑠 𝑡 + 2𝑥(𝑡) 𝑐𝑜𝑠2𝜔𝑠 𝑡 + 2 𝑥(𝑡)𝑐𝑜𝑠3𝜔𝑠 𝑡 + ⋯ ]
𝑇𝑠
Now to find the 𝑌(𝜔), we have to take the fourier transform of both the sides,
1
𝑌(𝜔) = [𝑋(𝜔) + 𝑋(𝜔 − 𝜔𝑠 ) + 𝑋(𝜔 + 𝜔𝑠 ) + 𝑋(𝜔 − 2𝜔𝑠 ) + 𝑋(𝜔 + 2𝜔𝑠 ) + ⋯ ]
𝑇𝑠
1
𝑌(𝜔) = 𝑇 ∑∞ 𝑛=−∞ 𝑋(𝜔 − 𝑛𝜔𝑠 ) Where 𝑛 = 0, ±1, ±2, …
𝑠
The Fourier spectrum 𝑌(𝜔) is shown in figure 2.1.3.1. Now If we have to recover 𝑥(𝑡) from 𝑦(𝑡), we
should be able to recover 𝑋(𝜔) from 𝑌(𝜔), and it is possible only if there is no overlapping between the
successive cycles of 𝑌(𝜔), and for this condition
𝑓𝑠 > 2𝐵 𝐻𝑧
Therefore the sampling interval
1
𝑇𝑠 <
2𝐵

Therefore as long as the sampling frequency 𝑓𝑠 is greater than 2𝐵, 𝑌(𝜔), will consist of non overlapping
repetitions of 𝑋(𝜔), and 𝑥(𝑡) can be recovered from 𝑦(𝑡) by passing 𝑦(𝑡) by an ideal low pass filter with
cut off frequency B Hz.
Nyquist Rate
The minimum sampling rate 𝑓𝑠 = 2𝐵 required to recover 𝑥(𝑡) from its samples 𝑦(𝑡) is called the Nyquist
rate and the corresponding sampling interval is called Nyquist Interval for 𝑦(𝑡).

Page 5 of 24

Page no: 5 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Effect of Sampling Rate on 𝒀(𝝎)

Possibility of sampled frequency spectrum with different conditions is given by the following diagrams:

(a) Over Sampling

(b) Under Sampling

(c) Perfect sampling

Figure 2.1.3.2 Sampling Conditions

Aliasing Effect
The overlapped region in case of under sampling represents aliasing effect, which can be removed by
 considering 𝑓𝑠 > 2𝑓𝑚
 By using anti-aliasing filters.

2.2 Sampling of Band Pass Signals

In case of band pass signals, the spectrum of band pass signal X[ω] = 0 for the frequencies outside the
range f1 ≤ f ≤ f2. The frequency f1 is always greater than zero. Plus, there is no aliasing effect when f s > 2f2.
But it has two disadvantages:
 The sampling rate is large in proportion with f2. This has practical limitations.
 The sampled signal spectrum has spectral gaps.
To overcome this, the band pass theorem states that the input signal x(t) can be converted into its samples
and can be recovered back without distortion when sampling frequency 𝑓𝑠 < 2𝑓2.
Also
1 2𝑓2
𝑓𝑠 = =
𝑇 𝑚
Where m is the largest integer < f2 /B and B is the bandwidth of the signal. If f2=KB, then for band pass
signals of bandwidth 2fm and the minimum sampling rate fs=2B=4fm, the spectrum of the sampled signal is
given by
∞
1
𝑌[𝜔] = ∑ 𝑋[𝜔 − 2𝑛𝐵]
𝑇
𝑛=−∞

Aliasing
Aliasing can be referred to as “the phenomenon of a high-frequency component in the spectrum of a
signal, taking on the identity of a low-frequency component in the spectrum of its sampled version.”
The corrective measures taken to reduce the effect of Aliasing are −
 In the transmitter section of PCM, a low pass anti-aliasing filter is employed, before the sampler, to
eliminate the high frequency components, which are unwanted.

Page 6 of 24

Page no: 6 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The signal which is sampled after filtering is sampled at a rate slightly higher than the Nyquist rate.
This choice of having the sampling rate higher than Nyquist rate, also helps in the easier design of the
reconstruction filter at the receiver.

2.3 Types of Modulation:

In pulse width modulation, there are different types of modulation for analog and digital as shown below:
 PCM: Pulse Code Modulation for Analog Modulation.
 PPM: Pulse Position Modulation for Digital Modulation
 PDM: Pulse Duration Modulation for Digital Modulation.
 PAM: Pulse Amplitude Modulation for Digital Modulation.
Types of Modulation – Tree Diagram:
Types of Modulation

Continuous Wave Modulation Pulse Digital Modulation

Amplitude Modulation Angular Modulation Digital Modulation Analog Modulation

Frequency Modulation Phase Modulation

Pulse Code Modulation

Pulse Amplitude Modulation Pulse Duration Modulation Pulse Position Modulation

Figure 2.3.1 Types of Modulation

PCM is an important method of analog –to-digital conversion. In this modulation

the analog signal is converted into an electrical waveform of two or more levels. A
simple two level waveform is shown in fig 2.3.2.

Figure 2.3.2 A Simple Binary PCM Waveform

The PCM system block diagram is shown in fig 3.2. The essential operations in the transmitter of a PCM
system are Sampling, Quantizing and Coding. The Quantizing and
encoding operations are usually performed by the same circuit, normally referred to as
analog to digital converter. The essential operations in the receiver are regeneration, decoding and

Page 7 of 24

Page no: 7 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

demodulation of the quantized samples. Regenerative repeaters are used to reconstruct the transmitted
sequence of coded pulses in order to combat the accumulated effects of signal distortion and noise.

2.4 Pulse Amplitude Modulation (PAM):

In pulse amplitude modulation, the amplitude of regular interval of periodic pulses or electromagnetic
pulses is varied in proposition to the sample of modulating signal or message signal. This is an analog type
of modulation. In the pulse amplitude modulation, the message signal is sampled at regular periodic or
time intervals and this each sample is made proportional to the magnitude of the message signal. These
sample pulses can be transmitted directly using wired media or we can use a carrier signal for transmitting
through wireless. There are two types of sampling techniques for transmitting messages using pulse
amplitude modulation, they are
 FLAT TOP PAM: The amplitude of each pulse is directly proportional to instantaneous modulating
signal amplitude at the time of pulse occurrence and then keeps the amplitude of the pulse for the
rest of the half cycle.
 Natural PAM: The amplitude of each pulse is directly proportional to the instantaneous modulating
signal amplitude at the time of pulse occurrence and then follows the amplitude of the modulating
signal for the rest of the half cycle.
Flat top PAM is the best for transmission because we can easily remove the noise and we can also easily
recognize the noise. When we compare the difference between the flat top PAM and natural PAM, flat top
PAM principle of sampling uses sample and hold circuit. In natural principle of sampling, noise interference
is minimum. But in flat top PAM noise interference maximum. Flat top PAM and natural PAM are practical
and sampling rate satisfies the sampling criteria.
There are two types of pulse amplitude modulation based on signal polarity
1. Single polarity pulse amplitude modulation, 2. Double polarity pulse amplitude modulation
In single polarity pulse amplitude modulation, there is fixed level of DC bias added to the message signal or
modulating signal, so the output of modulating signal is always positive. In the double polarity pulse
amplitude modulation, the output of modulating signal will have both positive and negative ends.

Figure 2.4.1 Pulse Amplitude Modulation

Advantages of Pulse Amplitude Modulation (PAM):
 It is the base for all digital modulation techniques and it is simple process for both modulation and
demodulation technique.
 No complex circuitry is required for both transmission and reception. Transmitter and receiver
circuitry is simple and easy to construct.

Page 8 of 24

Page no: 8 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

 PAM can generate other pulse modulation signals and can carry the message or information at
same time.

Disadvantages of Pulse Amplitude Modulation (PAM):

 Bandwidth should be large for transmitting the pulse amplitude modulation signal. Due to Nyquist
criteria also high bandwidth is required.
 The frequency varies according to the modulating signal or message signal. Due to these variations
in the signal frequency, interferences will be there. So noise will be great. For PAM, noise immunity
is less when compared to other modulation techniques. It is almost equal to amplitude modulation.
 Pulse amplitude signal varies, so power required for transmission will be more, peak power is also,
even at receiving more power is required to receive the pulse amplitude signal.

Applications of Pulse Amplitude Modulation (PAM):

 It is mainly used in Ethernet which is type of computer network communication, we know that we
can use Ethernet for connecting two systems and transfer data between the systems. Pulse
amplitude modulation is used for Ethernet communications.
 It is also used for photo biology which is a study of photosynthesis.
 Used as electronic driver for LED lighting.
 Used in many micro controllers for generating the control signals etc.

2.5 Pulse Position Modulation (PPM):

In the pulse position modulation, the position of each pulse in a signal by taking the reference signal is
varied according to the sample value of message or modulating signal instantaneously. In the pulse
position modulation, width and amplitude is kept constant. It is a technique that uses pulses of the same
breath and height but is displaced in time from some base position according to the amplitude of the signal
at the time of sampling. The position of the pulse is 1:1 which is propositional to the width of the pulse and
also propositional to the instantaneous amplitude of sampled modulating signal. The position of pulse
position modulation is easy when compared to other modulation. It requires pulse width generator and
mono stable multi vibrator.
Pulse width generator is used for generating pulse width modulation signal which will help to trigger the
mono stable multi vibrator; here trial edge of the PWM signal is used for triggering the mono stable multi
vibrator. After triggering the mono stable multi vibrator, PWM signal is converted into pulse position
modulation signal. For demodulation, it requires reference pulse generator, flip-flop and pulse width
modulation demodulator.

Advantages of Pulse Position Modulation (PPM):

 Pulse position modulation has low noise interference when compared to PAM because amplitude
and width of the pulses are made constant during modulation.
 Noise removal and separation is very easy in pulse position modulation.
 Power usage is also very low when compared to other modulations due to constant pulse
amplitude and width.
Disadvantages of Pulse Position Modulation (PPM):
 The synchronization between transmitter and receiver is required, which is not possible for every
time and we need dedicated channel for it.
 Large bandwidth is required for transmission same as pulse amplitude modulation.
 Special equipments are required in this type of modulations.
Applications of Pulse Position Modulation (PPM):
 Used in non-coherent detection where a receiver does not need any Phase lock loop for tracking
the phase of the carrier.
 Used in radio frequency (RF) communication.
 Also used in contactless smart card, high frequency, RFID (radio frequency ID) tags and etc.

Page 9 of 24

Page no: 9 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

2.6 Pulse Duration Modulation (PDM) or Pulse Width Modulation (PWM):

It is a type of analog modulation. In pulse width modulation or pulse duration modulation, the width of the
pulse carrier is varied in accordance with the sample values of message signal or modulating signal or
modulating voltage. In pulse width modulation, the amplitude is made constant and width of pulse and
position of pulse is made proportional to the amplitude of the signal. We can vary the pulse width in three
ways
1. By keeping the leading edge constant and vary the pulse width with respect to leading edge
2. By keeping the tailing constant.
3. By keeping the center of the pulse constant.
We can generate pulse width using different circuitry. In practical, we use 555 Timer which is the best way
for generating the pulse width modulation signals. By configuring the 555 timer as mono stable or a stable
multi vibrator, we can generate the PWM signals. We can use PIC, 8051, AVR, ARM, etc. microcontrollers
to generate the PWM signals. PWM signal generation has n number of ways. In demodulation, we need
PWM detector and its related circuitry for demodulating the PWM signal.

Advantages of Pulse Width Modulation (PWM):

 As like pulse position modulation, noise interference is less due to amplitude has been made
constant.
 Signal can be separated very easily at demodulation and noise can also be separated easily.
 Synchronization between transmitter and receiver is not required unlike pulse position modulation.
Disadvantages of Pulse Width Modulation (PWM):
 Power will be variable because of varying in width of pulse. Transmitter can handle the power even
for maximum width of the pulse.
 Bandwidth should be large to use in communication, should be huge even when compared to the
pulse amplitude modulation.
Applications of Pulse Width Modulation (PWM):
 PWM is used in telecommunication systems.
 PWM can be used to control the amount of power delivered to a load without incurring the losses.
So, this can be used in power delivering systems.
 Audio effects and amplifications purposes also used.
 PWM signals are used to control the speed of the robot by controlling the motors.
 PWM is also used in robotics.
 Embedded applications.
 Analog and digital applications etc.

PART II DIGITAL TRANSMISSION OF ANALOG SIGNALS

2.7 Quantization
In the process of quantization we create a new signal 𝑚𝑞 (𝑡), which is an approximation to 𝑚(𝑡). The
quantized signal 𝑚𝑞 (𝑡), has the great merit that it is separable from the additive noise.
The operation of quantization is represented in figure 2.7.1. Here we have a signal m(t), whose amplitude
varies in the range from VH to VL as shown in the figure.
We have divided the total range in to M equal intervals each of size S, called the step size and given by
(𝑉𝐻 − 𝑉𝐿 )
𝑆 = ∆=
𝑀
In our example M=8. In the centre of each of this step we located quantization levels 𝑚0, 𝑚1, 𝑚2, … 𝑚7.
The 𝑚𝑞 (𝑡) is generated in the following manner-
Whenever the signal 𝑚(𝑡) is in the range ∆0 , the signal 𝑚𝑞 (𝑡) maintains a constant level 𝑚0 , whenever
the signal 𝑚(𝑡) is in the range ∆1 , the signal 𝑚𝑞 (𝑡) maintains a constant level 𝑚1 and so on. Hence the

Page 10 of 24

Page no: 10 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

signal 𝑚𝑞 (𝑡) will found all times to one of the levels 𝑚0, 𝑚1, 𝑚2, … 𝑚7. The transition in 𝑚𝑞 (𝑡) from
𝑚0 to 𝑚1 is made abruptly when 𝑚(𝑡) passes the transition level 𝐿01 , which is mid way between 𝑚0
and 𝑚1 and so on.

Figure 2.7.2 Quantization Process

Using quantization of signals, the effect of noise can be reduced significantly. The difference between 𝑚(𝑡)
and 𝑚𝑞 (𝑡) can be regarded as noise and is called quantization noise.
𝑞𝑢𝑎𝑛𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛 𝑛𝑜𝑖𝑠𝑒 = 𝑚(𝑡) − 𝑚𝑞 (𝑡)
Also the quantized signal and original signal differs from one another in a ransom manner. This difference
or error due to quantization process is called quantization error and is given by
𝑒 = 𝑚 ( 𝑡 ) − 𝑚𝑘
( )
when 𝑚 𝑡 happens to be close to quantization level 𝑚𝑘 , quantizer output will be 𝑚𝑘 .

The process of transforming sampled amplitude values of a message signal into a discrete amplitude value
is referred to as Quantization. The quantization Process has a two-fold effect:
1. the peak-to-peak range of the input sample values is subdivided into a finite set of decision levels or
decision thresholds that are aligned with the risers of the staircase, and
2. The output is assigned a discrete value selected from a finite set of representation levels that are aligned
with the treads of the staircase.
A quantizer is memory less in that the quantizer output is determined only by the value of a corresponding
input sample, independently of earlier analog samples applied to the input.

Types of Quantizers:
1. Uniform Quantizer
2. Non- Uniform Quantizer
0 Ts 2Ts 3Ts Time Analog Signal Discrete Samples (Quantized)

In Uniform type, the quantization levels are uniformly spaced, whereas in non-uniform type the spacing
between the levels will be unequal and mostly the relation is logarithmic. Types of Uniform Quantizers:
(based on I/P - O/P Characteristics)
1. Mid-Rise type Quantizer

Page 11 of 24

Page no: 11 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

2. Mid-Tread type Quantizer

In the stair case like graph, the origin lies the middle of the tread portion in Mid –Tread type whereas the
origin lies in the middle of the rise portion in the Mid-Rise type. Mid – tread type: Quantization levels – odd
number. Mid – Rise type: Quantization levels – even number.

Figure 2.7.2 IO Characteristics of Mid-Rise type Quantizer

`
Figure 2.7.3 IO Characteristics of Mid-Tread type Quantizer

2.7.1 Quantization Noise and Signal-to-Noise:

“The Quantization process introduces an error defined as the difference between the input signal, x(t) and
the output signal, y(t). This error is called the Quantization Noise.”

𝑞 (𝑡) = 𝑥(𝑡) − 𝑦(𝑡)

Quantization noise is produced in the transmitter end of a PCM system by rounding off sample values of an
analog base-band signal to the nearest permissible representation levels of the quantizer. As such
quantization noise differs from channel noise in that it is signal dependent.
Let ‘Δ’ be the step size of a quantizer and L be the total number of quantization levels.
Quantization levels are 0, ± Δ., ± 2 Δ., ±3 Δ . . . . . . .
The Quantization error, Q is a random variable and will have its sample values bounded
by [-(Δ/2) < q < (Δ/2)]. If Δ is small, the quantization error can be assumed to a
uniformly distributed random variable.

Page 12 of 24

Page no: 12 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Consider a memory less quantizer that is both uniform and symmetric.

L = Number of quantization levels
X = Quantizer input
Y = Quantizer output
The output y is given by
𝑌 = 𝑄(𝑥) …(2.7.1.1)
it is a staircase function that befits the type of mid tread or mid riser quantizer of interest. Suppose that
the input ‘X’ lies inside the interval
𝐼𝑘 = {𝑋𝑘 < 𝑋 ≤ 𝑋𝑘−1 } k=1,2,….L …(2.7.1.2)
Where Xk and Xk−1 are the decision thresholds of the intervals Ik as shown in figure 2.7.1.1.
Correspondingly, the quantizer output y takes on a
discrete value
Y = 𝑦𝑘 if x lies in the interval Ik.
Let q = quantization error with values in the range
∆ ∆
− 2 ≤ 𝑞 ≤ 2 then
Yk = 𝑥 + 𝑞 if ‘n’ lies in the interval Ik
Assuming that the quantizer input ‘n’ is the sample
value of a random variable ‘X’ of zero mean with
variance𝜎𝑥 2 . Figure 2.7.1.1 Decision Thresholds
The quantization noise uniformly distributed out the signal band, its interfering effect on a signal is similar
to that of thermal noise.

2.7.2 Expression for Quantization Noise and SNR in PCM:-

Let Q = Random Variable denotes the Quantization error
q = Sampled value of Q
Assuming that the random variable Q is uniformly distributed over the possible range
(−∆/2 ≤ 𝑞 ≤ ∆/2), as
1 …(2.7.1.3)
𝑓𝑄 (𝑞) = {∆ − ∆/2 ≤ 𝑞 ≤ ∆/2
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Where fQ(q) = probability density function of the Quantization error. If the signal does not overload the
Quantizer, then the mean of Quantization error is zero and the variance 𝜎𝑥 2 .

Therefore
𝜎𝑄 2 = 𝐸{𝑄2 }
∞
𝜎𝑄 2 = ∫ 𝑞 2 𝑓𝑞 (𝑞 )𝑑𝑞 …(2.7.1.4)
−∞

1 ∆/2 2 ∆2
𝜎𝑄 2 = ∫ 𝑞 𝑑𝑞 = …(2.7.1.5)
∆ −∆/2 12

Page 13 of 24

Page no: 13 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Thus the varience of the Quantization noise produced by a Uniform Quantizer, grows as the square of the
step size. Equation (2.7.1.5) gives an expression for Quantization noise in PCM system.
Let 𝜎𝑥 2 = Variance of the base band signal x(t) at the input of the quantizer.
When the base band signal is reconstructed at the receiver output, we obtain original signal plus
Quantization noise. Therefore output signal to Quantization noise ration (SNR) is given by

𝑆𝑖𝑔𝑛𝑎𝑙 𝑃𝑜𝑤𝑒𝑟 𝜎𝑥 2 𝜎𝑥 2
(𝑆𝑁𝑅)𝑄 = = 2= 2 …(2.7.1.6)
𝑁𝑜𝑖𝑠𝑒 𝑃𝑜𝑤𝑒𝑟 𝜎𝑄 ∆ /12
Smaller the step size ∆, larger will be the SNR.

2.7.3 Signal to Quantization Noise Ration:- [Mid Tread Type]

Let x = Quantizer input, sampled value of random variable X with mean X, variance 𝜎𝑥 2. The Quantizer is
assumed to be uniform, symmetric and mid trade type.
Xmax=absolute value of the overload level of the Quantizer.
∆= Step size,
Then L= No. of Quantization level given by
2𝑋𝑚𝑎𝑥
𝐿= +1 …(2.7.1.7)
∆
Let n = no. of bits used to represent each level.
In general 2n = L, but in the mid trade quantizer, since the number of representation levels is odd,
𝐿 = 2𝑛 − 1
…(2.7.1.8)

From equations (2.7.1.7) and (2.7.1.8),

2𝑋𝑚𝑎𝑥
2𝑛 − 1 = +1
∆
Or
𝑋𝑚𝑎𝑥
∆= 𝑛−1
…(2.7.1.9)
2 − 1
𝑋𝑚𝑎𝑥
The ration is called the loading factor. To avoid significant overload distortion, the amplitude of the
𝜎𝑥
Quantizer input x extend from −4𝜎𝑥 to 4𝜎𝑥 , which correspond to loading factor of 4. Thus with 𝑋𝑚𝑎𝑥 =
4𝜎𝑥 , we can write equation (2.7.1.9) as,
4𝜎𝑥
∆= 𝑛−1 …(2.7.1.10)
2 −1
𝜎𝑥 2 3
(𝑆𝑁𝑅)𝑄 = 2 = [2𝑛−1 − 1]2 …(2.7.1.11)
∆ /12 4
For larger value of n(typically n>6), we may approximate the result as
3 3 2𝑛
(𝑆𝑁𝑅)𝑄 = [2𝑛−1 − 1]2 ≈ (2 ) …(2.7.1.12)
4 16
Hence expressing the SNR in db
10𝑙𝑜𝑔10 (𝑆𝑁𝑅)𝑄 = 6𝑛 − 7.2
…(2.7.1.13)
This formula states that each bit in code word of a PCM system contributes 6db to the
signal to noise ratio. For loading factor of 4, the problem of overload i.e. the problem that the sampled
value of signal falls outside the total amplitude range of Quantizer, 8σx is less than 10-4.
The equation 2.7.1.11 gives a good description of the noise performance of a PCM
system provided that the following conditions are satisfied.
1. The Quantization error is uniformly distributed

Page 14 of 24

Page no: 14 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

2. The system operates with an average signal power above the error threshold so
that the effect of channel noise is made negligible and performance is there by
limited essentially by Quantization noise alone.
3. The Quantization is fine enough (say n>6) to prevent signal correlated patterns in
the Quantization error waveform
4. The Quantizer is aligned with input for a loading factor of 4
Note: 1. Error uniformly distributed
2. Average signal power
3. n > 6
4. Loading factor = 4
From (2.7.1.13): 10 log10 (SNR)O = 6n – 7.2

In a PCM system, Bandwidth B = nW or [n=B/W], substituting the value of ‘n’ we get

𝐵
10𝑙𝑜𝑔10 (𝑆𝑁𝑅)𝑄 = 6( ) − 7.2 …(2.7.1.14)
𝑊

2.7.4 Signal to Quantization Noise Ratio:- [Mid Rise Type]

Let x = Quantizer input, sampled value of random variable X with mean X, variance 𝜎𝑥 2. The Quantizer is
assumed to be uniform, symmetric and mid rise type.
Let Xmax=absolute value of the overload level of the Quantizer.
2𝑋𝑚𝑎𝑥
𝐿= …(2.7.1.15)
∆
Since the number of representation levels is odd,
𝐿 = 2𝑛 (Mid rise only) …(2.7.1.16)

From equations (2.7.1.15) and (2.7.1.16),

𝑋𝑚𝑎𝑥
∆= …(2.7.1.17)
2𝑛
𝜎𝑥 2
(𝑆𝑁𝑅)𝑄 = 2 …(2.7.1.18)
∆ /12
Where 𝜎𝑥 2 represent the variance or the signal power.

Consider a special case of sinusoidal signals:

2
Let the signal power be Ps, then 𝑃𝑠 = 0.5𝑋𝑚𝑎𝑥
𝑃𝑠 12𝑃𝑠
(𝑆𝑁𝑅)𝑄 = 2
= = 1.5𝐿2 = 1.5𝐿2𝑛 …(2.7.1.19)
∆ /12 ∆2
In decibels (𝑆𝑁𝑅)𝑄 = 1.76 + 6.02𝑛
…(3.20)
Improvement of SNR can be achieved by increasing the number of bits, n. Thus for ‘n’ number of bits/
sample the SNR is given by the above equation 2.7.1.19. For every increase of one bit / sample the step
size reduces by half. Thus for (n+1) bits the SNR is given by
(𝑆𝑁𝑅)(𝑛+1)𝑏𝑖𝑡 = (𝑆𝑁𝑅)(𝑛)𝑏𝑖𝑡 + 6 𝑑𝐵

Therefore addition of each bit increases the SNR by 6dB.

2.7.5 Classification of Quantization Noise:

The Quantizing noise at the output of the PCM decoder can be categorized into four types depending on
the operating conditions, Overload noise, Random noise, Granular Noise and Hunting noise
Over Load Noise:- The level of the analog waveform at the input of the PCM encoder needs to be set so
that its peak value does not exceed the design peak of V max volts. If the peak input does exceed Vmax, then

Page 15 of 24

Page no: 15 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

the recovered analog waveform at the output of the PCM system will have flat – top near the peak values.
This produces overload noise.
Granular Noise:- If the input level is reduced to a relatively small value w.r.to to the design level
(quantization level), the error values are not same from sample to sample and the noise has a harsh sound
resembling gravel being poured into a barrel. This is granular noise. This noise can be randomized (noise
power decreased) by increasing the number of quantization levels i.e. Increasing the PCM bit rate.
Hunting Noise:- This occurs when the input analog waveform is nearly constant. For these conditions, the
sample values at the Quantizer output can oscillate between two adjacent quantization levels, causing an
undesired sinusoidal type tone of frequency (0.5fs) at the output of the PCM system. This noise can be
reduced by designing the quantizer so that there is no vertical step at constant value of the inputs.

