Demodulation and Detection Techniques

EEGR 453
Communications Theory
Module #3 (Lessons 14-17):
Baseband Demodulation
and Detection (Sklar: Chapt 3)
Dr. Yacob Astatke
Fall 2011
(Original Lesson done by: Sorour Falahati)
Introduction
In Module 2- chapter 2 we saw that the baseband
signals will be sent using waveforms that are shaped
like pulses that will be affected by the channels
characteristics and by noise (AWGN). This effect is
known as Inter-symbol Interference or (ISI).
In this chapter, we will discuss the design of
demodulators that are capable of recovering
baseband pulses with the best possible SNR
(using a receiving filter) and free of ISI (using an
Equalizing filter).
Learning Objectives
After completing this lesson the student will
be able to:
Understand the vector representation of signals and
noise in a signal space.
Understand the design of a typical receiver structure
using a two step process:
Step#1: Demodulation and sampling
Step#2: Detection
Understand the design of an important receiver circuit
called the optimum receiving filter also known as the
matched filter or correlator.
Last time we talked about:
Transforming the information source to a
form compatible with a digital system
Sampling
Aliasing
Quantization
Uniform and non-uniform
Baseband modulation
Binary pulse modulation
M-ary pulse modulation
M-PAM (M-ay Pulse amplitude modulation)
7/21/2011 Lecture 14-18 4
Formatting and transmission of baseband
signal
Information (data) rate:
Symbol rate :
For real time transmission:
7/21/2011 Lecture 14-18 5
[bits/sec] / 1
b b
T R =
ec] [symbols/s / 1 T R =
mR R
b
=
Sampling at rate
(sampling time=Ts)
Quantizing each sampled
value to one of the
L levels in quantizer.
Encoding each q. value to
bits
(Data bit duration Tb=Ts/l)
Encode
Pulse
modulate Sample Quantize
Pulse waveforms
(baseband signals)
Bit stream
(Data bits)
Format
Digital info.
Textual
info.
Analog
info.
source
Mapping every data bits to a
symbol out of M symbols and transmitting
a baseband waveform with duration T
s s
T f / 1 =
L l
2
log =
M m
2
log =
Overview of Chapter 3
In chapter 2 we saw that the baseband signals
will be sent using waveforms that are shaped like
pulses.
The baseband pulses will be affected by the
channels characteristics (filtering) and by noise
(electrical and thermal AWGN).
Therefore, the received baseband pulses will not
have an ideal pulse shape, and will not exactly
occupy their original symbol interval. This effect
is known as :Inter-symbol Interference or
(ISI).
7/21/2011 Lecture 14-18 6
7/21/2011 Lecture 14-18 7
The goal of the demodulator is to recover a
baseband pulse with the best possible SNR
(using a receiving filter) and free of ISI (using
an Equalizing filter) (see Fig 3.1).
The optimum receiving filter is called the
matched filter or correlator.
In some cases the receiving and equalizing
filters are grouped together to form a
receiving and equalizing filter.
7/21/2011 Lecture 14-18 8
Receiver Structure :a two step process
Step#1-Demodulation and sampling:
Waveform recovery and preparing the
received signal for detection:
Improving (SNR): the signal power to the noise
power using matched filter
Reducing ISI using equalizer
Sampling the recovered waveform
Step#2-Detection:
Estimate the transmitted symbol based on
the received sample
7/21/2011 Lecture 14-18 9
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.
7/21/2011 Lecture 14-18 10
Receiver structure
7/21/2011 Lecture 14-18 11
Step 1 waveform to sample transformation Step 2 decision making
Frequency
down-conversion
Receiving
filter
Equalizing
filter
Threshold
comparison
For bandpass signals
Compensation for
channel induced
ISI
Baseband pulse
(possibly distorted)
Sample
(test
statistic)
Baseband pulse
Received waveform
) (t r
) (T z
i
m
Demodulate & Sample Detect
z(T) has a voltage value that is directly proportional to the energy of the
received symbol and inversely proportional to the noise.
Receiver structure
#1) (Hypotesis ) ( then ) (
1
H T z T z if = >
1
H T z T z if = >
7/21/2011 Lecture 14-18 12
2
H T z T z if = <
T = Symbol duration
m
i
no
error
correction
used
u
i
error
correction
used
Topics covered in Chapter 3
Vector representation of signals and noise
(signal space), an important tool to
facilitate
Signals presentations, receiver structures
Detection operations
Receiver structure
Demodulation (and sampling)
Detection
First step for designing the receiver
Matched filter receiver
Correlator receiver
7/21/2011 Lecture 14-18 13
Signals and Noise:
Sources of Performance Degradation
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.
7/21/2011 Lecture 14-18 14
Format
Pulse
modulate
Bandpass
modulate
Format Detect
Demod.
& sample
) (t s
i
) (t g
i
i
m
i
m
) (t r ) (T z
channel
) (t h
c
) (t n
transmitted symbol
estimated symbol
M i , , 1 =
M-ary modulation
Signals and Noise:
Example: Impact of the channel
7/21/2011 Lecture 14-18 15
Signals and Noise:
Example: Channel impact
7/21/2011 Lecture 14-18 16
) 75 . 0 ( 5 . 0 ) ( ) ( T t t t h
c
= o o
A Vectorial View of Signals and Noise
What is a signal space?
Vector representations of signals in an N-dimensional
orthogonal space
Why do we need a signal space?
It is a means to convert signals to vectors and vice versa.
It makes it easy to calculate signals energy and
Euclidean distances between signals.
Why are we interested in Euclidean distances
between signals?
For detection purposes: The received signal is transformed
to a received vector. The reference signal which has the
minimum distance to the received signal is estimated as
the transmitted signal.
7/21/2011 Lecture 14-18 17
Vectorial view of Signals and Noise
Any arbitrary finite set of waveforms
where each member of the set is of duration T,
can be expressed as a linear combination of N
orthogonal waveforms where
The basis functions must satisfy the following:
7/21/2011 Lecture 14-18 18
{ }
M
i
i
t s
1
) (
=
{ }
N
j
j
t
1
) (
=
M N s
N 1,..., k j, T t 0 ) ( ) (
0
= s s =
}
jk j
T
j i
K dt t t o

