EE6640 Wmu

ECE 6640
Digital Communications
Dr. Bradley J. Bazuin

Assistant Professor
Department of Electrical and Computer Engineering
College of Engineering and Applied Sciences
Course/Lecture Overview
• Syllabus
• Personal Intro.
• Textbook/Materials Used
• Additional Reading
• ID and Acknowledgment of Policies
• Textbook
• Chapter 1
ECE 6640 2
Syllabus
• Everything useful for this class can be found on Dr. Bazuin’s web site!
– http://homepages.wmich.edu/~bazuinb/
• The class web site is at

– http://homepages.wmich.edu/~bazuinb/ECE6640/ECE6640_Sp14.htm
• The syllabus …
– http://homepages.wmich.edu/~bazuinb/ECE6640/Syl_6640.pdf
ECE 6640 3
Who am I?
• Dr. Bradley J. Bazuin

– Born and raised in Grand Rapids Michigan
– Undergraduate BS in Engineering and Applied Sciences, Extensive
Electrical Engineering from Yale University in 1980
– Graduate MS and PhD in Electrical Engineering from Stanford
University in 1982 and 1989, respectively.
– Industrial Experience – Digital, ASIC, System Engineering
• Part-time ARGOSystems, Inc. (purchased by Boeing) 1981-1989
• Full-time ARGOSystems, Inc. 1989-1991
• Full-time Radix Technologies 1991-2000
– Academic Experience – Electrical and Computer Engineering
• Term-appointed Faculty, WMU ECE Dept. 2000-2001
• Tenure track Assistant Professor, WMU ECE Dept. 2001-2007
ECE 6640 • Tenured Associate Professor, WMU ECE Dept. 2007- present 4
Research Activities and Interests
• Sunseeker
– Adviser to solar car team
– Electrical Systems: Li battery protection system, Controller Area Network (CAN)
based sensors and controllers, Solar Array Energy Collection and Conversion
• Center for the Advancement of Printed Electronics (CAPE)
– Printed electronic device design, fabrication and testing
– Semiconductor Physics
• Physical Layer Communication Signal Processing
– Software Defined Radios (SDR)
– Mulitrate Signal Processing (digital channel bank analysis and synthesis, filter-decimation and
interpolation-filter design methods)
– Adaptive Filtering and Systems (channel equalization, smart-antenna spatial beamforming)
• Communication-based Digital Signal Processing Algorithm Implementation
– Xilinx programmable devices
– Parallel processing, hosts including NVIDIA GPUs with CUDA and multithreaded applications
ECE 6640 5
Required Textbook/Materials
• Bernard Sklar, Digital Communications, Fundamentals

and Applications, Prentice Hall PTR, Second Edition,
2001. ISBN: 0-13-084788-7.
• SystemView by ELANIX CD with textbook
• MATLAB, Student Edition

• MATLAB Signal Processing Toolbox
– The MATH Works,
MATLAB and Signal Processing Toolbox
http://www.mathworks.com/
ECE 6640 6
Supplemental Books and Materials
• John G. Proakis and Masoud Salehi, “Digital Communications, 5th
ed.,” McGraw Hill, Fifth Edition, 2008. ISBN: 978-0-07-295716-7.
• John G. Proakis and Masoud Salehi, “Communication Systems
Engineering, 2nd ed.”, Prentice Hall, 2002. ISBN: 0-13-061793-8.
• A. Bruce Carlson, P.B. Crilly, “Communication Systems, 5th ed.”,
McGraw-Hill, 2010. ISBN: 978-0-07-338040-7.
• Leon W. Couch II, “Digital and Analog Communication Systems, 7th
ed.”, Prentice Hall, 2007. ISBN: 0-13-142492-0.
• Stephen G. Wilson, “Digital Modulation and Coding, ” Prentice-Hall,
1996. ISBN: 0-13-210071-1.
• Ezio Biglieri, D. Divsalar, P.J. McLane, M.K. Simon, “Introduction
to Trellis-Coded Modulation with Applications”, Macmillan, 1991.
ISBN: 0-02-309965-8.
ECE 6640 7
Identification and Acknowledgement
• Identification for Grade Posting,

Course and University Policies, and
Acknowledgement
• Please read, provide unique identification, sign and date,

and return to Dr. Bazuin.
ECE 6640 8
Course/Text Overview
1. Signals and Spectra.
Digital Communication Signal Processing. Classification of Signals.

Spectral Density. Autocorrelation. Random Signals. Signal
Transmission through Linear Systems. Bandwidth of Digital Data.
2. Formatting and Baseband Modulation.
Baseband Systems. Formatting Textual Data (Character Coding).

Messages, Characters, and Symbols. Formatting Analog
Information. Sources of Corruption. Pulse Code Modulation. Uniform
and Nonuniform Quantization. Baseband Modulation. Correlative
Coding.
ECE 6640 9
Course/Text Overview (2)
3. Baseband Demodulation/Detection.
Signals and Noise. Detection of Binary Signals in Gaussian Noise.

Intersymbol Interference. Equalization.
4. Bandpass Modulation and Demodulation/Detection.
Why Modulate? Digital Bandpass Modulation Techniques. Detection

of Signals in Gaussian Noise. Coherent Detection. Noncoherent
Detection. Complex Envelope. Error Performance for Binary
Systems. M-ary Signaling and Performance. Symbol Error
Performance for M-ary Systems (M>>2).
Exam #1
ECE 6640 10
5. Communications Link Analysis.
What the System Link Budget Tells the System Engineer. The
Channel. Received Signal Power and Noise Power. Link Budget
Analysis. Noise Figure, Noise Temperature, and System
Temperature. Sample Link Analysis. Satellite Repeaters. System
Trade-Offs.
ECE 6640 11
6. Channel Coding: Part 1.
Waveform Coding. Types of Error Control. Structured Sequences.

Linear Block Codes. Error-Detecting and Correcting Capability.
Usefulness of the Standard Array. Cyclic Codes. Well-Known Block
Codes.
Convolutional Encoding. Convolutional Encoder Representation.

Formulation of the Convolutional Decoding Problem. Properties of
Convolutional Codes. Other Convolutional Decoding Algorithms.
Exam #2
ECE 6640 12
Reed-Solomon Codes. Interleaving and Concatenated Codes. Coding

and Interleaving Applied to the Compact Disc Digital Audio System.
Turbo Codes.
Appendix 8A. The Sum of Log-Likelihood Ratios.
9. Modulation and Coding Trade-Offs.
Goals of the Communications System Designer. Error Probability

Plane. Nyquist Minimum Bandwidth. Shannon-Hartley Capacity
Theorem. Bandwidth Efficiency Plane. Modulation and Coding
Trade-Offs. Defining, Designing, and Evaluating Systems.
Bandwidth-Efficient Modulations. Modulation and Coding for
Bandlimited Channels. Trellis-Coded Modulation.
Final Exam
ECE 6640 13
Advanced Topics (as time permits)
11. Multiplexing and Multiple Access.
Allocation of the Communications Resource. Multiple Access

Communications System and Architecture. Access Algorithms.
Multiple Access Techniques Employed with INTELSAT. Multiple
Access Techniques for Local Area Networks.
12. Spread-Spectrum Techniques.
Spread-Spectrum Overview. Pseudonoise Sequences. Direct-

Sequence Spread-Spectrum Systems. Frequency Hopping Systems.
Synchronization. Jamming Considerations. Commercial
Applications. Cellular Systems.
Final Exam
ECE 6640 14
Text Appendices
A. A Review of Fourier Techniques.
Signals, Spectra, and Linear Systems. Fourier Techniques for Linear

System Analysis. Fourier Transform Properties. Useful Functions.
Convolution. Tables of Fourier Transforms and Operations.
B. Fundamentals of Statistical Decision Theory.
Bayes' Theorem. Decision Theory. Signal Detection Example.
C. Response of a Correlator To White Noise.
D. Often-Used Identities.
E. s-Domain, z-Domain and Digital Filtering.
F. List of Symbols.
G. SystemView by ELANIX Guide to the CD.

ECE 6640 15
Comments from 2006 Offering
• A strong focus on themes and critical results for each chapter covered
is needed. The text author provides his own list of critical elements,
they can be incorporated into the instructors set.
• Matlab simulations of all significant concepts should be available.

They allow the students to perform theoretical computations and then
observe what the computations mean, particularly as it relates to bit-
error rate performance, digital modulation and coherent and non-
coherent demodulation, and channel encoding and decoding.
• The software that comes with the text provides demonstrations, but it
is not user friendly and the software is very out-of-data (no longer
supported).
ECE 6640 16
Chapter 1
1. Signals and Spectra.
1.1 Digital Communication Signal Processing.
1.2 Classification of Signals.
1.3 Spectral Density.
1.4 Autocorrelation.
1.5 Random Signals.
1.6 Signal Transmission through Linear Systems.
1.7 Bandwidth of Digital Data.
A review of prerequisite material that is critically important

when studying digital communication systems.
ECE 6640 17
Sklar’s Communications System
Notes and figures are based on or taken from materials in the course textbook:
ECE 6640 Bernard Sklar, Digital Communications, Fundamentals and Applications, 18
Prentice Hall PTR, Second Edition, 2001.
Simplified Communications System
• Format: making the message compatible with digital processing
• Source Coding: efficient descriptions of information sources
• Channel Coding: signal transformation enabling improved reception
performance after expected channel impairments
• Modulation: formation of the baseband waveform
• RF Mixing: frequency domain translation of baseband signal
• Transmit/Receive: RF Amplifiers and Filters
Information Source Channel Mod- Antenna

Format RF Mixing Transmitter
Message Encode Encode ulation
RF Signal
Noise
Bits Symbols Signals

Interference
Information Source Channel Demod- Antenna

Reformat RF Mixing Receiver
Message Decode Decode ulation
ECE 6640 19
Communication Channel
Linear Nonlinear Atten-

Filtering Distortion uation
Noise
Interference
Transmitting Receiving
Antenna Antenna
RF Communication Channel
• The channel greatly effects received RF signals

– Frequencey, Bandwidth, Transmitted Signal Power, RF Propagation
– Attenuation, Nonlinear Distortion, Multipath, Range, Direction
– Signal-to-Noise Ratio (SNR) and Signal-to-Interference Ratio (SIR)
–
ECE 6640 Minimum Detectable Signal Level (MDS), Noise Floor 20
Received Signal
r t   s t   hc t   s2 t   h2 t     s N t   hN t  nt 
• The receiver must extract the original message as best

possible!
• Multiple signals with similar channel characteristics may be
present
• The RF channel(s) must be allocated and efficiently utilized.
– Frequency band assignments and regulations (power, direction, etc.)
– Signal modulation structures have different characteristics
ECE 6640 21
Why Digital?
1. Noise, Interference, Path Loss, and Channel Impairments
(signal environment)
2. Cost
3. Inherent Availability
4. Reliability and Reconfigurability
Terminology
• Information Source
• Textual Message
• Character
• Binary Digit (Bit)
• Bit Stream
• Symbol
• Digital Waveform
• Data Rate
Signal Processing Functions
Classification of Signals
• Deterministic and Random
• Periodic and Non-periodic
• Analog and Discrete/Digital
• Energy and Power Signals
ECE 6640 25
SKLAR DSP Tutorial
• The CD that comes with the text includes a

“Concise DSP Tutorial” in pdf format
• Table of Contents:
– Frequency Domain Analysis – critical importance
– General Digital Filters – important
– Finite Impulse Response (FIR) Filters – critical importance
– Infinite Impulse Response (IIR) Filters – useful but …
– Filter Design Techniques – will be discussed and provided
– Adaptive Filters – saved for Dr. Bazuin’s ECE6950 course
• Also see Appendix B: Fundamentals of Statistical Decision
Theory
– Specific material from probability and statistics is required.
ECE 6640 (ECE 3800 or ECE5820 material) 26
Spectral Density
• Energy Spectral Density


EX   x t   dt
2

X f   X f   X f 
*
• Power Spectral Density

T0
2
 x t   dt
1
PX 
2
T0 T0

2
1 *
G X f   lim  X T f   X T f  

T  T

ECE 6640 27
Autocorrelation
• of an Energy Signal

R XX     xt  x t   dt

• Properties:
1. Energy  
R XX 0   E X 2  X 2
2. Symmetry R XX    R XX   
3. Maximum R XX    R XX 0
4. Transform Pair R XX   XX f 
ECE 6640 28
Autocorrelation
• of a Power Signal
T
1 2
 XX    lim  x t   x t     dt
T  T
T
2
• Properties: T0
2
 XX 0    x t 
1
 dt
2
1. Energy T0

T0
2
2. Symmetry XX   XX  
3. Maximum XX   XX 0
4. Transform Pair XX   G XX f 
ECE 6640 29
Random Signals
1 Distribution Functions
Probability Distribution Function (PDF) or
Cumulative Distribution Function (CDF) [preferred]
 0  FX  x   1, for    x  
 FX     0 and FX    1
 FX is non-decreasing as x increases
 Pr  x1  X  x 2   FX  x 2   FX  x1 
For discrete events
For continuous events
ECE 6640 30
Random Signals
2. Density Functions
Probability Density Function (pdf) Probability Mass Function (pmf)
 f X x   0, for    x    f X  x   0, for    x  


 f x   dx  1
X  f x   dx  1
X
   
x x
FX 
 f X u   du FX 
 f X u   du
   
x2
Pr x1  X  x 2    f x   dx
x2
Pr  x1  X  x 2    f x   dx
X
X
 x1  x1
Functions of random variables

f Y  y   f X x  
dx
dy
ECE 6640 31
Random Signals
Mean Values and Moments
1st, general, nth Moments
 
X  EX  

x f X x   dx or X  E  X    x  Pr X  x
x  
 
E g  X  
 gX  f

X x   dx or E g  X    g  X   Pr X  x
x  
 
 E X   x  E X    x
  f X  x   dx or X  Pr  X  x 
n n n n n n
X
 x  
Central Moments
     x  X 

X  X  n
E XX
n n
 f X  x   dx

     x  X 

X  X  n
E XX
n n
 Pr  X  x 
x  
Variance and Standard Deviation
     x  X 

  X X
2
 
2
E X X
2 2
 f X  x   dx

     x  X 

2  X X  
2
E XX
2 2
 Pr  X  x 
x  
ECE 6640 32
Random Signals
The Gaussian Random Variable
f X x  
1
 exp

 x X 2  
, for    x  
 
2    2 
2

where X is the mean and  is the variance
x

 v X 2  
FX x  
  exp   dv
1
2    2  2 
v    
Unit Normal
x
 u2 
x  

1
 exp    du
2  2 
u   
 x   1  x 
 x X  x X 
FX  x     or FX  x   1  
   

    
The Q-function is the complement of the normal function, :
(Appendix B)

 u2 
Q x  

1
 exp    du
2  2 
ECE 6640
ux  33
Random Processes
5. Random Processes
5.1. Introduction
Ensemble
5.2. Continuous and Discrete Random Processes
5.3. Deterministic and Nondeterministic Random Processes
5.4. Stationary and Nonstationary Random Processes
5.5. Ergodic and Nonergodic Random Processes

A Process for Determining Stationarity and Ergodicity
a) Find the mean and the 2nd moment based on the probability
b) Find the time sample mean and time sample 2nd moment based on time
averaging.
c) If the means or 2nd moments are functions of time … non-stationary
d) If the time average mean and moments are not equal to the probabilistic mean
and moments or if it is not stationary, then it is non ergodic.
From: George R. Cooper and Clare D. McGillem, Probabilistic Methods of Signal and
ECE 6640 System Analysis, 3rd ed.,Oxford University Press Inc., 1999. ISBN: 0-19-512354-9 34
Random Processes: Continuous,
Discrete and Mixed
Continuous and Discrete Random Processes
A continuous random process is one in which the random variables, such as X t1 , X t 2 ,  X t n  ,
can assume any value within the specified range of possible values. A more precise definition for a
continuous random process also requires that the cumulative distribution function be continuous.
A discrete random process is one in which the random variables, such as X t1 , X t 2 ,  X t n  ,
can assume any certain values (though possibly an infinite number of values). A more precise
definition for a discrete random process also requires that the cumulative distribution function
consist of numerous discontinuities or steps. Alternately, the probability density function is better
defined as a probability mass function … the pdf is composed of delta functions.
A mixed random process consists of both continuous and discrete components. The probability
distribution function consists of both continuous regions and steps. The pdf has both continuous
regions and delta functions.
ECE 6640 From: George R. Cooper and Clare D. McGillem, Probabilistic Methods of Signal and 35
System Analysis, 3rd ed.,Oxford University Press Inc., 1999. ISBN: 0-19-512354-9
Random Processes: Deterministic and
Nondeterministic
Deterministic and Nondeterministic Random Processes
A nondeterministic random process is one where future values of the ensemble cannot be predicted
from previously observed values.
A deterministic random process is one where one or more observed samples allow all future values
of the sample function to be predicted (or pre-determined). For these processes, a single random
variable may exist for the entire ensemble. Once it is determined (one or more measurements) the
sample function is known for all t.
Random Processes: Stationary and
Nonstationary (1)
Stationary and Nonstationary Random Processes
The probability density function for random variables in time as been discussed, but what is the
dependence of the density function on the value of time, t, when it is taken?
If all marginal and joint density functions of a process do not depend upon the choice of the time
origin, the process is said to be stationary (that is it doesn’t change with time). All the mean values
and moments are constants and not functions of time!
For nonstationary processes, the probability density functions change based on the time origin or in
time. For these processes, the mean values and moments are functions of time.
In general, we always attempt to deal with stationary processes … or approximate stationary by

assuming that the process probability distribution, means and moments do not change significantly
during the period of interest.
ECE 6640 37
Random Processes: Stationary and
Nonstationary (2)
Stationary and Nonstationary Random Processes
The requirement that all marginal and joint density functions be independent of the choice of time
origin is frequently more stringent (tighter) than is necessary for system analysis.
A more relaxed requirement is called stationary in the wide sense: where the mean value of any
random variable is independent of the choice of time, t, and that the correlation of two random
variables depends only upon the time difference between them.
That is
E X t   X   X and
E X t1   X t 2   E X 0   X t 2  t1   X 0   X    R XX   for   t 2  t1
You will typically deal with Wide-Sense Stationary Signals.
ECE 6640 38
Random Processes: Ergodicity
Ergodic and Nonergodic Random Processes
Ergodicity deals with the problem of determining the statistics of an ensemble based on
measurements from a sample function of the ensemble.
For ergodic processes, all the statistics can be determined from a single function of the process.
This may also be stated based on the time averages. For an ergodic process, the time averages
(expected values) equal the ensemble averages (expected values).
That is to say,
 T
 x n  f  x   dx  lim
X t   dt
1
Xn  n
T   2T
 T
Note that ergodicity cannot exist unless the process is stationary!
ECE 6640 System Analysis, 3rd ed.,Oxford University Press Inc., 1999. ISBN: 0-19-512354-9 39
Random Processes
The power spectral density is the Fourier Transform of the autocorrelation:

S XX w  R XX   
 EX t   X t    exp iw   d

For an ergodic process,
T
 XX    lim
 xt   xt     dt  xt   xt   
1
T   2T
T
  T 
 XX    E X t   X t    
 
xt   xt     dt   exp iw   d
 1
lim
T   2T 
  T 
T  
 XX    lim
 
xt   exp iwt   xt     exp iwt     d   dt
1 
T   2T
T   
T
 XX    lim
 xt   exp iwt X w  dt
1
T   2T
T
T
 XX    X w  lim
 xt   exp i wt  dt
1
T   2T
T
 XX    X w  X  w  X w
2
ECE 6640 40
From: George R. Cooper and Clare D. McGillem, Probabilistic Methods of Signal and System Analysis, 3rd ed.,Oxford
University Press Inc., 1999. ISBN: 0-19-512354-9
Binary Sequence, Low Bit Rate
Binary Autocorrelation and PSD
Bandwidth Consideration
• The first spectral null occurs are 1/T. Therefore one

measure of bandwidth could be the null.
• Are there others bandwidth measures?
– 3dB bandwidth
– 99% Power
– If it were a rectangle with Gx(0) given, how wide would it be
(Noise Equivalent Bandwidth)
– Etc.
Bandwidth Consideration
White Noise
Noise is inherently defined as a random process.
You may be familiar with “thermal” noise, based on the energy of an atom and the mean-free path
that it can travel.
As a random process, whenever “white noise” is measured, the values are uncorrelated with each
other, not matter how close together the samples are taken in time.
Further, we envision “white noise” as containing all spectral content, with no explicit peaks or
valleys in the power spectral density.
As a result, we define “White Noise” as

R XX    S 0   t 
S XX w   S 0
This is an approximation or simplification because the area of the power spectral density is infinite!
Band Limited White Noise
Thermal noise at the input of a receiver is defined in terms of kT, Boltzmann’s constant times
absolute temperature, in terms of Watts/Hz. Thus there is kT Watts of noise power in every Hz of
bandwidth.
For communications, this is equivalent to –174 dBm/Hz or –144 dBW/Hz.
For typical applications, we are interested in Band-Limited White Noise where

 S 0 f W
S XX w   
0 W  f
The equivalent noise power is then:
 
W
E X 2  R XX 0  
 S 0  dw  2  W  S 0
W
For communications, we use kTB.
How much noise power, in dBm, would I say that there is in a 1 MHz bandwidth?
dBkTB   dBkT   dBB   174  60  114 dBm
ECE 6640 46
White Noise in Comm.
• From the text
Noise as A Gaussian Random Process
A Gaussian Random Variable
f X x  
1
 exp
 
 x X 2 
, for    x  
2    2  2

 
x
 
 v X 2 
FX x  
  exp   dv
1
2    2  2 
v    
• What is so special about a Gaussian Distribution?

– Result of summing a large number of random variables
– Linear systems produce Gaussian Outputs
– Well know/studied characteristics
– Used to define the characteristics of numerous natural, real-world signals
ECE 6640 48
Linear Systems
Linear transformation of signals:
y t   ht   xt 
Ys   Hs   Xs 
Convolution Integrals

y t  
 xt     h   d
0
or
t
y t  
 ht     x   d

where for physical realizability and stability constraints we require

ht   0 for t  0

ECE 6640  ht   dt   49


Transfer Function
Hf   Hf   exp j  f 
 ImHf 
f   tan 1  
 ReHf 
• For linear systems: A sinusoidal input results in sinusoidal

output modified in magnitude and phase.
x t   A  cos2  f 0  t 
yt   h t   x t 
yt   A  Hf 0   cos2  f 0  t  f 0 
ECE 6640 50
Filtering a Random Process
• The PSD of a filtered response is

    
   

 0
   
RYY    E  xt  1   h1   d1    xt    2   h2   d2 

 0 
 
 
 
RYY    d1  d2  h1   h2   R XX   1  2 
 
0 0 
   
 
  
SYY w  RYY    d1  d2  h1   h2   
 
R XX   1  2   exp iw   d 


0 0   
SYY w   R YY   SXX w   Hw   H w 
SYY w  RYY    S XX w  H w

2
ECE 6640 51
Distortionless Transmission and
the Ideal Filter
• To receive a signal without distortion, only changes in the
magnitude and/or a time delay is allowed.
yt   K  x t  t 0 
Yf   K  Xf   exp 2  f  t 0 
• The transfer function is

Hf   K  exp 2  f  t 0 
• A constant gain with a linear phase

Hf   K f   2  f  t 0
ECE 6640 52
Ideal Filter (1)
• For no distortion, the ideal filter should have the following

properties:
Hf   Hf   exp j  f 
1, for f  f u 2  f  t 0 , for f  f u

Hf    f   
0, for f  f u arbitrary, for f  f u
• The impulse response is

fu
h t    1 exp j2  f  t 0  exp j2  f  t  df

f u
fu
h t    exp j2  f  t  t 0  df

f u
ECE 6640 53
Ideal Filter (2)
• Continuing
fu
h t    exp j2  f  t  t 0  df

f u
exp j2  f  t  t 0 
fu
h t  
j2  t  t 0  f u
exp j2  f u  t  t 0  exp j2   f u   t  t 0 

h t   
j2  t  t 0  j2  t  t 0 
2  sin2  f u  t  t 0 
h t  
2   t  t 0 
h t   2  f u  sinc2  f u  t  t 0 
• The sinc function

– A non-causal filter
Ideal Filters in the Freq. Domain
Realizable Filters, RC Network
1st order
Butterworth
Filter
White Noise in an RC Filter
• The noise PSD has been modified

