Multiple-Antenna Techniques in LTE-Advanced

Transcription

1 TOPICS IN RADIO COMMUNICATIONS Multiple-Antenna Techniques in LTE-Advanced Federico Boccardi, Bell Labs, Alcatel-Lucent Bruno Clerckx, Imperial College London Arunabha Ghosh, AT&T Labs Eric Hardouin, Orange Labs George Jöngren, Ericsson Katsutoshi Kusume, DOCOMO Euro-Labs Eko Onggosanusi, Texas Instruments Yang Tang, Samsung Electronics 1 Codebook-based transmit processing assumes that the actual transmit precoder belongs to a codebook known to both transmitter and receiver. On the other hand, with non-codebook based transmit processing the transmitter can use precoding coefficients that are not known by the receiver. ABSTRACT In this article we discuss the different multiple antenna techniques introduced in LTE- Advanced. Rather than describing the technical details of the adopted solutions, we approach the problem starting from the design targets and the antenna deployments prioritized by the operators. Then we present the main enabling solutions introduced for downlink and uplink transmissions, and subsequently assess the performance of these solutions in different scenarios. Finally, we discuss some possible future developments. INTRODUCTION In the last 20 years multi-input multi-output (MIMO) has been identified as a key technology to improve communications over a wireless channel. Although the first beamforming concepts are 50 years old, it is only since the mid- 1990s that MIMO has been capturing large-scale attention, first in the research community and then in industry. Single-user MIMO (SU- MIMO) techniques were first studied, and they can be categorized into spatial diversity, beamforming, and spatial multiplexing techniques. Spatial diversity techniques were identified to transmit or receive the same information over different antennas. Beamforming techniques were studied to maximize the signal-to-noise ratio (SNR) or signal-to-interference-plus-noise ratio (SINR) over a given set of links. Spatial multiplexing techniques were introduced to send multiple data layers in parallel over different spatial dimensions. Multi-user MIMO (MU- MIMO) techniques were then studied starting in the early 2000s, with the goal of allowing simultaneous communications of multiple users over the same time-frequency resources. More recently, coordinated multipoint (CoMP) techniques have been proposed to allow coordinated transmission/reception from antennas belonging to different sites or sectors. The Third Generation Partnership Project (3GPP) standard for Long Term Evolution (LTE) was one of the first wireless standards designed with MIMO in mind from the beginning. LTE adopts various MIMO technologies. Without relying on channel-reciprocity-based techniques, LTE downlink transmission supports up to four antennas at the base station. It allows transmit diversity, codebook-based 1 beamforming, and spatial multiplexing with up to four layers per user, and, moreover, codebook-based MU-MIMO. In addition, user-specific beamforming is supported for any number of antennas at the base station, relying on channel reciprocity. For uplink transmissions, only one antenna is supported for transmission at the user side. However, there is an option for performing antenna switching with up to two transmit antennas. MU-MIMO is also supported in the uplink. The first release of LTE is termed LTE Release (Rel.) 8, which indicates the release year of In Rel. 9, two new user-specific reference signals are introduced in order to enable noncodebook-based dual-layer MU-MIMO beamforming. Inspired by the requirements from the International Telecommunication Union Radio Communication Sector (ITU-R) and 3GPP, LTE-Advanced (LTE-A, also known as LTE Rel. 10, which indicates the release year of 2010) introduces enhanced MIMO technologies. In the downlink new reference signals (user-specific reference signals [RS]) are adopted to enable the use of so-called non-codebook-based (e.g., zero-forcing [ZF]) MIMO transmissions. As a matter of fact, user-specific RS allow the estimation of an equivalent channel, including the pre /12/$ IEEE IEEE Communications Magazine March 2012