otherwise 0
k j for 1
=
=
jk
o
Kronecker Delta Function
When the K
j
constants are non zero , the
signal space is called orthogonal.
When the K
j
constants are normalized
such that K
j
= 1, the signal space is called
orthonormal.
7/21/2011 Lecture 14-18 19
1 because k j if ) ( ) (
0
= = =
}
o
j
T
j i
K dt t t
0 because k j if 0 ) ( ) (
0
= = =
}
o dt t t
T
j i
Example of an orthonormal basis functions
Example: 2-dimensional orthonormal signal space
Example: 1-dimensional orthonornal signal space
7/21/2011 Lecture 14-18 20
1 ) ( ) (
0 ) ( ) ( ) ( ), (
0 ) / 2 sin(
2
) (
0 ) / 2 cos(
2
) (
2 1
2
0
1 2 1
2
1
= =
= >= <
< s =
< s =
}
t t
dt t t t t
T t T t
T
t
T t T t
T
t
T

t
t
T t
) (
1
t
T
1
0
) (
1
t
) (
2
t
0
1 ) (
1
= t
) (
1
t
0
) , , ( ) ( ) ( ) ( ) (
3 2 1 3 3 2 2 1 1 m m m m m m m m
a a a t a t a t a t s = + + = s
7/21/2011 Lecture 14-18 21
The relationship between
s
i
(t),
i
(t), and coefficients
a
ij
(t) is given by equations
3.10 and 3.11.
Review example 3.1 on how to
find the coefficients a
ij
(t).
Why should we use this approach ?
It allows the receiver/detector to compare
the received signal with the known
prototypes or reference signals that belong
to the set of M waveforms {s
i
(t)}.
It will decide which of the prototypes (i.e.
reference signals s
i
(t)) is closest in distance to
the received vector r.
See Fig 3.4 for more information
7/21/2011 Lecture 14-18 22
7/21/2011 Lecture 14-18 23
Reference signals s
i
Received signal r
Detector decision: is r close in resemblance to reference signal s
j
or s
k
?
A signal space with 3 reference waveforms
7/21/2011 Lecture 14-18 24
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
2 1 2 2 1 1
32 31 3 2 32 1 31 3
22 21 2 2 22 1 21 2
12 11 1 2 12 1 11 1
z z t z t z t z
a a t a t a t s
a a t a t a t s
a a t a t a t s
= + =
= + =
= + =
= + =
z
s
s
s

