Course Name : ECT 305

Analog and
Digital Communication
(Credit – 4)

Prepared by
Riyaz R
 Course Outcomes
After the completion of the course the student will be
able to
• CO 1: Explain the existent analog communication
systems.
• CO 2: Apply the concepts of random processes to LTI
systems.
• CO 3: Apply waveform coding techniques in digital
transmission.
• CO 4: Apply GS procedure to develop digital
receivers.
• CO 5: Apply equalizer design to counteract ISI.
• CO 6: Apply digital modulation techniques in signal
transmission.
 Module 4
G-S Procedure and Effects in the channel
• Gram-Schmitt procedure.Signal space.
• Baseband transmission through AWGN channel.
• Mathematical model of ISI.
• Nyquist criterion for zero ISI. Signal modeling for ISI
• Raised cosine and Square-root raised cosine spectrum,
• Partial response signaling and duo binary coding.
• Equalization. Design of zero forcing equalizer.
• Vector model of AWGN channel. Matched filter and
correlation receivers.
• MAP receiver, Maximum likelihood receiver and probability
of error.
• Capacity of an AWGN channel (Expression only) --
significance in the design of communication schemes.
 Signal Space Analysis
• In baseband transmission, no carrier is
involved while in bandpass transmission, a
sinusoidal carrier signal is used for
transmitting digital data over a channel

• For the performance comparison of bandpass

transmission system, knowledge of signal
space is highly essential
• Consider a source emitting M distinct symbols
represented by m1,m2…mM with probabilities
p1,p2,…pM

• The transmitter takes the message source output

mi and maps it into a signal si(t) suitable for
transmission over the channel

• The signal si(t) is assumed to be a real valued

energy signal of duration T seconds
• While transmitting through the channel, channel noise w(t)
(which is considered to be Additive White Gaussian
Noise,AWGN) is added

• Accordingly the received signal x(t) is

• The performance evaluation of the receiver systems

become easier by representing each signal si(t) in
signal space
 Geometric Representation of Signals
• Objective: To represent ‘M’ energy signals
s1(t),s2(t)..sM(t) as a linear combination of ‘N’
orthonormal basis function

Synthesis
Eqn

Where the co-efficients of the expansion are defined by

Analysis
Eqn
• The basis functions Φ1(t),Φ2(t)…ΦN(t) are
called orthonormal basis functions since it
satisfies two conditions

• The signal vector si(t) can be represented as

Synthesizer Analyzer
Norm, Energy and Euclidean distance
of a Signal vector
• Norm: The length of the signal vector ‘si‘ denoted by the
symbol ‫׀׀‬si ‫ ׀׀‬is called the norm
• The squared length of the signal vector si is the inner product
of si with itself
• Energy: Energy of the signal si(t) is equal to
the squared length of the signal vector si
E=
Proof:
• Euclidean distance: The distance between two
signal vectors is called the Euclidean distance
and is defined as
 GRAM-SCHMIDT
ORTHOGONALIZATION PROCEDURE
 GS Orthogonalization Procedure
• The Gram-Schmidt orthogonalization procedure is
a straightforward way by which an appropriate set
of orthonormal functions can be obtained from
any given signal set

• The benefit of orthogonalizing the signal

waveforms is that they don’t overlap with each
other and are easy to identify at the demodulation
side
 GS Procedure
• Consider ‘M’ signals s1,s2…sM which are to be represented in
an N dimensional space

• GS Orthogonalization procedure can be used to find the

complete orthonormal basis function required to represent
‘M’ signals in an N dimensional space
GS Procedure
GS Procedure
• Step 1: Start by choosing an arbitarary signal s1(t).
The first basis function is defined as

where E1 is the energy of signal s1

where the co-efficients s11= 𝐸1 and Φ1(t) has unit energy

• Step 2: Choose another signal s2(t). Find an intermediate
function g2(t) such that g2(t) is orthogonal to Φ1(t)
ie, where s21 is the projection of s2(t)
on Φ1(t)
• Step 3: Φ3(t) can be found by choosing the
signal s3(t)
Intermediate function g3(t)
 • Continuing like this, in general

where

Now the basis function Φi(t)

forms an orthonormal set

• The dimension N is less than or equal to the number of given

signal M depending upon whether the set of signals form a
linearly independent set(N=M) or not linearly independent(N<M)
b)
c)
Example: Apply Gram Schmidt orthogonalisation to obtain
orthonormal basis functions for the signals shown below. Express
the signals in terms of orthonormal basis functions.
The signals are not linearly independent since
s1(t)=s2(t)+s3(t)
Example
Example
Example