2 coding weights between base station and user. New codebook and feedback designs are introduced to support spatial multiplexing with up to eight independent spatial streams and enhanced MU-MIMO transmissions. Dynamic switching between SU-MIMO and MU-MIMO is also supported. In the uplink, SU-MIMO has been introduced with up to four transmit antennas at the user side and transmit diversity is supported for the control channel. In this article we discuss the different MIMO technologies introduced in LTE-Advanced. We present the design targets and prioritized deployment scenarios. We introduce the downlink and uplink MIMO design, focusing on reference signals, feedback, and multi-antenna algorithms. We present some performance assessment. Finally, we discuss some possible future developments. DESIGN TARGET AND PRIORITIZED DEPLOYMENT SCENARIOS REQUIREMENTS AND IMPACTS ON THE NEEDED FUNCTIONALITIES A key design objective of LTE-Advanced was to meet the requirements defined by ITU-R in order to be recognized as an IMT-Advanced technology [3]. These requirements imposed, in particular, minimum values for the peak rate (defined as the highest theoretical data rate under error-free conditions when the maximum available radio resources for the corresponding link direction are utilized for a single user) in the uplink and downlink, for four antennas at the base station and two antennas at the user side. Whereas the peak rate depicts maximum performance, it is not typically achieved in a real-world deployment. In this regard cell average spectral efficiency and cell edge user spectral efficiency (defined as the user spectral efficiency achieved by at least 95 percent of the users) are more reflective of a typical deployment. Therefore, in addition to the minimum peak rate requirements, minimum cell spectral efficiency and cell edge spectral efficiency were also specified for four different operation scenarios: indoor, microcellular, base coverage urban, and high speed. From the peak rate requirements, it was apparent that SU-MIMO needed to be introduced in the uplink. Simulations performed within 3GPP subsequently revealed that conventional LTE Rel. 8 techniques were able to meet the ITU-R spectral efficiency requirements in most scenarios, except in the downlink of the base coverage urban and microcellular scenarios. It was identified that in order to meet the requirements there, enhanced MU-MIMO techniques were needed. On the other hand, 3GPP defined its own performance targets [4], with peak rates defined for an 8 8 antenna configuration in the downlink and a 4 4 configuration in the uplink, thereby extending the antenna configurations support beyond the ITU-R requirements. Additional spectral efficiency requirements were defined for 3GPP case 1, a commonly used evaluation scenario representing a typical urban macro environment. Note that although targeting similar deployments, the 3GPP case 1 and ITU base coverage urban scenarios differ in the associated channel models, so the related performance requirements cannot be directly compared. In summary, the ITU and 3GPP requirements together motivated the LTE Rel. 8 MIMO functionality to be enhanced with the following components: extension of downlink SU-MIMO to 8 8, introduction of SU-MIMO in the uplink up to 4 4, and enhancement of downlink MIMO. In addition, a required enhancement of LTE Rel. 8 was the support of up to 100 MHz bandwidth, achieved via simultaneous operation on multiple carriers through the so-called carrier aggregation concept [4]. PRIORITIZED MULTIPLE-ANTENNA SETUPS The multiple-antenna setups at the transmitter and the receiver have a significant impact on the MIMO performance. In particular, the interantenna spacing and polarization affect the antennas correlation properties, which are key factors in the performance of multiple antennas techniques. Indeed, a high correlation between the transmit antennas is beneficial for beamforming, which directs the signal energy in the direction of the user or base station receiver; in this case a single spatial layer (i.e. a single data stream) is transmitted. Beamforming is especially suited to users that need a higher received power, typically those located at the cell edge. Besides, the directivity of the transmission eases the separation of different users in the spatial domain, and is thus beneficial for MU-MIMO. In addition, this directivity lowers average intercell interference levels. In contrast, a low correlation between the transmit antennas favors multilayer transmission to a single user (i.e., SU- MIMO) and transmit diversity. SU-MIMO is rather suited to users in good radio conditions. This is because the available transmit power is shared between the different layers, which moreover interfere with each other. Depending on the correlation properties, different multipleantenna transmit schemes can thus be favored. Popular multiple-antenna setups include copolarized and cross-polarized antennas. In a cross-polarized setup, the antennas with orthogonal polarization have low correlation (theoretically zero, but in practice the propagation introduces some depolarization of the signals). The antennas with the same polarization have high correlation when they are close to each other (typically half the wavelength apart), the correlation decreasing when the antenna spacing increases. At the base station, experimenting with polarization and antenna spacing leads to various antenna setups with various correlation properties, especially for four and eight antennas, as illustrated in Fig. 1. Among those setups, closely spaced crosspolarized antennas are particularly attractive. On one hand, they combine two uncorrelated sets of correlated antennas, thereby favoring both dual- The multiple-antenna setups at the transmitter and the receiver have a significant impact on the MIMO performance. In particular, the inter-antenna spacing and polarization affect the antennas correlation properties, which are key factors in the performance of multiple-antenna techniques. IEEE Communications Magazine March

