Diversity

© Ammar Abu-Hudrouss
Islamic University Gaza

1
 Passband Modulation Tradeoffs

 Want high rates, high spectral efficiency, high power efficiency,

robust to channel, cheap.
 Amplitude/Phase Modulation (MPSK,MQAM)
Information encoded in amplitude/phase
More spectrally efficient than frequency modulation
Issues: differential encoding, pulse shaping, bit mapping.
 Frequency Modulation (FSK)
Information encoded in frequency
Continuous phase (CPFSK) special case of FM
Bandwidth determined by Carson’s rule (pulse shaping)
More robust to channel and amplifier nonlinearities

Wireless Communications
Slide 2
 Amplitude/Phase Modulation

 Signal over ith symbol period:

s(t )  si1g (t ) cos(2f ct  0 )  si 2 g (t ) sin( 2f ct  0 )
Pulse shape g(t) typically Nyquist
Signal constellation defined by (si1,si2) pairs
Can be differentially encoded
M values for (si1,si2)log2 M bits per symbol
 Ps depends on
Minimum distance dmin (depends on gs)
# of nearest neighbors aM
Approximate expression:

Ps  a M Q  M g s 
Wireless Communications
Slide 3
 Signal to noise ratio

 The received SNR is given by

SNR  Pr /N0 B
 In system with interference, we use signal to interference plus
noise power ratio (SINR)- (PI power of interference)

Pr
SNR 
N 0 B  PI
 SNR is often expressed in term of Es or Eb
Pr Es Eb
SNR   
N0 B N 0 BTs N 0 BTb

Wireless Communications
Slide 4
 Signal to noise ratio

 with

Eb Es
gb  gs 
N0 N0
 For raised cosine (=1), Ts = 1/B then SNR = ES /N0

 In general Ts = k/B or k.SNR = ES /N0

 For M-array signaling:

gs Ps
gb  and Pb 
log 2 M log 2 M

Wireless Communications
Slide 5
 Error Probability for PSK

 For BPSK


Pb  Q 2g b 
 For QPSK

Ps  1  1  Q
   gs 
2
or 
Ps  2Q g s / 2 
 For MPSK


Ps  2Q 2g s sin  / M  

Wireless Communications
Slide 6
 Error Probability

 For MPAM
2M  1  6g s 
Ps  Q 

M  M 2
 1 
 For MQAM


 2 M  1  3g s  
Ps  1  1  Q 
 2

 M M  1 
  
 And using the nearest neabour approximation

Ps 
 Q

2 M  1  3g s 

M M  1 
 

Wireless Communications
Slide 7
 Example 1: Find the bit error probability Pb and symbol error
probability Ps of QPSK assuming γb = 7dB. Compare the exact Pb
with the approximation Pb = Ps/2 based on the assumption of
Gray coding. Finally, compute Ps based on the nearest-neighbor
bound using γs = 2γb, and compare with the exact Ps.
 Solution:
 We have γb = 107/10 = 5.012,

  
Pb  Q 2g b  Q 10.024  7.726 *10 4 
 The exact symbol error rate


Ps  1  1  Q  gs 2
 
 1  1  Q 10.02 
2
 1.55 *10 3
 The approximate symbol error rate

Ps  2Pb  1.545 *103

Wireless Communications
Slide 8
 Example 2: Compare the probability of bit error for 8-PSK and
16-PSK assuming γb = 15 dB and using the Ps approximation
 For 8-PSK, γs = (log2 8) · 1015/10 = 94.87

 
Ps  2Q 189.74 sin  / 8  1.355.10 7

 Pb = Ps /3 = 4.52 · 10−8.

 For 16-PSK we have γs = (log2 16) · 1015/10 = 126.49

 
Ps  2Q 252.98 sin  / 16  1.916.10 3

 Pb = Ps /4 = 4.79 · 10−4.

 The error in 16-PSK is much larger (WHY?)

