Chapter 5 - Website of Professor Po
Download
Report
Transcript Chapter 5 - Website of Professor Po
Chapter 5 Signal-Space Analysis
Signal space analysis provides a
mathematically elegant and highly insightful
tool for the study of data transmission.
5.1 Introduction
Statistical model for a genetic digital communication
system
Message source: A priori probabilities for information
source
pi P(mi ) for i 1,2,...,M
Transmitter: The transmitter takes the message source
output mi and (en-)codes it into a distinct signal si(t)
suitable for transmission over the channel. So:
pi P(mi ) P( si (t )) for i 1,2,...,M
© Po-Ning [email protected]
Chapter 5-2
5.1 Introduction
si(t) must be a real-valued energy signal (i.e., a
signal with finite energy) with duration T.
Ei 0 si2 (t )dt .
T
Channel: The channel is assumed (in this text) linear
and with a bandwidth wide enough to pass si(t) with no
distortion.
Zero-mean additive white Gaussian noise (AWGN)
is also assumed (to facilitate the analysis).
© Po-Ning [email protected]
Chapter 5-3
5.1 A Mathematical Model
We can then simplify the previous system block diagram to:
Upon the receipt of x(t) for a duration of T, the receiver makes
the best estimate of mi. (We haven’t defined what the best is.)
© Po-Ning [email protected]
Chapter 5-4
5.1 Criterion for the “Best” Decision
Best = Minimization of the average probability of symbol
error.
M
Pe pi P(mˆ mi | mi )
i 1
This is optimum in the minimum probability of error
sense.
Based on this criterion, we can begin to design the
receiver that can give the best decision.
© Po-Ning [email protected]
Chapter 5-5
5.2 Geometric Representation of Signals
Signal space concept
Vectorization of the (discrete or continuous) signals
removes the redundancy in the signals, and provides a
compact representation for them.
Determination of the vectorization basis
Gram-Schmidt orthogonalization procedure
© Po-Ning [email protected]
Chapter 5-6
5.2 Gram-Schmidt Orthogonalization Procedure
Given v1 , v2 ,, vk , how tofind an orthonorma
l basis for them?
v
(step i ) Let u1 1 .
|| v1 ||
'
u
(step ii ) u2' v2 (v2 u1 )u1. Set u2 2' .
|| u2 ||
(step iii ) For i 3,4,...,
Let ui' vi (vi ui 1 )ui 1 (vi ui 2 )ui 2 (vi u1 )u1.
ui'
Set ui ' .
|| ui ||
(step iv ) T henu1 , u2 ,, uk forman orthonorma
l basis.
© Po-Ning [email protected]
Chapter 5-7
P ropert ies:
(i) vect or: v ( v1, ,vn )
n
(ii) inner product: v1 v2 v1i v2i
i 1
(iii) ort hogonal
, if int er product 0.
(iv) norm: || v || v12 vn2
( v) ort hosnorm
al, if inner product 0, and indivual norm 1.
(vi) linearlyindependent , if nonecan be represent ed as a linear
combinat ion of ot hers.
(vii) t riangleinequalit y: || v1 v2 |||| v1 || || v2 || .
© Po-Ning [email protected]
Chapter 5-8
( viii) Cauchy Schwartz inequality:
| v1 v2 ||| v1 || || v2 ||
with equality holds, if v1 av2 .
( xi) normsquare :
2 2
2
|| v1 v2 || || v1 || || v2 || 2v1 v2 .
( x) P ythagore
an property: If orthogonal
,
2 2
2
|| v1 v2 || || v1 || || v2 || .
( xi) Matrix transformation w.r.t.matrixA :
v1 Av2 .
( xii) eigenvalues w.r.t.matrixA :
solution of detA - I 0.
( xiii) eigenvectors w.r.t.eigenvalue :
solutionv of Av v .
© Po-Ning [email protected]
Chapter 5-9
5.2 Signal Space Concept for Continuous
Functions
P ropertiesfor continuousfunctions
(i) (complex valued) signal : z (t )
(ii) inner product: z (t ), zˆ(t ) a z (t ) zˆ* (t )dt.
b
(iii) orthogonal
, if inter product 0.