) (
1
t
) (
2
t
) , (
12 11 1
a a = s
) , (
22 21 2
a a = s
) , (
32 31 3
a a = s
) , (
2 1
z z = z
Transmitted signal
alternatives
Received signal
at
matched filter
output
You choose the
correct waveform s
i
(t)
after you compute
the distances
between the received
signal z, and the
stored waveforms:
s
1
, s
2
, s
3
.
Choose the signal
with the minimum
distance.
Signal Space and Basis Functions
Why should we use this approach ?
In the previous example, both the transmitter
and receiver must be able to generate the three
signal waveforms s
1
(t), s
2
(t), s
3
(t)
The receiver needs to generate the three
waveforms because it has to use them as
reference signals to make its decision.
Is there a way of minimizing the number of
waveforms that need to be generated by both
the transmitter and receiver ?
Yes, see example 3.1
7/21/2011 Lecture 14-18 25
Orthogonal Representation of Waveforms
7/21/2011 Lecture 14-18 26
0 because k j if 0 ) ( ) (
0
= = =
}
o dt t s t s
T
k j
1 because k j if ) ( ) (
0
= = =
}
o
j
T
k j
K dt t s t s
7/21/2011 Lecture 14-18 27
7/21/2011 Lecture 14-18 28
7/21/2011 Lecture 14-18 29
Prove s
1
(t) =
1
(t) - 2
2
(t) Sol: s
1
(t) = +1 2 = -1 (for 0<T<0.5)
and s
1
(t) = -1 - 2 = -3 (for 0.5<t <1)
+
eq#3.11 ) ( ) (

1
0
dt t t s
K
a
T
j i
j
ij
}
=
2 )) 5 . 0 ( 3 3 ( ) 5 . 0 ( ) 1 )( 3 ( ) 1 )( 1 ( ) ( ) (
1
5 . 0
5 . 0
0 0
2 1 12
= + + = + = =
} } }
dt dt dt t t s a
T
1 )) 5 . 0 ( 3 3 ( ) 5 . 0 ( ) 1 )( 3 ( ) 1 )( 1 ( ) ( ) (
1
5 . 0
5 . 0
0 0
1 1 11
= + = + = =
} } }
dt dt dt t t s a
T
) ( ) 2 ( ) ( ) 1 ( ) ( ) ( ) (
2 1 2 12 1 11 1
t t t a t a t s + = + =
7/21/2011 Lecture 14-18 30
A better method of finding the coefficients a
ij
for the
basis functions is to use equation 3.11.
For our case Kj = 1 (orthonormal function)
) ( ) ( ) (
2 22 1 21 2
t a t a t s + = ) ( ) ( ) (
2 32 1 31 3
t a t a t s + =
eq#3.11 ) ( ) (

1
0
dt t t s
K
a
T
j i
j
ij
}
=
= + = + = =
} } }
) ( ) ( ) ( ) (
1
5 . 0
5 . 0
0 0
1 2 21
dt dt dt t t s a
T
= + = + = ) ( ) ( ) ( ) ( ) ( ) ( ) (
2 1 2 22 1 21 2
t t t a t a t s
7/21/2011 Lecture 14-18 31
ij
for the
= + = + = =
} } }
) ( ) ( ) ( ) (
1
5 . 0
5 . 0
0 0
2 2 22
dt dt dt t t s a
T

eq#3.11 ) ( ) (

1
0
dt t t s
K
a
T
j i
j
ij
}
=
= + = + = =
} } }
) ( ) ( ) ( ) (
1
5 . 0
5 . 0
0 0
1 3 31
dt dt dt t t s a
T
= + = + = ) ( ) ( ) ( ) ( ) ( ) ( ) (
2 1 2 32 1 31 3
t t t a t a t s
7/21/2011 Lecture 14-18 32
ij
for the
= + = + = =
} } }
) ( ) ( ) ( ) (
1
5 . 0
5 . 0
0 0
2 3 32
dt dt dt t t s a
T

