Copyright Blind Selected Mapping Techniques for Space-Time Block Coded Filtered Single-Carrier Signals

Transcription

1 Blind Seleted Mapping Tehniques for Spae-Time Blok Coded Filtered Single-Carrier Signals IEEE VTS AWCS 6 6 August Tokyo, Japan Amnart Boonkaay Fumiyuki Adahi Wireless Signal roessing Researh Group Researh Organization of Eletrial Communiation (ROEC) Tohoku University Copyright 6 Aknowledgement: This work inludes a part of results of The researh and development proet for realization of the fifth-generation mobile ommuniations system (#55-9, April 6) ommissioned to Tohoku University by The Ministry of Internal Affairs and Communiations (MIC), Japan.

2 resentation outline Introdution Challenges of wireless ommuniations igh AR signal / SLM Side-information problem / blind SLM Multipath propagation hannel / SC-FDE/STTD AR problem in SC-FDE/STTD Researh obetive Transeiver system model Transmitter: SC-FDE + STBC enoding + FD-SLM or TD-SLM Reeiver: MMSE-FDE + STBC deoding + MLD erformane evaluation Conlusion & Future works Time-domain Frequeny-domain The following figures are used throughout this presentation: d d d D D D 6 August 6 AWCS Tokyo, Japan d 3 D 3 Copyright 6 Time-domain signal vetor (pulse) Analog waveform Frequeny-domain omponents (subarriers)

3 Introdution: igh AR signal [A] D. Faloner, IEEE Trans. Commun., vol.59, no.4, pp.54-6, Apr. [B] J. Joung et al., IEEE Commun. Surveys&Tutorials, vol.7, no., pp , Q 5. [C] S. Okuyama et al., in ro. VTC -Spring, Taipei, Taiwan, May. roblem: high peak-to-average power ratio (AR) signal Re{s(t)exp(f t)} AR - - max{ s( t) N N t s( t) Time index, t } peak average input V out (volt).5.5 igh-ar signal igh-power amplifier (A) V in (volt) easily distorted by non-linear amplifiation [A] high-spe amplifier onsumes muh power = energy effiieny (EE) degrades [B] Single-arrier (SC) signal has lower AR than OFDM signal appropriate for uplink transmission But AR inreases due to the following reasons igh-level data modulation [C] (e.g. 4QAM to 6QAM +. db) Copyright 6 output Re{s(t)exp(f t)} - - Time index, t 6 August 6 AWCS Tokyo, Japan 3

4 Introdution: igh AR signal [A] D. Faloner, IEEE Trans. Commun., vol.59, no.4, pp.54-6, Apr. [B] J. Joung et al., IEEE Commun. Surveys&Tutorials, vol.7, no., pp , Q 5. [C] S. Okuyama et al., in ro. VTC -Spring, Taipei, Taiwan, May. roblem: high peak-to-average power ratio (AR) signal Re{s(t)exp(f t)} AR - - max{ s( t) N N t s( t) Time index, t } peak average input V out (volt).5.5 igh-ar signal igh-power amplifier (A) V in (volt) easily distorted by non-linear amplifiation [A] high-spe amplifier onsumes muh power = energy effiieny (EE) degrades [B] Single-arrier (SC) signal has lower AR than OFDM signal appropriate for uplink transmission But AR inreases due to the following reasons igh-level data modulation [C] (e.g. 4QAM to 6QAM +. db) Conventional solution : Transmit filtering e.g. square-root raised osine (SRRC) filtering output Re{s(t)exp(f t)} - - Time index, t But it needs to equip with high roll-off fator (typially higher than.5) [C] spetrum effiieny (SE) degrades Roll-off fator = (high-ar, high-se) Copyright 6 Roll-off fator > (low-ar, low-se) A spetrum-effiient AR redution for SC signals is needed N (+)N 6 August 6 AWCS Tokyo, Japan 4

