Вы находитесь на странице: 1из 265

UNIT 1: Sampling and

Quantization
Introduction
• Digital representation of analog signals

Digital
Analog Waveform
signal
signal Coding
destination
source (Codec)

Analog-to-Digital
Digital Encoding
Advantages of Digital Transmissions

Noise immunity
Error detection and correction
Ease of multiplexing
Integration of analog and digital data
Use of signal regenerators
Data integrity and security
Ease of evaluation and measurements
More suitable for processing ……..
Disadvantages of Digital Transmissions

More bandwidth requirement

Need of precise time synchronization


Additional hardware for encoding/decoding

Integration of analog and digital data

Sudden degradation in QoS

Incompatible with existing analog facilities


A Typical Digital Communication Link

Fig. 2 Block Diagram


Basic Digital Communication Transformations
– Formatting/Source Coding
– Transforms source info into digital symbols (digitization)
– Selects compatible waveforms (matching function)
– Introduces redundancy which facilitates accurate decoding despite errors
It is essential for reliable communication
– Modulation/Demodulation
– Modulation is the process of modifying the info signal to facilitate transmission
– Demodulation reverses the process of modulation. It involves the detection and retrie
of the info signal
• Types
• Coherent: Requires a reference info for detection
• Noncoherent: Does not require reference phase information
Basic Digital Communication Transformations
– Coding/Decoding
Translating info bits to transmitter data symbols
Techniques used to enhance info signal so that they are less vulnerable to channel
impairment (e.g. noise, fading, jamming, interference)
• Two Categories
– Waveform Coding
• Produces new waveforms with better performance
– Structured Sequences
Involves the use of redundant bits to determine the occurrence of error (and
sometimes correct it)
– Multiplexing/Multiple Access Is synonymous with resource sharing with other users
– Frequency Division Multiplexing/Multiple Access (FDM/FDMA
Practical Aspects of Sampling
1. Sampling Theorem

2 .Methods of Sampling

3. Significance of Sampling Rate

4. Anti-aliasing Filter

5 . Applications of Sampling Theorem – PAM/TDM


Sampling

Sampling is the processes of converting continuous-time


continuous analog signal, xa(t), into a discrete-time si
by taking the “samples” at discrete-time
time intervals
– Sampling analog signals makes them discrete in time but still continuous valued
– If done properly (Nyquist theorem is satisfied), sampling does not introduce distortion
Sampled values:
– The value of the function at the sampling points
Sampling interval:
– The time that separates sampling points (interval b/w samples), Ts
– If the signal is slowly varying, then fewer samples per second will be required than if the wavef
is rapidly varying
– So, the optimum sampling rate depends on the maximum frequency component present in the
signal
Analog-to-digital conversion is (basically) a 2 step process:
– Sampling
• Convert from continuous-time
time analog signal xa(t) to discrete-time continuous value signal x
– Is obtained by taking the “samples” of xa(t) at discrete-time
discrete intervals, Ts

Quantization
– Convert from discrete-time
time continuous valued signal to discrete time discrete valued signal
Sampling

ling Rate (or sampling frequency fs):


he rate at which the signal is sampled, expressed as the number of samples per second
eciprocal of the sampling interval), 1/Ts = fs

st Sampling Theorem (or Nyquist Criterion):


the sampling is performed at a proper rate, no info is lost about the original signal and it can be
operly reconstructed later on
atement:
“If a signal is sampled at a rate at least, but not exactly equal to twice the max frequency
onent of the waveform, then the waveform can be exactly reconstructed from the samples
without any distortion”

f s  2 f max
….. Sampling Theorem
Sampling Theorem for Bandpass Signal - If an analog information
signal containing no frequency outside the specified bandwidth W
Hz, it may be reconstructed from its samples at a sequence of points
spaced 1/(2W)
W) seconds apart with zero-mean
zero squared error.

The reciprocal of Nyquist rate, 1/(2W), is


The minimum sampling called the Nyquist interval, that is, Ts =
rate of (2W) samples per 1/(
/(2W).
second, for an analog The phenomenon of the presence of
signal bandwidth of W Hz, high
high-frequency component in the
is called the Nyquist rate. spectrum of the original analog signal is
called aliasing or simply foldover.
Sampling Theorem
Sampling Theorem for Baseband Signal - A baseband signal having
no frequency components higher than fm Hz may be completely
recovered from the knowledge of its samples taken at a rate of at
least 2 fm samples per second, that is, sampling frequency fs ≥ 2 fm.

A baseband signal having no frequency


The minimum sampling components higher than fm Hz is
rate fs = 2 fm samples per completely described by its sample
second is called the values at uniform intervals less than or
equal to 1/(2fm) seconds apart, that is,
Nyquist sampling rate. the sampling interval Ts ≤ 1/(2fm)
seconds.
Methods of Sampling

Ideal
sampling -
an
impulse at
each
sampling
instant

Ideal Sampling
Ideal Sampling ( or Impulse Sampling)

Is accomplished by the multiplication of the signal x(t) by the uniform train of impulses (comb
function)
Consider the instantaneous sampling of the analog signal x(t)

Train of impulse functions select sample values at regular intervals



x s (t )  x (t ) 
n  
 (t  n Ts )

Fourier Series representation:


 
1 jn s t 2

n 
 (t  n T s ) 
Ts

n 
e ,  s 
Ts
Ideal Sampling ( or Impulse Sampling)

his shows that the Fourier Transform of the sampled signal is the Fourier Transform of the original
nal at rate of 1/Ts
Ideal Sampling ( or Impulse Sampling)
As long as fs> 2fm,no overlap of repeated replicas X(f - n/Ts) will occur in Xs(f)
Minimum Sampling Condition:
fs  fm  fm  fs  2 fm
Sampling Theorem: A finite energy function x(t) can be completely reconstructed from
ampled value x(nTs) with
  2 f (t  n Ts )  
  s in   
  2 T s  
(t )   Ts x ( n Ts )  
n     (t  n Ts ) 
 
  
  T s x ( n T s ) s in c ( 2 f s ( t  n T s ) )
1 1
n    Ts 
provided that => fs 2 fm
Ideal Sampling ( or Impulse Sampling)

his means that the output is simply the replication of the original signal at discrete intervals, e.g
Ts is called the Nyquist interval: It is the longest time interval that can be used for sampling a ban
signal and still allow reconstruction of the signal at the receiver without distortion
….. Methods of Sampling

Natural
sampling - a
pulse of
short width
with varying
amplitude
with natural
tops

Natural Sampling
Natural Sampling

If we multiply x(t) by a train of rectang


pulses xp(t), we obtain a gated wavefo
that approximates the ideal sampled
waveform, known as natural sampling
gating (see Figure 2.8)

x s (t )  x (t ) x p (t )

j 2 nfs
 x (t ) 
n
cne

X s ( f )  [ x (t ) x p (t )]

j2
 
n 
c n  [ x (t )e

 
n  
cn X [ f  nfs
Each pulse in xp(t) has width Ts and amplitude 1/T /Ts
The top of each pulse follows the variation of the signal being sampled
Xs (f) is the replication of X(f) periodically every fs Hz
Xs (f) is weighted by Cn  Fourier Series Coeffiecient
The problem with a natural sampled waveform is that the tops of the sample pulses are not flat
It is not compatible with a digital system since the amplitude of each sample has infinite number o
possible values
Another technique known as flat top sampling is used to alleviate this problem
….. Methods of Sampling

Flat-top
sampling - a
pulse of
short width
with varying
amplitude
with flat
tops

top Sampling
Flat-top
Flat-Top
Top Sampling
ere, the pulse is held to a constant height for the whole sample period
at top sampling is obtained by the convolution of the signal obtained after ideal
mpling with a unity amplitude rectangular pulse, p(t)
is technique is used to realize Sample-and
and-Hold (S/H) operation
S/H, input signal is continuously sampled and then the value is held for as long as i
kes to for the A/D to acquire its value
Flat top sampling (Time Domain)
x '( t )  x ( t )  ( t )
x s ( t )  x '( t ) * p ( t )

 
 p ( t ) * x ( t ) ( t )  p ( t ) *  x ( t )   ( t  n T s ) 
 n   
Taking the Fourier Transform will result to

X s ( f )   [ x s (t )]

 
 P ( f )   x (t )   (t  n T s )
 n   

 1 
 P( f ) X ( f )*
Ts
  ( f  nfs )
 n  

1
 P( f )
Ts

n
X ( f  nfs )

where P(f) is a sinc function


Flat top sampling (Frequency Domain)

