Presentation held at EUSIPCO'2010 about low-complexity detection of Golden and SS codes in BICM-OFDM scenarios such as DVB-T2..

1. FRFD SFBC
for DTV
Broadcasting
Iker Sobron
EUSIPCO
Full-Rate Full-Diversity Space-Frequency Block
2010
Coding for Digital TV Broadcasting
Introduction
MIMO system
model
SFBC schemes Iker Sobron, Mikel Mendicute and Jon Altuna
DVB-T2
Soft detection Dpt. of Electronics and Computer Science
University of Mondragon - Mondragon Goi Eskola Politeknikoa
Simulation
results
August 26, 2010 Aalborg (Denmark)
Conclusions isobron@eps.mondragon.edu
and future
work
1 / 21

2. Outline
FRFD SFBC
for DTV
Broadcasting
1 Introduction
Iker Sobron
EUSIPCO 2 MIMO system model
2010
Introduction
3 SFBC schemes
MIMO system
model
SFBC schemes 4 DVB-T2
DVB-T2
Soft detection 5 Soft detection
Simulation
results
6 Simulation results
Conclusions
and future
work
7 Conclusions and future work
2 / 21

3. FRFD SFBC
for DTV
Broadcasting
Iker Sobron
Main objetives
EUSIPCO
2010
Analysis of the FRFD codes performance in DTV
Introduction
scenarios: Rayleigh and Rice channels.
MIMO system
model Review of the soft detection for LDPC-based DTV
SFBC schemes systems.
DVB-T2
Performance comparison among the Golden code, a
Soft detection
Simulation
low-complexity FRFD code and the current DVB-T2
results SFBC code for the same bit rate.
Conclusions
and future
work
3 / 21

4. MIMO System Model
FRFD SFBC
for DTV
Broadcasting The MIMO model with M = 2 transmit and N = 2 receive
Iker Sobron antennas
EUSIPCO y11 y12 h11 h12 x11 x12 z11 z12
2010 = +
y21 y22 h21 h22 x21 x22 z21 z22
Introduction
MIMO system Y H X Z
model
SFBC schemes
DVB-T2
[Y]N×T Received signal matrix
Soft detection [H]N×M Channel matrix
Simulation
results [X]M×T Transmitted matrix or codeword with frequential
Conclusions depth T = 2
and future
work [Z]N×T Additive white Gaussian noise (AWGN) matrix
with CN 0, 2σ 2 per component
4 / 21

5. MIMO System Model
FRFD SFBC
for DTV
Broadcasting
Iker Sobron
EUSIPCO
The codeword:
2010
x11 x12
Introduction X= = f (s1 , . . . , sQ )
x21 x22
MIMO system
model A linear combination of Q symbols
SFBC schemes
DVB-T2 Q
The spatial rate: L = T
Soft detection
Simulation
Single-input single-output (SISO) system L = 1
results
Full-rate involves Q = 4 and L = 2
Conclusions
and future
work
5 / 21

6. SFBC schemes
FRFD SFBC
for DTV The DVB-T2 SFBC code
Broadcasting
Based on the Alamouti scheme:
Iker Sobron
s1 s2
Xal = ∗ ∗
EUSIPCO −s2 s1
2010
Introduction
Provides FD but not FR, L = 1 as the SISO case.
MIMO system In a 2x2 scheme, maximum ratio combining (MRC) at the
model
reception side.
SFBC schemes
DVB-T2 Payload cells within an OFDM symbol
(Frequency axis)
Soft detection SFN
Transmitter 1
Simulation ... s1 s2 s3 s4 s5 s6 s7 s8 ...
results
Data
Cells Both OFDM
Conclusions SFBC OFDM symbols are
...s1s2s3s4... transmitted
and future sinchronously
work
... -s * s *
2 1
-s4* s3* -s6* s5* -s8* s7* ...
SFN
Transmitter 2
Encoding process is done on
pairs of OFDM payload cells
6 / 21

