Pre-requisites to Understand Stochastic

Processes (SPs)
Module – Probability and Random Processes Notes

1. Introduction
Before learning Stochastic Processes (SPs), one must be comfortable with the lan-
guage of Probability and Statistics, as these form the mathematical foundation of
randomness.
In engineering, especially in fields like Electronics, Communication, and Signal
Processing, most real-world signals (like noise, voice, or sensor data) are random in
nature.
To analyze and predict their behavior, we model them statistically using probability
theory.

2. Basic Probability Theory

2.1. Sample Space and Events

• The sample space (S) is the complete set of all possible outcomes of a random
experiment. For example, for tossing a coin: S = {H, T }.

• An event is a subset of S. E.g., “getting a head” is an event in the above experi-

ment.

2.2. Definition of Probability

If n is the total number of outcomes and m are favorable, then
m
P (E) =
n
More generally, for any event E in sample space S,

0 ≤ P (E) ≤ 1 and P (S) = 1

2.3. Addition and Multiplication Theorems

• For two events A and B,

P (A ∪ B) = P (A) + P (B) − P (A ∩ B)

• For independent events,

P (A ∩ B) = P (A)P (B)

1
2.4. Conditional Probability
Conditional probability tells us the probability of event A given that B has already
occurred:
P (A ∩ B)
P (A|B) =
P (B)

2.5. Bayes’ Theorem

Used to update or revise probabilities based on new information — widely applied in
signal detection and decision theory:

P (B|Ai )P (Ai )
P (Ai |B) = P
j P (B|Aj )P (Aj )

3. Random Variables (RVs)

A Random Variable (RV) is a numerical quantity whose value depends on the outcome
of a random experiment. In engineering, we deal with random quantities like current,
voltage, noise, etc., which can be modeled as random variables.

3.1. Types of Random Variables

• Discrete Random Variable: Takes countable values (e.g., number of defective
items).

• Continuous Random Variable: Takes real, continuous values (e.g., amplitude

of noise).

3.2. Distribution Functions

• Probability Mass Function (PMF): For discrete RVs, P (X = xi ) gives the
probability that X takes value xi .

• Probability Density Function (PDF): For continuous RVs, fX (x) represents

density, satisfying Z b
P (a < X < b) = fX (x) dx
a

• Cumulative Distribution Function (CDF):

Z x
FX (x) = P (X ≤ x) = fX (u) du
−∞

2
3.3. Properties of PDF
• fX (x) ≥ 0
Z ∞
• fX (x) dx = 1
−∞

4. Expectation and Moments

The expected value represents the “average outcome” if an experiment is repeated
many times.

4.1. Mean or Expectation

(P
i xi P (X = xi ), (discrete)
E[X] = R∞
−∞
xfX (x) dx, (continuous)

4.2. Variance and Standard Deviation

Variance measures the spread of data around the mean:

V ar(X) = E[(X − E[X])2 ] = E[X 2 ] − (E[X])2

p
Standard deviation is σX = V ar(X)

4.3. Covariance and Correlation

When two random variables X and Y vary together:

Cov(X, Y ) = E[(X − E[X])(Y − E[Y ])]

Cov(X, Y )
ρXY =
σX σY
If ρXY = 0, the variables are uncorrelated.

5. Joint and Conditional Distributions

Often, two or more RVs are related (for instance, input and output voltages in a system).

5.1. Joint Distribution

The joint PDF/PMF fX,Y (x, y) gives the probability of X and Y taking specific values
simultaneously.

3
5.2. Marginal Distribution
Obtained by summing or integrating out the other variable:
Z ∞
fX (x) = fX,Y (x, y) dy
−∞

5.3. Conditional Distribution

Gives the probability of one variable given another:

fX,Y (x, y)
fX|Y (x|y) =
fY (y)

6. Random Vectors and Correlation

In real-world systems, multiple random quantities exist together — e.g., multi-sensor
data or multiple antennas in communication.

• Mean Vector: E[X] = [E[X1 ], E[X2 ], ..., E[Xn ]]T

• Covariance Matrix: CX = E[(X − E[X])(X − E[X])T ]

This concept is widely used in analyzing signals from sensor arrays, radar, and MIMO
systems.

