FLOWS 1ST-2001-32125 Deliverable No:D14 LIST OF TABLES alized system model Table 4.3 Mean valuesofdetermined analytically by(4.119)and by simulations.9 Table 4.4 Dimensions of the vectors and matrices used in the structure of Fig.2.1..........6 Table 4.5 Possible assignment of 2 bits for signalling the selection of the demodulation procedure 121 Table 4.6 Used CDMA codes of [3GP00b].. .125 Snapshot of channel impulseso n conditions for the considered channel model 12 Table 4.9 Relative transmitted energy for a snapshot with different scenarios... 128 Table 4.10 Parameters for an exemplary scenario... 132 Table 4.11 Relative transmit energies for the support of one MT .133 179 Table 6-2:Performance and complexity comparison of SM decoding techniques.....201 Table 6-3:MIMO-OFDM simulation parameters.. …205 31December 2003 Page 17
FLOWS IST-2001-32125 Deliverable No: D14 31st December 2003 Page 17 LIST OF TABLES Table 4.1 Matrices for normalized system model .............................................................................81 Table 4.2 Fixed but randomly chosen channel coefficients............................................................89 Table 4.3 Mean values ( ) , rel k η n of ( ) , rel k ηn determined analytically by (4.119) and by simulations ......89 Table 4.4 Dimensions of the vectors and matrices used in the structure of Fig. 2.1 ...................96 Table 4.5 Possible assignment of 2 bits for signalling the selection of the demodulation procedure ....................................................................................................................................121 Table 4.6 Used CDMA codes ( ) k c of [3GP00b] ...............................................................................125 Table 4.7 Propagation conditions for the considered channel model .........................................126 Table 4.8 Snapshot of channel impulse responses .......................................................................126 Table 4.9 Relative transmitted energy for a snapshot with different scenarios .........................128 Table 4.10 Parameters for an exemplary scenario .........................................................................132 Table 4.11 Relative transmit energies for the support of one MT µ1 ...........................................133 Table 6-1: Classification of MIMO-OFDM techniques according to the available channel knowledge ...................................................................................................................................179 Table 6-2: Performance and complexity comparison of SM decoding techniques ....................201 Table 6-3: MIMO-OFDM simulation parameters..............................................................................205
FLOWS 1ST-2001-32125 Deliverable No:D14 1.Introduction This documents FLOWS deliverable 1Report on selected MMi and their bes wo Multi-stand d MIMO tech ation of MIMO techniques).The BoPeiewfeag2rmtaaueodeunane8eeMopmetotaeetectabegorA5mpgmnthRng requires its performance evaluation. The document divides into three main parts:chapters 2 and 3describe work under A5.1.while discuss the ssue of cha opme ues oonrp9nsgepearesonmettersueswhehesennnPsneheeo2aSeg ide ant for a uch sche apters 8 A5 Chapter 10 add of MIMO tec 1800.UMTS AN/Z wirele ed in FLOW multi-st ard friendliness gomheonceptosandadme dliness identifies channel identification asa vital,and under-researched a Chapter 3 describes the road cssification of MIM tech a de d friendliness of standards. of sta eeoeaahesage nle concep may re uire fullv multi-m e terminals rsevera Accordin iven the nsiderable diffe standards such as UMTs and the WLA to ards such as HIPERLAN/2.11a.and to WCDMA-based standards.particularly UMTS it for true napter 6 co ility of channel kr mitter an os schemes which ca aspects of mu ce cancellation.primarily applicable to the up-ink.and chip equalisation.particularly for the 31December 2003 Page 18
FLOWS IST-2001-32125 Deliverable No: D14 31st December 2003 Page 18 1. Introduction This document is FLOWS deliverable D14 “Report on selected MIMO techniques and their performance”. It describes work performed in FLOWS workpackage 5 (“Baseband processing”) under Activities A5.1 (“Survey of Multi-standard MIMO techniques”), A5.2 (“Development of selected MIMO techniques”), and part of that under A5.3 (“Performance evaluation of MIMO techniques”). The purpose of A5.1 was to survey and select MIMO techniques suitable for use within the FLOWS project, while A5.2 has carried out further development of these techniques. A5.3, in part taking place simultaneously with A5.2, is evaluating performance of the techniques. Of course