Diversity

Pukyong University, EE
Park, Jaehyun
 Diversity

 One of best techniques to mitigate the effects of fading

 Diversity combining of independently fading signal paths
• Time diversity
• Frequency diversity
• Space diversity

http://www.digikey.com
2
 Time diversity

 Time diversity can be obtained by interleaving and coding over

symbols across different coherent time periods.

From Fundamentals of Wireless communication by D. Tse

3
 Receiver diversity

 System model
 Multiple receive antennas
s  t  : Transmitted signal
 Linear combining
• selection or weighted sum

Co-phasing:  i  ai e  ji

 Without co-phasing or coherent

combining,
• The output exhibits significant fading
due to constructive and destructive
addition of the signals

M
Output: s  t   ai ri
i 1

Linear combiner
4
 The array gain

 In the absence of fading ri  1 Noise PDF: N 0 / 2

 SNR of each branch  i  Es / N 0 Es : Energy per symbol of s  t 

 If we set ai  1/ N 0 Array gain

2
M 
2  M
1 
  ai ri  Es  

N0
 Es
   i 1    i 1  
MEs
M M
1
N 0  ai N0 
2 N0
i 1 i 1 N0
SNR increase in the absence of fading!
 Array gain
 The increase in the average combined SNR over the average branch
SNR
   : The average combined SNR
Ag 
  : The average branch SNR

5
 The diversity gain

 With fading,
 The combining of multiple independent fading paths => more favorable
distribution for   => Decrease in SEP or outage probability (Diversity
gain)

SEP: P   Ps    p    d 
0

0
Outage Probability: Pout  p      0    p    d 
0

SEP: P  c  M ,  : The average SNR per branch

Diversity order: How the slope of the SEP as a function

of average SNR changes
6
 SEP

 Selection combining (SC), Maximal-ratio combining (MRC)

SEP for SC with i.i.d. Rayleigh fading SEP for MRC with i.i.d. Rayleigh fading

7
 Selection combining

 The combiner selects the signal on the branch with the highest SNR
 When the noise power is the same, ri 2 Es / N i
Choose the signal on the branch with the highest ri 2 Es  N i

 Cumulative distribution function of   Outage probability

M
P      P        P  max  1 ,...,  M       Pout   i   
i 1

 For i.i.d. Rayleigh fading channel

1
PDF: p   i   e  i / i => CDF: Pout   0   p   i   0   1  e  0 / i
i
Exponential distribution

 
M
P    0   p      0    Pout   i     1  e  0 /
M

i 1

8
 Selection combining (cont’d)

 Probability density function of  

M
1  e 
M 1
p       /
e  /


1  e 
 M
M M 1 1
     /
e  /
d   
0  i 1 i

 M 1
Ag   Array gain
 i 1 i


SEP: P   Ps    p    d  : No closed form solution
0

9
 1
Exponential Dist
0.9 PDF of max val. among 5 Exp RVs

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10
 Maximal Ratio Combining Weight:  i  ai e  ji

 Weighted sum of all branches with co-phasing

M
Output: s  t   ai ri
i 1
2
M 
Es   ai ri 
    i M1 
N 0  ai2
i 1

 Optimal weighting
From Cauchy-Schwartz inequality 2
M  M
2 Es   ai ri  Es  ri 2 M
M  M M

 i i   i  i
         i M1   i 1  

2 2
a r a r
 i 1  i
N 0  ai
i 1 i 1 N0
2 i 1
Equality holds when ai  cri for any constant c i 1

 Output SNR : Sum of the SNRs on each branch

11
 Maximal Ratio Combining
M
 Distribution of    i
i 1

Exponential distribution
Chi-square distribution with 2M degrees of freedom
M
Mean:      i  M 
i 1

Variance: 2M  Diversity order: M

 M 1e  /
PDF: p    
 M  M  1 !

 0 /  
k 1
0  M 1e / M
CDF: Pout   0   
 
0  M M 1 !
d   1  e  /
 0

k 1  k  1 !

