Enhanced HARQ Technique Using Self-Interference Cancellation Coding (SICC)

Transcription

1 MITUBIHI ELECTRIC REEARCH LABORATORIE ttp:// Enanced HARQ Tecnique Using elf-interference Cancellation Coding (ICC) Wataru Matsumoto, Tosiyuki Kuze, igeru Ucida, Yosida Hideo, Pilip Orlik, Andreas Molisc, Zifeng Tao, Jinyun Zang TR July 008 Abstract Te paper provides a metod for combining HARQ along wit elf-interference Cancellation Coding (ICC), so tat te reliability of spatial multiplexing MIMO transmissions can be improved. Te simulation results sow tat significant gain is acieved over te traditional Case Combining. Tis work may not be copied or reproduced in wole or in part for any commercial purpose. Permission to copy in wole or in part witout payment of fee is granted for nonprofit educational and researc purposes provided tat all suc wole or partial copies include te following: a notice tat suc copying is by permission of Mitsubisi Electric Researc Laboratories, Inc.; an acknowledgment of te autors and individual contributions to te work; and all applicable portions of te copyrigt notice. Copying, reproduction, or republising for any oter purpose sall require a license wit payment of fee to Mitsubisi Electric Researc Laboratories, Inc. All rigts reserved. Copyrigt c Mitsubisi Electric Researc Laboratories, Inc., Broadway, Cambridge, Massacusetts 09

2 MERLCoverPageide

3 Enanced HARQ tecnique using elf-interference Cancellation Coding(ICC) Wataru Matsumoto, Tosiyuki Kuze, igeru Ucida, Yosida Hideo Information Tecnology R & D Center Mitsubisi Electric Corporation 5-- Ofuna Kamakura, Kanagawa 47850, JAPAN Matsumoto.Wataru@aj.MitsubisiElectric.co.jp Pilip V. Orlik, Andreas F. Molisc, Zifeng (Jeff) Tao, Jinyun Zang Mitsubisi Electric Researc Laboratories 0 Broadway Cambridge, MA 09 { porlik, molisc, tao, jzang}@merl.com Abstract Te paper provides a metod for combining HARQ along wit elf-interference Cancellation Coding (ICC), so tat te reliability of spatial multiplexing MIMO transmissions can be improved. Te simulation results sow tat significant gain is acieved over te traditional Case Combing. Keywords-MIMO, HARQ, elf Interference Cancellation I. INTRODUCTION In current and evolving mobile cellular communication systems, te use of MIMO transmission tecnology is becoming more widespread. Te Worldwide Interoperability for Microwave Access (WiMAX) forum as well as rd Generation Partnersip Project (GPP) ave released standard specifications tat make use of MIMO and continue to enance te tecnology to improve transmission capacity and reliability. MIMO systems increase capacity by transmitting multiple data symbols over several antennas simultaneously, in a tecnique usually termed patial Multiplexing (M). M is a transmission tecnique in MIMO to transmit different data signals, so called streams, from eac of te multiple transmission antennas.a MIMO receiver can use advanced signal processing and te properties of te cannel to separate out and decode te individual symbols. A tecnique to improve reliability is termed pace Time Block Coding (TBC), in wic a MIMO system transmits copies of te data symbols from multiple antennas. Te IEEE 80.6 standard Part 6: Air interface for Broadband Wireless Access ystems 80.6[], upon wic WiMAX is based, employs bot M and TBC tecniques. In addition to MIMO, tese newer standards also make use of Hybrid Automatic Repeat request (HARQ). As in traditional Automatic Repeat request (ARQ), a receiver requests a retransmission of a message tat failed to be decoded correctly, but wit HARQ rater tan discarding te original (corrupted) message, te receive retains a copy and combines te original message [],[]. Currently, two types of HARQ are widely used: (i) repetition coding, i.e., te transmitter repeats exactly te same signals tat were previously transmitted; te receiver adds up te received signals and acieves an improvement in te signal-to-noise ratio troug noise averaging; (ii) incremental redundancy encoding, were te transmitter sends different parity-ceck bits during eac retransmission. Bot HARQ scemes ave drawbacks. Incremental-redundancy metods require more complicated decoders, wile repetition coding sows poor performance wen applied to spatial-multiplexing systems. Tis is due to te self-interference between spatial streams, and te absence of diversity in time-invariant cannels. In te current paper, we propose an alternative tat offers a simple decoding sceme as well as excellent performance in MIMO systems. Our sceme is a combination of HARQ wit space-time block coding. Te retransmitted signal is a spacetime encoded version of te original signal tat allows elimination of self-interference troug simple linear receivers and (for some of te proposed scemes) offers enanced diversity as well. Te simplest form of our metod occurs in a x MIMO sceme: during a first transmission, we just transmit two symbols from te two TX antennas. If an HARQ transmission is necessary, we transmit te same symbol from antenna, but te 80 degree pase sifted symbol from antenna. Tus, wen te receiver adds up te signals of te first and second transmission, te contribution of symbol cancels out (and similarly, te contribution of symbol can be eliminated by subtracting te signals). Te simulation results