2.7.6 Quantization Error

For any system, during its functioning, there is always a difference in the values of its input and output. The
processing of the system results in an error, which is the difference of those values.
The difference between an input value and its quantized value is called a Quantization Error. A Quantizer is
a logarithmic function that performs Quantization (rounding off the value). An analog-to-digital converter
(ADC) works as a quantizer.
The following figure illustrates an example for a quantization error, indicating the difference between the
original signal and the quantized signal.

Figure 2.7.6.1 Quantization Error

Quantization Noise
It is a type of quantization error, which usually occurs in analog audio signal, while quantizing it to digital.
For example, in music, the signals keep changing continuously, where regularity is not found in errors. Such
errors create a wideband noise called as Quantization Noise.

2.8 Pulse Code Modulation:

A signal which is to be quantized before transmission is sampled as well. The quantization is used to reduce
the effect of noise and the sampling allows us to do the time division multiplexing. The combined
operation of sampling and quantization generate a quantized PAM waveform i.e. a train of pulses whose
amplitude is restricted to a number of discrete levels.
Rather than transmitting the sampled values itself, we may represent each quantization level by a code
number and transmit the code number. Most frequently the code number is converted in to binary

Page 16 of 24

Page no: 16 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

equivalent before transmission. Then the digits of the binary representation of the code are transmitted as
pulses. This system of transmission is called binary Pulse Code Modulation. The whole process can be
understood by the following diagram.

Figure 2.8.1 Pulse Code Modulation

PCM Transmitter:
Basic Blocks:
1. Anti aliasing Filter, 2. Sampler, 3. Quantizer, 4. Encoder
The block diagram of a PCM transmitter is shown in figure (a). An anti-aliasing filter is basically a filter used
to ensure that the input signal to sampler is free from the unwanted frequency components. For most of
the applications these are low-pass filters. It removes the frequency components of the signal which are
above the cutoff frequency of the filter. The cutoff frequency of the filter is chosen such it is very close to
the highest frequency component of the signal.
The message signal is sampled at the Nyquist rate by the sampler. The sampled pulses are then quantized
by the quantizer. The encoder encodes these quantized pulses in to binary equivalent, which are then
transmitted over the channel. During the channel the regenerative repeaters are used to maintain the
signal to noise ratio.
Continuous time
PCM Wave
message signal

LPF Sampler Quantizer Encoder

(A) Transmitter

Distorted Regenerative Regenerative

PCM wave Repeater Repeater

(b) Transmission Path

Page 17 of 24

Page no: 17 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Input
Message
Quantizer Holding Circuit LPF
Decoder

(c) Receiver

Figure 2.8.2 PCM System Basic Block Diagram

Figure (c) shows the receiver. The first block is again the quantizer, but this quantizer is different from the
transmitter quantizer sa it has to take the decision regarding the presence or absence of the pulse only.
Thus there are only two quantization levels. The output of the quantizer goes to the decoder which is an
D/A converter that performs the inverse operation of the encoder. The decoder output is a sequence of
quantized pulses. The original signal is reconstructed in the holding circuit and the LPF.

Figure 2.8.3 PCM Encoding

2.8.1 Advantages of Pulse Code Modulation:

 Pulse code modulation will have low noise addition and data loss is also very low.
 We can repeat the exact transmitted signal at the receiver. This is called repeatability. And we can
retransmit the signal with any distortion loss also.
 Pulse code modulation is used in music play back CD’s and also used in DVD for data storing whose
sampling rate is bit higher.
 Pulse code modulation can be used in storing the data.
 PCM can encode the data also.
 Multiplexing of signals can also be done using pulse code modulation. Multiplexing is nothing for
adding the different signals and transmitting the signal at same time.
 Pulse code modulation requires large bandwidth
 Pulse code modulation permits the use of pulse regeneration.

2.8.2 Disadvantages of Pulse Code Modulation:

 Specialized circuitry is required for transmitting and also for quantizing the samples at same
quantized levels. We can do encoding using pulse code modulation but we need to have complex
and special circuitry.

Page 18 of 24

Page no: 18 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

 Pulse code modulation receivers are cost effective when we compared to other modulation
receivers.
 Developing pulse code modulation is bit complicated and checking the transmission quality is also
difficult and takes more time.
 Large bandwidth is required for pulse code modulation when compared to bandwidth used by the
normal analog signals to transmit message.
 Channel bandwidth should be more for digital encoding.
 PCM systems are complicated when compared to analog modulation methods and other systems.
 Decoding also needs special equipment’s and they are also too complex.

2.8.3 Applications of Pulse Code Modulation (PCM):

 Pulse code modulation is used in telecommunication systems, air traffic control systems etc.
 Pulse code modulation is used in compressing the data that is why it is used in storing data in
optical disks like DVD, CDs etc. PCM is even used in the database management systems.
 Pulse code modulation is used in mobile phones, normal telephones etc.
Remote controlled cars, planes, trains use pulse code modulations.

2.9 Companding in PCM

The word Companding is a combination of Compressing and Expanding, which means that it does both.
This is a non-linear technique used in PCM which compresses the data at the transmitter and expands the
same data at the receiver. The effects of noise and crosstalk are reduced by using this technique.
There are two types of Companding techniques. They are −
A-law Companding Technique
 Uniform quantization is achieved at A = 1, where the characteristic curve is linear and no
compression is done.
 A-law has mid-rise at the origin. Hence, it contains a non-zero value.
 A-law companding is used for PCM telephone systems.

2.9.1 µ-law Companding Technique

 Uniform quantization is achieved at µ = 0, where the characteristic curve is linear and no
compression is done.
 µ-law has mid-tread at the origin. Hence, it contains a zero value.
 µ-law companding is used for speech and music signals.
µ-law is used in North America and Japan.
For the samples that are highly correlated, when encoded by PCM technique, leave redundant information
behind. To process this redundant information and to have a better output, it is a wise decision to take a
predicted sampled value, assumed from its previous output and summarize them with the quantized
values. Such a process is called as Differential PCM (DPCM) technique.

2.10 Differential Pulse Code Modulation:

Differential Pulse Code Modulation is an alternative to PCM. Instead of transmitting the sampled values
itself at each sampling time; we can transmit the difference between the two successive samples. For
example, we can transmit the difference between the sample value 𝑚(𝑘) at sampling time K and sample
value 𝑚(𝑘 − 1) at sampling time k-1. If such changes are transmitted then at the receiving end we can
generate a waveform identical to the m(t) by simply adding up these changes.
The DPCM has the special merit that when these differences are transmitted by PCM. The differences
𝑚(𝑘) − 𝑚(𝑘 − 1) will be smaller than the sample values themselves and fewer levels will be required to

Page 19 of 24

Page no: 19 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

quantize 𝑚(𝑘), and corresponding fewer bits will be needed to encode the signal. The basic principle of
DPCM is shown in figure 2.10.1.

(a) Transmitter

(b) Receiver
Figure 2.10.1 Differential PCM

The receiver consists of an accumulator which adds-up the receiver quantized differences ∆𝑄 (𝑘) and a
filter which smoothes out the quantization noise. The output of accumulator is the signal approximation
𝑚̂ (𝑘) which becomes 𝑚 ̂ (𝑡) at the filter output.
At the transmitter we need to know whether the 𝑚 ̂ (𝑡) is larger or smaller than 𝑚(𝑡) and by how much
amount. We may than determine whether the next difference ∆𝑄 (𝑘) needs to be positive or negative and
of what amplitude in order to bring 𝑚 ̂ (𝑡) as close as possible to 𝑚(𝑡). For this reason we have a duplicate
accumulator at transmitter.
At each sampling time the transmitter difference amplifier compares 𝑚(𝑡) and ̂(𝑡), 𝑚 and the sample and
hold circuit holds the result of that comparison ∆(𝑡), for the duration of interval between sampling times.
The quantizer generates the signal 𝑆0 (𝑡) = ∆𝑄 (𝑘) both for the transmission to the receiver and to provide
the input to the receiver accumulator in the transmitter.
The basic limitation of the DPCM scheme is that the transmitted differences are quantized and are of
limited values.
Need for a predictor:
There is a correlation between the successive samples of the signal 𝑚(𝑡). To take the advantage of this
correlation a predictor is included. It needs to incorporate the facility for storing past differences and
carrying out some algorithm to predict then next required increment.

2.11 Delta Modulation:

Delta Modulation is a DPCM scheme in which the difference signal ∆(𝑡) is encoded into just a single bit.
The single bit providing just for two possibilities is used to increase or decrease the estimate ̂𝑚(𝑡)[𝑚𝑞 (𝑡)].
The Linear Delta Modulator is shown in figure 2.11.1.
The baseband signal 𝑚(𝑡) and its quantized approximation ̂ 𝑚(𝑡) are applied as input to a comparator. The
comparator has one fixed output V(H) when 𝑚(𝑡) > 𝑚𝑞 (𝑡) and a difference output V(L) when 𝑚(𝑡) <
𝑚𝑞 (𝑡). Ideally the transition between V(H) and V(L) is arbitrarily abrupt as 𝑚(𝑡) − 𝑚𝑞 (𝑡) passes through

Page 20 of 24

Page no: 20 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

zero. The up-down counter increments or

decrements its count by 1 at each active edge of
the clock waveform. The count direction i.e.
incrementing or decrementing is determined by
the voltage levels at the “Count direction
command” input to the counter. When this binary
input is at level V(H), the counter counts up and
when this binary input is at level V(L), the counter
counts down.
The digital output of the counter is converted into
analog quantized approximation 𝑚𝑞 (𝑡) by a D/A
converter. The waveforms for the delta
modulator of figure 2.11.1 is shown in figure
2.11.2, assuming that the active clock edge is
falling edge.
It may be noted that at startup there is a brief
interval when 𝑚𝑞 (𝑡) may be a poor
approximation to 𝑚(𝑡), as shown in figure 2.11.3.
The initial large discrepancy between 𝑚(𝑡) and
𝑚𝑞 (𝑡) and stepwise approach of 𝑚𝑞 (𝑡) to 𝑚(𝑡) is
Figure 2.11.1 Delta Modulator
shown in figure 2.11.3.

Figure 2.11.2 The response of the delta modulator to a baseband signal m(t)

Figure 2.11.3 Startup response of DM

Figure 2.11.4 Slope Overload in a linear DM

Page 21 of 24

Page no: 21 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

It should be noted that when 𝑚𝑞 (𝑡) has caught up 𝑚(𝑡) and even though 𝑚(𝑡) remains constant, 𝑚𝑞 (𝑡)
hunts, swinging up and down to 𝑚(𝑡).
Slope Overload
The excessive disparity between 𝑚(𝑡) and 𝑚𝑞 (𝑡) is described as a slope overload error and occurs
whenever 𝑚(𝑡) has a slope larger than the slope 𝑆/𝑇𝑠 which can be sustained by the waveform 𝑚𝑞 (𝑡). The
slope overload as shown in figure 3.11.4 is developed due to the small size of S. To overcome the overload
we have to increase the sampling rate above the rate initially selected to satisfy the Nyquist criterion. The
sampling rate 𝑓𝑠 must satisfy the following condition
𝑠𝑓𝑠 = 2𝜋𝑓𝐴

Features of DM
Following are some of the features of delta modulation.
 An over-sampled input is taken to make full use of the signal correlation.
 The quantization design is simple.
 The input sequence is much higher than the Nyquist rate.
 The quality is moderate.
 The design of the modulator and the demodulator is simple.
 The stair-case approximation of output waveform.
 The step-size is very small, i.e., Δ (delta).
 The bit rate can be decided by the user.
 This involves simpler implementation.

2.11.3 Advantages of DM Over DPCM

 1-bit quantizer
 Very easy design of the modulator and the demodulator
 However, there exists some noise in DM.
 Slope Over load distortion (when Δ is small)
 Granular noise (when Δ is large)

2.12 Adaptive Delta Modulation (ADM):

In digital modulation, we have come across certain problem of determining the step-size, which influences
the quality of the output wave.
A larger step-size is needed in the step slope of modulating signal and a smaller step size is needed where
the message has a small slope. The minute details get missed in the process. So, it would be better if we
can control the adjustment of step-size, according to our requirement in order to obtain the sampling in a
desired fashion. This is the concept of Adaptive Delta Modulation.
Following is the block diagram of Adaptive delta modulator.

Page 22 of 24

Page no: 22 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 2.12.1 Adaptive Delta Modulation (ADM)

The step size S is not of fixed size but it is always a multiple of basic step size 𝑆0 . The basic step size 𝑆0 is
either added or subtracted by the accumulator as required to move 𝑚𝑞 (𝑡) more close to 𝑚(𝑡). If the
direction of the step at the clock edge K is same as at edge K-1, then the processor increases the step size
by an amount 𝑆0 . If the directions are opposite then the processor decreases the magnitude of the step
by 𝑆0 .

Figure 2.12.2 Waveforms comparing the response of DM and ADM

In figure 2.12.1, the output 𝑆0 (𝑡) is called 𝑒(𝑘), which represents the error i.e. the discrepancy between
the 𝑚(𝑡) and 𝑚𝑞 (𝑡), and it is either V(H) or V(L).

𝑒(𝑘) = +1, 𝑖𝑓 𝑚(𝑡) > 𝑚𝑞 (𝑡) immediately before Kth edge,

𝑒(𝑘) = −1, 𝑖𝑓 𝑚(𝑡) < 𝑚𝑞 (𝑡) immediately before Kth edge,

The features of ADM are shown in figure 2.12.2. As long as the condition 𝑚(𝑡) > 𝑚𝑞 (𝑡) persists the jumps
in 𝑚𝑞 (𝑡) becomes larger, that’s why 𝑚𝑞 (𝑡) catches up with 𝑚(𝑡) sooner than in the case of linear DM, as
shown by 𝑚′𝑞 (𝑡).
On the other hand, when the response to the large slope in 𝑚(𝑡), 𝑚𝑞 (𝑡) develops large jumps and large
number of clock cycles are required for these jumps to settle down. Therefore the ADM system reduces
the slope overload but it increases the quantization error. Also when 𝑚(𝑡) is constant 𝑚𝑞 (𝑡) oscillates
about 𝑚(𝑡) but the oscillation frequency is half of the clock frequency.

Page 23 of 24

Page no: 23 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

2.13 Voice coders:

A vocoder i.e. voice encoder is an analysis/synthesis system, used to reproduce human speech. In the
encoder, the input is passed through a multiband filter, each band is passed through an envelope follower,
and the control signals from the envelope followers are communicated to the decoder. The decoder
applies these control signals to corresponding filters in the (re)synthesizer.

It was originally developed as a speech coder for telecommunications applications in the 1930s, the idea
being to code speech for transmission. Its primary use in this fashion is for secure radio communication,
where voice has to be encrypted and then transmitted. The advantage of this method of "encryption" is
that no 'signal' is sent, but rather envelopes of the band pass filters. The receiving unit needs to be set up
in the same channel configuration.
Information, and recreates it, The Voder i.e. Voice Operating Demonstrator generates synthesized speech
by means of a console with fifteen touch-sensitive keys and a pedal, basically consisting of the "second
half" of the vocoder, but with manual filter controls, needing a highly trained operator.

The human voice consists of sounds generated by the opening and closing of the glottis by the vocal cords,
which produces a periodic waveform with many harmonics. This basic sound is then filtered by the nose
and throat (a complicated resonant piping system) to produce differences in harmonic content (formants)
in a controlled way, creating the wide variety of sounds used in speech. There is another set of sounds,
known as the unvoiced and plosive sounds, which are created or modified by the mouth in different
fashions.
The vocoder examines speech by measuring how its spectral characteristics change over time. This results
in a series of numbers representing these modified frequencies at any particular time as the user speaks.
In simple terms, the signal is split into a number of frequency bands (the larger this number, the more
accurate the analysis) and the level of signal present at each frequency band gives the instantaneous
representation of the spectral energy content. Thus, the vocoder dramatically reduces the amount of
information needed to store speech, from a complete recording to a series of numbers. To recreate
speech, the vocoder simply reverses the process, processing a broadband noise source by passing it
through a stage that filters the frequency content based on the originally recorded series of numbers.
Information about the instantaneous frequency (as distinct from spectral characteristic) of the original
voice signal is discarded; it wasn't important to preserve this for the purposes of the vocoder's original use
as an encryption aid, and it is this "dehumanizing" quality of the vocoding process that has made it useful
in creating special voice effects in popular music and audio entertainment.

Since the vocoder process sends only the parameters of the vocal model over the communication link,
instead of a point by point recreation of the waveform, it allows a significant reduction in the bandwidth
required to transmit speech.

Page 24 of 24

Page no: 24 Get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]
Program : B.Tech
Subject Name: Digital Communication
Subject Code: EC-502
Semester: 5th
Downloaded from www.rgpvnotes.in

Department of Electronics and Communication Engineering

Sub. Code: EC 5002 Sub. Name: Digital Communication

Unit 3
Digital Transmission Techniques: Phase shift Keying (PSK)- Binary PSK, differential PSK, differentially
encoded PSK, Quadrature PSK, M-ary PSK. Frequency Shift Keying (FSK)- Binary FSK (orthogonal and non-
orthogonal), M-ary FSK. Comparison of BPSK and BFSK, Quadrature Amplitude Shift Keying (QASK),
Minimum Shift Keying (MSK)

3.1 Digital Modulation

Digital Modulation provides more information capacity, high data security, quicker system availability with
great quality communication. Hence, digital modulation techniques have a greater demand, for their
capacity to convey larger amounts of data than analog modulation techniques.
There are many types of digital modulation techniques and also their combinations, as listed below.

ASK – Amplitude Shift Keying

The amplitude of the resultant output depends upon the input data whether it should be a zero level or a
variation of positive and negative, depending upon the carrier frequency.

FSK – Frequency Shift Keying

The frequency of the output signal will be either high or low, depending upon the input data applied.

PSK – Phase Shift Keying

The phase of the output signal gets shifted depending upon the input. These are mainly of two types,
namely Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK), according to the
number of phase shifts. The other one is Differential Phase Shift Keying (DPSK) which changes the phase
according to the previous value.

M-ary Encoding
M-ary Encoding techniques are the methods where more than two bits are made to transmit
simultaneously on a single signal. This helps in the reduction of bandwidth.
The types of M-ary techniques are M-ary ASK, M-ary FSK & M-ary PSK.

Amplitude Shift Keying (ASK) is a type of Amplitude Modulation which represents the binary data in the
form of variations in the amplitude of a signal.
Any modulated signal has a high frequency carrier. The binary signal when ASK modulated, gives a zero
value for Low input while it gives the carrier output for High input.
The figure 3.1.1 represents ASK modulated waveform along with its input.

(a) ASK Modulation

Page 1 of 34

Page no: 1 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

(b) ASK Modulated Wave

Figure 3.1.1 ASK Modulation
To find the process of obtaining this ASK modulated wave, let us learn about the working of the ASK
modulator.

3.2 ASK Modulator

The ASK modulator block diagram comprises of the carrier signal generator, the binary sequence from the
message signal and the band-limited filter. Following is the block diagram of the ASK Modulator.

Figure 3.2.1 ASK Modulator

The carrier generator sends a continuous high-frequency carrier. The binary sequence from the message
signal makes the unipolar input to be either High or Low. The high signal closes the switch, allowing a
carrier wave. Hence, the output will be the carrier signal at high input. When there is low input, the switch
opens, allowing no voltage to appear. Hence, the output will be low.
The band-limiting filter, shapes the pulse depending upon the amplitude and phase characteristics of the
band-limiting filter or the pulse-shaping filter.

ASK Demodulator
There are two types of ASK Demodulation techniques. They are −
 Asynchronous ASK Demodulation/detection
 Synchronous ASK Demodulation/detection
The clock frequency at the transmitter when matches with the clock frequency at the receiver, it is known
as a Synchronous method, as the frequency gets synchronized. Otherwise, it is known as Asynchronous.

Asynchronous ASK Demodulator

The Asynchronous ASK detector consists of a half-wave rectifier, a low pass filter, and a comparator.
Following is the block diagram for the same.

Figure 3.2.2 ASK Demodulator

Page 2 of 34

Page no: 2 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The modulated ASK signal is given to the half-wave rectifier, which delivers a positive half output. The low
pass filter suppresses the higher frequencies and gives an envelope detected output from which the
comparator delivers a digital output.

Synchronous ASK Demodulator

Synchronous ASK detector consists of a Square law detector, low pass filter, a comparator, and a voltage
limiter. Following is the block diagram for the same.

Figure 3.2.3 Synchronous ASK Demodulator

The ASK modulated input signal is given to the Square law detector. A square law detector is one whose
output voltage is proportional to the square of the amplitude modulated input voltage. The low pass filter
minimizes the higher frequencies. The comparator and the voltage limiter help to get a clean digital
output.

3.3 Frequency Shift Keying

Frequency Shift Keying (FSK) is the digital modulation technique in which the frequency of the carrier
signal varies according to the digital signal changes. FSK is a scheme of frequency modulation. The output
of a FSK modulated wave is high in frequency for a binary High input and is low in frequency for a binary
Low input. The binary 1s and 0s are called Mark and Space frequencies. The following image is the
diagrammatic representation of FSK modulated waveform along with its input.

Figure 3.3.1 Frequency Shift Keying (FSK)

To find the process of obtaining this FSK modulated wave, let us know about the working of a FSK
modulator.

FSK Modulator

The FSK modulator block diagram comprises of two oscillators with a clock and the input binary sequence.
Following is its block diagram.

Page 3 of 34

Page no: 3 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

FSK Transmitter

F1
FSK Modulated
Wave
F2

Binary
Message

Figure 3.3.2 FSK Modulator

The two oscillators, producing a higher and a lower frequency signals, are connected to a switch along with
an internal clock. To avoid the abrupt phase discontinuities of the output waveform during the
transmission of the message, a clock is applied to both the oscillators, internally. The binary input
sequence is applied to the transmitter so as to choose the frequencies according to the binary input.

FSK Demodulator
There are different methods for demodulating a FSK wave. The main methods of FSK detection are
asynchronous detector and synchronous detector. The synchronous detector is a coherent one, while
asynchronous detector is a non-coherent one.

Asynchronous FSK Detector

The block diagram of Asynchronous FSK detector consists of two band pass filters, two envelope detectors,
and a decision circuit. Following is the diagrammatic representation.

Figure 3.3.3 Asynchronous FSK Detector

The FSK signal is passed through the two Band Pass Filters (BPFs), tuned to Space and Mark frequencies.
The output from these two Band Pass Filters look like ASK signal, which is then applied to the envelope
detector. The signal in each envelope detector is modulated asynchronously.
The decision circuit chooses which output is more likely and selects it from any one of the envelope
detectors. It also re-shapes the waveform to a rectangular one.

Synchronous FSK Detector

The block diagram of Synchronous FSK detector consists of two mixers with local oscillator circuits, two
band pass filters and a decision circuit. Following is the diagrammatic representation.

Page 4 of 34

Page no: 4 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.3.4 Synchronous FSK Detector

The FSK signal input is given to the two mixers with local oscillator circuits. These two are connected to two
band pass filters. These combinations act as demodulators and the decision circuit chooses which output is
more likely and selects it from any one of the detectors. The two signals have a minimum frequency
separation.
For both of the demodulators, the bandwidth of each of them depends on their bit rate. This synchronous
demodulator is a bit complex than asynchronous type demodulators.

3.4 Phase Shift Keying (PSK)

Phase Shift Keying (PSK) is the digital modulation technique in which the phase of the carrier signal is
changed by varying the sine and cosine inputs at a particular time. PSK technique is widely used for
wireless LANs, bio-metric, contactless operations, along with RFID and Bluetooth communications.

3.4.1 Binary Phase Shift Keying (BPSK)

In binary phase-shift keying (BPSK) the transmitted signal is a sinusoid of fixed amplitude. It has one fixed
phase when the data is at one level and when the data is at the other level the phase is different by 180°. If
the sinusoid is of amplitude A it has a power
1
𝑃𝑠 = 𝐴2 𝑡ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒 𝐴 = √2𝑃𝑠
2

Thus the transmitted signal is either

𝑉𝐵𝑃𝑆𝐾 (𝑡) = √2𝑃𝑠 . cos(𝜔𝑜 𝑡) …3.4.1.1
or 𝑉𝐵𝑃𝑆𝐾 (𝑡) = −√2𝑃𝑠 . cos(𝜔𝑜 𝑡 + 𝜋) …3.4.1.2(a)
𝑉𝐵𝑃𝑆𝐾 (𝑡) = −√2𝑃𝑠 . cos(𝜔𝑜 𝑡) …3.4.1.2(b)

In BPSK the data b(t) is a stream of binary digits with voltage levels which, as a matter of convenience, we
take to be at + 1V and - 1V. When b(t) = 1V we say it is at logic level 1 and when b(t) = -I V we say it is at
logic level 0.
Hence 𝑉𝐵𝑃𝑆𝐾 (𝑡) can be written, as
𝑉𝐵𝑃𝑆𝐾 (𝑡) = 𝑏(𝑡)√2𝑃𝑠 . cos(𝜔𝑜 𝑡) …3.4.1.3

In practice, a BPSK signal is generated by applying the waveform cos(𝜔𝑜 𝑡), as a carrier, to a balanced
modulator and applying the baseband signal b(t) as the modulating waveform. In this sense BPSK can be
thought of as an AM signal.

