Optimal CSD and Truncated SVD for Channel Estimation Month Year doc.: IEEE 802.11-yy/xxxxr0 September 2008 Optimal CSD and Truncated SVD for Channel Estimation Date: 2008-09-08 Authors: Wahyul Syafei, KIT JAPAN John Doe, Some Company
Month Year doc.: IEEE 802.11-yy/xxxxr0 September 2008 Abstract Adjusting the cyclic shift of long training field will minimize the condition number. Truncating the singular value decomposition of the estimated channel will minimize the noise effect. Combination of both technique will enhance the channel estimation quality Wahyul Syafei, KIT JAPAN John Doe, Some Company
Optimal CSD and Truncated SVD for Channel Estimation September 2008 Optimal CSD and Truncated SVD for Channel Estimation Singular-Value truncation: minimizing noise effect on estimated CIR. Cyclic shift adjustment of Long preamble on the successive antenna: minimizing Condition-number between singular values of the estimated CIR. Wahyul Syafei, KIT JAPAN
… … MIMO-OFDM Tx system September 2008 Channel Coding Mux CP Time domain Cyclic shift Channel Coding Mux CP Long Preamble 1 d1 Mapper IFFT Mux CP Long Preamble 2 d2 Mapper IFFT … … Mux CP Long Preamble N dN Mapper IFFT Wahyul Syafei, KIT JAPAN
Maximum Likelihood Decoder September 2008 MIMO-OFDM Rx system Error detection Maximum Likelihood Decoder de Mux CP Channel estimation 1 d1 FFT de Mux CP Channel estimation 2 d2 FFT … … de Mux CP Channel estimation N dN FFT Wahyul Syafei, KIT JAPAN
Cyclic Shifting on LTF September 2008 Tx Antenna # LTF 1 STF C0 C1 C2 SIG DATA 2 STF cs C0 C1 C2 SIG DATA N STF C2 cs C0 C1 SIG DATA Wahyul Syafei, KIT JAPAN
MISO case September 2008 L1 h1 Long preamble 1 rlp Channel estimator Error of ML estimation will be minimized if: or with L are the long preamble symbols and already known by the receiver. Wahyul Syafei, KIT JAPAN
September 2008 MISO case 32 32 145 Wahyul Syafei, KIT JAPAN
SV-Truncation September 2008 To minimize the noise effect, SV-truncation ( ) is proposed. From the known LTF: The new term of known LTF becomes: Estimated CIR is: Exp. the truncated SVD for q =2, 32 32 145 Wahyul Syafei, KIT JAPAN
SV-Truncation V X+ L U rlp,j September 2008 (32 X 32) Proposed Add zeros FFT S/P S D (32 X 145) (145 X 1) (32 X 1) (16 X 1) 64 (64 X 1) Inner product (32 X 145) (32 X 145) (145 X 145) Wahyul Syafei, KIT JAPAN
Condition Number September 2008 The ratio between the biggest singular value to the smallest one, written as: If the CN =1, then is the constant X identity matrix and zeros matrix, which is easier to be handled. Regarding the SVD truncation, there is a Partial Condition Number: Find the smallest value (PCN 1), by adjusting the cyclic shift, as shown in Simulation 1. Wahyul Syafei, KIT JAPAN
Simulation 1, find the optimal CS September 2008 Simulation 1, find the optimal CS CS=80 CS=18 CS=64 No truncation Wahyul Syafei, KIT JAPAN
Find the optimal q September 2008 1. Consider the noise part. 2. Consider the channel error. where and… Wahyul Syafei, KIT JAPAN
Find the optimal q September 2008 where Example for q = 0, 32 32 Considering the channel error and noise part, we have. min as shown in Simulation 2. Wahyul Syafei, KIT JAPAN
Simulation 2, find the optimal q September 2008 Simulation 2, find the optimal q CS=18 samples; min J is found at q=4 CS=80 samples; min J is found at q=5 CS=64 samples; min J is found at q=5 Wahyul Syafei, KIT JAPAN
Simulation 3, Performance September 2008 Simulation 3, Performance LTF type IEEE802.11a FEC Viterbi R=1/2, K=7 Configuration 2X2 MIMO OFDM Channel ,Exponential attenuation Antenna diversities STBC and Rx diversity Fading Independent, slow fading Modulation BPSK CS;q 18;4 , 64;5 , 80;5 Wahyul Syafei, KIT JAPAN
September 2008 Conclusion The truncation of SVD of the estimated channel will minimize the noise effect. The cyclic shift adjustment will minimize the partial condition number of SVD of the estimated channel. The combination of both will increase the quality of channel estimation. Wahyul Syafei, KIT JAPAN
References IEE802.11a IEE802.11n draft September 2008 Wahyul Syafei, KIT JAPAN