7. FRFD SFBC schemes
FRFD SFBC
for DTV
Broadcasting
Iker Sobron The Golden code
EUSIPCO 1 α (s1 + θs3 )
α (s2 + θs4 )
2010
Xg = √
¯ ¯
5 i α s2 + θs4
¯ α s1 + θs3
¯
Introduction
√ √
¯
with θ = 1+2 5 (the golden number), θ = 1− 5
α = 1 + i − iθ
MIMO system 2 ,
model ¯ ¯
and α = 1 + i − i θ.
SFBC schemes
Maximal coding gain
DVB-T2
Soft detection
Maximum likelihood (ML) detection: exhaustive search or
Simulation sphere decoder (SD)
results
ˆml = arg min Y − HX 2
s
Conclusions s
and future
work Complexity O(P 4 ) being P, the modulation order.
7 / 21

8. FRFD SFBC schemes
FRFD SFBC
for DTV
Broadcasting
Iker Sobron
EUSIPCO
Low-complexity FRFD code designed by Sezginer-Sari
2010
as1 + bs3 as2 + bs4
Introduction Xfrfd = ∗ ∗ ∗ ∗
MIMO system
−cs2 − ds4 cs1 + ds3
√ √
model √ 1− 7+i (1+ 7)
SFBC schemes
with a = c = 1/ 2, b = √
4 2
and d = −ib.
DVB-T2 Lower coding gain than the Golden code’s
Soft detection
Detection combines one symbol-by-symbol stage plus an
Simulation
results ML stage with complexity O(P 2 ).
Conclusions
and future
work
8 / 21

9. DVB-T2
FRFD SFBC
for DTV
Broadcasting
Cell, Time and
LDPC Bit Symbol
Iker Sobron Frequency SFBC OFDM
Binary Encoder Interleaving Mapper
Interleavers
Data
EUSIPCO
2010
Introduction
MIMO system Second generation of the European DVB-T standard
model
SFBC schemes
Low-Density Parity Check (LDPC) encoding
DVB-T2 Several interleavers included a time interleaver for mobile
Soft detection reception
Simulation
results High modulation orders (up to 256-QAM)
Conclusions The first DTV standard with multi-antenna processing
and future
work
Single Frequency Networks ⇒ distributed MISO or MIMO
as optional
9 / 21

10. DVB-T2: MIMO distributed
FRFD SFBC
for DTV
Broadcasting
Iker Sobron
Tx 1 Tx 2
EUSIPCO
2010
Introduction
MIMO system
model
SFBC schemes SFN Cell 1 SFN Cell 2
DVB-T2
Soft detection
Simulation Every transmit antenna of the MIMO system is a transmitter of
results
the SFN network.
Conclusions
and future Both antennas transmit synchronously the encoded signal at
work
the same carrier frequency.
10 / 21

11. Soft detection
FRFD SFBC
for DTV
Broadcasting
Maximum a posteriori (MAP) detection
Iker Sobron
EUSIPCO
LD (bk |Y) = ln Pr[bk =+1|Y] = LA (bk ) + LE (bk |Y)
Pr[bk =−1|Y]
2010
where LA (bk ) and LE (bk |Y) denote the a priori and extrinsic
Introduction
information, respectively and k = 0, . . . , MT log2 P − 1.
MIMO system
model
SFBC schemes Extrinsic information (LE -value)
DVB-T2  
Soft detection p(Y|b) exp LA (bj )
b∈Bk,+1 j∈Jk,b
Simulation
results LE (bk |Y) = ln  
Conclusions p(Y|b) exp LA (bj )
and future b∈Bk,−1 j∈Jk,b
work
where p(Y|b) represents the likelihood function.
11 / 21