Apply Gram Schmidt orthogonalisation to obtain orthonormal

basis functions for the signals shown below. Express the signals
in terms of orthonormal basis functions.
MATCHED FILTER
Detection of Transmitted Pulse Over a
Baseband Channel
• A major problem in the study of communication
system is the detection of pulse transmitted over
a channel corrupted by channel noise

• Simple solution is to sample the signal at a

particular time instant in a bit duration Tb and
compare it with a threshold

• When the effect of noise is more, this is not the

optimum solution
 Matched Filter

Proof:
 Matched Filter
Summary
• The impulse response of the optimum filter
except for the scaling factor k is a time reversed
and delayed version of the input signal g(t), ie, it
is matched to the input signal.

• Matched filters are commonly used in radar, in

which a known signal is sent out, and the
reflected signal is examined for common elements
of the out-going signal.
Inter Symbol Interference
 Inter Symbol Interference
• Another source of bit error in a baseband transmission system
• ISI arises when there is multipath or when the communication
channel is dispersive or bandlimited
• Dispersive means the frequency component contained in the
input pulse are differentially attenuated delayed by the
channel

Effect of ISI in transmission of pulse through a band limited

channel
 Mathematical Model of ISI
• Inorder to illustrate the concept of ISI, consider a
baseband Pulse amplitude modulation system as
shown in fig

a(t)

c(t)

Proof: Baseband data transmission system

 Nyquist Criterion For Distortion less
Transmission (Nyquist Criterion for Zero ISI)
• Inorder to avoid the effect of ISI in the received signal, we need
to select a suitable pulse shape p(t)

where p(0)=1 by normalization

• Thus the Nyquist criteria for distortionless transmission state

that the pulse shaping function p(t) with fourier transform P(f)
satisfies
has zero ISI.
That is Proof
 Pulse Shaping
• Aim is to identify whether there is any particular class of
received pulse shapes ,p(t) which would allow as to
correctly determine the amplitude of the received pulse
pertaining to a particular time slot even in the presence
of ISI

• After identifying such a pulse we can design the transmit

and receive filter so as to obtain such a pulse p(t)
p(t) = Tx Filter * Channel * Receive Filter
(Transfer function of the channel is known a priori)
• In such a pulse shape p(t), for ISI to be zero
P(iTb-kTb)=1 for i=k ie,p(0)=1
=0 i≠k

• The above conditions transforms to the below

expression for p(t) in frequency domain
Σ p(f – n/Tb) = Tb

• A rectangular function in frequency domain

correspondingly a sinc function in time domain
satisfies the above condition
• The ideal Nyquist channel is the one in which the output
pulses from the receive filter have the shape of a sinc
function
• The overall system bandwidth is given by
W= Rb/2= 1/2Tb

• Thus the bit rate Rb=2W is called the Nyquist

rate and W is the minimum bandwidth
required for this rate called the Nyquist
bandwidth

• But the magnitude response p(f) is not

physically realizable due to sudden transitions.
Also the sinc pulses should be sampled exactly
at t=0,Tb,2Tb etc..even a slight sampling jitter
may introduce error
Practical Solution: The Raised Cosine
Pulses
• Where f1 is the maximum frequency
upto which spectrum has a constant
value ie, 1/2W

• ρ is called as the roll off factor

• ρ = 1-(f1/W)

• Transmission Bandwidth, B= 2W-f1

 Statistical characterization of
Correlator(Analyzer) Outputs
• To develop statistical characterization of correlator(analyzer/N
product integrator) outputs
• Let x(t) be the received signal corresponding to the
transmitted signal si(t)

• Assume x(t) is a sample function of the random

process X(t)

• Let Xj denote the random variable whose sample

value is represented by the correlator output xj,
j=1,2,…N
Mean of Xj
Let wj be the jth correlator output in response to the
Gaussian noise component w(t) and w(t) is assumed to be a zero
mean Gaussian noise process
Variance of Xj
Thus the output of the jth correlator Xj is a Gaussian random
variable with mean sij and variance N0/2
MAXIMUM LIKELIHOOD
DECODING
 Likelihood Function
• Consider the received signal x(t) given as
x(t)=si(t)+w(t) 0 ≤t ≤ T and i=1,2,…M