in practice it is difficult to separate these two activities: the development of a technique of course simultaneously requires its performance evaluation. The document divides into three main parts: chapters 2 and 3 describe work under A5.1, while chapters 4-7 describe further development of techniques under A5.2 and A5.3. Chapters 8 and 9 discuss the issue of channel estimation, identified as important for all such scheme during A5.1. Chapter 10 additionally covers some further issues which have arisen in WP5 which relate to linkages to other parts of the project. The criteria for selection of MIMO techniques in A5.1 included the compatibility of the techniques with multiple wireless standards, and especially the standards selected for work in FLOWS, namely GSM- 1800, UMTS and HIPERLAN/2 (subsequently extended to include other wireless LAN standards operating at 5.2 GHz and using OFDM, i.e. IEEE 802.11a). This issue was described in the FLOWS Technical Annexe as “multi-standard friendliness”, building on the concept of “standard friendliness” introduced in the METRA project. Accordingly chapter 2 describes a literature survey which has been undertaken to ensure familiarity with the significant parts of the by now voluminous literature of this subject, including relevant previous IST programmes. This focuses on different types of channel, and the MIMO techniques designed for them, considering also the capacity limits for the channel. It also identifies channel identification as a vital, and under-researched area. Chapter 3 describes the remaining work undertaken under A5.1 to assist in the selection of MIMO techniques, including a broad classification of MIMO techniques, a description in more detail of some candidate MIMO techniques, an analysis of the concept of “multi-standard friendliness”, and finally a description of the selection process and a list of categories of technique for further consideration. The techniques have been selected both on the basis of their multi-standard friendliness and of their performance. Chapter 3 acknowledges that the central concept of FLOWS, namely simultaneous use of standards, inherently requires multiple standards to operate at the same time, which to some extent precludes extensive sharing of functions between standards. Thus implementation of this concept may require fully multi-mode terminals, including signal processing functions for several standards. Under these circumstances separate implementations of standards may be used. Accordingly, given the considerable differences between standards such as UMTS and the WLAN standards it is worthwhile to consider MIMO approaches aimed specifically at the these types of standard. Thus chapters 6 and 7 consider, respectively, schemes most appropriate to OFDM-based standards such as HIPERLAN/2 or IEE 802.11a, and to WCDMA-based standards, particularly UMTS. Chapters 4 and 5 describe further work to date on one promising transmission scheme, Joint Transmission, aimed originally at WCDMA-TDD but more widely applicable, and especially to develop it for true MIMO systems. Chapter 6 covers the work done on techniques applicable to OFDM systems like HIPERLAN/2, beginning with a classification of these schemes according to the availability of channel knowledge at transmitter and receiver, and develops schemes which can provide a compromise between such schemes. Chapter 7 considers two aspects of multi-user detection applied to MIMO-CDMA systems, particularly relevant to UMTS, namely iterative parallel interference cancellation, primarily applicable to the up-link, and chip equalisation, particularly for the downlink. Note that chapters 5, 6 and 7 introduce new techniques which have been developed under FLOWS: namely Channel Oriented Joint Transmission (CO-JT), a MIMO version of JT, a compromise scheme
FLOWS 1ST-2001-32125 Deliverable No:D14 P8vientig8ralehniecomposionandELUAST.anmdspace-ineturbocodedwucDMAwtmtutbo r2 has all MIMO tech nsiSb9greheyesnaion This has up to now bee estimation.including the devel ment of Cramer-Rao bounds on the a of channel estimatio mao-n0eyenaonecmques.anmanieraep0aciom2s6hgeon8aeaa in Y1M9 o 2002 dor to that pont.This included the sent ch nd nate awork in WP5 in A5 3 will in futur shift towards the evaluation of MIMO context of LOWS scen making use of th channel mo develd ed in WP ance to on the pe complete wireless network.The work described in this report lays a sound basis for this. 31December 2003 Page 19
