Adapive Anenna Array for Reliable OFDM Transmission Shinsuke Hara Deparmen of Elecronics, Informaion Sysem And Energy Engineering, Graduae School of Engineering, Osaka Universiy
Conens of This Presenaion * Wha are 4G Sysems? * Moivaion, * Adapive Anenna Array for Suppression of Doppler-Shifed Signals, * Adapive Anenna Array for Suppression of Delayed Signals beyond Guard Inerval, * Conclusions.
Wha are 4G Sysems? 4G Sysems should suppor 20~0Mbps ransmission in downlink, and 2~20 Mbps ransmission in uplink, even for high-speed cruising mobiles. Where in frequency band can we provide he services? User Mobiliy 3 (< 5~6) GHz bands? Fas Mobile Saionary Movable Slow Mobile IMT-2000 (3G Sysems) 4G Sysems High-Rae Wireless LANs IEEE802.11a, HIPERLAN/2, MMAC 20 2000 0.1 1.0.0 0.0 Transmission Rae [Mbis/sec]
OFDM scheme is a candidae as a physical layer proocol for 4G sysems. J.Chuang and N.Sollenberger, "Beyond 3G: Wideband Wireless Daa Access Based on OFDM and Dynamic Packe Assignmen,'' IEEE Commun. Mag., Vol.38, No.7, pp.78-87, July 2000. τ The BER of OFDM scheme is degraded by Doppler-Shifed Signals f Delayed Signals beyond Guard Inerval
Moivaion Here, focusing our aenion on Doppler-shifed signals and delayed signals beyond Guard inerval, How can we suppress hem? Temporal Equalizaion high compuaional complexiy (agains concep of OFDM) Adapive Anenna Array mus be a good soluion!
OFDM Basics Single-Carrier Modulaion Muli-Carrier Modulaion B B SCM MCM f F k (f;)...... f f 0 f 0 1 f 2 f k f NSC (a) Frequency Specra of Transmied Signals Transfer Funcion H(f;) H k (f;) Adapive Equalizaion Frequency Selecive Fading Channel f f k B k f...... f 0 0 (b) Frequency Specra of Received Signals f
+1-1 +1-1 +1-1 +1-1 +1-1 +1-1 +1-1 c 1 =+1 c 2 =-1 c 3 =-1 c 4 =+1 c 5 =-1 c 6 =+1 c 1 (a) c 2 cos(2π/t s ) c 3 cos(4π/t s ) c 4 cos(6π/t s ) c 5 cos(8π/t s ) c 6 cos(π/t s ) (b) T s T s T s T s T s T s
F(f;) f f 0 (a) Non-Overlapped Band-Limied Orhogonal Signals f=1/t s F(f;) f 0 (b) Overlapped Time-Limied Orhogonal Signals f F(f;) f 0 (c) Overlapped Band-Limied Orhogonal Signals
(i-1)-h Symbol i-h Symbol Symbol Period T s (i-1)-h Symbol Desired Signal i-h Symbol Symbol Period T s (a) No Guard Inerval Inserion i-h DFT Window Timing (Observaion Period) Delayed Signal (a) No Guard Inerval Inserion i-h DFT Window Timing (Observaion Period) Symbol Period T s Desired Signal Symbol Period T s (b) Guard Inerval Inserion s G i-h DFT Window Timing (Observaion Period) s Delayed Signal G i-h DFT Window Timing (Observaion Period) (b) Guard Inerval Inserion Symbol Period T s Desired Signal Symbol Period T s s G i-h DFT Window Timing (Observaion Period) (c) Guard Inerval Inserion wih Cyclic Prefix Firs Pah Second Pah τ max Delayed Signal s G i-h DFT Window Timing (Observaion Period) (c) Guard Inerval Inserion wih Cyclic Prefix τ
2/T s T s (b) Frequency Specrum of Pulse Waveform f G s copy (a) Cyclic Exension Technique f=1/ s 1 2 3... N SC 0 (c) Frequency Specrum of OFDM Signal f
Suppression of Doppler-Shifed Signals Weigh per User/Pre-FFT Type OFDM Adapive Anenna Array u 1 (i) K: Number of Anenna Elemens u 2 (i) u K (i) w * 1 (i) w * 2 (i) w * K (i) y(i)=w H (i)u(i) Σ FFT Demodulaion Weigh Conroller Calculaion of Virual Subcarrier Oupus u(i)=[u 1 (i),..., u K (i)] T (Received Signal Vecor) w(i)=[w 1 (i),..., w K (i)] T (Array Weigh Vecor)
OFDM Adapive Anenna Array Pah#1 Received Signals from one User Pah#1 Σ Pah#2 Pah#3 w 1* 1 (i) w 1* 2 (i) w 1* K (i) Pah#1 Pah#2 Σ Pah#2 Σ Σ Pah#3 w * 1 (i) w* 2 (i)w* K (i) Pah#3 w 2* 1 (i) w 2* 2 (i) Pah#1 w 2* K (i) Pah#2 Σ Pah#3 w 3* 1 (i) w 3* 2 (i) w 3* K (i) Weigh-per-User Type Weigh-per-Pah Type
