Development of Complex Curricula for Molecular Bionics and Infobionics Programs within a consortial* framework**
Consortium leader
PETER PAZMANY CATHOLIC UNIVERSITY
Consortium members
SEMMELWEIS UNIVERSITY, DIALOG CAMPUS PUBLISHER
The Project has been realised with the support of the European Union and has been co-financed by the European Social Fund ***
**Molekuláris bionika és Infobionika Szakok tananyagának komplex fejlesztése konzorciumi keretben
***A projekt az Európai Unió támogatásával, az Európai Szociális Alap társfinanszírozásával valósul meg.
Ad hoc Sensor Networks
Detection and channel equalization
Érzékelő mobilhálózatok
Detekció és csatornakiegyenlítés
Dr. Oláh András
Lecture 4 review
• Advantage of digital modulation
• Bandwidth of a signal
• ISI-free system requirements
• IQ modulator
• Constellation and eye diagrams
• Tradeoff between spectral efficiency and power efficiency
• Linear and constant envelope modulation scheme
• Spread Spectrum Modulation
Outline
• Signal space representation
• Optimal detection of signal in AWGN (Bayesian decision)
• Probability of error (BER and SER)
• Demodulation and detection for modulation schemes
• BER in fading channel
• Equalization
Structure of a wireless communications link
Receiver tasks
• Demodulation and sampling:
– Waveform recovery and preparing the received signal for detection – Improving the signal-to-noise ratio (SNR) using matched filter – Reducing ISI by using equalizer
– Sampling the recovered waveform
• Detection:
– Estimate the transmitted symbol based on the received samples
Receiver stucture
• Note: the bandpass model of detection process is equivalent to the baseband model because the received bandpass waveform is first transformed into a baseband waveform.
• Received signal in fading channel:
Model of the received signal
• Received signal in AWGN:
Simplifying the model
Ideal channel:
hc(τ) = δ(τ)
Signal space representation
• What is a signal space?
– Vector representations of signals in an N-dimensional space
• Why do we need a signal space?
– It is a mean to convert signals to vectors and vice versa.
– It is a mean 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 vectors. The signal which has the minimum distance to the received signal is estimated as the transmitted signal (in the case of Gaussian noise Euclidean metric provides the best – minimum error probability – decision).
Signal space representation (cont’)
• The inner product between two signals (functions):
• The distance in signal space is measured by calculating the norm:
– Norm of a signal:
– The norm between two signals:
( ), ( ) ( ) *( ) x t y t x t y t dt
∞
−∞
< >=
∫
(It is the cross-correlation between x(t) and y(t) )( ) ( ), ( ) ( ) 2 x
x t x t x t ∞ x t dt E
= < > =
∫
−∞ =, ( ) ( )
dx y = x t − y t
( It is the “length” of x(t))
(We refer to it as the Euclidean distance.)
Example of signal space representation
1 11 1 12 2 1 11 12
2 21 1 22 2 2 21 22
3 31 1 32 2 3 31 32
1 1 2 2 1 2
( ) ( ) ( ) ( , )
( ) ( ) ( ) ( , )
( ) ( ) ( ) ( , )
( ) ( ) ( ) ( , )
s t a t a t a a
s t a t a t a a
s t a t a t a a
r t r t r t r r
ψ ψ
ψ ψ
ψ ψ
ψ ψ
= + ⇔ =
= + ⇔ =
= + ⇔ =
= + ⇔ =
s s s r
Transmitted signal Received signal at matched filter output The Euclidean distance between signals r(t) and s(t)
2 2
, ( ) ( ) ( 1 1) ( 2 2) 1, 2,3
s ri i i i
d = s t −r t = a −r + a −r i =
Orthogonal signal space
• N-dimensional orthogonal signal space is characterized by N linearly independent functions { Ψ
j
(t)}
j=1,..,Ncalled basis functions. The basis functions must satisfy the orthogonality condition
where 0 ≤ t ≤ T, and i,j=1,…,N.
• If all K
i=1 , the signal space is orthonormal.