sow significant gain is acieved over te traditional Case Combining Te idea of combining HARQ wit space-time codes was treated for convolutional codes in [4], and for turbo codes in [5], toug none of tose papers discusses space-time block codes. Tarok [] introduced an HARQ sceme for x MIMO systems based on Alamouti TBC codes tat gives diversity benefits as well as self-interference cancellation. Tis sceme was adopted on te IEEE80.6e standard []. In tis paper, we compare te ICC-based approac to []; we also give generalizations to larger MIMO arrays and sow ow te ICC approac can be combined wit TBCs. Te remainder of te paper is organized as follows: ection II presents te system model. Our new sceme is described for x MIMO in ection III; we detail various encoding matrices tat can be used wit our sceme. Generalizations to larger MIMO systems are presented in ec. IV, followed by simulation results and conclusions /08/$ IEEE 650 Autorized licensed use limited to: Polytecnic Inst of New York Univ. Downloaded on eptember, 009 at 4:4 from IEEE Xplore. Restrictions apply.

4 II. MIMO WITH PATIAL MULTIPLEXING AND HARQ A. MIMO-OFDM wit patial Multiplexing Figure sows a block diagram of a MIMO-OFDM transmitter wit transmit antennas. Te transmitter consists of a source of modulated data symbols, a space-time encoder, and two OFDM cains tat consist of an OFDM modulator tat performs an IFFT on te input data and ten filters, amplifies and upconverts te time domain signal to te pass-band in te RF block. Te exact nature of te space-time encoder determines te type of MIMO transmission. In te case of spatial multiplexing (M), different bits are mapped onto te two transmit antennas, tus increasing te spectral efficiency. For example, in te case of vertical encoding, two consecutive symbols and are transmitted during one cannel use, since is transmitted on antenna and is transmitted on antenna. A receiver of te signal typically needs to ave te same or more receive antennas to enable te separation and decoding of te symbols. Many receiver types ave been developed in te literature, including te optimal (Maximum Likeliood) detector as well as suboptimal receivers suc as te Minimum Mean quare Error (MME) and Zero Forcing (ZF) receivers [6]. in te estimation of and can implement decoding scemes suc as Minimum Mean quared Error (MME) or Zero Forcing (ZF) to estimate. We see tat te terms, s and, s are te interference terms at receive antenna from transmit antenna and receive antenna from transmit antenna respectively. Tis type of interference is typically called self interference, since it is due to te transmission of multiple streams from multiple antennas. B. Case combining in MIMO-OFDM systems Even wit appropriate forward error correction coding and adaptive modulation, not all data packets arrive at te receiver error-free. Te receiver can find out weter coding failed, e.g., from CRC (cyclic redundancy ceck) bits or from te consistency messages passed in an LDPC (low-density parity ceck) decoder. In any case, if te decoding fails, te receiver can initiate an HARQ procedure, in wic te receiver retains a copy of R and sends retransmission request. In conventional case-combining, te transmitter ten sends an exact duplicate of te vector. Tx Antenna Rx Antenna Modulated symbols ) pace-time encoder IFFT RF Tx Antenna Rx Antenna Fig. MIMO cannel model Let us next compute te received signal. Te MIMO cannel seen by eac of te OFDM subcarriers is denoted as a x matrix,, H =, (),, were te element i,j is te cannel gain from te j t transmit antenna to te i t receive antenna as sown in Figure. We can write te received signal at te two antennas as r = r Figure: General MIMO OFDM transmitter,,,, n + n wic is equivalent, in matrix notation, to R = H+n, were n is an additive wite Gaussian noise vector and is te vector of transmitted signals. Under normal operation, te receiver operates on te vector R to estimate te transmitted vector. It is assumed te receiver also as knowledge of te cannel coefficients H, wic aides IFFT RF We denote te two successive transmissions as [ () () ] were () = () After reception of te retransmission te receiver as two copies of te signals R (), R (). Tese can be expressed as () () () () r r, s +, s + n, s +, s + n = () () (4) () () r r,s +,s + n,s +,s + n Were te term r j (i), represents te signal at te j t antenna element due to te i t transmission, and n j (i), is te noise at te j t antenna element associated wit te i t transmission. It sould be noted tat n j (i), {j =,, i =,} are all independent identically distributed Gaussian wit variance σ. Te receiver now as two copies of te data wic can be combined so as to improve te decoding probability. One common way to combine R () and R () is to simply average te two vectors to obtain ( ) ( ) ' R + R R = (5) Tis operation as te effect of reducing te noise variance/power by a factor of two and will aid in decoding. However, te self-interference discussed in ec. II.A is not improved by tis procedure. An analysis of te INR for a spatial multiplexing MIMO-OFDM system using an MME receiver and case combining is found in [7]. 65 Autorized licensed use limited to: Polytecnic Inst of New York Univ. Downloaded on eptember, 009 at 4:4 from IEEE Xplore. Restrictions apply.

