Beruflich Dokumente
Kultur Dokumente
Francois Horlin
1
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
2
References
3
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
4
Introduction
5
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
6
Outline: Signal demodulation
Correlation demodulator
Matched filter
7
Correlation demodulator
8
Correlation demodulator
9
Gaussian noise variables nk
Mean of nk : Z
E [nk ] = E [n(t)] fk (t)dt = 0
t=
Covariance of nk :
Z Z
E [nk nl ] = E [n(t)n(t )] fk (t)fl (t )dtdt
t= t =
Z Z
N0
= (t t )fk (t)fl (t )dtdt
t= t = 2
Z
N0
= fk (t)fl (t)dt
2 t=
N0
= kl
2
10
Vector model
r = sm + n
with:
h iT
sm := sm1 smK
h iT
r := r1 rK
h iT
n := n1 nK
11
Conditional probability
K
Y
p(r|sm ) = p(rk |smk )
k=1
K
Y 1 1
= p exp (rk smk )2
(N0 ) N0
k=1
K
" #
1 1 X
= p exp (rk smk )2
(N0 )K N0
k=1
12
Remaining noise
Because the functions fk (t) span entirely the signal space but not the
noise space, there is a remaining noise process n (t):
K
X K
X
r(t) = smk fk (t) + nk fk (t) + n (t)
k=1 k=1
K
X
= rk fk (t) + n (t)
k=1
where:
K
X
n (t) := n(t) nk fk (t)
k=1
13
Remaining noise
14
Sufficient statistics
Therefore the remaining noise n (t) does not contain any relevant
information and the correlator outputs rk contain all the available
information for the symbol detection
15
Concept of matched filter
The matched filter h(t) to the signal s(t) is defined as h(t) := s(t)
The response of the filter h(t) to the signal s(t) is the auto-correlation
of the signal s(t):
Z +
y(t) = s( )h(t )d
=
Z +
= s( )s( t)d
=
16
Matched filter demodulator
17
Matched filter demodulator
18
Maximization of the SNR
19
Maximization of the SNR (demonstration)
The received signal is the initial signal s(t) corrupted by AWGN n(t):
20
Maximization of the SNR (demonstration)
The SNR is the signal power (the signal s(t) is deterministic) divided
by the noise power (the noise n(t) is random):
(y S )2
SNR :=
E [(y N )2 ]
21
Maximization of the SNR (demonstration)
hR i2
=
s( )h( )d
SNR := N0
R
2 =
h2 ( )d
Cauchy-Schwartz inequality:
Z 2 Z Z
g1 (t)g2 (t)dt g12 (t)dt g22 (t)dt
t= t= t=
22
Maximization of the SNR (demonstration)
2
Z
SNR = s2 ( )d
N0 =
2E
=
N0
23
Matched filter in the frequency domain
Z
H(f ) = s(t)ej2f t dt
t=
Z
= s(t)ej2f t dt
t=
Z
= s(t)ej2f t dt
t=
= S (f )
24
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
25
Outline: Symbol detection
Maximum a posteriori
Maximum likelihood
26
Detection of memoryless modulations
27
Maximum a posteriori detector
sM
m
AP
= max p(sm |r)
sm
28
Maximum a posteriori detector
Bayes rule:
p(r|sm ) p(sm )
p(sm |r) =
p(r)
The denominator is independent on the hypothetic transmitted signal
and can be neglected in the MAP criterion:
sM
m
AP
= max p(r|sm ) p(sm )
sm
29
Maximum likelihood detector
sM
m
L
= max p(r|sm )
sm
30
Minimize the Euclidian distance
sM
m
L
= max (ln p(r|sm ))
sm
K
!
X
= min (rk smk )2
sm
k=1
31
Select the signal with the largest correlation
Equivalently:
K K
!
