Enhanced Two-Dimensional Data-aided Channel Estimation for TDS-OFDM
INSTITUT D’ÉLECTRONIQUE ET DE TÉLÉCOMMUNICATIONS DE RENNES
Ming LIU,
[email protected],
Matthieu CRUSSIÈRE, Jean-François HÉLARD
[email protected] [email protected] European University of Brittany (UEB), INSA-Rennes, Institute of Electronics and Telecommunications of Rennes (IETR), France
Context Time Domain Synchronous OFDM (TDS-OFDM): Classical Cyclic Prefix (CP) is replaced by a known pseudo-random (PN) sequence PN sequence is reused as training sequence for channel estimation & synchronization
~ H2
Xˆ
Hˆ 2
H2
Hˆ1 ˆ H
Channel estimation problem: interference from OFDM data symbols degrades the PN-based channel estimation; unreliable channel estimation causes imperfect PN removal; residual PN sequence introduces interference to the OFDM data symbols.
cˆ
Block diagram of TDS-OFDM system.
TDS-OFDM requires quite reliable channel estimation.
Algorithm Description
Data-aided channel estimation
PN-based channel estimation
Rebuilding soft data symbols:
a priori
1 10
probability
(0 81) (0.81)
(1 19) (1.19)
(1)
5 42
(0 62) (0.62)
(0.2)
3 42
(0.24)
(0.43)
1 42
(0.024)
−1 10
−
1 10
3 10 Real
virtual training sequence
3 10
D Real
1 42
Instantaneous data-aided channel estimation:
(2.33)
D 1 − 10
MSE:
(1.76)
10
− 3 10
Estimator:
(1.38)
(1.8)
7 42
(1.19)
(1)
3
LLR:
PN structure:
Imaginary
Imaginary
5 42
3 42
7 42
Data symbol used in channel estimation Data symbol excluded from channel estimation
Data symbol used in channel estimation Data symbol excluded from channel estimation
coarse estimates for all active subcarriers
2-Dimensional averaging: refined estimates for virtual pilot positions
MMSE combination Linear combination: 2-Dimensional Wiener filtering based interpolation:
Frequency
Minimize the MSE:
Coherence region
Lf Virtual pilot Available estimate Non-available estimate
MSE:
MMSE combination factor:
Lt
Time B
2-D virtual pilot pattern -1
10
Simulation Results
data-aided channel estimation 1 iteration 2 iterations 3 iterations PN base channel estimation
Simulation Configurations -2
10
7.56 MHz
-2
10 BER
420 (1/9)
0.1 to 0.2 dB gap
7.8dB gain MSE
3 780
-1
10
-3
10
64QAM
16QAM
16QAM & 64QAM
16QAM
-3
10
LDPC (R = 0.6) + BCH(762, 752)
perfect CIR data-aided channel estimation PN based channel estimation method in [5]
-4
B=52, M=240 (170 OFDM symbols)
10
-4
500 MHz 30 km/h Rectangular shape with Lt = 2, Lf = 9 21 OFDM symbols
10 5
10
15
20
25
30
SNR (dB)
MSE of the proposed data-aided channel estimator comparing with the PN-based method in TU-6 channel.
Conclusion The proposed data-aided channel estimation method: achieves near optimal performance in terms of the BER in the specification of DTMB system; requires low computational complexity compared to the Turbo-like channel estimation algorithm; adapts to long channel impulse response cases; also adapts to the traditional CP-OFDM based systems.
12
64QAM
13
14
15
16
17 18 SNR (dB)
19
20
21
BER of the DTMB system using the proposed dataaided channel estimation method in TU-6 channel.
22