7. Important Probability Distributions

Distribution Mean Variance

Bernoulli(p) p p(1 − p)
Binomial(n, p) np np(1 − p)
Poisson(λ) λ λ
a+b (b − a)2
Uniform(a, b)
2 12
1 1
Exponential(λ)
λ λ2
Gaussian(µ, σ 2 ) µ σ2

8. Random Signals and Autocorrelation

In electronics and communication systems, noise and information signals are random in
nature. To analyze their properties, we use correlation functions.

4
 • Deterministic Signal: Exactly known (e.g., x(t) = sin(2πf t))

• Random Signal: Value at any time is uncertain, e.g., thermal noise.

8.1. Autocorrelation Function (ACF)

Autocorrelation measures the similarity between signal values at different time instants:

RX (t1 , t2 ) = E[X(t1 )X(t2 )]

For stationary processes:

RX (τ ) = E[X(t)X(t + τ )]
The autocorrelation function provides insights about signal predictability and bandwidth.

RX (τ )
Typical ACF of Random Signal

9. Power and Energy of Signals

Power and energy are important measures of signal strength.

• Energy of a Signal: Z ∞
E= |x(t)|2 dt
−∞

• Average Power: Z T
1
P = lim |x(t)|2 dt
T →∞ 2T −T

In communication, energy signals represent data bursts, while power signals represent
continuous noise or carrier waves.

10. Connection to Stochastic Processes

A Stochastic Process is simply a time-indexed collection of random variables:

X(t), t∈T

where T is the parameter space (time) and the possible values of X(t) form the state
space.

5
 • Probability theory describes uncertainty in one experiment.

• Random variables extend it to numerical outcomes.

• Random processes extend it over time.

Probability Theory Random Variables Random Processes Stochastic Systems

11. Conclusion
A solid grasp of the above topics ensures smooth understanding of advanced stochastic
process concepts like:

• Classification of processes based on parameter and state space

• Stationary and ergodic processes

• Power spectral density (PSD)

These topics help engineers analyze noise, design optimal filters, and improve communi-
cation reliability.

6
 EC503 – Digital Communication
Module I – Stochastic Processes

Introduction to Stochastic Processes

Formal Definition
A stochastic process (SP) is defined as a family of random variables:

{X(t) : t ∈ T }

where • T is the index (parameter) space, often representing time • X(t) is a random variable
defined on the same probability space
(Ω, F, P )

Each ω ∈ Ω generates a sample function or realization X(t, ω).

Intuitive Meaning
A stochastic process describes how uncertainty evolves over time. Useful in:

• noise modeling

• signal variations

• digital data sequences

• queue lengths, traffic models

Random Variables vs Random Processes

• A random variable describes randomness at a single instant.

• A random process describes randomness over time.

State Space (Definition)

The set of all possible values a process may take:

S = {x : P (X(t) = x) > 0}

Examples:

• {+A, −A} → binary waves

• R → Gaussian noise

1
Parameter (Time) Space (Definition)
The set of all index values for which the process is defined:
T = {t} (discrete or continuous)

Classification of Stochastic Processes

• Discrete-time, discrete-state: digital data sequences, Markov chains
• Continuous-time, continuous-state: Brownian motion
• Continuous-time, discrete-state: Poisson process
• Discrete-time, continuous-state: sampled Gaussian noise

Moments of a Random Process

Mean:
mX (t) = E[X(t)]

Second moment:
E[X 2 (t)]

Variance:
2
σX (t) = E[(X(t) − mX (t))2 ]

Autocorrelation:
RX (t1 , t2 ) = E[X(t1 )X(t2 )]

Higher-order moments describe shape, skewness, kurtosis.

Stationary and Ergodic Processes

A random process is stationary if its statistical characteristics do not change with time. It
ensures that the system behavior is predictable based on time differences, not the exact time.

Strict-Sense Stationary (SSS) Process

A process is SSS if:
FX(t1 ),...,X(tn ) = FX(t1 +τ ),...,X(tn +τ ) , ∀τ
All joint PDFs remain unchanged under time shift.