(iv) norm: z (t )
2
|
z
(
t
)
|
dt
a
b
( v) orthonorma
l, if inner product 0, and indivual norm 1.
(vi) linearlyindependent vectors,if nonecan be represented as a linear
combination of others.
(vii) triangleinequality: z (t ) zˆ(t ) || z (t ) || || zˆ(t ) || .
© Po-Ning [email protected]
Chapter 5-10
( viii) Cauchy Schwartz inequality:
z (t ), zˆ(t ) || z (t ) || || zˆ(t ) ||
with equality holds, if z (t ) a zˆ(t ). ( a is a complexnumber.)
( xi) normsquare :
|| z (t ) zˆ(t ) ||2 || z (t ) ||2 || zˆ(t ) ||2 z (t ), zˆ(t ) zˆ(t ), z (t ) .
( x) P ythagorean property: If orthogonal
, || z (t ) zˆ(t ) ||2 || z (t ) ||2 || zˆ(t ) ||2 .
( xi) T ransformation w.r.t.a functionC (t,τ ) :
zˆ(t ) a C (t , ) z ( )d
b
n
(Recall v1 j a jn v2 i .)
i 1
(xii.a)eigenvalues and eigenfunctions w.r.t.a continuousfunctionC (t, ) :
solutionsk and { k (t )} of k k (t ) a C (t , ) k ( )d
k 1
b
and C (t , ) can be represented as C (t , ) k (t ) k k ( ).
k 1
© Po-Ning [email protected]
Chapter 5-11
(xii.b) Give a deterministic function{s(t ), t [0, T )} and a set of
orthonorma
l basis { k (t )}1k thatcan span s(t ). T hen
s(t ) ak k (t ), 0 t T ,
k 0
where ak 0 s(t ) k (t )dt.
T
(xii.c) If orthonorma
l set { k (t )}1k K does not span thespace, then
K
it is possible thatsˆ(t ) ak k (t ) s(t ) for all choicesof {ak }1k K .
k 0
© Po-Ning [email protected]
Chapter 5-12
Problem : How to minimize the “energy” of e(t ) s(t ) sˆ(t ) ?
T oselect thecoefficients{ak } thatminimize e 2 (t )dt,
K
2
2
[
s
(
t
)
a
(
t
)
]
dt
k k
e (t )dt
k 1
a j
a j
K
2 s(t ) ak k (t ) j (t )dt
k 1
K
2 s(t ) j (t )dt 2 ak k (t ) j (t )dt
k 1
2 s(t ) j (t )dt 2a j 0. a j s(t ) j (t )dt.
© Po-Ning [email protected]
Chapter 5-13
a2 s(t ), 2 (t )
2 (t )
s(t )
a1 s(t ),1 (t )
e(t )
sˆ(t )
1 (t )
Hence, e(t ), sˆ(t ) e(t ) sˆ(t )dt 0
Interpretation
aj is the projection of s(t) onto the Yj(t)-axis.
(aj)2 is the energy-projection of s(t) onto the Yj(t)-axis.
© Po-Ning [email protected]
Chapter 5-14
s(t )
a1 s(t ),1 (t )
e(t )
sˆ(t )
a2 s(t ), 2 (t )
2 (t )
1 (t )
2
e
(t )dt e(t )[s(t ) sˆ(t )]dt e(t )s(t )dt e(t )sˆ(t )dt
e(t ) s(t )dt 0 e(t ) s(t )dt [ s(t ) sˆ(t )]s(t )dt
K
s (t )dt ak k (t ) s(t )dt
k 1
2
K
s (t )dt ak s(t ) k (t )dt
2
k 1
K
s 2 (t )dt ak2
k 1
© Po-Ning [email protected]
K
Notably, sˆ (t )dt ak2 .
2
k 1
Chapter 5-15
5.2 Signal Space Concept for Continuous
Functions
Completeness
If every finite energy signal satisfies
K
2
s
(
t
)
dt
a
k,
2
k 1
{ k (t )}1k K is a com pleteorthonorma
l set.