• The autocorrelation is spread in time
Signal Filtering in the Real World
Signal Filtering in the Real World (2)
Bandwidth Considerations, Easy
Bandwidth Considerations, Harder
• If the spectrum extends to infinity, where do you assume

that it can be cut off?
Bandwidth Considerations
• Note 1 that as soon as time is limited, the signal has been
multiplied by a rect function in the time domain.
– A rect in the time domain creates an infinite sinc convolution in the
frequency domain!
• Note 2 that a bandlimited frequency domain signal can be
generated by multiplying by a rect function in the
frequency domain.
– A rect in the frequency domain results in a non-causal, infinite
time convolution in the time domain!
• For mathematicians, a real signal can not be both time

limited and frequency band limited?!
ECE 6640 62
Bandwidths that are Used
Bandwidth Definitions
(a) Half-power bandwidth. This is the interval between frequencies at
which Gx(f ) has dropped to half-power, or 3 dB below the peak value.
(b) Equivalent rectangular or noise equivalent bandwidth. The noise
equivalent bandwidth was originally conceived to permit rapid
computation of output noise power from an amplifier with a wideband
noise input; the concept can similarly be applied to a signal bandwidth.
The noise equivalent bandwidth WN of a signal is defined by the
relationship WN = Px/Gx(fc), where Px is the total signal power over
all frequencies and Gx(fc) is the value of Gx(f ) at the band center
(assumed to be the maximum value over all frequencies).
(c) Null-to-null bandwidth. The most popular measure of bandwidth for
digital communications is the width of the main spectral lobe, where
most of the signal power is contained. This criterion lacks complete
generality since some modulation formats lack well-defined lobes.
ECE 6640 64
Bandwidth Definitions (2)
(d) Fractional power containment bandwidth. This bandwidth criterion
has been adopted by the Federal Communications Commission (FCC
Rules and Regulations Section 2.202) and states that the occupied
bandwidth is the band that leaves exactly 0.5% of the signal power
above the upper band limit and exactly 0.5% of the signal power below
the lower band limit. Thus 99% of the signal power is inside the
occupied band.
(e) Bounded power spectral density. A popular method of specifying
bandwidth is to state that everywhere outside the specified band, Gx(f )
must have fallen at least to a certain stated level below that found at
the band center. Typical attenuation levels might be 35 or 50 dB.
(f) Absolute bandwidth. This is the interval between frequencies, outside
of which the spectrum is zero. This is a useful abstraction. However,
for all realizable waveforms, the absolute bandwidth is infinite.
ECE 6640 65
Spectrum and Time Domain of a
Band-limited Bandpass Signal
ECE 6640 Notes and figures are based on or taken from materials in the course textbook: 66
Bernard Sklar, Digital Communications, Fundamentals and Applications,
Summary
• Communication must consider a number of aspects
– Time and Frequency Domain Signals
– Discrete and Continuous Time Signal Constructs
– Deterministic and Random Signal Properties
– Models of Signal Propagation
• Simple time and amplitude changes
• Complex channel impairments
– Models of Other Signals in the Environment
• Noise (white, Gaussian, or more complex)
• Interference
• Multipath
• To successfully model and analyze modern communication
systems, there is a lot of prerequisite knowledge required.
ECE 6640 67
ECE 6640

Assistant Professor
Chapter 2
2. Formatting and Baseband Modulation.
1. Baseband Systems.
2. Formatting Textual Data (Character Coding).
3. Messages, Characters, and Symbols.
4. Formatting Analog Information.
5. Sources of Corruption.
6. Pulse Code Modulation.
7. Uniform and Nonuniform Quantization.
8. Baseband Modulation.
9. Correlative Coding.
ECE 6640 2
Formatting
• Insure that the message is compatible with digital

processing
• Transmit formatting is where the source information is
translated into digital symbols
• When data compression is also employed, the process is
called source coding. (see Chap. 13)
ECE 6640 5
Baseband Signaling
• Generation of the baseband waveform from the digital

symbols provide by formatting or source coding.
• This could take the form of pulse modulation or pulse code
modulation.
• The baseband signal may be sent using a wired connection
or network to a receiver.
ECE 6640 6
Formatting and Transmission
Digital info.
Textual Format
source info.
Pulse
Analog Transmit
Sample Quantize Encode modulate
info.
Pulse
Bit stream waveforms Channel
Format
Analog
info. Low-pass
Decode Demodulate/
filter Receive
Textual Detect
sink
info.
Digital info.
ECE 6640 7
Textual Data
• 5-bit coding – Baudot: 32 characters, alphabet plus 6

• 7-bit coding – ASCII: American Standard Code for
Information Interchange
– Originally for telegraph therefore extra fields
• 8-bit coding – EBCDIC: Extended Binary Coded Decimal
Interchange Code
– IBM system
• 16-bit coding – Unicode
• Code may be sent serially with start, parity and stop bits
• Code may be structures as words/symbols
ECE 6640 8
Data Format for Asynchronous
Data Communication
• Data is transmitted character by character bit-serially.
• A character consists of
– one start bit (0-level)
– 7 to 8 data bits (often, an ASCII character plus a parity bit)
– an optional parity bit
– one, or one and a half, or two stop bits (1-level)
– least significant bit is transmitted first
– most significant bit is transmitted last
The transmission is a burst occurring at an unknown time but

ECE 6640 9
with known bit periods.
The EIA-232E Electrical
Specifications
Electrical signal on a pair of wires … signal and ground.

ECE 6640 10
Messages, Characters and Symbols
• Message is encoded into a sequence of bits

– The bit stream may be a basedband signal
– ASCII can generate a continuous bit stream if “idle characters are
“1’s”
• Grouping of k-bits can be formed into symbols

– M-ary systems use symbols sets where M=2^k
– For k=1, the bit rate and symbol rate are the same
– Defined waveforms represent each of the symbols
• Therefore a message based bit streams can be represented

as a string of Octal or Hex characters in sequences!
ECE 6640 11
– See Text Figure 2.5
A Review of Sampling Theorem
• We use digital signal processing to transmit and receive all forms of
communications.
• Digital communications inherently describes bit values and symbol

values that “conceptually” exist for a defined period of time and the
“instantaneously” switch to another value.
– The transmitted signals can not physically do this!
– Transmitted signals must exist at defined frequencies and within defined
bandwidths … limited bandwidths often start at baseband.
• We may not discuss or simulate all the “real world” effects even in this
class.
ECE 6640 12
Analog to Digital Conversion
• Sampling 1 ws
fs  
– Sampling Theorem T 2
• Nyquist rate  fs>=2*fmax
1
– Sample Rate T
– Sample Period
fs
• Impulse Sampling Function

x  t    t  n  T  s
k  

x s t   x  t   x   t    x t  t  n  T 
s
n  

x s t    x n  T  t  n  T 
s s
ECE 6640 n   13
Fourier Domain – Replicated Spectra
1 
X  f     f  n  f s 
Ts n  
X s f   Xf   X  f 

X s f    X     X f     d


1 
X s f        n  f s   Xf     d
T
  s k  

1 
X s f        n  f s   Xf     d
Ts k   
1 
ECE 6640 X s f     Xf  n  f s  14
Ts k  
Fourier Domain
• Spectral replication at steps of fs

• Appears as the convolution of the original spectrum by a
comb waveform spaced as fs
– If the Nyquist rate is not maintained, the convolved elements will
overlap and become distorted
• See Figure 2.6 on p. 64
• Note: Signals are not typically band limited; therefore,

there will be some aliasing whenever sampling is
performed
ECE 6640 15
Sampling Pulses and Filters
• While Nyquist Theory and Impulse Sampling is

mathematically wonderful ….
– Sampling rates above Nyquist are more practical (Fig. 2.7)
• 2.2 fmax for audio example (20 kHz vs. 44.1 ksps CD rate)
– Impulses must be approximated by signals with real duration and
magnitude (Section 2.4.1.2 Natural Sampling and Fig. 2.8)
• Sample by infinite sequence of rects
• Math equivalent of convolving sampling impulses with rect in time
In frequency, convolve infinite replicas with sinc … amp mod
impulses
• When this sampling signal is used (mult. in time, conv. In freq.)
you get Fig. 2.8
1 
X s  f     ck  X  f  n  f s 
ECE 6640 Ts k   16
Sample and Hold, Zero Order Hold
• Typical ADCs use a “sample and hold” prior to the ADC

• Sampling is typically an integration of the signal for a
fixed sampling period
• Hold is to insure the ADC has a stable signal for a defined
period of time (conversions time)
  
x sp t   pt   x t   x  t   pt     x t   t  n  Ts 
 n   
 t 
pt   rect 
 Ts 
ECE 6640 17
ZOF Spectral Domain
1  
X sp f   Pf      X f  n  f s 
 Ts k   
1  
X sp f   Ts  sincf  Ts      X f  n  f s 
 Ts k   

X sp f   sincf  Ts    Xf  n  f 
s
k  
• The spectrum is shaped by the sinc function.

– Note that if spectral analysis is being performed, an inverse sinc
weighting should be applied to “correct” the output.
ECE 6640 18
Filter and Aliasing
• If a signal is “under-sampled” the output will have spectral

content that is not desired.
– “Engineering Nyquist” 2.2 x fmax
– With digital post filtering, sample at 4 x fmax (or 4.4) and then use
a half-band filter decimator (may be less expensive)
• If additional digital filtering will be employed, aliased
regions of the spectrum may be digitally removed.
– This allows the transition bands to overlap and the stopband to be
placed nearer to the passband edge.
ECE 6640 19
Filter Terminology
• Passband
– Frequencies where signal is meant to
pass
• Stopband
– Frequencies where some defined
level of attenuation is desired
BWPB
• Transition-band
BWSB
– The transitions frequencies between
the passband and the stopband
• Filter Shape Factor
– The ratio of the stopband bandwidth
to the passband bandwidth
BWSB
SF  See FilterNotes and FIR_Filter_DSPNotes
BWPB
or
ECE 6640 MRSP Chap 4 Nyquist/Raised Cosine Filter 20
Reducing the Sample Rate
If additional digital filtering

will be employed, aliased
regions of the spectrum may be
digitally removed.
This allows the transition
bands to overlap and the
stopband to be placed
Dynamic
nearer to the passband edge.
Range
ECE 6640 21
Oversampling
• Without Oversampling
– High performance LPF
– Nyquist Sampling still includes some aliased components
• With Oversampling
– Lower performance LPF
– Aliased components can be significantly reduced
– High performance digital filter may be employed
– Identical or similar data rate can be achieved
ECE 6640 22
2.5 Sources of Corruption
• Quantization Noise
– Saturation
– Timing Jitter
– See Analog Devices – Radio 101

– http://www.analog.com/static/imported-files/tech_articles/480501640radio101.pdf
• Intersymbol Interference (ISI)
ECE 6640 23
Quantization Noise
• Round-off Error
– +/- one half of the LSB q
2
– Uniform error distribution about the     pe de
2 2
e
quantized value q
2
– Error mean  0 q
1 2 2
– Error Variance  q^2/12    e  de
q q
2
q
• Truncation Error 1 e 3 2
1  q3  q3 
      
q 3 q  24 24 
– 0 to +1 LSB q
2
– Uniform error distribution from one q2


quantized value to the next 12
– Error mean  1/2
– Error Variance  q^2/12
ECE 6640 24
Quantization Levels
• The levels defined for a typical L-level (2^k=L) ADC

Quantized
qL
rmsVp  
values
ECE 6640 25
Quantized Peak SNR
• For an L level quantized system, letting power be the

square of one half the rms value of a maximum sine wave
2
  
rmsVp   
2 q L
 
rmsVp 
 2 2  2
• The estimated signal to noise ratio is SNR q 

2
2
 qL 
2 2  q 2  L2 12 3 2
SNR q      2  L
q2 8 q 2
12
For 8-bits or 256 levels SNR q  1.5  256 2  49.9 dB
ECE 6640 Nominally 6 dB per bit 26
Intersymbol Interference
• Web
– InterSymbol Interference (ISI)
• http://en.wikipedia.org/wiki/Intersymbol_interference
– Nyquist ISI Criterion
• http://en.wikipedia.org/wiki/Nyquist_ISI_criterion
– Inter Symbol Interference (ISI) and Raised cosine filtering
• http://complextoreal.com/wp-content/uploads/2013/01/isi.pdf
• From C. Langton “Complex to Real” web site
• Other classes (ECE6950 MRSP Chap. 4)

– A raised cosine window/filter is a form of Nyquist filtering
• As combined transmitting and receiving filters, each uses a square-
root raised cosine filter.
ECE 6640 27
2.6 Pulse Code Modulation
• Use the digital word/symbol generated for each character

or ADC value
• Note that the more information or accuracy per symbol, a
higher bit rate is required to maintain the symbol rate
– Fs demands a fixed, constant communication rate
– Text may be sent at any rate that is acceptable (non real-time)
• This can also be referred to as Amplitude Shift Keying

– http://en.wikipedia.org/wiki/Amplitude-shift_keying
• Note that ADC values if sent as pulse would be called

PAM or pulse amplitude modulation
ECE 6640 28
– PAM may be discrete or quantized
2.7 Uniform vs. Nonuniform
• Uniform (linear) quantizing:

– No assumption about amplitude statistics and
correlation properties of the input.
– Robust to small changes in input statistic by not finely
tuned to a specific set of input parameters
– Simply implemented
• Non-uniform quantizing:
– Using the input statistics to tune quantizer parameters
– Larger SNR than uniform quantizing with same number
of levels
– Non-uniform intervals in the dynamic range with same
ECE 6640 quantization noise variance 29
Uniform vs. Nonuniform (2)
• Application of linear quantizer:

– Signal processing, graphic and display
applications, process control applications
• Application of non-uniform quantizer:

– Commonly used for speech
– u-law in US, A-law in Europe
ECE 6640 30
Nonuniform Quantization
• When some portions of the voltage range are not often

used, additional emphasis can be given to those that are.
• μ-law algorithm (North America μ=255)

– http://en.wikipedia.org/wiki/Mu-law
• A-law algorithm (Stnadard Value A=87.6)

– http://en.wikipedia.org/wiki/A-law_algorithm
ECE 6640 31
Baseband Signaling
• Generation of the baseband waveform from the digital

symbols provide by formatting or source coding.
• This could take the form of pulse modulation or pulse code
modulation.
• The baseband signal may be sent using a wired connection
or network to a receiver.
ECE 6640 32
PCM Transmission
• Pulse code modulation (PCM) is used when a binary data

stream is to be sent
• In PCM the binary sequence is used to define logical signal

levels for transmission.
– A logical level may map to bits (e.g. 0-High, 1-Low)
– A bit value may define whether a level changes or not
Mark : change whenever the bit is a one
Space: change whenever the bit is a zero
– Period half-cycles can take on various structures based on a bit
value or the sequence of bits
• See Figure 2.22 on p. 87
ECE 6640 33
PCM Common Waveform Types
• Marks (1’s) and Spaces (0’s)
• Non-return-to-zero (NRZ) – Level, Mark, Space

• Return-to-zero (RZ) – unipolar, bipolar, AMI (alternate
mark inversion)
• Manchester – biphase level, biphase mark, biphase space
ECE 6640 34
PCM Types
NRZ Bipolar RZ Manchester
AMI-Bipolar Encoding
Biphase Mark Code
(Alternate Mark Inversion)
ECE 6640 35
PCM Type Selection
• Spectral characteristics
(power spectral density and bandwidth efficiency)
• Bit synchronization capability
• Error detection capability
• Interference and noise immunity
• Implementation cost and complexity
ECE 6640 36
Spectral Attributes of PCM
ECE 6640 37
M-ary Pulse-Modulation Waveforms
• M-ary modulation is used when symbol data stream is to

be sent
• M-ary waveform include:
– PAM: Pulse-Amplitude Modulation
– PPM: Pulse-Position Modulation
– PDM: Pulse-Duration Modulation or PWM: Pulse-Width
Modulation
– Multiple “level” can be transmitted as one symbol

• Other M-ary waveforms
– QAM: Quadrature-Amplitude Modulation
ECE 6640 38
Correlative Codes
• Web site: “Complex technology made real • Complex

communications technology made easy” or
“Complex to Real” by Charan Langton
– http://complextoreal.com/
• Tutorial 16 – Partial Response signaling and Quadrature
Partial Response (QPR) modulation
– http://complextoreal.com/wp-content/uploads/2013/01/qpr.pdf
ECE 6640 39
ECE 6640

Assistant Professor
Chapter 3
3. Baseband Demodulation/Detection.
1. Signals and Noise.
2. Detection of Binary Signals in Gaussian Noise.
3. Intersymbol Interference.
4. Equalization..
ECE 6640 2
Chapter Goals
• Detection of Binary Signals plus Gaussian Noise

• The decision process
• Define Intersymbol Interference
• Error Performance Degradation
• Equalization Techniques
ECE 6640 4
Demodulation and Detection
• Focus on Symbols, Samples, and Detection

• In the presence of Gaussian Noise and Channel Effects
ECE 6640 5
Demodulation and detection
• Major sources of errors:

– Signal Path Loss
• Friis equation relates the received signal power to the transmitted
power, antenna gains, and distance that the signal travels in free space
– Thermal noise (AWGN)
• disturbs the signal in an additive fashion (Additive)
• has flat spectral density for all frequencies of interest (White)
• is modeled by Gaussian random process (Gaussian Noise)
– Inter-Symbol Interference (ISI)
• Due to the filtering effect of transmitter, channel and receiver,
symbols are “smeared”.
• Time spreading effects cause symbols to “overlap”
– Total symbol “length” may easily be 3+ symbol periods!
6
Receiver Block Diagram
• Receive the transmitted symbol plus noise

– Symbol filter by channel
• Frequency down-conversion to baseband
– Receiver filtering and equalization (if needed) applied
• Symbol filter
– Matched filtering with Nyquist shaping for ISI
– Optimize the pre-detected signal prior to sampling
• Optimal Time Sampling
ECE 6640 – Peak filter response time 7
Review Slides from ECE5640
• Chapter 9: Noise
• Chapter 10: Noise in analog modulated signals
• Chapter 11: Baseband Digital Transmission
ECE 6640 8
Noise Approximation
• Uniform Noise Spectral Density
– Resistor description (Thevenin Model)
G vv f   2  R    T
• Available Power from the “noise source”
– Source output power into a matched load
v sout 
R
vs Psout
v 
 sout
2 2
v  1 v
 s    s
2
2R R  2  R 4R
G vv f  2  R    T   T N 0
G ss f     
4R 4R 2 2
R ss     
N0
2 9
System Noise
• Since the noise power spectrum is uniform, a systems
noise power is the product of the noise power and the
integral of the filter power.
SNN f   Hf   SN 0 N 0 f    Hf 

2 N0 2
 
R NN 0    Hf   df  N 0   Hf   df
N0 2 2
2  0
10
Noise Equivalent Bandwidth
• If we want the total noise power after the filter, we can
integrate the PSD for all frequencies or use the Filtered
noise autocorrelation function at zero.
– Both of these approaches may be difficult
– Could we great a more simple “noise equivalent bandwidth for
filters” that is rectangular?
11
 
R NN 0  0   Hf   df  0  2   Hf   df
N 2 N 2
2  2 0
• When filtering, it is convenient to think of band-limited

noise, where the filter is a rect function with bandwidth
BEQN
 
H rect _ mod el  f   GainDC _ Power  f 
 rect Gain DC _ Power  H0 
2
 2 B 
 EQN 
 
 H  f   df   H  f rect _ mod el  df  GainDC _ Power  BEQN  H 0   BEQN

2 2 2
0 0

 Hf   df
2
BEQN  0
H0
2
12
• Low pass filter
coherent _ gain  H0  Gain DC _ Power  H0 
2
 Hf   df
2
B EQN  0
H0 
2
PN  R NN 0    2  H0   BEQN  H0   N 0  BEQN

N0 2 2
• For a unity gain filter

– assumed when computing receiver input noise power

B EQN   Hf   df
2
PN  R NN 0  
N0
 2  BEQN  N 0  BEQN
2 13
Model of Received Signal
with Noise
© 2010 The McGraw-Hill Companies

Signal Plus Noise
Copyright © The McGraw-Hill Companies, Inc. Permission required for reproduction or display.
Analog baseband transmission system with noise: Figure 9.4-2
 At  
Pr eDt     cos2  f c  t  t   nt   ht 
x t   At   cos2  f c  t  t   L 
At  At 
xR t    cos2  f c  t  t  st    cos2  f c  t  t   nt 
L L
• Additive Gaussian White Noise 15

Signal-to-Noise Ratio
• Comparing the desired signal power to the undesired noise
power.
y t   xc t   nt 
• To compare signal and noise power, we must assume a
filtering operations
Linear Nonlinear Atten-

Filtering Distortion uation
xc t 
nt 
y t   xc t   nt 
Noise
Transmitting Receiving
Antenna Antenna
RF Communication Channel
16
y D t   xR t   nt  ht 
• Equivalent receiver input signal and noise (ER)

y R t   xER t   nER t 
• Equivalent destination signal and noise (D) or pre-
demodulation (PreD)
yPr eD t   xPr eD t   nPr eD t  17

• Equivalent receiver input SNR (ER)
SNRR 

E xER t 
2

 
E xER t 
2

 SR

E nER t 
2

N 0  BEQN N 0  BEQN
 can be used to
• Equivalent destination SNR represent receiver
 
noise figure
E xPr eD t 
2 contributions
SD SD
SNRR   

E nPr eD t 
2

N D   N 0  BEQN 18
Increase in SNR with filtering
• If a filter matched to the input signal is applied, the noise
power would be reduced to the smallest equivalent noise
bandwidth that is allowed.
– Filter to minimize noise power
– Importance of the IF filter in a super-het receiver!
• Front-end filtering goals – a dilemna

– Minimize signal power loss (wider bandwidth)
– Minimize filter equivalent noise bandwidth
(narrower bandwidths)
– A trade-off must be made!
19
Typical Transmission Requirements
Signal Type Freq. Range SNR (dB)
Intelligible Voice 500 Hz to 2 kHz 5-10
Telephone Quality 200 Hz to 3.2 kHz 25-35
AM Broadcast Audio 100 Hz to 5 kHz 40-50
High-fidelity Audio 20 Hz to 20 kHz 55-65
Video 60 Hz to 4.2 MHz 45-55
Spectrum Analyzer 100 kHz-1.8 GHz 65-75
20
CW Communication with Noise
Model of a CW communication system with noise: Figure 10.1-1
At 
x t   At   cos2  f c  t  t  vt    cos2  f c  t  t   n t 
L
A t   At  
x c t    cos2  f c  t  t  PreDt     cos2  f c  t   t   nt   hR t 
L  L 
21
Signal and Noise Power
• What are the signal and noise powers at the receiver?
T
1 2
Ps  lim   x t   x t   dt

T  T
T
2

Pn  E n t 
2

• What is the receiver input power
T T
   
2
1 2
Pv  lim   E vt   vt   dt  lim   E x c t   n t   x c t   n t   dt
1  
T  T T  T
T T
2 2
22
Receiver Signal plus Noise Power
• What is the receiver input power
T T
   
2
1 2
Pv  lim   E vt   vt   dt  lim   E x c t   n t   x c t   n t   dt
1  
T  T T  T
T T
2 2
T
1 2
 
Pv  lim   E x c t   2  x c t   n t   n t   dt
T  T
2 2
T
2
T
1 2
  
Pv  lim   x c t   2  x c t   En t   E n t   dt
T  T
2 2
T
2
 T T T

1 
  
2 2 2
Pv  lim    xc t   dt   2  xc t   0  dt   E nt   dt 
2 2
T  T
T 2 T
2
T
2

T
1 2
 
Pv  lim   x c t   dt  E n t   Ps  Pn
T  T
2 2
T
2
23
Signal-to-Noise Ratio (SNR)
• The SNR is a measure of the signal power to the noise
power at a point in the receiver.
– Typically described in dB
– The above computation was performed at the input
Ps
SNR 
Pn
• Matlab SNR example: SNR_AM_Example.m
– Pre-D “AM” SNR based in filter Beqn
• Effective BEQN RF due to sampling spectrum (B=Fis/2)
• AM signal power based on carrier plus signal
24
• Since the noise power spectrum is uniform, a systems
average noise power is the product of the noise power and
the integral of the filter power.
SNN f   Hf   SN 0 N 0 f    Hf 

2 N0 2
 
R NN 0     Hf   df  N 0   Hf   df
N0 2 2
2  0
25
 
R NN 0   0   Hf   df  0  2   Hf   df
N 2 N 2
2  2 0
• When filtering, it is convenient to think of band-limited

noise, where the filter is a rect function with bandwidth
BEQN
 
H rect _ mod el  f   GainDC _ Power Gain DC _ Power  H0 
f
 rect   2
 2 B 
 EQN 
 
 Hf   df   Hf rect _ mod el  df  Gain DC _ Power  BEQN  H0  BEQN

2 2 2
0 0

 H  f   df
2
BEQN  0
H 0 
2
26
• Low pass filter
coherent _ gain  H0  Gain DC _ Power  H0 
2
 Hf   df
2
BEQN  0
H0 
2
PN  RNN 0   2  H 0  BEQN  H 0   N 0  BEQN

N0 2 2
• For a unity gain filter

– assumed when computing receiver input noise power

B EQN   Hf   df
2
PN  R NN 0 
N0
 2  BEQN  N 0  BEQN
2 27
Filtering
• What happens if the receiver input is filtered?
v f t   x c t   n t  h1 t   h 2 t 
v f t   x c t   h 1 t   h 2 t   n t   h 1 t   h 2 t 
• What effect does the filter have on the signal?

– None or slight band edge de-emphasis, if and only if the filter is
“wider” than the signal bandwidth
– Now you know why a 3dB bandwidth isn’t that useful,
(3dB1/2 power point)!
Ps
SNR 
Pn
28
Filtering
• What effect does the filter have on the noise?
– Normally you would expect for two filters
Ps
PN _ Post _ Filter1  N 0  BEQN _ Filter1 SNRPost _ Filter1 
N 0  BEQN _ Filter1
PN _ post _ filter 2  N 0  BEQN _ Filter 2 Ps

SNRPost _ Filter 2 
N 0  BEQN _ Filter 2
– Assume that the filters follow each other and that the first filter is
narrower than the second filter
PN _ Post _ Filter 2  N 0  min BEQN _ Filter1 , BEQN _ Filter 2 

 N 0  BEQN _ Filter1
Ps
SNRPost _ Filter 2 
N 0  BEQN _ Filter1 29
Filters Provide SNR “Gain”
• If filter 2 Beq < filter 1 Beq:
Ps  BEQN _ Filter1 
SNR Post _ Filter 2   SNR Post _ Filter1   
N 0  BEQN _ Filter 2 B 
 EQN _ Filter 2 
• You expect the IF filter to be smaller than the front-end RF

or “pre-filtering” performed
 B EQN _ Filter1 
Gain Filter   
B 
 EQN _ Filter 2 
– Think about kTB at different bandwidths and you will derive the
same “gain”
– In typical receivers, the IF filter sets the
Pre-Demodulation Bandwidth
30
Bandpass Noise Processing
cos2  f c  t 
Band Pass LowPass
Filter Filter
Bandpass filter bandwidth

may not be centered on fc
• an alpha offset
BT N0 BT
2
 fc fc
 f c    BT f c    BT
• What happens after mixing and lowpass filtering?

