emad alsusa & christos masouros dept. of electrical & electronic engineering university of...
Post on 21-Dec-2015
216 views
TRANSCRIPT
Emad Alsusa & Christos Masouros
Dept. of Electrical & Electronic Engineering
University of Manchester
Adaptive Code Allocation for Interference Exploitation on the Downlink of MC-CDMA
Systems
Principles of the Proposed Method
For PSK Modulation, Interference can be Separated to Constructive and Destructive
Interference Depends on Users Crosscorrelations as well as the instantaneous Data
By Reallocating the Codes According to the Current Data, the Crosscorrelations and hence Interference Amongst Users can be Manipulated
By Exploiting Constructive Interference the Effective SINR can be Increased and Performance can be Improved Without the Need to Increase Transmitted per-User Power
MC-CDMA Downlink Employing post-Equalization (K users)
Received Signal at the u-th Mobile Unit (MU) at the i-th symbol period:
Decision Variable:
1,
K
H
iu iu u uu iu ku ik iuk k u
x x nd C
r1,
( ) Hiu iu u k
K
u u uu iu iuku ikk k u
a x a x n
d r CE
1, ( ) H
iu iu u k
K
u u uu iu iuku ikk k u
a x a x n
d r CE
= [ · · ] + iu i u ur x A C H N
=( )· Hpq p q q q C H E C
Constructive - Destructive Interference Separation
User-to-User Constructive MAI:
0iu k ik kux a x
Cumulative Constructive MAI:
1,
0K
iu k ik kuk k u
ax x
1
-1
-1
1
c1 axis
c2 axis
x2 . c2
x1 . c1
r = x1c1 + x2c2
c1 axis
c2 axis
1
1
-1
-1
x2 . c2
x1 . c1
r = x1c1 + x2c2
d1
d2 1
1
-1
-1
c1 axis
c2 axis
x2 . c2
x1 . c1
r = x1c1 + x2c2
d2
d1
d2
d1
a b c
Constructive - Destructive Interference Separation
Instantaneous per Symbol Effective SINR:
constructivee
destructive
S MAISINR
MAI N
Im
ReRe
Im
1-1
BPSK QPSK 8PSK
1+i
1-i
-1+i
-1-i
a b c
Decision Variables Distributions for pc=8 Different Allocation Patterns for K=5, L=16
x1 x2 x3 x4 x5
1 -1 1 1 -1
allocation pattern No. d1 d2 d3 d4 d5
s=1 0.75 -1 0.5 0.75 -1
s=2 0.5 -1.75 -0.5 1.25 -2
s=3 2.5 -2.75 2.25 3 -0.5
s=4 0.5 -1 0.75 0.75 -1
s=5 0.75 -0.5 -0.25 0 -1
s=6 2.25 -0.5 3.25 2.5 -2.5
s=7 -0.25 -0.5 0.75 0 -1
s=8 1.25 -2 0.5 -0.5 -1.75
b
x 1 x 2 x 3 x 4 x 5
- 1 - 1 1 - 1 -1 allocation
pattern No. d 1 d 2 d 3 d 4 d 5
s =1 - 0.25 0 0.5 - 0.25 - 1
s =2 - 1 - 1.25 1 - 0.25 - 0.5
s =3 - 0.5 0.25 0.25 0 - 0.5
s =4 - 3 - 1. 7 5 3.25 - 2.75 - 2. 2 5
s =5 0.25 0 0.25 - 0.5 - 0.5
s =6 - 0.25 - 1 0.5 0 - 0.25
s =7 - 2.75 - 3 3.25 - 2.5 - 1.5
s =8 0.75 - 1.5 1 - 1 - 1.25
a
Code-to-User Allocation (CUA) Technique (1/9)
1. Create Code Sets 2. Evaluate Code
Sets3. Select Optimum
Code Set4. Spread and
Transmit5. Transmit SI6. Detect SI and
select the correct code
7. Dispread and Detect
Nc codes
Random Shuffling
Code-to-user allocation pattern 1
Code-to-user allocation pattern 2
Code-to-user allocation pattern pc
ˆsR
ˆ1R
pre-estimation
ˆi,sd
Pre-estimation processing
Pre-estimation processing
ˆi,1d
Pre-estimation processing
ˆi,2d
ˆci,pd
xi
A
Channel Estimation
allocation pattern selection
copt
BS Transmitter Processing
u-th channel
Nu
SUD / MUD SI
detection
u-th MU receiverriu
SI transmission
ˆoptc
estimation
arg max min ˆs
i,sd
iud
OFDM
ˆuH
IOFDMPost-
Equalization
Code-to-User Allocation (CUA) Technique (2/9)
Nc codes
Random Shuffling
Code-to-user allocation pattern 1
Code-to-user allocation pattern 2
Code-to-user allocation pattern pc
ˆsR
ˆ1R
pre-estimation
ˆi,sd