3 4-antenna setups (1) Closely spaced crosspolarized setup 8-antenna setups (1) Closely spaced crosspolarized setup (2) Widely spaced crosspolarized setup (3) Widely spaced cross-polarized setup (3) Closely spaced copolarized setup (2) Closely spaced co-polarized setup Figure 1. Base station antenna setups considered for the LTE-Advanced MIMO design. The numbers in parentheses indicate priority. layer SU-MIMO and good beamforming/mu- MIMO capabilities (although not as good as in the case of closely spaced co-polarized setups with the same number of antennas). On the other hand, a four-antenna setup of two closely spaced cross-polarized antennas occupies roughly the same space as a setup with two co-polarized antennas. For these reasons, closely spaced cross-polarized antennas are expected to be the most deployed four-antenna setup, and they have been prioritized in the LTE-Advanced MIMO design. Nevertheless, LTE-Advanced has been designed to be efficient with all the antenna setups shown in Fig. 1. DOWNLINK DESIGN BACKGROUND Closed-loop MIMO techniques in LTE Rel. 8 support two and four transmit antennas and are mainly optimized for SU-MIMO transmissions such as open-loop transmit diversity and open/closed-loop spatial multiplexing with up to four layers. Closed-loop (CL) MU-MIMO with up to two users with one layer per user is also supported but the performance is relatively limited as Rel. 8 typically relies on cell-specific reference signals (CRS) and codebook-based precoding. The same CRS are used by all users in the cell, and they are exploited for both channel estimation and demodulation. One user-specific RS is also supported to enable single-user single-layer beamforming for base stations with more than four transmit antennas. With the intention to improve MU-MIMO performance, Rel. 9 introduced two completely new user-specific RSs in order to enable duallayer beamforming based on non-codebookbased precoding. Further details on Rels. 8 and 9 are available in [1]. REFERENCE SIGNAL DESIGN In LTE Release 10 the whole RS paradigm is shifted from being typically based on CRS to the use of user-specific RS. This fundamental change in RS design was originally triggered by increased interest in multi-antenna techniques for MU-MIMO and CoMP, for which considerable flexibility in determining precoders is needed to limit interference between co-scheduled users. The user-specific RS offers a clear advantage in that respect by providing the base station with complete freedom in determining a precoder for the data transmission to a user as long as the same precoder is used for the user-specific RS and the data within a scheduling unit. The way the transmission is performed hence becomes transparent from a user point of view, offering flexibility to even transmit signals from two widely separated sites (i.e., CoMP) without the user needing to be aware of it. Release 10 RS is based on separate RS for demodulation and channel state information (CSI) feedback. The user-specific RS, also called demodulation reference signal (DMRS), is used for demodulation, while channel estimation measurements for determining user CSI feedback is performed on so-called channel state information reference signals (CSI-RS) typically shared by all users in the cell. One reason for introducing such a dual-rs concept was to exploit that demodulation typically requires much more accurate channel estimates than the estimates needed for CSI feedback. There is consequently a difference in density between the DMRS and CSI-RS, which balances the signaling overhead by exploiting the fact that the transmission rank to a user is never higher, and is typically much lower, than the number of transmit antennas. DMRS allows both the demodulation and CSI feedback to support dynamic rank adaptation with up to as much as eight layers scheduled to a single user for 8 8 MIMO. CSI-RS also enables a user to estimate the CSI for multiple cells, to allow future multicell cooperative transmission schemes. FEEDBACK DESIGN AND MULTIANTENNA ALGORITHMS The support of eight transmit antennas in LTE- A Rel. 10 for CL MIMO represents a major downlink enhancement compared to Rel. 9. The system is designed such that a single transmission mode enables CL SU-MIMO, CL MU- MIMO, and dynamic switching between SU and MU-MIMO based on the same feedback mechanism. This