Wide-Sense Stationary (WSS) Process

A weaker form of stationarity, defined by:
E[X(t)] = m = constant
RX (t1 , t2 ) = RX (τ ), τ = t2 − t1

2
Ergodic Process — Formal Definition
A process is ergodic if time averages (from one long signal realization) equal ensemble averages
(statistical averages taken over all possible signals).
A process is ergodic in mean if:
Z T
1
lim X(t) dt = E[X(t)]
T →∞ T 0

A process is ergodic in autocorrelation if:

1 T
Z
lim X(t)X(t + τ ) dt = RX (τ )
T →∞ T 0

Importance of Ergodicity
Ergodic processes allow:

• estimation of statistical properties from one long observation

• practical measurement of PSD and ACF

Most communication noise models assume ergodicity.

Correlation, Covariance and Related Definitions

Autocorrelation (Definition)
RX (t1 , t2 ) = E[X(t1 )X(t2 )]

For WSS:
RX (τ ) = E[X(t)X(t + τ )]

Autocovariance (Definition)
CX (t1 , t2 ) = E[(X(t1 ) − m)(X(t2 ) − m)]
For WSS:
CX (τ ) = RX (τ ) − m2

Correlation Coefficient
CXY
ρXY = p
CX (0)CY (0)
Measures similarity between signals.

3
Properties of Autocorrelation
• RX (0) is maximum

• RX (τ ) = RX (−τ ) (even)

• |RX (τ )| ≤ RX (0)

• Non-negative definite kernel

These ensure the PSD is real and non-negative.

Random Binary Wave (Formal Treatment)

A binary wave is defined as:
X(t) ∈ {+A, −A}

Definition
A random binary waveform is a discrete-state, continuous-time stochastic process generated
by mapping random bits to amplitude levels.

Mean
m = E[X(t)] = A(2p − 1)

Autocorrelation
If symbols are independent: (
A2 , |τ | < T
RX (τ ) =
m2 , |τ | ≥ T

Usage
• Baseband signaling (NRZ, RZ)

• PCM signals

• Modeling digital data streams

Power Spectral Density (PSD)

Definition
The PSD describes how signal power is distributed in the frequency domain.

4
 Z ∞
SX (f ) = RX (τ )e−j2πf τ dτ
−∞

Key Theoretical Properties

• SX (f ) ≥ 0

• Even function: SX (f ) = SX (−f )

• Total average power: Z ∞

P = RX (0) = SX (f )df
−∞

• Wiener–Khinchin theorem links PSD and ACF.

Example — White Noise

RX (τ ) = N0 δ(τ ) ⇒ SX (f ) = N0

Markov Chains

Definition
A discrete-time stochastic process is a Markov chain if:

P (Xn+1 = j | Xn = i, Xn−1 , ...) = P (Xn+1 = j | Xn = i)

This property is called the Markov (memoryless) property.

Transition Probability Matrix

 
P11 P12 · · ·
P = P21 P22 · · ·
 
.. .. . .
. . .
where
Pij = P (Xn+1 = j | Xn = i)
Each row sums to 1.

Chapman–Kolmogorov Equation (Definition)

P (n) = P n
(n)
X (n−1)
Pij = Pik Pkj
k

5
n-Step Transition Probability
(n)
Pij = (P n )ij

Classification of States
Recurrent State:
P (return to state i) = 1

Transient State:
P (return to state i) < 1

Periodic State:

∃k > 1 : P (Xn+k = i | Xn = i) = 0 unless n ≡ 0 mod k

Aperiodic State: no such restriction.

Steady-State Distribution
A vector π satisfying: X
πP = π, πi = 1
i

represents long-run state probabilities.

Applications:

• Bursty channel models (Gilbert–Elliott)

• Queueing networks

• Wireless fading state transitions

6
 EC503 – Digital Communication
Module II – Signal Vector Representation

Introduction to Signal Vector Representation

Digital communication signals are finite-energy waveforms defined over a finite interval [0, T ]. A
powerful method of representing such signals is to treat them as vectors in an N -dimensional Eu-
clidean space. This approach is foundational for modern detection theory, constellation design,
and receiver optimization.

