Mimo lte advanced


Published on

Published in: Business, Technology
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Mimo lte advanced

  1. 1. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® TOPICS IN RADIO COMMUNICATIONS Recent Trend of Multiuser MIMO in LTE-Advanced Chaiman Lim, Korea University Taesang Yoo, Qualcomm Incorporated Bruno Clerckx, Imperial College, London Byungju Lee and Byonghyo Shim, Korea University ABSTRACT 2009. Standardization of the latest release (Rel. 10), popularly called LTE-Advanced (LTE-A), Recently, the mobile communication industry was closed in June 2011, and LTE Rel-11 is cur- is moving rapidly toward Long Term Evolution, rently under development. or LTE, systems. The leading carriers and ven- Among many features in the LTE-A, which dors are committed to launching LTE service in supports up to 3 Gb/s throughput in downlink, the near future; in fact, a number of major oper- the multiuser multiple-input-multiple-output ators such as Verizon have initiated LTE service (MU-MIMO) scheme has been identified as one already. LTE aims to provide improved service of the key enablers for achieving high spectral quality over 3G systems in terms of throughput, efficiency [3]. From both the theory and design spectral efficiency, latency, and peak data rate, perspectives, MU-MIMO systems have several and the MIMO technique is one of the key unique features distinct from single-user MIMO enablers of the LTE system for achieving these (SU-MIMO) systems. Although both systems diverse goals. Among several operational modes provide spatial multiplexing gains that are effec- of MIMO, multiuser MIMO (MU-MIMO), in tive in improving the throughput, cell coverage, which the base station transmits multiple streams and reliability of mobile communication systems, to multiple users, has received much attention as the SU-MIMO systems are vulnerable in scenar- a way of achieving improvement in performance. ios where the spatial multiplexing capability of a From the initial release (Rel. 8) to the recent single user’s channel is limited either due to sig- release (Rel. 10), so called LTE-Advanced, MU- nal-to-interference-plus-noise ratio (SINR) or MIMO techniques have evolved from their pre- fading correlations among antenna elements or mature form to a more elaborate version. In this by the number of receive antennas at the user article, we provide an overview of design chal- side. To make up for the shortcomings of SU- lenges and the specific solutions for MU-MIMO MIMO, early LTE standards (Rels. 8 and 9) systems developed in the LTE-Advanced stan- defined a primitive form of the MU-MIMO dard. mode. The operations of MU-MIMO are elabo- rated and enhanced in the LTE-A standard INTRODUCTION (Rel. 10). Our purpose in this article is to pro- vide an overview of the key features of MU- The recent explosion of smart phone users and MIMO systems defined/adopted in the LTE and ever increasing demand for high-quality multi- LTE-A standards. media over wireless are fueling the deployment of Long Term Evolution (LTE) mobile commu- KEY DISTINCTIONS BETWEEN MU-MIMO AND nication systems [1, 2]. Initiated in 2004, the SU-MIMO SYSTEMS LTE standardization effort focuses on enhancing Universal Terrestrial Radio Access (UTRA) and Information theory highlights some key aspects optimizing the Third Generation Partnership of MU-MIMO systems over SU-MIMO systems. Project’s (3GPP’s) radio access architecture. Ini- First, it is well known that when the channel tial design targets of the LTE were to have the matrix is full rank, the capacity gain of SU- average user throughput be about three times MIMO systems is scaled by min {M, N} at high that of Release 6 high-speed downlink packet signal-to-noise ratio (SNR), where M and N are access’s (HSDPA’s) in the downlink (100 Mb/s) the number of antennas at the transmitter and and about three times that of the HSUPA’s in the receiver, respectively. In typical cellular sys- the uplink (50 Mb/s). Its initial release (Rel. 8), tems, the number of receive antennas at the bat- which has been the basis for the LTE standard, tery powered user (user equipment; UE) is often was finalized in December 2008, and a subse- smaller than the number of transmit antennas at quent release (Rel. 9) was frozen in December the base station (enhanced node-B; eNB), limit- IEEE Communications Magazine • March 2013 0163-6804/13/$25.00 © 2013 IEEE 127C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  2. 2. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® accuracy directly affects the multiplexing gain of LTE standard MU-MIMO terminology the MU-MIMO downlink [4]. Another important problem related to the CSI feedback is how to Reference signal (RS) Pilot signal determine channel quality information (CQI) to be fed back. Unlike SU-MIMO systems, UE in User equipment (UE) User terminal, mobile MU-MIMO cannot accurately estimate its SINR, due to the lack of knowledge of eNB’s beam Enhanced NodeB (eNB) Base station selections for co-scheduled UE devices. Such knowledge is typically unavailable to the UE at Precoding matrix indicator (PMI) Precoding codebook index the time of CQI calculation, as it is a function of the eNB’s scheduling decision that will in turn be Rank indication (RI) MIMO channel rank dependent on the UE’s CSI feedback. In the rest of this article, we provide an Physical downlink shared channel (PDSCH) Downlink data channel overview of MU-MIMO techniques adopted in LTE and LTE-A. In particular, we summarize Physical downlink control channel (PDCCH) Downlink control channel features of MU-MIMO and the rationale behind them defined in Rel. 8/9 LTE. We also devote a Table 1. The nomenclature of MU-MIMO related terminologies in the LTE section to MU-MIMO techniques adopted in standard. LTE-A as well as the technical grounds behind each decision. Finally, we conclude the article. Since the terminologies of LTE standards might ing the gain of SU-MIMO systems to the num- be unfamiliar to the reader, we put a summary ber of antennas at the receiver. In contrast, of commonly used technical terms and their under the assumption that the transmitter has mappings to LTE standards in Table 1. perfect channel state information at the trans- mitter (CSIT), the sum capacity of MU-MIMO is scaled by min {M, Nn}, where n is the number MU-MIMO IN RELEASE 8/9 LTE of users multiplexed into the MU-MIMO trans- LTE Rel. 8 supports a primitive form of the mission, so an M-fold increase in the sum rate MU-MIMO as a direct extension of its SU- can be obtained as long as Nn is larger than M. MIMO closed-loop spatial multiplexing mode. In Second, it has been shown that SU-MIMO a nutshell, the MU-MIMO in Rel. 8 LTE systems are susceptible to the poor behavior of employs a codebook-based precoding, wherein propagation channels such as a strong line of each UE device measures its spatial channel sight (LOS) path or strong correlations among using the common RS (CRS) broadcasted from antenna elements. In such cases, the effective the eNB, selects the rank-1 precoder that best rank of the channel matrix decreases, and so represents the channel from a predefined code- does the capacity gain of SU-MIMO. On the book set, and then feeds back the index (PMI) other hand, with the additional degrees of free- of the codebook as well as the resulting CQI to dom introduced by the co-scheduling of multiple the eNB via an uplink channel. The eNB collects UE devices, MU-MIMO systems can be designed the PMI and CQI reports from different UE to be less sensitive to the poor behavior of the devices and performs user grouping onto the channel matrix, allowing MU-MIMO systems to given time/frequency resource such that the achieve the full multiplexing gain regardless of paired UE devices experience a minimum level the effective rank of each individual user. of inter-stream interference among themselves. It is important to note that as far as the feed- CHALLENGES OF MU-MIMO SYSTEMS back is concerned, the UE does not differentiate From a theoretical perspective, one of the major MU-MIMO from SU-MIMO. That is, the CQI benefits of MU-MIMO systems is the realization at the UE is computed based on the assumption of the full spatial multiplexing gain. Supporting that there is no inter-stream interference caused MU-MIMO operations, however, requires by concurrent transmission from the eNB to a appropriate resource allocation, in particular, co-scheduled user. Since the CQI generated in among users. The aim of the user allocation is to this way would be a bit optimistic, the eNB may choose a group of UE devicess that, when co- need to apply a backoff to the reported CQI to scheduled, maximizes a predefined performance choose a