12. Soft detection: some definitions
FRFD SFBC
for DTV
Broadcasting
Iker Sobron Rearrangement of X, Y and Z as x = [x11 , x21 , x12 , x22 ]T ,
¯
¯ = [y11 , y21 , y12 , y22 ] and ¯ = [z11 , z21 , z12 , z22 ]T ,
y T
z
EUSIPCO
2010 respectively.
Introduction
Definition of the real-value column vector:
MIMO system
model ¯R = [ {x11 } , {x11 } , {x21 } , . . . , {x22 }]
x
SFBC schemes
DVB-T2 Y = HX + Z can be rewritten as
Soft detection ˘x ˘ s
¯R = H¯R + ¯R = HG¯R + ¯R
y z z
Simulation
results where G is the code generator matrix and the real-value channel
1 j
is H = (1/2)I2 ⊗ (H ⊗ E + H∗ ⊗ E∗ ) with E =
˘
Conclusions
and future .
work −j 1
12 / 21

13. Soft detection
FRFD SFBC
for DTV
Broadcasting
The likelihood function for MAP detection
˘ s 2
¯R −HG¯R
y
Iker Sobron exp
2σ 2
p(Y|¯R = map(b)) =
s
EUSIPCO (2πσ 2 )NT
2010
Introduction
Only the term inside the exponent is relevant for the
MIMO system
calculation of the LE -value.
model
SFBC schemes
Max-log approximation for the LE -value
DVB-T2
Soft detection
1 1 ˘ s
2
Simulation LE (bk |Y) ≈ max − 2 ¯R − HG¯R
y + bT LA,[k]
[k]
results
2 b∈Bk,+1 σ
Conclusions
and future 1 1 ˘ s
2
work − max − 2 ¯R − HG¯R
y + bT LA,[k]
[k]
2 b∈Bk,−1 σ
13 / 21

14. Soft detection
FRFD SFBC
for DTV
Broadcasting
Iker Sobron In order to reduce the complexity of MAP detection, extrinsic
information is calculated through a list L of Ncand candidates
EUSIPCO
2010 with the smallest ML metrics.
Introduction
List ML detection
MIMO system
model
SFBC schemes 1 1 ˘ s
2
DVB-T2
LE (bk |Y) ≈ max − 2 ¯R − HG¯R
y + bT LA,[k]
[k]
2 b∈L∩Bk,+1 σ
Soft detection
1 1 ˘ s
2
Simulation − max − 2 ¯R − HG¯R
y + bT LA,[k]
[k]
results 2 b∈L∩Bk,−1 σ
Conclusions
and future
work
14 / 21

15. Simulation results: parameters
FRFD SFBC
for DTV The DVB-T2 parameters are:
Broadcasting
Iker Sobron Length of LDPC block: 64800 bits
EUSIPCO
LDPC code rate: R = 2/3
2010
Constellation sizes: QPSK, 16-QAM and 256-QAM
Introduction
MIMO system FFT size: 2048 carriers (2K)
model
Guard interval: 1/4
SFBC schemes
DVB-T2 Non-iterative reception (without a priori information).
Soft detection Number of candidates: Ncand = 100
Simulation
results Channel parameters:
Conclusions
and future
Rayleigh channel (Typical Urban of six path, TU6) and a
work Ricean channel (Rural Area of six paths, RA6).
Perfect channel state information (CSI).
15 / 21

16. Simulation results: TU6
FRFD SFBC
for DTV
Broadcasting −1
10
X 16QAM
frfd
Iker Sobron X 16QAM
g
−2
X 256QAM
al
EUSIPCO 10
X QPSK
2010 frfd
X QPSK
g
Introduction
−3
Xal 16QAM
10
MIMO system
model
BER
SFBC schemes −4
10
DVB-T2
Soft detection
−5
10
Simulation
results
Conclusions −6
and future 10
0 1 2 3 4 5 6 7 8 9
work Eb/No (dB)
16 / 21