7/21/2011 Lecture 14-18 33
By using orthogonal basis functions, the
three waveforms can be recreated
using only two basis functions (see
example 3.1)
The Gram-Schmidt
orthogonalization procedure allows
the user to find the appropriate set of
orthogonal basis functions {
j
(t)} for
any given signal set {s
i
(t)} (see slides
at the end of the lecture).
L11 Begin
Orthogonal Representation of
Waveforms
The distance in signal space is measured by
calculating the norm.
What is norm?
Norm of a signal:
Norm between two signals:
We refer to the normbetween two signals as
the Euclidean distance between two signals.
7/21/2011 Lecture 14-18 34
x
E dt t x t x t x t x = = > < =
}

2
) ( ) ( ), ( ) (
) ( ) ( t x a t ax =
) ( ) (
,
t y t x d
y x
=
= length of x(t)
L15 Begin
Example of distances in signal space
7/21/2011 Lecture 14-18 35
The Euclidean distance between signals z(t) and s
i
(t):
3 , 2 , 1
) ( ) ( ) ( ) (
2
2 2
2
1 1 ,
=
+ = =
i
z a z a t z t s d
i i i z s
i
) (
1
t
) (
2
t
) , (
12 11 1
a a = s
) , (
22 21 2
a a = s
) , (
32 31 3
a a = s
) , (
2 1
z z = z
z s
d
,
1
z s
d
,
2
z s
d
,
3
1
E
3
E
2
E
SNR parameter for DCS
The figure of merit for analog communication
systems is the ratio of average signal power (S)
to average noise power (N) also known as :
SNR = (S)/(N)
In DCS, we will be using (E
b
/N
o
) as a figure of
merit, which is a normalized version of SNR.
E
b
is defined as bit energy E
b
= S * Tb
where S= signal power, and Tb = bit time
N
o
is defined as the noise power spectral density
N
o
=N/W,
where N=noise power,W=Bandwidth
7/21/2011 Lecture 14-18 36
Relationship between SNR and (E
b
/N
o
):
Why do we use (E
b
/N
o
) for DCS instead of
SNR that is used for Analog Comm Systems ?
The answer lies in the types of signals that are
commonly used in analog and digital communication
systems (note: Eb/No is a unit-less ratio like SNR).
sec *
sec *
Watt/Hz
Joules
unit
Watt
Watt
N
S
N
E
o
b
= = =
7/21/2011 Lecture 14-18 37
|
.
|
\
|
= = =
R
W
N
S
N
E
W N
R S
W N
T S
N
E
o
b b b
o
b
*
Most analog signals can be classified as
power signals they have a finite
average power but have infinite energy.
Thats because most analog waveforms have
infinite duration (ex: cos function).
Since we cannot compute the energy of
analog waveforms, it is better to use their
power (rate of delivering energy) to
evaluate their performance.
7/21/2011 Lecture 14-18 38
Most digital signals can be classified as energy
signals they have a finite energy but have
zero average power.
Thats because most digital waveforms have a
finite duration because they are transmitted and
received within a time frame or window known as
the symbol time Ts .(see Pb1.1(b))
The average power of a symbol over a symbol
time Ts is zero. Therefore, it is better to use the
energy of a symbol (integral of power over Ts)
to evaluate the performance of a DCS.
7/21/2011 Lecture 14-18 39
Insert Pb 1.1 (a) and (b) solutions
7/21/2011 Lecture 14-18 40
Example of distances in signal space
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
) , ( ) ( ) ( ) (
2 1 2 2 1 1
32 31 3 2 32 1 31 3
22 21 2 2 22 1 21 2
12 11 1 2 12 1 11 1
z z t z t z t z
a a t a t a t s
a a t a t a t s
a a t a t a t s
= + =
= + =
= + =
= + =
z
s
s
s

7/21/2011 Lecture 14-18 41
) (
1
t
) (
2
t
) , (
12 11 1
a a = s
) , (
22 21 2
a a = s
) , (
32 31 3
a a = s
) , (
2 1
z z = z
z s
d
,
1
z s
d
,
2
z s
d
,
3
1
E
3
E
2
E
Does expressing s
i
(t) and
r(t)=z(t) in terms of the
orthogonal basis
functions help us in
computing the energy
of the signals ?
Yes !!!!
Waveform Energy
The energy Ei of a waveform can be
expressed as a function of its orthogonal
components (see equation 3.17 for derivation):
If orthonormal basis functions are used (i.e. Kj=1)
the normalized energy over a symbol duration T
is (see equation 3.17):
7/21/2011 Lecture 14-18 42
{ }
M
i
i
t s
1
) (
=
2
1
ij
N
j
j i
a K E