Sub-Carrier#l Sub-Carrier#1 w *1 1 (i) w *l 1 (i) Sub-Carrier#L w*l 1 (i) FFT w *1 2 (i) w *l 2 (i) w*l 2 (i) Σ Sub-Carrier Oupu#1 Σ FFT FFT Σ Sub-Carrier Oupu#l w * 1 (i) w * 2 (i) w* K (i) w *1 K (i) FFT w *l K (i) w*l K (i) Σ Sub-Carrier Oupu#L Pre-FFT Type Pos-FFT Type
Subcarrier Arrangemen 0 5 6 1112 24 25 26 31 32 33 38 39 40 52 53 54 58 59 63frequency Daa Subcarrier Pilo Subcarrier Virual Subcarrier Signal Burs Forma 80 samples 80 samples 160 samples OFDM symbols Paern A Paern B Paern C Payload (Daa) for AGC for FFT Window Synchronizaion for Subcarrier Recovery and Array Weigh Conrol
Array Weigh Conrol Principle Simple Null Seering for Doppler-Shifed Signals Based on Observaion of Virual Subcarrier Oupus frequency v 1 =0 v 2 =0 v 3 =0 5 32 59 When received signals conain no Doppler-shifed signals. frequency v 1 0 =0 v 2 0 =0 v 3 0 =0 5 32 59 When received signals conain some no Doppler-shifed signals.
Weigh Conrol Crierion Number of Virual Subcarriers Considered Array Weigh Vecor 3 minimize e v (i)= Σ 0 w H (i)v j 2 (i) j=1 Desired v j Response (i): he j-h Virual Subcarrier Oupu Vecor v j (i)=[v j 1 (i),..., vj K (i)]t LMS Algorihm Paern C Payload ime This algorihm is workable in payload par, because i requires no pilo signal! Sar of Algorihm Down-conversion requires 64 samples o give he Algorihm he firs virual subcarrier oupu.
Numerical Resuls by Compuer Simulaion Envelope: Rayleigh-disribued wih he same average power Arrival Time: uniformly-disribued wihin guard inerval 8-Elemen Circular Anenna Array Anenna Elemen DOA: uniformly-disribued in (0deg, 360deg] λ/2 λ: Wavelengh Received Signal Number of Subcarriers Modulaion/Deecion FFT Lengh/Guard Inerval Lengh Forward Error Correcion Over-sampling Facor Number of Virual Subcarrier Used 52 (including 4 Pilo Subcarriers) QPSK/Coheren Deecion 64[samples]/16[samples] Convoluional Encoding/ Vierbi Decoding (R=1/2, K=7) 4 3 (#5, #32, #59)
2 Received Signals (f D =0[%], f D =3[%]) µ=0.001 µ=0.01 f D =3[%] 1 w/o Adapive Array Anenna µ=0.1-1 w/ Null-Seering (Preamble) BER -2 w/ Null-Seering (Preamble+Daa) -3 f D =0[%] -4 µ=0.01 0 5 15 20 25 Average E b /N 0 per Signal [db]
BER agains Normalized Doppler Shif 2 Received Signals f D =x[%] 1 Average E b /N 0 per Signal=20[dB] f D =0[%] BER -1-2 µ=0.01 0 1 2 3 4 5 6 7 8 9 Normalized Doppler Shif (x) [%] -3 Larger ISI Higher deecabiliy w/ Null-Seering (Preamble)
f D =0[%], f D =0[%], f D =0[%], f D =0[%], 8 Received Signals ( f D =1[%], f D =3[%], f D =5[%], f D =7[%] ) f D =7[%] f D =0[%] f D =0[%] f D =1[%] f D =0[%] f D =5[%] f D =3[%] f D =0[%] BER 1-1 -2-3 w/o Adapive Array Anenna w/ Null-Seering (Preamble+Daa) w/ Null-Seering (Preamble) -4 µ=0.01 0 5 15 20 25 Average E b /N 0 per Signal [db]
BER agains Number of Received Signals Toal Number of Received Signals 0 2 4 6 8 1 Average E b /N 0 per Signal=20[dB] BER -1-2 w/o Adapive Array Anenna w/ Null-Seering (Preamble) w/ Null-Seering (Preamble+Daa) -3 µ=0.01 0 1 2 3 4 5 Number of Received Signalswihou Doppler Shifs
Discussions * Longer observaion of virual subcarrier oupus does no always resul in beer BER performance. w=[0,..., 0] T is a soluion! * Diversiy gain is no obained, when he number of received signals wihou Doppler shifs increases. This mehod jus seers nulls oward undesired signals, namely, i ries o suppress Doppler-shifed signals bu does no ry o effecively cach desired signals. * Can we always receive signals wihou Doppler shifs? No!
