IST-2000-30148 I-METRA: D3.2 Implementation of relevant algorithms

IST-2000-30148 I-METRA D3.2 Implementation of Relevant algorithms Contractual Date of Delivery to the CEC: 31 October, 2002 Actual Date of Delivery to the CEC: 31 October, 2002 Author(s): Javier R Fonollosa, Roger Gaspa, Markku Heikkilä, Jean Philippe Kermoal, Kari Majonen, Xavier Mestre, Alba Pagès, Adam Pollard, Laurent Schumacher, Ami Wiesel, Juha Ylitalo Participant(s): UPC, AAU, NET, NMP, VOD Workpackage: WP3: Re-configurable Algorithm study Est. person months: 3 Security: Public Nature: Report Version: 0.1 Total number of pages: 33 Abstract: This deliverable provides a high level description of the software developed within the IMETRA project following the selection reported in D3.1 "Design, Analysis and Selection of Suitable Algorithms". Keyword list: Adaptive Antennas, MIMO Systems, Space-Time Coding, Adaptive Modulation and Coding, HSDPA, HSUPA, Turbo Coding. IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Design, Analysis and Selection of suitable algorithms 3 July, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 2 of 33 0 EXECUTIVE SUMMARY This deliverable provides a high level description of the software developed within the IMETRA project for the HSDPA and HSUPA of UMTS following the selection reported in D3.1 "Design, Analysis and Selection of Suitable Algorithms". With respect to the HSDPA services, only the high level description of link level simulations is presented. The transmitter follows the HS-DSCH specifications already reported in D3.1 in such a way that this deliverable is mainly focused to describe the software of the space-time block codes implemented in the transmitter as well as the reception schemes. The herein reported HSDPA transmission techniques include the transmit antenna switching technique, four different space-time block codes based on the STTD (i.e., CSTD, Trombi, DSTTD and DABBA), the well-known V-BLAST system and the more general Linear Dispersion Codes. The channel is implemented using the MIMO channel model developed in WP2. Regarding the receiver, several schemes are described: the Interference Cancelling and Nulling technique developed for V-BLAST, the LMMSE Space-Time Equalizer technique and different suboptimal ML receivers (such as, the “Turbo algorithm”). With respect to HSUPA, the simulation model includes full 3GPP features and has been designed for both a low-complexity UE (1 antenna UE) and a dual-antenna with MIMO capability. This approach allows a common reference point for advanced MIMO concept as well as realistic signal models and receiver implementation schemes. A standard Rake receiver structure is assumed but the model allows also the implementation of more complex receivers with interference suppression in a straight-forward manner. IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 3 of 33 DISCLAIMER The work associated with this report has been carried out in accordance with the highest technical standards and the I-METRA partners have endeavoured to achieve the degree of accuracy and reliability appropriate to the work in question. However since the partners have no control over the use to which the information contained within the report is to be put by any other party, any other such party shall be deemed to have satisfied itself as to the suitability and reliability of the information in relation to any particular use, purpose or application. Under no circumstances will any of the partners, their servants, employees or agents accept any liability whatsoever arising out of any error or inaccuracy contained in this report (or any further consolidation, summary, publication or dissemination of the information contained within this report) and/or the connected work and disclaim all liability for any loss, damage, expenses, claims or infringement of third party rights. IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 5 of 33 0 EXECUTIVE SUMMARY .......................................................................................................................... 2 1 INTRODUCTION ........................................................................................................................................ 6 2 HIGH SPEED DOWNLINK PACKET ACCESS (HSDPA) .................................................................... 7 2.1 HSDPA LINK-LEVEL SIMULATOR ............................................................................................................. 