Finite Energy Signal (Definition)

A signal s(t) is a finite-energy signal if:
Z T
Es = |s(t)|2 dt < ∞
0

Signal Space (Definition)

Signal space is the set of all finite-energy signals defined over [0, T ]. This forms a Hilbert space
with inner product: Z T
⟨s1 , s2 ⟩ = s1 (t) s2 (t) dt
0

Vector Analogy (Definition)

Each signal can be represented uniquely by a coordinate vector:
s = (s1 , s2 , ..., sN )
in an orthonormal basis.

Orthogonality, Orthonormality and Basis Functions

Orthogonality (Definition)
Two signals s1 (t) and s2 (t) are orthogonal if:
⟨s1 , s2 ⟩ = 0

Orthonormality (Definition)
A set {ϕk (t)} is orthonormal if: (
1, i = j
⟨ϕi , ϕj ⟩ =
̸ j
0, i =

1
Basis Function (Definition)
A basis function is an elementary, independent waveform used for expanding any signal in the
space:
N
X
s(t) = sk ϕk (t)
k=1

Coordinate of a Signal (Definition)

Z T
sk = s(t)ϕk (t) dt
0

Orthogonal Signal Space and Geometric Interpretation

Orthogonal Signal Space (Definition)

The N -dimensional space spanned by mutually orthonormal basis functions.

Geometric Interpretation (Definition)

Each signal si (t) corresponds to a unique point (vector) si in signal space.

Euclidean Distance Between Signals

dij = ∥si − sj ∥

Distinguishability of Signals (Definition)

Two signals are distinguishable iff dij > 0.

Message Point and Signal Constellation

Message Point (Definition)

The coordinate vector corresponding to a transmitted signal si (t):
si = (si1 , si2 , . . . , siN )

Signal Constellation (Definition)

The set of all message points:
C = {s1 , s2 , ..., sM }

Examples: BPSK (2 points), QPSK (4 points), QAM (M -point grids), etc.

2
Likelihood Functions and Detection Theory

Likelihood Function (Definition)

L(r|Hi ) = p(r(t) | si (t))

AWGN Channel Model

r(t) = si (t) + n(t), n(t) ∼ N (0, N0 /2)

Likelihood Under AWGN

 
1 2
Li ∝ exp − ∥r − si ∥
N0

Log-Likelihood Function (Definition)

1
ln Li = − ∥r − si ∥2 + constant
N0

Schwartz (Cauchy–Schwarz) Inequality

Statement

|⟨s, x⟩|2 ≤ ⟨s, s⟩ ⟨x, x⟩

Geometric Interpretation
Angle θ between vectors:
⟨s, x⟩
cos θ = √
Es Ex

Gram–Schmidt Orthogonalization Procedure

Definition
A systematic method to convert linearly independent signals into orthonormal basis functions.

Procedure
Given s1 , s2 , ..., sN :

3
 s1
ϕ1 =
∥s1 ∥

ŝ2
ŝ2 = s2 − ⟨s2 , ϕ1 ⟩ϕ1 ϕ2 =
∥ŝ2 ∥

General step:
k−1
X
ŝk = sk − ⟨sk , ϕi ⟩ϕi
i=1

ŝk
ϕk =
∥ŝk ∥

Response of a Noisy Signal at the Receiver

r(t) = si (t) + n(t)

Projection on basis: Z T
rk = r(t)ϕk (t)dt = sik + nk
0

Noise projection:
nk ∼ N (0, N0 /2)

Maximum Likelihood (ML) Decision Rule

Definition
Choose the hypothesis with maximum likelihood:

îM L = arg max L(r|Hi )

i

ML Rule in AWGN

î = arg min ∥r − si ∥
i

Thus ML = minimum Euclidean distance rule.

Decision Regions and Decision Boundaries

Decision Region (Definition)

The set of all received vectors for which the receiver decides a specific signal.

4
Decision Boundary (Definition)
The hypersurface satisfying:
∥r − si ∥2 = ∥r − sj ∥2

A (N − 1)-dimensional hyperplane orthogonal to (si − sj ).

Optimum Correlation Receiver

Definition
A receiver that computes:
zi = ⟨r(t), si (t)⟩
and chooses the largest correlation value.
Equivalent to ML in AWGN.