=
=
Waveform energy
=
=
N
j
ij i
a E
1
2
Normalized Waveform energy
P
b
versus (E
b
/N
o
) Curves for DCS
The performance of a DCS can be
characterized by looking at the bit
error probability P
b
versus (E
b
/N
o
)
curves as shown in Figure 3.6.
The curves indicate the minimum
required (E
b
/N
o
) in order to meet the
bit error rate P
b
required for proper
detection of the received signals.
7/21/2011 Lecture 14-18 43
P
b
versus (E
b
/N
o
) Curves for DCS
7/21/2011 Lecture 14-18 44
For a given P
b
= Po, select
the minimum required
(E
b
/N
o
)= x
o
for accurate
detection of the received
signal.
For a given P
b
= P
o
, the smaller the required (E
b
/N
o
) the more efficient
the detection of the received signal.
Detection of Binary Signals in AWGN
Maximum Likelihood
receiver
The received signal is:
z(T)= a
i
(T) + n
o
(T)
a
i
(T)desired signal
n
o
(T)noise signal
If we assume that the noise
signal is AWGN, then z(T) will
also be a Gaussian random
variable with a mean of either
a
1
(s
1
=1 was sent) or a
2
(s
2
=0
was sent).
7/21/2011 Lecture 14-18 45
Maximum Likelihood receiver
Assume P(s
1
) prob that signal s
1
(t) was sent.
Assume P(s
2
) prob that signal s
2
(t) was sent.
Both P(s
1
) and P(s
2
) are known a priori.
Assume p(z/s
1
) the conditional probability
density function that indicates that z is received
assuming s1 was sent.
Assume p(z/s
2
) the conditional probability
density function that indicates that z is received
assuming s
2
was sent.
H
1
and H
2
are the two possible hypotheses.
7/21/2011 Lecture 14-18 46
Conditional Probability Density Functions (pdf)
7/21/2011 Lecture 14-18 47
z(T) is a voltage signal that is proportional to the energy of
the received symbol.
The detection decision is based on the energy and not the shape
of the received signal.
p (z /s
i
)
Conditional
Probabilities
Shaded area is prob of error
s1 sent but classified as s2
The detection piece of the
receiver is described as :
The main question is how to
choose the threshold ( ) such
that the probability of error
is minimized ?
7/21/2011 Lecture 14-18 48

The likelihood ratio test can be rewritten as:
If the product on the left is greater than the
one on the right, then the detector chooses
H
1
it assumes signal s
1
(t) was sent.
If not, the detector chooses H
2
it assumes
signal s
2
(t) was sent.
) ( * ) ( ) ( * ) (
2 2 1 1
2
1
s P s z p s P s z p
H
H
s
>
7/21/2011 Lecture 14-18 49
If P(s
1
)= P(s
2
) , we obtain the optimum
threshold ( ) that will minimize the
probability of making an incorrect decision
The detector will choose the hypothesis

(H) of the signal with the maximum
likelihood max likelihood detector.
) ( ) (
2 1
2
1
s z p s z p
H
H
s
>
o
7/21/2011 Lecture 14-18 50

2
2 1
0
a a +
=
Error Probability
An error of a binary detector can occur
under two conditions:
Case 1=P(e/s
1
): Signal s
1
(t) is sent, but the
channel noise forces the receiver output to
choose s
2
(t) because : z (T) <
Case 2=P(e/s
2
): Signal s
2
(t) is sent, but the
channel noise forces the receiver output to
choose s
1
(t) because : z (T) >
o
7/21/2011 Lecture 14-18 51

o

Error Probability P
B
The probability P
B
represents the area under the tail of
the likelihood pdfs p(z/s
1
) or p(z/s
2
) that falls on the
incorrect side of the threshold.
P
B
can be computed by integrating p(z/s
1
) from - to
( ) or p(z/s
2
) from ( ) to +
o
7/21/2011 Lecture 14-18 52