– assume LPF passes the entire baseband. 31
Quadrature Noise (1)
• Question: Is Quadrature noise a different power than “real”
baseband noise
• Noise in a quadrature process
nt   ni t   cos2  fc  t   nq t   sin2  fc  t 
• Noise power is related as
 2
  2
 
E n t   E n i t   E n q t  
2
 N0
2
• What about ?
  
E nt   E ni t   cos2  fc  t     nq t   sin2  fc  t   
2 2

32
Quadrature Noise (2)
• Noise in a quadrature process
  
E n t   E n i t   cos2  f c  t    n q t   sin 2  f c  t  
2 2

n t 2  cos 2 2  f  t   
 
 
i c
E n t   E  2  n i t   n q t   cos2  f c  t    sin 2  f c  t  

2
 
 n q t   sin 2  f c  t  
2

2
   2 1  2 1 
E nt   E ni t     cos2  2 f c  t  2     nq t     cos2  2 f c  t  2   
2
2  2 

   2 1
2
 

2 1
2
  
E nt   E ni t     E cos2  2 f c  t  2     E nq t     E cos2  2 f c  t  2   
2
 
E nt  
2 1
2
 2

1
2

 E ni t    E nq t   0
2 N
2

33
Mixing Noise (1)
• Think of the two noise bands as
1. The band of interest
2. The image band
nt   cos2  f LO  t   hIF t   n1i t   cos2   f LO  f IF   t   n1q t   sin 2   f LO  f IF   t  cos2  f LO  t   hIF t 
 n2i t   cos2   f LO  f IF   t   n2 q t   sin 2   f LO  f IF   t  cos2  f LO  t   hIF t 
1 n1i t   cos2  f IF  t   n1q t   sin 2   f IF   t  

nt   cos2  f LO  t   hIF t      hIF t 
2 n1i t   cos2  2 f LO  f IF   t   n1q t   sin 2  2 f LO  f IF   t 
1 n2i t   cos2   f IF   t   n2 q t   sin 2   f IF   t  

    hIF t 
2 n2i t   cos2  2 f LO  f IF   t   n2 q t   sin 2  2 f LO  f IF   t 
nt   cos2  f LO  t   hIF t    n1i t   cos2  f IF  t   n1q t   sin 2  f IF  t  hIF t 
1
2
  n2i t   cos2  f IF  t   n2 q t   sin 2  f IF  t  hIF t 
1
2 34
Mixing Noise (2)
• Defining the equivalent IF noise
nt   cos2  f LO  t   hIF t    n1i t   cos2  f IF  t   n1q t   sin 2  f IF  t  hIF t 
1
2
  n2i t   cos2  f IF  t   n2 q t   sin 2  f IF  t  hIF t 
1
2
 1  

 2 1i  n t  
1
 n2i t    cos 2  f IF  t   
 2  
nt   cos2  f LO  t   hIF t      hIF t 
  1  n t   1  n t   sin 2  f  t  
 2 1i 2
2i

IF


E ni t 
2
   
  E n1i t    E n2i t   0
1 2 1 2 N
 
E nq t  
2
 1
2
 2

1
2

 E n1q t    E n2 q t   0
2 N
2

2 2 2
• But this is the same as quadrature noise
nt   hIF t   ni t   cos2  fIF  t   nq t   sin2  fIF  t  hIF t 
• Mixing doesn’t change the noise power 35

Mixing Noise to Baseband
• What if we split bandpass noise into two
distinct noise bands, BT/2 above and below
the carrier/IF? B
for W  T
 
nt   n1i t   cos2   f c  f B   t    n1q t   sin 2   f c  f B   t    h1 t 
2
 n2i t   cos2   f c  f B   t    n2 q t   sin 2   f c  f B   t   h2 t 
 
E nt 
2
carrier 
N0
2
 2  BT  N 0  BT  2  N 0 W

E n1 t 
2

AboveC

 E n2 t 
2

BelowC
N0
2
  2 
BT
2
 N 0 
BT
2
 N 0 W
  
E nt 
2
  
 E n1 t   E n2 t   N 0  BT  2  N 0 W
2 2
• Noise bands get added … 36

Mixing Noise to Baseband
BT BT
N0 (a) General case;
2
(b) symmetric-sideband case;
 fc fc
 f c    BT f c    BT (c) suppressed-sideband case: Figure 10.1-3
Not Desired 0    0.5
Mix to Baseband   0.5

W=½ BT
N0BT=2N0W
Mix to IF
  0 or   1
BT
N0BT 37
Why did we do these derivations?
• The past derivations were all about mixing and filtering.
– Quadrature noise is the noise that gets mixed to the intermediate. The
bandwidth and noise power do not change
 
E nt  
2 N0
2
 2  BT  N 0  BT  2  N 0  W
– Quadrature noise is the noise that gets mixed to baseband. The

bandwidth is halved and the noise power is doubled the LPF
bandwidth standard noise power.
BT
W  
E nt   N 0  BT  2  N 0 W
2
38
Complex Noise
• Noise in a complex process
nt   ni t   j  nq t 
  
E n t   E n i t   j  n q t  n i t   j  n q t 
2
 H
E n t    E n t   j  n t  n t   j  n t 

2
i q i q
E n t    E n t   j  n t   n t   j  n t   n t   n t  
2 2 2
i i q q i q
     
E n t   E n i t   E n q t 
2 2 2
  E n t    E n t   
N
E n t 
N 2 2 0
 0
2
i q
2 4
MATLAB : n  randn ( m , n )
MATLAB : n  randn ( m , n )  i  randn ( m , n )  sqrt 2 
39
Noise Envelope and Phase (1)
• Noise as a magnitude and phase
n t   A n t   cos2  f c  t   n t 
n i  A n  cos n  n q  A n  sin  n 
• The magnitude is a Rayleigh distribution

– Mean and moment
 An 2 
p A A n   n   u A n 
A
 exp  
NR  2  NR 
n
EA n  
  NR
2
 
E An  2  NR
2
40
Noise Envelope and Phase (2)
• Probability of An exceeding a
P An  a   exp  a 
2
 2  N R 
• Phase Distribution
  NR
E  An  
2
p  n  
1
0   n  2
2
for
 
E An  2  N R
2
• Noise Power
    
E nt   E An t   E cos2  f c  t   n t 
2 2 2

E nt    2  N     N 
2 1 N 0
R R
2 2
41
Noise Characteristics
• The noise power does not change based on the
representation, the center frequency, or due to mixing.
• The noise power will change when the bandwidth is

further limited in some way!
42
CW Communication with Noise
Model of a CW communication system with noise: Figure 10.1-1
At 
x t   At   cos2  f c  t  t  vt    cos2  f c  t  t   n t 
L
A t   At  
x c t    cos2  f c  t  t  Pr eDt     cos2  f c  t   t   nt   hR t 
L  L 
43
Chapter 11
• Baseband Transmission of PAM Symbols

• PAM Symbol Autocorrelation and Power Spectral Density
• Symbol Detection in Noise
ECE 6640 44
Digital Pulse-Amplitude Modulation
(PAM)
• Also referred to as pulse-code modulation (PCM)
• The amplitude of pulse take on discrete number of
waveforms and/or levels within a pulse period T.
x t    a k  pt  kT 
k
• p(t) takes on many different forms, a rect for example
1 0tT
pt   
0 else
x mT     a k  pmT    kT   a m  p  a m , for 0    T

k
45
Digital Signaling Rate
• For symbols of period T,
the symbol rate is 1/T=R
• The rate may be in bits-per-second when bits are sent. A

bps rate is usually computed and defined.
• The rate may be in symbols-per-second when symbols are
sent. When there are a defined number of bits-per-symbol,
the rate may be defined in bits-per-second.
– If parity or other non-data bits are sent, the messaging rate and the
signaling rate may differ.
46
Transmission
(a) Baseband transmission system (b) signal-plus-noise waveform: Figure 11.1-2
yt    a k  ~
p t  t d  kT   n t 
k
47
Transmission
y t    a k  ~
p t  t d  kT   n t 
k
• The digital signal is time delayed

td
• The pulse is “filtered” and/or distorted by the channel
p t   fn pthe
• Recovering or Regenerating
~
c t 
t   hsignal may not be trivial
– Signal plus inter-symbol interference (ISI) plus noise
yˆ mT  t d   am   ak  ~
p mT  kT   nmT  t d 
k m
48
ABC Binary PAM formats
(a) Unipolar RZ & NRZ
(b) Polar RZ & NRZ
(c) Bipolar NRZ
(d) Split-phase Manchester
(e) Polar quaternary NRZ
49
PAM Power Spectral Density:
Polar NRZ

 t  Td  k  Tb 
vt    a k  rect  
k    Tb 
 
Ea n   0, E a n   2
2
pTd   ,
1
Tb
0  Td  Tb
 
E a j  a k  0, for j  k
 
Rvv   E vt   vt      1  ,
 2
Tb    Tb
 Tb 
S vv w  E vt   vt     2  Tb  sinc 2  f  Tb 
• For a zero mean, polar NRZ of amplitude +/- A and

symbol duration Tb
S vv w  A2  Tb  sinc 2  f  Tb 
 
E an   2  A 2
2

A2 f 
 sinc 2  
rb  rb  50
PAM Power Spectral Density:
Arbitrary Pulse

 t  Td  k  D  pTd  
1
vt    a k  p   , 0  Td  D
k    D  D
 
E an   ma , E an   a  ma
2 2 2

S vv  f    P f    Ra n   exp j  2  f  D 
1 2
D n  
 a 2  ma 2 , n0 Tb  D, rb 
1
Ra n    2 D
ma , n0
rb is symbol rate
• Using Poisson’s sum formula
2
a
2
 ma    n   n
2
S vv  f    P f       P    f  
2
D  D  n    D   D

S vv  f    a  rb  P f   ma  rb    Pn  rb    f  n  rb 
2 2 2 2
n   51
Power spectrum of
Unipolar, binary RZ signal
Figure 11.1-5
 
 t 
pt   rect   rect2  rb  t 
 Tb 
2
 f  A2 
2
A2 n
S vv  f    sinc     sinc    f  n  rb   2 
16  rb  2  rb  16 n    2
 f 
P f  
1
 sinc 
2  rb  b
2  r
E an   , E an 
A
2
2 A2
2
 
 2 A2
 a  ma  ,n  0
2
Ra n   
2
2
m 2  A , n0
 a 4
52
Power spectrum of
Unipolar, binary RZ signal
2
 f  A2 
2
A2 n
S vv  f    sinc     sinc    f  n  rb 
16  rb  2  rb  16 n    2
Unipolar Binary RZ
0.07
0.06
• For rb=2
0.05
0.04
0.03
0.02
0.01
0
-8 -6 -4 -2 0 2 4 6 8
freq (f)
Plots from PSD_PCM.m 53

Power spectrum of
Unipolar, binary NRZ signal
2
A2  f  A2   t 
S vv  f    sinc     sincn    f  n  rb  pt   rect   rectrb  t 
2
4  rb  rb  4 n    Tb 
2
2
 f  A2
S vv  f     f 
A f
 sinc   P f  
1
 sinc 
4  rb  b
r 4 rb  rb 
Unipolar Binary NRZ
 
0.25
A2
E an   , E an 
A 2
0.2
2 2
• For rb=2
 2 A2
 a  ma  ,n  0
0.15 2
Ra n   
2
2
0.1
m 2  A , n0
 a 4
0.05
0
-8 -6 -4 -2 0 2 4 6 8
freq (f)
54
Power spectrum of
Polar, binary RZ signal (+/- A/2)
 
 t 
pt   rect   rect2  rb  t 
2
A2  f 
S vv  f    sinc 
 Tb 
16  rb  2  rb   2 
 f 
P f  
1
• For rb=2  sinc 
0.035
Polar Binary RZ 2  rb  2  rb 
 
0.03
E an   0, E an  A
2 2
0.025 4
0.02
0.015  2  m 2  A2 , n  0
Ra n    4
a a
0.01
ma2  0, n0
0.005
0
-8 -6 -4 -2 0 2 4 6 8
freq (f)
55
Power spectrum of
Polar, binary NRZ signal (+/- A/2)
f 
2
 t 
pt   rect   rectrb  t 
2
S vv  f  
A
 sinc 
4  rb  rb   Tb 
f
P f  
• For rb=2 1
 sinc 
0.14
Polar Binary NRZ rb  rb 
0.12
0.1
2
 
E an   0, E an  A
2
0.08
 2  m 2  A2 , n  0
Ra n    4
0.06 a a
0.04 ma2  0, n0

0.02
0
-8 -6 -4 -2 0 2 4 6 8
freq (f)
56
Spectral Attributes of PCM
If Bandwidth W=1/T,
then WT=1
Note that WT=0.5 or a

bandwidth equal to ½ the
symbol rate can be used!
57
Baseband Binary Receiver
Figure 11.2-1
 
yt    a k  pt  kT  h t   n in t   h t 
k 
y t k   ak  n t k 
• Synchronous Time sampling of maximum filter output
58
Regeneration of a unipolar signal
(a) signal plus noise (b) S/H output (c) comparator output: Figure 11.2-2
59
Unipolar NRZ Binary Error
Probability
• Hypothesis Testing using a voltage threshold
– Hypothesis 0
• The conditional probability distribution expected if a 0 was sent
pY  yk | H 0   pY ak  n t k  | ak  0  pY n t k 
– Hypothesis 1
pY  yk | H 0   p N  yk 
• The conditional probability distribution expected if a 1 was sent
pY  yk | H1   pY ak  n t k  | ak  A  pY A  n t k 
pY  yk | H1   p N  yk - A 
60
Decision Threshold and
Error Probabilities
Conditional PDFs Figure 11.2-3

y t k   ak  n t k 
V
Pe1  PY  V    p Y y | H1   dy


Pe 0  PY  V    p Y y | H 0   dy
V
• Use Hypothesis to establish a decision rule

– Use threshold to determine the probability of correctly and
incorrectly detecting the input binary value
61
Average Error Probability
• Using the two error conditions:
– Detect 1 when 0 sent
– Detect 0 when 1 sent
Perror  PH 0   Pe 0  PH1   Pe1
• Selecting an Optimal Threshold

PH   p V | H   PH   p V
0 Y opt 0 1 Y opt | H1 
• For equally likely binary values
Perror   Pe 0  Pe1 

1
2
PH 0   PH1  
1
pY Vopt | H 0   pY Vopt | H1 
2
62
Threshold regions for conditional PDFs
Figure 11.2-4
P H 0   P H1  
1 A
Vopt 
2 2
63
For AWGN
• The pdf is Gaussian
 y2 
p Y y | H 0   p N y  
1
 exp  
2 
2   2
 2 
for 
 2 
Q x  
1
  exp    d
2 x  2

V  A 
Pe 0  PY  V    p N y   dy  Q   Q 
V     2   
AV  A 
V
Pe1  PY  V    p N y  A   dy  Q   Q 
     2   
for
P H 0   P H 1  
1 A
Vopt 
2 2
 A V   A 
Pe1  Q   Q   Pe 2 64
    2  
Modification for
Polar NRZ Signals (+/- A/2)
– Hypothesis 0
• The conditional probability distribution expected if a 0 was
sent
 A  A 
pY  yk | H 0   pY  ak  n t k  | ak     pY    n t k 
 2  2 
 A
pY  yk | H 0   p N  yk  
 2
– Hypothesis 1
sent
 A A 
pY  yk | H1   pY  ak  n t k  | ak    pY   n t k 
 2 2 
 A
pY  yk | H1   p N  yk - 
 2
A A
Vopt    0 65
2 2
Modification for
Polar NRZ Signals (+/- A/2)
• Determining the error probability
  A V 
 A   Q A 
Pe 0  PY  V    p N  y    dy  Q 2  
 2     2   
V
 
 A V 
 A   Q A 
V
Pe1  PY  V    p N  y    dy  Q 2  
 2     2   

 
• Notice that the error is the same as Unipolar NRZ
– The distance between the expected signal values is the
same
66
Modification for
Bipolar NRZ Signals (+/- A)
– Hypothesis 0
sent
pY  yk | H 0   pY ak  n t k  | ak   A  pY  A  n t k 
pY  yk | H 0   p N  yk  A
– Hypothesis 1
sent
pY  yk | H1   pY ak  n t k  | ak  A  pY  A  n t k 
pY  yk | H1   p N  yk - A 
Vopt  A  A  0 67
Modification for Bipolar NRZ Signals
• Determining the error probability

AV A
Pe 0  PY  V    p N y  A   dy  Q   Q 
V    
AV A
V
Pe1  PY  V    p N y  A   dy  Q   Q 
    
• Notice that the error has been reduced

– The distance between the expected signal values may
be twice as large as the unipolar case (using +/- A)
68
Relationship to signal power
• Defining the average received signal power
– Unipolar NRZ
1
S R   A2 , 0, A
2
  T2 
 A A  1 
1 S R  E  lim   xc t   dt  
2
– Polar NRZ S R   A2 ,  2 , 2 
  T T 2 
T 
4 
– Bipolar NRZ S R  A2 ,  A, A
• In terms of SNR
1  S 
2 2  N  for Unipolar
 A  A 2
  R
   
 2  4  N R  S 
  for Polar
 N  R
A 2  S 
2
A
      for Bipolar

  N R  N  R 69
Probability of error
• The probability of detecting a transmitted symbol
correctly is dependent upon the received signal-to-
PH   PH  
noise ratio …. assuming 1
0 1
2
– Unipolar NRZ (orthogonal)
 A   1 S   A 
2
A2 1 S
Pe  Q   Q   
      
 2    2  N R   2   4  NR 2  N R
– Polar NRZ (antipodal)

 A   S   A 
2
A2 S
Pe  Q   Q        
 2     N R   2   4  N R  N R
– Bipolar NRZ (antipodal)

 A  S   A
2
A2  S 
Pe  Q   Q        
    N R    N R  N R
  70
Power Ratio vs. Bit Energy
• For continuous time signals, power is a normal
way to describe the signal.
• For a discrete symbol, the “power” is 0 but the
energy is non-zero
– Therefore, we would like to describe symbols in terms
of energy not power
• For digital transmissions how to we go from
power to energy?
– Power is energy per time, but we know the time
duration of a bit. Noise has a bandwidth.
S R  Eb 
1
N R  N 0 W S SR
   ?
Tb  N R N R 71
SNR to Eb/No
• For the Signal to Noise Ratio
– SNR relates the average signal power and average noise
power (Tb is bit period, W is filter bandwidth)
1
Eb 
S Tb  E b  1  Eb  R b
  
 
     
 N  N 0  W  N 0  Tb  W  N 0  W
– Eb/No relates the energy per bit to the noise energy

(equal to S/N times a time-bandwidth product)
 Eb   S  W  S 
          Tb  W 
 N0   N  R b  N 
72
Relationship to Eb/No
• Defining the energy per bit to noise power ratio

for a time-bandwidth product of W  T  R  T  1b
b
b
2 2
 Eb 
2
 A  A2 1 S
– Unipolar          
 2  4  N R 2  N R  N0 
 2  Eb 
2
 A  A2 S
– Polar         
 2  4  N R  N R  N0 
 2  Eb 
2
– Bipolar A A2  S 
        
 N R  N R  N0 
73
Relationship to Bit Error Probability
• Defining the binary bit error probability

for a time-bandwidth product PH   PH  
1
0 1
2
A  Eb 
– Unipolar Perror  Q   Q 

 2   N0 
A  2  Eb 
– Polar Perror  Q   Q 

 2   N0 
A  2  Eb 
– Bipolar Perror  Q   Q 

  N0 
74
Bit Error Rate Plot
Classical Bit Error Rates
0.5
Orthogonal
0.45 Antipodal
0.4
0.35 EbNo=(0:10000)'/1000;
0.3
Bit Error Rate
% Q(x)=0.5*erfc(x/sqrt(2))
0.25
Ortho=0.5*erfc(sqrt(EbNo)/sqrt(2));
0.2 Antipodal=0.5*erfc(sqrt(2*EbNo)/sqrt(2));
0.15
semilogx(EbNo,[Ortho Antipodal])
0.1 ylabel('Bit Error Rate')
xlabel('Eb/No')
0.05 title('Classical Bit Error Rates')
legend('Orthogonal','Antipodal')
0
-3 -2 -1 0 1
10 10 10 10 10
Eb/No
75
BER Performance, Classical Curves
log-log plot
0
10
Orthogonal
-1 Antipodal
10
-2
10
Bit Error Rate
-3
10
-4
10
-5
10
-6
10
-7
10
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Eb/No
76
Antipodal and Orthogonal Signals
• Antipodal
– Distance is twice “signal voltage”
– Only works for one-dimensional signals
T
 1 for i  j
zij    si t   s j t   dt  
1
d  2  Eb
E 0  1 for i  j
• Orthogonal
– Orthogonal symbol set
– Works for 2 to N dimensional signals
T
1 for i  j
z ij    s i t   s j t   dt  
1
d  2  Eb
E 0 0 for i  j
77
M-ary Signals
• Symbol represents k bits at a time M  2k

– Symbol selected based on k bits
– M waveforms may be transmitted
• Allow for the tradeoff of error probability for
bandwidth efficiency
• Orthogonality of k-bit symbols
– Number of bits that agree=Number of bits that disagree
 sumb   
K N
i
 b   sum b ik  b kj
j
k k
1 for i  j
z ij  k 1 k 1

K 0 for i  j
78
Example 11.2-1
• Unipolar computer network with
Rb  106 bps N 0  4 10 20 W / Hz  194dB / Hz
– Desired BER is one bit per hour
Pe  1  3 10 10
3600  Rb
• Solve for the signal energy
 A A
Perror  Q   3 10 10 From p. 790  6.2
2
 2 
1 S 
2
 A  A2
      R  for Unipolar
 2  4  NR 2  NR 
2  R 
2  6.2   N R  2  6.2   N 0  b   S R
2
 2
 
S R  2  38.44  4 10 20  0.5 106 W  1.54 10 12W 79
Exercise 11.2-1 (1)
• Unipolar system with equally likely digits and
SNR = 50
• Calculate the error probabilities when the
threshold is set to V=0.4 x A
1 S 
2
 A 
     R  for Unipolar
 2  2  NR 
 A
 50  10
1
   2

  2
 A V 
V
Pe1  PY  V    N
p  y  A   dy  Q  
  

V 
Pe 0  PY  V    p N  y   dy  Q 
V  80
Exercise 11.2-1
• Calculate the error probabilities when the
threshold is set to V=0.4 x A PH   PH  
1
0 1
V  2
Pe 0  PY  V   Q   Q0.4 10

Pe 0  Q4.0  3.5 10 5 V=0.5 x A
 A V 
Pe1  PY  V   Q   Q0.6 10 Pe 0  Pe1  Q0.5 10  Q5.0
  
Pe1  Q6.0  1.5 10 9
Pe 0  Pe1  Perror  3.5 10 7
Perror   Pe 0  Pe1 

1
2
Perror   3.5 10 5  1.5 10 9   1.75 10 5

1
2
81
Gaussian Distribution
The Gaussian probability density function (pdf)
The Gaussian or Normal probability density function is defined as:
f X x  
1
 exp
 
 x X 2 
, for    x  
2    
 2 
2

The Gaussian Probability Distribution Function (PDF)
 
Gaussian PDF and pdf
x  v X 2  1
FX x  
  exp   dv
1
2    
 2 
2 0.9
v   
0.8
The PDF can not be represented in a closed form solution! 0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-8 -6 -4 -2 0 2 4 6 8
ECE 6640 82
Gaussian Distribution
The Gaussian Probability Distribution Function is
x

 v X 2  
FX x  
  exp   dv
1
2    
 2 
2
v   
The PDF can not be represented in a closed form solution!
The PDF is tabulated for a zero mean, unit variance pdf.

For these values, it is often described as “normalized” and is defined as
x
 u2 
x  

1
 exp    du
2  2 
u    
The distribution function is then defined as

x X 
FX  x    

  
When using Appendix D, the negative values in x are derived as

  x   1   x 
ECE 6640 83
Q Function
Another defined function that is related to the Gaussian (and used) is the Q-function.:

 u2 
Qx  
1


exp 
 2 
  du Q Function Table p. 858
2  
ux
The Q-function is the complement of the normal function, :

Q x   1    x 
Therefore not that:

Q x   1  Q x 
x X 
FX  x   1  Q 

  
84
Using MATLAB
Another way to find values for the Gaussian
The error function
 exp u  du
x
erf  x  
2 2


u 0
1   x 
Q x    1  erf  
2   2 
1   x X  1 1  xX 
FX x   1   1  erf      erf 
  2 


2   2   2 2  
The error function (Y = ERF(X)) is built-in to MATLAB. .
From MATLAB:
ERF Error function.
Y = ERF(X) is the error function for each element of X. X must be
real. The error function is defined as:
erf(x) = 2/sqrt(pi) * integral from 0 to x of exp(-t^2) dt.
See also erfc, erfcx, erfinv.
Reference page in Help browser

doc erf
ECE 6640 85
Using MATLAB (2)
The complementary error function
erfcx   1  erf x 
 x 
Q x  
1
 erfc 
2  2
The error function (Y = ERFC(X)) is built-in to MATLAB. .
From MATLAB:
ERFC Complementary error function.
Y = ERFC(X) is the complementary error function for each element
of X. X must be real. The complementary error function is
defined as:
erfc(x) = 2/sqrt(pi) * integral from x to inf of exp(-t^2) dt.