Probability of Error

Error Probability in Binary Detection

For two signals separated by Euclidean distance d in AWGN, the probability of detection error
is !
d
Pe = Q p .
2 N0 /2

Q-Function (Definition)
The Q-function represents the tail probability of the standard Gaussian distribution:
Z ∞
1 2
Q(x) = √ e−u /2 du.
2π x

Error Function (Definition)

The error function is defined as
Z x
2 2
erf(x) = √ e−u du.
π 0

Complementary Error Function

The complementary error function is defined as

erfc(x) = 1 − erf(x).

5
Type-I and Type-II Errors

Type-I Error (False Alarm / False Positive)

Definition: A Type-I error occurs when the receiver decides in favor of hypothesis H1 even
though the true hypothesis is H0 .

PType-I = P (decide H1 | H0 is true)

Interpretation: The receiver falsely detects the presence of signal s1 (t) when s0 (t) was actually
transmitted.
Also called: False Alarm Probability

PF A = PType-I

Example: Detecting a binary “1” when the transmitted symbol was “0”.

Type-II Error (Miss / False Negative)

Definition: A Type-II error occurs when the receiver decides in favor of hypothesis H0 even
though the true hypothesis is H1 .

PType-II = P (decide H0 | H1 is true)

Interpretation: The receiver fails to detect the transmitted signal s1 (t) and incorrectly decides
that s0 (t) was sent.
Also called: Miss Detection Probability

PM D = PType-II

Example: Detecting a binary “0” even though “1” was transmitted.

6
 EC503 – Digital Communication
Module III (Part 1) – Digital Data Transmission

Digital Data Transmission

This module introduces the fundamental principles of converting analog signals into digital form
and transmitting them efficiently. The key stages include sampling, pulse modulation, quantiza-
tion, encoding, companding, and differential waveform coding.

Concept of Sampling
Sampling (Definition): Sampling is the process of converting a continuous-time signal x(t) into
a discrete-time sequence by taking samples at equal intervals:

x(nTs ), n = 0, 1, 2, . . .

Sampling Frequency:
1
fs =
Ts
Nyquist Sampling Theorem: A bandlimited signal of bandwidth B can be perfectly recon-
structed from its samples if:
fs ≥ 2B
This minimum value 2B is called the Nyquist rate.

Pulse Amplitude Modulation (PAM)

PAM (Definition): PAM is a modulation technique in which the amplitudes of regularly spaced
pulses are varied in proportion to the message signal.
General PAM representation:
∞
X
s(t) = x(nTs )p(t − nTs )
n=−∞

Applications:

• Digital telephony

• PCM systems

• Time-division multiplexing (TDM)

1
Interlacing and Multiplexing of Samples
Multiplexing (Definition): Multiplexing combines multiple sampled message signals into a
single composite stream.
Interlacing (Definition): Interlacing arranges samples from different signals in alternating time
slots.
Time Division Multiplexing (TDM): Each channel is assigned a separate time slot within a
periodic frame:
Frame = [x1 , x2 , . . . , xK ]
TDM is widely used in telephony and PCM-based systems.

Pulse Code Modulation (PCM)

PCM (Definition): PCM converts an analog signal into a binary sequence through:

1. Sampling

2. Quantization

3. Binary encoding

PCM is widely used in:

• Digital telephony (e.g., 64 kbps PCM voice)

• Digital audio

• Communication networks

Quantization
Quantization (Definition): The process of mapping a continuous-valued sample to a discrete
level:
Q(x) = qi for x ∈ [bi , bi+1 )

Uniform Quantization
Definition: Equal spacing between quantization levels:
xmax − xmin
∆=
L
where L is the number of quantization levels.
Useful for signals with nearly uniform amplitude distribution.

Non-Uniform Quantization
Definition: Quantization intervals are narrower for small amplitudes and wider for large ampli-
tudes. Used for voice/speech signals, where low amplitudes are more probable.

2
Quantization Noise
Quantization Noise (Definition): Error between the input sample and quantized output:

eq (n) = x(n) − Q(x(n))

For a uniform mid-rise quantizer with step size ∆, assuming eq is uniformly distributed in
[−∆/2, ∆/2]:
∆2
σq2 = E[e2q ] =
12

Binary Encoding
Binary Encoding (Definition): Mapping each quantization level to a unique binary word.
For L quantization levels:
Bits per sample = log2 L

Example: L = 256 ⇒ 8 bits/sample (standard telephony PCM).