Pre-estimation processing
Pre-estimation processing
ˆi,1d
Pre-estimation processing
ˆi,2d
ˆci,pd
xi
A
Channel Estimation
allocation pattern selection
copt
BS Transmitter Processing
estimation
arg max min ˆs
i,sd
ˆuH
OFDM
SI DATA
Code-to-User Allocation (CUA) Technique (3/9)
Decision Variables pre-Estimation
Code Allocation Selection Criteria
ˆarg max minjs
i,sd̂arg max min
js
i,sd
ˆarg max minjs
i,sd
,1, 2,ˆ ˆ ˆˆ ˆ
iK si s i sd d d i si,s i,s i,sd x AR d e ,1, 2,ˆ ˆ ˆˆ ˆ
iK si s i sd d d i si,s i,s i,sd x AR d e
,1, 2,ˆ ˆ ˆˆ ˆ
iK si s i sd d d i si,s i,s i,sd x AR d e
Code-to-User Allocation (CUA) Technique (4/9)
For Correct Dispreading According to the Updated Codes, Transmission of
SI bits is NecessarySI is Common for all Users
If Code Allocation s=7 is [3, 5, 2, 1, 4] then User k=3 Should Employ Code with Index 2 from the reference set for Correct Dispreading
Control Symbols (SI)
Data bits – Symbol 1
Data bits – Symbol 2
Data bits – Symbol N
FRAMELENGTH
2log ( )s cpN 2log ( )s cpN 2log ( )s cpN
2log ( )s cpN
Code-to-User Allocation (CUA) Technique (5/9)
SUD / MUD SI
detection
u-th MU receiverriu
ˆoptc
iud
Post-Equalization
SI
DATA
• Enhanced Received SINR, Improved Reliability
• Data Detection Very Sensitive to SI Errors
Code-to-User Allocation (CUA) Technique (6/9)
CUA with MRC, EGC, SU- MMSE post-Equalization
•Performance Improvement of an Order of Magnitude Without Increase in Transmitted per-User Power
•Efficiency Reduction to 91% due to Transmission of Side Information (SI)
Number of paths=4, K=20, L=32, pc=16
0 2 4 6 8 10 12 14 16 18
10-4
10-3
10-2
10-1
100
Transmitted Eb/No per user (dB)
BE
R
MRCEGC
MMSE
MRC - CUA
EGC - CUAMMSE - CUA
Code-to-User Allocation (CUA) Technique (7/9)
CUA with EGC post-Equalization and SIC Detection
Number of paths=4, K=20, L=32
0 2 4 6 8 10 12 14 16 18
10-4
10-3
10-2
10-1
100
Transmitted Eb/No per user (dB)
BE
R
EGCEGC - CUA Nc=K
EGC - CUA Nc=32
EGC - SIC
EGC - SIC - CUAEGC - SIC - CUAperfect SI detection
• Limited Improvement for Increased NC
• Performance Loss for Low SNR due to Unreliable SI
Code-to-User Allocation (CUA) Technique (8/9)CUA with EGC
•Limited Improvement for Increased Number of Available Allocation Patterns (pc)
Number of paths=3, K=16, L=16, SNR=7dB
5 10 15 20 25 3010
-4
10-3
10-2
10-1
100
pc
BE
R
MRC
MRC SICEGC
EGC SIC
Code-to-User Allocation (CUA) Technique (9/9)
CUA with pre-decorrelation employing MRC Equalization
•Significant Performance Improvement
•Transmission Efficiency of 32/34=94.2%
Number of paths=3, K=32, L=32, SNR=7dB
0 2 4 6 8 10 12 14 16 18 2010
-4
10-3
10-2
10-1
100
Transmitted Eb/N
0 per user
BE
R
TP-MRCTP-MRC-CUAJT-MRCJT-MRC-CUA
Conclusions In Conventional Systems Energy Inherent in the System is
Wasted due to Data-Code MisalignmentPart of the Existent Interference can be Exploited to
Enhance the Received SINRBy Optimizing the Code Allocation Amongst the Users
with CUA the Constructive Component of Interference can be Maximized Improved Received SINR without Transmitted per-User Energy
IncreaseApplication of CUA can Enhance the Performance of a Number of
Conventional MultiUser Precoding and Detection SchemesThe Dependency on SI Detection Limits the CUA Overall
Performance for Low Transmitted SNRFor High SNR Values Performance Improvement of an Order of
Magnitude is Attained
Thank you
QuestionsCommentsSuggestions