• The observation vector ‘x’ is computed from the received

signal x(t)

• Aim is to find the most likely value of transmitted symbol

mi given the observation vector ‘x’

• At the receiver side, the estimate 𝑚 ෝ of the transmitted

mi is so chosen that it maximizes the conditional
probability density function 𝑓𝑥 (x/mi)
• The conditional probability density function 𝑓𝑥 (x/mi)
i=1,2..M is also known as the likelihood function
L(mi)= 𝑓𝑥 (x/mi)

• It can be interpreted as the likelihood of ‘mi’ for the

observed vector ‘x’
• The observation vector ‘x’ has N components and each ‘Xj’ is a
Gaussian random variable with mean sij and variance N0/2

• Hence

• Since Xj’s are independent Gaussian random variables

• For convenience, the likelihood function is converted to log
likelihood function as it reduces the series of multiplications
to additions

• Ignoring the constant term as it bears no relation

to the fixed message symbol mi, is given as
 Coherent Detection of Signals in
Noise
• Consider ‘M’ signals s1(t),s2(t),…sM(t) is transmitted with equal
probability 1/M

• Each signal is fed to ‘N’ correlators which are also supplied

with ‘N’ orthonormal basis functions

• The resulting correlator outputs define the signal vector si

• The set of message points(in the Euclidean space)

corresponding to the set of transmitted signals si(t)i=1,..M is
called a signal constellation(message point)
• Now when the received signal x(t) is applied to the bank of ‘N’
correlators, the correlator output is defined by the
observation vector ‘x’

• The vector ‘x’ differs from the signal ‘si’ by the noise vector ‘w’
x= si + w

• The representation of observation vector ‘x’ in the Euclidean

space is called received signal point
 Design of Optimum Receiver
• Given the observation vector ‘x’, perform a mapping from x to an
estimate 𝑚
ෝ of the transmitted symbol mi in such a way to minimize
the probability of error in decision making process

Pe(mi/x) = P(mi not sent / x)

= 1- P(mi sent / x)
• Minimizing the probability of error is equivalent to maximizing the a
posteriori probability

ෝ = mi if
set 𝒎

P(mi sent / x) ≥ P(mK sent / x) for all k≠ i ; This decision rule is

referred to as the maximum a posteriori probability (MAP) rule
• Applying Baye’s rule, P(mi sent /x) can be re-written as

• Where
✓ P(mi) is the a prior probability; Pk= Pi when all source
symbols are transmitted with equal probability
✓The denominator fx(x) is independent of transmitted symbol

Thus the MAP rule can be restated as

ෝ = mi if
set set 𝒎
• Since conditional density function bears a one to one
relationship with the log likelihood function, the decision rule
may be restated in terms of
ෝ = mi if
set 𝒎

• This decision rule is referred to as the Maximum Likelihood

(ML) rule

• The receiver based on ML rule is called the ML decoder

• A log-likelihood decoder computes the log likelihood
functions as metrics for all M possible message signals,
compares them and then decides in favour of the maximum
 Graphical Interpretation of ML Rule
• Let ‘Z’ denote the space of all observation vectors ‘x’
called as the observation space

• Based on the ‘M’ symbols the observation space is

partitioned into ‘M’ decision regions Z1,Z2,…ZM

• The decision rule may be

restated as
• But we know that

• Where I is the Euclidean distance between the

received signal point and message point represented by ‘x’
and ‘sK’ resp.

• Thus the ML decision rule is simply to choose the message

point closest to the received signal point, which is intuitively
satisfying

• The Euclidean distance is minimum for k=i

• The need for squarer's in the decision rule can be avoided by
using

✓ The first summation term in the expansion is independent of

the index ‘k’ and therefore may be ignored
✓ The second summation term is the inner product of
observation vector ‘x’ and signal vector ‘sk’
✓ The third summation term is the energy of the transmitted
signal ‘sK’
• Accordingly after rearranging we may restate the decision rule
as
Observation ‘x’ lies in the region Zi if
Conversion of the Continuous AWGN
Channel into a Vector Channel
Correlation Receiver