Wireless Communications
Slide 9
 Example 3: For 16=QAM with γb = 15 dB , compare the exact
probability of symbol error with the nearest neighbor
approximation , and with the symbol error probability for 16PSK
with the same γ b that was obtained in the previous example
 The exact symbol error rate is
2
 24  1  3 *126.49  
Ps  1  1  Q    7.37 *10 7

 4 15
  
 And the approximate

24  1  3 *126.49 
Ps  Q   3.68 *10 7

4  15 

Wireless Communications
Slide 10
 Alternate Q Function Representation

 Traditional Q function representation

 1  x2 / 2
Q( z )  p ( x  z )   e dx, x ~ N (0,1)
z
2
Infinite integrand
Argument in integral limits
 New representation (Craig’93)

1  /2


 z 2 /(sin 2  )
Q( z)  e d
0

Leads to closed form solution for Ps in PSK

Very useful in fading and diversity analysis
Wireless Communications
Slide 11
 Probability of error in Fading

Ts
Ps   Ps (g s ) p(g s )dg s
Ps
Ps
t or d
Average probability :
 Expected value of random variable Ps
 Used when Tc~Ts
 Error probability much higher than in AWGN alone

Wireless Communications
Slide 12
 Outage Probability

Ts
Ps Outage
Ps(target)

t or d

 Probability that Ps is above target

 Equivalently, probability gs below target
 Used when Tc >>Ts

Wireless Communications
Slide 13
 Combined outage and average Ps

Outage
Ps(gs)
Ps(gs)
Pstarget

Ps(gs)
 Used in combined shadowing and flat-fading
 Ps varies slowly, locally determined by flat fading
 Declare outage when Ps above target value

Wireless Communications
Slide 14
 Doppler Effects

 High Doppler causes channel phase to de-correlate between

symbols

 Leads to an irreducible error floor for differential modulation

Increasing power does not reduce error

 Error floor depends on BdTs

Wireless Communications
Slide 15
 ISI Effects

 Delay spread exceeding a symbol time causes ISI (self

interference).
1 2 3 4 5

0 Ts Tm

 ISI leads to irreducible error floor

Increasing signal power increases ISI power
 ISI requires that Ts>>Tm (Rs<<Bc)

Wireless Communications
Slide 16
 Introduction to Diversity

 Basic Idea
Send same bits over independent fading paths
 Independent fading paths obtained by time, space,
frequency, or polarization diversity
Combine paths to mitigate fading effects

Tb

Multiple paths unlikely to fade simultaneously t

Wireless Communications
Slide 17
 Diversity Types

microdiversity mitigates the effect of multipath fading are

called, and that is the focus of this lecture.

macrodiversity mitigates the effects of shadowing from

buildings.

Wireless Communications
Slide 18
 Diversity Techniques

Micro-diversity
• Space diversity (Transmit and receive)
• Time Diversity
• Frequency Diversity
• Polarization
• Angle

Wireless Communications
Slide 19
 Diversity Techniques

• Macro-diversity

Wireless Communications
Slide 20
 Receiver Combining Techniques

 Selection Combining
Fading path with highest gain used

 Threshold selection Combining.

 Maximal Ratio Combining

All paths co-phased and summed with optimal
weighting to maximize combiner output SNR.

 Equal Gain Combining

All paths co-phased and summed with equal
weighting
Wireless Communications
Slide 21
 System model

Wireless Communications
Slide 22
 Array gain

 Let us assume that there is no fading

 The received symbol at each branch equal to ri  Es

 Where Es is the symbol energy of the transmitted signal.

 Assume AWGN and pulse shaping BTs = 1, then snr at each

branch is γ0 = Es /N0.

 For maximum combining: ai  ri / N 0

 Then the received SNR is

g 

M
i 1
ai ri  2


 E 
M
i 1 s
2


MEs
N 0 i 1 ai2 N 0 i 1 Es
M M
N0

Wireless Communications
Slide 23
 Thus there is an M fold increase in the snr.