Flattop sampling becomes identical to ideal sampling as the width of the pulses become
shorter
Recovering the Analog Signal
One way of recovering the original signal from sampled signal Xs(f) is to pass it through a Low Pass
ilter (LPF) as shown below

f fs > 2B then we recover x(t) exactly


Else we run into some problems and signal is not fully recovered
Significance of Sampling Rate

When fs < 2fm,


spectral
components of
adjacent samples
will overlap,
known as aliasing

An Illustration of Aliasing
Undersampling and Aliasing
– If the waveform is undersampled (i.e. fs < 2B)) then there will be spectral overlap in the sample
signal

he signal at the output of the filter will be


different from the original signal spectrum

This is the outcome of aliasing!!


his implies that whenever the sampling condition is not met, an irreversible overlap of the spectral
licas is produced
This could be due to:
1. x(t) containing higher frequency than were expected
2. An error in calculating the sampling rate
Under normal conditions, undersampling of signals causing aliasing is not recommended
Solution 1: Anti-Aliasing Analog Filter

– All physically realizable signals are not completely bandlimited


– If there is a significant amount of energy in frequencies above half the sampling frequency
(fs/2), aliasing will occur
– Aliasing can be prevented by first passing the analog signal through an anti-aliasing filter (als
called a prefilter)) before sampling is performed
– The anti-aliasing
aliasing filter is simply a LPF with cutoff frequency equal to half the sample rate
Antialiasing Filter
An anti-aliasing
filter is a low-pass
filter of sufficient
higher order
which is
recommended to
be used prior to
sampling.

Minimizing Aliasing
• Aliasing is prevented by forcing the bandwidth of the sampled signal to satisfy the requirement
of the Sampling Theorem
Solution 2:: Over Sampling and Filtering in the Digital Domain
– The signal is passed through a low performance (less costly) analog low-pass
low filter to lim
the bandwidth.
– Sample the resulting signal at a high sampling frequency.
– The digital samples are then processed by a high performance digital filter and down
sample the resulting signal.
Summary Of Sampling 
Ideal Sampling
(or Impulse Sampling)
x s (t )  x (t ) x (t )  x (t ) 
n  
 (t  n T s )

 
n  
x ( n T s ) ( t  n T
Natural Sampling
(or Gating) 
j 2 n
x s (t )  x (t ) x p (t )  x (t ) 
n  
cne
Flat-Top Sampling

 
x s ( t )  x '( t ) * p ( t )   x ( t )   ( t  n T s )  * p
For all sampling techniques  n   
– If fs > 2B then we can recover x(t) exactly
– If fs < 2B) spectral overlapping known as aliasing
liasing will occur
Quantization
Quantization is a non linear transformation which maps elements from a continuous
o a finite set. It is also the second step required by A/D conversion.

og Signal Sample Quantize Digital Signal


ontinuous time - Discrete time
ontinuous value - Discrete time - Discrete value
- Continuous value
Uniform Quantization
output w2(t)
V

-V V
input w1(t)

-V
on of operation
For M=2n levels, step size :
 = 2V /2n = V(2-n+1)
Figure 3.10 Two types of quantization: (a)
( midtread and (b) midrise.
Quantization Error, e
output w2(t)
V

-V V
input w1(t)

-V

Error, e
/2
-/2 input w1(t)

Error is symmetric 
around zero. 0

Average error power :


    3 
  
1
V
2

22
2

2  2   2  
V  2  n 1
2
V 2 2n
2 V V
e ( s ) ds
 0
x dx
  3   12  12

3
2
 
 
Suppose the input signal is a triangula r wave between  V and  V .
V 2
Then the average signal power is .
3
 S 
    2 2n
 N  out
Definition. The dynamic range of an input signal is the ratio of the largest to
the smallest power levels which the input signal can take on and be reproduced
with the acceptable signal distortion.

The dynamic range of the quantizer input in the PCM system is 6n dB.
Nonuniform Quantizer

Used to reduce quantization error and increase the dynamic range when input signa
not uniformly distributed over its allowed range of values.

owed
lues input

for
t
pressing-and-expanding”
expanding” is called “companding.”
Nonuniform quantizer

rete Uniform digital


ples Compressor Quantizer

Channel
••••

Decoder Expander outp


ved
al signals
Line codes:

1. Unipolar nonreturn-to-zero
zero (NRZ) Signaling
2. Polar nonreturn-to-zero(NRZ)
zero(NRZ) Signaling
3. Unipor nonreturn-to-zero
zero (RZ) Signaling
4. Bipolar nonreturn-to-zero
zero (BRZ) Signaling
5. Split-phase
phase (Manchester code)
Figure 3.15 Line codes for the electrical representations of binary data.
(a) Unipolar NRZ signaling. (b)) Polar NRZ signaling.
(c) Unipolar RZ signaling. (d)) Bipolar RZ signaling.
(e) Split-phase or Manchester code.
Application of Sampling Theorem –
PAM/TDM

Design of
PAM/TDM
System
UNIT-II
II DIGITAL MODULATION
Pulse Code Modulation (PCM)

1. Block Diagram of PCM

2 PCM Sampling

3 Quantization of Sampled Signal

4 Encoding of Quantized Sampled Signal


Pulse Code Modulation

The basic elements of a PCM system.


PCM Sampling

he Process
of Natural
Sampling
Quantization of Sampled Signal
s(t)
sq(t)

VH
s7 Δ7
L67

s6 Δ6
L56
s5 Δ5
s(t)
L45 sq(t)

peration of L34
s4 Δ4

uantization L23
s3 Δ3

s2 Δ2
Δ/2
L12

s1 Δ1
L01

Δ s0 Δ0
VL
t
Quantization Error and Classification

Quantization is the Quantization error is defined as the


conversion of an analog difference between rounding off sample
values of an analog signal to the nearest
sample of the
permissible level of the quantizer during
information signal into the process of quantization.
discrete form. Thus, an
infinite number of
Classification of Quantization
possible levels are Process
converted to a finite
Uniform Non-uniform
number of conditions. quantization quantization
Characteristics of Compressor, Uniform
and Non-
Non-uniform Quantizer
μ-law and A-
A-law Compression
Characteristics
Encoding of Quantized Sampled Signal

PCM – Functional Blocks


PCM System Parameters

PCM Data Rate (bps) = 2nfm

PCM Bandwidth (Hz) = (1/2)) PCM Data Rate = nfm

Dynamic Range (dB) = 20 log (2n – 1)

Coding Efficiency (%) = [(minimum bits)/(actual bits)] x 100

Where n is number of PCM encoding bits and fm is the highest


frequency component of information signal
DELTA MODULATION
Essence of Delta Modulation (DM)
Delta modulation (DM) uses a single-bit
single DPCM code to achieve
digital transmission of analog signals

An Ideal Delta Modulation Waveform


DM system. (a)) Transmitter. (b)
( Receiver.
lator consists of a comparator, a quantizer, and an accumulator
utput of the accumulator is

n
m q n     sgn( e i )
i 1
n
  e q i  (3.55)
i 1

wo types of quantization errors :


ope overload distortion and granular noise
a-Sigma modulation (sigma-delta
delta modulation)
 modulation
  which has an integrator can
eve the draw back of delta modulation (differentiator
differentiator)
eficial effects of using integrator:
Pre-emphasize the low-frequency
frequency content
Increase correlation between adjacent samples
educe the variance of the error signal at the quantizer input )
Simplify receiver design
ause the transmitter has an integrator , the receiver
sists simply of a low-pass filter.
differentiator in the conventional DM receiver is cancelled by the integrator )
ivalent versions of delta-sigma modulation system.
Differential Pulse-Code
Code Modulation (DPCM)
PCM has the sampling rate higher than the Nyquist rate .The encode signal contains redundant information. D
ciently remove this redundancy.

Figure 3.28 DPCM system. (a) Transmitter. (b)) Receiver.


Adaptive Differential Pulse--Code Modulation (ADPCM)
eed for coding speech at low bit rates , we have two aims in mind:
Remove redundancies from the speech signal as far as possible.
Assign the available bits in a perceptually efficient manner.

Figure 3.29 Adaptive quantization with backward estimation (AQB).

Adaptive prediction with backward estimation (APB).