5 III. HARQ WITH ICC FOR XMIMO Our proposed coding sceme tat can be used to eliminate te self interference after an HARQ transmission is one wic we term elf-interference Cancellation Coding (ICC). Tis is based on Hadamard matricesdiscrete Fourier Transform, or Alamouti coding and is simple to implement. We again consider a x system and denote = [ ] T as a vector of signals transmitted from te two antennas. Once again after te reception of te signal R = H + n and a decoding failure te HARQ process is initiated and a request for a retransmission is sent to te transmitter. However, te retransmission occurs in a sligtly different form, wic enables easy cancellation of te self-interference. We stress tat tere is a key difference between te abovementioned sceme and te conventional TBC sceme. In a conventional TC, te transmitter always sends te st and nd transmissions, and any receiver (single-antenna or multipleantenna) can decode te received signal. In te sceme discussed ere, te receiver as at least antennas, and often can decode te signal from te st transmission alone; te nd transmission is sent only if required. A. Hadamard matrix type Te retransmission is of te form: () =. (7) were in tis case te signal transmitted from te second antenna is simply sent wit a negative sign. At receiver, te signals for te original transmission along wit te HARQ retransmission can be expressed as (,) () () (,) R = H[ ] + n ( ) ( ) (9), +,,, (,) = + n,,, ( ), ( ) Te combining sceme for ICC begins wit te multiplication of te received matrix (R (,) ) by a x Hadamard matrix yielding, R (,)' (,) = R = + n ~ (0) were n ~ is again an iid Gaussian matrix wose entries ave (,) twice te variance of te entries of n. Tus we see tat te signal component of te matrix, R (,), contains two columns were te first column depends only on te signal and te second column depends only on te signal. We can combine te signals by multiplying te first column of R (,) by te vector () =[ ] T and te second column of R (,) by te vector () =[ ] T. Tis yields,, +,,,, + were () n = n ~ and n = n ~,, = (, = ( (), +, + ), ) Tus we see tat te ICC combining yields signals were te self interference as been eliminated. Tis Hadamard matrix m type ICC can be applied for m ( m ) MIMO systems. For te x case, if after te initial HARQ retransmission, (), te receiver still detects a decoding error on te signals, and ten te additional retransmission can be requested. Denoting (j) as te j t HARQ transmission at receiver, we ave (,, ) () () () (4) (, ) r = H[ ] + n () By applying te same sceme for ICC, we process te signals arriving at eac antenna wit a repeated Hadamard matrix (,, ) r Witin eac Hadamard repetition, te received signals are combined according to te metod described above; te results from tose repeated matrices are ten case-combined. It is noteworty tat in a time-invariant cannel, te repetitions of te Hadamdard matrix do not increase te diversity order, but just increase te NR by db for eac doubling of te number of repetitions. B. Discrete Fourier Transform type As a generalization of te ICC sceme, it is noted tat we can also describe te ICC sceme using discrete Fourier transform matrices. Te k t transmission symbol at t -t transmission antenna is given by multiplying st transmission π j ( t ) ( k ) NT symbol by e. Te st and nd transmissions are π j 0 0 () () = =, π = j 0 π j 0 π j (4) =, and at receiver, te signals for te original transmission along wit te HARQ retransmission can be expressed same as in ec. III.A. Te combining sceme for ICC begins wit te multiplication of te received matrix (R (,) ) by a x Inverse π j a b Discrete Fourier Transform matrix suc as NT e (for NT a( 0 a NT )-t row, b( 0 b N R ) t column) Tis representation is useful for MIMO systems in wic te number of transmit and receive antennas are greater tan two, and it enables extensions of te ICC combining sceme to tese systems. However, inorder to completely remove self-interference square MIMO systems are required. C. Alamouti coding type We note tat te ICC sceme is very similar to te pace Time Block Code (TBC) first presented by Alamouti [4]. wic also completely eliminates self-inference between antennas. Essentially, if we encode te transmitted signals 65 Autorized licensed use limited to: Polytecnic Inst of New York Univ. Downloaded on eptember, 009 at 4:4 from IEEE Xplore. Restrictions apply.