Page 5 of 34

Page no: 5 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

BPSK Modulator:
The block diagram of Binary Phase Shift Keying consists of the balance modulator which has the carrier sine
wave as one input and the binary sequence as the other input.
Following is the diagrammatic representation.
BPSK Modulator

PSK Wave

Career Wave
generator
Binary
Sequence Data
Figure 3.4.1 BPSK Modulator

The modulation of BPSK is done using a balance modulator, which multiplies the two signals applied at the
input. For a zero binary input, the phase will be 0° and for a high input, the phase reversal is of 180°.
Following is the diagrammatic representation of BPSK Modulated output wave along with its given input.

Figure 3.4.2 BPSK Modulated Waveform

The output sine wave of the modulator will be the direct input carrier or the inverted (180° phase shifted)
input carrier, which is a function of the data signal.

Reception of BPSK:
The received signal has the form
𝑉𝐵𝑃𝑆𝐾 (𝑡) = 𝑏(𝑡)√2𝑃𝑠 . cos(𝜔𝑜 𝑡 + 𝜃) = 𝑏(𝑡)√2𝑃𝑠 . cos 𝜔𝑜 (𝑡 + 𝜃/𝜔𝑜 ) …3.4.1.4
Here 𝜃 is a nominally fixed phase shift corresponding to the time delay θ/ωo which depends on the length
of the path from transmitter to receiver and the phase shift produced by the amplifiers in the" front-end"
of the receiver preceding the demodulator. The original data b(t) is recovered in the demodulator. The
demodulation technique usually employed is called synchronous demodulation and requires that there be
available at the demodulator the waveform cos(ωo t + θ). A scheme for generating the carrier at the
demodulator and for recovering the baseband signal is shown in Fig. 3.4.1.1.
The received signal is squared to generate the signal
1 1 …3.4.1.5
cos 2 (ωo t + θ) = + cos 2(ωo t + θ)
2 2
The DC component is removed by the band pass filter whose pass band is centered around 2fo and we
then have the signal whose waveform is that of cos 2(ωo t + θ). A frequency divider (composed of a flip-
flop and narrow-band filter tuned to fo is used to regenerate the waveform cos(ωo t + θ). Only the
waveforms of the signals at the outputs of the squarer, filter and divider are relevant, not their amplitudes.

Page 6 of 34

Page no: 6 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.4.1.1 Reception of BPSK

Accordingly in Fig. 3.4.1.1 we have arbitrarily taken amplitudes to be unity. In practice, the amplitudes will
be determined by features of these devices which are of no present concern. In any event, the carrier
having been recovered, it is multiplied with the received signal to generate
1 1 …3.4.1.6
𝑏(𝑡)√2𝑃𝑠 cos 2 (𝜔𝑜 𝑡 + 𝜃) = 𝑏(𝑡)√2𝑃𝑠 [ + cos 2(𝜔𝑜 𝑡 + 𝜃)]
2 2
which is then applied to an integrator as shown.

We have included in the system a bit synchronizer. This device, whose operation is able to recognize
precisely the moment which corresponds to the end of the time interval allocated to one bit and the
beginning of the next. At that moment, it closes switch S, very briefly to discharge (dump) the integrator
capacitor and leaves the switch S, open during the entire course of the ensuing bit interval, closing switch
S, again very briefly at the end of the next bit time, etc. (This circuit is called an "integrate-and-dump"
circuit.) The output signal of interest to us is the integrator output at the end of a bit interval but
immediately before the closing of switch 𝑆𝑜. This output signal is made available by switch S; which
samples the output voltage just prior to dumping the capacitor.
Let us assume for simplicity that the bit interval 𝑇𝑏 is equal to the duration of an integral number n of
cycles of the carrier of frequency 𝑓𝑜 that is, 𝑛 . 2𝜋 = 𝜔0 𝑇𝑏 In this case the output voltage 𝑣0 (𝑘𝑇𝑏 ) at the
end of a bit interval extending from time (𝑘 − 1)𝑇𝑏 to (𝑘)𝑇𝑏 is, using Eq. (3.4.1.6).
𝑘𝑇𝑏 𝑘𝑇𝑏
1 1 …3.4.1.7(a)
𝑣0 (𝑘𝑇𝑏 ) = 𝑏(𝑘𝑇𝑏 )√2𝑃𝑠 ∫ dt + 𝑏(𝑘𝑇𝑏 )√2𝑃𝑠 ∫ cos 2(𝜔𝑜 𝑡 + 𝜃) dt
(𝑘 – 1)𝑇𝑏 2 (𝑘 − 1)𝑇𝑏 2
…3.4.1.7(b)
𝑃𝑠
𝑣0 (𝑘𝑇𝑏 ) = 𝑏(𝑘𝑇𝑏 )√ 𝑇𝑏
2

Since the integral of a sinusoid over a whole number of cycles has the value zero. Thus we see that our
system reproduces at the demodulator output the transmitted bit stream b(t). The operation of the bit
synchronizer allows us to sense each bit independently of every other bit. The brief closing of both
switches, after each bit has been determined, wipes clean all influence of a preceding bit and allows the
receiver to deal exclusively with the present bit.

Page 7 of 34

Page no: 7 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

3.4.2 Differential Phase Shift Keying (DPSK)

In BPSK, to regenerate the carrier we start by squaring 𝑏(𝑡)√2𝑃𝑠 . cos(𝜔𝑜 𝑡). Accordingly, if the received
signal were instead −𝑏(𝑡)√2𝑃𝑠 . cos(𝜔𝑜 𝑡) the recovered carrier would remain as before. Therefore we
shall not be able to determine whether the received baseband signal is the transmitted signal 𝑏(𝑡) or it’s
negative − 𝑏(𝑡).
Differential phase-shift keying (DPSK) and differential encoded PSK (DEPSK) are modifications of BPSK
which have the merit that they eliminate the ambiguity about whether the demodulated data is or is not
inverted. In addition DPSK avoids the need to provide the synchronous carrier required at the demodulator
for detecting a BPSK signal.
A means for generating a DPSK signal is shown in Fig. 3.4.2.1. The data stream to be transmitted, 𝑑(𝑡), is
applied to one input of an exclusive-OR logic gate. To the other gate input is applied the output of the
exclusive or gate 𝑏(𝑡) delayed by the time 𝑇𝑏 , allocated to one bit. This second input is then 𝑏(𝑡 − 𝑇𝑏 ). In
Fig. 3.4.2.2 we have drawn logic waveforms to illustrate the response b(t) to an input d(t). The upper level
of the waveforms corresponds to logic 1, the lower level to logic O. The truth table for the exclusive-OR
gate is given in Fig 3.4.2.1, and with this table we can easily verify that the waveforms for (𝑡), 𝑏(𝑡 − 𝑇𝑏 ),
and 𝑏(𝑡) are consistent with one another. We observe that, as required, 𝑏(𝑡 − 𝑇𝑏 ) is indeed b(t) delayed
by one bit time and that in any bit interval the bit b(t) is given 𝑏(𝑡) = 𝑑(𝑡) ⊕ 𝑏(𝑡 − 𝑇𝑏 ).

Figure 3.4.2.1 Means for generating DPSK

Because of the feedback involved in the system of Fig. 3.4.2.2 there is a difficulty in determining the logic
levels in the interval in which we start to draw the waveforms (interval 1 in Fig. 3.4.2.2). We cannot
determine b(t) in this first interval of our waveform unless we know b(k = 0). But we cannot determine b(0)
unless we know both d(0) and b( -1), etc. Thus, to justify any set of logic levels in an initial bit interval we
need to know the logic levels in the preceding interval. But such a determination requires information
about the interval two bit times earlier and so on. In the waveforms of Fig. 3.4.2.2 we have circumvented
the problem by arbitrarily assuming that in the first interval b(0) = 0. It is shown below that in the
demodulator, the data will be correctly determined regardless of our assumption concerning b(0).

Page 8 of 34

Page no: 8 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.4.2.2 Logic waveforms to illustrate the response b(t) to an input d(t).

We now observe that the response of b(t) to d(t) is that b(t) changes level at the beginning of each interval
in which d(t) = 1 and b(t) does not change level when d(t) = 0. Thus during interval 3, d(3) = 1, and
correspondingly b(3) changes at the beginning at that interval. During intervals 6 and 7, d(6) = d(7) = 1 and
there are changes in b(t) at the beginnings of both intervals. During bits 10, 11, 12, and 13 d(t) = 1 and
there are changes in b(t) at the beginnings of each of these intervals. This behavior is to be anticipated
from the truth table of the exclusive-OR gate. For we note that when 𝑑(𝑡) = 0, 𝑏(𝑡) = 𝑏(𝑡 − 𝑇𝑏 ) so that,
whatever the initial value of 𝑏(𝑡 − 𝑇𝑏 ) ,it reproduces itself. On the other hand when d(t) = 1, then 𝑏(𝑡) =
𝑏(𝑡 − 𝑇𝑏 ). Thus, in each successive bit interval b(t) changes from its value in the previous interval. Note
that in some intervals where d(t) = 0 we have b(t) = 0 and in other intervals when d(t) = 0 we have b(t) = 1.
Similarly, when d(t) = 1 sometimes b(t) = 1 and sometimes b(t) = 0. Thus there is no correspondence
between the levels of d(t) and b(t), and the only invariant feature of the system is that a change
(sometimes up and sometimes down) in b(t) occurs whenever d(t) = 1, and that no change in b(t) will occur
whenever d(t) = 0.
Finally, we note that the waveforms of Fig. 3.4.2.2 are drawn on the assumption that, in interval 1, b(0) = 0.
As is easily verified, if not intuitively apparent, if we had assumed b(0) = 1, the invariant feature by which
we have characterized the system would continue to apply. Since b(0) must be either b(0) = 0 or b(0) = 1,
there being no other possibilities, our result is valid quite generally. If, however, we had started with b(0) =
1 the levels b(l) and b(0) would have been inverted.
As is seen in Fig. 3.4.2.1 b(t) is applied to a balanced modulator to which is also applied the carrier
√2𝑃𝑠 . cos(𝜔𝑜 𝑡). The modulator output, which is the transmitted signal, is

𝑉𝐷𝑃𝑆𝐾 (𝑡) = 𝑏(𝑡)√2𝑃𝑠 . cos(𝜔𝑜 𝑡)

= ±√2𝑃𝑠 . cos(𝜔𝑜 𝑡) …3.4.2.1

Thus altogether when d(t) = 0 the phase of the carrier does not change at the beginning of the bit interval,
while when d(t) = 1 there is a phase change of magnitude 𝜋.
Reception:
A method of recovering the data bit stream from the DPSK signal is shown in Fig. 3.4.2.3. Here the received
signal and the received signal delayed by the bit time 1'" are applied to a multiplier. The multiplier output
is

Page 9 of 34

Page no: 9 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 )2𝑃𝑠 cos(𝜔𝑜 𝑡 + 𝜃) . cos[𝜔𝑜 (𝑡 − 𝑇𝑏 ) + 𝜃]

𝑇
= 𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 )𝑃𝑠 {cos 𝜔𝑜 𝑇𝑏 + cos [2𝜔𝑜 (𝑡 − 𝑏) + 2𝜃]}
2
…3.4.2.2
and is applied to a bit synchronizer and integrator as shown in Fig. 6.2-1 for the BPSK demodulator. The
first term on the right-hand side of Eq.(3.4.2.1) is, aside from a multiplicative constant, the waveform
𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 ) which, as we shall see is precisely the signal we require. As noted previously in connection
with BPSK, and so here, the output integrator will suppress the double frequency term. We should select
𝜔𝑜 𝑇𝑏 so that 𝜔𝑜 𝑇𝑏 = 2𝑛𝜋 with n an integer. For, in this case we shall have cos 𝜔𝑜 𝑇𝑏 = +1 and the signal
output will be as large as possible.

Figure 3.4.2.3 Methods of recovering data from DPSK

Further, with this selection, the bit duration encompasses an integral number of clock cycles and the
integral of the double-frequency term is exactly zero.
The transmitted data bit d(t) can readily be determined from the product 𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 ). If d(t) = 0 then
there was no phase change and 𝑏(𝑡) = 𝑏(𝑡 − 𝑇𝑏 ) both being + 1V or both being - 1V. In this case
𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 )= 1. If however, d(t) = 1 then there was a phase change and either b(t) = 1V with 𝑏(𝑡 − 𝑇𝑏 ) =
- I V or vice versa. In either case 𝑏(𝑡). 𝑏(𝑡 − 𝑇𝑏 ) = −1.

OR
Differential Phase Shift Keying (DPSK) the phase of the modulated signal is shifted relative to the previous
signal element. No reference signal is considered here. The signal phase follows the high or low state of the
previous element. This DPSK technique doesn’t need a reference oscillator.
The following figure represents the model waveform of DPSK.

Figure 3.4.9 Differential Phase Shift Keying (DPSK)

It is seen from the above figure that, if the data bit is Low i.e., 0, then the phase of the signal is not
reversed, but continued as it was. If the data is a High i.e., 1, then the phase of the signal is reversed, as
with NRZI, invert on 1 (a form of differential encoding).
If we observe the above waveform, we can say that the High state represents an M in the modulating
signal and the Low state represents a W in the modulating signal.

DPSK Modulator

Page 10 of 34

Page no: 10 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

DPSK is a technique of BPSK, in which there is no reference phase signal. Here, the transmitted signal itself
can be used as a reference signal. Following is the diagram of DPSK Modulator.

Figure 3.4.10 DPSK Modulator

DPSK encodes two distinct signals, i.e., the carrier and the modulating signal with 180° phase shift each.
The serial data input is given to the XNOR gate and the output is again fed back to the other input through
1-bit delay. The output of the XNOR gate along with the carrier signal is given to the balance modulator, to
produce the DPSK modulated signal.

DPSK Demodulator
In DPSK demodulator, the phase of the reversed bit is compared with the phase of the previous bit.
Following is the block diagram of DPSK demodulator.

Figure 3.4.11 DPSK Demodulator

From the above figure, it is evident that the balance modulator is given the DPSK signal along with 1-bit
delay input. That signal is made to confine to lower frequencies with the help of LPF. Then it is passed to a
shaper circuit, which is a comparator or a Schmitt trigger circuit, to recover the original binary data as the
output.
The word binary represents two bits. M represents a digit that corresponds to the number of conditions,
levels, or combinations possible for a given number of binary variables.
This is the type of digital modulation technique used for data transmission in which instead of one bit, two
or more bits are transmitted at a time. As a single signal is used for multiple bit transmission, the channel
bandwidth is reduced.

3.4.3 Differentially Encoded Phase Shift Keying (DEPSK)

The DPSK demodulator requires a device which operates at the carrier frequency and provides a delay of
Tb. Differentially-encoded PSK eliminates the need for such a piece of hardware. In this system,
synchronous demodulation recovers the signal b(t), and the decoding of b(t) to generate d(t) is done at
baseband.

Page 11 of 34

Page no: 11 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.4.3.1 Baseband Decoder to obtain d(t) from b(t)

Figure 3.4.3.2 Errors in DEPSK occurs in pairs

The transmitter of the DEPSK system is identical to the transmitter of the DPSK system shown in Fig.
3.4.2.1, The signal b(t) is recovered in exactly the manner shown in Fig. 3.4.2.1 for a BPSK system. The
recovered signal is then applied directly to one input of an exclusive-OR logic gate and to the other input
is applied 𝑏(𝑡 − 𝑇𝑏 ) (see Fig. 3.4.3.1). The gate output will be at one or the other of its levels depending on
whether 𝑏(𝑡) = 𝑏(𝑡 − 𝑇𝑏 ) or 𝑏(𝑡) = ̅̅̅̅̅̅̅̅̅̅̅̅̅
𝑏(𝑡 − 𝑇𝑏 ) . In the first case b(t) did not change level and therefore
the transmitted bit is d(t) = 0. In the second case d(t) = 1.
We have seen that in DPSK there is a tendency for bit errors to occur in pairs but that single bit errors are
possible. In DEPSK errors always occur in pairs. The reason for the difference is that in DPSK we do not
make a hard decision, in each bit interval about the phase of the received signal. We simply allow the
received signal in one interval to compare itself with the signal in an adjoining interval and, as we have
seen, a single error is not precluded. In DEPSK, a firm definite hard decision is made in each interval about
the value of b(t). If we make a mistake, then errors must result from a comparison with the preceding and
succeeding bit. This result is illustrated in Fig.3.4.3.2. It is shown the error-free signals 𝑏(𝑘), 𝑏(𝑘 − 1) and
𝑑 (𝑘) = 𝑏(𝑘) ⊕ 𝑏(𝑘 − 1). We have assumed that b'(k) has a single error. Then b'(k - 1) must also have a
single error. We note that the reconstructed waveform d'(k) now has two errors.

3.4.4 Quadrature Phase Shift Keying (QPSK)

This is the phase shift keying technique, in which the sine wave carrier takes four phase reversals such as
0°, 90°, 180°, and 270°.

Page 12 of 34

Page no: 12 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

If these kinds of techniques are further extended, PSK can be done by eight or sixteen values also,
depending upon the requirement.

QPSK Modulator

The mechanism by which a bit stream b(t) generates a QPSK signal for transmission is shown in Fig. 3.4.4.1
and relevant waveforms are shown in Fig. 3.4.4.2. In these waveforms we have arbitrarily assumed that in
every case the active edge of the clock waveforms is the downward edge. The toggle flip-flop is driven by a
clock waveform whose period is the bit time Tb. The toggle flip-flop generates an odd clock waveform and
an even waveform. These clocks have periods 2Tb. The active edge of one of the clocks and the active edge
of the other are separated by the bit time Tb. The bit stream 𝑏(𝑡) is applied as the data input to both type-
D flip-flops, one driven by the odd and one driven by the even clock waveform. The flip-flops register
alternate bits in the stream b(t) and hold each such registered bit for two bit intervals, that is for a time Tb.
In Fig. 3.4.4.2 we have numbered the bits in 𝑏(𝑡). Note that the bit stream 𝑏(𝑡) (which is the output of the
flip-flop driven by the odd clock) registers bit 1 and holds that bit .for time 2Tb, then registers bit 3 for time
2Tb, then bit 5 for 2Tb, etc. The even bit stream be(t) holds, for times 2Tb each the alternate bits numbered
2.4, 6, etc.
The bit stream 𝑏(𝑡) (which. as usual, we take to be 𝑏𝑒 (𝑡)= ± 1 volt) is superimposed on a carrier
√𝑃𝑠 . sin(𝜔𝑜 𝑡) by the use of two multipliers (i.e., balanced modulators) as shown, to generate two signals
Se(t) and So(t). These signals are then added to generate the transmitted output signal Vm(t) which is
𝑣𝑚 (𝑡) = √𝑃𝑠 . 𝑏𝑜 (𝑡) sin(𝜔𝑜 𝑡) + √𝑃𝑠 . 𝑏𝑒 (𝑡) cos(𝜔𝑜 𝑡)
As may be verified, the total normalized power of 𝑣𝑚 (𝑡) is Ps.

The QPSK Modulator uses a bit-splitter, two multipliers with local oscillator, a 2-bit serial to parallel
converter, and a summer circuit. Following is the block diagram for the same.

Figure 3.4.6 QPSK Modulator

At the modulator’s input, the message signal’s even bits (i.e., 2 nd bit, 4th bit, 6th bit, etc.) and odd bits (i.e.,
1st bit, 3rd bit, 5th bit, etc.) are separated by the bits splitter and are multiplied with the same carrier to
generate odd BPSK (called as PSKI) and even BPSK (called as PSKQ). The PSKQ signal is anyhow phase shifted
by 90° before being modulated.

The QPSK waveform for two-bits input is as follows, which shows the modulated result for different
instances of binary inputs.

Page 13 of 34

Page no: 13 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.4.7 QPSK Waveforms

QPSK Demodulator
The QPSK Demodulator uses two product demodulator circuits with local oscillator, two band pass filters,
two integrator circuits, and a 2-bit parallel to serial converter. Following is the diagram for the same.

Figure 3.4.8 QPSK Demodulator

The two product detectors at the input of demodulator simultaneously demodulate the two BPSK signals.
The pair of bits are recovered here from the original data. These signals after processing, are passed to the
parallel to serial converter.
3.5 M-ary Equation
If a digital signal is given under four conditions, such as voltage levels, frequencies, phases, and amplitude,
then M = 4. The number of bits necessary to produce a given number of conditions is expressed
mathematically as N=log2M. Where N is the number of bits necessary M is the number of conditions,
levels, or combinations possible with N bits.
The above equation can be re-arranged as
2N=M
2
For example, with two bits, 2 = 4 conditions are possible.

Types of M-ary Techniques

In general, Multi-level (M-ary) modulation techniques are used in digital communications as the digital
inputs with more than two modulation levels are allowed on the transmitter’s input. Hence, these
techniques are bandwidth efficient.
There are many M-ary modulation techniques. Some of these techniques, modulate one parameter of the
carrier signal, such as amplitude, phase, and frequency.

M-ary ASK
This is called M-ary Amplitude Shift Keying (M-ASK) or M-ary Pulse Amplitude Modulation (PAM).
The amplitude of the carrier signal, takes on M different levels.

Page 14 of 34

Page no: 14 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Representation of M-ary ASK

𝑆𝑚 (𝑡) = 𝐴𝑚 cos(2𝜋𝑓𝑐 𝑡) 𝐴𝑚 ∈(2𝑚 − 1 − 𝑀)∆, 𝑚 = 1,2, … … 𝑀 and 0 ≤ 𝑡 ≤ 𝑇𝑠
Some prominent features of M-ary ASK are −
 This method is also used in PAM.
 Its implementation is simple.
 M-ary ASK is susceptible to noise and distortion.

M-ary FSK
This is called as M-ary Frequency Shift Keying (M-ary FSK).
The frequency of the carrier signal, takes on M different levels.

Representation of M-ary FSK

2𝐸 𝜋
𝑆𝑖 (𝑡) = √ 𝑇 𝑠 cos (𝑇 (𝑛𝑐 + 𝑖)𝑡) 0 ≤ 𝑡 ≤ 𝑇𝑠 𝑖 = 1,2, … … 𝑀
𝑠 𝑠

Where 𝑓𝑐 = 𝑛𝑐/ 2𝑇𝑠

for some fixed integer n.
Some prominent features of M-ary FSK are −
 Not susceptible to noise as much as ASK.
 The transmitted M number of signals are equal in energy and duration.
 The signals are separated by 12Ts
 Hz making the signals orthogonal to each other.
 Since M signals are orthogonal, there is no crowding in the signal space.
 The bandwidth efficiency of M-ary FSK decreases and the power efficiency increases with the
increase in M.

M-ary PSK
This is called as M-ary Phase Shift Keying (M-ary PSK).
The phase of the carrier signal, takes on M different levels.
Representation of M-ary PSK
2𝐸
𝑆𝑖 (𝑡) = √ 𝑇 cos(𝜔0𝑡 + 𝜙𝑖 𝑡) 0 ≤ 𝑡 ≤ 𝑇 𝑎𝑛𝑑 𝑖 = 1,2, … … 𝑀
2𝜋𝑖
𝜙𝑖 (𝑡) = 𝑀 where 𝑖 = 1,2, … … 𝑀
Some prominent features of M-ary PSK are −
 The envelope is constant with more phase possibilities.
 This method was used during the early days of space communication.
 Better performance than ASK and FSK.
 Minimal phase estimation error at the receiver.
 The bandwidth efficiency of M-ary PSK decreases and the power efficiency increases with the
increase in M.
So far, we have discussed different modulation techniques. The output of all these techniques is a binary
sequence, represented as 1s and 0s. This binary or digital information has many types and forms, which are
discussed further.

3.6 Comparison of BPSK and BFSK

The two modulation techniques can be compared in the following manners:

1. The bandwidth required for transmitting the BFSK signal is 4f (f is the frequency of the data signal),
whereas the bandwidth requirement for the BPSK signal is only 2f. Therefore BPSK is a better option.

Page 15 of 34

Page no: 15 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

2. In BPSK the information of the message is stored in the phase variations of the career wave whereas in
case of the BFSK scheme, the information is available as the frequency variations of the career wave. Now
we know that the noise can affect the frequency of the career wave but cannot affect the phase of the
career signal. Therefore the BPSK Scheme is again a better option.

3. The noise will also occupy some frequency it may damage the signal flatly or frequency selectively. But
in PSK there is very less chance in the changing of phase f the signal. Hence for noisy channel PSK is better
than FSK.

3.7 Quadrature Phase Shift Keying (QPSK)

We have seen that when a data stream whose bit duration is T b is to be transmitted by BPSK the channel
bandwidth must be nominally 2fb where fb = I/Tb.
Quadrature phase-shift keying allows bits to be transmitted using half the bandwidth. In QPSK system we
use the type-D flip-flop as a one bit storage device.