right modulation order and code rate metric (e.g., the sum throughput). To achieve for the UE. On top of this, an additional 3 dB 1 In general, it may be this objective, channel state information (CSI) of backoff is required as the power needs to be beneficial for the eNB to UE should be fed back to the eNB. A practical shared by the two UE devices. choose a precoding matrix problem related to the CSI feedback is the bal- Release 8 MU-MIMO (a.k.a transmission outside of the predefined ancing of the feedback accuracy and the uplink mode 5) does not define a dedicated pilot. codebook, an approach resource consumption caused by the feedback. Hence, the UE should rely on the CRS for chan- also known as non-code- Since feeding back real numbered CSI would nel estimation and demodulation of data. As the book precoding. For require an infinite number of bits, the CSI needs CRS is un-precoded, the UE should explicitly example, the eNB may to be represented (e.g., via appropriate quantiza- derive the precoded channel by multiplying the want to use zero-forcing tion) with a limited number of bits. Note that estimated channel with the precoding informa- beams computed by while the accuracy of the CSI affects the SNR tion separately signaled to the UE through a inverting the composite offset but not the multiplexing gain in SU- control channel. A drawback of this approach is channel matrix construct- MIMO systems, such is not the case for MU- that the eNB’s choice of precoding matrix is con- ed by stacking the rank-1 MIMO systems. In fact, since MU-MIMO fined to those defined in the codebook. 1 This channel vectors corre- systems are interference limited under the finite- limitation is relaxed in Rel. 10, as discussed in sponding to PMIs. rate feedback constraint, the level of feedback the next section. 128 IEEE Communications Magazine • March 2013C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  3. 3. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Release 8 MU-MIMO relies on a single wide- Rel. 9, is extended in Rel. 10 to support up to band PMI reporting per UE, representing the rank 8 transmissions in new transmission modes. To make up for the average channel direction over the entire system The CSI-RS, newly introduced in Rel. 10, is bandwidth. Such a lack of frequency-selective used for CSI measurements in the new transmis- shortcomings of PMI reporting essentially limits the usefulness of sion modes. Second, dynamic switching between SU-MIMO, early LTE MU-MIMO to cells with a highly correlated uni- SU-MIMO and MU-MIMO is adopted. With standards (Rels. 8 form linear array (ULA) antenna configuration, the use of DM-RS, an eNB can flexibly switch in which UE devices typically experience rank-1 the MIMO operation mode of UE with no need and 9) defined a channels with rather non-frequency-selective to report the precoding information to the UE. primitive form of the eigenbeam directions. This helps the eNB to promptly respond to the Also, in Rel. 8, the SU-MIMO and MU- variations in the channel and system conditions MU-MIMO mode. MIMO modes are semi-statically switched. such as the traffic type and the number of UE The operations of Although a dynamic switching between two devices. Third, as an effort to reduce feedback MU-MIMO are modes may in principle allow more flexibility in load, a dual codebook structure is adopted for the scheduling and increase multiuser diversity, the 8Tx configuration, where one codebook tar- elaborated and the lack of MU-MIMO-specific CQI reporting gets capturing wideband and long-term channel enhanced in the and the limited precoding mechanism in Rel. 8 properties, while the other is designed to capture diminish the benefit of dynamic switching. In frequency-selective and/or short-term channel LTE-A standard fact, Rel. 8 MU-MIMO is seen as beneficial only properties. (Rel. 10). in certain scenarios (i.e., for heavily loaded cells with highly correlated ULA antennas). Consider- REFERENCE SIGNAL DESIGN ing these factors and also design simplicity, semi- Over the past years, practical schemes achieving static switching was chosen in Rel. 8. a fair amount of capacity yet requiring reason- In addition to CRS-based transmissions, Rel. able precoding cost have been suggested [6, 7]. 