* 0
( ), ( ) ( ) ( )
0
T
i
i j i j i ji
K i j
t t t t dt K
i j
ψ ψ ψ ψ δ =
< >= = =
≠
∫
Orthogonal signal space (cont’)
• Example: 1 dimensional orthonormal signal space
• Example: 2 dimensional orthonormal signal space
Signal space representation
• Any arbitrary finite set of waveforms {si(t)}i=1,..,M, where each member of the set is of duration T, can be expressed as a linear combination of N orthonogal waveforms {Ψ
j(t)}j=1,..,N where N ≤ M : where
Waveform to vector conversion
Vector to waveform conversion
1
( ) ( )
N
i ij j
j
s t a ψ t
=
=
∑
* 0
1 1
( ), ( ) ( ) ( )
T
ij i j i j
j j
a s t t s t t dt
K ψ K ψ
= < >=
∫
0 ≤ ≤t T
2 1
N
i j ij
j
E K a
=
=
∑
Waveform energy
1 2
( , ,..., )
i = a ai i aiN
Vector rep. s
of waveform
1,..., i = M 1,...,
j = N
Gram-Schmidt procedure
• To find an orthonormal basis functions for a given set of signals, the Gram- Schmidt procedure can be used:
• given a signal set {si(t)}i=1,..,M , compute an orthonormal basis {Ψ
j(t)}j=1,..,N
1. Define
2. For i=1,…,M compute
If bi(t)≠0 let
If bi(t)=0, do not assign any basis function.
3. Renumber the basis functions such that basis is {Ψ
1(t), Ψ
2(t),…, Ψ
N(t)}
This is only necessary if bi(t)=0 for any i in step 2. Note that N≤M.
1( )t s t1( ) / E1 s t1( ) / s t1( )
ψ = =
1
1
( ) ( ) ( ), ( ) ( )
i
i i i j j
j
b t s t − s t ψ t ψ t
=
= −
∑
< >( ) ( ) / ( )
i t b ti b ti
ψ =
• Find the basis functions and plot the signal space for the following transmitted signals:
• Using Gram-Schmidt procedure:
2 2
1 1
0
1 1 1 1
2 1 2 1
0
2 2 1
( )
( ) ( ) / ( ) /
( ), ( ) ( ) ( )
( ) ( ) ( ) ( ) 0
T
T
E s t dt A
t s t E s t A
s t t s t t dt A
b t s t A t
ψ
ψ ψ
ψ
= =
= =
< >= = −
= − − =
∫
∫
Gram-Schmidt procedure (cont’)
White noise in the orthonormal signal space
• AWGN, n(t), can be expressed as
• Vector representation of noise:
n=(n1,n2,…,nN)
• {Ψ
j(t)}j=1,..,N independent zero-mean Gaussain random variables with variance
σ(ni)=N0/2
( ) ˆ( ) ( ) n t = n t + n tɶ
Noise projected on the signal space
which has an impact on the detection process.
Noise outside of the signal space
( ), ( )
j j
n =< n t1 ψ t > < n tɶ( ),ψ j( )t >= 0
ˆ( ) ( )
N
j j j
n t n ψ t
=
=
∑
1,..., j = N
Detection of signals at the output of AWGN
• Detection problem: given the observation vector r, perform a mapping from r to an estimate mˆ of the transmitted symbol, mi , such that the average probability of error in the decision is minimized.
– Signal vector si is deterministic.
– Elements of noise vector n are i.i.d Gaussian random variables with zero-mean and variance N0/2. The noise vector pdf is
– The elements of observed vector r are independent Gaussian random variables.
Its pdf is
( )
2 / 2
0 0
( ) 1 N exp
p πN N
= −
n
n n
( )
2 / 2
( | )i 1 N exp i
p πN N
−
= −
r
r s r s
Detection of signal at the output of AWGN (cont’)
• Optimum decision rule (maximum a posteriori probability):
– Applying Bayes’ rule gives:
– For equal probable symbols, the optimum decision rule (maximum posteriori probability) is simplified to:
– or equivalently:
which is known as maximum likelihood.