Comparison of PCM and DM Techniques
S. No. Parameter PCM DPCM DM ADM
1. Number of bits per 4/8/16 bits More than one bit but One bit One bit
sample less than PCM

2. Number of levels Depends on number of bits Fixed number of levels Two levels Two levels

3. Step size Fixed or variable Fixed or variable Fixed Variable

4. Transmission bandwidth More bandwidth needed Lesser than PCM Lowest Lowest

5. Feedback Does not exist Exists Exists Exists


6. Quantization Quantization noise depends Quantization noise & slope overload & Quantization noise only
noise/distortion on number of bits slope overload granular noise

7. Complexity of Complex Simple Simple Simple


implementation
UNIT-III
UNIT
Basband Pulse Transmission
Transmit and Receive Formatting
Sources of Error in received Signal
Major sources of errors:
– 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”.
Receiver Structure
AW GN

DETECT
DEM ODULATE & SAM PLE
SAM PLE
at t = T
R ECEIV E D
W AVEFO RM FR EQ U EN C Y
R E C E IV IN G E Q U A L IZ IN G
DOWN
F IL T E R F IL T E R TH R ESH O LD M ESSA
A N S M ITT ED C O N V E R S IO N
AVEFORM C O M P A R IS O N SYM BO
OR
CHANN
FOR CO M P EN S A T IO N
SYM BO
BAN DPASS FO R CH ANN EL
SIG N A L S IN D U CED ISI

O P T IO N A L

E S S E N T IA L

Demodulation/Detection of digital signals


Receiver Structure contd
The digital receiver performs two basic functions:
functions
– Demodulation
– Detection
Why demodulate a baseband signal???
– Channel and the transmitter’s filter causes ISI which “smears” the
transmitted pulses
– Required to recover a waveform to be sampled at t = nT.
Detection
– decision-making
making process of selecting possible digital symbol
Steps in designing the receiver
Find optimum solution for receiver design with the following goals:
1. Maximize SNR
2. Minimize ISI
Steps in design:
– Model the received signal
– Find separate solutions for each of the goals.
Detection of Binary Signal in Gaussian Noise

covery of signal at the receiver consist of two parts


ilter
• Reduces the received signal to a single variable z(T)
• z(T) is called the test statistics
Detector (or decision circuit)
• Compares the z(T) to some threshold level 0 , i.e.,
H 1

z (T )  H1 and H0 are the two
where
possible binary hypothesis
 0
H 0
Finding optimized filter for AWGN channel

Assuming Channel with response equal to impulse


function
Detection of Binary Signal in Gaussian Noise
• For any binary channel, the transmitted signal over a symbol interval (0,T)
( is:

 s 0 (t ) 0  t  T for a binary 0
si (t )  
 s1 ( t ) 0  t  T for a binary 1
• The received signal r(t) degraded by noise n(t) and possibly degraded by the impulse resp
the channel hc(t), is

r (t )  s (t ) * h (t )  n (t ) i  1,2
Where n(t) is assumed toi be zero mean
c AWGN process

• For ideal distortionless channel where hc(t) is an impulse function and convolution with
produces no degradation, r(t) can be represented as:

r (t )  si (t )  n (t ) i  1, 2 0  t T
Design the receiver filter to maximize the SNR

Model the received signal


t) h c (t ) r (t ) r (t )  si (t )  h c (t )  n (t )

n (t )
AWGN

Simplify the model:


– Ideal
Received
channels
signal in AWGN
hc (t )   (t )
si (t ) r (t ) r (t )  si (t )  n (t )

n (t )
AWGN
Find Filter Transfer Function H0(f)

ctive: To maximizes (S/N)T and find h(t)


sing signal ai(t) at filter output in terms of filter transfer function H(f)

j 2  ft
a i (t )   H ( f ) S( f ) e df


the filter transfer funtion and S(f) is the Fourier transform of input signal s(t)
ded PSD of i/p noise is N0/2
t noise power can be expressed as:
2 N 0 
 0   | H ( f ) | 2 df
2 

(S/N)T : 2

j 2  fT
 H ( f ) S( f ) e df
 S  
  
 N T N 0 
 | H ( f ) | 2 df
2 
• For H(f) = Hopt (f) to maximize (S/N)T use Schwarz’s Inequality:
Inequality

 2  2  2

 f 1 ( x ) f 2 ( x ) dx   f 1 ( x ) dx  f 2 ( x ) dx
  

• Equality holds if f1(x) = k f*2(x) where k is arbitrary constant and * indicates complex conjugate
• Associate H(f) with f1(x) and S(f) ej2 fT with f2(x) to get:

 2  2  2
j 2  fT
 H ( f ) S( f ) e df   H ( f ) df  S ( f ) df
  
• Substitute yields to:

2
 S  2 
    S ( f ) df
 N T N0 
 S  2 E
max  and energy E of the input signal s(t):
 N T N 0 2

E   S ( f ) df

s (S/N)T depends on input signal energy
power spectral density of noise and
T on the particular shape of the waveform

lity for  S  for optimum


holds 2E filter transfer function H0(f)
max   
 N T N 0

that: H ( f )  H 0 ( f )  kS * ( f ) e  j 2  fT

h (t )   1
kS * ( f(3.55)
)e  j 2  fT

eal valued s(t):
 kS ( T  t ) 0  t  T
h (t )  
0 else where
mpulse response of a filter producing maximum output signal-to-noise
signal ratio is the
or image of message signal s(t), delayed by symbol time duration T.
filter designed is called a MATCHED FILTER

 kS ( T  t ) 0  t  T
h (t )  
0 else where
ned as:
a linear filter designed to provide the maximum
signal-to-noise
noise power ratio at its output for a given
transmitted symbol waveform
Matched Filter Output of a rectangular Pulse
Replacing Matched filter with Integrator
Implementation of matched filter receiver
Bank of M matched filters

*
z 1 (T )
s1 (T  t )  z1  Matched filter output:
r (t )    z Observation
 z vector
 
*  z M 
s M (T  t ) zM (T
T)

z i  r ( t )  s  i (T  t ) i  1 ,..., M
z  ( z 1 ( T ), z 2 ( T ),..., z M ( T ))  ( z 1 , z 2 ,..., z M )
Detection

Max. Likelihood Detector


Probability of Error
Detection
ed filter reduces the received signal to a single variable z(T), after which the detection of symbol is carried out
ncept of maximum likelihood detector is based on Statistical Decision Theory
ws us to
mulate the decision rule that operates on the data
imize the detection criterion

H 1

z (T ) 
 0
H 0
Probabilities Review

P[s1]  a priori probabilities


hese probabilities are known before transmission

obability of the received sample


), p(z|s1)
nditional pdf of received signal z, conditioned on the class si
], P[s1|z]  a posteriori probabilities
fter examining the sample, we make a refinement of our previous knowledge
0], P[s0|s1]
rong decision (error)
1], P[s0|s0]
rrect decision
How to Choose the threshold?

mum Likelihood Ratio test and Maximum a posteriori (MAP)


( criterion:

p ( s 0 | z )  p ( s1 | z )   H 0

p ( s1 | z )  p ( s 0 | z )   H 1
em is that a posteriori probabilities are not known.
on: Use Bay’s theorem:
p (z |s ) p (s )
p (s | z)  i i
i p(z)
H H
p ( z | s1 ) P ( s1 ) 
1
p (z | s0 )P (s0 ) 
1

 
 p ( z | s1 ) P ( s1 ) 
p ( z | s0 ) P ( s0 )
P (z) H0
P (z) H0
s means that if received signal is positive, s1 (t) was sent,
sent else
s0 (t) was sent
Likelihood of So and S1

1
MAP criterion:
H1
p ( z | s1 )  P (s0 )
L(z)  
 likelihood ratio test ( LRT )
p(z |s0 ) H
P ( s1 )
0

When the two signals, s0(t) and s1(t),, are equally likely, i.e., P(s0) = P(s1) = 0.5, then the decision rule
becomes

H 1
p ( z | s1 ) 
L(z)  
1  max likelihood ratio test
p(z |s0) H 0

s is known as maximum likelihood ratio test because we are selecting


hypothesis that corresponds to the signal with the maximum likelihood.

erms of the Bayes criterion, it implies that the cost of both types of error is the same
tuting the pdfs
2
1  1  z  a0  
H 0 : p ( z | s0 )  exp    
 0 2  2 0 
2
1  1  z  a1  
H1 : p ( z | s1 )  exp     
 0 2  2  0  