leaves full flexibility at the base station to select the most appropriate transmission scheme and incurs low latency as no higher-layer reconfigurations are required. No transmit diversity scheme has been standardized for eight transmit antennas as the expected performance gain is marginal. Nevertheless, transmit diversity standardized in Rel. 8 for two and four transmit antennas can be used even when eight transmit antennas are deployed by using so-called antenna virtualization. Antenna virtualization consists of precoding the physical array such that the user effectively perceives only two or four antennas. With eight transmit and eight receive antennas, a single user can be scheduled in SU-MIMO 116 IEEE Communications Magazine March 2012

4 and receive up to eight simultaneous layers using spatial multiplexing. MU-MIMO can be used to enable simultaneous transmission of up to four layers with up to two layers per user and is motivated by the practical dual-polarized antenna deployment, explained earlier. CL MIMO, and especially MU-MIMO, is sensitive to the feedback accuracy [2]. MU- MIMO with non-codebook-based precoding and precoded user-specific RS enables the use of advanced transmit filtering at the base station and advanced feedback mechanisms at the user side. While the exact base station transmit precoder design is an implementation issue, an appropriate feedback mechanism is designed to fully benefit from the use of non-codebook based precoding. Release 10 relies on the Rel. 8 feedback mechanism for both frequency-division duplex (FDD) and time-division duplex (TDD) (uplink sounding can also be used for TDD to take advantage of the uplink-downlink channel reciprocity property as in Rel. 8/9). Hence, Rel. 10 inherits the well established and time-tested Rel. 8/9 feedback framework, which relies on reports from a precoding matrix indicator (PMI), a channel quality indicator (CQI), and a rank indicator (RI). The PMI is chosen from a finitesize codebook as a recommendation on which precoding weights to apply at the transmitter. Note that although a codebook is used for CSI feedback reports, the actual transmit precoding is not constrained to use any codebook when user-specific RS are employed. The user typically makes the hypothesis that it will be scheduled in SU-MIMO and decides on RI, PMI and CQI accordingly. For two and four transmit antennas, the feedback relies on the Rel. 8 codebooks. The introduction of enhanced feedback mechanisms for those antenna configurations was shown to provide moderate performance enhancements. For eight transmit antennas, a feedback framework based on a double-codebook (4 bits per codebook) structure is introduced. The recommended PMI is obtained as the multiplication of one matrix targeting long-term and wideband feedback information and another matrix targeting short-term and subband feedback. This scheme results in a good performance vs. overhead trade-off in practical closely spaced crosspolarized antenna scenarios. The first matrix has a block diagonal structure to cope with the channel statistics properties of cross-polarized antennas. The overall precoder is constant modulus to fully utilize power amplifier resources in SU- MIMO, is drawn from a phase shift keying (PSK) alphabet to reduce the PMI search complexity, and is designed to perform well for both SUand MU-MIMO. For MU-MIMO, the base station scheduler is tasked to derive a suitable MU- MIMO precoder based on SU-MIMO feedback. UPLINK DESIGN BACKGROUND LTE Rel. 8/9 allows transmit antenna selection at the user equipment (UE) side, whereas multiple-antenna transmissions are not supported. MU-MIMO is supported as a base station and is free to schedule uplink transmission from multiple users within the same time and frequency resources. Release 8/9 specifies two types of reference signals: DMRS and sounding reference signal (SRS). DMRS is transmitted by each user to enable uplink channel estimation at the base station for the purpose of demodulating uplink transmission. Consequently, DMRS occupies the same frequency resource blocks as the uplink transmission. SRS, on the other hand, allows the base station to estimate the uplink channel of the transmitting user for enabling uplink frequency-selective scheduling and link adaptation. For such purpose, SRS is transmitted across different frequency resource blocks with lower frequency resolution than DMRS. REFERENCE SIGNAL DESIGN Different from the downlink case where the Rel. 10 RS design is fundamentally different from the Rel. 8/9 one, in uplink the Rel. 10 design follows the Rel. 8/9 philosophy. As a matter of fact, the main difference is the extension of DMRS and SRS to enable uplink SU-MIMO in Rel. 10. As explained later, uplink multi-antenna transmission supports codebook-based precoding for data transmission. This affects how DMRS and SRS