D Flip Flop

The type-D flip-flop represented in Fig3.7.4.1 has a single data input

terminal (D) to which a data stream d(t) is applied. The operation of
the flip-flop is such that at the" active" edge of the clock waveform
the logic level at D is transferred to the output Q. Representative
waveforms are shown in Fig. 3.7.4.2 We assume arbitrarily that the
negative-going edge of the clock waveform is the active edge.
Figure 3.7.4.1 D Flip Flop

Figure 3.7.4.2 Waveforms showing D Flip Flop Characteristics

QPSK Transmitter (Modulator):

The mechanism by which a bit stream b(t) generates a QPSK signal for transmission is shown in Fig. 3.7.4.3
and relevant waveforms are shown in Fig. 3.7.4.4. In these waveforms we have arbitrarily assumed that in
every case the active edge of the clock waveforms is the downward edge. The toggle flip-flop is driven by a
clock waveform whose period is the bit time Tb. The toggle flip-flop generates an odd clock waveform and
an even waveform. These clocks have periods 2Tb. The active edge of one of the clocks and the active edge
of the other are separated by the bit time Tb. The bit stream 𝑏(𝑡) is applied as the data input to both type-
D flip-flops, one driven by the odd and one driven by the even clock waveform. The flip-flops register
alternate bits in the stream b(t) and hold each such registered bit for two bit intervals, that is for a time Tb.
In Fig. 3.4.4.4 we have numbered the bits in 𝑏(𝑡). Note that the bit stream 𝑏(𝑡) (which is the output of the
flip-flop driven by the odd clock) registers bit 1 and holds that bit .for time 2Tb, then registers bit 3 for time
2Tb, then bit 5 for 2Tb, etc. The even bit stream be(t) holds, for times 2Tb each the alternate bits numbered
2.4, 6, etc.

Page 16 of 34

Page no: 16 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The bit stream 𝑏(𝑡) (which. as usual, we take to be 𝑏𝑒 (𝑡)= ± 1 volt) is superimposed on a carrier
√𝑃𝑠 . sin(𝜔𝑜 𝑡) by the use of two multipliers (i.e., balanced modulators) as shown, to generate two signals
Se(t) and So(t). These signals are then added to generate the transmitted output signal Vm(t) which is
𝑣𝑚 (𝑡) = √𝑃𝑠 . 𝑏𝑜 (𝑡) sin(𝜔𝑜 𝑡) + √𝑃𝑠 . 𝑏𝑒 (𝑡) cos(𝜔𝑜 𝑡)
As may be verified, the total normalized power of 𝑣𝑚 (𝑡) is Ps.

Figure 3.7.4.3 QPSK Transmitter

Figure 3.7.4.3 Waveforms for the QPSK Transmitter

In BPSK, the bit duration is Tb, and the generated signal has a nominal bandwidth of 2x1/T b. In the
waveforms of bo(t) and be(t), the bit times are each 1/2T b, hence both bo(t) and be(t) have nominal
bandwidth which are half of the bandwidth in BPSK.
Phasor Diagram:
When b0 = 1 the signal so(t) = √𝑃𝑠 . sin(𝜔𝑜 𝑡), and so(t) = −√𝑃𝑠 . sin(𝜔𝑜 𝑡) when bo = -1. Correspondingly, for
be(t) = ± 1, se(t) = ±. √𝑃𝑠 . (𝑡) cos(𝜔𝑜 𝑡) These four signals have been represented as phasors in Fig. 3.7.4.4.
They are in mutual phase quadrature. Also drawn are the phasors representing the four possible output
signals 𝑣𝑚 (𝑡) = 𝑠𝑜(𝑡) + 𝑆𝑒(𝑡). These four possible output signals have equal amplitude √2𝑃𝑠 and are in
phase quadrature; they have been identified by their corresponding values of bo and be. At the end of each
bit interval (i.e., after each time Tb) either bo, or be can change, but both cannot change at the same time.
Consequently, the QPSK system shown in Fig. 3.7.4.3 is called offset or staggered QPSK and abbreviated
OQPSK. After each time Tb, the transmitted signal, if it changes, changes phase by 90° rather than by 180°
as in BPSK.

Page 17 of 34

Page no: 17 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.7.4.4 Phasor diagrams for the sinusoids of Fig 3.7.4.2

Non-offset QPSK
Suppose that in Fig. 3.7.4.3 we introduce an additional flip-flop before either the odd or even flip-flop. Let
this added flip-flop be driven by the clock which runs at the rate fb. Then one or the other bit streams, odd
or even, will be delayed by one bit interval. As a result, we shall find that two bits which occur in time
sequence (i.e., serially) in the input bit stream b(t) will appear at the same time (i.e., in parallel) at the
outputs of the odd and even flip-flops. In this case be(t) and bo(t) can change at the same time, after each
time 2Tb, and there can be a phase change of 180° in the output signal. There is no difference, in principle,
between a staggered and non-staggered system.
In practice, there is often a significant difference between QPSK and OQPSK. At each transition time, T" for
OQPSK and 2Tb for QPSK, one bit for OQPSK and perhaps two bits for QPSK change from 1V to -1V or -1V to
1V. Now the bits be(t) and bo(t) can, not change instantaneously and, in changing, must pass through zero
and dwell in that neighborhood at least briefly. Hence there will be brief variations in the amplitude of the
transmitted waveform. These variations will be more pronounced in QPSK than in OQPSK since in the first
case both be(t) and bo(t) may be zero simultaneously so that the signal amplitude may actually be reduced
to zero temporarily.

Symbol versus Bit Transmission

In BPSK we deal individually with each bit of duration Tb. In QPSK we lump two bits together to form what
is termed a symbol. The symbol can have anyone of four possible values corresponding to the two-bit
sequences 00, 01, 10, and 11. We therefore arrange to make available for transmission four distinct
signals. At the receiver each signal represents one symbol and, correspondingly, two bits. When bits are
transmitted, as in BPSK, the signal changes occur at the bit rate. When symbols are transmitted the
changes occur at the symbol rate which is one-half the bit rate. Thus the symbol time is Ts=2Tb.

The QPSK Receiver

A receiver for the QPSK signal is shown in Fig. 3.7.4.5. Synchronous detection is required and hence it is
necessary to locally regenerate the carriers cos 𝜔𝑜 𝑡 and sin 𝜔𝑜 𝑡. The scheme for carrier regeneration is
similar to that employed in BPSK. In that earlier case we squared the incoming signal, extracted a
waveform at twice the carrier frequency by filtering, and recovered the carrier by frequency dividing by
two. In the present case, it is required that the incoming signal be raised to the fourth power after which
filtering recovers a waveform at four times the carrier frequency and finally frequency division by four
regenerates the carrier. In the present case, also, we require both cos 𝜔𝑜 𝑡 and sin 𝜔𝑜 𝑡.
The incoming signal is also applied to two synchronous demodulators consisting, as usual, of a multiplier
(balanced modulator) followed by an integrator. The integrator integrates over a two-bit interval of
duration Ts=2Tb and then dumps its accumulation. As noted previously, ideally the interval 2Tb = Ts should
encompass an integral number of carrier cycles. One demodulator uses the carrier cos 𝜔𝑜 𝑡 and the other
the carrier sin 𝜔𝑜 𝑡. We recall that when sinusoids in phase quadrature are multiplied, and the product is
integrated over an integral number of cycles, the result is zero. Hence the demodulators will selectively
respond to the parts of the incoming signal involving respectively be(t) or bo(t).

Page 18 of 34

Page no: 18 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.7.4.5 A QPSK Receiver

Of course, as usual, a bit synchronizer is required to establish the beginnings and ends of the bit intervals
of each bit stream so that the times of integration can be established. The bit synchronizer is needed as
well to operate the sampling switch. At the end of each integration time for each individual integrator, and
just before the accumulation is dumped, the integrator output is sampled. Samples are taken alternately
from one and the other integrator output at the end of each bit time Tb and these samples are held in the
latch for the bit time Tb Each individual integrator output is sampled at intervals 2 Tb. The latch output is
the recovered bit stream b(t).
The voltages marked on Fig. 3.7.4.5 are intended to represent the waveforms of the signals only and not
their amplitudes. Thus the actual value of the sample voltages at the integrator outputs depends on the
amplitude of the local carrier, the gain, if any, in the modulators and the gain in the integrators. We have
however indicated that the sample values depend on the normalized power P s of the received signal and
on the duration Ts of the symbol.

3.8 M ARY PSK

In BPSK we transmit each bit individually. Depending on whether b(t) is logic 0 or logic 1, we transmit one
or another of a sinusoid for the bit time Tb, the sinusoids differing in phase by 2π/2 = 1800. In QPSK we
lump together two bits. Depending on which of the four two-bit words develops, we transmit one or
another of four sinusoids of duration 2Tb the sinusoids differing in phase by amount 2π/4 = 90°. The
scheme can be extended. Let us lump together N bits so that in this N-bit symbol, extending over the time
NTb, there are 2N = M possible symbols. Now let us represent the symbols by sinusoids of duration NTb= Ts
which differ from one another by the phase 2 π / M. Hardware to accomplish such M-ary communication is
available.
Thus in M-ary PSK the waveforms used to identify the symbols are
𝑣𝑚 (𝑡) = √2𝑃𝑠 cos(𝜔𝑜 𝑡 + ∅𝑚 ) (m=0, 1, …, M-1) …3.8.1
Where phase angle is given by
𝜋 …3.8.2
∅𝑚 = (2𝑚 + 1)
𝑀

The waveforms of Eq. (3.8.1) are represented by the dots in Fig. 3.8.1 in a signal space in which the
coordinate axes are the orthonormal waveforms 𝑢1(𝑡) = √2/𝑇𝑠 cos(𝜔𝑜 𝑡) and 𝑢2(𝑡) = √2/𝑇𝑠 sin(𝜔𝑜 𝑡).
The distance of each dot from the origin is √𝐸𝑠 = √𝑃𝑠 𝑇𝑠
From Eq. (3.5.1) we have
𝑣𝑚 (𝑡) = (√2𝑃𝑠 cos ∅𝑚 ) cos(𝜔𝑜 𝑡) − (√2𝑃𝑠 sin ∅𝑚 ) sin(𝜔𝑜 𝑡) …3.8.3
Defining pe and po by

Page 19 of 34

Page no: 19 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝑝𝑒 = √2𝑃𝑠 cos ∅𝑚 …3.8.4a

𝑝𝑜 = √2𝑃𝑠 sin ∅𝑚 …3.8.4b
Equation 3.5.3 becomes
𝑣𝑚 (𝑡) = 𝑝𝑒 cos(𝜔𝑜 𝑡) − 𝑝𝑜 sin(𝜔𝑜 𝑡) …3.8.5

Figure 3.8.1 Graphical representation of M-ary PSK Signals

M-ary Transmitter and Receiver

Figure 3.8.2 M Ary Transmitter

The transmitter, the bit stream b(t) is applied to a serial-to-parallel converter. This converter has facility for
storing the N bits of a symbol. The N bits have been presented serially, that is, in time sequence, one after
another. These N bits, having been assembled, are then presented all at once on N output lines of the
converter, that is they are presented in parallel. The converter output remains unchanging for the duration
NTb of a symbol during which time the converter is assembling a new group of N bits. Each symbol time
the converter output is updated.
The converter output is applied to a D/A converter. This D/A converter generates an output voltage which
assumes one of 2N = M different values in a one to-one correspondence to the M possible symbols applied
to its input. That is, the D/A output is a voltage v(Sm) which depends on the symbol Sm (m = 0, 1,... ,M - 1).
Finally v(Sm) is applied as a control input to a special type of constant amplitude sinusoidal signal source
whose phase 4>m is determined by v(Sm). Altogether, then, the output is a fixed amplitude, sinusoidal

Page 20 of 34

Page no: 20 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

waveform, whose phase has a one-to-one correspondence to the assembled N-bit symbol. The phase can
change once per symbol time.

Figure 3.8.3 M Ary Transmitter

The carrier recovery system requires, in the present case a device to raise the received signal to the Mth
power, filter to extract the Mfo component and then divide by M.
Since there is no staggering of parts of the symbol, the integrators extend their integration over the same
time interval. Of course, again, a bit synchronizer is needed.
The integrator outputs are voltages whose amplitudes are proportional to TsPe and TsPo respectively and
change at the symbol rate. These voltages measure the components of the received signal in the directions
of the quadrature phasors sinw0t and cosw0t. Finally the signals TsPe and TsPo are applied to a device
which reconstructs the digital N-bit signal which constitutes the transmitted signal.
Current operating systems are common in which M = 16. In this case the bandwidth is B = 2fh/4 =fb/2 in
comparison to B =fb for QPSK. PSK systems transmit information through signal phase and not through
signal amplitude. Hence such systems have great merit in situations where, on account of the vagaries of
the transmission medium, the received signal varies in amplitude.

3.9 Quadrature Amplitude Shift Keying (QASK)

In BPSK, QPSK, and M-ary PSK we transmit, in any symbol interval, one signal or another which are
distinguished from one another in phase but are all of the same amplitude. In each of these individual
systems the end points of the signal vectors in signal space falls on the circumference of a circle. Now we
have noted that our ability to distinguish one signal vector from another in the presence of noise will
depend on the distance between the vector end points. It is hence rather apparent that we shall be able to
improve the noise immunity of a system by allowing the signal vectors to differ, not only in their phase but
also in amplitude. We now describe such an amplitude and phase shift keying system. Like QPSK it involves
direct (balanced) modulation of carriers in quadrature (i.e., cos 𝜔𝑜 𝑡 and sin 𝜔𝑜 𝑡) in quadrature and hence
abbreviated as QAPSK or simply QASK.
For Example consider to transmit a symbol for every 4 bits. There are then 2 4= 16 different possible
symbols and we shall have to be able to generate 16 distinguishable signals. One possible geometrical
representation is shown in figure 3.9.1. In this configuration each signal point is equally distant from its
neighbors, the distance being d=2a.
Let us assume that all 16 signals are equally likely. Because of the symmetrical placement around the
origin, we can determine the average energy associated with a signal, from the four signals in the first
quadrant. The average normalized energy of a signal is
1
𝐸𝑠 = [(𝑎2 + 𝑎2 ) + (9𝑎2 + 𝑎2 ) + (𝑎2 + 9𝑎2 ) + (9𝑎2 + 9𝑎2 )] …3.9.1
4
𝐸𝑠 = 10𝑎2

Page 21 of 34

Page no: 21 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.9.1 Geometrical Representation of 16 signals QASK

𝑎 = √0.1𝐸𝑠 …3.9.2
𝑑 = 2√0.1𝐸𝑠 …3.9.3
In present case each symbol represent 4 bits, the normalized symbol energy is 𝐸𝑠 = 4𝐸𝑏 where 𝐸𝑏 is the
normalized bit energy. Therefore
𝑎 = √0.1𝐸𝑠 = √0.4𝐸𝑏
and 𝑑 = 2√0.4𝐸𝑏 …3.9.4
This distance is significantly less than the distance between adjacent QPSK signals where, 𝑑 = 2√𝐸𝑏 ;
however the distance is greater than 16 MPSK where
𝜋
𝑑 = √16𝐸𝑏 𝑠𝑖𝑛2 = 2√0.15𝐸𝑏 …3.9.5
16
Thus 16 QASK will have a lower error rate than 16 MPSK, but higher than QPSK.
Generation of QPSK

Figure 3.9.2 Generation of QPSK Signal

Page 22 of 34

Page no: 22 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

QASK generator for 4 bit symbol is shown. The 4 bit symbol b k+3 bk+2 bk+1 bk is stored in the 4 bit register
made up of four flip flops. A new symbol is presented once per interval T s= 4Tb and the content of the
register is correspondingly updated at each active edge of the clock which also have a period T s. Two bits
are presented to one D/A converter and two to second converter. The converter output Ae(t) modulates
the balanced modulator whose input career is the even function √𝑃𝑠 cos(𝜔𝑜 𝑡) and Ao(t) modulates the
modulator whose input career is the odd function career. Then the transmitted signal is
𝑣𝑄𝐴𝑆𝐾 (𝑡) = 𝐴𝑒 (𝑡)√𝑃𝑠 . cos(𝜔𝑜 𝑡) + 𝐴𝑜 (𝑡)√𝑃𝑠 . sin(𝜔𝑜 𝑡) …3.9.6

Bandwidth of QASK
The Bandwidth of the QASK signal is
B=2fb/N
Which is the same as in the case of M ary PSK. With N=4 corresponding to 16 possible distinguishable
signals we have BQASK(16) =fb/2 which is one fourth of the bandwidth required for binary PSK.

QASK Receiver

Figure 3.9.3 The QASK Receiver

It is similar to QPSK receiver, where a set of quadrature careers for synchronous demodulation is
generated by raising the received signal to the power 4, extracting the components at frequency 4f o and
then dividing the frequency by 4.
In present case since the coefficients A e and Ao are not of fixed value we have to enquire whether the
career is still recoverable. We have
4 (𝑡) = 𝑃𝑠2 [𝐴𝑒 (𝑡). cos(𝜔𝑜 𝑡) + 𝐴𝑜 (𝑡). sin(𝜔𝑜 𝑡)]4
𝑣𝑄𝐴𝑆𝐾 …3.9.7
Neglecting all the terms not at the frequency 4fo, we have

4
𝑣𝑄𝐴𝑆𝐾 (𝑡 ) 𝐴4𝑒 (𝑡) + 𝐴4𝑜 (𝑡) − 6𝐴2𝑒 (𝑡)𝐴2𝑜 (𝑡)
=[ ] cos(4𝜔𝑜 𝑡)
𝑃𝑠 8
𝐴𝑒 (𝑡)𝐴𝑜 (𝑡)[𝐴2𝑒 (𝑡) − 𝐴2𝑜 (𝑡)]
+[ ] sin(4𝜔𝑜 𝑡)
2 …3.9.8
The average value of the coefficient of cos 4𝜔𝑜 𝑡 is not zero whereas the average value of the coefficient of
sin 4𝜔𝑜 𝑡 is zero. Thus a narrow filter centered at 4f o will recover the signal at 4fo.
After getting the careers, two balanced modulators together with two integrators recover the signals A e(t)
and Ao(t) as shown in figure. The integrators have an integration time equal to the symbol time T s. Finally
the original input bits are recovered by using A/D Converter.

Page 23 of 34

Page no: 23 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

3.10 Binary Frequency Shift Keying (BFSK)

In binary frequency-shift keying (BFSK) the binary data waveform d(t) generates a binary signal

𝑣𝐵𝐹𝑆𝐾 (𝑡) = √2𝑃𝑠 cos[𝜔𝑜 𝑡 + 𝑑(𝑡)Ω𝑡] …3.10.1

Here d(t) = + 1 or -1 corresponding to the logic levels 1and 0 of the data waveform. The transmitted signal
is of amplitude √2𝑃𝑠 and is either
𝑣𝐵𝐹𝑆𝐾 (𝑡) = 𝑆𝐻 (𝑡) = √2𝑃𝑠 cos(𝜔𝑜 + Ω) 𝑡 …3.10.2
𝑣𝐵𝐹𝑆𝐾 (𝑡) = 𝑆𝐿 (𝑡) = √2𝑃𝑠 cos(𝜔𝑜 − Ω) 𝑡 …3.10.3
and thus has an angular frequency ωo + Ω or ωo - Ω with Ω a constant offset from the nominal carrier
frequency ωo. We shall call the higher frequency ωH( = ωo + Ω) and the lower frequency ωL.( = ωo - Ω). We
may conceive that the BFSK signal is generated in the manner indicated in Fig. 3.7.1. Two balanced
modulators are used, one with carrier ωH and one with carrier ωL. The voltage values of PH(t) and of PL(t)
are related to the voltage values of d(t) in the following manner

d(t) PH(t) PL(t)

+1V +1V 0V
-1V 0V +1V

Thus when d(t) changes from +1 to -1 PH changes from 1 to 0 and PL from 0 to 1. At any time either PH or PL
is 1 but not both so that the generated signal is either at angular frequency ωH or at ωL.

Figure 3.10.1 A representation of a manner in which a BFSK signal can be generated.

Spectrum of BPSK
In terms of variables PH and PL the BFSK signal is given by

𝑣𝐵𝐹𝑆𝐾 (𝑡) = √2𝑃𝑠 𝑃𝐻 cos[𝜔𝐻 𝑡 + 𝜃𝐻 ] + √2𝑃𝑠 𝑃𝐿 cos[𝜔𝐿 𝑡 + 𝜃𝐿 ] …3.10.4

where we have assumed that each of the two signals are of independent and random, uniformly
distributed phase. Each of the terms in Eq. (3.10.4) looks like the signal √2𝑃𝑠 𝑏(𝑡) cos 𝜔𝑜 𝑡 which we
encountered in BPSK and for which we have already deduced the spectrum, but there is an important
difference. In the BPSK case, b(t) is bipolar, i.e., it alternates between + 1 and – 1 while in the present case
PH and PL are unipolar, alternating between + 1 and 0. We may, however, rewrite PH and PL as the sums of a
constant and a bipolar variable, that is

1 1 ′ …3.10.5a
𝑃𝐻 (𝑡) = + 𝑃𝐻 (𝑡)
2 2
1 1 …3.10.5b
𝑃𝐿 (𝑡) = + 𝑃𝐿′ (𝑡)
2 2
In above equations 𝑃𝐻′ (𝑡) and 𝑃𝐿′ (𝑡) are bipolar alternating between +1 and -1 and are complementary i.e.
when 𝑃𝐻′ (𝑡) = +1, 𝑃𝐿′ (𝑡) = −1 and vice versa. Then from equation 4,

Page 24 of 34

Page no: 24 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

…3.10.6
𝑃𝑠 𝑃𝑠 𝑃𝑠
𝑣𝐵𝐹𝑆𝐾 (𝑡) = √ cos[𝜔𝐻 𝑡 + 𝜃𝐻 ] + √ cos[𝜔𝐿 𝑡 + 𝜃𝐿 ] + √ 𝑃𝐻′ cos[𝜔𝐻 𝑡 + 𝜃𝐻 ]
2 2 2

𝑃𝑠
+ √ 𝑃𝐿′ cos[𝜔𝐿 𝑡 + 𝜃𝐿 ]
2

The first two terms in Eq. (3.10.6) produce a power spectral density which consists of two impulses, one at
fH and one at fL. The last two terms produce the spectrum of two binary PSK signals one centered about fH
and one about fL. The individual power spectral density patterns of the last two terms are for the case fH - fL
= 2fb. For this separation between fH and fL we observe that the overlapping between the two parts of the
spectra is not large and we may expect to be able, to distinguish the levels of the binary waveform d(t). In
any event, with this separation the bandwidth of BFSK is
𝐵𝑊𝐵𝐹𝑆𝐾 = 4𝑓𝑏 …3.7.7
which is twice the bandwidth of BPSK.

Figure 3.10.2 Power Spectral Densities of Equation 3.7.6

BFSK Receiver:

Figure 3.10.3 A Receiver for BFSK Signal

A BFSK signal is typically demodulated by a receiver system as in Fig.3.10.3. The signal is applied to two
bandpass filters one with center frequency at fH the other at fL. Here we have assumed, that fH - fL =
2(Ω/2π) = 2fb. The filter frequency ranges selected do not overlap and each filter has a passband wide

Page 25 of 34

Page no: 25 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

enough to encompass a main lobe in the spectrum of Fig. 3.10.2. Hence one filter will pass nearly all the
energy in the transmission at fH the other will perform similarly for the transmission at fL. The filter outputs
are applied to envelope detectors and finally the envelope detector outputs are compared by a
comparator. A comparator is a circuit that accepts two input signals. It generates a binary output which is
at one level or the other depending on which input is larger. Thus at the comparator output the data d(t)
will be reproduced.
When noise is present, the output of the comparator may vary due to the systems response to the signal
and noise. Thus, practical systems use a bit synchronizer and an integrator and sample the comparator
output only once at the end of each time interval Tb.

Geometrical Representation of Orthogonal BFSK

In M-ary phase-shift keying and in quadrature-amplitude shift keying, any signal could be represented as
C1u1(t) + C2u2(t). There u1(t) and u2(t) are the orthonormal vectors in signal space, that is, 𝑢1(𝑡) =
2 2
√𝑇 . cos(𝜔𝑜 𝑡)and 𝑢2(𝑡) = √𝑇 . sin(𝜔𝑜 𝑡). The functions u1 and u2 are orthonormal over the symbol
𝑠 𝑠

interval TS. And, if the symbol is a single bit, TS = Tb. The coefficients C1 and C2 are constants. The
normalized energies associated with C1u1(t) and with C2u2(t)are respectively C12 and C22 and the total signal
energy is C12 + C22. In M-ary PSK and QASK the orthogonality of the vectors u1 and u2 results from their
phase quadrature. In the present case of BFSK it is appropriate that the orthogonality should result from a
special selection of the frequencies of the unit vectors. Accordingly, with m and n integers, let us establish
unit vectors
2
𝑢1 (𝑡) = √ cos 2𝜋𝑚𝑓𝑏 𝑡
𝑇𝑏 …3.10.8
2
and 𝑢2 (𝑡) = √ cos 2𝜋𝑛𝑓𝑏 𝑡
𝑇𝑏 …3.10.9
th th
Where fb=1/Tb. The vectors U1 and U2 are the m and n harmonics of the (fundamental) frequency fb. As
we are aware, from the principles of Fourier analysis, different harmonics (m ± n) are orthogonal over the
interval of the fundamental period Tb = 1/fb.
If now the frequencies fH and fL in a BFSK system are selected to be (assuming m > n)
𝑓𝐻 = 𝑚𝑓𝑏 …3. 10.10a
and 𝑓𝐿 = 𝑛𝑓𝑏 …3. 10.10b
Then corresponding signal vectors are
𝑆𝐻 (𝑡) = √𝐸𝑏 𝑢1 (𝑡) …3. 10.11a
and 𝑆𝐿 (𝑡) = √𝐸𝑏 𝑢2 (𝑡) …3. 10.11b

The signal space representation of these signals is

shown in Fig. 3. 10.4. The signals, like the unit vectors
are orthogonal. The distance between signal end
points is therefore

𝑑 = √2𝐸𝑏

Note that this distance is considerably smaller than

the distance separating end points of BPSK signals,
which are antipodal.