A-Law and µ-Law Companding

Companding (Definition): A nonlinear operation (compressor at transmitter + expander at
receiver) applied to reduce quantization noise for low-level signals.
Companding = COMPressing + exPANDING.

A-Law (Used in Europe)

For normalized input x with |x| ≤ 1:


A|x| 1

 , 0 ≤ |x| <
C(x) = 1 + ln A A
1 + ln(A|x|) 1
, ≤ |x| ≤ 1


1 + ln A A
where A is typically 87.6.

µ-Law (Used in USA and Japan)

ln(1 + µ|x|)
C(x) = , |x| ≤ 1
ln(1 + µ)
where µ = 255 for telephony.
Purpose of Companding:

• Improves Signal-to-Quantization-Noise Ratio (SQNR) for weak signals.

• Adapts to non-uniform amplitude distributions of speech.

3
Differential Pulse Code Modulation (DPCM)
DPCM (Definition): A predictive coding technique that encodes the difference between the
present sample and a predicted sample:

e(n) = x(n) − x̂(n)

Here x̂(n) is obtained from past reconstructed samples using a predictor.

Advantages:

• Reduces number of bits required.

• Exploits correlation between successive samples (especially in speech).

Delta Modulation (DM)

DM (Definition): A special case of DPCM where:

• Predictor is a one-step integrator.

• Only the sign of the difference is transmitted (1-bit per sample).

(
+1, x(n) > x̂(n)
s(n) =
−1, x(n) < x̂(n)

Problems in DM:

• Slope overload distortion when signal slope is high and step size is too small.

• Granular noise when step size is too large for slowly varying signals.

Adaptive Delta Modulation (ADM)

ADM (Definition): Improved DM where the step size ∆(n) is varied according to the recent
behavior of the signal.
Example rule:
(
α∆(n), if two successive steps have same sign
∆(n + 1) =
β∆(n), otherwise

with α > 1 and 0 < β < 1.

Advantages of ADM:

• Reduces slope overload by increasing step size during rapid changes.

• Reduces granular noise by decreasing step size when signal is slowly varying.

• Better suitable for speech and audio signals.

4
Digital Data Transmission
Digital communication involves converting analog or digital information into discrete-time, dig-
itized signals suitable for reliable transmission through physical channels. This module covers
sampling, pulse modulation, quantization, encoding, companding, and the components of a digi-
tal transmission system, including line coding, pulse shaping, ISI, eye patterns, and equalization.

Concept of Sampling
Sampling (Definition): Sampling converts a continuous-time signal x(t) into a discrete-time
sequence:
x[n] = x(nTs )
1
where Ts is the sampling interval and fs = Ts
is the sampling frequency.
Nyquist Sampling Theorem: A bandlimited signal of bandwidth B Hz can be uniquely re-
constructed if:
fs ≥ 2B

Pulse Amplitude Modulation (PAM)

PAM (Definition): PAM represents sampled values by scaling the amplitudes of pulses:
∞
X
s(t) = x[n]p(t − nTs )
n=−∞

Used in PCM systems, multiplexing, and digital telephony.

Interlacing and Multiplexing of Samples

Multiplexing (Definition): Combining multiple sampled message signals into a single data
stream for transmission.
Interlacing (Definition): Assigning each signal a unique time slot in a repeating frame.
TDM Frame:
Frame = [x1 (n), x2 (n), . . . , xM (n)]

Pulse Code Modulation (PCM)

PCM (Definition): A process that converts continuous-valued samples into binary sequences
through:

1. Sampling

2. Quantization

3. Binary encoding

Widely used in telephone networks (64 kbps PCM voice).

5
Quantization
Quantization (Definition): Mapping a continuous sample x to a discrete output level:

Q(x) = qi , x ∈ [bi , bi+1 )

Uniform Quantization
Definition: Equal-width quantization intervals:
xmax − xmin
∆=
L

Non-Uniform Quantization
Definition: Unequal quantization intervals; finer resolution near zero. Used for speech where
small amplitudes occur frequently.