H1 1  1  H1
exp   2
z  a 1   2

p ( z | s1 )   0 2  2 o  
L(z)  1 1
p ( z | s0 )  1  1 2  
exp   2
z  a 0  
H 0  0 2  2 0  H 0
Hence:

 z (a1  a 0 ) ( a 12  a 02 )  
exp  2
   1
  0 2  02 
Taking the log, both sides will give
H1
z ( a1  a 0 ) ( a 12  a 02 ) 
ln{ L ( z )}  2
 0
0 2  02 
H 0

H 1

z (a1  a 0 )  ( a 12  a 02 ) ( a 1  a 0 )( a 1  a 0 )
 
 02  2  02 2  02
H 0
• Hence

H1 H1
  02 ( a 1  a 0 )( a 1  a 0 )
z  ( a1  a 0 )
 2 02 ( a 1  a 0 ) z   0
 2
H0 H0
where z is the minimum error criterion and  0 is optimum
ptimum threshold
• For antipodal signal, s1(t) = - s0 (t)  a1 = - a0
H1

z 0

H0
Probability of Error

will occur if
s sent  s0 is received
P(H 0 | s1 )  P ( e | s1 )
 0
P ( e | s1 )   p ( z | s 1 ) dz

s sent  s1 is received
P (H 1 | s0 )  P (e | s0 )

P (e | s0 )   p ( z | s 0 ) dz
0

The total probability of error is sum of the errors


2
PB  
i 1
P ( e , s i )  P ( e | s1 ) P ( s1 )  P ( e | s 0 ) P ( s 0 )

 P (H 0 | s1 ) P ( s1 )  P ( H 1 | s0 ) P (s0 )
nals are equally probable

PB  P ( H 0 | s1 ) P ( s1 )  P ( H 1 | s0 ) P (s0 )
1
 P ( H 0 | s1 )  P ( H 1 | s0 )
2
e, the probability of bit error PB, is the probability that an incorrect hypothesis is made
rically, PB is the area under the tail of either of the conditional distributions p(z|s1) or p(z|s0)

 
PB   P (H 1 | s 0 ) dz   p ( z | s 0 ) dz
0 0

2
 1  1  z  a  
0
  exp     dz
0  0 2  2   0  
Inter-Symbol
Symbol Interference (ISI)
SI in the detection process due to the filtering effects of the
ystem
Overall equivalent system transfer function

H ( f )  H t ( f )H c ( f )H r ( f )
– creates echoes and hence time dispersion
– causes ISI at sampling time

zk  sk  nk  
i k
i si
Inter-symbol
symbol interference
Baseband system model
x2
Channel Rx. filter zk
Tx filter r (t )
ht (t ) hc (t ) hr (t ) Detector
t  kT
T H t( f ) Hc( f ) Hr( f )
x3 T n (t )

Equivalent model
x2
Equivalent system z (t ) zk
h (t ) Detector
t  kT
H (f )
x3 T nˆ ( t )
filtered noise
H ( f )  H t ( f )H c ( f )H r ( f )
Nyquist bandwidth constraint
Nyquist bandwidth constraint:
• The theoretical minimum required system bandwidth to detect Rs [symbols/s]
without ISI is Rs/2 [Hz].
• Equivalently, a system with bandwidth W=1/2T=Rs/2 [Hz] can support a maxim
transmission rate of 2W=1/T=Rs [symbols/s] without ISI.

1 Rs Rs
 W   2 [symbol/s/ Hz]
2T 2 W
Bandwidth efficiency, R/W [bits/s/Hz] :
• An important measure in DCs representing data throughput per hertz of bandw
• Showing how efficiently the bandwidth resources are used by signaling techniq
Ideal Nyquist pulse (filter)
Ideal Nyquist filter Ideal Nyquist pulse

H (f ) h ( t )  sinc( t / T )
T 1

0 f  2T  T 0 T 2T
1 1
2T 2T
1
W 
2T
Nyquist pulses (filters)
Nyquist pulses (filters):
– Pulses (filters) which results in no ISI at the sampling time.
Nyquist filter:
– Its transfer function in frequency domain is obtained by convolving a
rectangular function with any real even-symmetric
even frequency function
Nyquist pulse:
– Its shape can be represented by a sinc(t/T) function multiply by another
time function.
Example of Nyquist filters: Raised-Cosine
Raised filter
Pulse shaping to reduce ISI
Goals and trade-off in pulse-shaping
shaping
– Reduce ISI
– Efficient bandwidth utilization
– Robustness to timing error (small side lobes)
The raised cosine filter
aised-Cosine Filter
– A Nyquist pulse (No ISI at the sampling time)

1 for | f | 2 W 0  W
   | f |  W  2W 0 
2
H ( f )   cos   for 2 W 0  W  | f |  W
  4 W  W 0 
 0 for | f |  W
cos[ 2  (W  W 0 ) t ]
h ( t )  2 W 0 (sinc( 2 W 0 t ))
1  [ 4 (W  W 0 ) t ] 2
W W0
Excess bandwidth: W W0 Roll-off factor r 
W0
0  r 1
The Raised cosine filter – cont’d
| H ( f ) | | H RC (f )| h ( t )  h RC ( t )
1 r  0 1

r  0 .5
0.5 0.5 r 1
r 1 r  0 .5
r

1 3 1 0 1 3 1  3T  2T T 0 T 2T
T 4T 2T 2T 4T T

Rs
Baseband W sSB  (1  r ) Passband W DSB  (1  r ) R s
2
Pulse shaping and equalization to remove ISI
No ISI at the sampling time

H RC ( f )  H t ( f )H c ( f )H r ( f )H e ( f )

quare-Root
Root Raised Cosine (SRRC) filter and Equalizer
H RC ( f )  H t ( f )H r ( f )
Taking care of ISI
H r( f )  Ht( f )  H (f)  H (f) caused by tr. filter
RC SRRC

1
H e( f )  Taking care of ISI
H c( f ) caused by channel
Example of pulse shaping
Square-root Raised-Cosine
Cosine (SRRC) pulse shaping
mp. [V]

Baseband tr. Waveform

Third pulse

t/T

First pulse
Second pulse

Data symbol
Example of pulse shaping …
Raised Cosine pulse at the output of matched filter
Amp. [V]

Baseband received waveform at


the matched filter output
(zero ISI)

t/T
Eye pattern
Eye pattern:Display on an oscilloscope which sweeps the system response to
a baseband signal at the rate 1/T (T symbol duration)

istortion
due to ISI
Noise margin
amplitude scale

Sensitivity t
timing erro

Timing jitter
time scale
Example of eye pattern:
Binary-PAM,
PAM, SRRQ pulse
Perfect channel (no noise and no ISI)
Correlative Coding

Transmit 2W
W symbols/s with zero ISI, using the theoretical minimum bandwidth of W Hz, without
infinitely sharp filters.
Correlative coding (or duobinary signaling or partial response signaling) introduces some
controlled amount of ISI into the data stream rather than trying to eliminate ISI completely
Doubinary signaling
Duobinary signaling
Duobinary signaling (class I partial response)
Duobinary signal and Nyguist Criteria
Nyguist second criteria: but twice the bandwidth
Differential Coding

The response of a pulse is spread over more than one signalin


interval.
The response is partial in any signaling interval.
Detection :
– Major drawback : error propagation.
To avoid error propagation, need deferential coding
(precoding).
Modified duobinary signaling
Modified duobinary signaling
– In duobinary signaling, H(f) is nonzero at the origin.
– We can correct this deficiency by using the class IV partial response
Modified duobinary signaling
Spectrum
Modified duobinary signaling
Time Sequency: interpretation of receiving 2, 0, and -2?
Duobinary Transfer Function
Comparison of Binary with Duobinary
Signaling
Binary signaling assumes the transmitted pulse amplitude are independent of one another
Duobinary signaling introduces correlation between pulse amplitudes
Duobinary technique achieve zero ISI signal transmission using a smaller system bandwidth
Duobinary coding requires three levels, compared with the usual two levels for binary coding
Duobinary signaling requires more power than binary signaling (~2.5
(~ dB greater SNR than binary
signaling)
Pass-band
band Data Transmission
Block Diagram

Functional model of pass-band


band data transmission system.
Signaling
Illustrative waveforms for the three basic forms of signaling binary information.
(a) Amplitude-shift keying. (b) Phase-shift
shift keying. (c)
( Frequency-shift keying with
continuous phase.
What do we want to study?