1,...,
: arg max Pr( | )i
i M
k s
= = r
( ) ( )
( )
1,...,
: arg max i i
i M
p s
k p s
p
= = r
r
r r
1,...,
( )
: arg max i
i M
k p s
= = r r
( ( ) )
1,...,
: arg max ln i
i M
k p s
= = r r
ˆ k
m = m
ˆ k
m = m
ˆ k
m = m
ˆ k
m = m
Detection of signal at the output of AWGN (cont’)
• Partition the signal space into M decision regions, D1,D2,…,DM.
• Restate the maximum likelihood decision rule as follows:
vector r lies inside region Dk if
for all k≠i, that means m~=mk.
( ( ) ) ( )
( ) { }
2
1,..., 1,..., 0 / 2 0
2
2 2
1,..., 0 / 2 0 1,..., 1,...,
: arg max ln arg max ln 1 exp
arg max ln 1 arg max arg min
i
i N
i M i M
i
i i
i M N i M i M
k p s
N N
N N
π π
= =
= = =
−
= = −
−
= − = − − = −
r
r r s
r s r s r s
ˆ k
m = m
k i
− < − r s r s
Detection of signal at the output of AWGN (cont’)
• The optimal decision (cont’):
( ( ) ) { 2} {
2 2}
1,..., 1,..., 1,...,
1,..., 1
: arg max ln arg max arg max 2
arg max
2
i i i i
i M i M i M
N
i j ij
i M j
k p s
r a E
= = =
= =
= = − − = − + ⋅ − =
= −
∑
r r r s r r s s
The probability of symbol error
• Erroneous decision: for the transmitted symbol mi or equivalently signal vector si, an error in decision occurs if the observation vector r does not fall inside region Di.
– Probability of erroneous decision for a transmitted symbol:
or equivalently
– Probability of correct decision for a transmitted symbol:
thus the probability of erroneous decision:
e( i) Pr( ˆ i and i sent) P m = m ≠ m m
Pr(mˆ ≠ mi) = Pr(mi sent)Pr( does not lie inside r D mi i sent)
( )
( )
c Pr( ˆ ) Pr( sent)Pr( lies inside sent) ( )
i
i i i i i
i i
D
P m m m m D m
p m p m d
∈
= = = =
= ⋅
∫
rr
r
r r
( )
e( i) ( i) i 1 c( i)
P m = p m ⋅
∫
pr r m dr = − P mThe probability of symbol error (cont’)
• Average probability of symbol error :
• Example for binary PAM:
( )
E e c
1 1
1
( ) 1 ( )
1 1 ( | )
i
M M
i i
i i
M
i
i D
P M P m P m
p m d M
= =
= ∈
= = − =
= −
∑ ∑
∑ ∫
rr
r r
b
B E
0
(2) 2E
P P Q
N
= =
1 2
e 1 e 2
0
( ) ( ) / 2
P m P m Q / 2
N
−
= =
s s
( )
2
1 2
2
u
x
Q x e du
π
∞ −
=
∫
Recall:
Example of Symbol error prob. for PAM signals
Demodulation and detection
• Demodulation: The receiver signal is converted into a baseband signal, and then filtered and sampled.
• Detection: Sampled values are used for detection and we use a decision rule such as the ML detection rule.