are designed in Release 10. To support multilayer transmission, one DMRS resource is allocated for each of the scheduled transmission layers to allow the base station to estimate the uplink channels associated with all the available transmission layers. The DMRS across all the transmission layers are to be precoded in the same manner as the corresponding data transmission. This means that the same precoding vector/matrix is used for data and DMRS across all the scheduled transmission layers. While DMRS is defined with respect to transmission layer, SRS is defined with respect to transmit antennas. For uplink SU-MIMO, link adaptation encompasses precoding matrix and rank (number of transmission layers) adaptation in addition to modulation/coding rate at the base station. Since SRS is used for rank and precoding adaptation, SRS is not precoded with the same precoding matrix/vector as data. For most practical purposes, it suffices to state that SRS across physical transmit antennas are not precoded. To illustrate the above concept, let us assume a user with four transmit antennas scheduled to transmit two layers. In this case, the user is assigned two DMRS resources where the DMRS across the two layers are precoded in the same manner as data. In addition, four SRS resources are assigned to the user where the SRS across four transmit antennas are not precoded. Other enhancements to the SRS with respect to Rel. 8/9 are represented by the possibility of allowing channel state estimation of the same user in different cells (by exploiting the uplinkdownlink channel reciprocity) to enable CoMP transmissions in future LTE releases. MULTIANTENNA ALGORITHMS Compared to the downlink case, the design of multiple-antenna techniques for uplink transmissions is limited by power amplifier efficiency and The DMRS across all the transmission layers are to be precoded in the same manner as the corresponding data transmission. This means that the same precoding vector/matrix is used for data and DMRS across all the scheduled transmission layers. IEEE Communications Magazine March

5 Scheme Cell spectral efficiency (b/s/hz/cell) Urban micro (UMi) Cell edge user spectral efficiency (b/s/hz/user) Cell spectral efficiency (b/s/hz/cell) Urban macro (UMa) Cell-edge user spectral efficiency (b/s/hz/user) ITU-R Requirement Rel. 8 SU-MIMO 2.14 (100%) (100%) 1.61 (100%) (100%) Advanced MU-MIMO (non-codebook-based) 2.9 (136%) (128%) 2.4 (149%) (132%) Table 1. FDD downlink spectral efficiency of 4 2 MIMO setup in urban micro (UMi) and urban macro (UMa) environments. The antennas at the base station are co-polarized and closely spaced (Fig. 1). cost in the user s hardware. This is particularly due to the large power variation of multicarrier signals that reduces the amplifier efficiency or requires a more expensive amplifier with large dynamic range. In Release 10, discrete Fourier transform (DFT)-precoded orthogonal frequency-division multplexing (OFDM) has been selected as the uplink multiple access transmission scheme for its inherently lower power variations than regular OFDM. It is an extension of the Release 8/9 single-carrier FDM (SC-FDM) concept where the frequency resource assignment is now allowed to be non-contiguous. LTE Rel. 10 users can support both two and four transmit antennas with up to four independent layers. Closed-loop MIMO is used for the uplink in a similar way to downlink MIMO: rank and precoding matrix are fixed by the base station and signaled to the user by means of a grant over the control channel. There are different codebooks for two and four transmit antennas, requiring 3 and 6 bits to signal, respectively. The precoding codebooks are designed based on several major considerations including but not limited to PMI searching complexity, cubic metric (a measure of the power variation impact on the amplifier efficiency) property, and quantization accuracy/efficiency. In particular, all the non-zero elements in a precoding codeword are constrained to be QPSK and thus have the same modulus. The benefit of this constraint is twofold. First, it reduces the computational complexity required for the precoding codeword selection by simplifying complex multiplications with real summations. Second, it enables the power amplifiers to transmit with the same power level, which is an essential property to maximize the amplifier efficiency. Another critical design constraint is the need of not increasing the cubic metric with respect to single-layer transmissions. This is achieved by allowing each antenna to transmit no more than one layer; that is, by avoiding linear combinations of different layers over the same antenna. Another important factor considered in the codebook design is quantization accuracy/efficiency. Due to the fact that users are typically operating in an ample scattering environment, it would be