Error Probability
The probability of errors P(e/s
1
) and P(e/s
2
)
are written as follows:
( )dz s z p s H P s e P
}
= =
0
2 2 1 2
) ( ) (
( )dz s z p s H P s e P
}

= =
0
1 1 2 1
) ( ) (

7/21/2011 Lecture 14-18 53
Error Probability
The total probability of bit error P
B
is the
sum of probabilities of all the ways that an
error can occur. For the binary case:
) ( ) ( ) , (
2
1
2
1
i
i
i
i
i B
s P s e P s e P P

= =
= =
) ( ) ( ) ( ) (
2 2 1 1
s P s e P s P s e p P
B
+ =
7/21/2011 Lecture 14-18 54
Error Probability
If P(s
1
)= P(s
2
) = , then P
B
becomes:
Since the pdfs are symmetrical we get:
) (
2
1
) (
2
1
2 1 1 2
s H P s H P P
B
+ =
) (
2
1
) (
2
1
2 1
s e P s e P P
B
+ =
7/21/2011 Lecture 14-18 55
) ( ) (
2 1 1 2
s H P s H P P
B
= =
Error Probability
P
B
can be computed as follows:
We obtain the solution:
Q(x) is called the complementary error
function or co-error function (Appendix B).

2
0
is the variance of the output noise
(average noise power).
( )dz s z p s H P P
a a
B
}
+
=
= =
2
2 2 1
2 1
0
) (
7/21/2011 Lecture 14-18 56

|
|
.
|
\
|

=
0
2 1
2o
a a
Q P
B
Error Probability
Q(x) can be approximated for values of
x>3 using equation 3.44
At this stage, we have chosen the value of
the threshold ( ) that optimizes the
system by minimizing P
B
.
o
7/21/2011 Lecture 14-18 57

( )
|
|
.
|
\
|
H
~
2
exp
2
1
2
x
x
x Q
The Matched Filter
Next, we will attempt to optimize the
receiving filter by maximizing the argument
of Q(x) . Thats because P
B
decreases as x
increases.
The matched filter is a linear filter designed
to provide the maximum signal-to-noise
power ratio at the output, for a given
transmitted symbol waveform.
7/21/2011 Lecture 14-18 58
L12 Begin
Receiver structure
7/21/2011 Lecture 14-18 59
T = Symbol duration
The Matched Filter
The output of the sampler in step 1 is:
a
i
(t): is the signal at the filter output

2
0
: is the average noise power .
Our goal is to find the transfer function
H
o
(f) of a filter that maximizes the above
equation (3.45).
3.45 Equ
2
0
2
o
i
T
a
N
S
=
|
.
|
\
|
7/21/2011 Lecture 14-18 60
The Matched Filter
The maximum output (S/N)
T
depends on
the input signal energy and the power
spectral density of the noise.
The max (S/N)
T
does not depend on the
particular shape of the waveform.
3.52 Equ
2
max
0
N
E
N
S
T
=
|
.
|
\
|
3.53 Equ df S(f)
2
-
}
= E
7/21/2011 Lecture 14-18 61
The Matched Filter
The maximum output (S/N)
T
equation holds
true iff we employ a filter with the optimum
transfer function shown below:
The optimum filter transfer function is a
mirror image of the message signal s(t),
delayed by the symbol time duration T.
s s
=
Eq3.56 elsewhere 0
T t 0 t) - ks(T
(t) h
0
7/21/2011 Lecture 14-18 62
Matched filter receiver
Problem:
Design the receiver filter such that the SNR is
maximized at the sampling time when
is transmitted.
Solution:
The optimum filter, is the Matched filter, given by
which is the time-reversed and delayed version of the conjugate of
the transmitted signal
7/21/2011
Lecture 14-18
63
) (t h
) ( ) ( ) (
*
t T s t h t h
i opt
= =
) 2 exp( ) ( ) ( ) (
*
fT j f S f H f H
i opt
t = =
M i t s
i
,..., 1 ), ( =
T 0 t
) (t s
i
T 0 t
) ( ) ( t h t h
opt
=
The Matched Filter
The output z(T) of step 1 is therefore the
convolution of the received input waveform r(t)
with the impulse response of the filter h(t).
If we choose t = T, and k=1 the equation is:
( ) ( ) t t t d t h r t h t r t z
t
}
= - =
0
) ( ) ( ) (
( ) ( ) 3.59 Eq# ) (
0
t t t d s r T z
T
}
=
7/21/2011 Lecture 14-18 64
( ) ( ) | | ( ) | | t t t t t t d t T s r d t T s r t z
t t
} }
+ = =
0 0
) (
The Matched Filter = The Correlator
Equ # 3.59 indicates that z(T) can be
obtained by computing the cross -
correlation of r(t) and s(t).
Therefore, the receiver filter can be
implemented as follows:
The received signal r(t) is correlated with each
prototype signal s
i
(t) (i=1 M) using a bank of M
matched filters or M correlators.
The correct signal s
i
(t) will be chosen as the one
that yields the maximum output z
i
(t).
7/21/2011 Lecture 14-18 65
Implementation of matched filter receiver
7/21/2011 Lecture 14-18 66
Bank of M matched filters
Matched filter output:
Observation
vector
) ( ) ( t T s t r z i
i
- =
-
M i ,..., 1 =
) ,..., , ( )) ( ),..., ( ), ( (
2 1 2 1 M M
z z z T z T z T z = = z
(
(
(
M
z
z
1
z =
) (t r
) (
1
T z
) (
*
1
t T s
) (
*
t T s
M