Example. Fourier series
2
2kt 2
2kt
cos
sin
,
T T
T 0 k
T
is a completeorthonorma
l set for signals
defined over[0, T ].
© Po-Ning [email protected]
Chapter 5-16
5.2 Gram-Schmidt Orthogonalization Procedure
Given v1 (t ), v2 (t ),, vk (t ), how tofind an orthonorma
l basis for them?
(step i )
v1 (t )
Let u1 (t )
.
|| v1 (t ) ||
'
u
(t )
(step ii ) u2' (t ) v2 (t ) (v2 (t ), u1 (t )) u1 (t ). Set u2 (t ) 2'
.
|| u2 (t ) ||
(step iii ) For i 3,4,...,
Let ui' (t ) vi (t ) vi (t ), ui 1 (t ) ui 1 (t ) vi (t ), u1 (t ) u1 (t ).
ui' (t )
Set ui (t ) '
.
|| ui (t ) ||
(step iv ) T henu1 (t ), u2 (t ),, uk (t ) forman orthonorma
l basis.
© Po-Ning [email protected]
Chapter 5-17
5.2 Geometric Representation of Signals
N
si (t ) sij f j (t ), 0 t T , i 1,2,...,M
j 1
{ fi }iN1 orthonorma
l
sij 0 si (t ) f j (t )dt,
T
i 1,2,...M , j 1,2,...,N
© Po-Ning [email protected]
Chapter 5-18
5.2 Geometric Representation of Signals
Through the signal space concept, si(t) (where 1 i M)
can be unambiguously represented by an N-dimensional
signal vector (si1, si2,…, siN) over an N-dimensional signal
space.
The design of transmitters becomes the selection of M
points over the signal space, and the receivers make a guess
about which of the M points was transmitted.
In the N-dimensional signal space,
length of the vector = energy of the signal
angle between vectors = energy correlation between
N
T
signals
2
2
2
si (t ), sk (t )
||
s
(
t
)
||
s
(
t
)
dt
s
i
ij
0 i
cos(ik )
j 1
|| si (t ) || || sk (t ) ||
© Po-Ning [email protected]
Chapter 5-19
5.2 Geometric Representation of Signals
the angle between vectors is independent of the basis
used.
From this view,
the transmitter may be viewed as a synthesizer, which
synthesizes the transmitted signal by a bank of N
multipliers.
the receiver may be viewed as an analyzer, which
correlates (product-integrate) the common input into
individual informational signal.
© Po-Ning [email protected]
Chapter 5-20
5.2 Geometric Representation
of Energy Signals
Illustration the geometric
representation of signals for the
case when N = 2 and M = 3
© Po-Ning [email protected]
Chapter 5-21
5.2 Euclidean Distance
After vectorization, we can then calculate the Euclidean
distance between two signals:
T
0
N
( si (t ) sk (t )) dt || si (t ) sk (t ) || ( sij skj )2
2
2
j 1
1, i j
Kroneckerdelta function: ij
0, i j
Applications : We may say thatorthonorma
lity means i (t ), j (t ) ij .
© Po-Ning [email protected]
Chapter 5-22
Example 5.1 Schwarz Inequality
Cauchy-Schwarz inequality and angle between signals
Cauchy-Schwarz inequality said that
s1 (t ), s2 (t )
2
|| s1 (t ) ||2 || s2 (t ) ||2 with equality holds if s1 (t ) cs2 (t ).
Also, angles between signals give that
s1 (t ), s2 (t )
cos(12 )
|| s1 (t ) || || s2 (t ) ||
Hence, Cauchy-Schwarz inequality is equivalently
stated as:
| cos(12 ) |2 1 with equality holdsif 12 0 or
© Po-Ning [email protected]
Chapter 5-23
5.2 Basis
The (complete) orthonormal basis for a signal space is not
unique!
So the synthesizer and analyzer for the transmission of
the same informational messages are not unique!
One way to determine a set of orthonormal basis is the
Gram-Schmidt orthogonalization procedure.
Try and practice Example 5.2 yourself!
