Cooperative communications
Direct transmission
Cooperative communications
User cooperation
relay
RSC
1
systematic
parity 1
information
bits
RSC
2
parity 2
1
H
RSC
1
systematic
parity 1
information
bits
RSC
2
parity 2
1
H
RSC
decoder
User (S)
Partner (R)
Cooperative turbo
coding structure
conventional
turbo coding
structure
Diversity vs. Multiplexing Tradeoff
U
1
1
T
D
R
1
U
T
R
T
(a)
Encoder Mapping
H
Encoder Mapping
decoded
information
from U
t
U
t R
t
(b)
(a) System model
(b) Signal processing
for a pair of the
user t and the
relay t.
Disadvantage: low bandwidth efficiency
Distributed Turbo Coding
Consider a cooperative relaying system with T singleantenna sources, 1 Tantenna
destination and 1 Tantenna relay.
S1
ST
1
T
1
T
R
1
T
D
System model.
Timeslot 1 T T+1
Terminal S
1
S
T
R
Channel allocation for T sources and a relay in the proposed cooperative relaying schemes.
The cooperative relaying operates as follows:
The bandwidth efficiency of the proposed schemes is reduced only by T/(T+1)
(instead of as in the schemes in [1][9]).
Distributed Turbo Coding
Encoder Mapping
transmitted signal
s
t
information bits
I
t
s
t
= (s
t
[1], , s
t
[l], , s
t
[L
t
]) = {u{I
t
}}
{.}: modulation
u{.}: encoding
Signal Processing at Sources
Distributed Turbo Coding
Signal Processing at the relay: decoding
Distributed Turbo Coding
received signal
r
t
[l]
MRC Decoder Demapping
r'
t
[l]
I'
t
 
( )
'
, , t l p t
b r l A
bit metrics calculation
r
t
[l] = a
t
s
t
[l] + n
t
[l]
   
'
[ ]
[ ] = = +
H
' ' t t
t t t t
H
t t
l
r l a s l n l
a r
a a
 
( )
   
   
1,
0,
2
' '
0
'
, ,
2
' '
0
exp
log
exp
x p
x p
t t x
s
t l p t
t t x
s
r l a l s
N
b r l
r l a l s
N
_
_
A
e
e
 
 




\ .

=

 




\ .
\ .
Signal Processing at the relay: retransmission
Distributed Turbo Coding
Parallel Transmission (PT)
H
I'
1
Encoder Mapping
1
H
I'
T
Encoder Mapping
T
x
1
x
T
Simple for the singleuser iterative decoding (SUID)
Signal Processing at the relay: retransmission
Distributed Turbo Coding
Multiplexing Transmission (MT)
Encoder
Mapping
Mapping
S/P
M
U
X
1
T
H
I'
1
H
I'
T
x
1
x
T
Multiplexing increases the volumes (in bits) but also makes longer
parity segments, and hence stronger codes.
Facilitate the highperformance multiuser joint iterative decoding
(MUJID).
Signal Processing at the Destination
Distributed Turbo Coding
For parallel transmission (PT)
Bit metrics
calculation
from S
t
SISO
H
( )
, , t l p
b A
( )
,
L
j
t e
SISO
Bit metrics
calculation
from the relay
( )
, , , r t l p
b A
( ) 1
,
L
j
t a
( ) '
,
L
j
t e
1
H
Iterative decoding the signal of the source S
t
(SUID)
Signal Processing at the Destination
Distributed Turbo Coding
For multiplexing transmission (MT)
Bit metrics
calculation
from S
1
( )
1, , l p
b A
SISO
H
MUX
Bit metrics
calculation
from S
T
SISO
H
( )
, , T l p
b A
SISO
Bit metrics
calculation
( )
, , , r t l p
b A
from the
relay
DeMUX
1
H
1
H
( )
L
j
e
( 1)
L
j
a
( ) 1
1,
L
j
a
( ) 1
,
L
j
T a
( )
1,
L
j
e
( )
,
L
j
T e
( ) '
1,
L
j
e
( ) '
,
L
j
T e
P/S
Iterative decoding for all sources (MUJID)
Bit metrics calculation
A(b
t
,
l,p
) at D is similar to that
at R.
How to compute bit metrics
A(b
r,t,l,p
)?
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
Relaydestination channel:
y[l] = Hx[l] + n[l], l=1, , L
 
( )
 
 
1
0
2
,
2
exp
log
exp
, ,
, ,
x
x
y Hx
y
y Hx
t p
t p
0
r t,l, p
0
l
N
b l
l
N
_
_
A
e
e
 
 




\ .