7 2.2 DIVERSITY MIMO TECHNIQUES ................................................................................................................ 9 2.2.1 STTD ................................................................................................................................................ 9 2.2.2 Circular Shifted Transmit Diversity (CSTD) ................................................................................. 10 2.2.3 Trombi ............................................................................................................................................ 11 2.2.4 Transmit Antenna Switching .......................................................................................................... 11 2.3 LAYERED MIMO TECHNIQUES ................................................................................................................ 12 2.3.1 DSTTD ........................................................................................................................................... 12 2.3.2 V-BLAST ....................................................................................................................................... 13 2.3.3 Transformed 4-Antenna Double-Rate Block Code (DABBA) ...................................................... 14 2.4 ADVANCED RECEIVER STRUCTURES ........................................................................................................ 15 2.4.1 Space-Time Equalizers ................................................................................................................... 15 2.4.2 Maximum Likelihood Receivers .................................................................................................... 17 2.5 COMBINATION OF LINEAR DISPERSION CODES WITH TURBO SPACE TIME DECODING. ............................ 19 3 HIGH SPEED UPLINK PACKET ACCESS (HSUPA) ......................................................................... 25 3.1 INTRODUCTION ......................................................................................................................................... 25 3.2 SIMULATION MODEL ................................................................................................................................. 25 3.2.1 3GPP compatibility ........................................................................................................................ 25 3.2.2 High-level modelling of uplink high-speed data with MIMO approach ........................................ 25 3.2.3 Signal structure ............................................................................................................................... 27 3.2.4 Transmitter structure ...................................................................................................................... 28 3.2.5 Receiver structure ........................................................................................................................... 29 3.2.6 MIMO channel properties .............................................................................................................. 30 3.2.7 Performance measure ..................................................................................................................... 31 3.2.8 Simulation cases ............................................................................................................................. 31 4 REFERENCES ........................................................................................................................................... 33 IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 6 of 33 1 INTRODUCTION [UPC] IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 7 of 33 2 HIGH SPEED DOWNLINK PACKET ACCESS (HSDPA) 2.1 HSDPA Link-Level Simulator Figure 1 illustrates the HSDPA link-level simulator which is applicable with most of the selected MIMO transmission schemes. The simulator is implemented in CoCentric System Studio and COSSAP environments. Transmitter Data bits are first grouped into packets which are then channel encoded. At the same time, the packet data is stored in a memory slot for possible later retransmissions. After puncturing (rate matching) and interleaving the signal is modulated using the selected modulation scheme (QPSK or 16-QAM). The symbols are either space-time encoded using the selected