5 Introdution: igh AR signal / SLM [A] D. Faloner, IEEE Trans. Commun., vol.59, no.4, pp.54-6, Apr. [B] J. Joung et al., IEEE Commun. Surveys&Tutorials, vol.7, no., pp , Q 5. [D] A. Boonkaay et al., Wirel. Commun. Mob. Comput., early view, Jun. 6. roblem: high peak-to-average power ratio (AR) signal Re{s(t)exp(f t)} AR - - max{ s( t) N N t s( t) Time index, t } peak average Solution: seleted mapping (SLM) [D] input V out (volt).5.5 igh-ar signal igh-power amplifier (A) V in (volt) hase-rotating the original signal for getting low-ar waveform Frequeny-domain SLM (FD-SLM) rotating the ueny omponents Time-domain SLM (TD-SLM) rotating the -domain symbols d d d Codebook easily distorted by non-linear amplifiation [C] high-spe amplifier onsumes muh power = energy effiieny (EE) degrades [D] 6 August 6 AWCS Tokyo, Japan 5 output = diag{,,,} = diag{,-,-,} Copyright 6 d 3 Transmit blok hase rotation = diag{-,-,,} d d d d 3 Re{s(t)exp(f t)} - - Tx. filtering Time index, t TD-SLM, 3 patterns, 4 symbols hase-rotated signal (Low-AR signal)

6 Introdution: Blind SLM roblem: side-information transmission Original SLM Codebook = diag{,,,} = diag{,-,-,} = diag{-,-,,} Data Side-information (pattern number) Additional transmission bits -> spetrum effiieny (SE) degrades TD-SLM, 3 patterns, 4 symbols Equalization I Codebook = diag{,,,} = diag{,-,-,} = diag{-,-,,} Copyright 6 De-mapping Original (omplex-onugate) 6 August 6 AWCS Tokyo, Japan 6

7 Introdution: Blind SLM [E] A. Boonkaay et al., in ro. ICICS5, Singapore, De. 5. [F] A. Boonkaay et al., to be presented at VTC6-Fall, Montreal, Canada, Sept. 6. roblem: side-information transmission Original SLM = diag{,,,} = diag{,-,-,} = diag{-,-,,} Data Side-information (pattern number) Additional transmission bits -> spetrum effiieny (SE) degrades Solution: Blind SLM [E,F] Imaginary part - 6QAM, 6-path Rayleigh fading Avg. reeived E b /N = 3 db Real part Blind FD-SLM reeived signal Codebook Imaginary part - 6QAM, 6-path Rayleigh fading Avg. reeived E b /N = 3 db Real part TD-SLM, 3 patterns, 4 symbols = diag{,,,} = diag{,-,-,} = diag{-,-,,} (omplex-onugate) 6 August 6 AWCS Tokyo, Japan 7 Equalization I Codebook De-mapping Original Utilizing the fat that the reeived signal onstellation obtained from orret de-mapping and inorret de-mapping are largely different Classifying them by alulating the mean-square error from the nearest original onstellations Copyright 6 Blind TD-SLM reeived signal Remarks Blind FD-SLM binary phase pattern {, 8} Blind TD-SLM polyphase pattern {,, 4} : orret de-mapping : inorret de-mapping

8 Introdution: SC-FDE/STTD [G] D. Faloner et al., IEEE Commun. Mag., vol.4, no.4, pp.58-66, Apr.. [] S. M. Alamouti, IEEE J. Sel. Areas. Commun., vol.6, no.8, pp , Ot nsmitter roblem: multipath propagation hannel Transmit omponents obstales reeiver Channel gain (f) (db) - - transmitter L6-path uniform power delay profile t l(l )ns f (Mz) obstales reeiver (f) (db) - - Reeived omponents L6-path uniform power delay profile t l (l )ns f (Mz) Bit-error rate (BER) degrades due to ueny-seletive fading Solution: SC with -domain equalization (SC-FDE) & spae- blok oded transmit diversity (STTD) SC-FDE [G] Effetive fading mitigation, obtaining ueny diversity gain STTD [] Utilizing multi-input multi-output (MIMO) transmission for obtaining spatial diversity gain D D STBC enoder spae D D D D Copyright 6 R, R, R, R, FDE & STBC deoder ˆD ˆD transmit antennas, reeive antennas 6 August 6 AWCS Tokyo, Japan 8

9 AR problem in SC-FDE/STTD [E] A. Boonkaay et al., in ro. ICICS5, Singapore, De. 5. [F] A. Boonkaay et al., to be presented at VTC6-Fall, Montreal, Canada, Sept. 6. [I] E. S. assen et al., IET Commun., vol.3, Iss., pp , Ot. 9. AR harateristi in SC-FDE/STTD D D STBC enoding & Transmit filtering IFFT IFFT Transmitter with N t transmit antennas Antenna # Antenna # revious works [E], [F] onsidered only single-antenna transmission Re{s(t)exp(f t)} Re{s(t)exp(f t)} transmit antennas, Square-root raised osine (SRRC) filter, roll-off fator = D D igh peak Time index, t igh peak Time index, t D D AR inreases, similar to single-antenna ase Low-AR waveform design is neessary for uplink SC-FDE/STTD Side-information transmission needs to be eliminated, espeially in MIMO transmission sine side-information inreases [I] Blind SLM tehniques for SC-FDE/STTD should be determined Copyright 6 - roblems -. Where to applied the phase rotation? Before STBC or after STBC enoding?. ow to selet the phase rotation pattern? 3. ow to detet the signal without side-information? 6 August 6 AWCS Tokyo, Japan 9

