Toward MIMO MC-CDMA
Toward MIMO MC-CDMA
Speaker : Pei-Yun Tsai
Advisor : Tzi-Dar Chiueh
Outline
Outline
• Motivation
• MIMO MC-CDMA transmitter
– Allocation of system resource
– STBC+MC-CDMA
– V-BLAST+MC-CDMA
• MIMO MC-CDMA receiver
– Synchronization
– Channel estimation
– MIMO decoding
Motivation
Motivation
• MC-CDMA
– A promising solution for future wireless cellular
communication systems.
• MIMO
– One technique to improve capacity and diversity gain.
• MIMO MC-CDMA
Considerations in Transmitter
Requirement for Channel Estimation
Requirement for Channel Estimation
(1/2)
(1/2)
• For some subcarrier k,
– Tx antenna 0, S
0[n]
– Tx antenna 1, S
1[n]
• At the Rx,
Antenna 0 Antenna 1S
0[0]
S
0[1]
S
1[1]
S
1[0]
Antenna 0 Antenna 1R
0[0
]R
0[1
]
R
1[1
]
R
1[0
]
H
00H
11H
01H
10
]
1
[
]
0
[
]
1
[
]
1
[
]
0
[
]
0
[
]
1
[
]
0
[
0 0 01 00 1 0 1 0 0 0V
V
H
H
S
S
S
S
R
R
V
SH
R
V
S
H
R
S
H
ˆ
1
1Requirement for Channel Estimation
Requirement for Channel Estimation
(2/2)
(2/2)
• MMSE Channel estimation [3]
subject to
)}
(
)
{(
1
}
||
ˆ
{||
1
2
1 1
H
H
VS
HVS
T TE
M
E
M
MSE
2 1 2}
]
{[
H TTr
M
SS
T M iS
in
M
T
1 2|
]
[
|
S, known training symbols or pilot subcarriers, should
System Resources
System Resources
• Training symbol and pilot subcarriers
24 800 sub-carriers tim e training symbol pilot subcarrier 24 800 sub-carriers 9 18
Pattern
Pattern
• Training symbol
• Pilot subcarriers
Antenna 0 Antenna 0 Antenna 1 Antenna 1-1-j
1+j
-1-j
1+j
differentially encoded by one PN sequence
Problem
Problem
• Channel estimation in mobile MIMO systems
– Time-invariant requirement of channel response.
– Impossible in fast-fading channel.
User u
STBC + MC-CDMA
STBC + MC-CDMA
Pilot Insertion OFDM Modulation Training Symbol Insertion Constellation Mapping STBC Spreading Spreading Pilot Insertion OFDM Modulation Training Symbol Insertion Antenna 0 output Antenna 1 output User 0 Spreading C0 C0 0 , 0d
d
0,L1 * , 0 Ld
* 1 2 , 0 Ld
Ld
0,d
0,2L1 * 1 , 0
d
L * 0 , 0d
Time i Time i+1 Time i Time i+1V-BLAST + MC-CDMA
V-BLAST + MC-CDMA
User u Pilot Insertion OFDM Modulation Training Symbol Insertion Constellation Mapping V-BLAST Spreading Spreading Pilot Insertion OFDM Modulation Training Symbol Insertion Antenna 0 output Antenna 1 output User 0 Spreading C0 C0 0 , 0d
d
0,L1 Ld
0,2d
0,3L1 Ld
0,d
0,2L1 1 4 , 0 Ld
Ld
0,3 Time i Time i+1 Time i Time i+1V-BLAST
Considerations in Receiver
Synchronization Tasks (1/2)
Synchronization Tasks (1/2)
• Coarse symbol boundary detection
– Training symbol 0 still has two
repetitions in the time domain.
• Fractional CFO acquisition
– Using training symbol 0
• Integer CFO acquisition
– Using training symbol 1
– Equivalent channel response H
00,k-H
01,k• Fine symbol boundary detection
– Using training symbol 0
– Equivalent channel response H
00,k+H
01,kAntenna 0 Antenna 1 Training symbol 0
Training symbol 1 Normal symbol 0 Normal symbol 1
Synchronization Tasks (2/2)
Synchronization Tasks (2/2)
• Estimation for residual CFO and
TFO
– Using pilot subcarriers.
– Using phase difference of conse
cutive symbols in the frequency
domain.
– Problem arises
due to alternativ
e pilot data transmitted by ante
nna 1.
