Channel Estimation Schemes for OFDM Relay-Assisted System Darlene Maciel, C. Ribeiro, A. Silva e Atílio Gameiro darlene@av.it.pt Workshop 2009
Outline Introduction Motivation PACE Schemes Simulation Scenario Results Conclusion Future Works
Introduction: Diversity Diversity is inherent in the physical layer: PHY diversity Time, frequency, space (antenna) and polarization diversity Combat the fading channel by trying to flatten the channel Diversity can also be achieved in the MAC or higher layer: Network diversity Multiuser diversity (by scheduling or routing) Cooperative diversity (by cooperative transmission)
Introduction: Cooperation Redundant transmission is realized via the cooperation of third party devices rather than solely from the originating device; D Node S cooperates with neighbors to send information to D S AF, DF, SDF or CF Half dupplex AF: 2 phases RN Transmitted signal S D Received signal Simple forms of cooperation involves 3 links
Introduction: Amplify-and-Forward Protocol RN S D In the half duplex AF protocol receiver at D needs First phase: Estimate channel S-D: Single link Conventional Channel Estimation Second Phase: Estimate channel S-RN-D: Compound Channel
Motivation: Equivalent Channel Compound Channel S h (1) () t RN h (2) () t D h h h h h h Eq (1) (2) (1) (2) Eq ( t) A ( t) (t) ( m) A ( m) ( m) The Power Delay Profile - PDP PDP = E{ h ( t) h ( t) } = PDP PDP (1) (2) 2 (1) (2) Maximum delay = Delay channel 1 + Delay channel 2 # Taps of the compound channel will depend on the both channels PDP
Motivation: Equivalent Channel S h (1) () t RN h (2) () t D (1) (2) h h ES, : 0, 1 Two sources of Noise. The total noise at the D: w m Ah m w m w m (2) (1) (2) ( ) ( ) ( ) ( ) t ( m) E{ w ( m) } A 2 2 2 2 2 t t n n Conditioned to a specific channel realization the noise variance: 2 ( m) A h ( m) 2 2 (2) 2 2 t n n The conventional channel estimation schemes should be adapted to this scenario LS, MMSE
Motivation: Questions Questions to be Solved: How does the statistics of the compound channel affect the performance of classical PACE in OFDM signalling? (2) How much can be gained through the knowledge of h?
DFT Classical Pilot Aided Channel Estimation Schemes FD LS (Least Square): h ˆP LS hˆ LS X Y 1 P Wh ˆP LS P W: Interpolator AWGN + LS Estimate Channel Estimator ĥ LS FD MMSE Filter ĥ MMSE Equalisation FD MMSE (Mean Minimum Square Error) : ĥ W MMSE W MMSE R R 1 MMSE HP P hˆ P LS R R P HP Autocorrelation; Cross-correlation;
Classical Pilot Aided Channel Estimation Schemes TD-MMSE AWGN CIR Group + h ˆ LS Channel Estimator TD MMSE Filter h ˆ MMSE Equalisation Example CIR Estimate CIR estimate [ n] hh 2 W n R [ n] n hh R R [ n] Auto-correlation Function Channel PDP hh PDP PDP PDP 1 2 2 2 n t
Simulation Scenario Scenario and Parameter S RN D Modulation Path Delay(ns) QPSK Relative Power(dB) Sampling frequency 1 (LTE) 0.0 15.36 MHz 0.0 # Subcarriers 2 65.1 (T) 1024-0.7 Link Analized 3 260.4 (4T) Compound -0.8 channel For reference 4 586.0 (9T) Conv. SISO -6.0 Channels 5Noise statistics 1041.67 (16T) identical -10.0 Channel 6 1627.6 (25T) 7 Taps -14.0 7 2474.0 (38T) -19.0
Simulation Scenario Simulation Parameters The pilots are multiplexing in the symbol: Frequency Data Pilot Nf= 32; 4 Nt= 1; 12 Time
MSE (db) MSE (db) Results FD LS Estimator LS Channel Estimation -2-3 -4-5 -6-2 -3-4 -5-6 -7-8 -9-10 -11-12 Nf=32, Nf=32, Nt=1,Relay Nt=1,Relay on on Nf=32, Nt=12,Relay on on Nf=4, Nf=4, Nt=1,Relay on on Nf=4, Nt=12,Relay on on Nf=32, Nt=1,Conv. SISO Nf=32, Nt=12,Conv. SISO Nf=4, Nt=1,Conv. SISO Nf=4, Nt=12,Conv. SISO 0 1 2 3 4 5 6 7 8 9 E b /N0 (db) -7-8 -9-10 -11-12
MSE (db) MSE (db) Results TD MMSE Estimator MMSE Channel Estimation -8-10 -12-14 -8-10 -12-14 -16-16 -18-20 -22-24 -26-28 Nf=32, Nt=1,Relay on Nf=32, Nt=1,Relay on Nf=32, Nt=12,Relay on Nf=32, Nt=12,Relay on Nf=4, Nt=1,Relay on Nf=4, Nt=1,Relay on Nf=4, Nf=4, Nt=12,Relay Nt=12,Relay on on Nf=32, Nt=1,Conv. SISO Nf=32, Nt=12,Conv. SISO Nf=4, Nt=1,Conv. SISO Nf=4, Nt=12,Conv. SISO 0 1 2 3 4 5 6 7 8 9-28 E b /N0 (db) -18-20 -22-24 -26 5 db
Magnitude Magnitude Example 0.4 0.2 Point-to-point Channel PDP P2P Channel N1 = 7 Taps 0 0 10 20 30 40 50 Compound Channel PDP 0.2 0.1 0 0 10 20 30 40 50 N2 can be quite larger than N1 Compound Channel conv (P2P Ch, P2P Ch) N2 = 27 Taps N2 N1 1 N1 2 Taps SNR per Tap MSE
ChEst MSE (db) How much can be gained through the Knowledge oh h2? -12-14 -16-18 -20 Channel Estimation MSE vs. Eb/N0 2 n t mm m n 2 n 0 5 10 Eb/N0 (db) 2 2 2 2 2 2 m A h m t n n No noticeable improvement by the knowledge of h2; Number of non-zero taps << Nc/Nf Filter design is robust to errors in the estimate of the noise variance
Conclusion In AF the equivalent channel S-RN-D has a larger delay than point-to-point Increases the minimum pilot density that can be used; Degrades the performance of the MMSE; The robustness of the TD-MMSE filter to errors in the estimate of noise variance The knowledge of individual P2P channels does not bring any noticeable improvement;
Future Works Consider a scenario which the channel statistics can bring improvements to the channel estimate: Antenna array at the BS; Equalize-and-Forward Protocol; Power constraints at the RN; Channels with different statistics.