) (T z
M
z
Example of implementation of
matched filter receivers
7/21/2011 Lecture 14-18 67
(
(
(
2
1
z
z
z =
) (t r
) (
1
T z
) (
2
T z
z
Bank of 2 matched filters
T
T 0
0
T
A
T
A
T t
) (
2
t s
T
A
0
T t
) (
1
t s
T
A
0
Example of matched filter receivers using basic
functions
Number of matched filters (or correlators) is reduced by 1 compared to
using matched filters (correlators) to the transmitted signal.
7/21/2011 Lecture 14-18 68
T t
) (
2
t s
T t
) (
1
t
T
1
0
| |
1
z
z =
) (t r z
1 matched filter
T t
) (
1
t
T
1
0
1
z
T
A
0
T t
) (
1
t s
T
A
0
Optimizing Error Performance
The Matched Filter
The matched filter is therefore the
optimum receiving filter that maximizes the
argument of Q(x) . Thats because P
B
decreases as x increases.
Therefore, the matched filter provides the
maximum distance between the two
prototype signals s
1
(t)=a
1
and s
2
(t)=a
2
at the
receiver.
|
|
.
|
\
|

=
0
2 1
2o
a a
Q P
B
7/21/2011 Lecture 14-18 69
Choosing the right signals/waveforms
P
B
can be re-written as follows:
Where () is defined as the time cross-correlation
coefficient between s
1
(t) and s
2
(t) .
() indicates the measure of similarity between two
signals s
1
(t) and s
2
(t) .
() is the angle between the s
1
(t) and s
2
(t) .
Eq#3.68
) 1 (
0
|
|
.
|
\
|