6 prior to transmission and use sligtly more complex (as compared to simple case combining) processing at te receiver, we effectively eliminate self interference and at te same time provide diversity benefits. Te st and nd transmissions are of te form: () ( ) () =, = (5) ( ) Te combining at te receiver is done as for standard Alamouti-decoders [8], after wic te receiver ten attempts to detect te transmitted signal. ince te combined signal no longer contains self interference te probability of correct detection increases. Tis sceme was proposed by Tarok [] and adopted into te Wimax specifications []. IV. ICC FOR GROUPING TC 4X4MIMO By combining te ICC and Alamouti TBC scemes we can acieve new MIMO space-time codes tat acieve te elimination of self interference for larger MIMO arrays wit ig rate. In te following discussion we assume 4 transmit antennas and 4 receive antennas. We combine te x Alamouti TC coding along wit te ICC coding to transmit te following sequence of vectors (5) = Here eac column of represents te symbols transmitted at eac transmission/retransmission interval. Te structure of te first columns of can be seen to be an Alamouti type code on te symbols and transmitted on antennas and, wile a second Alamouti type code on symbols and 4 transmitted on antennas and 4. Te next two columns repeat te Alamouti code owever te symbols on antennas and 4 ave been negated. Te advantage of tis sceme is tat after te transmission of te symbols in (5) a simple linear combining sceme can be employed at te receiver to eliminate te self interference. At te receiver, te following signal is received, r = H + n were r is a 4x4 matrix. For, te receiver sould combine te symbols from eac antenna as, [ ], r r r r + [ r r r r ] +, [ ], r r r r + [ r r r r ],,,,,,4,4,,,,,,, 4,,,,, 4, 4, 4, 4, (, +, +, +, +, +, + 4, + 4, ) + = n (7), 4,, 4,,4 4,4 similarly, for and 4 we obtain (, +, +, +, +, +, + 4, + 4, ) + n, (, +,4 +, +,4 +, +,4 + 4, + 4,4 ) + n, (, +,4 +, +,4 +, +,4 + 4, + 4,4 ) 4 + n4. (8) As we can see te combining yields 4 symbols tat contain no self-interference terms and tus simple detection scemes can be applied to estimate te transmitted symbols. V. IMULATION REULT Tis section evaluates te error rate performance at te nd transmission for bot retransmission metods te proposed HARQ wit Hadamard type and Alamouti coding type metod vs. te Case Combining (repetition and averaging). We consider a x MIMO system similar to WiMAX [] in wic eac antenna transmits an OFDM symbol consisting of 04 subcarriers. Te system is simulated over an WINNER Urban micro-cell [0] cannel model wit a mobile velocity of 0.5m/s. Modulation is assumed to be 6QAM on eac subcarrier and te WiMAX rate ½ convolutional turbo coder is used to protect data.m may be used under a low-mobility case, so we select Urban micro-cell in Winner cannel model [0] wit a mobile velocity 0.5m/s for te simulation. Figure. sows tat te proposed strategy ICC(Hadamard type and Alamouti coding type) as a gain over Case Combining(repeat) of about db and.5db (@BLER=0 - ), respectively, at te second retransmission. Figure. Performance comparisonfor te MIMO stream in a x M wit Urban micro-cell wit 0.5 m/s velocity and 6QAMat te nd transmission. Next, for a 4x4 case, we simulate our HARQ sceme from ec. IV using QPK modulation over a Rayleig fading cannel. We compare its performance to te sceme used in te IEEE 80.6 standard [], wic uses te following retransmission matrix 65 Autorized licensed use limited to: Polytecnic Inst of New York Univ. Downloaded on eptember, 009 at 4:4 from IEEE Xplore. Restrictions apply.