10 Researh obetive To introdue the blind SLM tehniques for SC-FDE/STTD We are going to propose the following transeiver models (assuming point-to-point MIMO transmission) Transmitter : Single-arrier + blind SLM + STBC enoding Blind SLM tehniques for SC-FDE/STTD hase rotation is applied before STBC enoding Reeiver : Joint MMSE-FDE & STBC deoding + MLD Joint MMSE-FDE & STBC deoding obtaining ueny & spatial diversity gain MLD alulates the mean-square error and selets the reeived symbols obtained from orret de-mapping no side-information transmission Real part 6 August 6 AWCS Tokyo, Japan Imaginary part - D D 6QAM, 6-path Rayleigh fading Avg. reeived E b /N = 3 db Copyright 6 STBC enoder Blind FD-SLM reeived signal Components of D and D spae Imaginary part - D D D D 6QAM, 6-path Rayleigh fading Avg. reeived E b /N = 3 db Real part Blind TD-SLM reeived signal

11 Transeiver system model: transmitter ( antennas) SC-FDE/STTD with blind FD-SLM Info. bits SC-FDE/STTD with blind TD-SLM Info. bits Data modulation Data modulation () d d D D hase rotation hase rotation FD-SLM module (pattern seletion) û û () STBC enoding & transmit filtering () STBC enoding & transmit filtering Copyright 6 I I I I +C +C +C +C # # # # () TD-SLM module (pattern seletion) 6 August 6 AWCS Tokyo, Japan

12 Transeiver system model: transmitter ( antennas) SC-FDE/STTD with blind FD-SLM Info. bits SC-FDE/STTD with blind TD-SLM Info. bits Data modulation Data modulation () () d d D D x x hase rotation TD-SLM module (pattern seletion) X X hase rotation FD-SLM module (pattern seletion) û û () SLM algorithm () Construt a phase rotation matrix, then multiply the phase rotation matrix to the transmit bloks before STBC enoding STBC enoding & transmit filtering e.g. transmit antennas uˆ () STBC enoding & transmit filtering pattern for d or D u I I I I +C +C hase rotation pattern seletion riterion. Minimax riterion Selet a pattern whih minimizes the maximum AR among J transmit bloks For FD-SLM: uˆ For TD-SLM uˆ ˆ, uˆ, pattern for d or D u~ U ~ J +C +C # # U = no. of phase rotation patterns J = no. of input transmit bloks of STBC enoding T = Transmit filtering arg min ( max (AR( FN T u, F Copyright 6 arg min ( max (AR( FN TFN u, d u~ U ~ J N d ))) ))) # # 6 August 6 AWCS Tokyo, Japan

13 Transeiver system model: transmitter ( antennas) SC-FDE/STTD with blind FD-SLM Info. bits SC-FDE/STTD with blind TD-SLM Info. bits Data modulation Data modulation () () d d D D x x hase rotation TD-SLM module (pattern seletion) X X hase rotation FD-SLM module (pattern seletion) û û () SLM algorithm () Construt a phase rotation matrix, then multiply the phase rotation matrix to the transmit bloks before STBC enoding STBC enoding & transmit filtering e.g. transmit antennas uˆ () STBC enoding & transmit filtering pattern for d or D u I I I I +C +C hase rotation pattern seletion riterion. Blok-by-blok minimization riterion Selet a pattern whih minimizes the AR of eah transmit blok For FD-SLM: uˆ, For TD-SLM uˆ, ˆ, uˆ, pattern for d or D arg min (AR( FN Tu, F Copyright 6 u~ U u~ U +C +C # # U = no. of phase rotation patterns J = no. of input transmit bloks of STBC enoding T = Transmit filtering arg min (AR( FN TFN u, d N d )) )) # # 6 August 6 AWCS Tokyo, Japan 3