=

 
 
 
\ .
\ .
Need to sum over 2
mT1
possible symbol vectors in the set _
1,t,p
.
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
ZFbased Bit Metrics Calculation (ZFC)
z[l] = Wy[l] = x[l] + q[l]
W=(H
H
H)
1
H
H
z
t
[l] = x
t
[l] + q
t
[l]
 
( )
 
 
 
 
1,
0,
2
,
2
exp
log
exp
x p
x p
t x
s
t
r t,l, p t
t x
s
t
z l s
l
b z l
z l s
l
_
_
o
A
o
e
e
 
 




\ .

=
  
 
 
\ .
\ .
Need to sum over 2
m1
possible symbols in the set _
1,p
.
o
t
[l]=W(t, :)W(t, :)
H
N
0
: noise enhancement.
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
QRbased Bit Metrics Calculation (QRC)
Using QR decomposition: H=QR
k[l] = Q
H
y[l] = Rx[l] + v[l]
v[l] = Q
H
n[l] has the same probability distribution of n[l] since Q is a
unitary matrix: no noise enhancement.
Explicitly
k
T
[l] = r
T,T
x
T
[l] + v
T
[l]
       
, ,
1
, = 1,...,1
T
t t t t t j j t
j t
k l r x l r x l l t T v
= +
= + +
the bit metrics calculation in accompany with the soft interference
cancellation.
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
 
( )
   
   
1,
0,
2
,
0
, , ,
2
,
0
exp
log
exp
x p
x p
T T T x
s
r T l p T
T T T x
s
k l r l s
N
b k l
k l r l s
N
_
_
A
e
e
 
 




\ .

=

 




\ .
\ .
LLRs for the coded bits transmitted on the antenna T of the relay
Need to sum over 2
m1
possible symbols in the
set _
1, p
.
A(b
r,T,l,p
k
t
[l])s are used to compute the soft symbols, m
T
[l]s, corresponding
to x
T
[l]s for the transmit antenna T and the variances,
T
[l]s, of these soft
symbols
    { }   { }
1
E Pr
M
T T c T c
c
m l x l x x l x
=
= = =
   
{ }
    { }
2 2
1
E Pr
M
T c T c T T c
c
l x m l x m l x l x
=
= = =
  { } { }
, , ,
1
Pr Pr
m
T c r T l p
p
x l x b
=
= =
[
{ }
 
( ) { }
, , ,
, , ,
, , ,
1
Pr
1 exp ( 1)
r T l p
r T l p
b
r T l p T
b
b k l A
=
+
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
Calculate the LLRs for the coded bits on the remaining transmit antennas in
the order t=T1,,1 in two steps.
Step 1: symbol interference cancellation
Need to sum over 2
m1
possible symbols in the
set _
1, p
.
           
( )
 
 
, , ,
1 1

t
T T
t t t j j t t t t j j j t
j t j t
v l
k l k l r m l r x l r x l m l v l
= + = +
'
'
= = + +
Assume v'
t
[l] e CN(0,o'
t
[l]) with    
2
, 0
1
T
t t j j
j t
l r l N o
= +
'
= +
Step 2: compute the LLRs for the coded bits
 
( )
   
 
   
 
1,
0,
2
,
, , ,
2
,
exp
log
exp
x p
x p
t t t x
s
t
r t l p t
t t t x
s
t
k l r l s
l
b k l
k l r l s
l
_
_
o
A
o
e
e
 
 
'


 ' 
\ .

'
=

 
'


 ' 
\ .
\ .
Bit Metrics Calculation in the Timeslot (T+1)
Distributed Turbo Coding
Calculate the LLRs for the coded bits on the remaining transmit antennas in
the order t=T1,,1 in two steps.
Step 1: symbol interference cancellation
Need to sum over 2
m1
possible symbols in the
set _
1, p
.
           
( )
 
 
, , ,
1 1

t
T T
t t t j j t t t t j j j t
j t j t
v l
k l k l r m l r x l r x l m l v l
= + = +
'
'
= = + +
Assume v'
t
[l] e CN(0,o'
t
[l]) with    
2
, 0
1
T
t t j j
j t
l r l N o
= +
'
= +
Step 2: compute the LLRs for the coded bits
 
( )
   
 
   
 
1,
0,
2
,
, , ,
2
,
exp
log
exp
x p
x p
t t t x
s
t
r t l p t
t t t x
s
t
k l r l s
l
b k l
k l r l s
l
_
_
o
A
o
e
e
 
 
'


 ' 
\ .