7 = (9). As we can see, te conventional sceme also employs Alamouti coding between antennas and, but after tat retransmits (witout modifications). Tis is in contrast to our sceme tat uses ICC for retransmissions and 4, and tus is able to cancel self-interference between streams / on one and, and streams /4 on te oter and. For te simulation of te conventional TC coding sceme, it is assumed tat received subpackets are case combined and ten a Zero-forcing MIMO receiver is employed after Alamouti combining. In tis case, te first and tird receptions are case combined and te second and fourt receptions are case combined. Alamouti decoding is ten performed on te signals at te first and second antennas as well as on te tird and fourt antennas. At tis point a Zero-forcing receiver can be employed to finally recover te symbols. Te bit error rates for te transmission scemes in equations (5) and (9) are sown in te Figure 4 and we see tat te proposed ICC+TBC sceme outperforms te existing sceme [] by about.5-db after 4 transmissions. Table. Comparison table for HARQ scemes No Hadama DFT Alamouti coding rd Performance 0dB db db 4.5dB (gain on (ref.) Fig..) Diversity order 0 4 Encoder complexity No coding Decoder complexity Remarks addition of sign(+,-) ame as Hadamard type for x MIMO. (6) (0) (0) for x. number No of restriction antenna for te is power number of of. antenna. Excange symbols and complex conju -gation see [8] Furtermore, a combining te ICC and TBC sceme is proposed to acieve new MIMO space-time codes tat acieve te elimination of self interference.tis proposed ICC+TBC sceme outperforms te existing sceme in section of [] by about.5-db over a Rayleig fading cannel. REFERENCE Figure. Performance comparison of 4x4 ICC+TBC sceme wit 4x4 80.6e sceme. VI. CONCLUION In tis paper a new HARQ metod using ICC is presented, wic extends and generalizes te ingenious sceme of Tarok []. Wen using HARQ in a MIMO system, tis new metod uses Hadamard matrices, Discrete Fourier Transform, and Alamouti coding to generate retransmission packets and performs a combining and cancellation using all te received symbols. Te proposed metod as significant gains over te conventional case combining HARQ scemes. A tradeoff in performance and complexity of te different scemes is possible, as sown in te Table. [] IEEE80.6e-005-IEEE tandard for Local and Metropolitan area networks Part 6: Air Interface for Fixed and Mobile Broadband Wireless Access ystems. Amendment : Pysical and Medium Access Control Layers for Combined Fixed and Mobile Operation in Licesed Bands and Corrigendum. [] D. Case, "A combined coding and modulation approac for communications over dispersive cannels," IEEE Trans. Commun., voi., no., Mar. 97. [] D. Case, "Code combining-a maximum-likeliood decoding approac for combining an arbitrary number of noisy packets", IEEE Transactions Communication, vol. COM-, No. 5, Marc 985, pp [4] A. Van Nguyen and M. A. Ingram, M.A.; Hybrid ARQ protocols using space-time codes, VTC 00 Fall pp.:64 68, 00. [5] Y. Q. Zou and J. Wang, Optimum sub-packet transmission for turbocoded ybrid ARQ systems, ICC 0 pp. 4 8, May 00. [6] A. F. Molisc, Wireless Communications, IEEE-Press Wiley, 005. [7]. Zang, et al, Proposed Text Modification in ection 4.7: PHY Abstraction for H-ARQ, IEEE C80.6m-07/89, Malaga, pain, eptember 007. [8].. M. Alamouti, A simple transmit diversity tecnique for wireless communications, elected Areas in Communications, IEEE Journal on Volume 6, Issue 8, Oct. 998 Page(s): [9] ITU-R recommendation M.5, Gaidelines for evaluation of radio transmission tecnologies for IMT-000, 997. [0] IT-WINNER II Deliverable D.. v.0, WINNER II Interim cannel Models, 0 December 006. [] V. Tarok, Autorized licensed use limited to: Polytecnic Inst of New York Univ. Downloaded on eptember, 009 at 4:4 from IEEE Xplore. Restrictions apply.