= 1 - erf(x).
Class support for input X:

float: double, single
See also erf, erfcx, erfinv.
Reference page in Help browser

doc erfc
ECE 6640 86
Qfn and Qfninv
• These function are now in the Misc_Matlab zip file on the
web site
function [Qout]=Qfn(x)
% Qfn(x) = 0.5 * erfc(x/sqrt(2));
Qout = 0.5 * erfc(x/sqrt(2));
function [x]=Qfninv(Pe)
% For Qfn(x) = 0.5 * erfc(x/sqrt(2));
% The inverse can be found as
x=sqrt(2)*erfcinv(2*Pe);
87
Properties of Matched
• See ECE3800 Notes
– Review from Chapter 9
• Wikipedia
– http://en.wikipedia.org/wiki/Matched_filter
– “The matched filter is the optimal linear filter for maximizing the
signal to noise ratio (SNR) in the presence of additive stochastic
noise.”
88
Defining a Filter for Pulses
y t    ak  pt  t d  k  T 
k
• We want to minimize or zero inter-symbol interference

(ISI)
1 t 0
pt   
0 t  T ,2T , 
• We want a frequency band limited filter

P f   0 B f
r r 1
where B   , with 0    and B  BT r
2 2 T
– Allowable signal rates with  as the excess bandwidth
r  2  B  , for B  r  2  B
89
Defining a Filter for Pulses
• Possible solutions
1 t 0 pt   p t   sincr  t 
pt   
0 t  T ,2T , 
P f   0 B f  p t   P  f   0 for   f

p 0   P  f  df
 1

• Therefore we select
1  f  1
P f   P  f     rect  r
r  r  T
These are considered the Nyquist conditions for the filter

90
Cosine Spectral Shaping
Raised cosine pulse.
(a) Waveform
From Chap 2
(b) Derivatives Raised cosine
(c) Amplitude pulse
spectrum
Figure 2.5-7
• A candidate filter is (with

with  as the excess BW)
  2  f   f 
P  f    cos   rect 
4  4   2   
91
Convolving
• Raised Cosine Convolution with Bandlimited Spectrum
1  f 
P f   P  f     rect 
r  r 
1 r
r f  
2

1  2   r
P f     cos 2 
r r
  f        f  
r  4   2  2 2
 r
0 f  
 2
r r
where B   , with 0    and B  BT
2 2
• Transforming to the time domain filter

cos2    t 
pt    sincr  t 
1  4    t 
2
92
Nyquist/Raised Cosine Pulse Shaping
http://en.wikipedia.org/wiki/Raised-cosine_filter
 ABC

r
2
1
r
T
GNU FDL:Oli Filth, Raised Cosine Filter Response , en.wikipedia.org, 3 November 2005, Oli Filth
 ABC

r
2
1
r
T
GNU FDL:Oli Filth, Raised Cosine Filter, Impulse Response, en.wikipedia.org, 3 November 2005, Oli Filth 93
Nyquist Filter
 n 
cos     
cos2    t   M  n 
pt    sincr  t  pn    sinc  , for k  M  n  k  M
1  4    t 
2
M 
2
 n 
1   2   
 M
% function hnyq=nyquistfilt(alpha,M)
2    r
% or
r
0  0  1 % function hnyq=nyquistfilt(alpha,fsymbol,fsample,k)
2 %
cos    r  t 
% alpha roll-off
pt    sincr  t  % fsample rate
1  2    r  t 
2
% fsymbol rate
% M = fsample/fsymbol (an integer value)
% k is 1/2 the number of symbols in the filter
n fs
t r % The filter length is euqal to 2*ceil(k*M)+1
fs M %
% A discrete time cosine taperd Nyquist filter
n
r t  % Based on frederic harris, Multirate Signal Processing for
Communications
M
% Prentice-Hall, PTR, 2004. p. 89 94
MATLAB Raised Cosine Filters (1)
• Rcosine (obsolete)
– [NUM, DEN] = RCOSINE(Fd, Fs, ‘fir’, R)
– FIR raised cosine filter to filter a digital signal with the digital
transfer sampling frequency Fd. The filter sampling frequency is
Fs. Fs/Fd must be a positive integer. R specifies the rolloff factor
which is a real number in the range [0, 1].
• rcosfir
– B = RCOSFIR(R, N_T, RATE, T)
– Raised cosine FIR filter. T is the input signal sampling period, in
seconds. RATE is the oversampling rate for the filter (or the
number of output samples per input sample). The rolloff factor, R,
determines the width of the transition band. N_T is a scalar or a
vector of length 2. If N_T is specified as a scalar, then the filter
length is 2 * N_T + 1 input samples.
95
MATLAB Raised Cosine Filters (2)
• firrcos
– B=firrcos(N,Fc,DF,Fs)
– Returns an order N low pass linear phase FIR filter with a raised
cosine transition band. The filter has cutoff frequency Fc,
sampling frequency Fs and transition bandwidth DF (all in Hz).
– The order of the filter, N, must be even.
– Fc +/- DF/2 must be in the range [0,Fs/2]
– The coefficients of B are normalized so that the nominal passband
gain is always equal to one.
– B=firrcos(N,Fc,R,Fs,'rolloff') interprets the third argument, R, as
the rolloff factor instead of as a transition bandwidth.
– R must be in the range [0,1]
96
Textbook Waveform Energy
• Waveform Energy
T
Ei   si t   dt
2
• Matched Filter
t
z t   r t   ht    r    ht     d

ht   u t   s * T  t 
t
z t    s   s * T  t     d
0
T T T
z T    s   s T  T     d   s   s    d   s    d
* * 2
0 0 0
ECE 6640 Correlation 97

Optimum binary detection
(a) parallel matched filters (b) correlation detector: Figure 14.2-3
98
Symbols and Matched Filters
y (t )  si (t ) h opt (t )
si (t ) h opt (t ) A2
A A
T T
T t T t 0 T 2T t
y (t )  si (t ) h opt (t )
si (t ) h opt (t ) A2
A A
T T
T/2 T t T/2 T t 0 T/2 T 3T/22T t


2
A A A
2
T T
2006-01-31 Lecture 3 99
Optimized Error Performance
• Maximize the “distance”, Ed, between the sampled values

that are used for detection.
– The distance is based on the correlation of the symbols
T T T
Ed   s1 t   dt  2   si t   s j t   dt   s2 t   dt
2 * 2
0 0 0
T T
Eb   si t   dt   Eb   si t   s j * t   dt
2
0 0
Ed  Eb  2   Eb  Eb  Eb  1   
ECE 6640 100

Optimized Error Performance
• To maximize the distance:  = -1 Ed  2  Eb

– Symbols are said to be antipodal
– Examples: +/- 1 Symbols (Bipolar), BPSK
– Not always achievable
• Useful performance: :  = 0 Ed  Eb
– Symbols are said to be orthogonal
– Examples: On-Off Keying (ASK), FSK, “independent symbols”
ECE 6640 101

• Antipodal
T
 1 for i  j
zij    si t   s j t   dt  
1
d  2  Eb
E 0  1 for i  j
• Orthogonal
T
1 for i  j
z ij    s i t   s j t   dt  
1
d  2  Eb
E 0 0 for i  j
102
Relationship to Bit Error Probability
• Defining the binary bit error probability

for a time-bandwidth product PH   PH  
0 1
1
2
– Orthogonal  Eb 
Perror  Q 

 N0 
– Antipodal  2  Eb 

 N0 
103
Bit Error Rate Plot-Linear BER
0.5
Orthogonal
0.45 Antipodal
0.4
0.35 EbNo=(0:10000)'/1000;
0.3
Bit Error Rate
% Q(x)=0.5*erfc(x/sqrt(2))
0.25
Ortho=0.5*erfc(sqrt(EbNo)/sqrt(2));
0.2 Antipodal=0.5*erfc(sqrt(2*EbNo)/sqrt(2));
0.15
semilogx(EbNo,[Ortho Antipodal])
0.1 ylabel('Bit Error Rate')
xlabel('Eb/No')
0.05 title('Classical Bit Error Rates')
legend('Orthogonal','Antipodal')
0
-3 -2 -1 0 1
10 10 10 10 10
Eb/No
ECE 6640 104

BER Performance Fig. 3.14
0
10
Orthogonal
-1 Antipodal
10
 Eb 
 Q 
-2
10 Perror 
 N0 
Bit Error Rate
-3
10
-4
10
 2  Eb 
-5
10

10
-6
 N0 
-7
10
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Eb/No
ECE 6640 105

ECE 6640

Assistant Professor
Chapter 4
4. Bandpass Modulation and Demodulation/Detection.
1. Why Modulate?
2. Digital Bandpass Modulation Techniques.
3. Detection of Signals in Gaussian Noise.
4. Coherent Detection.
5. Noncoherent Detection.
6. Complex Envelope.
7. Error Performance for Binary Systems.
8. M-ary Signaling and Performance.
9. Symbol Error Performance for M-ary Systems (M>>2).
ECE 6640 2
Bandpass Demodulation and Detection
• Focus on Signal of Symbol, Samples, and Detection

• In the presence of Gaussian Noise and Channel Effect
ECE 6640 5
Analog Bandpass Modulation
Includes the RF/IF Frequency
• AM , PM and FM Modulation
st   At   cos2  f 0  t     t 

 t 
 A  1    m1 t  cos2  f 0  t     p  m2 t   2   f   m3    d 
 
• The time varying phase components
t   2  f 0  t    t 
t
 2  f 0  t     p  m 2 t   2   f   m 3    d
ECE 6640 6
Phasor Representation
• Taking the positive spectrum complex representation
st   At   Reexpj  2  f 0  t  j    j  t 
• Think in terms of the “analytical signal” representation

– Complex, positive frequencies only
ECE 6640 7
Example: Bandpass Phasor Analysis
of Double Sideband (DSB)
• Given a tone message …
mt   A m  cos2  f m  t 
st   A c  A m  cos2  f m  t   cos2  f c  t 
Ac  Am
s t    cos2  f c  f m   t   cos2  f c  f m   t 
2
• A positive frequency phasor can be defined and drawn
– First define the complex signal as (cos  exp)
Ac  Am
s pos  f  t  
C
 exp j  2  f c  f m   t   exp j  2  f c  f m   t 
4
ECE 6640 8
Phasor Analysis DSB (2)
Ac  Am
s pos  f  t  
C
 exp j  2  f c  f m   t   exp j  2  f c  f m   t 
4
4
A
m
A
c
f fc 4
 A
m
m
A
f c

m
ECE 6640 9
Phasor Analysis AM
• Given a tone message …

A m t   1    cos2  f m  t 
st   A c  1    cos2  f m  t   cos2  f c  t 

Ac A
st   A c  cos2  f c  t      cos2  f c  f m   t   c    cos2  f c  f m   t 
2 2
s t  pos  f    exp j  2  f c  t   c    exp j  2   f c  f m   t   c    exp j  2   f c  f m   t 

C Ac A A
2 4 4
ECE 6640 10
Phasor Analysis AM (2)
s t  pos  f    exp j  2  f c  t   c    exp j  2   f c  f m   t   c    exp j  2   f c  f m   t 

C Ac A A
2 4 4
 4
A
fm

fm
c

2
A

4
c
A
fc
ECE 6640 11
Narrowband FM & PM Spectrum
s C t   A  expj  2  f c  t  j  t 
• Forming the Quadrature Representation and transforming
the series expanded rig functions
s C t   A  expj  2  f c  t  j  t   A  expj  2  f c  t expj  t 
 1 
s C t   A  expj  2  f c  t 1  j  t     j  t   
2
 2! 
• Maintaining the 1st order terms …
s C t   A  expj  2  f c  t 1  j  t 
ECE 6640 12
Narrowband FM & PM Spectrum (2)
• Taking the Fourier Trasnform of the 1st order

approximation
s C t   A  expj  2  f c  t 1  j  t 
SC f   A  f  f c   f   j   f 
SC f   A  f  f c   j   f  f c 
SC f   A  f  f c    f  f c 
2 2 2
ECE 6640 13
PM and FM Basis
• Based on the previous analysis, we need to determine the

transform of the phase components
t
 PM t    p  m 2 t   FM t    f   m 3    d
M 3  
 PM f    p  M 2 f   FM t    j   f 
f
ECE 6640 14
PM Phasor v1
s c t   A  expj  2  f 0  t  j   expj   p  m 2 t 
• The carrier can be removed to describe the baseband signal

as a bounded phase variation about the carrier
2 t  Ac
m
p
fo
ECE 6640 15
PM Phasor v2
s C t   A  expj  2  f c  t 1  j  t 
• For a cos wave message input
 j  p j  p 
s C t   A  expj  2  f c  t 1   expj  2  f m  t  exp j  2  f m  t
 2 2 
Ac
fo
ECE 6640 16
FM Phasor
s C t   A  expj  2  f c  t 1  j  t 
• For a cos wave message input
 t 
s C t   A  expj  2  f c  t 1  j   f   cos2  f m  t   d 
 
s C t   A  expj  2  f c  t 1  j   f  sin 2  f m  t 
   
s C t   A  expj  2  f c  t 1  f  exp j  2  f m  t   f  exp j  2  f m  t 
 2 2 
• See Figure 4.4, p. 173
ECE 6640 17
Why discuss phasors?
• We are about to describe digital modulation in terms of

one, two, and three dimensional constellation points.
– Amplitude Shift Keying: 1-D array of possible points
– Phase Shift Keying: 2-D circle with points equally spaced on the
circle
– Frequency Shift Keying: N-D space with one point on each of the
N axis
– Quadrature Amplitude Modulation: 2-D 2Mx2M array of points
ECE 6640 18
General Notes from ABC
• The following notes are based on Carlson Chapter 14.
• There is a notational difference between Sklar and Carlson

in describing a symbol. Sklar’s more easily lends itself to
defining Eb/No!
ECE 6640 19
Binary modulated waveforms
a) ASK
b) FSK
c) PSK
d) DSB with
baseband pulse
shaping
See Figure 4.5 on p. 174 20

Amplitude Shift Keying (ASK)
• Digital Symbol Amplitude Modulation
• On-Off Keying (OOK)
p 0 t   0 s0 t   0
p1 t   1 s1 t   Ac  cos2  f c  t   
• Auto-correlation
E s0 t   s0 t     0
Es 0 t   s1 t     0
2
 
E s1 t   s1 t         cos2  f c  
Ac
2 T 
• Average Power
POOK  P0   Rs0 s0    P1  Rs1s1  
2 E
1 1 Ac
2
0 2
Ac 
    cos0  
Ac
POOK  0   T
2 2 2 T  4 21
Amplitude Shift Keying (2)
• Auto-correlation
R s0s0    0 2 E
Ac 
Ac
2
 T
R s1s1        cos2  f c   
2 T
• Symbol Power Spectral Density
 
2
S OOK    c  T 2  sinc 2  f c  f   T   sinc 2  f c  f   T 
A
8
• Bandpass Bandwidth
– Nominally: BT=1/T, first null at Bnull=2/T
22
ASK Power Spectrum
• From ABC Chapter 11

2 2 2 2
n  
• Baseband or LPF analysis

pt   rectrb  t 
 
2
E an  
A 2 A
, E an  1 f
2 2 P f    sinc 
rb  rb 
2
A2  f  A2
S vv  f    sinc     f 
4  rb  rb  4
• RF Analysis
1
Gc  f    S vv  f  f c   S vv  f  f c 
4
23
ASK Power Spectrum (2)
2
A2  f  A2
S vv  f    sinc     f  1
4  rb  rb  4 rb 
Tb
1
4
Figure 14.1-2 24
ASK MATLAB Simulation
Symbol Sequenct in Time
0
Symbol Sequence Circular Auto-correlation
0.5
-50
Magnitude (dB)
Amplitude
-0.5 -100
-1
-150
1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6
Time -5
Frequency 8
x 10 x 10
OOK Demodulation Eye Diagram
2.5 Symbol Sequence Circular Auto-correlation
0
2 -10
-20
1.5
Magnitude (dB) -30
Amplitude
1 -40
-50
0.5
-60
0 -70
-80
-0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 25
Time -6
3 3.05 3.1 3.15 3.2 3.25 3.3 3.35 3.4
x 10 Frequency 7
x 10
ASK Transmission Capability
• Comparing the ratio of the bit rate to the required
signal bandwidth
rb
TP 
BT
– From the previous slide for the bandwidth

BT  rb
– Therefore, the transmission capability is
rb
TP   1 bit  per  second Hz
BT
26
M-ary ASK
• Use multiple amplitude levels to represent more than one
bit per symbol
• MASK
– M-1 one states and the off state
– All positive amplitudes (no phase reversals)
M 1
ma  E an  
2
 a  E an
2
  2 2
 ma 
M 2 1
12
S vv  f  
2

A  M 1
2
  f  A2  M  12
 sinc  
2
  f 
12  rb  b
r 4
Gc  f   S vv  f  f c   S vv  f  f c  27
M-ary ASK Transmission Capability
• Comparing the ratio of the bit rate to the required
signal bandwidth
– For m-ary, the bit rate is
bit  rate  rs  log 2 M 
– The symbol bandwidth remains

BT  rs

rs  log 2 M 
TP   log 2 M  bits  per  second Hz
BT
– Note that for m-ary ASK, the OOK system has the
smallest spectral efficiency
28
Binary QAM
(a) transmitter (b) signal constellation : Figure 14.1-3
xi t    a2k  pt  k  T   xi t   cos2  f c  t  

xc t   Ac   
   
k
  sin 2     
xq t    a2k 1  pt  k  T   q x t f c t 
k
ma  E an   0
 
 a 2  E an 2  A 2 29
Quadrature AM (QAM)
• An M-ary Signal – 4 complex symbols
• Quadrature
p 0 t   1 s 0 t   1  A c  cos2  f c  t  
p1 t   i s1 t   i  A c  cos2  f c  t  
p 2 t   1 s 2 t   1  A c  cos2  f c  t  
p 3 t   i s 3 t   i  A c  cos2  f c  t  
• Auto-correlation, Single Pulse Period


E sk t   sk t   
*
 
 
 Ac 2
 2  cos2  f c  

T 
• Average Power
0 1 Ac
2
 A c      cos0  
2
E QAM
T 2 2 30
QAM
• Symbol Cross Correlation
  
C0, 0 t   1    C0, 2 t   1   
T  T 
  
C0,1 t   i    C0,3 t   i   
T  T 
• Not that adjacent symbol average correlation is

zero for equal probability symbols …
E sk t   i   Ac  E cos2  f c  t     0
k 1
E sk t   sk 1 t   P0  C s0 s0    P1  C s0 s1    P2   C s0 s2    P3  C s0 s2  

1 1 1 1   
E sk t   sk 1 t    1   i    1    i    0

4 4 4 4  T 
31
Quadrature AM Power Spectrum
2 2
 t  1 f
pt   rect   rectrs  t  S vv  f   Ac  rs     sinc 
2
 Ts   rs   rs 
1 f
P f    sinc 
r 2
rs  b 2 1 f
rs  rs  2 S vv  f   Ac   sinc 
rs  rs 
Note that the symbol rate
is one-half the bit rate. Gc  f   S vv  f  f c   S vv  f  f c 
2
2
Ac  4  2 f 
S vv  f    sinc 
rb  rb 

S vv  f    a  r  P f   ma  r    Pn  r    f  n  r 
2 2 2 2
32
n  
QAM Transmission Capability
• Comparing the ratio of the symbol rate to the
required signal bandwidth
rs  log 2 M 
TP 
BT

TP  2 bits  per  second Hz
33
Phase Modulation Methods
• Phase shift keying (PSK) is digital PM
x t   A   cos2  f  t       p t  k  T 
c c c k D s
k
– Points on a unit circle of a constellation plot

– 4-QAM as previously described is using phase to
represent symbols. The magnitude is the same, but
successive symbols differ by 90 degrees in phase.
• Frequency shift keying (FSK) is digital FM

x t   A   cos2  f  t    2  a  f  t   p t  k  T 
c c c k d D s
k
– Multiple discrete frequencies

34
PSK Signal Constellations
This is QAM,
rotated by pi/4
for 4-PSK
M=4 M=8
4-PSK 8-PSK
35
M-PSK
• An M-ary Signal – M complex symbols
• Quadrature (2 possible representations)

s k t   A c  cos 2  f c  t   
2  k  1   , for k  0 to M  1

 M 
  2  k  1     2  k  1    
p k t   I k , Q k    cos , sin   , for k  0 to M  1
  M   M 
• Auto-correlation, single symbol Period

   1
E s k t   s k t    A c      cos2  f c  
* 2
T 2
• Average Power, Amplitude to Energy
0 1 2
2 E
 Ac      cos0  
2 Ac Ac 
PQAM
T  2 2 T
36
Binary PSK
• Signal Symbols
s 0 t   A c  cos2  f c  t      0   A c  cos2  f c  t  
s1 t   A c  cos2  f c  t     1  A c  cos2  f c  t  
• Autocorrelation

E s k t   s k t   
*
  1
 A c      cos2  f c  
2
T 2
• Cross Correlation (the definition of antipodal)
   1
E s 0 t   s1 t     A c      cos2  f c  
* 2
T 2
Rs0 s1    Rs0 s0 
37
Binary PSK
• Signal Symbols
s 0 t   A c  cos2  f c  t      0   A c  cos2  f c  t  
s1 t   A c  cos2  f c  t     1  A c  cos2  f c  t  
• Autocorrelation

E s k t   s k t   
*
  1
 A c      cos2  f c  
2
T 2
• Cross Correlation (the definition of antipodal)
   1
E s 0 t   s1 t     A c      cos2  f c  
* 2
T 2
Rs0 s1    Rs0 s0 
38
BPSK Power Spectrum
• From Chapter 11

2 2 2 2
n  
• Baseband or LPF analysis

pt   rectrb  t 
 
E an   0, E an  A2
2
1 f
P f    sinc 
rb  rb 
2
2
f
S vv  f  
A
 sinc 
rb  rb 
• RF Analysis
1
2
39
1
BPSK MATLAB Simulation -20
0.8
-40
0.6
0.4 -60
0.2
Magnitude (dB)
Amplitude
-80
0
-0.2 -100
-0.4
-120
-0.6
-0.8
-140
-1
0 0.5 1 1.5 2 2.5 3 3.5 4 -160

Time -6
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 10 Frequency 8
BPSK Demodulation Eye Diagram x 10
1.5 0
1 -20
0.5 -40
Magnitude (dB)
Amplitude
0 -60
-0.5 -80
-1 -100
-1.5 -120
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 40
Time -6 Frequency 7
x 10 x 10
Other Forms of PSK
• Differential PSK
– The symbols are the “encoding” of two adjacent bits
– Encoding the bit changes not the bit values
– Typically an exclusive-Or or Exclusive NOR
• QPSK
– Already shown as QAM
• Offset QPSK
– Offset the I and Q bits of QAM by one half the symbol period
– Phase changes at BPSK bit rate, bit absolute phase change is now
always pi/2 (orthogonal)
41
Differential Encoded PSK (DPSK)
• The binary data stream is differentially encoded
– The logical combination of the previous bit sent and the next bit to
be sent. An Exclusive NOR gate can be used.
– Provides an arbitrary start … only phase change by pi is required
to decode the message, not the absolute bit values!
Sample Index 0 1 2 3 4 5 6 7 8 9 10
Information m(k) 1 1 0 1 0 1 1 0 0 1
Diff. Encoding (0) 0 0 0 1 1 0 0 0 1 0 0

DPSK Phase 0 0 0 pi pi 0 0 0 pi 0 0
Detect 1 1 0 1 0 1 1 0 0 1
Diff. Encoding (1) 1 1 1 0 0 1 1 1 0 1 1

DPSK Phase pi pi pi 0 0 pi pi pi 0 pi pi
Detect 1 1 0 1 0 1 1 0 0 1
42
Offset-keyed QPSK transmitter
(OQPSK)
Figure 14.1-6
• Instead of changing I and Q at the same

time, delay the change by T/2.
• Visualize the phase changes … always to
an adjacent symbol!
43
Digital Frequency Modulation
Frequency Shift Keying Continuous Phase FSK

(FSK) (CPFSK)
44
Frequency Shift Keying
• Binary FSK
s0 t   Ac  cos2   f c  f d   t  
s1 t   Ac  cos2   f c  f d   t  
• M-ary FSK or MFSK

sk t   Ac  cos2   f start  f step  k  t  , for k  0 to M  1
• Desired Condition (makes the time signal

continuous at the symbol time boundaries)
2   f step  TS   m  2 , for m an interger
45
M-FSK
• An M-ary Signal – M complex symbols
s t   A  cos2  f  t    2   f  k  t ,
k c start for k  0 step to M  1
• Desired Condition (normally)

2   f step  T   m  2, for m an interger
Can make expected
value zero
• Crosscorrelation

E s0 t   sk t  
*
 2  1
 
 Ac      E cos2   f start  f step  k    2   f step  k  t 
T  2
• Autocorrelation

E sk t   sk t   
*
  1
 Ac      cos2   f start  f step  k  
2
T  2
46
BFSK
• Signal Symbols
s0 t   Ac  cos2   f c  f d   t  
s1 t   Ac  cos2   f c  f d   t  
• Autocorrelation
   1
E sk t   sk t    Ac      cos2   f c  f d   
* 2
T  2
• Cross Correlation
  
E s0 t   s1 t     Ac     E cos2   f c  f d   t    cos2   f c  f d   t     
* 2
T 
   1
E s0 t   s1 t     Ac      E cos2  2  f d   t  2   f c  f d   
* 2
T  2
orthogonal for 2  x2fdxT=2 47

BFSK Quadrature Representation (1)
rb
sk t   Ac  cos2   f c  ak  f d   t   ak  1 fd 
2
sk t   Ac  cos2  f c  t    cos2  ak  f d  t   Ac  sin 2  f c  t    sin 2  ak  f d  t 
sk t   Ac  cos2  f c  t    cos2  f d  t   ak  Ac  sin 2  f c  t    sin 2  f d  t 
sk t   Ac  cos2  f c  t    cos  rb  t   ak  Ac  sin 2  f c  t    sin   rb  t 
• The sign term for odd bits becomes

sk t   Ac  cos2  f c  t    cos  rb  t    1  ak  Ac  sin 2  f c  t    sin   rb  t 
k

bbk t   I k , Qk   cos  rb  t ,  1  ak  sin   rb  t 
k

48
BFSK Quadrature Representation (2)
 
bbk t   I k , Qk   cos  rb  t ,  1  ak  sin   rb  t 
k
• The baseband spectrum Glp