reasonable to consider the maximization of the minimum chordal distance as a metric to design the UL codebook. In fact, as the chordal distance criterion does not necessarily yield to a constellation with constant modulus, the UL codebook in Rel. 10 is actually designed as a trade-off between chordal distance minimization and constant modulus. Antenna turn-off is also supported by the UL codebook to accommodate the cases when the antenna gain imbalance is present and/or power saving by turning off some of the power amplifiers is desired. The rank-1 codewords are also designed to improve the MU-MIMO operations performed at the base station side. Transmit diversity is not used for the (data) uplink shared channel. This decision was based on simulation studies showing that transmission diversity-based multiantenna schemes were outperformed by long-term closed-loop precoding, in the meaningful SN region. On the other hand, in Rel. 10 transmit diversity is used for the uplink control channel. PERFORMANCE ASSESSMENT In this section we discuss the performance of the LTE-A technologies based on features presented in the earlier sections. The performance results are shown for some key scenarios that have been deemed typical for a deployment. The performance is assessed according to the three criteria explained earlier, which include peak spectral efficiency, cell average spectral efficiency, and cell-edge user spectral efficiency. DOWNLINK The downlink peak spectral efficiency of LTE-A can be found in [5]. It is shown that LTE Rel. 8 already fulfills the ITU-R requirement in terms of downlink peak spectral efficiency and further performance improvement is possible by the eight-layer spatial multiplexing in Rel. 10. With respect to cell spectral efficiency and cell edge user spectral efficiency, LTE Rel. 8 SU-MIMO is able to meet the downlink ITU-R requirements for 4 2 MIMO setup in all scenarios except the base coverage urban and microcellular scenarios, which use the urban macro (UMa) and urban micro (UMi) channel models, respectively, as mentioned earlier. Some selected results from [5, 6] are summarized in Table 1. Table 1 shows that the advanced noncodebook-based MU-MIMO precoding scheme based on user-specific reference signals enables the ITU-R requirements to be satisfied and pro- 118 IEEE Communications Magazine March 2012

6 Scheme Closely spaced cross-polarized setup at the base station Cell spectral efficiency (b/s/hz/cell) Cell edge user spectral efficiency (b/s/hz/user) Closely spaced co-polarized setup at the base station Cell spectral efficiency (b/s/hz/cell) Cell edge user spectral efficiency (b/s/hz/user) Rel SU-MIMO 2.41 (100%) (100%) 2.23 (100%) (100%) 4 2 SU/MU-MIMO dynamic switching 8 2 SU/MU-MIMO dynamic switching 2.56 (106%) (117%) 3.19 (143%) (131%) 3.72 (154%) (177%) 4.38 (196%) (175%) Table 2. FDD downlink spectral efficiency in 3GPP Case 1 (urban macro) scenario with angular spread of 8. vides significant gain compared to the 4 2 SU- MIMO scheme in LTE Rel. 8. In addition to the ITU-R requirements, 3GPP has also defined its own performance target [4], which is more aggressive than the ITU-R requirements. As explained earlier, the new CSI feedback codebook has been designed and adopted for supporting eight transmit antennas at the base station to further enhance the spectral efficiency. Table 2 shows the downlink spectral efficiency in 3GPP case 1 representing a typical urban macro environment with lower angular spread of 8 at the base station. The feedback information is assumed wideband for RI and PMI and subband for CQI. For Rel. 8 SU-MIMO, the cross-polarized setup leads to better cell spectral efficiency than the co-polarized one due to the less correlated nature, which is beneficial for the spatial multiplexing, whereas better cell edge user spectral efficiency is achieved with the co-polarized setup since the higher correlation can be exploited as better array gain. The advanced dynamic SU/MU- MIMO switching mechanism enables more flexible resource utilization 2 leading to significant performance improvements in both cell average and cell edge user spectral efficiencies, which are shown to be further enhanced by the use of eight antennas with the new double-codebook described earlier. UPLINK In LTE Rel. 10 several enhancements were incorporated in the uplink to deliver a more robust and efficient design, as discussed earlier. As already mentioned and different from the downlink case, LTE Rel. 8 did not meet the ITU-R requirements for peak spectral efficiency, and additional features such as multilayer transmission and codebook-based precoding were introduced to increase the peak data rate and spectral efficiency. Unlike the downlink case, it is hard to define a single spectral efficiency (for average and cell edge) in uplink since this is closely tied to the power control