Suppression of Delayed Signals beyond Guard Inerval S 1 u 1 S 2 S 3 Channel Impulse Response Guard Inerval u 2 u L The BER performance depends on y(i). w 1 w 2 w L Weigh Conroller r(i): Reference Signal d(i): Pilo Signal y(i): Impulse Response - w H u + OFDM Demodulaor r(i) = y(i) d(i)
How can we deermine y(i)? Guard Inerval S 3 cu! Combining Diversiy Effec Frequency Selecive Fading Guard Inerval S 2 S 3 S 1 cu! Selecion Diversiy Effec Frequency Non-Selecive Fading S 2 S 1
Number of Subcarriers Modulaion/Deecion FFT Lengh/Guard Inerval Lengh Forward Error Correcion Over-sampling Facor 52 (including 4 Pilo Subcarriers) QPSK/Coheren Deecion 64[samples]/16[samples] Convoluional Encoding/ Vierbi Decoding (R=1/2, K=7) 4 80 samples 80 samples 160 samples OFDM symbols Paern A Paern B Paern C Payload (Daa) for AGC for FFT Window for Subcarrier Recovery Synchronizaion and and Array Weigh Conrol Channel Esimaion
Channel Esimaion We canno apply he well-known maximum lengh shif-regiser code (2 N -1: N=number of sages) o 80 sample-long Paern B, which is composed of 52 subcarriers. We use a chaoic random sequence generaion mehod. The logisic map x n+1 =4x n (1-x n ) generaes a random sequence uniformly disribued in [0, 1.0) wih infinie lengh.
PN Sequence in Frequency Domain x n+1 =4x n (1-x n ) Q Q I I Q #1 #2 #52 64 Poin-IFFT PN Sequence in Time Domain I frequency 80 samples 64 samples 16 samples ime Guard Inerval Inserion
Auo-Correlaion Propery of Generaed Sequence Normalized Auo-Correlaion (Absolue Value) 1.0 0.8 0.6 0.4 0.2 Chaoic Random Sequence Generaion Guard Inerval Lengh 0 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 Lag [samples]
Impulse Responses for Reference Signal Generaion h 1 h 2 h 1 1 h 2 1 h 2 1 h2 2 h h L 2 1 h L L h 1 M h 2 M h M L u 1 u 2 u L w 1 w 2 w L h l =[h l 1,..., hl M ] T (l-h Impulse Response Vecor), w=[w 1,..., w L ] T (Array Weigh Vecor), H=[h 1,..., h L ] (Impulse Response Marix), L: Number of Anenna Elemens, M: Number of Samples in Guard Inerval.
Mehod A: L Elemen-Based Opimum Combining Any realizable impulse response, which can be used for reference signal generaion, can be wrien as a weighed sum of L impulse response vecors: h ref = [h ref 1,..., h ref M ] T = Hw. If OFDM prefers frequency selecive fading channels, a reference signal wih maximum oal power wihin guard inerval may give a beer performance.
Signal-o-Noise Power Raio (SNR) a array oupu is given by σ 2 x wh R H w SNR= σ 2 n wh w. R H =E[H H H]: Channel Correlaion Marix, σ 2 x : Signal Power, σ2 n : Noise Power. We can find w max (= arg max SNR ) by solving an equaion: (R H -p max )w max =0, p max : The larges eigenvalue of R H, w max : The eigenvecor corresponding o he eigenvalue. Impulse Response by L Elemen-Based Opimum Combining is calculaed as: h ref L-op =Hw max.
