Channel Estimation for Wired MIMO Communication Systems Final Presentation Daifeng Wang
Download ReportTranscript Channel Estimation for Wired MIMO Communication Systems Final Presentation Daifeng Wang
05/05/2005 Channel Estimation for Wired MIMO Communication Systems Final Presentation Daifeng Wang Dept. of Electrical and Computer Engineering The University of Texas at Austin [email protected] Introduction • Review – Wired MIMO Communication Systems – Multicarrier Modulation – Training-Based Channel Estimation • Today – – – – Which channel estimation strategy for wired communication systems? How to design the training sequence? What is the channel model? How to estimate mirror Multiadd D/A + the wired MIMO Bits S/P Modulation data cyclic P/S transmit and prefix filter Encoder N-IFFT channel? TRANSMITTER Data Transmission for ADSL, a wired communication system RECEIVER N/2 subchannels N real samples 5/25/2016 P/S decoder freq. domain equalizer (invert channel) N-FFT and remove mirrored data delete S/P cyclic prefix time domain equalizer (FIR filter) channel receive filter + A/D Training-Based Channel Estimation Strategy • Block-Type – – – – All subcarriers + Period Least Square (LS), Minimum Mean-Square (MMSE) Slow Fading/Varying Channels Decision Feedback Equalizer Tradeoff between performance and • Comb-Type complexity – Selective subcarriers + Interpolation – LS, MMSE, Least Mean-Square (LMS) – Fast Fading/Varying Channels – Interpolation • • • • • 5/25/2016 Linear Second order Low-pass Spline Cubic Time domain Training Sequences • y=Sh+n – h: L-tap channel – S: transmitted N x L Toeplitz matrix made up of N training symbols – n: AWGN Domain Method Time Periodic Minimum Complexity MSE Optimal Sequence* Yes High(2N) Yes No Medium(N2) Yes Almost Low(Nlog2N) Sometimes No Low(Nlog2N) Sometimes [Chen & Mitra, 2000] Aperiodic [Tella, Guo & Barton, 1998] L-Perfect (MIMO) [Fragouli, Dhahir & Turin, 2003] Frequency Periodic [Tella, Guo & Barton, 1999] * impulse-like autocorrelation and zero crosscorrelation 5/25/2016 Training-Based MIMO Channel Model • 2 X 2 MIMO Model TX 1 RX 1 h11 TX 2 h12 h21 RX 2 h22 y h ( L) y 1 Sh z [ S1 ( L, N t ) S 2 ( L, N t )] 11 y2 h21 ( L ) where y and z are of 2( N t L 1) 1 T hij ( L ) hij ( L 1) hij (0) , i or j 1, 2 si (0) si ( L 1) s ( L) s (1) i i , i 1, 2 S i ( L, N t ) 5/25/2016 s ( N 1) s ( N L ) t i t i h12 ( L ) z h22 ( L ) Training-Based Channel Estimation for MIMO • Least Square (LS) – Assumes S has full column rank hˆ11 ĥ= hˆ21 hˆ12 H -1 H =(S S) S y hˆ22 • Mean-Square (MSE) H 2 – zero-mean and white Gaussian noise: R z E zz 2 I Nt - L 1 H – 2 H -1 ˆ ˆ h h 2 Tr((S S) ) MSE E h h 2 2 L – H H S S S 1 1 2 S1 H MMSE = , iff S S= H ( Nt L 1)I 2 L H Nt L 1 S1 S2 S2 S2 – Sequences satisfy above are optimal sequences – Optimal sequences: impulse-like autocorrelation and zero crosscorrelation 5/25/2016