algorithm used. LTE Rel. 10, like its predecessor, allows the power control algorithm to use full or partial path loss compensation. Full path loss compensation allows the cell edge performance to be improved at the cost of causing high interference to the neighbor cells; and similarly, partial path loss compensation reduces the interference to the neighbor cells at the cost of cell edge performance. The uplink results presented in this section should be taken with the caveat that they represent full path loss compensation. Table 3 shows the average and cell edge spectral efficiencies of LTE Rel. 8 and LTE Rel. 10 for the UMi and UMa environments [5, 6] with four receive antennas at the enodeb. These results indicate that features such as multilayer transmission and codebook-based precoding provide significant improvement for both the average and cell edge spectral efficiencies of the system. Table 4 shows the spectral efficiency performance of LTE Rel. 8 and LTE Rel. 10 for 3GPP case 1. Unlike the downlink case, similar performance is observed with cross-polarized and copolarized antenna setups. FUTURE DEVELOPMENTS The Rels MIMO design mainly targets a homogenous macro base station deployment. Future investigations may consider MIMO in a heterogeneous network context, where layers of macro base stations, lower-power base stations, femtocells, and relays will interact. In such a scenario, different antenna configurations, propagation conditions, deployment types, use cases, and user behaviors will require a more flexible design of codebooks, reference signals, and feedback mechanisms. The case of indoor low-mobility users is an interesting area for further investigation. The low-mobility condition may provide better exploitation of closed-loop MIMO schemes by allowing more refined CSI to be accessible at the transmitter via enhanced feedback schemes. Antenna configurations play another critical role. For example, downlink codebook enhancements in Rel. 10 consider the case of eight transmit antennas, whereas for two and four transmit antennas Rel. 8 codebooks are still used. Future enhancements should focus on antenna deployments of high priority for network operators, such as two and four cross-polarized antenna elements. Another interesting area as part of future MIMO investigations will be to learn from practical experience in current and future Rel deployments using multiple antennas. New 2 In this context the word flexible has a broad meaning. For example, flexibility could refer to the possibility of tightly coupling the MIMO scheme to the scheduler. Or it could refer to the use of advanced user-pairing techniques. More in general, the way resources are utilized is a matter of the specific base station implementation; therefore, it is not defined in the standard. IEEE Communications Magazine March

7 Scheme Cell Spectral Efficiency [bps/hz/cell] Urban Micro (UMi) Cell Edge Spectral Efficiency [bps/hz/cell] Cell Spectral Efficiency [bps/hz/cell] Urban Macro (UMa) Cell Edge Spectral Efficiency [bps/hz/cell] ITU-R Requirements Rel SIMO 1.93 (100%) (100%) 1.55 (100%) (100%) Rel SU-MIMO 2.20 (114%) (111%) 1.78 (115%) (110%) Table 3. FDD uplink spectral efficiency for a 2 4 set-up in Urban Macro (UMa) and Urban Micro environments (UMi). The antennas are widely spaced co-polarized (Fig. 1). Scheme Closely-spaced cross-polarized set-up Cell Spectral Efficiency [bps/hz/cell] Cell Edge Spectral Efficiency [bps/hz/cell] Widely-spaced co-polarized set-up Cell Spectral Efficiency [bps/hz/cell] Cell Edge Spectral Efficiency [bps/hz/cell] Rel SIMO 2.00 (100%) (100%) 1.95 (100%) (100%) Rel SU-MIMO 2.27 (114%) (121%) 2.36 (121%) (129%) Table 4. FDD uplink spectral efficiency for 3GPP Case 1 (urban macro) scenario with angular spread of 8 degrees. challenges may arise, and further tuning of functionality may be necessary to further improve the performance in real-world network deployments. Interesting aspects important for efficient MIMO design include assumptions about antenna calibration, link adaptation and traffic patterns. Finally, MIMO investigations should take into account the possible interactions between MIMO and CoMP techniques, where antennas of multiple cell sites (also called transmission points) are utilized in such a way that they can contribute in improving the received signal quality. Issues such as efficient switching between single-point and multipoint transmissions/receptions, enhanced feedback techniques carrying the channel state information concerning multiple transmission points, and enhanced codebook design approaches are expected to play an important role in future investigations. CONCLUSIONS We discuss the different multiple antenna techniques adopted in LTE-Advanced. After describing the design targets and antenna deployments prioritized by the operators, we present the principles