We are going to study and compare different modulation


techniques in terms of
– Probability of errors
– Power Spectrum
– Bandwidth efficiency

Rb
  Bits/s/Hz
B
Coherent PSK

Binary Phase Shift Keying (BPSK)


– Consider the system with 2 basis functions
2
 1 t   cos 2  f c t
Tb
– and

2
 2 t   sin 2  f c t
Tb
BPSK
If we want to fix that for both symbols (0
( and 1) the transmitted energies are
equal, we have
2
s0
s1

1
s0

We place s0 to minimize
probability of error
BPSK
We found that phase of s1 and s0 are 180 degree difference.
We can rotate s1 and s0
2
s1

1
s0

Rotate
BPSK
2

s1
s0

1

We observe that 2 has nothing to do with signals. Hence, only one


basis function is sufficient to represent the signals
BPSK
Finally, we have

2Eb
s 1 t   E b 1 (t )  cos 2  f c t
Tb

2Eb
s 0 t    E b 1 (t )   cos 2  f c t
Tb
BPSK
Signal-space
space diagram for coherent binary PSK system. The waveforms depicting
the transmitted signals s1(t) and s2(t),
), displayed in the inserts, assume nc  2.
BPSK

• Probability of error calculation. In the case of equally likely


(Pr(m0)=Pr(m1)), we have

1  d ik 
Pe  erfc  
2 2 N 
 0 

1  Eb 
 erfc  
2  N 
 0 
BPSK
Block diagrams for (a)) binary PSK transmitter and (b)
( coherent binary PSK
receiver.
Quadriphase-Shift
Shift Keying (QPSK)

2E   
s i t   cos  2  f c t  2 i  1   ; 0  t  T
T  4
T is symbol duration
E is signal energy per symbol
There are 4 symbols for i = 1, 2, 3,, and 4
QPSK

   2    2
s i t   E cos  2 i  1   cos 2  f c t   E sin  2 i  1   sin 2  f c t 
 4 T  4 T
     
 E cos  2 i  1    1 t   E sin  2 i  1    2 t ; 0  t  T
 4  4 

Which we can write in vector format as

  
 E cos 2 i  1  
si   4
 
  E sin 2 i  1  
 4 
QPSK

i Input Dibit Phase of Coordinate of Message


QPSK point
signaling si1 si2
1 10  /4
E /2  E /2

2 00 3 / 4  E /2  E /2

3 01 5 / 4  E /2 E /2

4 11 7 / 4
E /2 E /2
QPSK

2
s3 s4
(01) (11)

1
s2 s1

(00) (10
10)
QPSK signals
QPSK
Block diagrams of (a)
QPSK transmitter and
(b) coherent QPSK
receiver.
QPSK: Error Probability QPSK

Consider signal Z3 2 Z4
constellation given in
s3 s4
the figure (10
10) (11)

E /2

 E /2 E /2
Z1 1
s2 s1
Z2
(10)
(00)  E /2
QPSK
can treat QPSK as the combination of 2 independent
K over the interval T=2Tb
ce the first bit is transmitted by 1 and the second bit is
nsmitted by 2.
bability of error for each channel is given by

1  d 12  1  E 

P  erfc    erf c  
2   2 2N 0 
2 N0   
QPSK
mbol is to be received correctly both bits must be received
ectly.
ce, the average probability of correct decision is given by
2
P 
ch gives the probability of errors equal
c 1
to P  

 E  1 2  E 
 PC  erfc    erfc 
 2N 0  4  2 N 0 
 E 
fc  

 2 N 0 
QPSK
e one symbol of QPSK consists of two bits, we have E = 2Eb.

 Eb 
Pe per symbol  erfc  
 N0 
 
above probability is the error probability per symbol. The avg.
bability of error per bit

1 1  Eb 
it  Pe per symbol erfc 
 N


2 2
ch is exactly the same as BPSK
 .0 
BPSK vs QPSK
P o w e r s p e c t ru m d e n s it y o f B P S K vs . Q P S K
2

1.8 BPSK
QPSK
1.6

1.4
b
Normalized PSD,Sf/2E

1.2

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
N o rm a liz e d fre q u e n c y , fT b
QPSK
Conclusion
– QPSK is capable of transmitting data twice as faster as BPSK with the
same energy per bit.
– We will also learn in the future that QPSK has half of the bandwidth
of BPSK.
OFFSET QPSK
90 degree shift in phase
2
(01) s3 s4 (11)

1
s2 s1

(00) (10
10)

180 degree shift in phase


OFFSET QPSK
OFFSET QPSK
Whenever both bits are changed simultaneously, 180 degree
phase-shift occurs.
At 180 phase-shift,
shift, the amplitude of the transmitted signal
changes very rapidly costing amplitude fluctuation.
This signal may be distorted when is passed through the filter
or nonlinear amplifier.
OFFSET QPSK
2

-1

-2
0 1 2 3 4 5 6 7 8

Original Signal
2

1 . 5

0 . 5

- 0 . 5

- 1

- 1 . 5

- 2
0 1 2 3 4 5 6 7 8

Filtered signal
OFFSET QPSK
To solve the amplitude fluctuation problem, we propose the
offset QPSK.
Offset QPSK delay the data in quadrature component by T/2
T/
seconds (half of symbol).
Now, no way that both bits can change at the same time.
OFFSET QPSK
In the offset QPSK, the phase of the signal can change by 90
or 0 degree only while in the QPSK the phase of the signal can
change by 180 90 or 0 degree.
OFFSET QPSK
hase 1

SK 0 . 5
0 1 0
- 0
0

. 5
1
- 1
0 1 2 3 4 5 6 7 8

phase 0 . 5

0
1 0 0 0
PSK - 0 . 5

- 1
0 1 2 3 4 5 6 7 8

PSK 10 00
0

- 1

- 2
0
01
1 2 3 4
10
5 6 7 8

0 . 5

0 0 1 1 0
SK
- 0 . 5

- 1
0 1 2 3 4 5 6 7 8

1 0
0 . 5

SK
0

- 0 . 5
0
- 1
0 1 2 3 4 5 6 7 8

SK
1

10 10
0

- 1
01 00
- 2
0 1 2 3 4 5 6 7 8
Offset QPSK

Possible paths for switching between the


message points in (a) QPSK and (b)) offset QPSK.
OFFSET QPSK
dwidths of the offset QPSK and the regular QPSK is the
me.
m signal constellation we have that
 E 
P e  erfc 
 2 N 0 

ich is exactly the same as the regular QPSK.


M-array
array PSK
At a moment, there are M possible symbol values being sent
for M different phase values,
 i  2 i  1  / M

2E  2 
s i t   cos  2  f c t  i 
 1 , i  1, 2 ,  , M
T  M 
M-array
array PSK
Signal-space diagram for octaphase-
shift keying (i.e., M  8). The decision
boundaries are shown as dashed
lines.
Signal-space diagram illustrating the
application of the union bound for
octaphase-shift keying.
M-array
array PSK
Probability of errors

 d 12  d 18  2 E sin  / M 

 E 
P e  erfc  sin  / M  ; M  4
 N0 
M-ary
ary PSK

0
10

-1 0
10
Probability of Symbol errors

-2 0
10

-3 0
10

-4 0
10
QPSK
8 -a ry P S K
1 6 -a ry P S K
-5 0
10
0 5 10 15 20 25 30
E /N dB
b 0
M-array
array PSK
Power Spectra (M-array)
S PSK ( f )  2 E sinc 2 Tf 
 2 E b log 2 M sinc 2 T b f log 2 M 
M=2, we have

S BPSK ( f )  2 E b sinc 2 T b f 
M-array
array PSK
Power spectra of M-ary
ary PSK signals for M  2, 4, 8.