1   r   r 
Glp  f   Gi  f   Gq  f     f  b    f  b   rb  Qk 
2
4   2  2 
2
1   r    r  
Qk 
2
  sinc  f  b   rb   sinc  f  b   rb 
4  rb   2    2 2  
  f  
cos 
4 
 rb  
Qk   2 2  
2
2

  rb   2  f  

    1 
rb 
2
 
 cos   f  
1   r   r  4  rb 
Glp  f   Gi  f   Gq  f     f  b    f  b   2   
4   2  2    rb   2  f  2


    1 
rb  49
Power spectrum of BFSK
Figure 14.1-8
2
    f  
cos
1   r   rb  4   rb  
Glp  f   Gi  f   Gq  f     f  b    f    2   
4   2  2    rb   2  f  2


    1 
rb 
2  2  f d  rb   2
rb
fd 
2
50
BFSK MATLAB Simulation 0
Not readily observable -50
Magnitude (dB)
The change in frequency -100
is too small
-150
BFSK Demodulation Eye Diagram 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
1.5 Frequency 8
x 10
0
1
-20
0.5
-40
Amplitude
Magnitude (dB)
0
-60
-0.5
-80
-1
-100
-1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-120
Time -6
x 10 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 51
Frequency 7
x 10
Spectrum of M-FSK
• As tones with equal spacing are required, MFSK requires

additional bandwidth for additional symbol tones.
– The bandwidth must grow as a multiple of M,
whereas for M-PSK the bandwidth is based on the symbol period.
– M-FSK is inherently wideband modulation.
– More bits per symbol requires more bandwidth
52
Special Versions of FSK
• Continuous Phase FSK (CPFSK)
 t

xc t   Ac  cos 2  f c  t    2  f d   x   d 

 0 
 a0  t , 0t T
a  T  a  t  T , T  t  2 T
 0 1

t
0 x    d  
 k 1 
 a j  T   ak  t  k  T , k  T  t  k  1  T
 j 0 
• Minimum-Shift Keying (MSK)

– The binary version of CPFSK
– Also called fast FSK
– Capable of using an rb/2 bandwidth 53
CPFSK
• Continuous Phase FSK (CPFSK)
 t

xc t   Ac  cos 2  f c  t    2  f d   x   d 

 0 
 a0  t , 0t T
a  T  a  t  T , T  t  2 T
 0 1

t
0 x    d  
 k 1 
 a j  T   ak  t  k  T , k  T  t  k  1  T
 j 0 
• The phase is continuous at the transitions between
bit.
– This is most easily accomplished if the phase is π or a
multiple of π at the start and end of each bit period.
54
Binary CPFSK
• The binary version of CPFSK is called

Minimum-Shift Keying (MSK)
– Also called fast FSK
– Capable of using an rb/2 bandwidth
55
MSK Baseband
bbk t   I k , Qk 
 
xi t    cos k  ak  ck   pt  k  T  xq t    sin  k  ak  ck   pt  k  T 
k k
m  , for k even
  rb 
ck   t  k  T  k   
2  n    , for k odd
2
• Frequency and phase (history) modulation
56
Illustration of MSK.
(a) phase path (b) i and q waveforms: Figure 14.1-11
• MSK includes the

phase history with the
frequency slope in time
of the new bit.
• Therefore the phase
plot in time can appear
as shown, with the
corresponding
quadrature
components.
57
Minimum Shift Keying (MSK)
MSK power spectrum: Figure 14.1-9
2   f step  T    Use 0.25 in BFSK Sim
58
Sklar Representations
• Amplitude Shift Keying s0 t   0

2  Ei 2 E
si t    cos2  f c  t    s1 t    cos2  f c  t   
T T
• Phase Shift Keying

2 E
2 E  2  k  1    s0 t    cos2  f c  t    0 
sk t    cos 2  f c  t     T
T  M 
2 E
for k  0 to M  1 s1 t    cos2  f c  t     
T
• Frequency Shift Keying

2 E   f step  
s0 t    cos 2   f c    t   

2 E T   2  
sk t    cos2   f min  k  f step  t   
T 2 E   f step  
s1 t    cos 2   f c    t   

for k  0 to M  1 T  2 
ECE 6640  59 
Textbook Waveform Energy
• Waveform Energy (Symbol Autocorrelation)

T
Ei   si t   dt
2
• Matched Filter
t
z t   r t   ht    r    ht     d

ht   u t   s * T  t 
t
z t    s   s * T  t     d
0
T T T
z T    s   s T  T     d   s   s    d   s    d
* * 2
0 0 0
ECE 6640 Correlation 60

Signal Power vs. Bit Energy
• For continuous time signals, power is a normal
way to describe the signal.
• For a discrete symbol, the “power” is 0 but the
energy is non-zero
– Therefore, we would like to describe symbols in terms
of energy not power
• For digital transmissions how to we go from
power to energy?
– Power is energy per time, but we know the time
duration of a bit. Noise has a bandwidth.
1
S R  Eb 
Tb 61
Energy and Power
2 E
• For s t    cos2  f c  t   
T
• The average power and energy per bit becomes

T  2 E 2

Eb   E   cos2  f c  t      dt
 T 
0
 
2 E
 
T 2
   E cos 2 2  f c  t     dt A2
1  2 E  E
T 0 P     

2 2  T  T
2 E  1 cos2  2 f c  t    
T
   E    dt
T 0 2 2 
2 E 1 2 E T
T
    dt   E
T 02 T 2
ECE 6640 62
SNR to Eb/No Reminder
• For the Signal to Noise Ratio
– SNR relates the average signal power and average noise
power (Tb is bit period, W is filter bandwidth)
1
Eb 
S Tb  Eb  1
     
 N  N 0  W  N 0  Tb  W
– Eb/No relates the energy per bit to the noise energy
(equal to S/N times a time-bandwidth product)
 Eb   S  W  S 
          Tb  W 
 N0   N  R b  N 
If you want a higher Eb/No, increase Tb.

(Changing W changes the SNR too!) 63
Symbol Detection
• Baseband detection and BER defined in the previous

chapter.
• The following are from ABC Chap. 14
ECE 6640 64
Optimum binary detection
(a) parallel matched filters (b) correlation detector: Figure 14.2-3
65
Conditional PDFs
Figure 14.2-2
66
Bandpass binary receiver
Figure 14.2-1
• Using superposition of the “parallel matched filters”, the

BPF is the difference of the two filters.
hBPF t   h1 t   h0 t 
• This results in an optimal binary detector

67
Binary Receiver
h1 t   K  s1 T  t 
hBPF t   h1 t   h0 t 
h0 t   K  s0 T  t 
• OOK
hBPF t   h1 t   K  s1 T  t   cos2  f c  T  t 
• BPSK
hBPF t   h1 t   h0 t   2  h1 t   2  K  s1 T  t   cos2  f c  T  t 
• BFSK
hBPF t   h1 t   h0 t   K  s1 T  t   K  s0 T  t 
hBPF t   cos2   f c  f d   T  t   cos2   f c  f d   T  t 
hBPF t   2  sin 2  f c  T  t   sin 2  2  f d  T  t  68
Correlation receiver for
OOK or BPSK
Figure 14.2-4
• Since both optimal filters consist of cosine waveforms,

mix and integrate instead of filter an optimally sample.
– Note that the integrator can be a rectangular window filter that is
optimally sampled. (Provides functionality near synchronization as
well.)
69
Optimal Parallel Matched Filter Receiver
Error Analysis
T 
E   s1 t   s0 t   dt 
2
2
 z1  z0 
   0 
 2    max 2  N0
• Evaluating the expected value
T  T  T  T 
E   s1 t   s0 t   dt   E   s1 t   dt   2  E   s1 t   s0 t   dt   E   s0 t   dt 
2 2 2
0  0  0  0 
T 
E   s1 t   s0 t   dt   E1  2  E10  E0
2 Eb  E1  E0  2
0 
2
 z1  z0  2  Eb  2  E10 Eb  E10
   
 2    max 2  N 0 N0
70
Optimal Parallel Matched Filter Receiver
Error Analysis
T 
 E   s1 t   s0 t   dt 
Eb
E10    Eb 
E1  E0 0 
2
 z1  z0  Eb
• OOK E10  0   
 2    max N 0
2
 z1  z0  2  Eb
• PSK E10   1  Eb   
 2    max N0
2
• FSK E10  0  z1  z0 

E
  b
 2    max N 0
71
Generalized Probability of Error
• Using the optimal BPF filter and sampling for
each symbol, the relationship will be based on:
Eb  E10 Eb  1  
2
 z1  z0 
   
 2    max N0 N0
• The BER is then based on

z  z   E  1   
Pe  Q  1 0   Q  b 
 2   N 0 
• Therefore picking arbitrary symbols is possible,

but the symbol correlation coefficient will drive
the BER performance.
72
Generalized FSK
s0 t   Ac  cos2   f c  f d   t  
s1 t   Ac  cos2   f c  f d   t  
T 
E10  Ac  E   cos2   f c  f d   t  cos2   f c  f d   t   dt 
2
0 
2 T
E10  c   cos2  2 f c  t   cos2  2 f d  t  dt
A
2 0
T
E 1
E10  b    expi 2  2 f d  t   exp i 2  2 f d  t  dt
T 2 0
Eb 1  expi 2  2 f d  T  exp i 2  2 f d  T  
  Eb     
T 2  i 2  2 f d i 2  2 f d 
Eb sin 2  2 f d  T   f 
   Eb   Eb  sinc4  f d  T   Eb  sinc 4  d 
T 2  2 f d  rb 
k
• There are multiple “orthogonal” tone separations. 2 f d  f step 
2T
• The correlation coefficient can go negative! The minimum occurs at
73
approximately sinc(1.22) = -0.166
MATLAB Coherent Receivers
• BASK example code

• BPSK example code
• BFSK example code
ECE 6640 74
Noncoherent Binary Systems
• Synchronous coherent receiver can be very difficult to

design.
• Can noncoherent systems be more easily designed without
giving up significant BER performance?
– For a 1-2 dB Eb/No performance loss, YES!
75
Noncoherent OOK receiver
Figure 14.3-2
• Using an envelope detector, the noise pdf for a zero

symbol becomes Rician and is non-longer Gaussian.
• The noise pdf for a one symbol remains Gaussian
76
Conditional PDFs for
noncoherent OOK
Figure 14.3-3
Pe 0 Vopt   Pe1 Vopt 
 E 
Pe 0  exp  b 
 2  N0 
1 1 1  E 
Pe   Pe 0  Pe1    Pe 0    exp  b 
2 2 2  2  N0 
Ac
Vopt 
2
77
Noncoherent detection of binary FSK
Figure 14.3-5
78
Noncoherent FSK
• Qualitative comments
– Using envelope detectors on each symbol output, the Rician error
distribution effects the z detection statistic.
1  E 
Pe   exp  b 
2  2  N0 
79
Binary error probability curves
(a) coherent BPSK (b) DPSK (c) coherent OOK or FSK (d) noncoherent
FSK (e) noncoherent OOK: Figure 14.3-4
0
BER Simulation for BPSK and BFSK
10
-1
10
-2
10
BER -3
10
-4
10
BPSK simulation
-5
10 BPSK (theoretical)
BFSK simulation
BFSK (theoretical)
-6
10
0 2 4 6 8 10 12 14 16
E b/No (dB)
80
Binary error probability curves
(a) coherent BPSK (b) DPSK (c) coherent OOK or FSK
(d) noncoherent FSK (e) noncoherent OOK
Figure 14.3-4
81
Detection for M-ary Systems
• Determine the detection statistic for all symbols
• Select the maximum statistic
• Decode the binary values from the selected symbol
• Notes:
– M-ASK and M-PSK symbols may no longer be orthogonal
– M-FSK symbols may be orthogonal, but the bandwidth W must
increase to contain the symbols.
82
Quadrature-carrier receiver with
correlation detectors
Figure 14.4-1
• Applicable for:
– M-QAM
– M-PSK
83
Carrier synchronization
for quad-carrier receiver
Figure 14.4-2
84
Coherent M-ary PSK receiver
Figure 14.4-3
PreDecode, Es /N0 (dB)=19

20
15
10
Imag
0
• MPSK_Demo.m -5
-10
– Fixed N0, varying signal Eb -15
-20
-20 -15 -10 -5 0 5 10 15 20
Real
85
Decision thresholds for M-ary PSK
Figure 14.4-4
PreDecode, Es /N0 (dB)=19

20
15
10
Imag
0
-5
-10
-15
-20
-20 -15 -10 -5 0 5 10 15 20
Real
86
PSK signal constellations
(a) M=4 (b) M=8
Figure 14.5-1
• MPSK Symbols are typically “Gray-code” encoded prior

to transmission
– In the Gray-code, adjacent symbols are only different by 1 bit
value!
87
MPSK Eb/N0 Examples
PreDecode, Es /N0 (dB)=1 PreDecode, Es /N0 (dB)=9
8 10
8
6
6
4
4
2
2
Imag
Imag
0 0
-2
-2
-4
-4
-6
-6
-8
-8 -10
-8 -6 -4 -2 0 2 4 6 8 -10 -8 -6 -4 -2 0 2 4 6 8 10
PreDecode, Es /N0 (dB)=19 Symbol Error Rate, M=8 Bit Error Rate, M=8
Real 0 Real 0
20 10 10
15
-1 -1
10 10
10
5 -2 -2
10 10
Imag
SER
BER
0
-3 -3
10 10
-5
-10
-4 -4
10 10
-15
-20
-20 -15 -10 -5 0 5 10 15 20 -20 -10 0 10 20 -20 -10 0 10 20
88
Real Es/No (dB) Eb/No (dB)
Simulated Performance MPSK
• MPSK_Ber and MPSK_PP_Plot
0
MPSK Symbol Error Rate 0
MPSK Bit Error Rate
10 10
-1 -1
10 10
-2 -2
10 10
-3 -3
10 10
SER
BER
-4 -4
10 M=2 Sim 10 M=2 Sim
M=2 Bound M=2 Bound
-5 M=4 Sim -5 M=4 Sim
10 M=4 Bound 10 M=4 Bound
M=8 Sim M=8 Sim
-6 M=8 Bound -6 M=8 Bound
10 10
M=16 Sim M=16 Sim
M=16 Bound M=16 Bound
-7 -7
10 10
-5 0 5 10 15 20 25 -5 0 5 10 15 20
Es /N0 (dB) Eb/N0 (dB)
89
Simulated Performance MFSK
• MFSK_Ber and MFSK_PP_Plot
0
MFSK Symbol Error Rate 0
MFSK Bit Error Rate
10 10
-1 -1
10 10
-2 -2
10 10
-3 -3
10 10
SER
BER
-4 -4
10 M=2 Sim 10 M=2 Sim
M=2 Bound M=2 Bound
-5 M=4 Sim -5 M=4 Sim
10 M=4 Bound 10 M=4 Bound
M=8 Sim M=8 Sim
-6 M=8 Bound -6 M=8 Bound
10 10
M=16 Sim M=16 Sim
M=16 Bound M=16 Bound
-7 -7
10 10
0 2 4 6 8 10 12 14 16 -5 0 5 10 15
Es /N0 (dB) Eb/N0 (dB)
90
Comparing MPSK and MFSK
• MPSK
– More Eb/N0 required for higher M for symbol error rate
– 2- and 4-PSK have the same BER
• Otherwise higher BER for higher M
• MFSK
– More Eb/N0 required for higher M for symbol error rate,
BUT it does not increase as fast as MPSK
– Less Eb/N0 required for higher M for BER!
– How could this be?

• The symbols are all orthogonal!
91
M-ary QAM system
(a) transmitter (b) receiver (c) square signal

constellation and thresholds with M=16
Figure 14.4-8
92
Performance comparisons of M-ary
modulation systems
Pbe  104
93
MATLAB Coherent Receivers
• MPSK example code

• MFSK example code
• QAM example code
ECE 6640 94
Notes on BER
• For MPSK and QAM

– Sklar
• QAM p. 565
• MPSK p. 229-230
– J.G. Proakis & M. Salehi, Digital Communications, 5th ed.
• QAM p. 196-200
• MPSK p. 190-195
– Jianhua Lu; Letaief, K.B.; Chuang, J. C-I; Liou, M.-L., "M-PSK
and M-QAM BER computation using signal-space concepts,"
Communications, IEEE Transactions on , vol.47, no.2, pp.181,184,
Feb 1999.
ECE 6640 95
QAM BER Computation
% Sklar (bit error rate)

PB1(:,ii) = 2*((1-L^-1)/log2(L))*Q_fn(sqrt(3*log2(L)*2*Es_No/((M-1)*bitpersym)) );
% Proakis (symbol error rate)
PB2(:,ii) = 2*(1-L^-1)*Q_fn(sqrt(3*log2(M)*Es_No/((M-1)*bitpersym)));
PB2(:,ii) = 2*PB2(:,ii).*(1-0.5*PB2(:,ii));
PB2(:,ii) = PB2(:,ii)/bitpersym;
% Lu, Lataief, Chuang, and Liou (bit error rate)
Qsum = 0;
for jj=1:L/2
Qsum=Qsum+Q_fn((2*jj-1)*sqrt(3*log2(M)*Es_No/((M-1)*bitpersym)));
end
PB3(:,ii) = 4*((1-L^-1)/log2(M))*Qsum;
ECE 6640 96
QAM BER Curves
0
BER Composite Plot
10
4 QMA
-1 16 QAM
10
64 QAM
256 QAM
-2
10
Bit Error Rate
-3
10
-4
10
-5
10
-6
10
-7
10
0 5 10 15 20 25 30
Eb/No (dB)
ECE 6640 97
QAM BER Curves Detail/Differences
0
BER Composite Plot
10
4 QMA
16 QAM
64 QAM
256 QAM
Bit Error Rate
-1
10
-2
10
-1 0 1 2 3 4 5 6 7 8 9 10
Eb/No (dB)
ECE 6640 98
MPSK Nyquist Filter BER
SER vs SNR
Sklar 1
10
MPSK Simulation: Theory vs. Simulation
Theory 0
Plot
10
-1
10
-2
10
Symbol Error Rate
-3
10 T4
S4
T8
-4 S8
10
T16
S16
T32
-5
10 S32
T64
S64
-6 T128
10
S128
T256
S256
-7
10
0 5 10 15 20 25 30 35 40 45 50 55
SNR (dB)
ECE 6640 99
MPSK Nyquist Filter BER
BER vs Eb/No
Sklar MPSK Simulation: Theory vs. Simulation
Theory
0
10
Plot -1
10
-2
10
-3
T4
Bit Error Rate
10
S4
T8
-4 S8
10
T16
S16
T32
-5
10 S32
T64
S64
-6 T128
10
S128
T256
S256
-7
10
-5 0 5 10 15 20 25 30 35 40 45
EbNo (dB)
ECE 6640 100

QAM Nyquist Filter BER
SER vs. SNR
Sklar 1
QAM Simulation: Theory vs. Simulation
10
Theory
Plot 0
10
-1
10
-2
Symbol Error Rate
10
-3
10
T4
-4
10 S4
T16
-5 S16
10
T64
-6 S64
10 T256
S256
-7
10
0 5 10 15 20 25 30 35
ECE 6640 SNR (dB) 101
QAM Nyquist Filter BER
BER vs Eb/No
Sklar 0
10
QAM Simulation: Theory vs. Simulation
Theory
Plot -1
10
-2
10
Bit Error Rate
-3
10
-4
10 T4
S4
-5 T16
10 S16
T64
-6 S64
10
T256
S256
-7
10
-5 0 5 10 15 20 25
ECE 6640 EbNo (dB) 102
ECE 6640

Assistant Professor
Chapter 5
5. Communications Link Analysis.
1. What the System Link Budget Tells the System
Engineer.
2. The Channel.
3. Received Signal Power and Noise Power.
4. Link Budget Analysis.
5. Noise Figure, Noise Temperature, and System
Temperature.
6. Sample Link Analysis.
7. Satellite Repeaters.
8. System Trade-Offs.
ECE 6640 2
What is a Link Budget
• An analysis of the entire communications path
– signal, noise, interference, ISI contributions, etc.
– Include gains and losses
• Link Budget
– An estimate of the input to output system performance
– Will the message get communicated?
– What trade-offs can be made and what effect will they have?
ECE 6640 4
The Channel
• The propagation medium of the communicated signal

• Between the transmitting device and the receiving device
(e.g. RF antennas, cable modems, fiber optic transceivers)
• For RF we think of “Free Space”

– An ideal approximation for near-ground, atmospheric RF
transmissions.
– Non ideal atmospheric impairments include:
absorption, reflection, diffraction, scattering.
ECE 6640 5
Error-Performance Degradation
• Established in Chapter 3
– Loss of SNR
– Intersymbol interference
• For Digital Communications

Eb S  W 
  
N0 N  R 
– The relationship between SNR and Eb/No
– SNR relates the average signal power and average noise power
– Eb/N0 relates the energy per bit to the noise energy
– Loss: refers to a loss in signal energy
– Noise: refers to an increase in noise or interference energy
ECE 6640 6
Sources of Signal Loss and Noise
1. Bandlimiting Loss 12. Atmospheric Loss and Noise
2. Intersymbol Interference (ISI) 13. Space Loss
3. Local Oscillator Phase Noise 14. Adjacent Channel Interference
4. AM/PM Conversion (Amplitude 15. Co-channel Interference
variations) 16. Intermodulation Noise
5. Limiter Loss or Enhancement 17. Galactic or Cosmic, Star and
6. Multiple-carrier Intermodulation Terrestrial Noise
Products (non-linear devices) 18. Feeder Line Loss
7. Modulation Loss (message content 19. Receiver Noise
power) 20. Implementation Loss
8. Antenna Efficiency 21. Imperfect Synchronization
Reference
9. Radome Loss and Noise
10. Pointing Loss See Figure 5.1, p. 246.
11. Polarization Loss
ECE 6640 7
Figure 5.1
ECE 6640 8
Gains and Losses to be Discussed
• Antenna Efficiency
• Pointing
• Atmospheric Noise
• Space Loss
• Receiver
ECE 6640 9
Range Equations
• The power density in a sphere
from a “point source” antenna
(surface area of a sphere)
Pt
pr  
4  r 2
• Receiving power collected by
an antenna (using the
effective area of the receiving
antenna so that p(d) can be • Effective Area
collected) total power extracted
A er 
Pt  A er incident power flux density
Pr  pr   A er 
4  r 2
ECE 6640 10
Antenna Efficiency and Gain
• The ratio of the effective area to the actual area

Ae

Ap
• Antenna Gain
maximum power intensity
G
average power intensity over 4 steradians
– From wikipedia: http://en.wikipedia.org/wiki/Steradian

Steradians the SI unit of solid angle. It is used to describe two-
dimensional angular spans in three-dimensional space, analogous to the
way in which the radian describes angles in a plane.
– Note: a sphere has 4 steradians
ECE 6640 11
Effective Radiated Power
• The effective radiated power is the product of the

transmitted power and the antenna gain
EIRP  Pt  Gt
– The same EIRP can be achieved in many ways
• In terms of received power using effective radiated power

Aer
Pr  EIRP 
4  r 2
Aer
Pr  Pt  Gt 
4  r 2
ECE 6640 12
Antenna Gain in terms of Area
• For antennas with a large area as compared to a signal

wavelength
Note:
4  Aer 2
4  Aer  f
Gr   Gr  2
 2
c2 Aer 
4
• Antenna Reciprocity
– For given antenna and carrier wavelength, the transmitting and
receiving antenna gains are identical.
• The effective area of an isotropic antenna (equal

transmission in all directions)
4  Aer 2 c2
Gr  1   Ae  
ECE 6640 2 4 4  f 2 13
Antenna Beamwidth
• Since an isotropic antenna is defined as having a gain of 1,

the area ratio of the antenna beam pattern from maximum
to -3dB to the area of the sphere is often an estimate of the
antenna gain.
• For an antenna with a half power beamwidths in two
planes the directivity, D, (and gain) are
4
DG
x  y
• For a /4 beam

4 43
DG   20.37
  
4 4
ECE 6640 14
Received Power in EIRP
• For an isotropic receiving antenna, the received power is

A e isotropic
Pr iso  EIRP 
4  r 2
2 EIRP EIRP
Pr iso  EIRP   

42  r 2 4  r 2 Ls


• Where Ls is called the “free-space” or “path” loss
– Note: It is defined based on an isotropic antenna with G=1!