Quantization Noise
Error:
eq (n) = x(n) − Q(x(n))

For uniform quantizers:

∆2
σq2 =
12

Binary Encoding
Definition: Mapping each quantization level to binary.
For L quantization levels:
Bits/sample = log2 L

Example: L = 256 ⇒ 8 bits/sample.

A-Law and µmu-Law Companding

Companding (Definition): Compressing dynamic range before quantization and expanding it
afterward.

A-Law

A|x|

 , |x| < 1/A
C(x) = 11 + ln A
 + ln(A|x|) , 1/A ≤ |x| ≤ 1

1 + ln A

6
µ-Law

ln(1 + µ|x|)
C(x) = , µ = 255
ln(1 + µ)
Used in North America and Japan.

Differential PCM (DPCM)

DPCM (Definition): Encodes differences between consecutive samples:

e(n) = x(n) − x̂(n)

Advantages:

• Lower bit rate

• Good for correlated signals like speech

Delta Modulation (DM)

DM (Definition): A 1-bit DPCM where only the sign of error is transmitted:
(
+1, x(n) > x̂(n)
s(n) =
−1, x(n) < x̂(n)

Issues:

• Slope overload (too small step size)

• Granular noise (too large step size)

Adaptive Delta Modulation (ADM)

ADM (Definition): The step size is varied adaptively:
(
α∆(n), same sign
∆(n + 1) =
β∆(n), change of sign

Digital Transmission Components

Source
Generates the analog or digital data to be transmitted.

Multiplexer
Combines multiple data streams into one using TDM/FDM/CDM.

7
Line Coder
Line Coding (Definition): Mapping bits {0, 1} to waveforms s(t).
Required properties:

• DC suppression

• Self-clocking

• Efficient bandwidth usage

Regenerative Repeater
Reconstructs digital pulses using:

• Sampling

• Threshold detection

• Re-transmission

It removes accumulated noise/distortion.

Line Coding Techniques

Unipolar NRZ

1 = +A, 0 = 0 Inefficient, strong DC component.

Polar NRZ

NRZ-L: 1 = +A, 0 = -A NRZ-I: 1 = transition, 0 = no transition.

Polar RZ

1 = half-bit positive pulse; 0 = negative/zero. Better timing; higher bandwidth.

Bipolar NRZ (AMI)

1 alternates between +A and –A; 0 = 0. Zero DC component.

Manchester Coding

1 = High→Low; 0 = Low→High. Self-clocking; doubles bandwidth.

8
Differential Encoding
Definition: Information encoded as transitions, not absolute levels:

1 = change polarity, 0 = same polarity

Used in Differential NRZ and Differential Manchester.

PSD of Line Codes

PSD (Definition): Distribution of signal power across frequencies.
Key behaviors:

• Unipolar NRZ: high DC content.

• Polar NRZ: suppressed DC.

• Manchester: notch at DC; high bandwidth.

• AMI: excellent DC suppression.

Pulse Shaping
Pulse Shaping (Definition): Designing pulses to limit bandwidth and reduce ISI.
Common pulses:

• Rectangular

• Raised cosine

• Root raised cosine

Inter-Symbol Interference (ISI)

ISI (Definition): Overlap of pulses into adjacent symbol intervals causing symbol distortion.

∞
X
r(n) = ak h(nT − kT )
k=−∞

Eye Pattern
Eye Pattern (Definition): A superposition of multiple received waveforms used to visualize
ISI and noise.
Interpretation:

• Large eye opening → low ISI, reliable timing.

• Closed eye → severe ISI.

9
Nyquist Criterion for Zero ISI
Time-domain criterion: (
1, n = 0
p(nT ) =
̸ 0
0, n =

Frequency-domain condition:
 
1
P (f ) + P f − = constant
T

Equalizer
Equalizer (Definition): A receive-side filter that compensates channel distortion to minimize
ISI.

Zero Forcing Equalizer (ZFE)

1
Heq (f ) =
Hc (f )

Advantage:

• Eliminates ISI completely (in theory)

Disadvantage:

• Amplifies noise when Hc (f ) is small

Timing Extraction (Clock Recovery)

Definition: Deriving symbol timing from the received signal.
Methods:

• Early–Late Gate

• Mueller–Müller algorithm

• PLL-based timing recovery

10