© Po-Ning [email protected]
Chapter 5-24
5.3 Conversion of the Continuous AWGN Channel
into a Vector Channel
Influence of the AWGN noise to the signal space concept
x(t ) si (t ) w(t )
where w(t) is zero-mean AWGN with PSD N0/2.
After the correlator at the receiver, we obtain:
x (t ), f j (t ) si (t ), f j (t ) w(t ), f j (t )
Or equivalently, x j sij wj .
Notably, there is no information loss
by the signal space representation.
© Po-Ning [email protected]
x1
,
x(t )
f1 ( t )
,
xN
f N (t )
Chapter 5-25
5.3 Conversion of the Continuous AWGN Channel
into a Vector Channel
Statistics of {wj}
x1 si1 w1
x N siN wN
Since {sij} is deterministic, the distribution of x is a meanshift of that of w.
Observe that w is Gaussian distributed because w(t) is
AWGN. The distribution of w can therefore be determined
by its mean vector and covariance matrix.
© Po-Ning [email protected]
Chapter 5-26
5.3 Conversion of the Continuous AWGN Channel
into a Vector Channel
Mean
E[ w j ] E 0 w(t ) f j (t )dt 0 E[ w(t )] f j (t )dt 0
Covariance
T
E[ wi w j ] E
0
T
T
w(s) f (s)ds w(t) f (t)dt
T
0
T
0
T
i
0
j
E[ w( s ) w(t )] f i ( s ) f j (t )dsdt
N0
0 0
( s t ) f i ( s) f j (t )dsdt
2
N0 T
N0
f i (t ) f j (t )dt
ij
0
2
2
T
© Po-Ning [email protected]
T
Chapter 5-27
5.3 Conversion of the Continuous AWGN Channel
into a Vector Channel
As a result, [w1, w2, …, wN] are zero-mean i.i.d. Gaussian
distributed with variance N0/2.
This shows that x is independent Gaussian distributed with
common variance N0/2 and mean vector si = [si1, si2, …, siN].
Equivalently,
N
1
1
2
f ( x | si )
exp
( x j sij )
N 0
j 1
N0
© Po-Ning [email protected]
Chapter 5-28
5.3 Conversion of the Continuous AWGN Channel
into a Vector Channel
Remainder term in noise
It is possible that
N
w' (t ) w(t ) wi f i (t ) 0
i 1
However, it can be shown that w’(t) is orthogonal to si(t)
for 1 i M. Hence, w’(t) will not affect the decision
error rate on message i.
w' (t ), si (t ) 0 with probability 1.
© Po-Ning [email protected]
Chapter 5-29
5.4 Likelihood Functions
An equivalent signal-space channel model
ˆ d ( x) {m1,...,mM }
m mi , 1 i M s c(m) x s w m
The best decision function d( ) that minimizes the decision
error is:
d ( x ) mi , if P{mi | x} P{mk | x} for all 1 k M
arg max P{m | x}
m{ m1 ,..., m M }
This is the maximum a posteriori probability (MAP)
decision rule.
© Po-Ning [email protected]
Chapter 5-30
5.4 Likelihood Functions
With equal prior probabilities,
d ( x ) arg max P{m | x}
m{ m1 ,..., m M }
arg maxP{m1 | x}, P{m2 | x},...,P{mM | x}
P{mM | x} f ( x )
P{m1 | x} f ( x ) P{m2 | x} f ( x )
arg max
,
,...,
1/ M
1/ M
1/ M
P{m1 | x} f ( x ) P{m2 | x} f ( x )
P{mM | x} f ( x )
arg max
,
,...,
P
(
m
)
P
(
m
)
P
(
m
)
1
2
M
arg max f ( x | m1 ), f ( x | m2 ),..., f ( x | mM )
f(x|mi) is named the likelihood function given that mi is transmitted
Hence, the above rule is named the maximum-likelihood decision rule.
© Po-Ning [email protected]
Chapter 5-31
5.4 Likelihood Functions
MAP rule = ML rule, if equal prior probability is assumed.