X X
sM
m
L
= max 2 rk smk s2mk
sm
k=1 k=1
H 2
= max 2 r sm ksm k
sm
Z
= max 2 r(t)sm (t)dt Em
sm t=
32
Alternative receiver structure
33
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM)
performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
34
Binary PAM
r = sm + n
with:
s 1 = Eb
s 2 = Eb
Conditional probability:
1 1
p(r|sm ) = p exp (r sm )2
(N0 ) N0
35
Decision rule
s2 0 s1
36
Performance
37
Performance
P (e|s2 ) = P (e|s1 )
r !
2Eb
= Q
N0
38
M-ary PAM
39
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
41
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
42
Orthogonal binary FSK
Relationship input symbol/output correlator:
r = sm + n
with:
h iT
s1 := Eb 0
h iT
s2 := 0 Eb
and
h iT
r := r1 r2
h iT
n := n1 n2
43
Orthogonal binary FSK
Binary FSK:
s2
2E b
0
s1
44
Performance
45
Performance
P (e|s2 ) = P (e|s1 )
r !
Eb
= Q
N0
46
Comparison binary PAM / binary FSK
47
M-ary FSK
48
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
49
PSK
r = sm + n
with:
h iT
sm := Es cos 2
M (m 1) Es sin 2
M (m 1)
and
h iT
r := r1 r2
h iT
n := n1 n2
50
Decision rule
PSK: compute the phase of the received signal based on r and select
the signal sm whose phase is the closest (phase detector)
r2
r = arctan
r1
51
Probability distribution function of r
52
Probability distribution function of r
Phase-only PDF:
Z
pr (r ) = pvr ,r (vr , r )dvr
vr =0
" #
r
vr Es Es
Z
2
= exp vr + 2 vr cos r dvr
vr =0 N0 N0
53
Performance
Ps = P (e|s1 )
54
Performance
Es
s :=
N0
55
Performance
The bit error probability is difficult to derive from the symbol error
probability due to its dependence on the mapping of the bits onto the
corresponding signal phases
When a Gray code is used, two symbols corresponding to adjacent
signal phases differ only in a single bit
Since the most probable errors due to the noise result in the erroneous
selection of an adjacent phase to the true phase, most symbol errors
contain only a single bit error
Therefore the bit error probability can be approximated as:
Ps
Pb
log2 M
56
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
57
Phase synchronization requirements
58
Orthogonal FSK
59
FSK non-coherent demodulation
60
FSK non-coherent demodulation
r2I = nI2
61
Optimum detector
62
Computation p(r|s1 )
with:
R
2 I
2
1 (r1 Eb cos ) + (r1 Eb sin )
p(r1R , r1I |s1 , ) = exp
N0 N0
R 2 I 2
1 (r2 ) + (r2 )
p(r2R , r2I |s1 , ) = exp
N0 N0
63
Computation p(r|s1 )
2
1
Z
p(r|s1 ) = p(r|s1 , ) d
2 =0
2
1 1
= exp ((r1R )2 + (r1I )2 + (r2R )2 + (r2I )2 + Eb )
N0 N0
q
2 Eb
I0 (r1R )2 + (r1I )2
N0
64
Modified bessel function of order 0
12
10
8
I0(x)
0
0 0.5 1 1.5 2 2.5 3 3.5 4
x
65
Computation p(r|s2 )
Similarly:
2
1 1
p(r|s2 ) = exp ((r1R )2 + (r1I )2 + (r2R )2 + (r2I )2 + Eb )
N0 N0
q
2 Eb
I0 (r2R )2 + (r2I )2
N0
66
Optimum detector
67
FSK non-coherent receiver
68
Performance
69
Impact of signal correlation
70
Outline
Introduction
Signal demodulation
Symbol detection
Pulse-amplitude modulation (PAM) performance
Quadrature-amplitude modulation (QAM) performance
Frequency-shift keying (FSK) performance
Phase-shift keying (PSK) performance
Non-coherent receiver
Exercise
71
Exercise
72