Page 26 of 34

Page no: 26 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 3.10.4 Signal Space representation of BFSK

Geometrical Representation of Non-Orthogonal BFSK

When the two FSK signals SH(t) and SL(t) are not orthogonal, the Gram-Schmidt procedure can still be used
to represent the signals of Eqs. 3.10.2 and 3.10.3.
Let us represent the higher frequency signal SH(t) as:
𝑆𝐻 (𝑡) = √2𝑃𝑆 cos 𝜔𝐻 𝑡 = 𝑆11 𝑢1 (𝑡) 0 ≤ 𝑡 ≤ 𝑇𝑏 …3. 10.12a
and 𝑆𝐿 (𝑡) = √2𝑃𝑆 cos 𝜔𝐿 𝑡 = 𝑆12 𝑢1 (𝑡) + 𝑆22 𝑢2 (𝑡) 0 ≤ 𝑡 ≤ 𝑇𝑏 …3. 10.12b
The representation of these two signals in signal space is shown in Fig. 3. 10.5.
Referring to this figure we see that the distance separating SH and SL is:
2 2 2 2 2
𝑑𝐵𝐹𝑆𝐾 = (𝑆11 − 𝑆12 )2 + 𝑆22 = 𝑆11 − 2𝑆11 𝑆12 + 𝑆12 + 𝑆22 …3. 10.13

In order to determine d2BFSK when the two signals are not

orthogonal we must evaluate S11, S12, and S22 using Eqs.
(3. 10.12). From Eq. 3. 10.12a we have:
2 T sin 2ωH Tb
S11 = 2PS ∫0 b cos 2 ωH t dt = Eb [1 + ]…3. 10.14
2ωH Tb

Using Eq. (3. 10.12b) we first determine S12 by multiplying

both sides of the equation by u1(t) and integrating from
0 ≤ 𝑡 ≤ 𝑇𝑏 . The result is:
𝑆12 = √2𝑃𝑆 𝑢1 (𝑡) cos 𝜔𝐿 𝑡 𝑑𝑡
𝐸 Sin(𝜔𝐻 −𝜔𝐿 )𝑇𝑏 Sin(𝜔𝐻 +𝜔𝐿 )𝑇𝑏
𝑆12 = 𝑆 𝑏 [ − ] …3.7.15a
11 (𝜔𝐻 −𝜔𝐿 )𝑇𝑏 (𝜔𝐻 +𝜔𝐿 )𝑇𝑏
By using equation 3.10.12a where,
𝑢1 (𝑡) = (√2𝑃𝑆 /𝑆11 ) cos 𝜔𝐻 𝑡 …3.7.15b
Finally, S22 is found from Eq. 3.10.12b by squaring both
sides of the equation and then integrating from 0 to Tb.
Since u1 and u2 are orthogonal, the result is: Figure 3.10.5 Signal Space representation of
BFSK when SH(t) and SL(t) are not orthogonal

𝑇𝑏 𝑇𝑏
∫ SL2 (t) 𝑑𝑡 = 2𝑃𝑆 ∫ 𝑐𝑜𝑠 2 𝜔𝐻 𝑡 𝑑𝑡 = 𝑆12
2 2
+ 𝑆22
0 0 …3. 10.16a
Therefore 2 2
sin 2𝜔L 𝑇𝑏
𝑆12 + 𝑆22 = 𝐸𝑏 [1 + ]
2𝜔𝐿 𝑇𝑏 …3. 10.16b

The distance d between SH and SL given in Eq. (6.8-14) can now be determined by substituting Eqs. 3.
10.14, 3. 10.15a, and 3. 10.16b into Eq. 3.10.13 The result is:

sin 2ωH Tb Sin(𝜔𝐻 − 𝜔𝐿 )𝑇𝑏 Sin(𝜔𝐻 + 𝜔𝐿 )𝑇𝑏 sin 2𝜔L 𝑇𝑏

𝑑 2 = Eb [1 + ] − 2Eb [ + ] +𝐸𝑏 [1 + ]
2ωH Tb (𝜔𝐻 − 𝜔𝐿 )𝑇𝑏 (𝜔𝐻 + 𝜔𝐿 )𝑇𝑏 2𝜔𝐿 𝑇𝑏 …3. 10.18
In above equation,
sin 2ωH Tb
| |≪1
2ωH Tb
sin 2𝜔L 𝑇𝑏
| |≪1
2𝜔𝐿 𝑇𝑏
And Sin(𝜔𝐻 + 𝜔𝐿 )𝑇𝑏 Sin(𝜔𝐻 − 𝜔𝐿 )𝑇𝑏
| |≪| |
(𝜔𝐻 + 𝜔𝐿 )𝑇𝑏 (𝜔𝐻 − 𝜔𝐿 )𝑇𝑏
Then the final result is then

Page 27 of 34

Page no: 27 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Sin(𝜔𝐻 − 𝜔𝐿 )𝑇𝑏
𝑑 2 ≅ 2𝐸𝑏 [1 − ]
(𝜔𝐻 − 𝜔𝐿 )𝑇𝑏 …3.10.19
Here SH(t) and SL(t) are orthogonal (ωH-ωL)Tb=2π(m-n)fbTb=2π(m-n) and the above equation given 𝑑 =
√2𝐸𝑏 .
Note that if (ωH-ωL)Tb=3π/2, the distance d increases and becomes,
2 1/2
𝑑𝑜𝑝𝑡 = [2𝐸𝑏 (1 + )] = √2.4𝐸𝑏 …3. 10.20
3π
d2 is increased by 20%.

3.11 Comparison of BFSK and BPSK

Let us start with the BFSK signal

𝑣𝐵𝐹𝑆𝐾 (𝑡) = √2𝑃𝑠 cos[𝜔𝑜 𝑡 + 𝑑(𝑡)Ω𝑡]

Using the trigonometric identity for the cosine of the sum of two angles and recalling that cos θ = cos ( - θ)
while sin θ = - sin ( -θ) we are led to the alternate equivalent expression

𝑣𝐵𝐹𝑆𝐾 (𝑡) = √2𝑃𝑠 cos Ω𝑡 cos 𝜔𝑜 𝑡 − √2𝑃𝑠 𝑑(𝑡) sin Ω𝑡 sin 𝜔𝑜 𝑡 …3.11.1

Note that the second term in above equation looks like the signal encountered in BPSK i.e., a carrier sinω0t
multiplied by a data bit d(t) which changes the carrier phase. In the present case however, the carrier is not
of fixed amplitude but rather the amplitude is shaped by the factor sin Ωt. We note further the presence of
a quadrature reference term cos Ωt cos ω0t which contains no information. Since this quadrature term
carries energy, the energy in the information bearing term is thereby diminished. Hence we may expect
that BFSK will not be as effective as BPSK in the presence of noise. For orthogonal BFSK, each term has the
same energy, hence the information bearing term contains only one-half of the total transmitted energy.

The Generation of BFSK is easier but it has many disadvantages…

• Bandwidth is greater in comparison with BPSK, almost double. (Because we are using two career
signals.)
• Error rate of BFSK is higher

3.12 M-ARY FSK

An M-ary FSK communications system is shown in Fig. 3.12.1. It is an obvious extension of a binary FSK
system. At the transmitter an N-bit symbol is presented each TS, to an N-bit D/A converter. The converter
output is applied to a frequency modulator, i.e., a piece of hardware which generates a carrier waveform
whose frequency is determined by the modulating waveform. The transmitted signal, for the duration of
the symbol interval, is of frequency f0 or f1 ...or fM-1 with M = 2N. At the receiver, the incoming signal is
applied to M paralleled bandpass filters each followed by an envelope detector. The bandpass filters have
center frequencies f0, f1, ... ,fM-1. The envelope detectors apply their outputs to a device which determines
which of the detector indications is the largest and transmits that envelope output to an N-bit AID
converter.
The probability of error is minimized by selecting frequencies f0, f1, ... ,fM-1 so that the M signals are
mutually orthogonal. One commonly employed arrangement simply provides that the carrier frequency be
successive even harmonics of the symbol frequency fS=1/TS. Thus the lowest frequency, say f0, is f0 = k fS,
while f1 = (k + 1) fS, f2 = (k + 2) fS etc. In this case, the spectral density patterns of the individual possible
transmitted signals overlap in the manner shown in Fig. 3.12.1, which is an extension to M-ary FSK of the
pattern of Fig. 3.9.2, which applies to binary FSK. We observe that to pass M-ary FSK the required spectral
range is

Page 28 of 34

Page no: 28 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝐵 = 2𝑀𝑓𝑆 …3.12.1
Since fS= fb/N and M=2N, we have
𝐵 = 2𝑁+1 𝑓𝑏 /𝑁 …3.12.2

Figure 3.12.1 An M-ARY Communication System

Note that M-ary FSK requires a considerably increased bandwidth in comparison with M-ary PSK.
However, as we shall see, the probability of error for M-ary FSK decreases as M increases, while for M-ary
PSK, the probability of error increases with M.

Figure 3.12.2 Power Spectral Density of an M-ARY FSK (Four Frequencies are shown)
Geometrical Representation of an M-ARY FSK
In Fig 3.7.4, we provided a signal space representation for the case of orthogonal binary FSK.

Figure 3.12.3 Geometrical representation of orthogonal M-ary FSK (M = 3) when the frequencies are
selected to generate orthogonal signals.

Page 29 of 34

Page no: 29 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The case of M-ary orthogonal FSK signals is clearly an extension of this figure. We simply conceive of a
coordinate system with M mutually orthogonal coordinate axes. The square of the length of the signal
vector is the normalized signal energy. Note that, as in Fig. 3.7.4, the distance between signal points is
3.9.3.
Note that this value of d is greater than the values of d calculated for M-ary PSK with the exception of the
cases M = 2 and M = 4. It is also greater than d in the case of 16-QASK.
𝑑 = √2𝐸𝑠 = √2𝑁𝐸𝑏 …3.12.3

3.13 Minimum Shift Keying (MSK)

There are two important differences between QPSK and MSK:
1. In MSK the baseband waveform, that multiplies the quadrature carrier, is much "smoother" than
the abrupt rectangular waveform of QPSK. While the spectrum of MSK has a main center lobe
which is 1.5 times as wide as the main lobe of QPSK; the side lobes -in MSK are relatively much
smaller in comparison to the main lobe, making filtering much easier.
2. The waveform of MSK exhibits phase continuity, that is, there are not abrupt. phase changes as in
QPSK. As a result we avoid the inter-symbol interference caused by nonlinear amplifiers.
The waveforms of MSK are shown in figure 3.13.1. In (a) we start with a typical data bit stream b(t). This bit
stream is divided into an odd an even bit stream in (b) and (c), as in the manner of OQPSK. The odd stream
bo(t) consists of the alternate bits b1, b3, etc., and the even stream be(t) consists of b2, b4; etc. Each bit in
both streams is held for two bit intervals 2 Tb = Ts, the symbol time. The staggering, which is optional in
QPSK, is essential in MSK. The staggering is that the changes in the odd and even stream do not occur at
the same time.
Also generated at the MSK transmitter are the waveforms sin 2π(t/4Tb) and cos 2π(t/4Tb) as in (d). These
waveforms, and their phases with respect to the bit streams bo(t) and be(t), meet the essential
requirements that sin 2π(t/4Tb) passes through zero precisely at the end of the symbol time in be(t) and cos
2π(t/4Tb) passes through zero at the end of the symbol time in b o(t). We now .generate the products be(t)
sin 2π(t/4Tb) and bo(t) cos 2π(t/4Tb) which are shown in (e) and (f).
In MSK the transmitted signal is
𝑡 𝑡
𝑣𝑀𝑆𝐾 (𝑡) = √2𝑃𝑠 [𝑏𝑒 (𝑡) sin 2𝜋 ( )] cos 𝜔𝑜 𝑡 + √2𝑃𝑠 [𝑏0 (𝑡) cos 2𝜋 ( )] sin 𝜔𝑜 𝑡
4𝑇𝑏 4𝑇𝑏 …3.13.1

In MSK, the carriers are multiplied by the" smoother" waveforms shown in Fig. 3.13.1(e) and (f). As we may
expect, the side lobes generated by these smoother waveforms will be smaller than those associated with
the rectangular waveforms and hence easier to suppress as is required to avoid interchannel interference.
In Eq. (3.13.1) MSK appears as a modified form of OQPSK, which we can call" shaped Q'PSK". We can,
however, rewrite the equation to make it apparent that MSK is an FSK system. Applying the trigonometric
identities for the products of sinusoids we find that Eq. (3.10.1) can be written:

𝑏𝑜 (𝑡) + 𝑏𝑒 (𝑡) 𝑏𝑜 (𝑡) − 𝑏𝑒 (𝑡)

𝑣𝑀𝑆𝐾 (𝑡) = √2𝑃𝑠 [ ] sin(𝜔𝑜 + Ω)𝑡 + √2𝑃𝑠 [ ] cos(𝜔𝑜 + Ω)𝑡
2 2 …3.13.2a
Where 𝛺 = 2𝜋/4𝑇𝑏 = 2𝜋(𝑓𝑏/4) …3.13.2b

If we define CH = (b0+be)/2, CL==(b0-be)/2, ωH = ω0 + Ω, ωH = ω0 – Ω then equations 3.10.2 becomes

vMSK (t) = √2Ps CH (t) sin ωH t + √2Ps CL (t)ωL t …3.13.3
Now b; = + 1 and be = + 1, so that as is easily verified, if b0, = be then CL = 0 while CH = b0 = ± 1. Further, if b0
= -be then CH = 0 and CL = b0 = + 1.
Thus, depending on the value of the bits b o and be in each bit interval, the transmitted signal is at angular
frequency ωH or at ωL precisely as in FSK and the magnitude of the amplitude is always equal to √2Ps .
In MSK, the two frequencies fH and fL, are chosen to insure that the two possible signals are orthogonal
over the bit interval Tb. That is, we impose the constraint that

Page 30 of 34

Page no: 30 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝑇𝑏
∫ sin 𝜔𝐻 𝑡 sin 𝜔𝐿 𝑡 𝑑𝑡 = 0 …3.13.4
0

Figure 3.13.1 MSK Waveforms

The equation 3.13.4 will be satisfied provided that it is arranged, with m and n integers, that
2𝜋(𝑓𝐻 − 𝑓𝐿 )𝑇𝑏 = 𝑛𝜋 …3.13.5a
and 2𝜋(𝑓𝐻 + 𝑓𝐿 )𝑇𝑏 = 𝑚𝜋 …3.13.5b

Page 31 of 34

Page no: 31 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Also
𝑓𝑏 …3.13.06a
𝑓𝐻 = 𝑓0 +
4
𝑓𝑏 …3.13.6b
and 𝑓𝐿 = 𝑓0 −
4
From equations 3.13.5 and 3.13.6
1 …3.13.7a
𝑓𝑏 𝑇𝑏 = 𝑓𝑏 =1=𝑛
𝑓𝑏
𝑚 …3.13.7b
and 𝑓0 = 𝑓𝑏
4
Equation (3.13.7a) shows that since n = 1, fH and fL are as close together as possible for orthogonality to
prevail. It is for this reason that the present system is called" minimum shift keying." Equation (3.13.7b)
shows that the carrier frequency fo is an integral multiple of fb/4. Thus
𝑓𝑏 …3.13.8a
𝑓𝐻 = (𝑚 + 1)
4
𝑓𝑏 …3.13.8b
and 𝑓𝐿 = (𝑚 − 1)
4
Signal Space Representation of MSK
The signal space representation of MSK is shown in Fig. 3.13.2 The orthonormal unit vectors of the
coordinate system are given by 𝑢H (t) = √2/𝑇𝑠 sin ωH t and 𝑢L (t) = √2/𝑇𝑠 sin ωL t. The end points of the
four possible signal vectors are indicated by dots. The smallest distance between signal points is
𝑑 = √2𝐸𝑠 = √4𝐸𝑏 …3.13.9
just as for the case of QPSK.

Figure 3.13.2 Signal Space Representation of MSK

We recall that QPSK generates two BPSK signals which are orthogonal to one another by virtue of the fact
that the respective carriers are in phase quadrature. Such phase quadrature can also be characterized as
time quadrature since, at a carrier frequency f0 a phase shift of π/2 is accomplished by a time shift in
amount 1/4f0, that is sin 2πf0(t+1/4f0) = sin (2πf0t + π/2) = cos (2πf0t). It may be noted that in MSK we have
again two BPSK signals. Here, however, the respective carriers are orthogonal to one another by virtue of
the fact that they are in frequency quadrature.

Page 32 of 34

Page no: 32 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Generation and Reception of MSK

One way to generate a MSK signal is the following: We start with sin Ωt and sin ω0t, and use 90° phase
shifters to generate sin (Ωt + π/2) = cosΩt and sin (ω0t + π/2) = cosω0t. We then use multipliers to form the
products sinΩt cosω0t and cosΩt sin ω0t. Additional multipliers generate √2𝑃𝑠 𝑏𝑒 (𝑡)sin Ωt cos 𝜔0 t and
√2𝑃𝑠 𝑏0 (𝑡)cos Ωt sin 𝜔0 t. Finally an adder is used to form the sum. An alternative and more favored
scheme is shown in Fig. 3.13.3a. This technique has the merit that it avoids the need for precise 90° phase
shifters at angular frequencies ωo and Ω.

Figure 3.13.3 MSK Modulation and demodulation

The MSK receiver is shown in Fig. 3.13.3b. Detection is performed synchronously, i.e., by determining the
correlation of the received signal with the waveform x(t) = cosΩt sin ω0t to determine bit b0(t), and with
y(t) =sinΩt cosω0t to determine bit be(t). The integration is performed over the symbol interval. The
integrators integrate over staggered overlapping intervals of symbol time Ts = 2Tb.

Figure 3.13.4 Regeneration of x(t) and y(t)

Page 33 of 34

Page no: 33 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

At the end of each integration time the integrator output is stored and then the integrator output is
dumped. The switch at the output swings back and forth at the bit rate so that finally the output waveform
is the original data bit stream dk(t).
At the receiver we need to reconstruct the waveforms x(t) and y(t). A method for locally regenerating x(t)
and y(t) is shown in Fig. 3.10.4. From Eq3.13.3 we see that MSK consists of transmitting one of two possible
HPSK signals, the first at frequency ω0 – Ω and the second at frequency ω0 + Ω. Thus as in BPSK detection,
we first square and filter the incoming signal. The output of the squarer has spectral components at the
frequency 2ωH = 2(ω0 + Ω) and at 2ωL = 2(ω0 – Ω). These are separated out by band pass filters. Division by
2 yields waveforms (1/2)sin 𝜔H t and (1/2)sin 𝜔L t from which, as indicated, x(t) and y(t) are regenerated by
addition and subtraction respectively. Further, the multiplier and low-pass filter shown regenerate a
waveform at the symbol rate Is =fb/2 which can 'be used to operate the sampling switches in Fig. 3.13.3b.

Page 34 of 34

Page no: 34 Get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]
Program : B.Tech
Subject Name: Digital Communication
Subject Code: EC-502
Semester: 5th
Downloaded from www.rgpvnotes.in

Department of Electronics and Communication Engineering

Sub. Code: EC 5002 Sub. Name: Digital Communication

Unit 4
Syllabus:
Other Digital Techniques: Pulse shaping to reduce inter channel and inter symbol interference- Duo binary
encoding, Nyquist criterion and partial response signaling, Quadrature Partial Response (QPR) encoder
decoder, Regenerative Repeater- eye pattern, equalizers
Optimum Reception of Digital Signals: Baseband signal receiver, probability of error, maximum likelihood
detector, Bayes theorem, optimum receiver for both baseband and pass band receiver- matched filter and
correlator, probability of error calculation for BPSK and BFSK.

4.1 Inter Symbol Interference

This is a form of distortion of a signal, in which one or more symbols interfere with subsequent signals,
causing noise or delivering a poor output.
Causes of ISI
The main causes of ISI are −
 Multi-path Propagation
 Non-linear frequency in channels
The ISI is unwanted and should be completely eliminated to get a clean output. The causes of ISI should
also be resolved in order to minimize its effect.
To view ISI in a mathematical form present in the receiver output, we can consider the receiver output.
The receiving filter output y(t)
is sampled at time ti=iTb
(with i taking on integer values), yielding –
∞

𝑦(𝑡𝑖 ) = 𝜇 ∑ 𝑎𝑘 𝑝(𝑖𝑇𝑏 − 𝑘𝑇𝑏 )]

𝑘=−∞
∞

= 𝜇𝑎𝑖 + 𝜇 ∑ 𝑎𝑘 𝑝(𝑖𝑇𝑏 − 𝑘𝑇𝑏 )]

𝑘=−∞
𝑘≠𝑖
In the above equation, the first term μai is produced by the ith transmitted bit.
The second term represents the residual effect of all other transmitted bits on the decoding of the ith bit.
This residual effect is called as Inter Symbol Interference.
In the absence of ISI, the output will be −
𝑦(𝑡𝑖 ) = 𝜇𝑎𝑖

This equation shows that the ith bit transmitted is correctly reproduced. However, the presence of ISI
introduces bit errors and distortions in the output.
While designing the transmitter or a receiver, it is important that you minimize the effects of ISI, so as to
receive the output with the least possible error rate.

Correlative Coding
So far, we’ve discussed that ISI is an unwanted phenomenon and degrades the signal. But the same ISI if
used in a controlled manner is possible to achieve a bit rate of 2W bits per second in a channel of
bandwidth W Hertz. Such a scheme is called as Correlative Coding or Partial response signaling schemes.

Correlative Level Coding:

Page 1 of 19

Page no: 1 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Correlative-level coding (partial response signaling) – adding ISI to the transmitted signal in a controlled
manner Since ISI introduced into the transmitted signal is known, its effect can be interpreted at the
receiver. A practical method of achieving the theoretical maximum signaling rate of 2W symbol per second
in a bandwidth of W Hertz.

Using realizable and perturbation-tolerant filters

Since the amount of ISI is known, it is easy to design the receiver according to the requirement so as to
avoid the effect of ISI on the signal. The basic idea of correlative coding is achieved by considering an
example of Duo-binary Signaling.

4.2 Duo-binary Signaling

If fM is the frequency of the maximum frequency spectral component of the baseband waveform, then, in
AM, the bandwidth is B = 2fM. In frequency modulation, if the modulating waveform were a sinusoid of
frequency fM, and if the frequency deviation was ∆f, then bandwidth would be

𝐵 = 2∆𝑓 + 2𝑓𝑀 …4.2.1

Altogether, it is apparent that bandwidth decreases with decreasing fM regardless of the modulation
technique employed. We consider now a mode of encoding a binary bit stream, called duobinary encoding
which effects a reduction of the maximum frequency in comparison to the maximum frequency of the un-
encoded data. Thus, if a carrier is amplitude or frequency modulated by a duobinary encoded waveform,
the bandwidth of the modulated waveform will be smaller than if the un-encoded data were used to AM or
FM modulate the carrier.
There are a number of methods available for duobinary encoding and decoding. One popular scheme is
shown in Fig. 4.2.1. The signal d(k) is the data bit stream with bit duration Tb. It makes excursions between
logic 1 and logic 0, and, as had been our custom, we take the corresponding voltage levels to be + 1V and -
1V. The signal b(k), at the output of the differential encoder also makes excursions between + 1V and -1V.
The waveform vd(k) is therefore
𝑣𝑑 (𝑘) = 𝑏(𝑘) + 𝑏(𝑘 − 1) …4.2.2

Figure 4.2.1 The Duobinary Encoder Decoder System

which can take on the values vD(k) = +2V, 0V and - 2V. The value of vD(k) in any interval k depends on both
b(k) and b(k - 1). Hence there is a correlation between the values of vD(k) in any two successive intervals.
For this reason the coding of Fig. 4.2.1 is referred to as correlative coding.
The correlation can be made apparent in another way. When the transition is made from one interval to
the next, it is not possible for vD(k) to change from +2V to - 2V or vice versa. In short, in any interval, vD(k)
cannot always assume any of the possible levels independently of its level in the previous intervals. Finally,
we note that the term duobinary is appropriate since in each bit interval, the generated voltage vD(k)
results from the combination of two bits.

Page 2 of 19

Page no: 2 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The decoder, shown in Fig. 4.2.1, consists of a device that provides at its output the magnitude (absolute
value) of its input cascaded with a logical inverter. For the inverter we take it that logic 1 is + 1V or greater
and logic 0 is 0V. We can now verify that the decoded data 𝑑̂ (k) is indeed the input data d(k). For this
purpose we prepare the following truth table:
Truth Table For Duobinary Signaling
Adder Input 1 Adder Input 2 Adder output Magnitude Output Inverter output
I1 I2 vD(k) (Inverter output) d(k)
Voltage Logic Voltage Logic Input Voltage Voltage Logic Logic
-1V 0 -1V 0 -2V 2V 1 0
-1V 0 1V 1 0 0V 0 1
1V 1 -1V 0 0 0V 0 1
1V 1 1V 1 2V 2V 1 0

From the table we see that the inverter output is 𝐼1 ⨁𝐼2 .The differential encoder (called a precoder in the
present application) output is:
𝐼1 = 𝑏(𝑘) = 𝑑(𝑘)⨁𝑏(𝑘 − 1) …4.2.3
̂
The input I2 = b(k - 1) so that the inverter output 𝑑 (k) is:

𝑑̂ (k) = 𝐼1 ⨁𝐼2 = 𝑑(𝑘)⨁𝑏(𝑘 − 1)⨁𝑏(𝑘 − 1) = 𝑑(𝑘 ) …4.2.4

Waveforms of Duobinary Signaling

The more rapidly d(k) switches back and forth between logic levels the higher will be the frequencies of the
spectral components generated. When d(k) switches at each time Tb, the switching speed is at a maximum.
The waveform d(k), under such circumstances, has the appearance of a square wave of period 2Tb and
frequency 1/2 Tb as shown in Fig. 4.2.2a.