FLOWS IST-2001-32125 Deliverable No: D14 31st December 2003 Page 19 between singular value decomposition and BLAST, and space-time turbo-coded W-CDMA with turboPIC multi-user detection. The literature review of chapter 2 has highlighted the need in nearly all MIMO techniques for channel estimation in general, which can be broadly interpreted to include parameter estimation in general, including frequency estimation. This is frequently significantly more complex in MIMO systems than in SISO, since there are effectively as many channels to estimate as there are antennas pairs, that is, the product of the number of transmit and the number of receive antennas. This has up to now been an under-researched area, and chapters 8 and 9 describe work done within the project on channel estimation, including the development of Cramer-Rao bounds on the accuracy of channel estimation, maximum likelihood estimation techniques, and an iterative approach to joint estimation and decoding in turbo-coded systems. Chapter 10 arose from a need from other workpackages considering propagation, antennas and RF systems for information on the requirements imposed by MIMO techniques on these areas. It is included because it gives some information on the requirements of MIMO terminals. It has been used in the design of antennas and the development of channel models, and it also covers the application of FLOWS channel models and antenna arrays, and will thus be further used in the simulation of MIMO terminals in realistic propagation environments. Note that a draft version of this deliverable was issued in Y1M9 of the project (September 2002), describing work done up to that point. This included the present chapters 2, 3, 5 and 10 and parts of chapters 6 and 7. The full version of the deliverable has updated these chapters, adding substantially to chapters 6 and 7, and additionally the material of chapters 4, 8 and 9 is substantially new. The focus of continuing work in WP5, in A5.3, will in future shift towards the evaluation of MIMO techniques in the context of FLOWS scenarios, making use of the channel models developed in WP2, the antennas of WP3, and the limitations of the RF system as determined in WP4. The objective is to provide information on link level performance to inform work in WP6 on the performance of a complete wireless network. The work described in this report lays a sound basis for this
FLOWS 1ST-2001-32125 Deliverable No:D14 2.MIMO Communication Systems:Literature Survey 2.1 Introduction In this report we present a survey on MIMO communication techniques. 2.2 Flat fading channels Most publications on MIMO syst ems are the flat fading channel preferably.chan be usefully teiwhspeti l-do Moreover CDMA system sp阳ogen5edP 2.2.1 Models of flat fading MIMO channels Y作eceidernatfhdinsMwehanmowtenemi9nteaen6agyrecTYeanema n ram to ey tanmited with variances Eh h B.- (2.1) The signal observed at receive antenna j is given by 当=∑g+ (2.2) is the r represented as y=Hs+n (2.3) where H is a nx nr matrix with entries hia.the nx 1 vector s contains the transmitted symbols si,the ng x 1 vector n contains the noise samples nj.and the nT x 1 vector y contains samples y;at receive antennas. 17 December 2003 Page 20
FLOWS IST-2001-32125 Deliverable No: D14 2. MIMO Communication Systems: Literature Survey 2.1 Introduction In this report we present a survey on MIMO communication techniques. We divide the techniques in accordance with the channel model used for their derivation. As a result, section 2.2 presents techniques for flat fading channels, section 2.3 describes techniques for more realistic frequency-selective channels, and section 2.4 considers MIMO techniques for time-varying channels. In addition, results of recent projects devoted to development and investigation of MIMO techniques for UMTS and HIPERLAN/2 are presented in sections 2.5 and 2.6, respectively. Finally, section 2.7 contains conclusions. 2.2 Flat fading channels Most publications on MIMO systems are concerning the flat fading channel, preferably, channels with Rayleigh channel coefficients. Although this channel model is not realistic, it can be usefully applied to systems with multicarrier modulation. Moreover, CDMA systems can also benefit from this model when spectral-domain signal processing techniques are used in the receiver. In this section we present a survey of MIMO techniques for such channels. 2.2.1 Models of flat fading MIMO channels We consider a flat fading MIMO channel with nT transmit antennas and nR receive antennas. The tap gain from transmit antenna i to receive antenna j is denoted by hij . The antennas are separated far enough to ensure nT · nR independent fading channels from transmit to receive antennas. The channel taps hij are independent zero-mean complex Gaussian random variables with variances E{|hij | 2} = 1. Here, we consider slow fading channels, so that the channel taps are not varying in time. The symbol transmitted from antenna i is denoted by si. The total energy transmitted by all antennas is Es = nT i=1 |si| 2 . (2.1) The signal observed at receive antenna j is given by yj = nT i=1 hijsi + nj (2.2) where the additive noise nj is white and Gaussian with E{njn∗ k} = N0δjk, N0 is the noise spectral density, and δjk is the Kronecker delta. In matrix notation, the channel model is represented as y = Hs + n (2.3) where H is a nR × nT matrix with entries hij , the nT × 1 vector s contains the transmitted symbols si, the nR × 1 vector n contains the noise samples nj , and the nT × 1 vector y contains samples yj at receive antennas. 17 December 2003 Page 20