14 Transeiver system model: transmitter ( antennas) [] S. M. Alamouti, IEEE J. Sel. Areas. Commun., vol.6, no.8, pp , Ot [J] G. Ganesan et al., IEEE Trans. Inform. Theory, vol.48, no., pp , Apr.. SC-FDE/STTD with blind FD-SLM () STBC enoding [,J] X X X X X X S N SC-FDE/STTD t 4 with blind X TD-SLM X X X X X Info. bits Data modulation Data modulation d d D D R N t J Q N STBC r =J/Q x x Note: hase rotation X X Simple enoding, employing only omplex onugate and blok exhange operations Info. bits e.g. ~4 transmit antennas SN t X X X X () SN t Relationship of N t and STBC oding rate X X X hase rotation 3 X X X X X X FD-SLM module (pattern seletion) arbitrary û no SE degradation û if N t = () STBC enoding & transmit filtering () STBC enoding & transmit filtering N t antennas X X X X Q slots X X X Copyright 6 N t antennas Q slots X I I I I +C +C +C +C # # # # () 4 3 4TD-SLM module.75 (pattern seletion) 6 August 6 AWCS Tokyo, Japan 4

15 Transeiver system model: reeiver Reeiver for blind FD-SLM # # -C -C R, R, R, R, Q slots Reeiver for blind TD-SLM # # -C -C R, R, R, R, Q slots N r antennas N r antennas (3) Joint MMSE-FDE & STTD deoding (3) Joint MMSE-FDE & STTD deoding ˆX I ˆx ˆX ˆx,, U,,, Copyright 6 U, I I I MLD MLD Rx. bits (4) Rx. bits (4) 6 August 6 AWCS Tokyo, Japan 5

16 Transeiver system model: reeiver [J]. Tomeba et al., IEICE Trans. Commun., vol.e9-b, no.3, pp , Mar. 7. Reeiver for blind FD-SLM # # # # -C -C -C -C R, R, R, R, Q slots Reeiver for blind TD-SLM R, R, R, R, Q slots N r antennas N r antennas (3) Joint MMSE-FDE & STTD deoding (3) Joint MMSE-FDE & STTD deoding (3) Joint MMSE-FDE & STBC deoding [J] ˆX ˆX I ˆx R, R, N t ~ ~ R, R,, MMSE-FDE mitigating the effet from ueny-seletive fading R ~ ( N MMSE-FDE weight: W n, n r t t Q) ( k) (, U, N N r t n n r t W n, n, I I I STBC deoding obtaining spatial diversity gain Xˆ ~ ~ ˆx r nr, nt nr, nt nr, nt nr nt N r Nt, n, W, r nt nr nt nr, nt nr nt spatial diversity gain U, ( N N But hase rotation regarding to SLM needs to be eliminated t n, n r ( k) ( k) ) / N t t t ( E s r ) / N ) R Nr N t W Wn, r nt Copyright 6 W n, n r t MLD MLD ( N r Rx. bits Q) (4) X Rx. bits X (4) 6 August 6 AWCS Tokyo, Japan 6

17 Transeiver system model: reeiver Reeiver (4) MLD for blind FD-SLM R, R Applying all possible de-mapping patterns, # # -C -C orret de-mapping by alulating the meansquare error (MSE) Corret de-mapping R, low R, MSE MLD is done for eah reeived blok # # Classifying the reeived symbols obtained from For FD-SLM: dˆ -C -C Q slots Reeiver for blind TD-SLM For TD-SLM: dˆ Inorret de-mapping high MSE arg arg min v~ U d mod min v~ U d F R, R, R, R v, FN X, N v, ICopyright mod De-mapping is done after Q slots Xˆ De-mapping is done before I ˆ spae d d spae (3) Joint MMSE-FDE & STTD deoding (3) Joint MMSE-FDE & STTD deoding ˆX I ˆx ˆX ˆx,, U,,, 6 U, I I I MLD MLD Rx. bits (4) Rx. bits (4) 6 August 6 AWCS Tokyo, Japan 7

18 erformane evaluation Simulation parameters Transmitter SLM module Channel Reeiver erformane indiators Data modulation Single-arrier, 6QAM No. of subarriers N = 56 C length N g = 6 Transmit filtering Square-root raised osine, = Blind FD-SLM Random binary hase-rotation sequene type {, 8} Blind TD-SLM Random polyphase ( ) {,, 4} No. of phase rotation patterns U = (no SLM) 5 Oversampling fator V = 8 Fading ower delay profile Channel estimation FDE AR.% -> AR value whih gives prob(ar > AR.% ) = -3 Bit-error rate (BER), assuming no hannel oding Frequeny-seletive blok Rayleigh Symbol-spaed, 6-path uniform Ideal MMSE-FDE ( ) The seleted set of polyphase {,, 4} is not optimal but suffiient for allowing blind TD-SLM Copyright 6 6 August 6 AWCS Tokyo, Japan 8