Tbf
M-array
array PSK
Bandwidth efficiency:
– We only consider the bandwidth of the main lobe (or null-to-null
null bandwidth)

2 2 2 Rb
B   
T T b log 2 M log 2 M
– Bandwidth efficiency of M-ary
ary PSK is given by

Rb Rb
   log 2 M  0 . 5 log 2 M
B 2Rb
M-ary
ary QAM
QAM = Quadrature Amplitude Modulation
Both Amplitude and phase of carrier change according to the
transmitted symbol, mi.

where a
s i t   i and
2 E 0b are integers.
ia i cos 2  f c t  
2E0
b i sin 2  f c t ; 0  t  T
T T
M-ary
ary QAM
Again, we have
2
 1 t   cos 2  f c t ;0  t  T
T
2
 2 t   sin 2  f c t 0  t  T
Tb

as the basis functions


M-ary
ary QAM
QAM square Constellation
– Having even number of bits per symbol, denoted by 2n.
– M=L x L possible values
– Denoting
L  M
16-QAM
QAM

 (  3 ,3 ) (  1, 3 ) (1 , 3 ) ( 3 ,3 ) 
 (  3 ,1 ) (  1 ,1 ) (1 ,1 ) ( 3 ,1 ) 
a i , b i   
 (  3 ,  1) (  1,  1 ) (1 ,  1 ) ( 3 ,  1) 
 
 (  3, 3) (  1,  3 ) (1 ,  3 ) (3, 3) 
L-ary, 4-PAM

16-QAM
16-QAM
QAM
Calculation of Probability of errors
– Since both basis functions are orthogonal, we can treat the 16-QAM
as combination of two 4-ary
ary PAM systems.
– For each system, the probability of error is given by

   E0 

Pe   1 
1 
 erfc  d    1  1 
 erfc  
 L  2 N   M   N0 
 0   
16-QAM
QAM
– A symbol will be received correctly if data transmitted on both 4-ary
PAM systems are received correctly. Hence, we have

Pc symbol   1  P e 2
– Probability of symbol error is given by

Pe symbol  1 Pc symbol   1  1  Pe 2


 1  1  2 Pe  Pe 2  2 Pe
16-QAM
QAM
– Hence, we have
 1   E0 
Pe symbol   21   erfc  
 M   N0 
 
– But because average energy is given by
2E0 L / 2 2  2 M  1 E 0
E av  2   2 i  1   
 L i 1  3
– We have
 1   3 E av 
Pe symbol   21   erfc  
 M   2 M  1 N 0 

Coherent FSK
FSK = frequency shift keying
Coherent = receiver have information on where the zero phas
of carrier.
We can treat it as non-linear
linear modulation since information is
put into the frequency.
Binary FSK
Transmitted signals are
 2Eb
 cos 2  f i t , 0  t  Tb
s i t    Tb

0, elsewhere

where
nc  i
fi  ; i  1, 2
Tb
Binary FSK
S1(t) represented symbol “1”.
S2(t) represented symbol “0”.
This FSK is also known as Sunde’s FSK.
It is continuous phase frequency-shift
frequency keying (CPFSK).
Binary FSK
There are two basis functions written as
 2
 cos 2  f i t , 0  t  Tb
 i t    T b

0, elsewhere

As a result, the signal vectors are

 Eb   0 
s1    and s2  
 0   E b 
BFSK
From the figure, we have d 12  2 E b

In case of Pr(0)=Pr(1),
), the probability of error is given by

1  Eb 
Pe  erfc  
2  2N 
We observe that at a given value
 of0 P
e, the BFSK system
requires twice as much power as the BPSK system.
TRANSMITTER
RECEIVER
Power Spectral density of BFSK
Consider the Sunde’s FSK where f1 and f2 are different by 1/Tb. We can write

2Eb  t 
s i t   cos  2  f c t  
Tb  Tb 
2Eb  t  2Eb  t 
 cos    cos 2  f c t   sin    sin 2  f c t 
Tb  Tb  Tb  Tb 
We observe that in-phase
phase component does not depend on mi since

2Eb  t  2Eb  t 
cos     cos  
Tb  T b  Tb T
 b 
Power Spectral density of BFSK
Half of the symbol power
We have
2
 2Eb  t   Eb   1   1 
S BI  f   F  cos       f      f   
 Tb  Tb   2Tb   2Tb   2Tb  

For the quadrature component

2Eb  t  8 E b T b cos 2  T b f 
g t   sin    S BQ 
Tb  Tb  
 2 4 T b2 f 2  1
2

Power Spectral density of BFSK
Finally, we obtain S B ( f )  S BI ( f )  S BQ ( f )
Phase Tree of BFSK
FSK signal is given by
2Eb  t 
s t   cos  2  f c t  
Tb  Tb 

At t = 0, we have
2Eb  0 2Eb
s 0   cos  2  f c 0    cos 0 
Tb  T b  Tb

The phase of Signal is zero.


Phase Tree of BFSK
At t = Tb, we have
2Eb  Tb  2Eb
s T b  cos  2  f c T b    cos   
Tb  Tb  Tb

We observe that phase changes by  after one symbol (Tb seconds). -


for symbol “1” and + for symbol “0
0”

We can draw the phase trellis as


Minimum-Shift
Shift keying (MSK)
MSK tries to shift the phase after one symbol to just half of
Sunde’s FSK system. The transmitted signal is given by

 2Eb
 cos 2  f 1 t   0  for "1"
2Eb  Tb
 cos 2  f c t   t   
Tb  2Eb
 cos 2  f 2 t   0  for "0"
 Tb
MSK
Where
h
 t    0   t
Tb
Observe that
h h
f1  f c  and f2  fc 
2T b 2T b

1
fc   f1  f2 
2
MSK
h = Tb(f1-f2) is called “deviation ratio.”
For Sunde’s FSK, h = 1.
For MSK, h = 0.5.
h cannot be any smaller because the orthogonality between cos(2f
cos( 1t)
and cos(2f2t) is still held for h < 0.5
5.
Orthogonality guarantees that both signal will not interfere each other
in detection process.
MSK
Phase trellis diagram for MSK signal 1101000
MSK
Signal s(t)
(t) of MSK can be decomposed into

2Eb
s t   cos 2  f c t   t 
Tb
2Eb 2Eb
 cos  t cos 2  f c t   sin  t sin 2  f c t 
Tb Tb
 s I t  cos 2  f c t   s Q t  sin 2  f c t 
where

 t    0   t ;0  t  T b
2Tb
MSK

Symbol (0) (Tb)

0 /2
1
 -/2

0 -/2
0

/2
MSK
For the interval –Tb < t  0,, we have

 t    0   t ; T b  t  0
2T b
Let’s note here that the  for the interval -Tb<t 0 and 0< tTb ma
not be the same.
We know that

 t   t   t 
cos  0     cos  0 cos    sin  0 sin  
 2T b   2Tb   2Tb 
MSK
Since (0) can be either 0 or  depending on the past history. We have

 t   t   t 
cos   0     cos  0 cos     cos  
 2Tb   2Tb   2T b 
“+” for (0) = 0 and “-” for (0) = 
Hence, we have

2Eb  t 
s I (t )   cos   ;Tb  t  Tb
Tb  2T b 
MSK
Similarly we can write

 t    T b   t  T b 
2T b
for 0< tTb and Tb < t2Tb. Note the “+” and “-”
“ may be different
between these intervals.
Furthermore, we have that (Tb) can be /2 depending on the
past history.
MSK
Hence, we have
  t  T b     t  T b     t  T b  
sin  T b     sin  T b cos    cos  T b sin  
 2 T b   2T b   2T b 
 t    t  
 sin  T b cos     cos  T b sin   
 2Tb 2   2Tb 2 

we have that (Tb) can be /2 depending on the past history.

  t  T b    t    t 
sin  T b      cos 
  
   sin  
 2Tb   2T b 2   2T b 
MSK
Hence, we have
2Eb  t 
s Q (t )   sin   ;0  t  2T b
Tb  2T b 
“+” for (Tb) = +/2 and “-” for (Tb) = -/2
The basis functions change to
2  t 
 1 t   cos   cos 2  f c t  ;0  t  T b
Tb  2Tb 

2  t 
 2 t   sin   sin 2  f c t  ;0  t  T b
Tb  2T b 
MSK
We write MSK signal as
2Eb 2Eb
s t   cos  t cos 2  f c t   sin  t sin 2  f c t 
Tb Tb
2Eb  2 t  2Eb  2 t 
 cos  0 cos   cos 2  f c t   sin  T b sin   sin 2  f c t 
Tb  Tb  Tb  Tb 
 E b cos  0  1 ( t )  E b sin  T b  2 ( t )
 s 1 1 ( t )  s 2  2 ( t )

Where s1  and
E b cos  0  s2   E b sin  T b 
MSK

Symbol (0) s1 (Tb) s2

0 Eb /2  Eb
1
  Eb -/2 Eb

0 Eb -/2  Eb
0

 Eb /2 Eb
1  Eb 
Pe  erfc  
2  N0 
 
Phase: 0 /2  /2  /2 0 -/2
MSK
We observe that MSK is in fact the QPSK having the
 t 
pulse shape cos  
 2T b 

Block diagrams for transmitter and receiver are given in the


next two slides.
Tb
x1   x ( t )  1 ( t ) dt
 Tb

2T b
x2   x ( t )  2 ( t ) dt
0
4
MSK
3.5 BPSK
QPSK

2.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
N o rm a liz e d F re q u e n c y , fT b
MSK
Probability of error of MSK system is equal to BPSK and QPSK
This due to the fact that MSK observes the signal for two symbol
intervals whereas FSK only observes for single signal interval.
Bandwidth of MSK system is 50% % larger than QPSK.