17. Simulation results: TU6
FRFD SFBC
for DTV
Broadcasting −1
10
X 16QAM
frfd
Iker Sobron X 16QAM
g
−2
X 256QAM
al
EUSIPCO 10
X QPSK
2010 frfd
X QPSK
g
Introduction
−3 I
@ Xal 16QAM
10
MIMO system i
P @
model
PP
BER
PP @
SFBC schemes P PP @
−4
10 P
DVB-T2 PP @
PP @
Soft detection PP
−5
10
P ≈ 0.3 dB
@
Simulation
results
Conclusions −6
and future 10
0 1 2 3 4 5 6 7 8 9
work Eb/No (dB)
16 / 21

18. Simulation results: RA6 LOS
FRFD SFBC
for DTV
Broadcasting −1
10
Xfrfd 16QAM
Iker Sobron Xg 16QAM
Xal 256QAM
EUSIPCO
−2 Xfrfd QPSK
2010 10
Xg QPSK
Introduction Xal 16QAM
MIMO system
model
BER
−3
10
SFBC schemes
DVB-T2
Soft detection −4
10
Simulation
results
Conclusions −5
and future 10
1 2 3 4 5 6 7 8 9 10
work Eb/No (dB)
17 / 21

19. Simulation results: RA6 LOS
FRFD SFBC
for DTV
Broadcasting −1
10
Xfrfd 16QAM
Iker Sobron Xg 16QAM
Xal 256QAM
EUSIPCO
−2 Xfrfd QPSK
2010 10
Xg QPSK
Introduction Xal 16QAM
MIMO system i
P
model
PP
BER
−3
10 PP I
@
SFBC schemes P PP @
P
DVB-T2 PP @
PP @
Soft detection −4
10
PP
@ ≈ 0.2-0.4 dB
P
Simulation
results
Conclusions −5
and future 10
1 2 3 4 5 6 7 8 9 10
work Eb/No (dB)
17 / 21

20. Conclusions
FRFD SFBC
for DTV
Broadcasting
Iker Sobron
FRFD codes fully exploit the MIMO channel capacity
EUSIPCO allowing an increase of the data rate.
2010
Introduction
The FRFD codes achieve a higher performance gain in
MIMO system
Rayleigh channels than the 2x2 scheme based on Alamouti
model
for high bit rates but that is reduced in low bit rates.
SFBC schemes
DVB-T2
The Sezginer-Sari code obtains a performance 0.3 dB lower
Soft detection
than the Golden code with reduced detection complexity.
Simulation Both Sezginer-Sari and Golden codes provide a reduction
results
of the performance in LOS environments having the 2 × 2
Conclusions
and future Alamouti scheme a higher gain.
work
18 / 21

21. Future work
FRFD SFBC
for DTV
Broadcasting
Iker Sobron
EUSIPCO
2010
Design of a low-complexity detection algorithm of FRFD SFBC
Introduction
codes
MIMO system
model Based on a fixed complexity tree search.
SFBC schemes
Complexity for Golden code: from O(P 4 ) to O(P 2 )
DVB-T2
Soft detection
Performance close to log-MAP approximation.
Simulation
results
Conclusions
and future
work
19 / 21

22. Future work
FRFD SFBC
for DTV Some results
Broadcasting
Detection @ Eb/N0 = 7 dB
Iker Sobron −1
10
Gold List ML
EUSIPCO Gold. nT=P2
2010
Gold. nT=4P2
Introduction
SS List ML
MIMO system −2 SS nT=P2
model 10
SS nT=4P2
SFBC schemes
BER
DVB-T2
Soft detection
−3
Simulation 10
results
Conclusions
and future
work
−4
10
0 10 20 30 40 50 60 70 80 90 100
Number of candidates Ncand
20 / 21

23. FRFD SFBC
for DTV
Broadcasting
Iker Sobron
EUSIPCO
2010
Introduction
MIMO system
Thanks for your attention
model Iker Sobron
SFBC schemes isobron@eps.mondragon.edu
DVB-T2
Soft detection
Simulation
results
Conclusions
and future
work
21 / 21