12
 Supp. Chi-square distribution

If Z1 ,..., Z k are independent, standard normal random variables, then the sum of their squares,
k
Q   Z i2
i 1

is distributed according to the chi-squared distribution with k degrees of freedom.

https://ko.wikipedia.org/wiki/%EC%B9%B4%EC%9D%B4%EC%A0%9C%EA%B3%B1_%EB%B6%84%ED%8F%AC
https://en.wikipedia.org/wiki/Chi-squared_distribution

13
 Maximal Ratio Combining

 For BPSK modulation with i.i.d. Rayleigh fading and MRC

 

Pb   Q 2 p    d 
0

 See p. 215
 Diversity order: M (the number of antennas)

M
  
Ps   M  M 
 2 

14
 Equal Gain Combining Weight:  i  e  ji

 Co-phasing only
 Gain on each branch are same
2
M 
Es   ri 
    i 1  The distribution and CDF do not exist in closed form
N0 M

 For two-branch case


P      1  e 2 /   /  e  / 1  2Q  2 /  
BPSK:
  1  
2 

 

Pb   Q 2 p     d   0.5 1  1    
0   1    

15
 Transmit diversity

 Channel known at transmitter

ji
 That is, path gain re
i is known at Tx
 Quite similar to Receiver diversity

 Consider MISO (multiple-input single-output)

 i  ai e  j i

M
r  t    ai ri s  t 
s t 
i 1

M
Total energy per symbol: Es a
i 1
2
i 1

16
 Channel known at transmitter

 Optimal weight

ri
ai 

M 2
i 1
ri

 Received SNR

M M
E Es 2
  s
N0
 ri    i ,
i 1
2

i 1
i 
N0
ri : The ith branch SNR

17
 Channel unknown at transmitter

 Consider 2x1 MISO case

 Tx signal
si  t   0.5s  t  For two antenna, we divide the Tx energy equally

 Channel
ji
hi  re
i : Complex Gaussian channel (Rayleigh flat fading)
with zero mean and unit-variance
 Rx signal

r  t   0.5  h1  h2  s  t 
Complex Gaussian with zero-mean
and variance 2
Complex Gaussian with zero-mean
and variance 1

No advantage from the two antennas.

No diversity gain!
18
 Channel unknown at transmitter

 Space-time coding (Alamouti scheme)

 2x1 MISO case

 0.5s2* 0.5s1 h1

0.5s1* 0.5s2 h2

Symbol Symbol
period 2 period 1

y1  0.5h1s1  0.5h2 s2  n1
y2   0.5h1s2*  0.5h2 s1*  n2

 y1   h1 h2   s1   n1 
y   *   0.5  *   0.5H eff s  n
 y2   h2 h1*   s2   n2* 

19
Space-time coding (Alamouti scheme)

H H
h h2   h1 h2   s1   h1 h2   n1 
z  H eff
H
y  0.5  1*  *
 h2 h1*   h2* *  
h1   s2   h2 h1*   n2* 

h2   h1  h2 
H 2 2
 h1 h2   h1
 
0
   
2 2
 h* h1*   h2* h1*  
h1 h2 I2
h1  h2 
2 2
 2  0

z  H eff
H

y  0.5 h1  h2
2 2
  s1   h1
 s   n , n   h*
 2  2
h2   n1 
h1*   n2* 


zi  0.5 h1  h2
2 2
  
si  ni , ni ~ CN 0, h1  h2
2 2
N  0

20
 Space-time coding (Alamouti scheme)

 The received SNR of each stream

 i ,eff 
 h1  h2
2

2 2
Es

 h1  h2
2 2
E s

1
 1   2 

2 h1  h2
2 2
N 0
2 N0 2
Chi-square distribution
with 2M degrees of
freedom
 Diversity gain: 2
 Array gain: 1 Diversity order: M(=2)

M
Es  ri 2 M
c. f ) MRC:    i 1
 i
N0 i 1

Array gain: M
Diversity gain: M
21
Implementation example

http://www.ni.com/white-paper/13878/en/

22