space-time coding (diversity) scheme or multiplexed into several parallel streams according to the chosen layered scheme. Each antenna symbol stream is spread using the selected number of spreading codes. At his phase, common pilot signal is added to the antenna signals as well as the selected number of interfering speech or HSDPA code channels. Channel METRA MIMO channel model generates the desired correlation between the antennas and simulates the desired multipath fading channel. Additive white Gaussian noise (AWGN) is used to model intercell interference. It is also possible to generate actual interfering base station (BS) signal(s). This is especially important when spacetime equalizing receivers are applied since they benefit from the structured interference compared to conventional receivers. Receiver A multiantenna receiver detects the data bits using channel estimates obtained using the common pilot channel. Only the complex channel gains are estimated while the channel delays are always assumed to be known. Estimation is performed by (moving-average) filtering the common pilot based raw channel estimates. One of the receiver algorithms described in the following sections is used. The receiver has perfect knowledge of the index of the packet it has received and, in case of a retransmitted packet, it combines the soft symbol estimates into a combiner memory slot. Chase-combining based hybrid-ARQ has been applied. Depuncturing operation (or inverse rate matching) restores the original code frame length for the channel decoder. Cyclic redundancy check (CRC) bits in the code frame are not used but the correctness or incorrectness of the decoded packet is detected ideally. This information is applied in generation of an ARQ command to the receiver. The transmission of ARQ is not simulated explicitly but the ACK/NACK message is given to the transmitter without errors. BS transmitter does not apply any scheduling but retransmits the requested packet after a fixed, parameterized delay. IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 8 of 33 The main parameters of the HSDPA link-level simulator are: • Channel coding rate (turbo code with puncturing or rate matching) • Modulation method (QPSK or 16-QAM) • Power and number of HSDPA code channels for the desired user • Power and number of interfering HSDPA code channels • Power and number of interfering speech users • Common pilot channel power • Maximum number of retransmissions per packet • Retransmission delay Various performance data is collected during the simulation including: • raw (non-decoded) bit error rate • decoded bit error rate • data throughput (kbit/s) as a function of both average geometry parameter (G) and maximum allowable number of packet retransmissions • residual average packet error rate (PER) as a function of both average geometry parameter (G) and maximum allowable number of retransmissions • distribution of packet-wise signal-to-noise ratio • PER as a function of packet-wise signal-to-noise ratio (only first transmission of each packet is included in this figure) The last performance figure is used for generating input for system simulations. IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 9 of 33 Figure 1. Structure of link-level HSDPA simulator. 2.2 Diversity MIMO Techniques Diversity MIMO techniques utilize the multiplicity of transmit antennas to obtain transmit diversity. In HSDPA, data rate and system capacity may increase indirectly through a lower packet error rate, higher supported modulation alphabet, or higher channel coding rate. 2.2.1 STTD In STTD transmission an orthogonal space-time block code is applied over two consecutive data symbols. STTD coding of symbols s1 and s2 is illustrated in Figure 2. Columns of the code matrix C(s1, s2) represent different symbol intervals and rows different transmit antennas. The code is used as a basis for several other multiantenna transmission schemes. Figure 2. Space-Time Transmit Diversity scheme for HSDPA. MIMO channel & AWGN Multiantenna modulator & spreading Bit source Packet gen. Retrans. memory Channel encoder Puncturing/ rate matching Interleaver Pilot signal gen. Interf. user gen. Multiantenna receiver Retrans. combiner memory Channel decoder Depuncturing/ rate matching Deinterl. ARQ command generator Performance evaluator Channel estimator ú û ù ê ë é - =** 12 21 21 ),( ss ss ssC IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 