 The array gain, Ag is defined as the increase in averaged

combined snr over the average branch snr
g
Ag 
g
 In fading combining the multiple fading path leads to a more
favorable distribution for combined snr. The metric that used to
measure the diversity gain are the average error probability and
outage probability

Ps   Ps (g ) pg  (g )dg

Pout  p g   g 0    Pg  g dg
g0

0
Wireless Communications
Slide 24
 Diversity Gain

 Diversity gain is change in BER slope (fading)

 In fading combining the multiple fading path leads to a more

favorable distribution for combined snr. The metric that used to
measure the diversity gain are the average error probability and
outage probability

Ps   Ps (g ) pg  (g )dg

Pout  p g   g 0    Pg  g dg
g0

Wireless Communications
Slide 25
 Selection Combining (SC)
Combiner SNR is the maximum of the branch
SNRs.
For M branch diversity the CDF of snr is given by.
M
Pg  g    pg i  g 
i 1

For Rayleigh fading channel, the PDF of the snr is

given by
1
pg i   e g i / g i
gi
And for outage probability target of γ0 .

Pout g 0   1  eg 0 / g i 
Wireless Communications
Slide 26
 Selection Combining (SC)
If the average snr for all branches are assumed to
be the same.


Pout g 0   1  e g 0 / g M

By differentiating the CDF,

pg  g  
M
g
1  e  g / g M 1 g / g
e
The average snr of the combiner
M
 1
g   gpg  g dg  g 
i 1 i
0

The average SNR increase by M but not linearly

Wireless Communications
Slide 27
 For DPSK, the average probability of symbol error is given by

 M  1

 
M  m 
M 1
 M
Pb   0.5e pg  g dg 
g
  1
0 2 m0 1 m  g

Wireless Communications
Slide 28
 Example: Find the outage probability of BPSK modulation at Pb
= 10 −3 for a Rayleigh fading channel with SC diversity for M = 1
(no diversity),M = 2, and M = 3. Assume equal average branch
SNRs of γ = 15 dB.
 For Pb = 10 −3 , γ0 = 7 dB= 10.7 and 𝛾ҧ = 1015/10 .

0.1466 M 1

Pout g 0   1  e 
g 0 / g M 
 0.0215 M 2
0.003 M 3


Wireless Communications
Slide 29
Outage Probability of Selection Combining in Rayleigh Fading
Wireless Communications
Slide 30
  
Pb   Q 2g pg  g dg

Wireless Communications
Slide 31
 Threshold Combining
 SC requires continuous mentoring for each antenna branch

 SSC avoids this by clinging to the first antenna with γ >γt.

 As long as the system does not change the branch. If its γ falls
below γt . The sequential searching starts until the system
connect to the first antenna with γ >γt .

This method is called

switch and stay
combining (ssc)

Wireless Communications
Slide 32
 If we consider two-antenna scheme, the CDF is found to be

 Pg 1 g T Pg 2 g  g  gT
Pg  g   
 pg T  g 1  g   Pg 1 g T Pg 2 g  g  g T

 For Rayleigh fading with equal average snr

g T / g  g T g  / g

1  e  e g / g
 e g  gT
Pg  g    g 0 / g g T g  / g

1  2 e  e g  gT
 The outage probability is then found to be
g T / g g 0 / g  g T g 0  / g

1  e  e  e g  gT
Pout g 0   Pg  g 0    g 0 / g g T g 0  / g

1  2 e  e g  gT

Wireless Communications
Slide 33
 Example: Find the outage probability of BPSK modulation at Pb =
10−3 for two-branch SSC diversity with i.i.d. Rayleigh fading on
each branch for threshold values of γT = 5, 7, and 10 dB. Assume
the average branch SNR is γ = 15 dB. Discuss how the outage
probability changes with γT . Also compare outage probability
under SSC with that of SC and no diversity from Example 7.1.