L s  4  r

 
2
 4  r  f
c

2
ECE 6640 15
The Friis Transmission Equation
• The received signal power can be defined as

EIRP Pt  G t  G r
Pr  Gr 
Ls 
4  r

2

• There is a family of relationships (pick your use)
Pt  G t  A er
Pr 
4  r 2 Pt  A et  A er Pt  A et  A er
Pr  
P A G
 r
2 2
  2
c  r2
f
Pr  t et 2 r
4  r
ECE 6640 16
Path Loss Considerations
• Path Loss is defined using an isotopic receiving antenna


L s  4  r

 
2
 4  r  f
c

2
• The received flux density is strictly a function of distance

EIRP
pd  
4  r 2
• For large “effective area” receiving antennas
EIRP  A er
Pr 
4  r 2
ECE 6640 17
Path Loss Considerations (2)
• The effective area for G=1 receiving antennas change with

frequency
A er 
2  
c
 f
2
4 4
Frequency Area Path Loss 1km

3 kHz 7.96E+08 meter^2 -18.02 dB
3 MHz 7.96E+02 meter^2 41.98 dB
3 GHz 7.96E-04 meter^2 101.98 dB
ECE 6640 18
Radio Receiver Consideration
• Receivers collect signals, interference, and noise

• Signals-of-Interest (SOI) will require gain and filtering
prior to or as part of the signal processing
• The noise collected by the receiver will be processed along
with the signal but will be limited by filtering
• The electrical components will add their own noise to the
processed signals.
• Therefore, we need to discuss:

– Cascade gain stages
– Cascaded noise effects and component noise figures
ECE 6640
– Bandwidth effects on thermal noise power 19
RFID Receiver Downconversion
• ISM Band
Downconversion
(902-928 MHz)
– Only mixing and filters
shown
• High-side Los
– Synthesizer provides center
frequency selection
• IF filter sets bandwidth
• LPF for ADC anti-aliasing
• Convert to fs/4 for post-
ADC complex processing
– Fs > 4 x fmax
ECE 6640 20
Cascaded Gain
• Multiple the gain (loss) of each stage together

– If gain in dB, add the gains (in dB) and subtract the losses (in dB)
G prede mod dB  G RF dB  G1stMixer dB  G IF1 dB  G 2 ndMixer dB  G IF 2 dB
– If the mixers have loss instead of gain (passive mixers)
G prede mod dB  G RF dB  L1stMixer dB  G IF1 dB  L 2 ndMixer dB  G IF 2 dB
Linear gain is multiplicative

Gain in dB is additive
ECE 6640 21
Noise Figure
• The noise figure is a measure of the additional noise that is

added by any circuit element.
– Effective additional input noise …
xt  y t 
Caution,
PSin Noise Figure is
F
SNRin

N in often referred to in
SNRout G  PSin dB instead as a
G  N in  N amp 
linear term
SNRin N in  N amp N amp

F   1
SNRout N in N in
ECE 6640 22
Cascaded Noise Figure
• The noise figure is a measure of the additional noise that is

added by any circuit element.
– Effective additional input noise …
xt  y t 
PSin
SNR in N in
F 
SNR out G1  G 2  PSin

G 2  G1  N in  N amp1   N amp 2 
N amp 2
F 

SNRin G2  G1  N in  N amp1   N amp 2

 N in  N amp1 
G1
SNRout G1  G2  N in N in
N amp1 1 N amp 2 1 F 1
ECE 6640
F  1    1  F1  1   F2  1  F1  2 23
N in G1 N in G1 G1
Basic Receiver
Bandpass Bandpass Lowpass
Amplifier
Filter Filter Filter
x c t  x Pr eD t  x M t 
Demod
• RF Filter removes images

• Low Noise Amplifier
Tuning • Mixer to IF
• IF BPF sets the system BW
• Mixer to baseband
cos2  f LO1  t  cos2  f LO2  t 
• Baseband LPF to remove mixing
products
GPr eD dB   GRF  BPF dB   GLNA dB   G1stMixer dB   GBPF dB   G2 ndMixer dB   GLPF dB 
F1stMixer  1 F 1 FBPF  1
FPr eD  FRF  BPF   1stMixer 
GRF  BPF G Amp GRF  BPF G Amp GRF  BPF G Amp G1stMixer
F2 ndMixer  1 FLPF  1
 
ECE 6640 GRF  BPF G Amp G1stMixer GBPF GRF  BPF G Amp G1stMixer GBPF G2 ndMixer 24
Thermal Noise Power
• Modeled as additive white Gaussian noise (AWGN)

N   T  B
– Where N is the noise power
– κ is Boltzmann’s constant
– T is absolute temperature in degrees Kelvin
– B is the bandwidth in Hertz
  228.6 dBW / K  Hz
T0  290K IEEE ref
N 0    T0  1.38e  23  290  4.00e  21
ECE 6640
N 0  204 dBW / Hz  174 dBm / Hz 25
Receiver Operating Characteristics
• Sensitivity – minimum input value

• Dynamic Range – usable signal range
• Selectivity – filter out adjacent noise and interference
• Adjacent Channel Interference (ACI) Rejection
and Image Rejection
• Noise Figure
Building a performance diagram for a software radio

Input to ADC input
ECE 6640 26
FM Radio Design Diagram
• FM receiver
• 200 kHz BW
• 12-bit ADC with 10-bit
performance
• Multiple signal environment
• SOI detection threshold
• ROC
– Sensitivity -103 dBm
– Dynamic Range 41 dB
– Gain 63 dB
– NF 10 dB
– Selectivity: based on IF filter
– ACI: filter attenuation at n
channels away (n x 200 kHz)
ECE 6640 27
FM Radio Design Diagram
• FM receiver
• 200 kHz BW
• 12-bit ADC with
10-bit
performance
• Multiple signal
environment
• SOI detection
threshold
ECE 6640 28
Putting It All Together
• For dedicated communication systems, link budgets are

defined
– System Engineer’s responsibility to guarantee that successful
communication will occur.
– Examples: WiFi access point locations, satellite communications,
FM radio station coverage areas (and transmitting antenna siting),
Cellular Telephone System Base Station Siting, etc.
• If you build it, will it be useful?

– Reliability, design margin, upgrades, component replacement
ECE 6640 29
Table 5.2 Earth Terminal to Satellite
Link Budget
ECE 6640 30
Table 5.3 Link Budget Example
ECE 6640 31
ECE 6640

Assistant Professor
Chapter 6
1. Waveform Coding.
2. Types of Error Control.
3. Structured Sequences.
4. Linear Block Codes.
5. Error-Detecting and Correcting Capability.
6. Usefulness of the Standard Array.
7. Cyclic Codes.
8. Well-Known Block Codes.
ECE 6640 2
Encoding and Decoding
• Data block of k-bits is encoded using a defined algorithm

into an n-bit codeword
• Codeword transmitted
• Received n-bit codeword is decoded/detected using a
defined algorithm into a k-bit data block
Data
Encoder Codeword
Block
Data
Decoder Codeword
Block
ECE 6640 5
Waveform Coding Structured Sequences
• Waveform Coding:
– Transforming waveforms in to “better” waveform representations
– Make signals antipodal or orthogonal
• Structured Sequences:
that contain redundant bits
– Use redundancy for error detection and correction
– Bit sets become longer bit sets (redundant bits) with better
“properties”
– The required bit rate for transmission increases.
ECE 6640 6
• Antipodal
T
1 for i  j
z ij    s i t   s j t   dt  
d  2  Eb 1
E 0  1 for i  j
• Orthogonal
T
1 for i  j
z ij    s i t   s j t   dt  
1
d  2  Eb
E 0 0 for i  j
ECE 6640 7
M-ary Signals Waveform Coding
• Symbol represents k bits at a time

– Symbol selected based on k bits
– M waveforms may be transmitted M  2k
• Allow for the tradeoff of error probability for bandwidth

efficiency
• Orthogonality of k-bit symbols

– Number of bits that agree=Number of bits that disagree
 sumb   
K N
i
 b   sum b ik  b kj
j
k k
1 for i  j
z ij  k 1 k 1

K 0 for i  j
ECE 6640 8
Hadamard Matrix Orthogonal Codes
H D 1 H D 1 
HD  
H D 1 H D 1 
• Start with the data set for one bit and generate a Hadamard
code for the data set
0  0 0 
1  H 1  0 1 
   
0 0 0 0 0 0
0 1 H1   0
  H1 1 0 1
 H2    
1 0  H1 H1  0 0 1 1
  0 1 1 0 
1 1
1 for i  j
z ij  
ECE 6640 0 for i  j 9
Hadamard Matrix Orthogonal Codes (2)
0 0 0 0 0 0 0 0 0 0 0
0 1 0
 0
 1 0 1 0 1 0 1
0 1 0 0 0 1 1 0 0 1 1
 
0 1 1 H H 2  0 1 1 0 0 1 1 0

Data Set Codeword
 H3   2  
1 0 0 H 2 H 2  0 0 0 0 1 1 1 1
  0
1 0 1 1 0 1 1 0 1 0
 
1 1 0 0 0 1 1 1 1 0 0
   
1 1 1 0 1 1 0 1 0 0 1
0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0

 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
See Hadamard in 0

0
0 1 0
0 1 1
 0

0 1 1 0 0 1 1 0 0 1 1 0 0 1 1

0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0
1 0 0
MATLAB 
0 1 0 1
 0
0
0
1
0
0
0
1
1
1
1
0
1
1
1
0
0
0
0
1
0
0
0
1
1
1
1
0
1
1
1
0
0  
1 1 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0
 
H 3  0 1 
H4=hadamard(16) 0
1
1 1 1
0 0 0
H
 H4   3
H3

H 3  0
1
0
1
0
0
0
1
0
0
0
0
0
1
0
0
1
1
1
1
1
0
1
1
1
0
1
0
1 1
 
1 0 0 1 0 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0
 
H4xcor=H4'*H4=H4*H4' 
1 0 1 0

0
0
0 1 1 0 0 1 1 1 1 0 0 1 1 0 0
1 0 1 1 1 1 0 0 1 1 0 1 0 0 1 1 0 0 1
   
0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0
ans = 16*eye(16) 1
1
1 0 0
1 0 1
0

1 0 1 1 0 1 0 1 0 1 0 0 1 0 1

 
1 1 1 0 0 0 1 1 1 1 0 0 1 1 0 0 0 0 1 1
1 0 0  10
1 1 1 
ECE 6640 1 1 0 1 0 0 1 1 0 0 1 0 1 1

Hadamard Matrix Orthogonal Codes (2)
• The Hadamard Code Set generates orthogonal code for the

bit representations Letting
1 for i  j
z ij   0 = +1V
0 for i  j
1 = -1V
• For a D-bit pre-coded symbol a 2^D bit symbol is
generated from a 2^D x 2^D matrix
– Bit rate increases from D-bits/sec to 2^D bits/sec!
• For equally likely, equal-energy orthogonal signals, the
probability of codeword (symbol) error can be bounded as
 Es  M  2D
PE M   M  1  Q 

 N0  Es  D  E b
ECE 6640 11
Symbol error to Bit Error
• Discussed in chapter 4 (Equ 4.112) – k-bit set (codeword)

– 4.9.3 Bit error probability versus symbol error probability for
orthogonal signals
PB k  2 k 1 PB M 
M
 k M  2k  2
PE k  2  1 PE M  M  1
• Substituting bit error probability for symbol error
probability bounds previously stated
 Es 
PE M   M  1  Q 

 N0 
 k  Eb   M   E S 
 
PB k   2 k 1 
 Q 
 PB M      Q
 N0   2   N 0 
ECE 6640 12
Biorthogonal Codes
• Code sets with cross correlation of 0 or -1

1 for i  j


z ij   1 for i  j , i  j  M
2

0 for i  j , i  j  M
2
• Based on Hadamard codes (2-bit data set shown)

0 0 0 0
H D 1  0 1  1
BD      H1   0
 D 1 
H  B2    
1 0   H1  1 1
   
1 1 1 0
ECE 6640 13
Biorthogonal Codes (2)
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
 0 0 1 0
 1 0 1 0 1 0 1
0 0 1 0 1 0 1
0

0 1 0

0

0 1 1 0 0 1 1

  0
0
0 1 1 0 1 1 0 0 1 1 0
1 0 0 0 0 0 0 1 1 1 1
0 1 0 0 0 1 1    
    0
0
1 0 1
1 1 0
0
0
1 0 1 1 0 1 0
0 1 1 1 1 0 0
0 1 1  H 2  0 1 1 0    
 B3     0 1 1 1  H 3  0 1 1 0 1 0 0 1
1 0 0   H 2  1 1 1 1 1

0 0 0

 B4     
 H 3  1 1 1 1 1 1 1 1

    1 0 0 1 1 0 1 0 1 0 1 0
1 0 1 1 0 1 0 
1 0 1 0
 
1 1 0 0 1 1 0 0

1 1 0 1 1 0 0
1

0 1 1

1

0 0 1 1 0 0 1

    1
1
1 0 0

1
1
1 1 1 0 0 0 0
0 1 0 0 1 0 1
1 1 1 1 0 0 1 
1 0 1
  
1 1 1 0 1 1 0 0 0 0 1 1
1  1 0 0 1 0 1 1 0
 1 1 1 
ECE 6640 14
B3 Example
H2=hadamard(4)
B3=[H2;-H2]
B3*B3'
ans =
4 0 0 0 -4 0 0 0
0 4 0 0 0 -4 0 0
0 0 4 0 0 0 -4 0
0 0 0 4 0 0 0 -4
-4 0 0 0 4 0 0 0
0 -4 0 0 0 4 0 0
0 0 -4 0 0 0 4 0
0 0 0 -4 0 0 0 4
ECE 6640 15
• Biorthogonal codes require half as many code bits per code

word as Hadamard.
– The required bandwidth would be one-half
– Due to some antipodal codes, they perform better
• For equally likely, equal-energy orthogonal signals, the
probability of codeword (symbol) error can be bounded as
 Es   2  Es  M  2 k 1
PE M   M  2   Q
 
  Q 

 N0   N0  Es  k  E b
 
PB M   PE M , for M  8
2
ECE 6640 16
• Comparing to Hadamard to Biorthogonal

– Note that for D bits, M is ½ the size for Biorthogonal
M  2 D 1
 M  2   E s   2  Es 
Biorthogonal
PB M      Q   Q 
 M  2k
 2   N0   N0 
Es  k  E b
 M   E S  M  2k
Hadamard
PB M      Q
 2   N 0  Es  k  E b
ECE 6640 17
Waveform Coding System Example
• For a k-bit precoded symbol, one of M generators creates

an encoded waveform.
• For a Hadamard Code, 2^k bits replace the k-bit input
ECE 6640 18
Waveform Coding System Example (2)
• This bit sequence can be PSK modulated (an antipodal

binary system) for transmission over M time periods of
length Tc. T  2 k  TC  M  TC
• The symbol rate is 1/M*Tc RS  1  1 k

T 2  TC
• For real-time transmission, the code word transmission rate

and the k-bit input data rate must be identical.
– Codeword bits shorter than original message bits
Rk  1  RS  1  1 k
 k  T
k  Tb T 2  TC k
Tb  2 C
ECE 6640
T  k  Tb  2  TC
k
19
• The coherent detector now works across multiple PSK bits

representing a symbol
ECE 6640 20
Stated Improvement Tb   k  T
2 k
C
• For k=5, detection can be accomplished with

about 2.9 dB less Eb/No!
– Required bandwidth is 32/5=6.4 times the initial data word rate.
• Homework problem 6.28
– Compare PB of BPSK to the PB for orthogonally encoded symbols
at a predefined BER. Solution set uses 10^-5.
ECE 6640 21
Types of Error Control
• Error detection and retransmission

– Parity bits
– Cyclic redundancy checking
– Often used when two-way communication used. Also used when
low bit-error-rate channels are used.
• Forward Error Correction

– Redundant bits for detection and correction of bit errors
incorporated into the sequence
– Structured sequences.
– Often used for one-way, noisy channels. Also used for real-time
sequences that can not afford retransmission time.
ECE 6640 22
Structured Sequence
• Encoded sequences with redundancy
• Types of codes
– Block Codes or Linear Block Codes (Chap. 6)
– Convolutional Codes (Chap. 7)
– Turbo Codes (Chap. 8)
ECE 6640 23
Channel Models
• Discrete Memoryless Channel (DMC)

– A discrete input alphabet, a discrete output alphabet, and a set of
conditional probabilities of conditional outputs given the particular
input.
• Binary Symmetric Channel
– A special case of DMC where the input alphabet and output
alphabets consist of binary elements. The conditional probabilities
are symmetric p and 1-p.
– Hard decisions based on the binary output alphabet performed.
• Gaussian Channels
– A discrete input alphabet, a continuous output alphabet
– Soft decisions based on the continuous output alphabet performed.
ECE 6640 24
Code Rate Redundancy
• For clock codes, source data is segmented into k bit data

blocks.
• The k-bits are encoded into larger n-bit blocks.
• The additional n-k bits are redundant bits, parity bits, or
check bits.
• The codes are referred to as (n,k) block codes.
– The ratio of redundant to data bits is (n-k)/k.
– The ratio of the data bits to total bits, k/n is referred to as the code
rate. Therefore a rate ½ code is double the length of the underlying
data.
ECE 6640 25
Binary Bit Error Probability
• Defining the probability of j errors in a block of n

symbols/bits where p is the probability of an error.
n j
P j , n      p  1  p 
n j
 j
 n!  j
P j , n      p  1  p n  j
 j!n  j ! 
Binomial Probability Law

• The probability of j failures in n Bernoulli trials
E x   n  p  
E x 2  n  p  1  p 
ECE 6640 26
Code Example Triple Redundancy
• Encode a 0 or 1 as 000 or 111

– Single bit errors detected and corrected
 3!  j
P j ,3     p  1  p 3 j
 j!3  j ! 
• Assume p = 10-3
 3! 
P0,3     0.999   0.999   0.997
3 3
 0!3  0 ! 
 3! 
P1,3     0.001  0.999   3  0.001  0.999  0.002994
1 2 1 2
 1!3  1! 
 3! 
P2,3     0.001  0.999   3  0.001  0.999   0.000002997
2 1 2 1
 2!3  2 ! 
 3! 
P3,3     0.001  0.001  10 9
3 3
 3!3  3! 
Code Example Triple Redundancy
• Probability of two or more errors

Pr  x  2   P2,3  P2,3  2.998 10 6  3 10 6
• Probability of one or no errors
Pr  x  2   P0,3  P1,3  1  Pr x  2   1  2.998 10 6  0.999997
• If the raw bit error rate of the environment is p=10-3 …

• The probability of detecting the correct transmitted bit
becomes 3*10-6
Parity-Check Codes
• Simple parity – single parity check code

– Add a single bit to a defined set of bits to force the “sum of bits” to
be even (0) or odd (1).
– A rate k/k+1 code
– Useful for bit error rate detection
• Rectangular Code (also called a Product Code)
– Add row and column parity bits to a rectangular set of bits/symbols
and a row or column parity bit for the parity row or column
– For an m x n block, m+1 x n+1 bits are sent
– A rate mn/(m+1)(n+1) code
– Correct signal bit errors!
ECE 6640 29
(4,3) Single-Parity-Code
• (4,3) even-parity code example

– 8 legal codewords, 8 “illegal” code words to detect single bit errors
– Note: an even number of bit error can not be detected, only odd
– The probability of an undetected message error
n / 2 ( n even ) n / 2 ( n even )
( n 1) / 2 ( n odd ) ( n 1) / 2 ( n odd )
  2j
 P2 j, n     p  1  p n  2 j
n!
 
j 1 j 1  2 j!n  2 j ! 
Message Parity Codeword
000 0 0 000
001 1 1 001
010 1 1 010
011 0 0 011
100 1 1 100
101 0 0 101
ECE 6640
110 0 0 110 30
111 1 1 111
Matlab Results
p = 1e-3;
n=
4.00000000000000e+000
n=4
Perror =
0.00000000000000e+000
Perror = 0
jfactor =
for j=2:2:n
6.00000000000000e+000
jfactor = factorial(n)/(factorial(j)*factorial(n-j))
jPe = 2 bit errors
jPe = jfactor * p^j * (1-p)^(n-j)
5.98800600000000e-006
Perror = Perror + jPe
Perror =
end
5.98800600000000e-006
jfactor =
1.00000000000000e+000
jPe = 4 bit errors
1.00000000000000e-012
Perror =
5.98800700000000e-006
Probability of an undetected message error
ECE 6640 31
Matlab Results
• Computing the probability of a message error

– Channel bit size the parity is added to 2 to 16
– Raw BER = 10^-3
– Compare with parity and without (n x p)
-1
Message Error Probability for BER = 10-3
10
no parity
parity
-2
10
Prob of Error Detection
-3
10
-4
10
-5
10
-6
ECE 6640 10 32
2 4 6 8 10 12 14 16
Channel Bits
Rectangular Code
• Add row and column parity bits to a rectangular set of

bits/symbols
and a row or column parity bit for the parity row or column
• For an m x n block, m+1 x n+1 bits are sent
• A rate mn/(m+1)(n+1) code
• Correct signal bit errors!
– Identify parity error row and column, then fix the “wrong bit”
ECE 6640 33
Rectangular Code Example
• 5 x 5 bit array  25 total bits

• 6 x 6 encoded array  36 total bits
• A (36,25) code
• Compute probability that there is an undetected error

message
– The block error or word error becomes: j errors in blocks of n
symbols and t the number of bit errors. (note t=1 corrected)
n
n j
PM       p  1  p n j
j t 1  j 
 n  t 1
PM     p  1  p n t 1
ECE 6640
 t  1 34
Error-Correction Code Tradeoffs
• Improve message error/bit error rate performance

•
• Tradeoffs
– Error performance vs. bandwidth
• More bits per sec implies more bandwidth
– Power Output vs. bandwidth
• Lowering Eb/No changer BER, to get it back, encode and increase
bandwidth
• If communication need not be in real time … sending more
bits increases BER at the cost of latency!
ECE 6640 35
Coding Gain
• For a given bit-error probability,

the relief or reduction in required Eb/No
 Eb   Eb 
CodeGaindB   
 
dB  
 dB 
 No uncoded  N o  coded
ECE 6640 36
Data Rate vs. Bandwidth
• From previous SNR discussions

P W 
 W  T   S   , R 
Eb PS 1

N 0 PN PN  R  T
Eb PS  W  PS  1 
     
N 0 N 0 W  R  N 0  R 
• The tradeoff between Eb/No and bit rate
ECE 6640 37
Linear Block Codes
• Encodeing a “message space” k-tuple into a “code space”

n-tuple with an (n,k) linear block code
– maimize the distance between code words
– Use as efficient of an encoding as possible
– 2k message vectors are possible (for example M k-tuples to send)

– 2n message codes are possible (M n-tuples after encoding)
• Encoding methodology
– A giant table lookup
– Generators that can span the subspace to create the desired linear
independent sets can be defined
ECE 6640 38
Generator Matrix
• We wish to define an n-tuple codeword U

U  m1  V1  m2  V2   mk  Vk
where mi are the message bits (k-tuple elements), and Vi

are k linearly independent n-tuples that span the space.
– This allows us to think of coding as a vector-matrix multiple
with m a 1 x k vector and V a k x n matrix
• If a generator matrix composed of the elements of V can be
defined
V1   v11 v12  v1n 
V  v v22  v2 n 
  
U m G m   2
 m   21
     
   
 k
V  vk 1 vk 2  vkn 
ECE 6640 39
Table 6.1 Example
% Simple single bit parity
n = 6; k = 3; % Set codeword length and message length.
msg = [0 0 0; 0 0 1; 0 1 0; 0 1 1; ...
1 0 0; 1 0 1; 1 1 0; 1 1 1]; % Message is a binary matrix.
gen = [ 1 1 0 1 0 0; ...
0 1 1 0 1 0; ...
1 0 1 0 0 1];
code = rem(msg * gen, 2)
0 0 0 0 0 0
1 0 1 0 0 1
0 1 1 0 1 0
1 1 0 0 1 1
1 1 0 1 0 0
0 1 1 1 0 1
1 0 1 1 1 0
0 0 0 1 1 1
ECE 6640 40
Systematic Linear Block Codes
• The generator preserves part of the message in the

resulting codeword
 p11 p12  p1,( n  k ) 1 0  0  v11 v12  v1n 
p 
p22  0 1  0 v21 v22  v2 n 
G  P I   
21 p2 ,( n  k )
              
  
 pk1 pk 2  pk , ( n  k ) 0 0  1 vk1 vk 2  vkn 
• P is the parity array and I regenerates the message bits
– Note the previous example uses this format
• Matlab generators
– hammgen or
– POL = cyclpoly(N, K, OPT) finds cyclic code generator polynomial(s) for a given
code word length N and message length K.
ECE 6640 41
Systematic Code Vectors
• Because of the structure of the generators
G  P I k 
• The codeword becomes

U  m  G   p | m
where p are the parity bits generated for the message m
ECE 6640 42
Parity Check Matrix
• The parity check matrix allows the checking of a codeword

– It should be trivial if there are no errors …
but if there are?! Syndrome testing

H  I nk PT 
I 
U  H T  m  P | I k  H T  m  P | I k   n  k   0 mr ,n  k
 P 
ECE 6640 43
Parity Check Matrix
• Let the message be I

I 
U  H T  I k  P | I k  H T  P | I k   n  k   P  P   0
 P 
• The parity check matrix results in a zero vector, representing the

“combination” of the parity bits.
[parmat,genmat] = hammgen(3)
rem(genmat * parmat', 2)
[parmat,genmat] = cyclgen(7,cyclpoly(7,4))
rem(genmat * parmat', 2)
ECE 6640 44
Syndrome Testing
• Describing the received n-tuples of U

r U e
– where e are potential errors in the received matrix
• As one might expect, use the check matrix to generate a

the syndrome of r
S  r  H T  U  e   H T  U  H T  e  H T
• but from the check matrix

S  eHT
– if there are no bit errors, the result is a zero matrix!

ECE 6640 45
Syndrome Testing
• If there are no bit error, the Syndrome results in a 0 matrix

• If there are errors, the requirement of a linear block code is
to have a one to one mapping between correctable errors
and non-zero syndrome results.
• Error correction requires the identification of the

corresponding syndrome for each of the possible errors.
– generate a 2^n x n array representing all possible received n-tuples
– this is called the standard array
ECE 6640 46
Standard Array Format (n,k)
• The actual codewords are placed in the top row.