16 of 33 where s m2 is variance of dm(i) and the covariance matrix of the multiantenna signal is (2.9) where E(×) denotes expectation. The covariance matrix is estimated as a time average. It is also possible to use adaptive techniques to implement an LMMSE equalizer. Figure 9 shows the structure of the LMMSE space-time equalizer for the signal from transmit antenna m followed by a code correlator. An M-antenna transmission scheme requires M parallel space-time equalizers the outputs of which are decoupled to code correlators and followed by a proper decoder or detector for the applied MIMO transmission scheme. The processing can assume that the channel has only one multipath. This can also be seen in Figure 9 where only a single code correlator is used for detection of the symbol transmitted from antenna m. Figure 9. Basic structure of LMMSE space-time equalizer followed by a spreading code correlator. A property of the LMMSE receiver structure in Figure 9 is that the receiver can be functionally split into two parts: (i) so-called prefiltering and (ii) conventional RAKE detection including decoding of the possibly used space-time block code. The prefiltering performs the function of the inverted Crr(i) matrix Figure 9. The inverse matrix can be shown to approach a block matrix with Toeplitz blocks. This implies that the matrix multiplication operation can be replaced by linear filtering. With ï ï þ ï ï ý ü ï ï î ï ï í ì ÷ ÷ ÷ ÷ ÷ ø ö ç ç ç ç ç è æ = )()()()()()( )()()()()()( )()()()()()( E)( HH 2 H 1 H 2 H 22 H 12 H 1 H 21 H 11 iiiiii iiiiii iiiiii i NNNN N N rrrrrr rrrrrr rrrrrr Crr     r(i) hm,1(i) Channel estimator P/S r1(i) rN(i) hm,N(i) dm(i)Correlator Multipath combiner Multipath combiner Multipath combiner + Crr -1(i) IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 17 of 33 multiple RX antennas this conversion from block-wise matrix operation to continuous linear filtering requires a bank of parallel linear filters, as shown in Figure 10. The benefit of the prefilter-RAKE structure is that, regardless of the number or the origin (transmit antenna, data layer, base station etc.) of the signals to be detected, the same prefilter bank can be used prior to conventional RAKE-type receiver to convert the receiver into an LMMSE space-time equalizer. Figure 10. Prefilter-RAKE structure of LMMSE space-time equalizer. 2.4.2 Maximum Likelihood Receivers Optimal maximum likelihood (ML) receivers cannot be applied in practice due to complexity reasons. By considering intersymbol, multiple access and intercell interference as additive noise, it is possible to formulate a suboptimal ML receiver which is useful especially with layered schemes (e.g. BLAST, DSTTD) when linear space-time equalizers do not suppress the interlayer interference effectively enough. The assumptions imply that the receiver can function in one-shot mode detecting each symbol interval or decoding each space-time code word separately. Assuming a general L-path channel, and N receive antennas, the receiver despreads all NL signals using as many parallel code correlators. The input signal of the ML receiver is formed by the correlator outputs and can be simply written as . (2.10) Symbol vector s(k) holds the unknown symbols and H(k) is the channel matrix. Vector n(k) includes all such signal components that are considered as noise in the estimation process. (It should be noted that the above model applies directly only to the very basic V-BLAST scheme. In most cases, the effect of space-time block coding have to be incorporated into the signal model as shown e.g. in Section 2.3.1.) r(i) h1(i) Channel estimator P/S r1(i) rN(i) hN(i) Prefilter 1 Prefilter 2 Prefilter N b(k) Conventional RAKE detector g1(i) g2(i) gN(i) )()()()( kkkk nsHy += IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 18 of 33 The unknown symbol vector s(k) can be estimated using ML principle as , (2.11) where Cnn is the noise covariance matrix. The noise covariance can either be assumed to be a unit matrix (I) or it can be estimated. ML solution is found through exhaustive search over all possible combinations of the unknown symbols. Since the receiver complexity grows exponentially as a function of the number of unknown parameters in vector s, it is possible to apply the ML principle to a subset of the unknown parameters while considering the remaining parameters as additive noise (by including them in Cnn). A