Figure 4.2.2 Waveforms of d(k), b(k) and vD(k)

If d(k) is the input to the duobinary encoder of Fig. 4.2.1 then, as can be verified, b(k) appears as in Fig.
4.2.2b and the waveform, vD(k) which is to be transmitted appears as in Fig. 4.2.2c. Observe that the period
of vD(k) is 4 Tb with corresponding frequency 1/4 Tb. Thus the frequency of vD(k) is half the frequency of the

Page 3 of 19

Page no: 3 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

original unencoded waveform d(k). The waveform d(k) may be a sinusoid of frequency 1/2 Tb and
waveform vD(k) as a sinusoid of frequency 1/4Tb. If we were free to select either d(k) or vD(k) as a
modulating waveform for a carrier, and if we were interested in conserving bandwidth, we would choose
vD(k). If amplitude modulation were involved, the bandwidth of the modulated waveform would be
2(1/4Tb) = fb /2 using vD(k) since the modulating frequency is fM = 1/4Tb and would be 2(1/2Tb) = fb using
d(k). With frequency modulation, if the peak-to peak carrier frequency deviation were 2∆f, then, the
modulated carrier would have a bandwidth 2(∆f) + 2(1/2Tb) with d(k) as the modulating signal, as in BFSK;
and 2(∆f) + 2(1/4Tb)with vD(k) as the modulating signal.

4.3 Partial Response Signaling

Suppose, that corresponding to each bit of duration Tb, of a data stream we generate a positive impulse of
strength +1 whenever the bit is at logic 1 and a negative impulse -1 whenever the bit is at logic 0. Suppose,
further, that these impulses are applied to the input of the cosine filter. In Fig. 6.4.3 we have drawn the
filter responses individually to five successive positive impulses. For simplicity, we have in each case drawn
only the central lobe-and we have indicated by dots all the places where the individual response
waveforms pass through zero. Where there is no dot, the waveforms has a finite value. The peaks of the
responses are separated by times Tb and the widths of the central lobes are 3Tb.
The total response is, of course, simply the sum of the individual responses.
We can make the following observations from
Fig. 4.3.1:
1. If we sample the total response at a time when
an individual response is at its peak, the sample
will have contributions from all the individual
responses.
2. There is no possible time at which a sample of
the total response is due only to a single
individual response.
3. Importantly, if we sample the total response
midway between times when the individual
responses are at peak value, i.e., at t = (2k-1)Tb/2,
then the sample value will have contributions in
equal amount from only the two individual
responses that straddle the sampling time. These
sampling times are indicated in Fig. 4.3.1, by the
light vertical lines. One such sampling time,
yielding contributions from individual responses
2 and 3 is explicitly marked. It can be calculated
that at the sampling time the contribution from
each of the straddling individual responses will
be a voltage Ifb. Note that in sampling at the
indicated times, we sample when the individual
responses are not at peak value. For this reason,
the present signal processing is referred to as
Partial Response Signaling.
Figure 4.3.1 Filter responses to Five Different Impulses

In partial-response signaling, we shall transmit a signal during each bit interval that has contributions from
two successive bits of an original baseband waveform. But this superposition need not prevent us from

Page 4 of 19

Page no: 4 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

disentangling the individual original baseband waveform bits. A complete (baseband) partial-response
signaling communications system is shown in Fig. 4.3.2.

Figure 4.3.2 Duo Binary Encoder and Decoder Using Cosine filter

It is seen to be just an adaptation of duobinary encoding and decoding. The cosine filter employed a delay
and an advance of the impulse by amount Tb/2, the total time between delayed and advanced impulses
being Tb. Since, in the real world, a time advance is not possible, we have employed only a delay by
amount Tb. The brickwall filter at the receiver input serves to remove any out of band noise added to the
signal during transmission. It can be shown, that the output data 𝑑̂ (k) = d(k).

4.4 Quadrature Partial Response (QPR) Encoder and Decoder

Amplitude Modulation of Partial Response Signal
The baseband partial response (duobinary) signal may be used to amplitude or frequency modulate a
carrier. If amplitude modulation is employed, either double sideband suppressed carrier DSB/SC or
quadrature amplitude modulation QAM can be employed.
For the case of DSB/SC the duobinary signal, VT(t), shown in Fig. 4.3.2a, is multiplied by the carrier
√2 cos 𝜔0 (𝑡). The resulting signal is
𝑣𝐷𝑆𝐵 (𝑡) = √2𝑣 𝑇 (𝑡) cos 𝜔0 (𝑡) …4.4.1

The bandwidth required to transmit the signal is twice the bandwidth of the baseband duobinary signal
which is fb/2. Hence the bandwidth BDSB of an amplitude modulated duobinary signal is
𝐵𝐷𝑆𝐵 = 2(𝑓𝑏 /2) = 𝑓𝑏 …4.4.2

If the duobinary signal is to amplitude modulate two carriers in quadrature, the circuit shown in Fig. 4.4.1 is
used and the resulting encoder is called a "quadrature partial response" (QPR) encoder.
Figure 4.4.1 shows that the data d(t) at the bit rate fb is first separated into an even and an odd bit stream
de(t) and do(t) each operating with the bit rate fb /2, Both de(t) and do(t) are then separately duobinary
encoded into signals VTe(t) and VTo(t).
Each duobinary encoder is similar to the encoder shown in Fig. 4.3.2a except that each delay is now 2Tb,
rather than Tb, the data rate of the input is fb/2 rather than fb and the bandwidth of the brick wall filter is
now (1/2)(fb/2)= fb/4 rather than fb/2. Thus the bandwidth required to pass VTe(t) and VTo(t) is fb/4. Each
duobinary signal is then modulated using the quadrature carrier signals cos ωot and sin ωot.

Page 5 of 19

Page no: 5 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.4.1 QPR Encoder

The bandwidth of each of the quadrature amplitude modulated signals is

𝐵𝑄𝑃𝑅 = 2(𝑓𝑏 /4) = 𝑓𝑏 /2 …4.4.3

Hence the total bandwidth required to pass a QPR signal is also BQPR, since the two quadrature components
occupy the same frequency band.
It should be noted that if QPSK, rather than QPR, were used to encode the data d(t), the bandwidth
required would be BQPSK =fb. However, if 16 QAM or 16 PSK were used to encode the data the required
bandwidth would be B16QAM = B16PSK =fb/2. Thus the spectrum required to pass a QPR signal is similar to
that required to pass 16 QAM or 16 PSK. However, the QPR signal displays no (or in practice very small)
side lobes which makes QPR the encoding system of choice when spectrum width is the major problem.
The drawback in using QPR is that the transmitted signal envelope is not a constant but varies with time.

QPR Decoder

Figure 4.4.2 QPR Decoder

A QPR decoder is shown in Fig. 4.4.2. As in 16-QAM and 16-PSK to decode the input signal, VQ(t) is first
raised to the fourth power, filtered and then frequency divided by 4. The result yields the two quadrature

Page 6 of 19

Page no: 6 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

carriers: cos ωot and sin ωot. Using the two quadrature carriers we demodulate VQ(t) and obtain the two
baseband duobinary signals VTe(t) and VTo(t). Duobinary decoding then takes place; each duobinary decoder
being similar to the decoder shown in Fig. 4.3.2b except that they operate at fb/2 rather than at fb. The
reconstructed data do(t) and de(t) is then combined to yield the data d(t).

4.5 Eye Pattern

An effective way to study the effects of ISI is the Eye Pattern. The name Eye Pattern was given from its
resemblance to the human eye for binary waves. The interior region of the eye pattern is called the eye
opening. The following figure shows the image of an eye-pattern.

Figure 4.6 Image of eye pattern

Jitter is the short-term variation of the instant of digital signal, from its ideal position, which may lead to
data errors.
When the effect of ISI increases, traces from the upper portion to the lower portion of the eye opening
increases and the eye gets completely closed, if ISI is very high.
An eye pattern provides the following information about a particular system.
 Actual eye patterns are used to estimate the bit error rate and the signal-to-noise ratio.
 The width of the eye opening defines the time interval over which the received wave can be
sampled without error from ISI.
 The instant of time when the eye opening is wide, will be the preferred time for sampling.
 The rate of the closure of the eye, according to the sampling time, determines how sensitive the
system is to the timing error.
 The height of the eye opening, at a specified sampling time, defines the margin over noise.
Hence, the interpretation of eye pattern is an important consideration.

4.6 Equalization
For reliable communication to be established, we need to have a quality output. The transmission losses of
the channel and other factors affecting the quality of the signal have to be treated. The most occurring
loss, as we have discussed, is the ISI.
To make the signal free from ISI, and to ensure a maximum signal to noise ratio, we need to implement a
method called Equalization. The following figure shows an equalizer in the receiver portion of the
communication system.
Sampled
Noise and
Received
Interference
Signal
Digital Pulse Analog Linear Decision
Source Shaping Channel digital Device
Received TS Equalizer
Analog
Signal

Page 7 of 19

Page no: 7 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.7 Equalization

The noise and interferences which are denoted in the figure are likely to occur, during transmission. The
regenerative repeater has an equalizer circuit, which compensates the transmission losses by shaping the
circuit. The Equalizer is feasible to get implemented.

Error Probability and Figure-of-merit

The rate at which data can be communicated is called the data rate. The rate at which error occurs in the
bits, while transmitting data is called the Bit Error Rate (BER).
The probability of the occurrence of BER is the Error Probability. The increase in Signal to Noise Ratio (SNR)
decreases the BER, hence the Error Probability also gets decreased.
In an Analog receiver, the figure of merit at the detection process can be termed as the ratio of output SNR
to the input SNR. A greater value of figure-of-merit will be an advantage.

Regenerative Repeater
For any communication system to be reliable, it should transmit and receive the signals effectively, without
any loss. A PCM wave, after transmitting through a channel, gets distorted due to the noise introduced by
the channel.
The regenerative pulse compared with the original and received pulse, will be as shown in the following
figure.

Original Pulse Resulting Pulse Restored Pulse

Figure 4.8 Regenerative Repeater

For a better reproduction of the signal, a circuit called as regenerative repeater is employed in the path
before the receiver. This helps in restoring the signals from the losses occurred. Following is the
diagrammatical representation.
Distorted Decision Regenerated
Amplifier and Making
PCM Wave Equalizer Device PCM Wave

Timing
Circuit

Figure 4.9 Block Diagram of Regenerative Repeater

This consists of an equalizer along with an amplifier, a timing circuit, and a decision making device. Their
working of each of the components is detailed as follows.

Equalizer
The channel produces amplitude and phase distortions to the signals. This is due to the transmission
characteristics of the channel. The Equalizer circuit compensates these losses by shaping the received
pulses.
Timing Circuit

Page 8 of 19

Page no: 8 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

To obtain a quality output, the sampling of the pulses should be done where the signal to noise ratio (SNR)
is maximum. To achieve this perfect sampling, a periodic pulse train has to be derived from the received
pulses, which is done by the timing circuit.
Hence, the timing circuit allots the timing interval for sampling at high SNR, through the received pulses.
Decision Device
The timing circuit determines the sampling times. The decision device is enabled at these sampling times.
The decision device decides its output based on whether the amplitude of the quantized pulse and the
noise, exceeds a pre-determined value or not.

4.7 Baseband Signal Receiver:

Consider that a binary-encoded signal consists of a time sequence of voltage levels +V or-V. If there is a
guard interval between the bits, the signal forms a sequence of positive and negative pulses. In either case
there is no particular interest in preserving the waveform of the signal after reception. We are interested
only in knowing within each bit interval whether the transmitted voltage was + V or - V. With noise
present, the received signal and noise together will yield sample values generally different from ± V. In this
case, what deduction shall we make from the sample value concerning the transmitted bit?
Suppose that the noise is Gaussian and therefore the noise voltage has a probability density which is
entirely symmetrical with respect to zero volts. Then the probability that the noise has increased the
sample value is the same as the probability that the noise has decreased the sample value. It then seems
entirely reasonable that we can do no better than to assume that if the sample value is positive the
transmitted level was + V, and if the sample value is negative the transmitted level was - V. It is, of course,
possible that at the sampling time the noise voltage may be of magnitude larger than V and of a polarity
opposite to the polarity assigned to the transmitted bit. In this case an error will be made as indicated in
Fig. 4.7.1. Here the transmitted bit is represented by the voltage + V which is sustained over an interval T
from t1 to t2. Noise has been superimposed on the level + V so that the voltage v represents the received
signal and noise. If now the sampling should happen to take place at a time t=t1+∆t; an error will have been
made.
We can reduce the probability of error by
processing the received signal plus noise in such
a manner that we are then able to find a sample
time where the sample voltage due to the signal
is emphasized relative to the sample voltage due
to the noise. Such a processer (receiver) is shown
in Fig. 4.7.2. The signal input during a bit interval
is indicated. As a matter of convenience we have
set t = q at the beginning of the interval. The
waveform of the signal s(t) before t =0 and after t
= T has not been indicated since, as will appear,
the operation of the receiver during each bit
interval is independent of the waveform during
past and future bit intervals.
Figure 4.7.1 Illustration that noise may cause an error
in determination of transmitted voltage level

The signal s(t) with added white gaussian noise n(t) of power spectral density η/2 is presented to an
integrator. At time t = 0 + we require that capacitor C be uncharged. Such a discharged condition may be
ensured by a brief closing of switch SW1 at time t = 0-, thus relieving C of any charge it may have acquired

Page 9 of 19

Page no: 9 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.7.2 A Receiver for Binary Coded Signal

during the 'previous interval. The sample is taken at the output of the integrator by closing this sampling
switch SW2. This sample is taken at the end of the bit interval, at t = T. The signal processing indicated in
Fig. 4.7.2 is described by the phrase integrate and dump, the term dump referring to the abrupt discharge
of the capacitor after each sampling.

Peak Signal to RMS Noise Output Voltage Ratio

The integrator yields an output which is the integral of its input multiplied by 1/RC. Using τ = RC, we have

1 𝑇 1 𝑇 1 𝑇 …4.7.1
𝑣𝑜 (𝑇) = ∫ [𝑠(𝑡) + 𝑛(𝑡)]𝑑𝑡 = ∫ 𝑠(𝑡)𝑑𝑡 + ∫ 𝑛(𝑡)𝑑𝑡
𝜏 0 𝜏 0 𝜏 0
The sample voltage due to the signal is
1 𝑇 𝑉𝑇 …4.7.2
𝑠𝑜 (𝑇) = ∫ 𝑉𝑑𝑡 =
𝜏 0 𝜏
The sample voltage due to the noise is
1 𝑇 …4.7.3
𝑠𝑛 (𝑇) = ∫ 𝑛(𝑡)𝑑𝑡
𝜏 0
This noise-sampling voltage no(T) is a Gaussian random variable in contrast with n(t) which is a Gaussian
random process.
The variance of no(T) is given by
𝑛𝑇 …4.7.4
𝜎02 = ̅̅̅̅̅̅̅
𝑛02 (𝑡) = 2
2𝜏
It has a Gaussian probability density.

The output, of the integrator, before the sampling switch, is v0(t) = S0(t) + n0(t). As shown in Fig. 4.7.3a, the
signal output So(t) is a ramp, in each bit interval, of duration T. At the end of the interval the ramp attains
the voltage S0(t) which is + VT/τ or - VT/τ, depending on whether the bit is a 1 or a 0. At the end of each
interval the switch SW1 in Fig. 4.7.2 closes momentarily to discharge the capacitor so that so(t) drops to
zero. The noise n0(t) shown in Fig. 4.7.3b, also starts each interval with no(0) = 0 and has the random value
n0(t) at the end of each interval. The sampling switch SW2 closes briefly just before the closing of SW1 and
hence reads the voltage
𝑣𝑜 (𝑇) = 𝑠𝑜 (𝑇) + 𝑛𝑜 (𝑇) …4.7.5

We would naturally like the output signal voltage to be as large as possible in comparison with the noise
voltage. Hence a figure of merit of interest is the signal-to-noise ratio
[𝑠𝑜 (𝑇)]2 2 2
= 𝑉 𝑇 …4.7.6
̅̅̅̅̅̅̅̅̅̅̅
[𝑛𝑜 (𝑇)]2 𝜂

Page 10 of 19

Page no: 10 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.7.3 (a) The Signal Output and (b) the Noise Output of the integrator

This result is calculated from Eqs. (4.7.2) and (4.7.4). Note that the signal-to noise ratio increases with
increasing bit duration T and that it depends on V2T which is the normalized energy of the bit signal.
Therefore, a bit represented by a narrow, high amplitude signal and one by a wide, low amplitude signal
are equally effective, provided V2T is kept constant. It is instructive to note that the integrator filters the
signal and the noise such that the signal voltage increases linearly with time, while the standard deviation
(rms value) of the noise increases more slowly, as √𝑇. Thus, the integrator enhances the signal relative to
the noise, and this enhancement increases with time as shown in Eq. (4.7.6).

4.8 Probability of Error:

Since the function of a receiver of a data transmission is to distinguish the bit 1 from the bit 0 in the
presence of noise, a most important characteristic is the probability that an error will be made in such a
determination. We now calculate this error probability P, for the integrate-and-dump receiver of Fig. 4.7.2.
We have seen .that the probability density of the noise sample no(T) is Gaussian and hence appears as in
Fig. 4.7.1.The density is therefore given by
2 2
𝑒 −𝑛0 (𝑇)/2𝜎0
𝑓 [𝑛𝑜 (𝑇)] = …4.8.1
√2𝜋𝜎02
Where σ02 the variance is given by 𝜎02 = ̅̅̅̅̅̅̅
𝜎02 (𝑡). Suppose, then, that during some bit interval the input-
signal voltage is held at, say, - V. Then, at the sample time, the signal sample voltage is So(T)= -VT/τ, while
the noise sample is no(T). If no(T) is positive and larger in magnitude than VT/ τ, the total sample voltage
vo(T) = so(T) + no(T) will be positive. Such a positive sample voltage will result in an error, since as noted
earlier, we have instructed the receiver to interpret such a positive sample voltage to mean that the signal
voltage was + V during the bit interval. The probability of such a misinterpretation, that is, the probability
that no(T) > VT/ τ, is given by the area of the shaded region in Fig. 4.8.1. The probability of error is, using
Eq. (4.8.1).
∞ ∞ 2 2
𝑒 −𝑛0 (𝑇)/2𝜎0
𝑃𝑒 = ∫ 𝑓 [𝑛𝑜 (𝑇)] 𝑑𝑛𝑜 (𝑇) = ∫ 𝑑𝑛𝑜 (𝑇) …4.8.2
2
𝑉𝑇/𝜏 𝑉𝑇/𝜏 √2𝜋𝜎0

𝑛𝑜 (𝑇)
Defining 𝑥 ≡ and using equations 4.7.4 equation 4.8.2 may be written as
√2𝜎0

1 2 ∞ −𝑥 2 1 𝑇 1 𝑉𝑇 1/2 1 𝐸𝑆 1/2
𝑃𝑒 = ∫ 𝑒 𝑑𝑥 = 𝑒𝑟𝑓𝑐 (𝑉√ ) = 𝑒𝑟𝑓𝑐 ( ) 𝑒𝑟𝑓𝑐 ( ) …4.8.3
2 √𝜋 𝑥=𝑉𝑇/𝜏 2 𝜂 2 𝜂 2 𝜂
In which Es=V2T is the signal energy of a bit.

Page 11 of 19

Page no: 11 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.8.1 The Gaussian Probability Density of the noise sample n0(T)
If the signal voltage were held instead at + V during some bit interval, then it is clear from the symmetry of
the situation that the probability of error would again be given by P, in Eq. (4.8.3). Hence Eq. (4.8.3) gives
P, quite generally.

Figure 4.8.2 Variation of Pe versus Es/η

The probability of error Pe as given in Eq. (4.8.3), is plotted in Fig. 4.8.2. Note that Pe decreases rapidly as
Es/η increases. The maximum value of Pe is 1/2. Thus, even if the signal is entirely lost in the noise so that
any determination of the receiver is a sheer guess, the receiver cannot be wrong more than half the time
on the average.
4.9 The Optimum Receiver
In the receiver system of Fig. 4.7.2, the signal was passed through a filter (i.e. the integrator), so that at the
sampling time the signal voltage might be emphasized in comparison with the noise voltage. We are
naturally led to ask whether the integrator is the optimum filter for the purpose of minimizing the
probability of error. We shall find that for the received signal contemplated in the system of Fig. 4.7.2 the
integrator is indeed the optimum filter.
We assume that the received signal is a binary waveform. One binary digit (bit) is represented by a signal
waveform S1(t) which persists for time T, while the other bit is represented by the waveform S2(t) which
also lasts for an interval T. For example, in the case of transmission at baseband, as shown in Fig. 4.7.2,
S1(t) = + V, while S2(t) = - V; for other modulation systems, different waveforms are transmitted. For
example, for PSK signalling, S1(t) = A cos ω0t and S2(t) = - A cos ω0t; while for FSK, S1(t) = A cos (ω0+Ω)t and
S2(t) = A cos (ω0- Ω)t.

Page 12 of 19

Page no: 12 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.9.1 A Receiver for binary coded signaling

As shown in Fig. 4.9.1 the input, which is S1(t) or S2(t), is corrupted by the addition of noise n(t). The noise
is gaussian and has a spectral density G(f). [In most cases of interest the noise is white, so that G(J) = η/2.
However, we shall assume the more general possibility, since it introduces no complication to do so.] The
signal and noise are filtered and then sampled at the end of each bit interval. The output sample is either
vo(T) = S01(T) + n0(T) or vo(T) = S02(T) + no(T). We assume that immediately after each sample, every energy-
storing element in the filter has been discharged.
We note that in the absence of noise the output sample would be vo(T) = So1(T) or S02(T). When noise is
present we have shown that to minimize the probability of error one should assume that S 1(t) has been
transmitted if v0(T) is closer to S01(T) than to S02(T). Similarly, we assume S2(t) has been transmitted if vo(T)
is closer to S02(T). The decision boundary is therefore midway between So1(T) and S02(T). For example, in
the baseband system of Fig. 4.7.2, where S01(T) = VT/τ and S02(T) = - VT/τ, the decision boundary is vo(T) =
0. In general. we shall take the decision boundary to be
𝑠𝑜1 (𝑇) + 𝑠𝑜2 (𝑇) …4.9.1
𝑣𝑜 (𝑇) =
2
The probability of error for this general case may be deduced as an extension of the considerations used in
the baseband case. Suppose that S01(T) > S02(T) and that S2(t) was transmitted. If, at the sampling time the
noise no(T) is positive and larger in magnitude than the voltage difference (1/2)[ S01(T)+ S02(T)] - S02(T), an
error will have' been made. That is, an error will result if

𝑠𝑜1 (𝑇) − 𝑠𝑜2 (𝑇) …4.9.2

𝑛𝑜 (𝑇 ) ≥
2
Hence the probability of error is
∞ 2 2
𝑒 −𝑛0 (𝑇)/2𝜎0
𝑃𝑒 = ∫ 𝑑𝑛𝑜 (𝑇) …4.9.3
[𝑠𝑜1 (𝑇)−𝑠𝑜2 (𝑇)]/2 √2𝜋𝜎02
𝑛𝑜 (𝑇)
If we make the substitution 𝑥 ≡ , then above equation becomes,
√2𝜎0

1 2 ∞ 2
𝑃𝑒 = ∫ 𝑒 −𝑥 𝑑𝑥
2 √𝜋 [𝑠𝑜1 (𝑇)−𝑠𝑜2 (𝑇)]/2√2𝜎0 …4.9.4a

1 𝑠𝑜1 (𝑇) − 𝑠𝑜2 (𝑇)

𝑃𝑒 = 𝑒𝑟𝑓𝑐 [ ] …4.9.4b
2 2√2𝜎0

Note that for the case S01(T) = VT/τ and S02(T) = - VT/τ, and, using Eq. (4.7.4), Eq. (4.9.4b) reduces to Eq.
(4.8.3) as expected.
The complementary error function is a monotonically decreasing function of its argument. (See Fig. 4.8.2.)
Hence, as is to be anticipated, Pe decreases as the difference S01(T) - S02(T) becomes larger and as the rms
noise voltage σ0 becomes smaller. The optimum filter, then, is the filter which maximizes the ratio

𝑠𝑜1 (𝑇) − 𝑠𝑜2 (𝑇) …4.9.5

𝛾=
𝜎0
We now calculate the transfer function H(f) of this optimum filter.