In practice, it is more convenient to work on the loglikelihood functions, defined by
d ( x ) arg max f ( x | m1 ), f ( x | m2 ),..., f ( x | mM )
arg maxlog f ( x | m1 ), log f ( x | m2 ),...,log f ( x | mM )
Why log-likelihood functions are more convenient? The
decision function becomes “sum of Euclidean distances” in
AWGN channel.
© Po-Ning [email protected]
Chapter 5-32
d ( x ) arg max log f ( x | mi ) arg max log f ( x | si )
1 i M
1 i M
N
arg max log
1 i M
j 1
1
1
2
exp
( x j sij )
N 0
N0
1
1
2
arg max logN 0
( x j sij )
1 i M
2
N0
j 1
N
N
2
arg 1min
(
x
s
)
j ij
i M
j 1
arg min || x si ||2
1 i M
Upon receipt of received signal point x, find the signal
point si that is closest in Euclidean distance to x.
© Po-Ning [email protected]
Chapter 5-33
5.5 Coherent Detection of Signals in Noise:
Maximum Likelihood Decoding
Signal constellation
Set of M signal points in the signal space
Example. Signal constellation for 2B1Q code
decision region
for s1
© Po-Ning [email protected]
decision region
for s2
decision region
for s3
decision region
for s4
Chapter 5-34
5.5 Coherent Detection of Signals in Noise:
Maximum Likelihood Decoding
Decision regions for
N = 2 and M = 4
© Po-Ning [email protected]
Chapter 5-35
5.5 Coherent Detection of Signals in Noise:
Maximum Likelihood Decoding
Usually, s1, s2, …, sM are named the message points.
The received signal point x then wanders about the
transmitted message point in a Gaussian-distributed random
fashion.
© Po-Ning [email protected]
Chapter 5-36
5.5 Coherent Detection of Signals in Noise:
Maximum Likelihood Decoding
Constant-energy signal constellation
The ML decision rule can be reduced to an innerproduct.
2
d ( x ) arg 1min
||
x
s
||
i
i M
arg min|| x ||2 2 x, si || si ||2
1i M
arg min 2 x, si Ei
1i M
arg max x, si , if Ei is constant.
1i M
© Po-Ning [email protected]
Chapter 5-37
5.6 Correlation Receiver
If signals do not have equal energy, we can use
1
d ( x ) arg max x, si Ei .
1i M
2
to implement the ML rule.
The receiver is coherent because the receiver requires to
be in perfect synchronization with the transmitter (more
specifically, the integration must begin at the right time
instance).
© Po-Ning [email protected]
Chapter 5-38
N productintegrators or
correlators
x1
x2
Correlation receiver
xN
demodulator
or detector
© Po-Ning [email protected]
decision maker
Chapter 5-39
N productintegrators or
correlators
x1
x2
xN
matched filter
© Po-Ning [email protected]
decision maker
Chapter 5-40
5.6 Equivalence of Correlation and Matched Filter
Receivers
The correlator and matched filter can be made equivalent.
Specifically,
xi 0 x(t )i (t )dt x( )hi (T )d
T
if hi (t ) i (T t ) and implicitlyi (t ) is zero outside0 t T .
© Po-Ning [email protected]
Chapter 5-41
5.7 Probability of Symbol Error
Average probability of symbol error
M
Pe 1 Pc 1 P ( mi ) P ( d ( x ) mi | mi transmitted)
i 1
1
1
M
1
1
M
1
1
M
M
P(d ( x ) m | m
i 1
M
i
i
transmitted)
2
2
P
r
||
x
s
||
min
||
x
s
||
mi transmitted
i
j
1 j M , j i
i 1
M
f ( x | s )dx
i 1 Z i
i
where Z i x N :|| x si ||2 min || x s j ||2 .
© Po-Ning [email protected]
1 j M , j i
Chapter 5-42
5.7 Invariance of Probability of Symbol Error
Probability of symbol error is invariant with respect to basis
change (i.e., rotation and translation of the signal space).
Specifically, SER (symbol error rate) only depends on the
relative Euclidean distances between the message points.