subset can be selected based on the instantaneous signal-to-noise ratios of the parameters. After detection, the subset is cancelled from y(k), and then the remaining unknown parameters can be estimated. Application to Space-Time Block Coding When ML reception is applied e.g. for DSTTD transmitted signal, a problem is that block coding cannot be presented as a linear transformation of input symbol vector s (due to the fact that complex conjugation is not a linear operation). In this case, the signal model have to be modified as shown in earlier sections after which the above ML solution is applicable. Two possible ML approaches can be considered when the modified signal model is used: (i) to apply signal y(k) directly, or (ii) to apply signal HH(k)y(k) as an input to the ML detector. The latter technique has the benefit that, by first decoding the block code, one or several interfering symbols cancel out from the signal (since some elements of HH(k)H(k) equal to zero) thus simplifying the ML search. However, at the same time, the noise term becomes very coloured implying that the noise covariance matrix cannot be neglected. The latter method also fits better for coherent combining of packet retransmissions to obtain chase-combining gain. ( ) ( ) { } )()()()()()()(minarg)( ˆ1 H )( kkkkkkkk ksHyCsHys nn s --=- IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 19 of 33 Figure 11. Maximum likelihood receiver structure. Soft Output Generation ML detection generates hard decisions for the symbols. Bit-wise soft outputs for the channel decoder can be generated e.g. by computing ratio (2.12) where e(k) is the best of all such symbol vectors which correspond to a bit vector in which the bit of the interest is other than in . This technique automatically takes into account the applied Gray bit-to-symbol mapping. 2.5 Combination of Linear Dispersion Codes with Turbo Space Time Decoding. The transmitter is based on a combination of a turbo encoder, linear dispersion space time codes [Hassibi01], and multicode spreading. First, a frame of raw data bits is encoded using a standard Turbo Code with rate R. The coded bits are interleaved, and modulated using Mc bits per real symbol. These symbols are then multiplexed into the P parallel ST encoders. h1(i) Channel estimator r1(i) rN(i) hN(i) Correlator 1 Correlator 2 Correlator NL s(k) Maximumlikelihood estimator y1(k) y2(k) yNL(k) r2(i) ( ) ( ) ( ) ( ) )( ˆ )()()()( ˆ )()( )()()()()()()( 1 H 1 H kkkkkkk kkkkkkk sHyCsHy eHyCeHy nn nn -- -- - - )( ˆks IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 20 of 33 Figure 12. – Transmitter structure. Each LDC ST encoder brakes its symbol sequence into B blocks of 2Q symbols. Each block will be transmitted through T time slots. Before transmission, the output of the encoders are spreaded using Hadamard codes. Figure 13. – Space time encoder and spreader. The overall output of the transmitter’s M antennas at the i’th chip can be expressed as a linear function of all the symbols in the associated block: (2.13) where is a matrix representing the spreading codes, is a matrix representing the LDC structure, denotes the Kronecker product, is a vector representing the transmitted symbols (of all the parallel encoders) at the b’th block, and we define the indexes , and follow the following structure: Encoder Interleaver Mod LDC x x x c(1) LDC : x x x c(2) LDC : x x x c(P) • • • • • • MUX #2 P branches #1 + + #M + M branches LDC x x x 2Q symbols . . . . . . T channel uses M branches c(p) MxT matrix MxT SF matrix ( )( ) ( )( ) ( ) ( )( ) BTSFiibitici T××=Ä=,...,1)( sDcx ( ) cc ( ) tD Ä ( ) bs 2QP ()ci ()ti ()bi IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 21 of 33 The optimal receiver for the previous problem formulation is impractical. It involves a Viterbi Algorithm of the effective super-trellis associated with the concatenated encoders, and the frequency selective channel. The complexity of such a solution is too high for implementation. We therefore resort to a suboptimal approach, usually referred to as the ``Turbo Algorithm''. The use of interleavers decouples the demodulation stage and the decoding stage. This allows for iterative exchange of soft information between them: Figure 14. – Iterative receiver. The soft information is expressed using log likelihood ratios (LLR). The input to each stage is the a priori LLRs denoted by . Using them, it calculates the a posteriori