Page 13 of 19

Page no: 13 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Optimum Filter Transfer Function H(f)

The fundamental requirement we make of a binary encoded data receiver is that it distinguishes the
voltages S1(t) + n(t) and S2(t) + n(t). We have seen that the ability of the receiver to do so depends on how
large a particular receiver can make γ. It is important to note that it is proportional not to S1(t) nor to S2(t),
but rather to the difference between them. For example, in the baseband system we represented the
signals by voltage levels + V and - V. But clearly, if our only interest was in distinguishing levels, we would
do just as well to use +2 volts and 0 volt, or +8 volts and +6 volts, etc, (The + V and - V levels, however,
have the advantage of requiring the least average power to be transmitted.) Hence, while S1(t) or S2(t) is
the received signal, the signal which is to be compared with the noise, i.e., the signal which is relevant in all
our error-probability calculations, is the difference signal

𝑝(𝑇) ≡ 𝑠1 (𝑇) − 𝑠2 (𝑇) …4.9.6

Thus, for the purpose of calculating the minimum error probability, we shall assume that the Input Signal
to the optimum filter is p(t). The corresponding output signal of the filter is then

𝑝0 (𝑇) ≡ 𝑠01 (𝑇) − 𝑠02 (𝑇) …4.9.7

Let P(f) and P0(f) ne the Fourier transform of p(f) and P0(f) respectively. If H(f) is the transfer function of the
filter
𝑃0 (𝑓) = 𝐻(𝑓)𝑃(𝑓) …4.9.8
∞ ∞
And
𝑝0 (𝑇) = ∫ 𝑃0 (𝑓) 𝑒 𝑗2𝜋𝑓𝑇 𝑑𝑓 = ∫ 𝐻 (𝑓)𝑃(𝑓) 𝑒 𝑗2𝜋𝑓𝑇 𝑑𝑓 …4.9.9
−∞ −∞
The input noise to the optimum filler is n(t). The output noise is n0(t) which has a power spectral density
Gn0(f) and is related to the power spectral density of the input noise Gn(f) by

𝐺𝑛0 (𝑓) = |𝐻(𝑓)|2 𝐺𝑛 (𝑓) …4.9.10

Using Parseval's theorem, we find that the normalized output noise power, i.e., the noise varianceσ02, is
∞ ∞
…4.9.11
𝜎02 = ∫ 𝐺𝑛0 (𝑓) 𝑑𝑓 = ∫ |𝐻(𝑓)|2 𝐺𝑛 (𝑓) 𝑑𝑓
−∞ −∞
From equation 9 and 11,
∞ 2
2
𝑝02 (𝑇) [∫−∞ 𝐻 (𝑓)𝑃(𝑓) 𝑒 𝑗2𝜋𝑓𝑇 𝑑𝑓]
𝛾 = = ∞ …4.9.12
𝜎02 ∫−∞|𝐻(𝑓)|2 𝐺𝑛 (𝑓) 𝑑𝑓
Equation 4.9.12 is unaltered by the inclusion or deletion of the absolute value sign in the numerator since
the quantity within the magnitude sign p0(T) is a positive real number. The sign has been included,
however, in order to allow further development of the equation through the use of the Schwarz inequality.
The Schwarz inequality state that given arbitrary complex functions X(f) and Y(f) of a common variable f,
then
∞ 2 ∞ ∞ …4.9.13
|∫ 𝑋(𝑓)𝑌(𝑓) 𝑑𝑓| ≤ ∫ |𝑋(𝑓)|2 𝑑𝑓 ∫ |𝑌(𝑓)|2 𝑑𝑓
−∞ −∞ −∞
The equal sign applies when
𝑋 (𝑓) = 𝐾𝑌 ∗ (𝑓) …4.9.14
where K is an arbitrary constant and Y*(f) is the complex conjugate of Y(f).
We now apply the Schwarz inequality to Eq. (4.9.12) by making the identification
𝑋(𝑓) ≡ √𝐺𝑛 (𝑓) 𝐻(𝑓) …4.9.15
and 1 …4.9.16
𝑌 (𝑓 ) ≡ 𝑃(𝑓)𝑒 𝑗2𝜋𝑇𝑓
√𝐺𝑛 (𝑓)
Using equation 15 and 16 and using Schwarz inequality equation 13, we may write equation 12 as,

Page 14 of 19

Page no: 14 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

∞ 2
∞
𝑝02 (𝑇) [∫−∞ 𝑋 (𝑓)𝑌(𝑓) 𝑑𝑓]
= ∞ ≤ ∫ |𝑌(𝑓)|2 𝑑𝑓 …4.9.17
𝜎02 | |
∫−∞ 𝐻(𝑓) 𝑑𝑓2
−∞
Using equation 16,
∞ ∞ [ ( )]2
𝑝02 (𝑇) 2
𝑃 𝑓
2 ≤∫
| |
𝑌(𝑓) 𝑑𝑓 = ∫ 𝑑𝑓 …4.9.18
𝜎0 −∞ −∞ 𝐺𝑛 (𝑓)
The ratio p02(T)/σ02; will attain its maximum value when the equal sign in Eq. (4.9.18) may be employed as
is the case when X(f) = K Y*(f). We then find from Eqs. (4.9.15) and (4.9.16) that the optimum filter which
yields such a maximum ratio p02(T)/σ02; has a transfer function
𝑃∗ (𝑓) −𝑗2𝜋𝑓𝑇
𝐻(𝑓) = 𝐾 𝑒 …4.9.19
𝐺𝑛 (𝑓)
Correspondingly, the maximum ratio is, from Eq. (4.9.18),

∞ [ ( )]2
𝑝02 (𝑇) 𝑃 𝑓
[ 2 ] = ∫ 𝑑𝑓 …4.9.20
𝜎0 𝑚𝑎𝑥 −∞ 𝐺𝑛 (𝑓)

4.10 White Noise: The Matched Filter

An optimum filter which yields a maximum ratio p02(T)/σ02is called a matched filler when the input noise is
white. In this case Gn(f) = η/2, and Eq. (4.9.19) becomes
𝑃∗ (𝑓) −𝑗2𝜋𝑓𝑇
𝐻(𝑓) = 𝐾 𝑒 …4.10.1
η/2
The impulsive response of this filter, i.e., the response of the filter to a unit strength impulse applied at t =
0, is
2𝐾 ∞ ∗
ℎ(𝑡) = 𝐹 −1 [𝐻(𝑓)] = ∫ 𝑃 (𝑓)𝑒 −𝑗2𝜋𝑓𝑇 𝑒 −𝑗2𝜋𝑓𝑡 𝑑𝑓 …4.10.2(a)
η −∞
2𝐾 ∞ ∗
= ∫ 𝑃 (𝑓)𝑒 𝑗2𝜋𝑓(𝑡−𝑇) 𝑑𝑓 …4.10.2(b)
η −∞

A physically realizable filter will have an impulse response which is real, i.e., not complex. Therefore h(t) =
h*(t). Replacing the right-hand member of Eq. (4.10.2b) by its complex conjugate, an operation which
leaves the equation unaltered, we have
2𝐾 ∞
( )
ℎ 𝑡 = ∫ 𝑃(𝑓)𝑒 𝑗2𝜋𝑓(𝑡−𝑇) 𝑑𝑓 …4.10.3(a)
η −∞
2𝐾
= 𝑝(𝑇 − 𝑡) …4.10.3(b)
η
Finally since p(t)=s1(t) – s2(t), we have
2𝐾
ℎ(𝑡) = [𝑠 (𝑇 − 𝑡) − 𝑠2 (𝑇 − 𝑡)] …4.10.4
η 1
As shown in Fig. 4.10.1a, the s1(t) is a triangular waveform of duration T, while s2(t), (Fig. 4.10.1b), is of
identical form except of reversed polarity. Then p(t) is as shown in Fig. 4.10.1c, and p(-t) appears in Fig.
4.10.1d. The waveform p(-t) is the waveform p(t) rotated around the axis t =0. Finally, the waveform p(T -
t) called for as the impulse response of the filter in Eq. (4.10.3b) is this rotated waveform p(-t) translated in
the positive t direction by amount T. This last translation ensures that h(t) = 0 for t < 0 as is required for a
causal filter.
In general, the impulsive response of the matched filter consists of p(t) rotated about t = 0 and then
delayed long enough (i.e., a time T) to make the filter realizable. We may note in passing, that any
additional delay that a filter might introduce would in no way interfere with the performance of the filter,
for both signal and noise would be delayed by the same amount, and at the sampling the ratio of signal to
noise would remain unaltered.

Page 15 of 19

Page no: 15 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Figure 4.10.1 The signals (a) s1(t), (b) s2(t), (c) p(t)=s1(t)- s2(t), (d) p(t) rotated about the axis t=0, (e) The
waveform of (d) translated to right by amount T.

4.11 Correlator
Coherent Detection: Correlation
Coherent detection is an alternative type of receiving system, which is identical in performance with the
matched filter receiver. Again, as shown in Fig. 4.11.1, the input is a binary data waveform S1(t) or S2(t)
corrupted by noise n(t). The bit length is T. The received signal plus noise vi(t) is multiplied by a locally
generated waveform S1(t) - S2(t). The output of the multiplier is passed through an integrator whose output
is sampled at t = T. As before, immediately after each sampling, at the beginning of each new bit interval,
all energy-storing elements in the integrator are discharged. This type of receiver is called a correlator,
since we are correlating the received signal and noise with the waveform S 1(t)- S2(t).

Figure 4.11.1 A Coherent System of Signal Reception

The output signal and noise of the correlator shown in Fig. 4.11.1 are
1 𝑇
𝑠0 (𝑡) = ∫ 𝑆𝑖 (𝑡)[𝑆1 (𝑡) − 𝑆2 (𝑡)] 𝑑𝑡 …4.11.1
τ 0
1 𝑇
𝑛0 (𝑡) = ∫ 𝑛(𝑡)[𝑆1 (𝑡) − 𝑆2 (𝑡)] 𝑑𝑡 …4.11.2
τ 0
where Si(t) is either S1(t) or S2(t), and where τ is the constant of the integrator (i.e., the integrator output is
l/τ times the integral of its input). We now compare these outputs with the matched filter outputs.
If h(t) is the impulsive response of the matched filter, then the output of the matched filter vo(t) can be
found using the convolution integral. We have

Page 16 of 19

Page no: 16 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

∞ 𝑇
𝑣0 (𝑡) = ∫ 𝑣𝑖 (𝜆)ℎ(𝑡 − 𝜆) 𝑑𝜆 = ∫ 𝑣𝑖 (𝜆)ℎ(𝑡 − 𝜆) 𝑑𝜆 …4.11.3
−∞ 0
The limits on the integral have been changed to 0 and T since we are interested in the filter response to a
bit which extends only over that interval. Using Eq. (4.10.4) which gives h(t) for the matched filter, we have
2𝐾
ℎ(𝑡) = [𝑠 (𝑇 − 𝑡) − 𝑠2 (𝑇 − 𝑡)] …4.11.4
η 1
2𝐾
So that ℎ(𝑡 − 𝜆) = [𝑠 (𝑇 − 𝑡 + 𝜆) − 𝑠2 (𝑇 − 𝑡 + 𝜆)] …4.11.5
η 1
Submitting equation 4.11.5, in equation 4.11.3
2𝐾 𝑇
𝑣0 (𝑡) = ∫ 𝑣 (𝜆)[𝑠1 (𝑇 − 𝑡 + 𝜆) − 𝑠2 (𝑇 − 𝑡 + 𝜆)] 𝑑𝜆 …4.11.6
η 0 𝑖
Since vi(λ) = si(λ)+n(λ), and v0(t) = s0(t)+n0(t), setting t=T yields,
2𝐾 𝑇
𝑠0 (𝑡) = ∫ 𝑠 (𝜆)[𝑠1 (𝜆) − 𝑠2 (𝜆)]𝑑𝜆 …4.11.7
η 0 𝑖
Where si(λ) is equal to s1(λ) or s2(λ). Simillarly,
2𝐾 𝑇
𝑛0 (𝑡 ) = ∫ 𝑛(𝜆)[𝑠1 (𝜆) − 𝑠2 (𝜆)]𝑑𝜆 …4.11.8
η 0
Thus as we can see from above equations so(T) and no(T), are identical. Hence the performances of the two
systems are identical.
The matched filter and the correlator are not simply two distinct, independent techniques which happen to
yield the same result. In fact they are two techniques of synthesizing the optimum filter h(t).

4.12 Probability of error calculation for BPSK and BFSK

(i) BPSK
The synchronous detector for BPSK is shown in figure 4.12.1(b). Since the BPSK signal is one dimensional,
The only relevant noise in the present case is
𝑛(𝑡) = 𝑛0 𝑢(𝑡) = 𝑛0 √2/𝑇𝑏 cos 𝜔0 𝑡 …4.12.1
2
where no is a Gaussian random variable of variance σ0 = η/2. Now let us suppose that S2 was transmitted.

Figure 4.12.1 (a) BPSK representation in signal space showing r1 and r2 (b)Correlator receiver for BPSK
showing that r=r1+n0 or r2+ n0

The error probability, i.e., the probability that the signal is mistakenly judged to be S1 is the probability that
𝑛0 > √𝑃𝑠 𝑇𝑏 . Thus the error probability Pe, is
∞ ∞
1 −𝑛02/2𝜎02
1 2
𝑃𝑒 = ∫ 𝑒 𝑑𝑛0 = ∫ 𝑒 −𝑛0 /𝜂 𝑑𝑛0 …4.12.2
2
√2𝜋𝜎 √𝑃𝑠𝑇𝑏 √𝜋𝜂 √𝑃𝑠 𝑇𝑏

Page 17 of 19

Page no: 17 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Let y2=n02/2 σ02, then

∞
1 2 1 2 ∞ 2 1
𝑃𝑒 = ∫ 𝑒 −𝑦 𝑑𝑦 = ∫ 𝑒 −𝑦 𝑑𝑦 = 𝑒𝑟𝑓𝑐 √𝑃𝑠 𝑇𝑏 /𝜂 …4.12.3
√𝜋 √𝑃𝑠 𝑇𝑏 /𝜂 2 √𝜋 √𝑃𝑠 𝑇𝑏/𝜂 2

The signal energy is Eb = PsTb and the distance between end points of the signal vectors in Fig. 4.12.1 is =
2√𝑃𝑠 𝑇𝑏 . Accordingly we find that
1 1
𝑃𝑒 = 𝑒𝑟𝑓𝑐 √𝐸𝑏 /𝜂 = 𝑒𝑟𝑓𝑐 √𝑑 2 /4𝜂 …4.12.4
2 2
The error probability is thus seen to fall off monotonically with an increase in distance between signals.

(ii) BFSK
The case of synchronous detection of orthogonal binary FSK is represented in Fig. 4.12.2. The signal space
is shown in (a). The unit vectors are
𝑢1 (𝑡) = √2/𝑇𝑏 cos 𝜔1 𝑡 …4.12.5a
and 𝑢2 (𝑡) = √2/𝑇𝑏 cos 𝜔2 𝑡 …4.12.5b

Figure 4.12.2 (a) Signal Space representation of BFSK (b) Correlator Receiver for BFSK

Orthogonality over the interval Tb having been insured by the selection of ω1 and ω2. The transmitted
signals s1 and s2 are of power Ps, and are
𝑠1 (𝑡) = √2𝑃𝑠 cos 𝜔1 𝑡 = √𝑃𝑠 𝑇𝑏 √2/𝑇𝑏 cos 𝜔1 𝑡 = √𝑃𝑠 𝑇𝑏 𝑢1 (𝑡) …4.12.6a
and 𝑠2 (𝑡) = √2𝑃𝑠 cos 𝜔2 𝑡 = √𝑃𝑠 𝑇𝑏 √2/𝑇𝑏 cos 𝜔2 𝑡 = √𝑃𝑠 𝑇𝑏 𝑢2 (𝑡) …4.12.6b

Detection is accomplished in the manner shown in Fig. 4.12.2 (b). The outputs are r1 and r2. In the absence
of noise when s1(t) is received, r2 = 0 and r1 = √𝑃𝑠 𝑇𝑏 . For S2(t), r1 = 0 and r2 =√𝑃𝑠 𝑇𝑏 . Hence the vectors
representing r1 and r2 are of length √𝑃𝑠 𝑇𝑏 as shown in Fig. 4.12.2(a).
Since the signal is two dimensional the relevant noise in the present case is
𝑛(𝑡) = 𝑛1 𝑢1 (𝑡) + 𝑛2 𝑢2 (𝑡) …4.12.7
2 2
In which n1 and n2 are Gaussian random variables each of variance σ1 = σ2 =η/2. Now let us suppose that
S2(t) is transmitted and that the observed voltages at the output of the processor are r’1 and r’2 as shown in
Fig. 4.12.2a. We find that r’2≠r2 because of the noise n2 and r’1≠0 because of the noise n1. We have drawn
the locus of points equidistant from r1 and r2 and suppose, that the received voltage r, is closer to r1 than
to r2. Then we shall have made an error in estimating which signal was transmitted. It is readily apparent
that such an error will occur whenever n1>r2-n2 or n1 + n2 > √𝑃𝑠 𝑇𝑏 . Since n1 and n2 are uncorrelated, the
random variable n0 = n1 + n2 has a variance σ02= σ12+ σ22=η and its probability density function is
1 2
𝑓 (𝑛0 ) = 𝑒 −𝑛0 /2𝜂 …4.12.8
√2𝜋𝑛

Page 18 of 19

Page no: 18 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

The probability of error is

∞
1 2
𝑃𝑒 = ∫ 𝑒 −𝑛0 /2𝜂 𝑑𝑛0 …4.12.9
√2𝜋𝜂 √𝑃𝑠 𝑇𝑏

Again we have Eb = PsTb and in the present case the distance between r1 and r2 is 𝑑 = √2√𝑃𝑠 𝑇𝑏 .
Accordingly, proceeding as in Eq. (4.12.2) we find that
1
𝑃𝑒 = 𝑒𝑟𝑓𝑐√𝐸𝑏 /2𝜂 …4.12.10a
2
1
= 𝑒𝑟𝑓𝑐√𝑑 2 /2𝜂 …4.12.10b
2
Comparing Eqs. (4.12.10b) and (4.12.4) we see that when expressed in terms of the distance d, the error
probabilities are the same for BPSK and BFSK.

----------X----------

Page 19 of 19

Page no: 19 Get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]
Program : B.Tech
Subject Name: Digital Communication
Subject Code: EC-502
Semester: 5th
Downloaded from www.rgpvnotes.in

Department of Electronics and Communication Engineering

Sub. Code: EC 5002 Sub. Name: Digital Communication

Unit 5
Syllabus:
Information Theory Source Coding: Introduction to information theory, uncertainty and information,
average mutual information and entropy, source coding theorem, Huffman coding, Shannon-Fano-Elias
coding, Channel Coding: Introduction, channel models, channel capacity, channel coding, information
capacity theorem, Shannon limit.

Encoding is the process of converting the data or a given sequence of characters, symbols, alphabets etc.,
into a specified format, for the secured transmission of data.
Decoding is the reverse process of encoding which is to extract the information from the converted
format.
Data Encoding
Encoding is the process of using various patterns of voltage or current levels to represent 1s and 0s of the
digital signals on the transmission link.
The common types of line encoding are Unipolar, Polar, Bipolar, and Manchester.
Encoding Techniques
The data encoding technique is divided into the following types, depending upon the type of data
conversion.
 Analog data to Analog signals − The modulation techniques such as Amplitude Modulation,
Frequency Modulation and Phase Modulation of analog signals, fall under this category.
 Analog data to Digital signals − This process can be termed as digitization, which is done by Pulse
Code Modulation (PCM). Hence, it is nothing but digital modulation. As we have already discussed,
sampling and quantization are the important factors in this. Delta Modulation gives a better output
than PCM.
 Digital data to Analog signals − The modulation techniques such as Amplitude Shift Keying (ASK),
Frequency Shift Keying (FSK), Phase Shift Keying (PSK), etc., fall under this category. These will be
discussed in subsequent chapters.
 Digital data to Digital signals − These are in this section. There are several ways to map digital data
to digital signals. Some of them are −
Information is the source of a communication system, whether it is analog or digital. Information theory is
a mathematical approach to the study of coding of information along with the quantification, storage, and
communication of information.

Conditions of Occurrence of Events

If we consider an event, there are three conditions of occurrence.
 If the event has not occurred, there is a condition of uncertainty.
 If the event has just occurred, there is a condition of surprise.
 If the event has occurred, a time back, there is a condition of having some information.
These three events occur at different times. The differences in these conditions help us gain knowledge on
the probabilities of the occurrence of events.

5.01 Information Theory:

Information theory provides a quantitative measure of the information contained in a message signal and
allow us to determine the capacity of a communication system to transfer this information from source to
destination.

Page 1 of 13

Page no: 1 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Communication
Source Destination
Channel
Figure 5.01 A Communication System
1. The information associated with any event depends upon the probability with which it exists.
2. Higher the probability of occurring, lower the information associated with it and vice versa.
3. If the probability of occurring is 1, the information associated with that event is zero since we are
certain that a particular bit actually exists at the input of the system.

5.02 Information Source:

An information source is an object that produces an event, the outcome of which is selected at random
according to a probability distribution. A discrete information source has only a finite set of symbols as
possible outputs.
A source with memory is one for which a current symbol depends on the previous one.
A memory less source is one for which each symbol produced is independent of the previous symbol.
Information: Anything to which some meaning or sense can be attached is called the information.
Example, a written message, a spoken word, a picture etc.

5.03 Unit of Information:

Consider a DMS(discrete memory less source), denoted by X, with alphabet {x1, x2, … xm}. The information
content of a symbol xi, denoted by I(xi), is defined by
1
𝐼(𝑥𝑖 ) = log b = − log b 𝑃(𝑥𝑖 ) …5.1.1
𝑃(𝑥𝑖 )
Where P(xi) is the probability of occurrence of symbol xi. The I(xi) satisfy the following conditions i.e.
properties
(1) 𝐼(𝑥𝑖 ) = 0 𝑓𝑜𝑟 𝑃(𝑥𝑖 ) = 1
(2) 𝐼(𝑥𝑖 ) ≥ 0
(3) 𝐼(𝑥𝑖 ) ≥ 𝐼(𝑥𝑗 ) if 𝑃(𝑥𝑖 ) ≥ 𝑃(𝑥𝑗 )
(4) 𝐼(𝑥𝑖 . 𝑥𝑗 ) = 𝐼(𝑥𝑖 ) + 𝐼(𝑥𝑗 ) if 𝑥𝑖 and 𝑥𝑗 are independent

The unit of I(xi) is bit if b=2, Hartley or decit if b=10 and nat (natural unit) if b=e.
𝐼(𝑥𝑖 ) = − log 2 𝑃(𝑥𝑖 ) bits
𝐼(𝑥𝑖 ) = − log10 𝑃(𝑥𝑖 ) hartley or decit
𝐼(𝑥𝑖 ) = − log e 𝑃(𝑥𝑖 ) nats
It is standard to use b=2.
The unit bit (b) is a measure of information content.

Conversion of Units
ln 𝑎 log 𝑎
log 2 𝑎 = =
ln 2 log 2
log10 6 1
log 2 6 = = log 6
log10 2 log10 2 10

5.03 Average Information or Entropy

In a practical Communication system, we usually transmit long sequences of symbol from an information
source. Hence it is more important to find the average information that a source produces than the
information content of a single symbol.
The mean value of I(xi) over the alphabet of source X, with in different symbol is given by-

Page 2 of 13

Page no: 2 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝐻(𝑥) = 𝐸[𝐼(𝑥𝑖 )] = ∑ 𝑃(𝑥𝑖 ). 𝐼(𝑥𝑖 )

𝑖=1
𝑚

𝐻(𝑥) = 𝐸[𝐼(𝑥𝑖 )] = − ∑ 𝑃(𝑥𝑖 ). log b 𝑃(𝑥𝑖 ) b/symbol

𝑖=1
The quantity H(x) is called the entropy of source X. It is a measure of the average information content per
source symbol. The source entropy H(x) can be considered as the average amount of uncertainty with in
the source X.
For a binary source X, that generates independent symbols 0 & 1, with equal probability, the source
entropy H(x) is
1 1 1 1
𝐻(𝑥) = − 𝑙𝑜𝑔2 ( ) − 𝑙𝑜𝑔2 ( )
2 2 2 2
1 1
𝐻(𝑥) = 𝑙𝑜𝑔2 2 + 𝑙𝑜𝑔2 2 1 b/symbol
2 2

The source entropy H(x) satisfy the following relation

0 ≤ 𝐻(𝑥) ≤ 𝑙𝑜𝑔2 𝑚
Where m is the number of symbols (also called size of alphabet of source (X)
Case 1 If in input only one digit is occurring i.e. m=1; P(i)=1
𝐻(𝑥) = 0
Case 2 If all m digits are equal probable i.e. P(i) = 1/m
𝑚 𝑚
1 1
𝐻(𝑥) = − ∑ 𝑃(𝑖). log 2 𝑃(𝑖) = − ∑ . log 2
𝑚 𝑚
𝑖=1 𝑖=1
1
= −log 2 = log 2 𝑚
𝑚
This is the maximum value, therefore
𝐻(𝑥)|𝑚𝑎𝑥 = log 2 𝑚
0 ≤ 𝐻(𝑥) ≤ 𝑙𝑜𝑔2 𝑚
Entropy
When we observe the possibilities of the occurrence of an event, how surprising or uncertain it would be,
it means that we are trying to have an idea on the average content of the information from the source of
the event.
Entropy can be defined as a measure of the average information content per source symbol. Claude
Shannon, the “father of the Information Theory”, provided a formula for it as –
𝐻 = − ∑ 𝑝𝑖 log𝑏 𝑝𝑖
𝑖
Where pi is the probability of the occurrence of character number i from a given stream of characters and
b is the base of the algorithm used. Hence, this is also called as Shannon’s Entropy.
The amount of uncertainty remaining about the channel input after observing the channel output, is called
as Conditional Entropy. It is denoted by 𝐻(𝑋 ∣ 𝑌).