1 M
2
2
Pe 1
Pr
||
x
s
||
min
||
x
s
||
mi transmitted
i
j
1 j M , j i
M i 1
© Po-Ning [email protected]
Chapter 5-43
5.7 Invariance of Probability of Symbol Error
Specifically, if Q is a reversible transform (matrix), such as
rotation, then
x :|| x s || min || x s ||
x :|| Qx Qs || min || Qx Qs ||
N
2
i
2
1 j M , j i
N
j
2
i
2
1 j M , j i
j
If a signal constellation is rotated by an orthonormal
transformation, where Q is an orthonormal matrix, then
the probability of symbol error Pe incurred in maximum
likelihood signal detection over an AWGN channel is
completely unchanged.
© Po-Ning [email protected]
Chapter 5-44
5.7 Invariance of Probability of Symbol Error
A pair of signal constellation for illustrating the principle of
rotational invariance.
© Po-Ning [email protected]
Chapter 5-45
5.7 Invariance of Probability of Symbol Error
The invariance in SER for translation can be likewise
proved.
Is the transmission power the same for both
constellation?
© Po-Ning [email protected]
Chapter 5-46
5.7 Minimum Energy Signals
Since SER is invariant to rotation and translation, we may
rotate and translate the signal constellation to minimize the
transmission power without affecting SER.
M
E g pi || si ||2
i 1
M
Find a and Q such that Eg (a, Q) pi || Q( si a) ||2 is minimized.
i 1
But Q does not change the norm (i.e., transmission
power). Thus, we only need to determine the right a.
© Po-Ning [email protected]
Chapter 5-47
5.7 Minimum Energy Signals
Determine the optimal a.
M
E g ( a) pi || si a ||2
i 1
pi || si ||2 2aT si || a ||2
M
i 1
M
pi || si || 2a pi si || a ||2
i 1
i 1
M
2
M
T
aoptimal pi si and Eg (aoptimal ) pi || si ||2
i 1
© Po-Ning [email protected]
M
i 1
M
ps
i 1
2
i i
Chapter 5-48
5.7 Minimum Energy Signals
So subfigure (a) below has minimum average energy.
© Po-Ning [email protected]
Chapter 5-49
5.7 Union Bound on the Probability of Error
Union bound P( A B) P( A) P( B)
1
Pe 1
M
P r|| x s ||
M
2
i
i 1
min || x s j ||2 mi transmitted
1 j M , j i
|| x si ||2 || x s1 ||2
1
1
1
P
r
m
transmitt
ed
2
2
|| x s || || x s || i
i
M
M i 1 M i 1 j i
M
M
|| x si ||2 || x s1 ||2
1
P
r
m
transmitt
ed
2
2
|| x si || || x sM || i
M i 1
j i
1 M M
P r || x si ||2 || x s j ||2 mi transmitted
M i 1 j 1, j i
M
© Po-Ning [email protected]
Chapter 5-50
5.7 Union Bound on the Probability of Error
|| x s1 ||2 || x s2 ||2
2
2
|| x s1 || || x s3 ||
|| x s ||2 || x s ||2
1
4
|| x s || || x s || || x s || || x s || || x s || || x s ||
2
1
© Po-Ning [email protected]
2
2
2
1
2
3
2
1
2
4
Chapter 5-51
5.7 Union Bound on the Probability of Error
1
Pe
M
M
M
P (s , s )
i 1 j 1, j i
2
i
j
where P2 ( si , s j ) P r || x si ||2 || x s j ||2 mi transmitted .
Notably,given mi transmitted, x is Gaussian distributed with mean si .
si
sj
w
1
si s j
2
© Po-Ning [email protected]
Chapter 5-52
5.7 Union Bound on the Probability of Erroror
Hence,
For N = 1,
P2 ( si , s j ) Pr || x si ||2 || x s j ||2 mi transmited
t
1
Pr w | si s j |
2
d
ij
/2
v2
1
dv, where d ij | si s j |
exp
N 0
N0
d ij
1
erfc
2 N
2
0
© Po-Ning [email protected]
, where erfc(u) 2
u
exp( z 2 )dz.