2
32 E b  cos 2  T b f  
S MSK ( f )   
 2  16 T b2 f 2  1 
Noncoherent Orthogonal Modulation
Noncoherent implies that phase information is not available to the
receiver.
As a result, zero phase of the receiver can mean any phase of the
transmitter.
Any modulation techniques that transmits information through the
phase cannot be used in noncoherent receivers.
Noncoherent Orthogonal Modulation
sin(2ft) sin(2ft)
cos(2ft)

cos(2ft)

Receiver
Transmitter
Noncoherent Orthogonal Modulation

It is impossible to draw the signal constellation since we do not know


where the axes are.
However, we can still determine the distance of the each signal
constellation from the origin.
As a result, the modulation techniques that put information in the
amplitude can be detected.
FSK uses the amplitude of signals in two different frequencies. Hence
non-coherent
coherent receivers can be employed.
Noncoherent Orthogonal Modulation

Consider the BFSK system where two frequencies f1 and f2 are used to
represented two “1” and “0”.
The transmitted signal is given by

2E
s (t )  cos 2  f i t    ; i  1, 2 , 0  t  T b
T
Problem is that  is unknown to the receiver. For the coherent receiver
 is precisely known by receiver.
Noncoherent Orthogonal Modulation

Furthermore, we have
2E
s (t )  cos 2  f i t   
T
2E 2E
 cos   cos 2  f i t   sin  sin 2  f i t 
T T
 s i1  1 ( t )  s i 2  2 ( t )

To get rid of the phase information (),


( we use the amplitude

s t   s i21  s i22  E cos 2   E sin 2   E


Noncoherent Orthogonal Modulation
Where
T T
s i 1   s ( t )  1 ( t ) dt  x 1   x ( t )  1 ( t ) dt
0 0
T T
s i 2   s ( t )  2 ( t ) dt  x 2   x ( t )  2 ( t ) dt
0 0
The amplitude of the received signal

1/2
 T 2 2 
 T 
l i     x ( t ) cos 2  f i t dt     x ( t ) sin 2  f i t dt  
    
 0  0  
Noncoherent Orthogonal Modulation

Probability of Errors

1  E 
Pe  exp   
2  2N 0 
Noncoherent: BFSK
For BFSK, we have

 2Eb
 cos 2  f i t ; 0  t  Tb
s i t    Tb

0 ; elsewhere
Noncoherent: BFSK
Noncoherent: BFSK
Probability of Errors

1  Eb 
Pe  exp   
2  2N 0 
DPSK
Differential PSK
– Instead of finding the phase of the signal on the interval 0<tTb. Thi
receiver determines the phase difference between adjacent time
intervals.
– If “1”” is sent, the phase will remain the same
– If “0”” is sent, the phase will change 180 degree.
DPSK
Or we have
 Eb
 cos 2  f c t ; 0  t  2Tb
 2Tb
s1 ( t )  
 Eb
 cos 2  f c t ; Tb  t  2T b
 2Tb
and
 Eb
 cos 2  f c t ; 0  t  2Tb
 2T b
s 2 (t )  
 Eb
 cos 2  f c t   ; T b  t  2T b
 2T b
DPSK
In this case, we have T=2Tb and E=2
2Eb
Hence, the probability of error is given by

1  Eb 
Pe  exp   
2  N0 
DPSK: Transmitter
d k  bk d k 1  bk d k 1
DPSK

b k} 1 0 0 1 0 0 0 1 1

dk-1} 1 1 0 1 1 0 1 0 0

Differential encoded
1 1 0 1 1 0 1 0 0 0
d k}

Transmitted Phase 0 0  0 0  0   
DPSK: Receiver
DPSK: Receiver
From the block diagram, we have that the decision rule as
say 1

l x   x I 0 x I 1  x Q 0 x Q 1 0

say 0
If the phase of signal is unchanged (send “ the sign (“+” or “-”) of
“1”)
both xi and xQ should not change. Hence, the l(x) should be positive.
If the phase of signal is unchanged (send “0”)
“ the sign (“+” or “-1”) of
both xi and xQ should change. Hence, the l(x) should be negative.
Signal-space
space diagram of received DPSK signal.
Unit-V –Introduction
Introduction to Spread Spectrum Techniques
M-ary signaling scheme:
In this signaling scheme 2 or more bits are grouped
together to form a symbol.

One of the M possible signals


s1(t) ,s2(t),s3(t),……sM(t)
is transmitted during each symbol period
of duration Ts.

The number of possible signals = M = 2n,


where n is an integer.
The symbol values of M for a given value of n:

n M = 2n Symbol
1 2 0, 1

2 4 00, 01, 10, 11

3 8 000, 001, 010,011,...

4 16 0000, 0001, 0010,0011,….

…. …… ……….
• Depending on the variation of amplitude, phase or frequency of the carrier, the modulation scheme is calle
M-ary ASK, M-ary PSK and M-ary FSK.

Fig: waveforms of (a) ASK (b) PSK (c)FSK


M-ary
ary Phase Shift Keying(MPSK)

In M-ary
ary PSK, the carrier phase takes on one of the M possible values, namely
i = 2 * (i - 1) / M
where i = 1, 2, 3, …..M.
The modulated waveform can be expressed as

where Es is energy per symbol = (log2 M) Eb


Ts is symbol period = (log2 M) Tb.
he above equation in the Quadrature form is

choosing orthogonal basis signals

efined over the interval 0  t  Ts


M-ary signal set can be expressed as

 Since there are only two basis signals, the constellation of M-ary
M PSK is two
dimensional.

 The M-ary
ary message points are equally spaced on a circle of radius Es, centered
at the origin.

 The constellation diagram of an 8-ary


ary PSK signal set is shown in fig.
Fig: Constellation diagram of an M-ary
ary PSK system(m=8)
Derivation of symbol error probability:
Decision Rule:

Fig: Constellation diagram for M=2 (Binary PSK)


If a symbol (0,0,0)) is transmitted, it is clear
that if an error occurs, the transmitted signal is most
likely to be mistaken for (0,0,1) and (1,1,11) and the
signal being mistaken for (1,1,0)) is remote.

The decision pertaining to (0,0,0)) is bounded by  = -


/8(below 1(t)- axis) to  = + /8 ( above 2(t)- axis)

The probability of correct reception is…


Fig: Probability density function of Phase .
The average symbol error probability of an coherent M-ary
M PSK system in
AWGN channel is given by

Similarly, The symbol error Probability of a differential M-ary


M PSK system in
AWGN channel is given by
Fig: The performance of symbol error probability for
-different values of M
M-ary
ary Quadrature Amplitude Modulation
(QAM)

It’s a Hybrid modulation

As we allow the amplitude to also vary with the phase, a new modulation scheme
called quadrature amplitude modulation (QAM) is obtained.

The constellation diagram of 16-ary


ary QAM consists of a square lattice of signal
points.
Fig: signal Constellation of M-ary
ary QAM for M=16
M=
M-ary
ary Frequency Shift
Keying(MFSK)
In M-ary
ary FSK modulation the transmitted signals are defined by:

where fc = nc/2Ts, for some fixed integer n.