• The matched filters output (observation vector r) is the detector input and the decision variable is a function of r. We know that for calculating the probability of symbol error, we need to determine
Pr( lies inside r D mi i sent)
M-ary Pulse Amplitude modulation
• One dimensional modulation, demodulation and detection
2
( )
( ) cos
i i c
s t t
α T ω
=
( )
( )
1
1
b
2 2
b 2
s b
( ) ( ) 1, , ( ) 2 cos
(2 1 )
2 1
( 1)
3
i i
c
i
i i
s t a t i M
t t
T
a i M E
E E i M
E M E
ψ
ψ ω
= =
=
= − −
= = − −
= − s
…
b 2
E 2
0
2( 1) 6log
( )
1 E
M M
P M Q
M M N
= −
−
M-ary Phase Shift Keying (M-PSK)
• Two dimensional modulation, demodulation and detection
2 s 2
( ) cos
i c
E i
s t t
T M
ω π
= +
( ) ( )
1 1 2 2
1 2
1 s 2 s
2 s
( ) ( ) ( ) 1, ,
2 2
( ) cos ( ) sin
2 2
cos sin
i i i
c c
i i
i i
s t a t a t i M
t t t t
T T
i i
a E a E
M M
E E
ψ ψ
ψ ω ψ ω
π π
= + =
= = −
= =
= = s
…
(
2)
s E0
2 log
( ) 2 M E sin
P M Q
N M
π
≈
M-ary Quadrature Amplitude Mod. (M-QAM)
• Two dimensional modulation, demodulation and detection
( ) ( )
1 1 2 2
1 2
1 2 s
( ) ( ) ( ) 1, ,
2 2
( ) cos ( ) sin
2( 1)
where and are PAM symbols and
3
i i i
c c
i i
s t a t a t i M
t t t t
T T
a a E M
ψ ψ
ψ ω ψ ω
= + =
= =
= −
…
(
1 2)
( 1, 1) ( 3, 1) ( 1, 1)
( 1, 3) ( 3, 3) ( 1, 3)
,
( 1, 1) ( 3, 1) ( 1, 1)
i i
M M M M M M
M M M M M M
a a
M M M M M M
− + − − + − − −
− + − − + − − −
=
− + − + − + − + − − +
⋯
⋯
⋮ ⋮ ⋮ ⋮
⋯
( )
( ) 2 i cos
i c i
s t E t
T ω ϕ
= +
M-QAM (cont’)
b 2
E
0
1 3log
( ) 4 1
1 E
P M Q M
M N
M
= −
−
( )
1
2 s
( ) ( ) 1, ,
2
( ) cos
0
M
i ij j
j
s
i i ij
i i
s t a t i M
E i j
t t a
T i j
E E
ψ
ψ ω
=
= =
=
= =
≠
= =
∑
s
…
M-ary Frequency Shift keying (M-FSK)
• Multi-dimensional modulation, demodulation and detection
( ) ( )
s s
2 2
( ) cos cos ( 1)
1
2 2
i i c
E E
s t t t i t
T T
f T
ω ω ω
ω π
= = + − ∆
∆ = ∆ =
( )
sE
0
( ) 1 E
P M M Q
N
≤ −
Coherent vs. non-coherent detection
Coherent detection:
• It requires carrier phase recovery at the receiver and hence, circuits to perform phase estimation.
• The sources of carrier-phase mismatch at the receiver:
– Propagation delay causes carrier-phase offset in the received signal.
– The oscillators at the receiver which generate the carrier signal, are not usually phased locked to the transmitted carrier.
Non-coherent detection:
• No need for a reference in phase with the received carrier.
• Less complexity compared to coherent detection at the price of
higher error rate.
Bit error probability vs. symbol error probability
• Number of bits per symbol:
k=log
2M
• For orthogonal M-ary signaling (M-FSK) :
• For M-PSK, M-PAM and M-QAM:
1 Bit
E Bit
E
2 / 2
2 1 1
lim 1
2
k k
k
P M
P M
P P
−
→∞
= =
− −
=
E
Bit P for E 1
P P
≈ k <<
Designing a Wireless Communication System
• Goals:
– Maximizing the transmission bit rate (R) – Minimizing probability of bit error (PBit) – Minimizing the required power (PTX or PE) – Minimizing required system bandwidth (B) – Maximizing system utilization (SE)
– Minimizing system complexity
• Limitations:
– The Nyquist theoretical minimum bandwidth requirement
– The Shannon-Hartley capacity theorem (and the Shannon limit) – Government regulations
– Technological limitations
– Other system requirements (e.g wireless sensor networks)
• Most famous result in communication theory.
• Channel capacity of AWGN channel:
– B : bandwidth
– C : channel capacity (bps) of real data (not retransmissions or errors). It is the maximum data rate at which error-free communication over the channel is performed.
• SEmax is fundamental limit that cannot be achieved in practice.
Maximum SE: Shannon’s theorem (1948)
Recall from Chapter 4
Claude Elwood Shannon (1916-2001)
S b
max 2 2
N 0
log 1 P log 1 E
C R
SE B P N B
= = + = +
Given a modulation scheme and a targeted BER, the communication system designer can determine the SE (spectral efficiency) and the PE (Eb/N0 required to maintain the average BER target).