– The 1st code word is an all zeros codeword. It also defines the 1st
coset that has zero errors
• Each row is described as a coset with a coset leader
describing a particular error.
– for an n-tuple, there will be n-k coset leaders, one of which is zero
errors.
ECE 6640 47
Figure 6.11
ECE 6640 48
Decoding with the Standard Array
• The n-tuple received can be located somewhere in the

standard array.
– If there is an error, the corrupted codeword is replaced by the
codeword at the top of the column.
– The standard array contains all possible n-tuples, it is 2^(n-k) x 2^k
in size; therefore, all n-tuples are represented.
– There are 2^(n-k) cosets
ECE 6640 49
The Syndrome of a Coset
• Computing the syndrome of the jth coset for the ith

codeword
S  U i  e j  H T  U i  H T  e j  H T  e j  H T
– the syndrome is identical for the entire coset

– Thus, coset is really a name defined for
“a set of numbers having a common feature”
– The syndrome uniquely defines the error pattern
ECE 6640 50
Error Correct
• Now that we know that the syndrome identifies a coset, we

can identify the coset and perform error correction
regardless of the n-tuple transmitted.
• The procedure
1. Calculate the Syndrome of r
2. Locate the coset leader whose syd4om is identified
3. This is the assumed corruption of a valid n-tuple
4. Form the corrected codeword by “adding” the coset to the
corrupted codeword n-tuple.
ECE 6640 51
Error Correction Example
• see Table61_example
– each step of the process is shown in Matlab for the (6,3) code of
the text
ECE 6640 52
Decoder Implementation
ECE 6640 53
Hamming Weight and Distance
• The Hamming weight of a code/codeword defines the

performance
• The Hamming weight is defined as the number of nonzero
elements in U.
– for binary, count the number of ones.
• The Hamming difference between two codewords, d(U,V),
is defined as the number of elements that differ.
Ucode = w(U)= d(U0,U)
0 0 0 0 0 0 0
1 0 1 0 0 1 3
0 1 1 0 1 0 3
1 1 0 0 1 1 4
1 1 0 1 0 0 3
0 1 1 1 0 1 4
1 0 1 1 1 0 4
0 0 0 1 1 1 3
ECE 6640 54
Minimum Distance
• The minimum Hamming distance is of interest

– dmin = 3 for the (6,3) example
• Similar to distances between symbols for symbol

constellations, the Hamming distance between codewords
defines a code effective performance in detecting bit errors
and correct them.
– maximizing dmin for a set number of redundant bits is desired.
– error-correction capability
– error-detection capability
ECE 6640 55
Hamming Distance Capability
• Error-correcting capability, t bits
 d  1
t   min 
 2 
– for (6,3) dmin=3: t = 1
– Codes that correct all possible sequences of t or fewer errors may

also be able to correct for some t+1 errors.
• see the last coset of (6,3) – 2 bit errors!
– a t-error-correcting (n,k) linear code is capable of correcting 2(n-k)
error patterns!
– an upper bound on the proabaility of message error is
n
n j
PM       p  1  p n j
j t 1  j 
ECE 6640 56
Hamming Distance Capability
• Error-detection capability
e  d min  1
– a block code with a minimum distance of dmin guarantees that all
error patterns of dmin-1 or fewer error can be detected.
– for (6,3) dmin=3: e = 2 bits
– The code may also be capable of detecting errors of dmin bits.

– An (n,k) code is capable of detecting 2n-2k error patterns of length n.
• there are 2k-1 error patterns that turn one codeword into another and
are thereby undetectable.
• all other pattersn should produce non-zero syndrom
• therefore we have the desired number of detectable error patterns
• (6,3)  64-8=56 detectable error patterns
ECE 6640 57
Codeword Weight Distribution
• The weight distribution involves the number of codewords

with any particular Hamming weight
– for (6,3) 4 have dmin=3, 3 have d = 4
– If the code is used only for error detection, on a BSC, the

probability that the decoder does not detect an error can be
computed from the weight distribution of the code.
n
Pnd   A j  p j  1  p 
n j
j 1
– for (6,3) A(0)=1, A(3)=4, A(4)=3, all else A(i)=0
Pnd  4  p 3  1  p   3  p 4  1  p 
4 3
 4  p   p 3  1  p 
3
ECE 6640 58
– for p=0.001: Pnd ~ 3.9 x 10-6
Simultaneous Error
Correction and Detection
• It is possible to trade capability from the maximum
guaranteed (t) for the ability to simultaneously detect a
class of errors.
• A code can be used for simultaneous correction of alpha
errors and detection of beta errors for beta>=alpha
provided
d min      1
ECE 6640 59
Visualization of (6,3) Codeword
Spaces
ECE 6640 60
More to Come
• see http://www.mathworks.com/help/comm/error-
detection-and-correction.html
ECE 6640 61
(8,2) Block Code
ECE 6640 62
ECE 6640

Assistant Professor
Chapter 7
7. Channel Coding: Part 2 .
1. Convolutional Encoding.
2. Convolutional Encoder Representation.
3. Formulation of the Convolutional Decoding Problem.
4. Properties of Convolutional Codes.
5. Other Convolutional Decoding Algorithms.
ECE 6640 2
Waveform Coding Structured Sequences
• Structures Sequences:
that contain redundant bits
– Use redundancy for error detection and correction
• Block Codes are memoryless
• Convolution Codes have memory!
ECE 6640 5
Convolutional Encodings
• The encoder transforms each sequence M into a unique

codeword sequence U=G(m). Even though the sequence m
uniquely defines the sequence U, a key feature of
convolutional codes is that a given k-tuple within m does
not uniquely define its associated n-tuple within U since
the encoding of each k-tuple is not only a function of that
k-tuple but is also a function of the K-1 input k-tuples that
precede is.
• Each k-tuple effects not just the codeword generated when
the value is input, by the next K-1 codewords as well.
– the system was memory
ECE 6640 6
Convolutional Encoder Diagram
• Each message, mi, may be
a k-tuple. (or k could be a
bit)
• K messages are in the
encoder
• For each message input,
an n-tuple is generated
• The code rate is k/n
• We will usually be
working with k=1 and
n=2 or 3
ECE 6640 7
Proakis Convolution Encoder
John G. Proakis, “Digital Communications, 4th ed.,” McGraw Hill, Fourth

ECE 6640 8
Edition, 2001. ISBN: 0-07-232111-3.
Representations
• Several methods are used for representing a convolutional

encoder, the most popular being;
– the connection pictorial
– connection vectors or polynomials
– state diagrams
– tree diagrams
– trellis diagrams.
ECE 6640 9
Connection Representation
• k=1, n=3
• Generator Polynomials
– G1 = 1 + X + X2
– G2 = 1 + X2
• To end a message, K-1 “zero” messages are transmitted.
This allows the encoder to be flushed.
– effective code rate is different than k/n … the actual rate would be
(2+k*m_length)/n*m_length
– a zero tailed encoder ….
ECE 6640 10
Impulse Response of the Encoder
• allow a single “1” to transition through the K stages
– 100 -> 11
– 010 -> 10
– 001 -> 11
– 000 -> 00
• If the input message where 1 0 1
– 1 11 10 11
– 0 00 00 00
– 1 11 10 11
– Bsum 11 10 00 10 11
– Bsum is the transmitted n-tuple sequence …. if a 2 zero tail follows
– The sequence/summation involves superpoition or linear addition.
• The impulse response of one k-tuple sums with the impulse responses of
successive k-tuples!
ECE 6640 11
Convolutional Encoding the Message
• As each k-tuple is input,

an n-tuple is output
• This is a rate ½ encoding
• The “constraint” length is
K=3, the length of the k-tuple
shift register.
• The effective code rate for
m_length = 3 is: 3/10
ECE 6640 12
Proakis (3,1), rate 1/3, K=3 Pictorial
• Generator Polynomials
– G1 = 1
– G2 = 1 + X2
– G3= 1 + X + X2

ECE 6640 13
Edition, 2001. ISBN: 0-07-232111-3.
Polynomial Representation
• Generator Polynomials (also represented in octal)

– G1 = 1 + X + X2  7octal
– G2 = 1 + X2 5octal
• It is assumed to be a binary input.
• There are two generator polynomials, therefore n=2
– each polynomials generates one of the elements of the n-tuple
output
• Polynomial Multiplication can be used to generate output
sequences
• m(X)*g1(X) = (1 + X2)* (1 +X+ X2) = 1 +X+X3+ X4
• m(X)*g2(X) = (1 + X2)* (1 + X2) = 1 + X4
• Output: (11 , 10 , 00, 10, 11) as before
ECE 6640 14
State Representation
• Using the same encoding:
• Solid lines represent 0 inputs
• Dashed lines represent 1 inputs
• The n-tuple output is shown with
the state transition
• It can be verified that 2 zeros
always returns to the same
“steady state”
• Note: two previous k-tuples
provide state, the new k-tuple
drives transitions
ECE 6640 15
Proakis (3,1) K=3
State Diagram
Solid Lines
are 0 inputs
Dashed Lines
are 1 inputs
ECE 6640 JohnG. Proakis, “Digital Communications, 4th ed.,” McGraw Hill, Fourth 16
Edition, 2001. ISBN: 0-07-232111-3.
Tree Diagram
• Input values define where

to go next.
• Each set of branch outputs
transmit complementary
n–tuples
• States can be identified by

repeated level operations
ECE 6640 17
Proakis (3,1) K=3
Polynomial and Tree
g1  1 0 0
g 2  1 0 1
g 3  1 1 1
Edition, 2001. ISBN: 0-07-232111-3.
Trellis Diagram
• The tree structure repeats itself.

• The tree/state diagrams define a finite number of states.
– the tree has ever increasing number of branches to show the
complete path of a message
– the state folds back on itself so observing the complete path is
difficeult
• Can we identify diagrammatically a figure that shows the
state transitions and the entire message path?
• Yes, the Trellis Diagram
ECE 6640 19
Trellis Diagram
• Initial state, state development, continuous pattern

observable, “fully engaged” after K-1 inputs, easily trace
tail zeros to “initial state”.
ECE 6640 20
Proakis (3,1) K=3
Trellis Diagram
Edition, 2001. ISBN: 0-07-232111-3.
A more complicated example follows
• Proakis (3,2) K=2 Encoder

– Pictorial
– Polynomial and Tree
– State Diagram
– Trellis
ECE 6640 22
Proakis (3,2) K=2
Pictorial

ECE 6640 23
Edition, 2001. ISBN: 0-07-232111-3.
Proakis (3,2) K=2
Polynomial and Tree
g1  1 0 1 1
g 2  1 1 0 1
g 3  1 0 1 0
Edition, 2001. ISBN: 0-07-232111-3.
Proakis (3,2) K=2
State Diagram
Solid Lines
are 0 inputs
Dashed Lines
are 1 inputs
Edition, 2001. ISBN: 0-07-232111-3.
Proakis (3,2) K=2
Trellis Diagram
Edition, 2001. ISBN: 0-07-232111-3.
Encoding
• Each approach can be readily implemented in hardware.

• Good codes have been found be computer searches for
each value of the constraint length, K.
• The easy part, now for decoding.
ECE 6640 27
Decoding Convolutional Codes
• As the codes have memory, we wish to use a decoder that

achieves the minimum probability of error … using a
condition called maximum likelihood.
ECE 6640 28
ECE 5820 MAP and ML
• There are various estimators for signals combined with

random variables.
• In general we are interested in the maximum a-posteriori
estimator of X for a given observation Y.
max PX  x | Y  y 
x
– this requires knowledge of the a-priori probability, so that
PY  y | X  x  PX  x 
P X  x | Y  y  
PY  y 
ECE 6640 29
ECE 5820 MAP and ML
• In some situations, we know

PY  y | X  x 
but not the a-priori probabilities.
• In these cases, we form a maximum likelihood estimate
For a maximum likelihood estimate, we perform
max PY  y | X  x 
x
ECE 6640 30
ECE 5820 Markov Process
• A sequence or “chain” of subexperiments in which the

outcome of a given subexperiment determines which
subexperiment is performed next.
Ps0 , s0 ,  , sn   Psn | sn 1  Psn | sn 1  Psn 1 | sn  2   Ps1 | s0  Ps0 
• If the output from the previous state in a trellis is known,

the next state is only based on the previous state and the
new input.
– the decoder can be computed one step at a time to determine the
maximum likelihood path.
• Viterbi’s improvement on this concept.
– In a Trellis, the is a repetition of states. If two paths arrive at the
same state, only the path with the maximum likelihood must be
ECE 6640 maintained … the “other path” can no longer become the ML path! 31
Maximum Likelihood Decodong
• If all input message sequences are equally likely, the

decoder that achieves the minimum probability of error is
the one that compares the conditional probabilities of all
possible paths against the received sequence.
 
P Z | U m '   max
m 
all U
P Z | Um 

– where U are the possible message paths
• For a memoryless channel we can base the computation on
the individual values of the observed path Z
    PZ    Pz 
  
m  m  m 
P Z |U i | Ui ji | u ji
i 1 i 1 j 1
– where Zi is the ith branch of the received sequence Z, zji is the jth
code symbol of Zi and similarly for U and u.
ECE 6640 32
ML Computed Using Logs
• As the probability is a product of products, computation

precision and the final magnitude is of concern.
• By taking the log of the products, a summation may be

performed instead of multiplications.
– constants can easily be absorbed
– similar sets of magnitudes can be pre-computed and/or even scaled
to more desirable values.
– the precision used for the values can vary as desired for the
available bit precision (hard vs. soft values)
ECE 6640 33
Channel Models:
Hard vs. Soft Decisions
• Our previous symbol determinations selected a detected symbol with
no other considerations … a hard decision.
• The decision had computed metrics that were used to make the
determination that were then discarded.
• What if the relative certainty of decision were maintained along with
the decision.
– if one decision influenced another decision, hard decisions keep certainty
from being used.
– maintaining a soft decision may allow overall higher decision accuracy
when an interactions exists.
ECE 6640 34
ML in Binary Symmetric Channels
• Bit error probability

– P(0|1)=P(1|0) = p
– P(1|1)=P(0|0) = 1-p
• Suppose Z and any possible message U differ in dm
positions (related to the hamming distance). Then the ML
probability for an L bit message becomes
 
P Z | U m   p dm  1  p 
L  dm
• taking the log

 
log P Z | U m   dm  log p   L  dm   log1  p 
log PZ | U     dm  log

1 p 
m
  L  log1  p 
ECE 6640  p  35
ML in Binary Symmetric Channels
• The ML value for each possible U is then
  1 p 
log P Z | U m   dm  log   L  log1  p 
 p 
– The constant is identical for all possible U and can be pre-computed
– The log of the probability ratios is also a constant
 
log P Z | U m   dm  A  B
• Overall, we are looking for the possible sequence with a

minimum Hamming distance.
– for hard decisions, we use the Hamming distance
– for soft decisions, we can use the “certainty values” shown in the
previous figure!
ECE 6640 36
Viterbi Decoding
• The previous slide suggested that all possible U should be

checked to determine the minimum value (maximum
likelihood).
– If we compute the “metrics” for each U as they arrive, the trellis
structure can reduce the number of computations that must be
performed.
– For a 2^K-1 state trellis, only that number of possible U paths need
to be considered.
• Each trellis state has two arriving states. If we compute path values
for each one, only the smallest one needs to be maintained. The larger
can never become smaller as more n-tuples arrive!
• Therefore, only 2^K-1 possible paths vs. 2^L possible paths for U
must be considered!
ECE 6640 37
Viterbi Decoder Trellis
• Decoder Trellis with Hamming distances shown for each

of the possible paths from “state to state”.
encoder trellis
ECE 6640 38
Viterbi Example
• m: 1 1 0 1 1
• U: 11 01 01 00 01
• Z: 11 01 01 10 01
merging
paths
ECE 6640 39
Viterbi Example
• m: 1 1 0 1 1
• U: 11 01 01 00 01
• Z: 11 01 01 10 01
ECE 6640 40
Add Compare Select
Viterbi Decoding Implementation
• Section 7.3.5.1, p. 406
Possible Connections
ECE 6640 41
Add Compare Select
Viterbi Decoding Implementation
• State Metric Update based on new Branch Metric Values
– Hard coding uses bit difference measure
– Soft coding uses rms distances between actual and expected branch
values
– The minimum path value is maintained after comparing incoming
paths.
– Paths are eliminated that are not maintained.
• When all remaining paths use the same branch, update the output
sequence
ECE 6640 42
Properties of Convolutional Codes
• Distance Properties
– If an all zero sequence is input and there is a bit error, how and
how long will it take to return to an all zeros path?
– Find the “minimum free distance”
• The number of code bit errors required before returning
• Note that this is not time steps and not states moved through
• This determines the error correction capability
 d  1
t f 
 2 
– Systematic and non systematic codes

• For linear block codes, any non-systematic code can be transformed
into a systematic code (structure with I and data in columns)
• This is not true for convolutional codes. Convolutional codes focus
ECE 6640 43
on free distance, making them systematic would reduce the distance!
Catastrophic Error Propagation
• A necessary and sufficient condition to have catastrophic

error propagation is that the generator polynomials have a
common factor.
– Catastrophic errors is when a finite number of code symbol errors
can generate a infinite number of decoded data bit errors.
– See Section 7.4.3 and p. 414
ECE 6640 44
Computing Distance, Number of Branches,
and Branch Transition caused by a One
• Split the state diagram to start at 00.. and end at 0..
– Show state transitions with the following notations
– D: code bit errors for a path
– L: one factor for every branch
– N: one factor for every branch taken due to a “1” input
• Define the state equations using the state diagram
– Determine the result with the smallest power of D and interpret
– See Figure 7.18 and page p. 412
ECE 6640 45
Performance Bounds
• Upper Bound of bit error probability
dTD, N 
PB 
dN N 1,D  2 p1 p 
D5  N
– for Figure 7.18 and Eq. 7.15 on p. 412 TD, N  
1 2  D  N
dTD, N  D5 D5  N
    2  D 
dN 1  2  D  N 1  2  D  N 2

1  2  D  N   2  N  D  D 5
1  2  D  N 2
D5

1  2  D  N 2
PB 
D 5

2  p  1  p  
5
ECE 6640
1  2  D  N 2 N 1, D  2 p1 p  1  4  p  1  p  
2
46
Performance Bounds
• For
EC E k E
 r b   b
N0 N0 n N0
• The bound becomes
 5  Eb   
PB  Q   exp 5  E b   1
  2 N  2
 N0   0     Eb 
1  2  exp 
  2  N  
  0 
ECE 6640 47
Coding Gain Bounds
• From Eq. 6.19

E   
G dB   b  dB   E b  dB, for Pb  Same Value
 N 0  uncoded  N 0  coded
• This is bounded by
– The 10 log base 10 of the code rate and the min. free distance
G dB  10  log10 r  d f 
• Coding Gains are shown in Tables 7.2 and 7.3, p. 417
ECE 6640 48
Soft Decision Viterbi
• The previous example used fractional “soft values”

– See the Viterbi example slides on line
• For digital processing hardware: use integer values and
change the observed code to the maximum integer values
– For 0, 1 in an 8-level system use 0,7
– Compute distances as the rms value from the desired received code
to the observed received code.
• Note that only 4 values need to be computed to define all branch
metric values,
• Example: see Fig. 7.22 for (0,0) and (7,7) now computed distances
from (0,7) and (7,0) and you have all 4!
– Apply computations and comparisons as done before.
ECE 6640 49
Sequential Decoding
• Existed prior to Viterbi

• Generate a hypothesis about the transmitted sequence
– Compute metric between hypothesis and received signal
– If metric indicates reasonable agreement, go forward
otherwise go backward, change the hypothesis and keep trying.
– See section 7.5 and p. 422-425.
• Complexity
– Viterbi grows exponentially with constraint length
– Sequential is independent of the constraint length
• Can have buffer memory problems at low SNR (many trials)
ECE 6640 50
Feedback Decoding
• Use a look-ahead approach to determine the minimum

“future” hamming distance
– Look ahead length, L, has received code symbols forward in time
– Compare look-ahead paths for minimum hamming distance and
take the tree branch that contains the minimum value ….
• Section 7.5.3 on p. 427-429
• Called a feedback decoder as detection decisions are

required as feedback to compute the next set of code paths
to search for a minimum.
ECE 6640 51
References
• http://home.netcom.com/~chip.f/viterbi/tutorial.html
• http://www.eccpage.com/
ECE 6640 52
ECE 6640

Assistant Professor
Chapter 8
1. Reed-Solomon Codes.
2. Interleaving and Concatenated Codes.
3. Coding and Interleaving Applied to the Compact Disc
Digital Audio System.
4. Turbo Codes.
5. Appendix 8A. The Sum of Log-Likelihood Ratios.
ECE 6640 2
Reed-Solomon Codes
• Nonbinary cyclic codes with symbols consisting of m-bit

sequences
– (n, k) codes of m-bit symbols exist for all n and k where
0  k  n  2m  2
– Convenient example
n, k   2m  1, 2m  1  2  t 
– An “extended code” could use n=2m and become a perfect length
hexidecimal or byte-length word.
• R-S codes achieve the largest possible code minimum
distance for any linear code with the same encoder input
and output block lengths!
d min  n  k  1  d  1  n  k 
t   min    
 2   2 
ECE 6640 4
Comparative Advantage to Binary
• For a (7,3) binary code:

– 2^7=128 n-tuples
– 2^3=8 3-“symbol” codewords
– 8/128=1/16 of the n-tuples are codewords
• For a (7,3) R-S with 3-bit symbols
– (2^7)^3 =2,097,152 n-tuples
– (2^3)^3= 512 3-“symbol” codewords
– 2^9/2^21=1/2^12=1/4,096 of the n-tuples are codewords
• Significantly increasing hamming distances are possible!
ECE 6640 5
R-S Error Probability
• Useful for burst-error corrections

– Numerous systems suffer from burst-errors
• Error Probability
2 1  m
2  1 j
m
  p  1  p 2 1 j
1
  j  
m
PE  m
2  1 j t 1  j 
• The bit error probability can be upper bounded by the

symbol error probability for specific modulation types. For
MFSK m 1
PB 2
 m
PE 2  1
ECE 6640 6
Burst Errors
• Result in a series of bits or symbols being corrupted.

• Causes:
– Signal fading (cell phone Rayleigh Fading)
– Lightening or other “impulse noise” (radar, switches, etc.)
– Rapid Transients
– CD/DVD damage
• See Wikipedia for references:
http://en.wikipedia.org/wiki/Burst_error
• Note that for R-S Codes, the t correction is for symbols,

not just bits … therefore, t=4 implies 3-4 n-tuples of
sequential errors.
ECE 6640 7
R-S and Finite Fields
• R-S codes use generator polynomials

– Encoding may be done in a systematic form
– Operations (addition, subtraction, multiplication and division) must
be defined for the m-bit symbol systems.
• Galois Fields (GF) allow operations to be readily defined
ECE 6640 8
R-S Encoding/Decoding
• Done similarly to binary cyclic codes

– GF math performed for multiplication and addition of feedback
polynomial
• U(X)=m(X) x g(X) with p(X) parity computed
• Syndrome computation performed
• Errors detected and corrected, but with higher complexity
(a binary error calls for flipping a bit, what about an m-bit
symbol?)
– r(X)=U(X) + e(X)
– Must determine error location and error value …
ECE 6640 9
Reed-Solomon Summary
• Widely used in data storage and communications protocols

• You may need to know more in the future
(systems you work with may use it)
ECE 6640 10
Interleaving
• Convolutional codes are suitable for memoryless channels
with random error events.
• Some errors have bursty nature:

– Statistical dependence among successive error events (time-correlation)
due to the channel memory.
• Like errors in multipath fading channels in wireless communications,
errors due to the switching noise, …
• “Interleaving” makes the channel looks like as a memoryless

channel at the decoder.
Digital Communications I: Modulation and Coding Course, Period 3 – 2006, Sorour Falahati, Lecture 13
ECE 6640 11
Interleaving …
• Interleaving is done by spreading the coded symbols in
time (interleaving) before transmission.
• The reverse in done at the receiver by deinterleaving the
received sequence.
• “Interleaving” makes bursty errors look like random.
Hence, Conv. codes can be used.
• Types of interleaving:
– Block interleaving
– Convolutional or cross interleaving
ECE 6640 12
Interleaving …
• Consider a code with t=1 and 3 coded bits.
• A burst error of length 3 can not be corrected.
A1 A2 A3 B1 B2 B3 C1 C2 C3
2 errors
• Let us use a block interleaver 3X3

A1 A2 A3 B1 B2 B3 C1 C2 C3 A1 B1 C1 A2 B2 C2 A3 B3 C3
Interleaver Deinterleaver
A1 B1 C1 A2 B2 C2 A3 B3 C3 A1 A2 A3 B1 B2 B3 C1 C2 C3
1 errors 1 errors 1 errors
ECE 6640 13
Convolutional Interleaving
• A simple banked switching and delay structure can be used

as proposed by Ramsey and Forney.
– Interleave after encoding and prior to transmission
– Deinterleave after reception but prior to decoding
ECE 6640 14
Forney Reference
• Forney, G., Jr., "Burst-Correcting
Codes for the Classic Bursty
Channel," Communication
Technology, IEEE Transactions on ,
vol.19, no.5, pp.772,781, October
1971.
ECE 6640 15
Convolutional Example
• Data fills the commutator

registers
• Output sequence (in repeating
blocks of 16)
– 1 14 11 8
– 5 2 15 12
– 9 6 3 16
– 13 10 7 4
– 1 14 11 8
– 5 2 15 12
– 9 6 3 16
– 13 10 7 4
ECE 6640 16
Concatenated codes
• A concatenated code uses two levels on coding, an inner
code and an outer code (higher rate).
– Popular concatenated codes: Convolutional codes with Viterbi
decoding as the inner code and Reed-Solomon codes as the outer
code
• The purpose is to reduce the overall complexity, yet
achieving the required error performance.
Input Outer Inner
Interleaver Modulate
data encoder encoder
Channel
Output Outer Inner
Deinterleaver Demodulate
data decoder decoder
ECE 6640 17
Practical example: Compact Disc
“Without error correcting codes, digital audio
would not be technically feasible.”
• Channel in a CD playback system consists of a

transmitting laser, a recorded disc and a photo-detector.
• Sources of errors are manufacturing damages, fingerprints
or scratches
• Errors have bursty like nature.
• Error correction and concealment is done by using a
concatenated error control scheme, called cross-interleaver
Reed-Solomon code (CIRC).
ECE 6640 18
CD CIRC Specifications
Maximum correctable burst length 4000 bits (2.5 mm track length)
Maximum interpolatable burst length 12,000 bit (8 mm)
Sample interpolation rate One sample every 10 hors at PB=10-4
1000 samples/min at PB=10-3
Undetected error samples (clicks) Less than one every 750 hours at PB=10-3
Negligible at PB=10-3
New discs are characterized by PB=10-4
ECE 6640 19
Compact disc – cont’d
• CIRC encoder and decoder:
Encoder
 C2 D* C1 D
interleave encode interleave encode interleave
 C2 D* C1 D
deinterleave decode deinterleave decode deinterleave
Decoder
ECE 6640 20
CD Encoder Process
16-bit Left Audio
16-bit Right Audio
(24 byte frame)
RS code 8-bit symbols
RS(255, 251)
24 Used Symbols
227 Unused Symbols
Equ. RS(28, 24)
RS(255, 251)
28 Used Symbols
223 Unused Symbols
Equ. RS(32, 28)
ECE 6640 Overall Rate 3/4 21

CD Decoder Process
ECE 6640 22
Advanced Topic: Turbo Codes
• Concatenated coding scheme for achieving large coding

gains
– Combine two or more relatively simple building blocks or
component codes. Often combined with interleaving.
– For example: A Reed-Solomon outer code with a convolutional
inner code
• May use soft decisions in first decoder to pass to next
decoder. Multiple iterations of decoding may be used to
improve decisions!
• A popular topic for research, publications, and
applications.
ECE 6640 23
Turbo Code MATLAB
• I have been trying to run a simulation ….