Mehod B: L Elemen-Based One Signal Selecion h 1 h 2 h 1 1 h 2 1 h 2 1 h2 2 h 1 M h 2 M h M L h h L 2 1 h L L Calculaion of Average Power If OFDM prefers frequency non-selecive fading channels... p av h ref L-slc p 2 p 1 1 p M [0,1,0,...,0] T Average Power Vecor Impulse Response by L Elemen-Based One Signal Selecion
Mehod C: 1 Elemen-Based Combining Arbirary Choice of an Anenna Elemen h h l 2 1 h l l h M l h 1 h 2 h 1 1 h 2 1 h 2 1 h2 2 h h L 2 1 h L L h 1 M h 2 M h M L h ref 1-cmb h l 1 h l 2 h M l Impulse Response by 1 Elemen-Based Combining
Mehod D: 1 Elemen-Based One Signal Selecion Arbirary Choice of an Anenna Elemen h 1 h 2 h 1 1 h 2 1 h 2 1 h2 2 h h L 2 1 h L L h 1 M h 2 M h M L h h l 2 1 h l l h M l h ref 1-slc 1 [0,0,0,...,1] T Impulse Response by 1 Elemen-Based One Signal Selecion
Array Weigh Conrol Mehods (Summary) Mehod A: h L-op ref Mehod B: h L-slc ref Mehod C: h 1-cmb ref Mehod D: h 1-slc ref L-Anenna Elemens 1-Anenna Elemens Combining Selecion
Numerical Resuls (by Compuer Simulaion) 8 Elemen-Circular Anenna Array (L=8) Envelope: Rayleigh disribued Received Signal Arrival Time: uniformly disribued Anenna Elemen wihin guard inerval λ/2 λ: Wavelengh DOA DOA: uniformly disribued in [0deg, 360deg)
One Signal beyond Guard Inerval (Average E b /N 0 per Signal=3 [db], DOA > 0 [deg]) Inerleaving Deph =1Subcarriers (No Inerleaving) BER Inerleaving Deph =4Subcarriers BER -2-3 -4-5 -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval -2-3 -4-5 Selecion Combining Selecion Combining -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval Inerleaving Deph =3Subcarriers BER Inerleaving Deph =6Subcarriers BER -2-3 -4-5 -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval -2-3 Selecion -4-5 Selecion Combining Combining -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval
One Signal beyond Guard Inerval (Average E b /N 0 per Signal=3 [db], DOA > 30 [deg]) Inerleaving Deph =1Subcarriers (No Inerleaving) BER Inerleaving Deph =4Subcarriers BER -2-3 Combining -4 Inerleaving Deph -5 =3Subcarriers Selecion -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval -2 BER -3-4 Combining Inerleaving Deph -5 =6Subcarriers Selecion -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval BER -2-3 -4-5 Selecion -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval -2-3 -4 Combining Combining -5 Selecion -6 1 2 3 4 5 6 Number of Received Signals wihin Guard Inerval
Two Signals wihin Guard Inerval/One Signal beyond Guard Inerval 8 Elemen-Opimum Combining 1 Elemen-One Signal Selecion Combining Diversiy Effec Selecion Diversiy Effec Frequency Non-Selecive Fading Signal wihin Guard Inerval wih Larges Power Signal wihin Guard Inerval Signal beyond Guard Inerval
Four Signals wihin Guard Inerval/One Signal beyond Guard Inerval 8 Elemen-Opimum Combining 1 Elemen-One Signal Selecion Selecion Diversiy Effec Frequency Non-Selecive Fading Combining Diversiy Effec Loss of Received Signal Power Frequency Selecive Fading
Conclusions Suppression of Doppler-Shifed Signals Our proposed array works well. There are many ineresing applicaions based on his mehod. Suppression of Delayed Signals beyond Guard Inerval Our resuls sugges ha change of array conrol crierion can improve he BER according o DOA paern.
Publicaions [1] A.Nishikawa, Y.Hara and S.Hara, OFDM Adapive Array for Doppler-Shifed Wave Suppression, Technical Repor of IEICE, RCS-2000-113, pp.57-62, Oc. 2000. [2]A.Nishikawa, Y.Hara and S.Hara, A Sudy on OFDM Adapive Array in Mobile Communicaions, Technical Repor of IEICE, RCS-2000-232, pp.73-78, Mar. 2001. [3] A.Nishikawa, Y.Hara and S.Hara, An OFDM Adapive Array for Doppler-Shifed Wave Suppression, Proceedings of he 2001 IEICE General Conference, B-5-151, p.549, Mar. 2001. [4] S.Hara, A.Nishikawa and Y.Hara, A Novel OFDM Adapive Anenna Array for Delayed Signal and Doppler-Shifed Signal Suppression,' in Proceedings of IEEE Inernaional Conference on Communicaions (ICC) 2001, pp.2302-2306, Helsinki, Finland, 11-14 June 2001. [5] S.Hane, Y.Hara, and S.Hara, Selecive Signal Recepion for OFDM Adapive Array Anenna,' Technical Repor of IEICE, AP-2001-69, pp.35-41, Aug. 2001. [6] S.Hara, S.Hane and Y.Hara, Adapive Anenna Array for Reliable OFDM Transmission, Proceedings of 6h Inernaional OFDM Workshop, pp.1.1-1.4, Hamburg, Germany, 18-19 Sepember 2001. [7] S.Hara, S.Hane and Y.Hara, Does OFDM Really Prefer Frequency Selecive Fading Channels, Proceedings of 2001 Third Inernaional Workshop on Muli-Carrier Spread- Specrum (MCSS2001) and Relaed Topics, pp.1-4, Oberpfaffenhofen, Germany, 26-28 Sepember 2001.