behind downlink and uplink multi-antenna techniques, and we assess the performance of these techniques in different scenarios. Finally, we discuss some possible future developments. ACKNOWLEDGMENT We would like to thank our 3GPP RAN 1 colleagues for the passionate discussions and the long hours spent together during the LTE- Advanced standardization works. REFERENCES [1] Q. Li et al., MIMO Techniques in WiMAX and LTE: A Feature Overview, IEEE Commun. Mag., vol. 48, no. 5, May 2010, pp [2] N. Jindal, MIMO Broadcast Channels with Finite-Rate Feedback, IEEE Trans. Info. Theory, vol. 52, no. 11, July 2006, pp [3] ITU-R M.2134, Requirements Related to Technical Performance for IMT-Advanced Radio Interface(s), Nov [4] 3GPP TR36.913, Requirements for Further Advancements for Evolved Universal Terrestrial Radio Access (E- UTRA) (LTE-Advanced), v , Mar [5] 3GPP TR36.912, Feasibility Study for Further Advancements for E-UTRA (LTE-Advanced), v9.3.0, June [6] 3GPP TR36.814, Evolved Universal Terrestrial Radio Access (E-UTRA); Further Advancements for E-UTRA Physical Layer Aspects, v , Mar BIOGRAPHIES FEDERICO BOCCARDI (Federico.Boccardi@alcatel-lucent.com ) received M.Sc and Ph.D. degrees in telecommunication engineering from the University of Padova, Italy, in 2002 and 2007, respectively. He has been a visiting researcher in EURECOM and Bell Labs US, Lucent Technologies. Since 2006 he has been with Bell Labs, Alcatel-Lucent. His research interests include PHY and MAC layer design in mobile networks. In 2010 and 2011 he participated in the 3GPP activity for LTE-Advanced. BRUNO CLERCKX received M.S. and Ph.D. degrees in applied science from Université Catholique de Louvain. He held visiting research positions at Stanford University and EURE- COM, and was with Samsung Electronics from 2006 to He actively contributed to 3GPP LTE/LTE-A and IEEE802.16m, and acted as Rapporteur for the 3GPP CoMP Study Item. He is now a lecturer (assistant professor) at Imperial College London and an editor for IEEE Transactions on Communications. ARUNABHA GHOSH received his B.S. from the Indian Institute of Technology Kanpur in 1992 and his Ph.D. from the University of Illinois at Urbana-Champaign in He is currently a lead member of technical staff in AT&T Labs. His main area of research is in MIMO wireless communication systems and has worked extensively on technologies such as LTE and WiMax. He is also the co-author of the best selling book Fundamentals of LTE. ERIC HARDOUIN [S 02, M 05] received a Ph.D. degree in signal processing and telecommunications from the University of Rennes I, France, in Since 2004 he has been with Orange Labs, where he has conducted or supervised research on physical layer and physical MAC layer techniques for interference mitigation in mobile networks. 120 IEEE Communications Magazine March 2012

8 Since 2008 he has been representing Orange in the physical layer standardization group of 3GPP (RAN WG1) for HSPA, LTE, and LTE-Advanced. GEORGE JÖNGREN received M.Sc. and Ph.D. degrees in electrical engineering from the Royal Institute of Technology (KTH), Stockholm, Sweden, in 1998 and 2003, respectively. He was a postdoctoral researcher at KTH before joining Qualcomm CDMA Technologies GmbH, Nuremberg, Germany. Since September 2005 he has been with Ericsson Research, Sweden, working on smart antenna research, implementation, and 3GPP standardization. He was elected Teacher of the Year in Electrical Engineering at KTH in KATSUTOSHI KUSUME [M 05] received M.Sc. and Dr.-Ing. degrees from Munich University of Technology, Germany, in 2001 and 2010, respectively. In 2002 he joined DOCOMO Euro-Labs, Munich, Germany. In he was a 3GPP delegate in the area of downlink MIMO transmission. Since June 2011 he has been with the Radio Access Network Development Department of NTT DOCOMO Inc., Yokosuka, Japan. He received the best paper award of the IEEE Global Telecommunications Conference in EKO ONGGOSANUSI received his Ph.D. degree in electrical engineering from the University of Wisconsin-Madison in He joined Texas Instruments in 2001 where he has been working on various projects in wireless communication systems. He has been a 3GPP RAN1 delegate since 2005 and the lead for TI team since Currently he is the systems-and-standardization manager and a senior member of technical staff at the Wireless Basestation Infrastructure business unit at Texas Instruments. YANG TANG received his B.E. from Beijing University of Aeronautics and Astronautics in 1999, and M.S. and Ph.D. degrees from the University of Sydney in 2002 and 2006, respectively, all in electrical engineering. He has actively contributed to both 3GPP LTE-A and IEEE802.16m since 2006, when he was with Samsung Electronics and Huawei(USA). He is now a staff engineer at Mobile Solution Lab in Samsung Information Systems America. IEEE Communications Magazine March