 For γavg = 15 dB, γ0 = 7 dB and γT = 5

Pout  1  2e 10.7 / 101.5

e  
 10.5 101.5 / 101.5  0.0654

 For γavg = 15 dB and γT = 7

Pout  1  2e 10.7 / 101.5
e  
 10.7 101.5 / 101.5  0.0215

 For γavg = 15 dB and γT = 10

Pout  1 e 10 / 101.5

e 10 7 / 101.5
e  
 10.7 10 / 101.5  0.0397

Wireless Communications
Slide 34
 Error probability for DPSK

 For Rayleigh distribution iid snr, the CDF can be differentiated

to get the PDF

 1  
e g T / g
g e
1 g / g
g  gT

Pg  g   

 2  e g T / g

g e
1 g / g
g  gT

 And the average bit error rate for QPSK

Pb  
0
g
0.5e pg  g dg 
g 1
21  g 

1  e g T / g  e g T e g T / g 

Wireless Communications
Slide 35
 Example: Find the average probability of error for DPSK
modulation under two-branch SSC diversity with i.i.d. Rayleigh
fading on each branch for threshold values of γT = 5, 7, and 10
dB. Assume the average branch SNR is γ = 15 dB. Discuss how
the average probability of error changes with γT . Also compare
average error probability under SSC with that of SC and with no
diversity.
 Solution
 For scc with γavg = 15 dB and γT = 5, 7, and 10 dB yields,
respectively, Pb = .0029, Pb =.0023, and Pb = .0042.

 For SC with M = 2

Pb  0.5 1  10 
1.5 1

 0.5 2  10 
1.5 1
 4.56.10  4
 For SC with M =1 (no diversity)


Pb  0.5 1  10 
1.5 1
 .0153
Wireless Communications
Slide 36
 Maximal Ratio Combining (MRC)

 Optimal technique (maximizes output SNR)

 Combiner SNR is the sum of the branch SNRs.

 Distribution of SNR hard to obtain.

 Can use MGF approach for simplified analysis.

 Exhibits 10- 40 dB gains in Rayleigh fading.

Wireless Communications
Slide 37
 The output snr of the MRC combiner

g 
1  M
ar
i 1 i i

2


M
N0 a 2
i 1 i

 For maximum combining ai  ri / N 0

 The resultant snr is

1
 r  i 1 g i
M 2 M
g  i 1 i
N0
 Using Moment generating function of sum of independent
variable . Assuming Rayleigh fading with equal average branch
γavg:

 The distribution of the sum is chi squared with 2M degree of

freedom expected value = M γavg and variance of 2M γavg

Wireless Communications
Slide 38
 So the pdf is expressed as

g M 1e g / g
pg g   M , g 0

g M  1!

 The outage probability error rate for snr0

M
g 0 / g 
k 1
Pout  1  e g 0 / g

i 1 k  1!
 For BPSK, the bit error rate can be given by

 
m
 1     M  1  m  1   
M 1
Pb   Q 2g pg  g dg       
 2  m 0  m  2 
 Where   g / g  1

Wireless Communications
Slide 39
 Equal Gain Combining

 EGQ simpler than MRC

Harder to analyze

Performance about 1 dB worse than MRC

The output snr is given by

g 
1
N0 M
 r  M
i 1 i
2

The CDF for two branches is derived as

g g / g   2g  
Pg  g   1  e  2g / g
e 1  2Q 

g  g
  

Wireless Communications
Slide 40
 The resulting outage probability

g 0 g / g   2g 0 
Pout g 0   1  e  2g 0 / g
e 1  2Q 

g 
  g 
 By differentiating the CDF we get the pdf

1  1 1 g   2g  
pg  g   e  2g / g
 e g / g   1  2Q 
 g 
g  4gg g g 
   
 And finally the probability error for BPSK is
 2 
    1
Pb   Q 2g pg  g dg  0.51  1  
 1 g

 

   

Wireless Communications
Slide 41
 Example: Compare the average probability of bit error of BPSK
under MRC and EGC two-branch diversity with i.i.d. Rayleigh
fading with average SNR of 10 dB on each branch.

 For MRC
2
 1  10 / 11 
Pb  
2
 
 2  10 / 11  1.6.10 3

 
 For EGC

 1 
2 

Pb  0.5 1  1     2.07.10 3
  11  
 

Wireless Communications
Slide 42
 Transmit Diversity

 Channel Known at transmitter

 It is similar to receive diversity:
M
r t    ai ri st 
i 1

 ri is the channel gain and ai is the gain at transmitter which can

optimized to achieve maximum snr by
ri
ai 
 j 1 j
M2
r

Wireless Communications
Slide 43
 And the resulting snr is given by,
M M
Es
g 
N0
 i  g i
r 2

i 1 i 1
 If we assume that the channel gains are identical
Es
g  Mr 2
N0
 There is M fold increase in snr over single antenna.

Wireless Communications
Slide 44