5.04 Mutual Information

Let us consider a channel whose output is Y and input is X
Let the entropy for prior uncertainty be X = H(x)
(This is assumed before the input is applied)
To know about the uncertainty of the output, after the input is applied, let us consider Conditional
Entropy, given that Y = yk
𝑗−1
1
𝐻(𝑥 ∣ 𝑦𝑘 ) = ∑ 𝑝(𝑥𝑗 ∣ 𝑦𝑘 ) 𝑙𝑜𝑔2 [ ]
𝑝(𝑥𝑗 ∣ 𝑦𝑘 )
𝑗=0

Page 3 of 13

Page no: 3 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

𝐻(𝑋 ∣ 𝑦 = 𝑦0 ) … … … … … … … . 𝐻(𝑋 ∣ 𝑦 = 𝑦𝑘 )

This is a random variable for p(y0 ) … … … … … … … . p(yk−1 ) with probabilities respectively. The mean
value of H(X ∕ y = yk ) for output alphabet y is –
𝑘−1

𝐻(𝑋 ∣ 𝑌) = ∑ 𝐻(𝑋 ∣ 𝑦 = 𝑦𝑘 ) 𝑝(𝑦𝑘 )

𝑘=0

𝑘−1 𝑗−1
1
= ∑ ∑ 𝑝(𝑥𝑗 ∣ 𝑦𝑘 ) 𝑝(𝑦𝑘 )𝑙𝑜𝑔2[ ]
𝑝(𝑥𝑗 ∣ 𝑦𝑘 )
𝑘=0 𝑗=0

𝑘−1 𝑗−1
1
= ∑ ∑ 𝑝(𝑥𝑗 , 𝑦𝑘 ) 𝑙𝑜𝑔2 [ ]
𝑝(𝑥𝑗 ∣ 𝑦𝑘 )
𝑘=0 𝑗=0

Now, considering both the uncertainty conditions (before and after applying the inputs), we come to know
that the difference, i.e.
𝐻(𝑥) − 𝐻(x ∣ y)

must represent the uncertainty about the channel input that is resolved by observing the channel output.
This is called as the Mutual Information of the channel.
Denoting the Mutual Information as I(x;y), we can write the whole thing in an equation, as follows

𝐼(𝑥; 𝑦) = 𝐻(𝑥) − 𝐻(x ∣ y)

Hence, this is the equational representation of Mutual Information.

Properties of Mutual information

These are the properties of Mutual information.
1. Mutual information of a channel is symmetric.

𝐼(𝑥; 𝑦) = 𝐼(𝑦; 𝑥)
2. Mutual information is non-negative.
𝐼(𝑥; 𝑦) ≥ 0

3. Mutual information can be expressed in terms of entropy of the channel output.

𝐼(𝑥; 𝑦) = 𝐻(𝑦) − 𝐻(𝑦 ∣ 𝑥)

Where 𝐻( 𝑦 ∣ 𝑥 )is a conditional entropy.
4. Mutual information of a channel is related to the joint entropy of the channel input and the channel
output.
𝐼(𝑥; 𝑦) = 𝐻(𝑥) + 𝐻(𝑦) − 𝐻(𝑥, 𝑦)

Where the joint entropy 𝐻(𝑥, 𝑦)is defined by

j−1 k−1
1
𝐻(𝑥, 𝑦) = ∑ ∑ p(xj , yk ) log 2 [ ]
p(xj , yk )
j=0 k=0

Page 4 of 13

Page no: 4 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

5.05 Conditional Entropy

Using the input probabilities P(xi), output probabilities P(yi), transition probabilities P(xi/ yi) and joint
probabilities P(xi, yi), we can define the following various entropy functions for a channel with m input and
n outputs,
𝑚

Source Entropy 𝐻(𝑋) = − ∑ 𝑃(𝑥𝑖 ) log 2 𝑃(𝑥𝑖 )

𝑖=1
𝑛
Destination
𝐻(𝑌) = − ∑ 𝑃(𝑦𝑗 ) log 2 𝑃(𝑦𝑗 )
Entropy
𝑗=1

The conditional entropy H(X/Y) is a measure of the average uncertainty about the channel input after the
channel output has been observed.
𝑛 𝑚

𝐻(𝑋/𝑌) = − ∑ ∑ 𝑃(𝑥𝑖 , 𝑦𝑗 ) log 2 𝑃(𝑥𝑖 /𝑦𝑗 )

𝑗=1 𝑖=1

The conditional entropy H(Y/X) is the average uncertainty of the channel output given that X was
transmitted.
𝑛 𝑚

𝐻(𝑌/𝑋) = − ∑ ∑ 𝑃(𝑥𝑖 , 𝑦𝑗 ) log 2 𝑃(𝑦𝑗 /𝑥𝑖 )

𝑗=1 𝑖=1

The joint entropy H(X,Y) is the average uncertainty of the communication channel as a whole
𝑛 𝑚

𝐻(𝑋, 𝑌) = − ∑ ∑ 𝑃(𝑥𝑖 , 𝑦𝑗 ) log 2 𝑃(𝑥𝑖 , 𝑦𝑗 )

𝑗=1 𝑖=1

5.06 Efficiency
The transmission efficiency or the channel efficiency is defined as
𝑎𝑐𝑡𝑢𝑎𝑙 𝑡𝑟𝑎𝑛𝑠𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛
𝜂=
𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑡𝑟𝑎𝑛𝑠𝑖𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛
𝐼(𝑋; 𝑌) 𝐼(𝑋; 𝑌)
𝜂= =
max 𝐼(𝑋; 𝑌) 𝐶𝑠
Case I If I(X;Y) = max I(X;Y)
Then η = 100%
The channel is fully utilized.

Case II If I(X;Y) < max I(X;Y)

Then η < 100%

Case III If I(X;Y) > max I(X;Y)

Then η > 100%
The situation is avoided.
In the second case, to increase η, we code the input data using
1. Shannon Fano Coding
2. Huffman Coding Procedure

5.07 Discrete Memory less Source

A source from which the data is being emitted at successive intervals, which is independent of previous
values, can be termed as discrete memory less source.

Page 5 of 13

Page no: 5 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

This source is discrete as it is not considered for a continuous time interval, but at discrete time intervals.
This source is memory less as it is fresh at each instant of time, without considering the previous values.
The Code produced by a discrete memory less source, has to be efficiently represented, which is an
important problem in communications. For this to happen, there are code words, which represent these
source codes.

5.08 Source Coding

A conversion of the output of a DMS into a sequence of binary symbols is called source coding. The device
that performs this conversion is called the source encoder.

Let us take a look at the block diagram.

Source Channel Receiver

Figure 5.08.01 (a) Without Source Encoder

Source Encoder Channel Decoder Receiver

Figure 5.08.01 (b) With Encoder- Decoder

An objective of the source coding is to minimize the average bit rate required for representation of the
source by reducing the redundancy (i.e. increasing the efficiency) of the information source.

Code Length
Let X be a DMS with finite entropy H(X) and an alphabet {x1, x2, …, xm} with the corresponding probabilities
of occurrence P(xi) {i=1,2,…,m}.
Let the binary code word assigned to symbol xi have ni length, measured in bits. Then the average code
word length L per source symbol is given by:
𝑚

L = ∑ 𝑃(𝑥𝑖 ) 𝑛𝑖
𝑖=1
L represent the average number of bits per source symbol.
The code efficiency η can be defined as
𝐿𝑚𝑖𝑛
𝜂=
L
Where Lmin is the minimum possible value of L.
Code redundancy 𝜸
𝛾 =1−𝜂

5.09 Source Coding Theorem

It states that for a DMS-X, with entropy H(X), the average code word length L per symbol is bounded as
𝐿 ≥ 𝐻(𝑋)
And further L can be made as close to H(X) as desired for some suitable chosen code.
Thus with
𝐿𝑚𝑖𝑛 = 𝐻(𝑋)
𝐻(𝑋)
𝜂=
𝐿

Page 6 of 13

Page no: 6 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Example 5.01 A source delivers six digits with the following probabilities:

i A B C D E F
P(i) 1/2 1/4 1/8 1/16 1/32 1/32

Find (1) H(X) (2) H’(X) = r.H(X) (3) H’(X)max=C (4) Efficiency
Solution:
(1)
𝑚

𝐻(𝑋) = − ∑ 𝑃(𝑥𝑖 ). log 2 𝑃(𝑥𝑖 ) Bits/ Symbol

𝑖=1
6

𝐻(𝑋) = − ∑ 𝑃(𝑖). log 2 𝑃(𝑖)

𝑖=1
1 1 1 1 1 1 1 1 1 1 1 1
= − [ 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 ]
2 2 4 4 8 8 16 16 32 32 32 32
1 1 1 1 2
= − [ × (−1) + × (−2) + × (−3) + × (−4) + × (−5)]
2 4 8 16 32
= 1.938 𝑏𝑖𝑡𝑠/𝑠𝑦𝑚𝑏𝑜𝑙
(2)
𝐻′(𝑋) = 𝑟. 𝐻(𝑋)
6

𝐻(𝑥) = − ∑ 𝑃(𝑖). log 2 𝑃(𝑖)

𝑖=1
𝐻′(𝑋) = 1.938 𝑏𝑖𝑡𝑠/𝑠𝑒𝑐
(3)
log10 6
𝐻(𝑋)|𝑚𝑎𝑥 = log 2 𝑚 = log 2 6 = = 2.58 𝑏𝑖𝑡𝑠/𝑠𝑦𝑚𝑏𝑜𝑙
log10 2
𝐻′(𝑋)|𝑚𝑎𝑥 = 2.58 𝑏𝑖𝑡𝑠/𝑠𝑦𝑚𝑏𝑜𝑙 r=1 symbol/sec
(4)

𝐻′(𝑋) 1.938
𝜂= × 100% = × 100%
𝐻′(𝑋)|𝑚𝑎𝑥 2.58
𝜂 = 75.11%

5.10 Shannon-Fano Coding

An efficient code can be obtained by the following simple procedure, known as Shannon-Fano coding:
1. First write the source symbols in order of decreasing probability,
2. Partition the set into two most equi-probable sub sets and assign a ‘0’ to the upper set and ‘1’ to
the lower one,
3. Continue this procedure, each time partitioning the sets with as nearly as equal probabilities as
possible until further partitioning is not possible.

Example 5.02 Apply the Shannon Fano coding procedure for the following message ensemble:

X x1 x2 x3 x4 x5 x6 x7 x8
P 1/4 1/8 1/16 1/16 1/16 1/4 1/16 1/8
Take M=2. Find the Code Efficiency.
Solution:
As per the procedure explained in section 5.10, The code can be obtained as under,

Page 7 of 13

Page no: 7 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Message Prob. Step 1 Step 2 Step 3 Step 4 Code Code Length

x1 0.25 0 0 00 2
x6 0.25 0 1 01 2
x2 0.125 1 0 0 100 3
x8 0.125 1 0 1 101 3
x3 0.0625 1 1 0 0 1100 4
x4 0.0625 1 1 0 1 1101 4
x5 0.0625 1 1 1 0 1110 4
x7 0.0625 1 1 1 1 1111 4

𝐻(𝑋) = − ∑ 𝑃(𝑖). log 2 𝑃(𝑖)

𝑖=1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
= − [ 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 + 𝑙𝑜𝑔2 ]
4 4 8 8 16 16 16 16 16 16 4 4 16 16 8 8
1 1 1 1 1 1
= − [ × (−2) + × (−3) + × (−4) × 3 + × (−2) + × (−4) + × (−3)]
4 8 16 4 16 8
1 3 1 1 1 1 1 3
= + + + + + + +
2 8 4 4 4 2 4 8
= 2.75 𝑏𝑖𝑡𝑠/𝑠𝑦𝑚𝑏𝑜𝑙

Now the average code word length L per source symbol is

𝑚

𝐿 = ∑ 𝑃(𝑥𝑖 ). 𝑛𝑖
𝑖=1
1 1 1 1 1
= ×2+ ×2+ ×3+ ×3+ ×4×4
4 4 8 8 16
= 2.75 𝑏𝑖𝑡𝑠/𝑠𝑦𝑚𝑏𝑜𝑙

Then Efficiency

𝐻(𝑋) 2.75
𝜂= = 100% = 100% (𝑨𝒏𝒔𝒘𝒆𝒓)
𝐿 2.75

Example 5.03 A DMS has seven messages with probabilities

X x1 x2 x3 x4 x5 x6 x7
P 0.4 0.2 0.12 0.08 0.08 0.08 0.04
Apply the Shannon Fano coding procedure and calculate the efficiency of the code. Take M=2
Solution:
Message Prob. Step 1 Step 2 Step 3 Step 4 Code Code Length
X1 0.4 0 0 0 1
X2 0.2 1 0 0 100 3
X3 0.12 1 0 1 101 3
X4 0.08 1 1 0 0 1100 4
X5 0.08 1 1 0 1 1101 4
X6 0.08 1 1 1 0 1110 4
X7 0.04 1 1 1 1 1111 4

Page 8 of 13

Page no: 8 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Then Entropy is
𝑚

𝐻(𝑋) = − ∑ 𝑃(𝑥𝑖 ). log 2 𝑃(𝑥𝑖 )

𝑖=1
= −[(0.4)𝑙𝑜𝑔2 (0.4) + (0.2)𝑙𝑜𝑔2 (0.2) + (0.12)𝑙𝑜𝑔2 (0.12) + (0.08)𝑙𝑜𝑔2 (0.08) + (0.08)𝑙𝑜𝑔2 (0.08)
+ (0.08)𝑙𝑜𝑔2 (0.08) + (0.04)𝑙𝑜𝑔2 (0.04)]
= 2.42 𝑙𝑒𝑡𝑡𝑒𝑟𝑠/𝑠𝑦𝑚𝑏𝑜𝑙

Now the average code word length L per source symbol is

𝑚

𝐿 = ∑ 𝑃(𝑥𝑖 ). 𝑛𝑖
𝑖=1
= 0.4 × 1 + 0.2 × 3 + 0.12 × 3 + 0.08 × 4 + 0.08 × 4 + 0.08 × 4 + 0.04 × 4
= 2.48 𝑏𝑖𝑡𝑠/𝑚𝑒𝑠𝑠𝑎𝑔𝑒

Then Efficiency

𝐻(𝑋) 2.42
𝜂= × 100% = 100% = 97.58% (𝑨𝒏𝒔𝒘𝒆𝒓)
𝐿𝑎𝑣𝑔 2.48

5.10 Huffman Encoding

Huffman Encoding results in a code that has the highest efficiency. The Huffman Coding procedure is as
under:
1. List the source symbol in decreasing probability,
2. Combine the probabilities of the two symbols having the least probabilities and reorder the
resultant probabilities. This step is called reduction. The same procedure is repeated until there are
two ordered probabilities remaining,
3. Start reduction with the last reduction, which consists of exactly two ordered probabilities. Assign
‘0’ to the first probabilities and a ‘1’ to the second probability.
4. Now assign ‘0’ and ‘1’ for the probabilities that were combined in the previous reduction step, until
the first reduction step.

Example 5.04 A DMS has seven messages with probabilities

X x1 x2 x3 x4 x5 x6 x7
P 0.4 0.2 0.12 0.08 0.08 0.08 0.04

Apply Huffman coding procedure and calculate the efficiency of the code.
Solution:
Applying the Huffman’s Coding procedure and determining the codes as under

Page 9 of 13

Page no: 9 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Message Probability 1st Reduction 2nd Reduction 3rd Reduction 4th Reduction 5th Reduction Code Code Length
X1 0.4 0.4 0.4 0.6 0 1 1
0.4 0.4
0.36 0 0.4 1
0.24 0.24 1
X2 0.2 0.2 000 3
0.2 0.2 0
0.16 0.16 1
X3 0.12 0.12 0.12 0 010 3
0.12 0.12 1

X4 0.08 0.08 0010 4

0
1 4
X5 0.08 0.08 0011

X6 0.08 0110 4
0
1
X7 0.04 0111 4

Now, the Entropy

𝑚

𝐻(𝑋) = − ∑ 𝑃(𝑥𝑖 ). log 2 𝑃(𝑥𝑖 )

𝑖=1
= −[(0.4)𝑙𝑜𝑔2 (0.4) + (0.2)𝑙𝑜𝑔2 (0.2) + (0.12)𝑙𝑜𝑔2 (0.12) + (0.08)𝑙𝑜𝑔2 (0.08) + (0.08)𝑙𝑜𝑔2 (0.08)
+ (0.08)𝑙𝑜𝑔2 (0.08) + (0.04)𝑙𝑜𝑔2 (0.04)]
= 2.42 𝑙𝑒𝑡𝑡𝑒𝑟𝑠/𝑠𝑦𝑚𝑏𝑜𝑙

Now the average code word length L per source symbol is

𝑚

𝐿 = ∑ 𝑃(𝑥𝑖 ). 𝑛𝑖
𝑖=1
= 0.4 × 1 + 0.2 × 3 + 0.12 × 3 + 0.08 × 4 + 0.08 × 4 + 0.08 × 4 + 0.04 × 4
= 2.48 𝑏𝑖𝑡𝑠/𝑚𝑒𝑠𝑠𝑎𝑔𝑒

Then Efficiency

𝐻(𝑋) 2.42
𝜂= × 100% = 100% = 97.58% (𝑨𝒏𝒔𝒘𝒆𝒓)
𝐿𝑎𝑣𝑔 2.48

5.11 Shannon’s Theorem

Given a source of equally likely messages, with M>>1, which is generating information at a rate R. Given a
channel with channel capacity C. Then if R≤C, there exists a coding technique such that the output of the
source may be transmitted over the channel with the probability of error of receiving the message signal
very small.

Thus according to the theorem if R≤C, then the noise free transmission is possible in the presence of noise.
The negative theorem states that if the information rate R exceeds C (R>C), error probability approaches to
unity as M increases.

Page 10 of 13

Page no: 10 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

5.12 Channel Capacity

We have so far discussed mutual information. The maximum average mutual information, in an instant of a
signaling interval, when transmitted by a discrete memory less channel, the probabilities of the rate of
maximum reliable transmission of data, can be understood as the channel capacity.
It is denoted by C and is measured in bits per channel use.

5.13 Shannon Limit

Shannon-Hartley equation relates the maximum capacity (transmission bit rate) that can be achieved over
a given channel with certain noise characteristics and bandwidth. For an AWGN the maximum capacity is
given by
𝑆
𝐶 = 𝐵 log 2 (1 + )
𝑁
𝑆
𝐶 = 𝐵 log 2 (1 + )
𝑛𝛽

Here C is the maximum capacity of the channel in bits/second otherwise called Shannon’s capacity limit for
the given channel, B is the bandwidth of the channel in Hertz, S is the signal power in Watts and N is the
noise power, also in Watts. The ratio S/N is called Signal to Noise Ratio (SNR). It can be ascertained that the
maximum rate at which we can transmit the information without any error, is limited by the bandwidth,
the signal level, and the noise level. It tells how many bits can be transmitted per second without errors
over a channel of bandwidth B Hz, when the signal power is limited to S Watt and is exposed to Gaussian
White Noise of additive nature.

Example 5.05 Calculate the channel capacity of a channel with BW 3kHz & S/N ratio given as 103,
assuming that the noise is white Gaussian noise.
Solution:
We know that
𝑆
𝐶 = 𝐵 log 2 (1 + )
𝑁
= 3000 log 2 (1 + 103 )
= 30000 𝑏𝑖𝑡𝑠/𝑠𝑒𝑐

5.14 Binary Symmetric Channel (BSC)

A symmetric channel is defined as the one for which
(i) H(Y/xj) is independent of j: the entropy corresponding to each row of P(Y/X) is the same. &
(ii) ∑𝑚
𝑗=1 𝑃(𝑦𝑘 /𝑥𝑗 ) is independent of k i.e. the sum of all columns of P(Y/X) is the same.
For a symmetric channel
𝐼(𝑋: 𝑌) = 𝐻(𝑌) − 𝐻(𝑌/𝑋 )
𝑚
𝑌
= 𝐻(𝑌) − 𝐴 ∑ 𝐻 ( ) 𝑃(𝑥𝑗 )
𝑥𝑗
𝑗=1
𝑚

= 𝐻(𝑌) − 𝐴 ∑ 𝑃(𝑥𝑗 )
𝑗=1
Where A=H(Y/xj) is independent of j and hence taken out of the summation sign. Also
𝑚

∑ 𝑃(𝑥𝑗 ) = 1
𝑗=1
therefore 𝐼(𝑋: 𝑌) = 𝐻(𝑌) − 𝐴
Hence the channel capacity
𝐶 = max 𝐼(𝑥: 𝑦)
= max[𝐻(𝑌) − 𝐴]

Page 11 of 13

Page no: 11 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

= max[𝐻(𝑌)] − 𝐴
𝐶 = log 𝑛 − 𝐴
Where n is the total number of receiver symbols, and max[H(Y)] = logn.

Binary Symmetric Channel;

The most important case of a symmetric channel is BSC. In this case m=n=2, and the channel matrix is
𝑌 𝑃 1−𝑃 𝑃𝑞
𝐷 = [𝑃 ( )] = [ ]=[ ]
𝑋 1−𝑃 𝑃 𝑞𝑃
BSC is shown graphically as shown
The channel is symmetric because the P
𝑋1 (0) 𝑌1 (0)
probability of receiving a 1, if 0 was
transmitted is same as the probability of 1-P
receiving a 0 if a 1 was transmitted. This
common transition probability is given by
𝑋2 (1) 1-P 𝑌2 (1)
P.
P
Figure 5.14.1 BSC

Example 5.06 For the Binary Symmetric Channel P

𝑋1 (0) 𝑌1 (0)
Calculate the channel capacity for
(i) P=0.9 and (ii) P= 0.6 1-P

𝑋2 (1) 1-P 𝑌2 (1)

P
Solution:
We know that
𝐶 = log 𝑛 − 𝐴
𝑌
𝐶 = log 2 − 𝐻 ( )
𝑥𝑗
2
𝑦𝑘 𝑦𝑘
𝐶 = log 2 2 − [− ∑ 𝑃 ( ) log 2 𝑃 ( )]
𝑥𝑗 𝑥𝑗
𝑗=1
𝐶 = log 2 + 𝑃 log 𝑃 + (1 − 𝑃) log(1 − 𝑃)
𝐶 = log 2 + (𝑃 log 𝑃 + 𝑞 log 𝑞)
𝐶 = 1 + (𝑃 log 𝑃 + 𝑞 log 𝑞)
𝐶 = 1 + 𝐻(𝑃) = 1 − 𝐻(𝑞)
(i) for p=0.9
𝐶 = 1 + (0.9 log 0.9 + 0.1 log 0.1)
𝐶 = 0.531 bits / message
(ii) for p=0.6
𝐶 = 1 + (0.6 log 0.6 + 0.4 log 0.4)
𝐶 = 0.029 bits / message
Example 5.07 Find the entropy of the source, information rate, average length and efficiency if the rate
of message generation is 300message per seconds and if the symbols and the probabilities are as under

X x1 x2 x3 x4 x5 x6 x7 x8 x9 x10
P 0.1 0.13 0.01 0.04 0.08 0.29 0.06 0.22 0.05 0.02

Page 12 of 13

Page no: 12 Get real-time updates from RGPV

Downloaded from www.rgpvnotes.in

Example 5.08 For the Channel shown in figure, 0.8

𝑋1 (0) 𝑌1 (0)
Calculate the channel capacity.
0.3

𝑋2 (1) 0.2 𝑌2 (1)

0.7
Solution:
The channel matrix is given by
0.8 0.2
[𝑃(𝑌/𝑋)] = [ ]
0.3 0.7
Since this is an un symmetric channel therefore the channel capacity of a binary channel is
𝐶 = log(2𝑄1 + 2𝑄2 )
Where Q1 and Q2 are defined by [P].[Q]=[H], therefore
0.8 0.2 𝑄1 𝑃11 log 𝑃11 + 𝑃12 log 𝑃12
[ ][ ] = [ ]
0.3 0.7 𝑄2 𝑃21 log 𝑃21 + 𝑃22 log 𝑃22
0.8 log 0.8 + 0.2 log 0.2
=[ ]
0.3 log 0.3 + 0.7 log 0.7
−0.2576 − 0.4644 −0.722
=[ ]=[ ]
−0.5211 − 0.3602 −0.8813
Therefore 0.8𝑄1 + 0.2𝑄2 = −0.722
and 0.3𝑄1 + 0.7𝑄2 = −0.8813
After solving 𝑄1 = −0.6568 and 𝑄2 = −0.9764
Then 𝐶 = log(2𝑄1 + 2𝑄2 ) = log(2−0.6558 + 2−0.9764 )
𝐶 = log(0.6343 + 0.5082) = log 2 (1.14255) = 0.192 𝑏𝑖𝑡/𝑚𝑒𝑠𝑠𝑎𝑔𝑒 𝑨𝒏𝒔𝒘𝒆𝒓

Example 5.09 For the Channel shown in figure, 0.7

𝑋1 (0) 𝑌1 (0)
Calculate the channel capacity.
0.4

𝑋2 (1) 0.3 𝑌2 (1)

0.8
Solution
After solving 𝑄1 = −1.061 and 𝑄2 = −0.456
Then 𝐶 = log(2−1.061 + 2−0.456 )
𝐶 = log(0.4793 + 0.729) = 0.273 𝑏𝑖𝑡/𝑠𝑦𝑚𝑏𝑜𝑙 𝑨𝒏𝒔𝒘𝒆𝒓

Example 5.10 For the Channel shown in figure, 0.7

𝑋1 (0) 𝑌1 (0)
Calculate the channel capacity.
0.4

𝑋2 (1) 0.3 𝑌2 (1)

0.6
Solution
After solving 𝑄1 = −0.79 and 𝑄2 = −1.09
𝐶 = 0.067 𝑏𝑖𝑡/𝑠𝑦𝑚𝑏𝑜𝑙 𝑨𝒏𝒔𝒘𝒆𝒓

Page 13 of 13

Page no: 13 Get real-time updates from RGPV

 We hope you find these notes useful.
You can get previous year question papers at
https://qp.rgpvnotes.in .

If you have any queries or you want to submit your

study notes please write us at
[email protected]