The M transmitted signals are of equal energy and equal duration, and
the signal frequencies are separated by 1/2Ts Hertz, making the signal
orthogonal to one another.
The average probability of error based on the union bound is given by

Using only the leading terms of the binomial expansion:


Power Efficiency and Bandwidth :
Bandwidth:

The channel bandwidth of a M-ary


ary FSK signal is :
The channel bandwidth of a noncohorent MFSK is :

This implies that the bandwidth efficiency of an M-ary


M FSK signal decreases
with increasing M. Therefore, unlike M--PSK signals, M-FSK signals are
bandwidth inefficient.
Introduction to Spread Spectrum

• Problems such as capacity limits, propagation effects, synchroniza


occur with wireless systems
• Spread spectrum modulation spreads out the modulated signal
bandwidth so it is much greater than the message bandwidth
• Independent code spreads signal at transmitter and despreads sig
at receiver
Multiplexing
channels ki
• Multiplexing in 4 dimensions
– space (si) k1 k2 k3 k4 k5 k6

– time (t) c
– frequency (f) t c
– code (c) t
s1
f
• Goal: multiple use s2
of a shared medium c
t
• Important: guard spaces needed!
s3
f
Frequency multiplex
• Separation of spectrum into smaller frequency bands
• Channel gets band of the spectrum for the whole time
• Advantages:
k3 k4 k5
– no dynamic coordination needed
– works also for analog signals c

• Disadvantages:
– waste of bandwidth
if traffic distributed unevenly
– inflexible
– guard spaces

t
Time multiplex
Channel gets the whole spectrum for a certain amount of time
Advantages:
– only one carrier in the
medium at any time
– throughput high even
for many users k1 k2 k3 k4 k5
c
Disadvantages:
– precise
synchronization
necessary

t
Time and frequency multiplex
• A channel gets a certain frequency band for a certain amount of time (e.g.
GSM)
• Advantages:
– better protection against tapping
– protection against frequency
selective interference
– higher data rates compared to k1 k2 k3 k4 k5
code multiplex c
• Precise coordination
required

t
Code multiplex
k1 k2 k3 k4 k5 k6

Each channel has unique code


c
All channels use same spectrum at same time
Advantages:
– bandwidth efficient
– no coordination and synchronization
– good protection against interference
Disadvantages:
– lower user data rates
– more complex signal regeneration
Implemented using spread spectrum technology
t
Spread Spectrum Technology
• Problem of radio transmission: frequency dependent fading can wipe out
narrow band signals for duration of the interference
• Solution: spread the narrow band signal into a broad band signal using a
special code

interference
spread signal power signal
ower spread
interferenc
detection at
receiver
f f
Spread Spectrum Technology
• Side effects:
– coexistence of several signals without dynamic coordination
– tap-proof
• Alternatives: Direct Sequence (DS/SS), Frequency Hopping (FH/SS)
• Spread spectrum increases BW of message signal by a factor N, Processing
Gain

B ss  B ss 
P r o c e s s in g G a in N   1 0 lo g 1 0  
B  B 
Effects of spreading and interference

user signal
broadband interference
narrowband interference
P P

i) ii)
f f
P sender P P

iii) iv) v)
f f f
receiver
Spreading and frequency selective fading
channel
quality

2 narrowband channels
1 5 6
3
4

Narrowband signal frequency


guard space

channel
quality
2
2
2
2
2
1 spread spectrum
channels
spread frequency
spectrum
DSSS (Direct Sequence Spread Spectrum) I

• XOR the signal with pseudonoise (PN) sequence (chipping sequence)


• Advantages
– reduces frequency selective
fading Tb
– in cellular networks user dat
• base stations can use the
same frequency range
0 1 XOR
• several base stations can
Tc
detect and recover the signal chippin
• But, needs precise power control sequen
01101010110101
=
resulting
signal
01101011001010
DSSS (Direct Sequence Spread Spectrum) II

transmitter
Spread spectrum
Signal y(t)=m(t)c(t) transmit
user data signal
X modulator
m(t)
chipping radio
sequence, c(t) carrier

receiver correlator
sampled
received products data
sums
signal demodulator X integrator decision
radio
carrier
Chipping sequence, c(t)
DS/SS Comments III
Pseudonoise(PN) sequence chosen so that its autocorrelation
is very narrow => PSD is very wide
– Concentrated around t < Tc
– Cross-correlation
correlation between two user’s codes is very small
DS/SS Comments IV
Secure and Jamming Resistant
– Both receiver and transmitter must know c(t)
– Since PSD is low, hard to tell if signal present
– Since wide response, tough to jam everything
Multiple access
– If ci(t) is orthogonal to cj(t), then users do not interfere
Near/Far problem
– Users must be received with the same power
FH/SS (Frequency Hopping Spread Spectrum)
• Discrete changes of carrier frequency
– sequence of frequency changes determined via PN sequence
• Two versions
– Fast Hopping:: several frequencies per user bit (FFH)
– Slow Hopping:: several user bits per frequency (SFH)
• Advantages
– frequency selective fading and interference limited to short period
– uses only small portion of spectrum at any time
• Disadvantages
– not as robust as DS/SS
– simpler to detect
FHSS (Frequency Hopping Spread Spectrum) I
Tb

user data

0 1 0 1 1 t
f
Td
f3 slow
f2 hopping
(3 bits/hop)
f1

Td t
f

f3 fast
f2 hopping
(3 hops/bit)
f1

t
Tb: bit period Td: dwell time
FHSS (Frequency Hopping Spread Spectrum) III

transmitter narrowband Spread transmit


signal signal
user data
modulator modulator

frequency hopping
synthesizer sequence
receiver

received data
signal demodulator demodulator

hopping frequency
sequence synthesizer
Applications of Spread Spectrum
Cell phones
– IS-95 (DS/SS)
– GSM
Global Positioning System (GPS)
Wireless LANs
– 802.11b
Performance of DS/SS Systems
Pseudonoise (PN) codes
– Spread signal at the transmitter
– Despread signal at the receiver
Ideal PN sequences should be
– Orthogonal (no interference)
– Random (security)
– Autocorrelation similar to white noise (high at t=0 and low for t not
equal 0)
PN Sequence Generation
• Codes are periodic and generated by a shift register and XOR
• length (ML) shift register sequences, m-stage shift register, length: n = 2m – 1
Maximum-length
bits

R(
R(t)

t 
-1/n Tc nTc
-nTc

Output
+
Generating PN Sequences

Output m Stages connected to


+ modulo-2 adder
Take m=2 =>L=3 2 1,2
cn=[1,1,0,1,1,0, . . .], usually written
as bipolar cn=[1,1,-1,1,1,-1, . . .] 3 1,3
4 1,4
5 1,4
1 L

m    cncnm 6 1,6
L n 1

1 m  0 8 1,5,6,7
 
 1 / L 1 m  L 1
Problems with m-sequences
Cross-correlations with other m-sequences
m generated by
different input sequences can be quite high
Easy to guess connection setup in 2m samples so not too
secure
In practice, Gold codes or Kasami sequences which combine
the output of m-sequences
sequences are used.
Detecting DS/SS PSK Signals
transmitter
Spread spectrum
Signal y(t)=m(t)c(t) transmit
Bipolar, NRZ signal
m(t) X X

PN
sequence, c(t) sqrt(2)cos
)cos (wct + )

receiver
received z(t) w(t)
signal
X X LPF integrator decision
x(t)

sqrt(2)cos (wct + ) c(t)


Optimum Detection of DS/SS PSK
Recall, bipolar signaling (PSK) and white noise give the optimum error
probability
 2 Eb 
Pb  Q  
 
 
Not effected by spreading
– Wideband noise not affected by spreading
– Narrowband noise reduced by spreading
Signal Spectra

B ss  B ss  Tb
P r o c e s s in g G a in N   1 0 lo g 1 0   
B  B  Tc
• Effective noise power is channel noise power plus jamming (NB)
signal power divided by N

Tb

Tc
Multiple Access Performance
Assume K users in the same frequency band,
Interested in user 1,, other users interfere

4 6

1
3 2
Signal Model
Interested in signal 1,, but we also get signals from other K-1
users:
xk t   2 mk  t   k  c k  t   k  c o s  c  t   k    k 
 2 mk  t   k  c k  t   k  c o s  c t   k   k   k 
At receiver,

K
x  t   x1  t    x k t 
k 2
Interfering Signal

After mixing and despreading (assume t1=0)


 t   2 m k  t   k  c k  t   k  c1  t  c o s  c t   k  c o s  c t 

After LPF
w k t   m k t   k  c k  t   k  c1  t  c o s  k  1 
After the integrator-sampler
Tb
Ik  c o s  k   1   mk  t   k  c k  t   k  c1  t
0
At Receiver
(t) =+/-1 (PSK), bit duration Tb
terfering signal may change amplitude at tk
 k Tb
 c o s  k   1   b  1  ck  t   k  c1  t  d t  b 0  c k
  t   k  c1 
 0 k
Tb
User 1: I1   m 1  t  c1  t  c1  t  d t
0
eally, spreading codes are Orthogonal:
Tb Tb
 0
c1  t  c1  t  d t  A
0
ck   t   k  c1  t  d t  0