Chapter 5-53
5.7 Union Bound on the Probability of Error
For N = 2,
P2 ( si , s j ) P r || x si ||2 || x s j ||2 mi transmitted
1
P rw1 d ij and w2 don't care, where d ij || si s j ||2
2
d
ij
/2
v2
1
dv
exp
N 0
N0
d ij
1
erfc
2 N
2
0
, where erfc(u) 2
u
exp( z 2 )dz.
The same formula is valid for any N.
© Po-Ning [email protected]
Chapter 5-54
5.7 Union Bound on the Probability of Error
Consequently, the union bound for symbol error rate is:
1
Pe
M
1
P
(
s
,
s
)
2 i j M
i 1 j 1, j i
M
M
d ij
1
erfc
2 N
i 1 j 1, j i 2
0
M
M
The above bound can be further simplified when additional
condition is given.
For example, if the signal constellation is circularly
symmetric in the sense that “{di1, di2, …, diM} is a
permutation of {dk1, dk2, …, dkM} for i k,” then
d ij
1
Pe erfc
2 N
j 1, j i 2
0
M
© Po-Ning [email protected]
Chapter 5-55
5.7 Union Bound on the Probability of Error
Another simplification of union bound
Define the minimum distance of a signal constellation
as:
d min
min
1 i M ,1 j M ,i j
d ij
Then by the strict decreasing property of erfc function,
d ij
erfc
2 N
0
1
Pe
M
d ij
1
erfc
2 N
i 1 j 1, j i 2
0
M
M
© Po-Ning [email protected]
erfc d min
2 N
0
1
M
d min
1
erfc
2 N
i 1 j 1, j i 2
0
M
M
M 1
d min
erfc
2 N
2
0
Chapter 5-56
5.7 Union Bound on the Probability of Error
We may use the bound for erfc function to realize the
relation between SER and dmin.
erfc(u)
exp(u 2 )
d min
M 1
Pe
erfc
2 N
2
0
for u 0.608131
2
M 1
d min
2
, if d min
exp
1.47929N 0 .
2
4N0
Conclusion: SER decreases exponentially as the squared
minimum distance.
© Po-Ning [email protected]
Chapter 5-57
5.7 Relation between BER and SER
The information bits are transmitted in group of log2M bits
to form an M-ary symbol.
This gives the result that a large symbol error rate (SER)
may not cause a large bit error rate (BER).
For example, a symbol error (for large M) may be due to
only 1 bit error.
Optimistically, if every symbol error is due to a single
bit error, then (assuming n symbols are transmitted)
n SER
SER
BER
.
n log2 ( M ) log2 ( M )
© Po-Ning [email protected]
SER
In general, BER
.
log2 ( M )
Chapter 5-58
5.7 Relation between BER and SER
Pessimistically, if every symbol error causes log2M bit
errors, then (assuming n symbols are transmitted)
BER
n log2 M SER
SER .
n log2 M
In general, BER SER.
Summary:
SER
BER SER
log2 M
© Po-Ning [email protected]
Chapter 5-59
5.7 Relation between BER and SER
If the statistics for “number of bit error patterns causes one
symbol error” is known, we can then determine the exact
relation between BER and SER.
M 1
BER
n SER # ( b j ) P( b j )
j 1
n log2 M
where # (b j ) number of 1's in b j ,
and b j representsone binary permutation of log2 M bit pattern.
Here, a 1’s in bj means a bit error is occurred in the corresponding position;
hence, all-zero pattern is excluded because it represents no symbol error.
© Po-Ning [email protected]
Chapter 5-60
5.7 Relation between BER and SER
Example. If all bit error patters are equally likely, then
M 1
M 1
BER
n SER # ( b j ) P ( b j )
j 1
n log2 M
SER
( M 1) log2 M
log 2 M
u 1
SER # ( b j ) (1 /( M 1))
j 1
log2 M
k
k
log2 M
u
(Note u k 2 k 1.)
u 1
u
u
log2 M 2log M 1
SER
( M 1) log2 M
2
M /2
SER
M 1
© Po-Ning [email protected]
Chapter 5-61