– Reed Solomon Examples
– Turbo Code Examples
ECE 6640 24
Turbo Code Performance
• The decoding operation

can be performed multiple
times or iterations.
• There is a degree of
improvement as shown.
ECE 6640 25
MATLAB Simulations
0
LTE Turbo-Coding LTE Turbo-Coding
0
10 10
N = 2048, 1 iterations N = 2048, 2 iterations
-2 -2
10 10
-4 -4
10 10
BER
BER
-6 -6
10 10
-8 -8
10 10
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Eb/N0 (dB) Eb/N0 (dB)
ECE 6640 26
MATLAB Simulations
LTE Turbo-Coding 0
LTE Turbo-Coding
0
10 10
N = 2048, 3 iterations N = 2048, 4 iterations
-2 -2
10 10
-4 -4
10 10
BER
BER
-6 -6
10 10
-8 -8
10 10
-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Eb/N0 (dB) Eb/N0 (dB)
ECE 6640 27
References
• http://home.netcom.com/~chip.f/viterbi/tutorial.html
• http://www.eccpage.com/
• http://www.csee.wvu.edu/~mvalenti/turbo.html
• http://www.eg.bucknell.edu/~kozick/elec47601/notes.html
• Digital Communications I: Modulation and Coding Course, Period 3 – 2006,
Sorour Falahati, Lecture 13
ECE 6640 28
ECE 6640

Assistant Professor
Chapter 9
1. Goals of the Communications System Designer.
2. Error Probability Plane.
3. Nyquist Minimum Bandwidth.
4. Shannon-Hartley Capacity Theorem.
5. Bandwidth Efficiency Plane.
7. Defining, Designing, and Evaluating Systems.
8. Bandwidth-Efficient Modulations.
9. Modulation and Coding for Bandlimited Channels.
10. Trellis-Coded Modulation.
ECE 6640 2
System Level Tradeoffs
• The Nyquist theoretical minimum bandwidth requirements

• The Shannon-Hartley capacity theorem
– The Shannon limit
• Government regulatory involvement
– frequency allocation, bandwidth limitations
• Technology limitations
– physically realizable components using current technology
• Other system requirements
– For satellite: orbits and energy limitations
ECE 6640 4
Error Probability Plane
• Error probability performance curves

– define acceptable BER
– determine required Eb/No
• We would prefer equivalent bandwidth performance
curves
– allows system level tradeoffs
– trade-off Eb/No for modulation type at fixed BER
– trade off BER vs modulation type at fixed Eb/No
– show range of expected BER as Eb/No varies
ECE 6640 5
BER vs Eb/No Curves
ECE 6640 6
Nyquist Minimum Bandwidth
• Nyquist showed that the theoretical minimum bandwidth

needed for baseband transmission of Rs symbols per
second without ISI is Rs/2 Hz.
– A theoretical minimum constraint on bandwidth required.
– Referred to as 2 symbols/sec/Hz
– Typical systems and filters are 10%-40% wider
– More likely 1.8 to ¼ symbols/s/Hz.
• Rs in terms of M symbol modulation
R R
R  k  Rs Rs  
k log 2 M
ECE 6640 7
Example 9.1: Digital Schemes
• Orthogonal Signaling
– expect improvement in BER as k or M increases
• Non-orthogonal signaling
– expect a decrease in BER as k or M increases
a) Does error-performance improve or degrade with

increasing M, for M-ary signaling?
b) The choices available almost always involve a tread-off.
If error performance improves, what price must we pay?
c) If error-performance degrades, what benefit is exhibited?
ECE 6640 8
Example 9.1
• Expected trade-offs
• M-FSK
– as M increases, the required transmission bandwidth increases for
minimum frequency spacing.
– to maintain a constant bit rate, the symbol transmission rate
decreases with increasing M
• M-PSK
– while there is degradation as M increases, the symbol transmission
rate may be decreased as M increases
– M-PSK systems plot equal-bandwidth curves, as the bit
transmission rate increases.
ECE 6640 9
Shannon-Hartley Capacity Theorem
• The capacity relation in

AWGN can be stated as
 S
C  W  log 2 1  
 N
– where S is the signal power,

N the noise power, and W the
bandwidth
– the value is defined in bits per
second
ECE 6640 10
Shannon-Hartley Capacity Theorem
• The normalized channel

bandwidth vs. SNR may
also be plotted
C  S
 log 2 1  
W  N
1
W   S 
 log 2 1  
C   N 
ECE 6640 11
S-H Equivalent Equations
• Rearranging and defining the noise power and signal power

C  S C  E R 
 log 2 1    log 2 1  b  b 
W  N W  N0 W 
• For
C
Eb S 1 Rb Eb Rb
  2  1
W

N0 N 1 W N0 W
• Letting C = Rb
Eb W  W 
C C
E C
2  1 b 
W
   2  1
N0 W N0 C  
C  Eb C 
 log 2 1   
ECE 6640 W  N0 W  12
Shannon Capacity Theorem
• There is a limiting case as C/W  0

– let Eb  C 
x  
N0  W 
C  E C Eb N W  E C
 log 2 1  b   1   0    log 2 1  b  
W  N0 W  N0  Eb C   N0 W 
E 1
1  b   log 2 1  x  1
Eb
N0

 log 2 1  x  x
1

N0 x
1  lim
x 0
Eb
N0
 1

 log 2 1  x  x  b  log 2 e
E
N0
Eb 1
  0.693  1.6dB
N 0 log 2 e
ECE 6640 13
Shannon Limit
Eb 1
  0.693  1.59dB
N 0 log 2 e
• As C/W  0 or W/C∞
• In practice, it is not possible to
reach the bound.
• Provides an improvement bound
for encoding and decoding.
• For example: raw BPSK requires
approximately 9.6 dB Eb/No to
achieve a BER of 10-5 which
suggests that up to an 11.2 dB
improvement is possible.
– Turbu Codes can achieve ~ 10 dB.
ECE 6640 14
Entropy
• To compute communication capacity, a metric for the

message content of a system is also important.
• Entropy is defined as the average amount of information
per source output.
• It is expressed by: n
H   pi  log 2  pi 
i 1
– where pi is the probability of the ith output and
the sum of all pi is 1.
• For a binary system, entropy can be expressed as:
H   p  log 2  p   1  p   log 2 1  p 
ECE 6640 15
Entropy for a Binary System
• The entropy is based on the
probability, p, of an event.
• This can also be looked at as
the randomness of successive
events or how correlated
individual events are.
• Note that maximum entropy is
achieved when the probability
is 50%
– A sample provides no information
about a succeeding sample.
ECE 6640 16
Example 9.2 English Language
• The English language is highly redundant.

– The probability of the next letter in a word is not equally likely for
all possible characters.
– Determine the Entropy based on the letter probabilities
– p=0.10 for the letters a, e, o, t n
– p=0.07 for the letters h, I, n, r, s H   pi  log 2  pi 
i 1
– p=0.02 for the letters c, d, f, l, m, p, u, y
– p=0.01 for the letters b, g, j, k, q, v, w, x, z
H  4  0.1  log 2 0.1  5  0.07  log 2 0.07  1  1 

H  26    log 2    log 2 26
 8  0.02  log 2 0.02   9  0.01  log 2 0.01  26  26 
 4.17 bits/char  4.70 bits/char
English Language Equal Probability

ECE 6640 17
Equivocation
• A term used by Shannon to account for the uncertainty in a

received signal. It is defined as the conditional entropy of
the message X (transmitted source message), given Y (the
received signal).
H  X | Y    P X , Y   log 2 P X | Y 
X ,Y
– based on conditional probability
H  X | Y    PY    P X | Y   log 2 P X | Y 
Y X
ECE 6640 18
Equivocation Example
• Consider a binary sequence, X, where the bits are equally

likely. Assume that the channel produces on error in a
received sequence of 100 bit (Pb=0.01).
H  X | Y    P X , Y   log 2 P X | Y 
X ,Y
H  X | Y   1  Pb   log 2 1  Pb   Pb  log 2 Pb 
H  X | Y   0.99  log 2 0.99  0.01  log 2 0.01
H  X | Y   0.081
• Interpretation: the channel introduces 0.081 bit/received

symbol of uncertainty.
ECE 6640 19
Effective Transmission Rate
• Using the equivocation computation, the effective

transmission rate of the channel can be computed as
H eff  H  X   H  X | Y 
– based on the previous example, the binary system would have an

effective transmission rate (in terms of bit/received symbol) of
H eff  1  0.081  0.919
– for a communication system with R = 1000 bits/sec,

the effective transmission rate would become
Reff  R  H eff  1000  0.919  919
ECE 6640 20
Pb vs Eb/No Curves
• It appears that Pb approaches 0.5 as Eb/No decreases …

but the Shannon limits is Eb/No=-1.6 dB.
Is this a contradiction or not?
• Shannon refers to received information bits based on
equivocations.
ECE 6640 21
Deriving an Effective Eb/No
• As an example, take Eb/No=-10 dB for coherent BPSK


PB  Q 2  Eb N 0 
PB  Q0.447   0.33
H  X | Y   1  0.33  log 2 1  0.33  0.33  log 2 0.33  0.915
H eff  1  0.915  0.085
– from this form an effective Eb/No
 Eb  E N 0.1
   b 0   1.176  0.7dB
 N 0  eff H eff 0.085
– Thus, he effective Eb/No is well above the Shannon limit, -1.6dB
ECE 6640 22
Bandwidth-Efficieny Plane
• Using Shannon-Hartley Capacity, the “normalized”

channel bandwidth versus Eb/No for different symbol
schemes can be compared.
– Typically performed for a defined bit-error probability and under
optimal symbol detection assumptions.
– Let R=C, then
R  E R
 log 2 1  b  
W  N0 W 
– The bounds and appropriate values for MPSK, MFSK and MQAM
symbol schemes are shown on Fig. 9.6
ECE 6640 23
Figure 9.6:
Bandwidth-Efficiency Plane
• Factors of note:
– MPSK and QAM nominally
maintain the same bandwidth
will increasing the bits per
symbol and required Eb/No
– MFSK uses an increasing
bandwidth as the bits per
symbol increases while the
Eb/No is decreasing
– BPSK and QPSK have the
same Eb/No but different bits
per symbol
ECE 6640 24
Bit and Symbol Rate Considerations
• For MPSK
R  k  Rs  log 2 M   Rs log 2 M   Rs
 log 2 M 
R
1 
WIF   Rs WIF Rs
Ts
– R/W increases with M
• For MFSK
R  k  Rs  log 2 M   Rs R log 2 M   Rs log 2 M 
M  
WIF   M  Rs WIF M  Rs M
Ts
– R/W decreases with M
ECE 6640 25
Bandwidth versus Power
• For a bandwidth-limited system

– spectral efficiency is important
– expect that signal power may be increases to offset the limitation
– study the bandwidth-efficient plane
– PSK allows for fixed bandwidths
• For a power-limited system
– a defined transmission power limit has been established
– expect that signal bandwidth may increase to offset the limit
– study the bit-error probability planes
– FSK allows for limited spectral power
ECE 6640 26
Digital Comm. System Engineering
• Defining, designing, and evaluating communication

systems.
• Comparing MPSK and MFSK (table 9.1)
MPSK Non‐Coherent MFSK
M k R Rs min W R/W Eb/No (dB) min W R/W Eb/No (dB)
bits/sec sym/sec (Hz) Pb=1e‐5 (Hz) Pb=1e‐5
2 1 9600 9600 9600 1 9.6 19200 0.5 13.4
4 2 9600 4800 4800 2 9.6 19200 0.5 10.6
8 3 9600 3200 3200 3 13.0 25600 0.375 9.1
16 4 9600 2400 2400 4 17.5 38400 0.25 8.1
32 5 9600 1920 1920 5 22.4 61440 0.15625 7.4
ECE 6640 27
System Example #1:
Bandwidth Limited
• W = 4000 Hz, Pr/No=53 dB-Hz, R=9600 bps, PB=1e-5
• Equations needed for the computations (assuming M-PSK)

Pr E E
 b  R  s  Rs
N0 N0 N0
 log 2 M 
Es Eb

N0 N0
   
PE M   2  Q 2  s  sin 
E

 N0 M 
PE M 
PB 
ECE 6640 log 2 M  28
System Example #1:
Bandwidth Limited
 log 2 M 
Pr E E Es Eb
 b  R  s  Rs 
W 4000 Hz
N0 N0 N0 N0 N0
Pr/No 53 dB‐Hz
PE M 
R 9600 bps
    PB 
PE M   2  Q 2  s  sin 
Pb 1.00E‐05 BER E
 log 2 M 
Pr/No 199526.23 Hz  N0 M 
Eb/No 20.78
Eb/No 13.18 dB
M‐PSK Rs lin dB sqrt(2*Es/No) sin(pi/M) x Q(x)=Pe Pb

2 9600 sym/s Es/No 20.78 13.18 6.45 1.00 6.45 1.14E‐10 1.14E‐10
4 4800 sym/s Es/No 41.57 16.19 9.12 0.71 6.45 1.14E‐10 5.69E‐11
8 3200 sym/s Es/No 62.35 17.95 11.17 0.38 4.27 1.92E‐05 6.42E‐06
16 2400 sym/s Es/No 83.14 19.20 12.89 0.20 2.52 1.19E‐02 2.97E‐03
ECE 6640 29
System Example #2:
Power Limited
• W = 45 kHz, Pr/No=48 dB-Hz, R=9600 bps, PB=1e-5
• Equations needed for the computations (assuming M-FSK)

Pr E E
 b  R  s  Rs
N0 N0 N0
 log 2 M 
Es Eb

N0 N0
M 1  1 E 
PE M    exp   s 
2  2 N0 
2 k 1
PB  PE M   k
2 1
ECE 6640 30
System Example #2:
Power Limited
• W = 45 kHz, Pr/No=48 dB-Hz, R=9600 bps, PB=1e-5
Pr E E
 log 2 M 
 b  R  s  Rs Es Eb

W 45000 Hz N0 N0 N0 N0 N0
Pr/No 48 dB‐Hz
R 9600 bps
Pb 1.00E‐05 BER
M 1  1 Es  2 k 1
Pr/No 63095.73 Hz  
PE M   exp    PB  PE M   k
2  2 N0  2 1
Eb/No 6.57
Eb/No 8.18 dB
M‐FSK k Rs Ws lin dB exp(‐Es/No/2) PE Pb

2 1 9600 sym/s 19200 Hz Es/No 6.57 8.18 0.04 1.87E‐02 1.87E‐02
4 2 4800 sym/s 19200 Hz Es/No 13.14 11.19 0.00 2.10E‐03 1.40E‐03
8 3 3200 sym/s 25600 Hz Es/No 19.72 12.95 0.00 1.83E‐04 1.05E‐04
16 4 2400 sym/s 38400 Hz Es/No 26.29 14.20 0.00 1.47E‐05 7.82E‐06
32 5 1920 sym/s 61440 Hz Es/No 32.86 15.17 0.00 1.13E‐06 5.85E‐07
ECE 6640 31
Coded System Example
• When the previous methods do not produce a valid

implementation, encoding and decoding will be required.
– Monitor the effect of code rates on symbols/sec and bandwidths
ECE 6640 32
System Example #3:
Encode-Decode
• Starting with the previous 8-PSK system, we need
additional coding gain
 R  log 2 M   Rs
n
Pr E E E Rc 
 b  R  c  Rc  s  Rs k
N0 N0 N0 N0
E k
 log 2 M   b     log 2 M 
E s Ec

N0 N0 N0  n 
   
PE M   2  Q 2  s  sin 
E

 N0 M 
1 n n j
PB    j     Pc  1  Pc 
n j
PE M  n j t 1  j 
PC 
ECE 6640 log 2 M  33
Solution is Steps
• Step 1: Compute the Es/No
E k
 log 2 M   b     log 2 M 
E s Ec
Pr E E E
 b  R  c  Rc  s  Rs 
N0 N0 N0 N0 N0 N0 N0  n 
• Step 2: Compute the codeword symbol error rate PE(M)

   
PE M   2  Q 2  s  sin 
E

 N0 M 
• Step 3: Compute the codeword-bit-error rate
PE M 
PC 
log 2 M 
• Step 4: Compute the decoded bit error probability
1 n n j
PB    j     Pc  1  Pc 
n j
n j t 1  j 
ECE 6640 34
Excel Computations
• An excel spreadsheet can be used for all of the examples.
• see results for Example #3
• Alternate Approach
– the coding gain formula can be used.
E   
G in dB    b  in dB    Eb  in dB 
 N 0 uncoded  N 0  coded
G in dB   16  13.2  2.8
– an encoding scheme that meets the bandwidth requirement and has

2.8 dB or more coding gain is sufficient for solving this problem.
ECE 6640 35
Bandwidth Efficient Modulations
• Modern communication is hungry for bandwidth,

demanding an every increasing communications capacity
within the fixed frequ3ency bands available,
• Additional requirements to allow for non-linear
amplification put a premium on using signals that are
minimally effected by AM to PM conversion, limiting the
amplitude variations of the signal (desiring a constant
modulus).
ECE 6640 36
QPSK and Offset QPSK
• Conventional QPSK uses

consecutive bits received to
determine I-Q pairs for transmission.
• Offset QPSK also uses the bits, but
directs them to the I and Q ports as
they arrive in time (next slide)
ECE 6640 37
QPSK versus Offset QPSK
• OQPSK makes 90
degree phase transitions
• 180 degrees phase
changes may result in
significant amplitude
variation
ECE 6640 38
Minimum Shift Keying (MSK)
• Avoiding discontinuous phase transitions of the signal

– maintain a constant amplitude
– use a form of continuous-phase FSK
– also a modified form of OQPSK
  d  
st   cos 2     f 0  k   t  xk , k  T  t  k  1  T
  4 T  
  k  
xk  mod  xk 1   d k 1  d k ,2   
 2  
ECE 6640 39
MSK Quadrature Representation
• Expanding the cosine term cos(a+b)

  t 
st   ak  cos   cos2    f 0  t 
 2 T 
  t 
 bk  sin   sin 2    f 0  t 
 2 T 
ak  cosxk   1
bk  d k  cos xk   1
  k  
xk  mod  xk 1   d k 1  d k ,2   
 2  
– the similarity to OQPSK is based on the

amplitude weighted quadrature
structure of this formulation
ECE 6640 40
Bandwidth Comparison:
BPSK, QPSK & OQPSK, & MSK
ECE 6640 41
Modulation and Coding for
Bandlimited Channels
• Research Areas (as of 2001 copyright):
– Optimum signal constellation boundaries (choosing a closely
packed signal subset from any regular array or lattice of candidate
points)
– Higher density lattice structures (adding improvement to the signal
subset choice by starting with the densest possible lattice for the
space)
– Trellis-coded modulation (combined modulation and coding
techniques for obtaining coding gain for bandlimited channels).
• Ungerboeck Partitioning
ECE 6640 42
Evolution of Telephone
Modem Standards (1)
• Telephone modems have dealt with the limited power and bandwidth
problem for a considerable time.
• Progress was made at different times for both leased-lines and dial-line
services.
ECE 6640 43
Evolution of Telephone
Modem Standards (2)
• Home modem standards
– Mostly replaced by telephony DSL or cable TV access
ECE 6640 44
Signal Constellation Boundaries
• Various QAM constellations that
have been investigated.
– optimal packing of points with
maximum separation
– reduce maximum amplitude
– optimize PE(M)
ECE 6640 45
Trellis-Coded Modulation (TCM)
• Developing combined modulation and coding schemes

• Use a redundant nonbinary modulation in combination
with a finite-state machine based encoding process.
– FSM could be similar to convolutional encoding
– A multi-level/phase modulation scheme
• The concept, when performing MATLAB simulations of

encoded bit streams using MPSK or QAM symbols, is
there an optimal combination?
– if you know the symbols being used, could one convolutional code
leading to an appropriate trellis decoding perform better than
another?
ECE 6640 46
TCM Encoding
• Ungerboeck, G., "Channel coding with multilevel/phase

signals," Information Theory, IEEE Transactions on,
vol.28, no.1, pp.55,67, Jan 1982.
• Initial paper describing trellis coded, soft decision
encoding and modulation technique for communications.
ECE 6640 47
ECE 6640

Assistant Professor
MATLAB Simulations
• Communication Objects
– see MatlabModDemod directory for trial code
– AM Modulation
• comm.PAMModulator
• comm.RectangularQAMModulator
– PM Modulation
• comm.BPSKModulator
• comm.PSKModulator
• comm.DBPSKModulator
• comm.DQPSKModulator
• comm.OQPSKModulator
– FM Modulation
• comm.FSKModulator
ECE 6640 2
MATLAB Simulations
• Communication Objects.
– see MatlabModDemod directory for trial code
– CPM Modulation
• comm.CPFSKModulator
• comm.CPMModulator
• comm.GMSKModulator
• comm.MSKModulator
ECE 6640 3
MATLAB Simulations
• Communication Objects.
– see ViterbiComm directory for demos
– TCM Modulation
• comm.PSKTCMModulator
• comm.RectangularQAMTCMModulator
• comm.GeneralQAMTCMModulator
– Convolutional Coding
• comm.ConvolutionalEncoder
• comm.ViterbiDecoder (Hard and Soft)
• comm.TurboEncoder – available from Matlab, no demo
ECE 6640 4

EE6640 Wmu

Uploaded by

Copyright:

Available Formats

EE6640 Wmu

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

EE6640 Wmu

Uploaded by

Copyright:

Available Formats

ECE 6640

Dr. Bradley J. Bazuin

• The class web site is at

• Dr. Bradley J. Bazuin

• Bernard Sklar, Digital Communications, Fundamentals

• MATLAB, Student Edition

• Identification for Grade Posting,

• Please read, provide unique identification, sign and date,

Digital Communication Signal Processing. Classification of Signals.

2. Formatting and Baseband Modulation.

Baseband Systems. Formatting Textual Data (Character Coding).

Signals and Noise. Detection of Binary Signals in Gaussian Noise.

4. Bandpass Modulation and Demodulation/Detection.

Why Modulate? Digital Bandpass Modulation Techniques. Detection

Waveform Coding. Types of Error Control. Structured Sequences.

7. Channel Coding: Part 2.

Convolutional Encoding. Convolutional Encoder Representation.

Reed-Solomon Codes. Interleaving and Concatenated Codes. Coding

Appendix 8A. The Sum of Log-Likelihood Ratios.

9. Modulation and Coding Trade-Offs.

Goals of the Communications System Designer. Error Probability

11. Multiplexing and Multiple Access.

Allocation of the Communications Resource. Multiple Access

12. Spread-Spectrum Techniques.

Spread-Spectrum Overview. Pseudonoise Sequences. Direct-

Signals, Spectra, and Linear Systems. Fourier Techniques for Linear

B. Fundamentals of Statistical Decision Theory.

Bayes' Theorem. Decision Theory. Signal Detection Example.

C. Response of a Correlator To White Noise.

E. s-Domain, z-Domain and Digital Filtering.

G. SystemView by ELANIX Guide to the CD.

• Matlab simulations of all significant concepts should be available.

A review of prerequisite material that is critically important

Information Source Channel Mod- Antenna

Bits Symbols Signals

Information Source Channel Demod- Antenna

Linear Nonlinear Atten-

• The channel greatly effects received RF signals

r t   s t   hc t   s2 t   h2 t     s N t   hN t  nt 

• The receiver must extract the original message as best

• Deterministic and Random

• Periodic and Non-periodic

• Analog and Discrete/Digital

• Energy and Power Signals

• The CD that comes with the text includes a

• Energy Spectral Density

• Power Spectral Density

4. Transform Pair R XX   XX f 

2. Symmetry XX   XX  

3. Maximum XX   XX 0

4. Transform Pair XX   G XX f 

Functions of random variables

5.2. Continuous and Discrete Random Processes

5.3. Deterministic and Nondeterministic Random Processes

5.4. Stationary and Nonstationary Random Processes

5.5. Ergodic and Nonergodic Random Processes

In general, we always attempt to deal with stationary processes … or approximate stationary by

You will typically deal with Wide-Sense Stationary Signals.

Note that ergodicity cannot exist unless the process is stationary!

• The first spectral null occurs are 1/T. Therefore one