'
=

 
'


 ' 
\ .
\ .
Cooperative relaying schemes
Scheme Description
PT_ZF PT, ZFC, SUID
PT_QR PT, QRC, SUID
PT_ML PT, MLC, SUID
MT_ZF MT, ZFC, MUJID
MT_QR MT, QRC, MUJID
MT_ML MT, MLC, MUJID
180bit length, CRC16CCITT, 5 iterations, T=3.
Distributed Turbo Coding
Distributed Turbo Coding
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
10
0
SNR (dB)
B
E
R
DT
Reference [9]
PT_ZF
PT_QRSource 1
PT_QRSource 2
PT_QRSource 3
PT_ML
MT_ZF
MT_QRSource 1
MT_QRSource 2
MT_QRSource 3
MT_ML
BER versus SNR for SNR
rd
=SNR+5dB, SNR
in
=SNR+10dB
Distributed Turbo Coding
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
SNR (dB)
B
E
R
DT
Reference [9]
PT_ZF
PT_QRSource 1
PT_QRSource 2
PT_QRSource 3
PT_ML
MT_ZF
MT_QRSource 1
MT_QRSource 2
MT_QRSource 3
MT_ML
BER versus SNR for SNR
rd
=SNR+5dB, SNR
in
=SNR+20dB
Distributed Turbo Coding
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
10
0
SNR (dB)
B
E
R
Reference [9]SNR
in
=SNR+10dB
Reference [9]SNR
in
=SNR+20dB
PT_ZFSNR
in
=SNR+10dB
PT_ZFSNR
in
=SNR+20dB
MT_ZFSNR
in
=SNR+10dB
MT_ZFSNR
in
=SNR+20dB
PT_MLSNR
in
=SNR+10dB
PT_MLSNR
in
=SNR+20dB
MT_MLSNR
in
=SNR+10dB
MT_MLSNR
in
=SNR+20dB
BER with different interuser channel qualities, SNR
rd
=SNR+5dB
Distributed Turbo Coding
30
BER versus SNR for SNR
rd
=SNR+15dB, SNR
in
=SNR+10dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
10
0
SNR (dB)
B
E
R
DT
Reference [9]
PT_ZF
PT_QRSource 1
PT_QRSource 2
PT_QRSource 3
PT_ML
MT_ZF
MT_QRSource 1
MT_QRSource 2
MT_QRSource 3
MT_ML
Distributed Turbo Coding
31
BER with different relaydestination channel qualities, SNR
in
=SNR+10dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
6
10
5
10
4
10
3
10
2
10
1
10
0
SNR (dB)
B
E
R
PT_ZFSNR
rd
=SNR+5dB
PT_ZFSNR
rd
=SNR+15dB
MT_ZFSNR
rd
=SNR+5dB
MT_ZFSNR
rd
=SNR+15dB
PT_MLSNR
rd
=SNR+5dB
PT_MLSNR
rd
=SNR+15dB
MT_MLSNR
rd
=SNR+5dB
MT_MLSNR
rd
=SNR+15dB
Distributed Turbo Coding
32
BER versus SNR for SNR
rd
=SNR+15dB, SNR
in
=SNR+20dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
SNR (dB)
B
E
R
DT
Reference [9]
PT_ZF
PT_QRSource 1
PT_QRSource 2
PT_QRSource 3
PT_ML
MT_ZF
MT_QRSource 1
MT_QRSource 2
MT_QRSource 3
MT_ML
Distributed Turbo Coding
BER for different iterations : PT_ZF
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
3
10
2
10
1
SNR (dB)
B
E
R
PT_ZF
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Distributed Turbo Coding
BER for different iterations : PT_QR
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
3
10
2
10
1
SNR (dB)
B
E
R
PT_QR
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Distributed Turbo Coding
BER for different iterations : PT_ML
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
4
10
3
10
2
10
1
SNR (dB)
B
E
R
PT_ML
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Distributed Turbo Coding
BER for different iterations : MT_ZF
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
3
10
2
10
1
SNR (dB)
B
E
R
MT_ZF
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Distributed Turbo Coding
BER for different iterations : MT_QR
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
4
10
3
10
2
10
1
SNR (dB)
B
E
R
MT_QR
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Distributed Turbo Coding
BER for different iterations : MT_ML
SNR
in
=SNR+10dB and SNR
rd
=SNR+5dB
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10
5
10
4
10
3
10
2
10
1
SNR (dB)
B
E
R
MT_ML
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5