FLOWS IST-2001-32125 Deliverable No:D14 2.2.2 Capacity of flat fading channels channd mtri whle the ranmter does not.the channel c={u+系]} (2.4) where I is Pout Pr(C(H)<R) (2.5) ility that t the dat age probabio ergodic capacity C=E(C(H)} (2.6) gives the maximum transmission rate afforded by the channel. According to ifG981.the shannon capacity for a system with one transmit and n receive antennas scales only logarithmically wit -or a system omtneceive antennas the aympotic y grows linearly prodfolesatnloecopmPtationelostsaHoneer nd num ect of recei e an ennas ading c nnels this been shown that the e capacity AC (1-r)ioge (1+SNR.F) +1og2（1+1+5NR-I阳F/ (2.7) code rate, is the squar ed Frobeni us norm.are s time it is ra ate(2 tha eigh channel with ng receiv antennas the rank is one only for 1 Wh multiple rec ny rate a ty.AI e )S TBC lim CsTBc log2(1+ngE./No), (2.8) while the channel capacity approaches the capacity of nR parallel AWGN subchannels lim C=nglog2(1+E/No). 2.9 The capacity difference increa rapidly with increasing SNR and increasing numbers of an- loss in capacity is significant. 17 December 2003 Page 21
FLOWS IST-2001-32125 Deliverable No: D14 2.2.2 Capacity of flat fading channels If the receiver knows the channel matrix H, while the transmitter does not, the channel capacity is given by [FG98] C(H) = log2 det I + Es nTN0 HHH (2.4) where I is a nR × nR identity matrix. The channel capacity C(H) is a random variable with respect to the random channel matrix H. For a given data rate R, the outage probability Pout = P r (C(H) < R) (2.5) is the probability that the channel cannot support the data rate R. The outage capacity is the maximum data rate with an outage probability less or equal to a predefined value Pout. This capacity is most suitable for delay sensitive applications. For delay insensitive applications, the ergodic capacity C = E {C(H)} (2.6) gives the maximum transmission rate afforded by the channel. According to [FG98], the Shannon capacity for a system with one transmit and nR receive antennas scales only logarithmically with nR when nR → ∞. For a system using nT transmit and one receive antennas, asymptotically there is no additional capacity to be gained. In a general case of nT transmit and nR receive antennas, the asymptotic capacity grows linearly with min{nT , nR}. Space-time block codes (STBCs) provide full diversity at low computational costs. However, they incur a loss in capacity with respect to the true channel capacity. The loss is a function of channel rank, code rate, and number of receive antennas. For flat fading channels this dependence was investigated in [SP00]. It has been shown that the difference between the channel capacity and the capacity of STBCs is ∆C = (1 − r)log2 (1 + SNR · ||H||F ) + log2 1 + S 1 + SNR · ||H||F (2.7) where r is the code rate, ||H||F = R i=1 σ2 i is the squared Frobenius norm, {σ2 i } are singular values of the channel matrix H, and S is a function of {σ2 i }. It follows from (2.7) that a space-time block code is optimal with respect to the capacity when it is rate one (r = 1) and it is used over a channel of rank one (R = 1). For the Rayleigh channel with nR receive antennas the rank is one only for nR = 1. When multiple receive antennas are used, a STBC of any rate always incurs a loss in capacity. A full-rate (r = 1) STBC with one receive antenna is always optimal. When the number of transmit antennas nT is increased the STBC capacity approaches the capacity of one AWGN subchannel with SNR scaled by nR lim nT →∞ CSTBC = log2(1 + nREs/N0), (2.8) while the channel capacity approaches the capacity of nR parallel AWGN subchannels lim nT →∞ C = nRlog2(1 + Es/N0). (2.9) The capacity difference increases rapidly with increasing SNR and increasing numbers of antennas. Even for small number of antennas, such as nR = 2 and nT = 2 or nT = 4 for SNR=15 dB we have a difference of 1-3 bits per channel use. So, when using the STBCs the loss in capacity is significant. 17 December 2003 Page 21