LLRs, denoted by , which are supposedly ``better''. The difference between these LLRs is usually referred to as extrinsic LLR and is denoted by . The extrinsic LLRs of each stage are used as the a priori LLRs for the other stage. This procedure is repeated in an iterative fashion. (2.14) The decoder uses the well known BCJR algorithm [Pollara96]. In brevity, this algorithm calculates the a posteriori LLR of each bit as a function of the a priori LLR of all the other bits and the correlations between them induced by the encoder's trellis. This algorithm is well known, therefore we leave its details to references. The demodulation algorithm calculates the extrinsic LLR of each bit as a function of the a priori LLR of all the other bits and the received samples. This algorithm is based on the soft input soft output, linear multiuser detector, e.g., [Wang99].                       Bb Ttt b Ttt SFccSFccSFccSFcc = == = == ======== 1111 1 1 1 Demod Deinterleaver Decod LE1 LE2 + Interleaver LA1 LA2 Hard Decision - Received data from the bth block s(b) ^ Estimated symbols y(b) () A Ld () D Ld () E Ld ( ) ( ) ( ) (1/) (1) (/ 1) log log log (1/) (1) (/ 1) DAE Ld Ld Ld Pd Pd P d Pd Pd P d === =+ =-=-=- yy yy    IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 22 of 33 Figure 15. – Soft input soft output demodulator. The demodulator is basically a weighted least squares linear estimator. The estimator approximates the received vector as a linear function of the unknown symbols, and then uses a priori statistical information on these symbols to derive the linear estimator. The estimates are then converted to extrinsic LLRs. The linear estimator uses the following approximation for the received vector structure: (2.15) with , , , and . Due to the frequency selective nature of the channel, this structure is only an approximation. However, as will be explained in the sequel, using the DFE block validates this approximation. y(b) LA1 DFE Linear Estimator s to LLR ^ LLR to s ^ y(b) ~ LA1 LE1 s(b-1) ^ A priori statistical information on the symbols s(b) ^ () ()() () 1bbbbbB»+=yGsw  { } { } Re ( ) () Im ( ) b b b éù =êú ëû y yy { } { } Re ( ) Im ( ) b b b éù =êú ëû HD G HD { } { } Re ( ) () Im ( ) b b b éù =êú ëû w w w ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (0) (1) (0) (1) (0) () 1 1 11 0 0 12 120 () 00 0 11 () 1 1 T TT L T TT T TT T T TT b b SF T b SF T b SF T b SF T b SF T b b SF T b SF T c t SF T b b SF T b SF T - éù =-×+×× ëû éù -×+ êú -×+-×+ êú =êú êú êú ×× ×× ëû éù =Ä Ä Ä êú ëû éù =-×+×× ëû yy y H HH H HH Dc D c D c D ww w        IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 23 of 33 Using the well known matrix inversion lemma, e.g., [Kay93], it is possible to show that the weighted least squares estimate for the unknown symbols is (for simplicity we omit the block indexes): (2.16) where and are the a priori expectation and covariance of the unknown symbols. These a priori symbols statistics can be easily derived from the a priori bits LLRs: (2.17) where denotes the vector of all the bits associated with the symbol . Using these moments it is easy to express and : (2.18) Therefore, the demodulator derives the a priori statistics, and then applies the weighted least squares estimators. Finally, the last stage of the demodulator transforms the symbols estimates into bits LLRs using the following formula: (2.19) where in the last equation, denotes the vector of all the associated bits excluding the current bit . As previously explained, the receiver assumes a block channel structure. In order to validate this assumption, we propose the use of a decision feedback equalizer (DFE), which will mitigate the inter block interference. Thus, before applying the estimator for each block, the DFE tries to eliminate the interference of the previous block: ( ) ( ) ( ) 1 2 11 ˆ 11 H SS H SS HH KK diag s - =-+ =+ éù éù =ëû ëû sΣG W y E E W GC G I Σ G WG G WG S E SS C ( ) ( ) ( ) ( ) [ ] ( ) ( ) ( ) ( ) /2 /2 /2 2 2 Ai Ai Ai ki Ld i iLd Ld ki i Es s Pd e Pd ee Es s Pd b b -+ éù =êú ëû == Þ +éù éù =êú ëû ëû åÕ åÕ d d d d d ( ) sd S E SS C [ ] [ ] [ ] [ ] 22 11 1 22 00 00 00 S SS KKK Es E s Es Es Es E s éù éù - éù ëû êú êú == êú êú êú êú éù - ëû êú ëû ëû EC ( ) ( ) ( ) ( ) ( ) 21 2 2 ˆ 21 2 2 ˆ 1ˆ log exp 1, 2 1ˆ log exp 1, 2 jk jk T Dj k j j A s T kj j A s Ld s sd ssd s s ¹ ¹ ¹ ¹ ìü éù ïï =--=++ êú íý êú ïï ëû îþ ìü éù ïï ---=-+ êú íý êú ïï ëû îþ å å d d ddL ddL j¹ d j d IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 24 of 33 (2.20) for and , where is obtained from equation (¡Error! No se encuentra el origen de la referencia.) by substituting the symbols with their previous estimates. ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 1 () ˆ 11 1 1 L l li bBi bBi bBil bBil - = -+= -+- -+- -+- å yy H x  2bB= 1iL= ˆ x IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 25 of 33 3 HIGH SPEED UPLINK PACKET ACCESS (HSUPA) 3.1 Introduction In the following we consider the simulation models for both a low-complexity UE (1antenna UE) and a dual-antenna UE with MIMO capability. The single antenna case serves as a reference case with "standard" 1-code transmission but multicode transmission is also considered. In addition, a multi-antenna UE with single Tx-chain may be interesting from viewpoint of implementation complexity and cost. This approach allows antenna switching in an open loop mode and antenna selection using feedback from Node B. After considering the results from the initial study and taking into account the complexity issues of the Node B, the following cases were selected for detailed simulation study. This approach was chosen also because it represents a straightforward evolution path from the UE capabilities of the current 3GPP specification. Therefore, in this study, we assume that the signal structure of Uplink Common Packet Access using the Random Access Channel (RACH) is employed. It must be noted, however, that concept proposals for uplink high-speed packet access has not been considered yet in 3GPP. For example, it is not possible to obtain uplink channel state information at Node B because no reference channel exists fro uplink. Moreover, there are no feedback channels from Node B to UE to carry the channel state information. 3.2 Simulation model 3.2.1 3GPP compatibility The simulation model has been designed so that it model includes full 3GPP features. This approach allows a common reference point for advanced MIMO concept as well as realistic signal models and receiver implementation schemes. In a first stage a standard Rake receiver structure is assumed but the model allows also the implementation of more complex receivers with interference suppression in a straightforward manner. 3.2.2 High-level modelling of uplink high-speed data with MIMO approach The detailed simulation model includes features such as: • packet mode data with ARQ • channel coding (convolutional/turbo ) with different code rates • interleaving • rate matching / puncturing • fast power control • realistic interference modelling (64 kb/s data users) IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 32 of 33 - Known channel state - Adaptivity on the link level: transmission during favorable channel state - Antenna switching w/wo ARQ - Scheduling not considered Case 4 (tentative): • Multi-code via 1 Tx-antenna vs. multi-codes via 2 Tx antennas - Reference case: 1x2, 1x4 - Evaluating uplink peak data rate with MIMO approach IST-2000-30148 I-METRA Doc.: IST-2000-30148/UPC-WP3-D3.2-V1.2.doc Title: I-METRA D3.1: Implementation of the Suitable algorithms 31 October, 2002 File: UPC-WP3-D3.2-V1.2.doc Page 33 of 33 4 REFERENCES [I-METRA D3.1] IST-2000-30148 I-METRA, Deliverable 3.1, "Design, Analysis and Selection of Suitable Algorithms", available at www.istimetra.org [Trombi] 3GPP TSG RAN WG1#19, “Trombi – an extension of STTD for more than two antennas”. [DSTTD] Texas Instruments (2001) Double-STTD Scheme for HSDPA Systems with Four Transmit Antennas: Link Level Simulation Results. In TSG-RAN WG1 meeting #20. [Hottinen] A. Hottinen and O. Tirkkonen, ”Non-Orthogonal Space-Time Block Code with Symbol Rate Two”, Conference on Information Science and Systems, Princeton University, March 20-22, 2002. [Wang99] X. Wang and H. V. Poor. “Iterative (turbo) soft interference cancellation and decoding for coded CDMA”, IEEE Transactions on communications, vol. 47:pp. 1046-1061, July 1999. [Pollara96] S. Benedetto, D. Divsalar, G. Motorsi, and F. Pollara. “A soft input soft output maximum posteriori module to decode parallel and serial concatenated codes”, TDA Progress Report 42-127, November 1996. [Kay93] S. M. Kay, “Fundamentals of Statistical Signal Processing”, Prentice Hall, 1993. [Hassibi01] B. Hassibi and Bertrand M. Hochwald. “High rate codes that are linear in space and time”, Submitted to IEEE Trans. Info. Theory, August 2000. Revised April 2001.