Recall from Chapter 4
Power and bandwidth limited systems
• Two major communication resources:
– PTX transmit power – B channel bandwidth
• In many communication systems, one of these resources is more precious than the other. Hence, systems can be classified as:
– Power-limited systems: save power at the expense of bandwidth (for example by using coding schemes)
– Bandwidth-limited systems: save bandwidth at the expense of power (for example by using spectrally efficient modulation schemes)
M-ary signaling
• Bandwidth efficiency:
– Assuming Nyquist (ideal rectangular) filtering at baseband, the required passband bandwidth is:
• M-PSK and M-QAM (bandwidth-limited systems)
– SE increases as M increases.
• MFSK (power-limited systems)
– SE decreases as M increases
s s
1/ [Hz]
B = T = R
2 s
log 1
[bits/s/Hz]
M R
B = BT = BT
/ log2 [bits/s/Hz]
R B = M
/ log2 / [bits/s/Hz]
R B = M M
Design example of uncoded systems
• Design goals:
– The bit error probability at the modulator output must meet the system error requirement.
– The transmission bandwidth must not exceed the available channel bandwidth.
Design example for uncoded systems (cont’)
• Task: choose a modulation scheme that meets the system requirements.
• Solution:
RX 5
Bit 0
AWGN channel with 4000 [Hz]
53 [dBHz] 9600 [bits/s] 10 BC
P R P
N
−
=
= = ≤
s 2
s b RX
2 2
0 0 0
5
E s 0
6 5
E Bit
2
Band-limited channel M-PSK modulation
8 / log 9600 / 3 3200 [sym/s] 4000 [Hz]
(log ) (log ) 1 62.67
( 8) 2 2 / sin( / ) 2.2 10
( )
7.3 10 10 log
c
c
R B
M R R M B
E E P
M M
N N N R
P M Q E N M
P P M
M
π −
− −
> ⇒ ⇒
= ⇒ = = = < =
= = =
= ≈ = ×
≈ = × <
b RX
0 0
c b 0
s 2 c
s b RX
2 2
0 0 0
E
1 6.61 8.2 [dB]
and relatively small / power-limited channel M-FSK
16 /(log ) 16 9600 / 4 38.4 [ksym/s] 45 [kHz]
(log ) (log ) 1 26.44
( 16)
E P
N N R
R B E N
M B MR MR M B
E E P
M M
N N N R
P M M
= = =
< ⇒ ⇒
= ⇒ = = = × = < =
= = =
= ≤ − s 5 Bit 1 E 6 5
0
1 2
exp 1.4 10 ( ) 7.3 10 10
2 2 2 1
k k
E P P M
N
− − − −
− = × ⇒ ≈ = × <
−
Design example for uncoded systems (cont’)
• Task: choose a modulation scheme that meets the system requirements.
• Solution:
RX 5
Bit 0
AWGN channel with 45 [kHz]
48 [dBHz] 9600 [bits/s] 10 BC
P R P
N
−
=
= = ≤
BER in fading channels
• Wireless channel impairments:
– system generated interference – Shadowing, large-scale path loss
– Multipath Fading, rapid small-scale signal variations (ISI) – Doppler Spread due to motion of mobile unit
• All can lead to significant distortion or attenuation of the received signal (SNR) which degrade Bit Error Rate (BER) of digitally modulated signal.
BER in fading channels (cont’)
• We have BER expressions for non-fading AWGN channels.
• The pdf of SNR in Rayleigh fading channel:
where is the average SNR.
• The BER in case of Rayleigh fading can be computed as:
( )
1 e SNRSNRpdf SNR
SNR
= −
SNR
( ) ( ) ( )
Rayleigh AWGN
0
BER SNR BER SNR pdf SNR dSNR
= ∞
∫
⋅This is a serious problem!