19 erformane evaluation: AR performane [D] A. Boonkaay et al., Wirel. Commun. Mob. Comput., early view, Jun. 6. AR.% (db) N t = N t = SC-FDE, 6QAM, N = 56, SRRC filtering (=) STBC N t =3 Minimax Blind FD-SLM Blind TD-SLM Blok-by-blok minimization No. of phase rotation patterns, U AR redues when U inreases (both FD- SLM and TD-SLM) Inreasing the probability to obtain low- AR waveform andidate TD-SLM ahieves lower AR than FD-SLM TD-SLM output waveform samples are in a fixed set [D] Minimax riterion: AR inreases when N t inreases due to lower degree of freedom Blok-by-blok minimization riterion: Copyright 6 AR does not hange when N t inreases 6 August 6 AWCS Tokyo, Japan 9

20 erformane evaluation: AR performane [E] A. Boonkaay et al., Wirel. Commun. Mob. Comput., early view, Jun. 6. AR.% (db) N t = N t = SC-FDE, 6QAM, N = 56, SRRC filtering (=) STBC N t =3 Minimax Blind FD-SLM Blind TD-SLM Blok-by-blok minimization No. of phase rotation patterns, U AR an be redued up to 3. db regardless of the number of transmit antennas AR redues when U inreases (both FD- SLM and TD-SLM) Inreasing the probability to obtain low- AR waveform andidate TD-SLM ahieves lower AR than FD-SLM.5 db TD-SLM output waveform samples are in a fixed set [E] Minimax riterion:.4 db AR inreases when N t inreases due to lower degree of freedom Blok-by-blok minimization riterion: AR does not hange when N t inreases AR.% redution at U=5, N t = Copyright 6 FD-SLM TD-SLM Minimax -.3 db -.7 db Blok-by-blok minimization -.8 db -3. db 6 August 6 AWCS Tokyo, Japan

21 erformane evaluation: BER performane Average BER (N t, N r ) = (, ) blind FD-SLM U= Average reeived E b /N per antenna (db) No hannel oding Blind FD-SLM, random binary phase rotation w/ side-information U=56 SC-FDE/STTD, N = 56, N g = 6, 6-path uniform Rayleigh fading, 6QAM modulation (N t, N r ) = (, ) (N t, N r ) = (, ) Average BER (N t, N r ) = (, ) blind TD-SLM U=56 w/ side-information U=56 BER improves when either N t or N r inreases due to an inreasing of spatial diversity gain Low E b /N region: BER of blind SLM degrades Effet from residual ISI and noise igh E b /N region: no signifiant BER degradation Blind TD-SLM, random polyphase rotation SC-FDE/STTD, N = 56, N g = 6, 6-path uniform Rayleigh fading, 6QAM modulation (N t, N r ) = (, ) (N t, N r ) = (, ) Average reeived E b /N per antenna (db) Copyright 6 6 August 6 AWCS Tokyo, Japan

22 Conlusion We extended the blind SLM tehniques to SC-FDE/STTD. FD-SLM and TD-SLM for STBC enoded signal transmission hase rotation is applied before STBC enoding riterions for phase rotation pattern seletion: minimax and blok-by-blok minimization. erformane evaluation by omputer simulation AR an be redued up to 3. db (independent from the number of transmit antennas) No BER degradation when reeived E b /N is suffiiently high eak throughput is exatly the same as the onventional SC-FDE/STTD sine there is no side-information sharing 3. Future works: omplexity redution of blind SLM signal detetion and performane evaluation, et. Copyright 6 6 August 6 AWCS Tokyo, Japan

23 Thank you - Q & A - IEEE VTS AWCS 6 6 August Tokyo, Japan Amnart Boonkaay Fumiyuki Adahi Wireless Signal roessing Researh Group Researh Organization of Eletrial Communiation (ROEC) Tohoku University Copyright 6 Aknowledgement: This work inludes a part of results of The researh and development proet for realization of the fifth-generation mobile ommuniations system (#55-9, April 6) ommissioned to Tohoku University by The Ministry of Internal Affairs and Communiations (MIC), Japan.