– Simple solution: using pilot data
separated by 2 symbols
Antenna 0 Antenna 1 Training symbol 0 Training symbol 1 Data symbol 0 Data symbol 1AWGN
ICI
H
A
H
A
Z
k,i
k 0,k
k 1,k
AWGN
ICI
e
H
A
H
A
Z
N k N j k k k k i k T
) ( 2 , 1 , 0 1 ,(
)
N Data symbol 2Channel Estimation (1/2)
Channel Estimation (1/2)
• Static channels
– Matrix inverse
– Linear interpolation
– Channel estimates apply to the following normal symbols
Training symbol 0 Training symbol 1 Antenna 0 Antenna 1
V
SH
R
V
S
H
R
S
H
ˆ
1
1Channel Estimation (2/2)
Channel Estimation (2/2)
• Dynamic channels
– Two data symbols grouped together
– Getting channel estimates in pilot subcarriers
– Raised-Cosine frequency-domain channel interpolator
Data symbol 1 Data symbol 0 Antenna 0 Antenna 1
V
SH
R
ˆ
Combining and Despreading
Combining and Despreading
- STBC (1/3)
- STBC (1/3)
• STBC[1]
Antenna 0 input FFT Channel Estimation STBC Decoding Despreading De-Mapping MRC Algorithm MMSE Algorithm EM-based Detection PIC Algorithm Antenna 1 input FFTCombining and Despreading
Combining and Despreading
- STBC (2/3)
- STBC (2/3)
• Received signal after DFT
– b
: Multi-users’ signal in two time slots (2LUx1)
– C
: Spreading matrix (NxLU)
– H00
,
H
01: Channel complex gain (NxN)
• MRC
– Can’t reduce MAI
• MMSE
– Minimize the mean squared error per user data
n
Φb
n
b
C
H
C
H
C
H
C
H
r
* 00 * 01 01 00)
(
)
(
)
(
ˆ
Φ
r
b
MRC
dec
H)
)
(
(
ˆ
Φ
ΦΦ
2I
1r
b
MMSE
dec
H H
N [1]
Combining and Despreading
Combining and Despreading
- STBC (3/3)
- STBC (3/3)
• EM (Expectation-Maximization)-based detection[4]
– Arbitrary positive real scalar
– E-step
– M-step
• PIC (Parallel-Interference Cancellation) detector
– Iterative
u u u uΦ
b
n
x
n
n
E
nn
E
u u
u
U u 1n
un
u
U u 1 u1
U u U u u u u U u 1x
u 1Φ
b
1n
r
U u n u u u n u u u 1 ) ( ) (ˆ
ˆ
ˆ
Φ
b
r
Φ
b
x
2 ) 1 (arg
min
||
ˆ
||
ˆ
n u u u u ub
Φ
x
b
b
[1]
U u j j n j j H n u , 1 ) ( ) 1 (ˆ
ˆ
C
G
r
Φ
b
b
Performance
Performance
• Simulation parameters
[1]
– Carrier frequency : 2.56
GHz
– Bandwidth : 5 MHz
– N: 512
– U=32 (full loaded)
• PIC and EM detection
– Initial : MRC
Combining and Despreading
Combining and Despreading
- V-BLAST (1/3)
- V-BLAST (1/3)
• V-BLAST[5]
Antenna 0 input FFT Channel Estimation V-BLAST Decoding Despreading De-Mapping ZF Algorithm MMSE Algorithm IC-ZF Algorithm IC-MMSE Algorithm Antenna 1 input FFT Channel EstimationCombining and Despreading
Combining and Despreading
- V-BLAST (2/3)
- V-BLAST (2/3)
• Received signal after DFT
– b
u,k: data of the user u at the subcarrier k of two transmit antenna (2x1)
– H
k: Channel complex gain (2x2)
– r
k: received signal at the subcarrier k (2x1)
• ZF (Zero-Forcing)
– Using channel estimates to solve the two linear equations.
• MMSE
– MMSE per subcarrier
k k k k U u u k k u k k
c
H
b
n
H
V
n
r
1 , ,)
(
ˆ
1 0 1,
C k k k k ZFdec
c
H
r
b
)
)
(
(
ˆ
1 0 1 2 2 , 1
G k k H k k H k k MMSEdec
c
U
H
H
I
H
r
b
[5]
Combining and Despreading
Combining and Despreading
- V-BLAST (3/3)
- V-BLAST (3/3)
• Interference cancellation (IC) – ZF algorithm
– Initial :
– Recursion :
• IC-MMSE algorithm
– Change the pseudo-inverse to MMSE coefficient
k kH
G
(0) 2 ) 0 ( ) 0 (arg
min
||
(
)
||
j k jG
a
) ()
(
( ) ) ( n a n k n k
G
w
b
ˆ
k,a(n)
dec
(
w
(kn)r
k(n))
) ( ) ( , ) ( ) 1 ((
)
ˆ
n n k a a k n k n kr
H
b
r
((
)
)
) ( ) 1 ( n a k n kH
G
2 ) 1 ( ) 1 (arg
min
||
(
)
||
j n k j nG
a
Maximize SNR
Nulling the column
Performance
Performance
• Simulation parameters [5]
– N: 64
– Spreading factor : 8
– U=4 (half-loaded)
– Antenna diversity : 4x4
• Iterative detection before
despreading suffers MAI a
nd error propagation
Conclusion
Conclusion
• MIMO techniques incorporated into MC-CDMA systems are considered.
• Transmitter modification includes
– Pattern of training symbol and pilot subcarriers (done)
– MIMO encoding block (done)
• Receiver modification includes
– Joint estimation of residual CFO and TFO (done)
– MIMO decoding block
• Performance of MIMO decoding in the MC-CDMA systems does not hav
e the same trend as in the OFDM systems due to MAI.
Reference
Reference
[1] S. Iraji and J. Lilleberg, “ Interference cancellation for space-time block-coded MC-CDMA systems over multipath fading channels,” in Proceeding of IEEE VTC’03, pp.1104-1108. [2] V. Nangia and K. L. Baum, “Experimental broadband OFDM systems field results for OFD
M and OFDM with frequency domain spreading,” in Proceeding of IEEE VTC’02, pp. 223-227.
[3] D. Wang, G. Zhu and Z. Hu, “Optimal pilots in frequency domain for channel estimation in MIMO-OFDM systems in mobile wireless channesl”, in Proceeding of IEEE VTC’04 Spri ng.
[4] M. Feder and E. Weinstein, “Parameter estimation of superimposed signals using the EM algorithm”, IEEE Trans. on Acoustics, Speech and Signal Processing, vol. 36, pp. 477-48 9, Apr. 1988.
[5] Z. Lei, X. Peng and F. P. S. Chin, “V-BLAST receiver for downlink MC-CDMA systems,” in Proceeding of IEEE VTC’03, pp.866-870.