=
N
E
Q P
b
B

Eq#3.64b ) cos(u =
7/21/2011 Lecture 14-18 70
If = 90 s
1
(t) and s
2
(t) are orthogonal to
each other = cos(90)= 0
If = 180 s
1
(t) and s
2
(t) are antipodal
signals = cos(180)= -1
Eq#3.71
0
|
|
.
|
\
|
=
N
E
Q P
b
B
Eq#3.70
2
0
|
|
.
|
\
|
=
N
E
Q P
b
B
7/21/2011 Lecture 14-18 71
7/21/2011 Lecture 14-18 72
Note: Antipodal signals are easier to detect because there is
more distance between the two signals less chance of error.
Matched Filter detection of Antipodal Signals
7/21/2011 Lecture 14-18 73
7/21/2011 Lecture 14-18 74
Error Probability for Binary Signaling
The bipolar signals are easier to detect
under AWGN compared to unipolar signals.
Eq#3.71
0
|
|
.
|
\
|
=
N
E
Q P
b
B
Eq#3.70
2
0
|
|
.
|
\
|
=
N
E
Q P
b
B
7/21/2011 Lecture 14-18 75
Eq#3.71
0
|
|
.
|
\
|
=
N
E
Q P
b
B
Given the same
(Eb/No)=10 dB the P
b
for Bipolar is 100
times better than that
of Unipolar
Pb = 10
-3
for Unipo
Pb = 10
-5
for Bipo
Given the same P
b
=
10
-4
Bipolar requires
3dB less power than
Unipolar
(Eb/No)= 11 dB for
Unipo
(Eb/No)= 8 dB for Bipo
7/21/2011 Lecture 14-18 76
Intersymbol Interference (ISI)
ISI can be reduced using an equalizing filter
7/21/2011 Lecture 14-18 77
The best ISI filter is the Nyquist filter H(f)
Minimum bandwidth required in order to detect
signals arriving at Rs symbols/sec without ISI is Rs/2
Hertz.
The Ideal Nyquist filter is the sync function shown in
Fig 3.16 h(t) = sinc (t/T)
To avoid ISI, the sampling timing must be done at each
time T.
At t=0 sampler detects only h(t) since h(t-T)=0
At t=T sampler detects only h(t-T) since h(t)= 0
7/21/2011 Lecture 14-18 78
The best ISI filter is the Nyquist filter H(f)
Minimum bandwidth required in order to detect
signals arriving at Rs symbols/sec without ISI is Rs/2
Hertz.
7/21/2011 Lecture 14-18 79
BW=1/2T
Practical Filter is a Raised Cosine filter H(f)
Insert Figure 3.17
7/21/2011 Lecture 14-18 80
Required Bandwidth
for baseband signals
W = 0.5(1+ r)Rs
Required Bandwidth
for bandpass signals
W = (1+ r)Rs
See Example 3.3 for
more information.
Bandwidth requirements
Insert Figure 3.17
7/21/2011 Lecture 14-18 81
Required Bandwidth
for baseband signals
W = 0.5(1+ r)Rs
Required Bandwidth
for bandpass signals
W = (1+ r)Rs
Signal space contd
To find an orthonormal basis functions for a given
set of signals, Gram-Schmidt procedure can be
used.
Gram-Schmidt procedure:
Given a signal set , compute an orthonormal
basis
1. Define
2. For compute
If let
If do not assign any basis function.
3. Renumber the basis functions such that basis is
This is only necessary if for any i in step 2.
Note that
7/21/2011 Lecture 14-18 82
{ }
M
i i
t s
1
) (
=
{ }
N
j
j
t
1
) (
=
) ( / ) ( / ) ( ) (
1 1 1 1 1
t s t s E t s t = =
M i ,..., 2 =

=
> < =
1
1
) ( ) ( ), ( ) ( ) (
i
j
j j i i i
t t t s t s t d
0 ) ( = t d
i
) ( / ) ( ) ( t d t d t
i i i
=
0 ) ( = t d
i
{ } ) ( ),..., ( ), (
2 1
t t t
N

0 ) ( = t d
i
M N s
Example of Gram-Schmidt procedure
Find the basis functions and plot the signal space for
the following transmitted signals:
Using Gram-Schmidt procedure:
7/21/2011 Lecture 14-18 83
T t
) (
1
t s
T t
) (
2
t s
) ( ) (
) ( ) (
) ( ) (
2 1
1 2
1 1
A A
t A t s
t A t s
= =
=
=
s s

) (
1
t
-A A 0
1
s
2
s
T
A
T
A
0
0
T t
) (
1
t
T
1
0
0 ) ( ) ( ) ( ) (
) ( ) ( ) ( ), (
/ ) ( / ) ( ) (
) (
1 2 2
0
1 2 1 2
1 1 1 1
0
2
2
1 1
= =
= >= <
= =
= =
}
}
t A t s t d
A dt t t s t t s
A t s E t s t
A dt t s E
T
T
1
2

Demodulation and Detection Techniques

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Demodulation and Detection Techniques

Transféré par

Droits d'auteur :

Formats disponibles

EEGR 453

Orthogonal Representation of Waveforms

Orthogonal Representation of Waveforms

7/21/2011 Lecture 14-18 48

The detector will choose the hypothesis

7/21/2011 Lecture 14-18 50

7/21/2011 Lecture 14-18 51

Detection of Binary Signals in AWGN

7/21/2011 Lecture 14-18 52

7/21/2011 Lecture 14-18 56

7/21/2011 Lecture 14-18 57

Vous aimerez peut-être aussi