BER in fading channels (cont’)
• Four techniques are used to improve received signal quality and lower BER:
– Equalization
– Diversity (expensive and resource consuming) – Channel Coding
– Interleaving
• They are used independently or together
• We will consider equalization
Inter-Symbol Interference (ISI)
• ISI in the detection process due to the filtering effects of the system
• Overall equivalent system transfer function G(f)=GT(f )H(f )GR(f ) – creates echoes and hence time dispersion
– causes ISI at sampling time: k k k i k i
i k
x y n h y −
≠
= + +
∑
Equalization
Equalization (cont’)
• MLSE (Maximum likelihood sequence estimation)
• Filtering
– Transversal filtering: A weighted tap delayed line that reduces the effect of ISI by proper adjustment of the filter taps
• Zero-forcing equalizer (Peak-Distortion Criterion)
• Minimum mean square error (MMSE) equalizer (MSE criterion)
– Decision feedback: using the past decisions to remove the ISI contributed by them
– Adaptive equalizer
Linear channel equalization
• The simplified model and notations:
0
0 0 0 0 0 0
L
k k n n k
n
J J L L J J
k j k j j k j n n k j j k j n n j k j
j j n n j j
x h y
y w x w h y w h y w
ν
ν ν
= −
− − − − − − −
= = = = = =
= +
= = + = +
∑
∑ ∑ ∑ ∑ ∑ ∑
ɶ
0
k k n n k
n
k k n n k
n k
y q y
q y q y
η
η
−
≠ −
= + =
= + +
∑
∑
ɶ
ISI
Equivalent filter
{ }
k2 0 2E η = N w
(
0, 0)
k N N
η ∼ w
Zero forcing equalizer
• The zero forcing equalizer minimizes the peak-distortion:
• The ISI can be completely eliminated if:
• In the case of finite-length equalizer:
( )ZF
opt
, 0
: arg min ( ) arg min n
n n
PD q
≠
= =
∑
w w
w w
0
1 0
0 0
n n
q n
δ n =
= =
≠
0 0
1 0
0 1,...,
J
j n j n
j
w h n
n J
− δ
=
=
= =
=
∑
0 0
1 0 1
2 1 0 2
0 0 0 1
0 0 0
0 0
0
0 0 0 0
h w
h h w
h h h w
h w
=
…
…
…
⋮ ⋮ ⋮ ⋱ ⋮ ⋮
… ( )ZF
(J+1) (x J+1) ⋅ opt =
H w δ
( )ZF 1 opt
= − ⋅
w H δ
Eye open condition:
1 0
1
L n n
h h
= <
∑
Zero forcing equalizer (cont’)
Problems:
1. It eliminates ISI, but enhances the effect of additive noise (SNR ↓) 2. The hn is unknown
( ) ( ) ( )
0
1
J
l l k j k j k l
j
w k w k y w k x − y −
=
+ = − ∆ − −
∑
with learning set τ(K)
MMSE equalizer
• The filter taps are adjusted such that the MSE of ISI and noise power at the equalizer output is minimized:
• Wiener filtering problem! (see DSP course) – Autocorrelation matrix
– Crosscorrelation vector:
• The linear equations in matrix form:
(MMSE)
(J+1) (x J+1) ⋅ opt =
R w r (optMMSE) 1
= − ⋅
w R r
(MMSE)
opt : arg min k j k j
j
E y w x −
= −
∑
w
w
{ }
ij k i k j
R = E x − x −
{ }
i k k i
r = E y x −
, 0,..., i j = J
0,..., i = J
MMSE equalizer (cont’)
Problems:
1. The hn is unknown
with learning set τ(K)
( ) ( ) ( )
0
1
J
l l k j k j k l
j
w k w k y w k x − x −
=
+ = − ∆ − −
∑
Summary
• In the signal space: calculate signals energy and Euclidean distances between signals.
• The receiver signal is converted into baseband, filtered and sampled (demodulation).
• Sampled values are used for detection using a decision rule such as the ML detection rule (detection).
• A communication system designer can determine the modulation scheme according to the SE and the PE.
• All of these can lead to significant distortion or attenuation of received signal (SNR) which degrade (BER) of digitally modulated signal.
• A weighted tap delay-line that reduces the effect of ISI by proper adjustment of the weights of the filter taps.
• Next lecture: Multiple channel access