9 supports rank-2 transmissions (called dual It is widely known that the gain of these tech- layer beamforming) using the demodulation RS niques over unitary precoding is considerable. (DM-RS). Unlike the CRS, which is common to This gain, however, is guaranteed only when the all users in a cell and hence cannot be precoded, accuracy of the CSI feedback is high enough. To the DM-RS is UE-specific, that is, dedicated to maintain a reasonable CSI estimation quality each UE and precoded along with accompanying while avoiding the excessive pilot overhead, data. Note that when the pilot is precoded, the instead of a simple extension of CRS, Rel. 10 precoder information does not need to be sepa- has adopted a separate DM-RS/CSI-RS rately delivered to the UE. In this sense, the use approach. of precoded DM-RS makes the operation of the The main features of the reference scheme in eNB transparent to the UE and enables the use Rel. 10 are as follows: of non-codebook-based precoding. In the dual • CRS is defined for only up to four transmit layer beamforming, an orthogonal cover code antennas to limit the pilot overhead. (OCC) of length two is used to differentiate • In the new transmission mode of Rel. 10 pilots on two layers. In addition, by defining two (transmission mode [TM] 9), the DM-RS is scrambling IDs, the Rel-9 dual layer beamform- used for data demodulation of up to rank 8. ing can support up to four-stream multiplexing, • In TM 9, CSI measurements are performed wherein the two groups are differentiated by dis- using the CSI-RS. The CSI-RS is defined tinct scrambling IDs, and the two layers within for up to eight transmit antennas but with each group are separated by an OCC. This much lower overhead compared to the allows the eNB to schedule two UE devices CRS. simultaneously in MU-MIMO mode, with two • In legacy transmission modes, which can layers assigned for each UE.2 support up to rank 4, channel estimation and CSI measurements are performed using the CRS. MULTIUSER MIMO IN Note that the DM-RS is not appropriate for LTE-ADVANCED CSI measurements as the DM-RS is precoded in a UE-specific manner. That is, the DM-RS is BRIEF OVERVIEW present only on time/frequency resources where LTE-A systems should meet various require- the UE is scheduling data. CSI-RS, on the other ments of the International Telecommunication hand, is unprecoded and defined over the entire Union Radiocommunication Standardization system bandwidth for the CSI measurement. In Sector (ITU-R), including peak, average, and order to minimize overhead, CSI-RS is transmit- cell edge spectrum efficiencies [5]. Several new ted in a fraction of subframes (Fig. 1). This is in features have been introduced in Rel. 10 to contrast to the CRS, which is used for both achieve these goals. First, in order to support demodulation and CSI measurements, and there- high-dimensional SU-MIMO operation (up to 8 fore needs to be transmitted every subframe. ¥ 8 MIMO) and also improve MU-MIMO oper- Unlike CRS, the DM-RS needs to be trans- ation, new RSs have been introduced in the mitted only for the active spatial layers where downlink. There are three different kinds of RSs data is present. Thus, the DM-RS is advanta- — CRS, DM-RS, and channel state information geous in terms of the pilot overhead if the rank (CSI)-RS. The CRS is used for CSI measure- of the data transmission is less than the number ments and demodulation in legacy (Rels. 8 and of transmit antennas, which is typically the case 9) transmission modes, as well as for control in 8-Tx antenna deployments. In fact, the DM- 2However, the DM-RS channel decoding, and various measurements RS patterns in Rel. 10 are optimized for each between two groups is only and UE procedures. The DM-RS, introduced in rank to minimize the pilot overhead. quasi-orthogonal. IEEE Communications Magazine • March 2013 129C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  4. 4. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® The DM-RS patterns for ranks 3–8 in Rel. 10 • No more than four layers are transmitted in Unlike CRS, the DM- are natural extensions of that for rank 2 in Rel. total. 9 and are based on hybrid code-/frequency-divi- Multiplexing four UE devices implies that RS needs to be sion multiplexing; for ranks 1 and 2, the DM-RS each device receives only a quarter of the total transmitted only for patterns for the first and second layers are multi- transmit power and that the devices are also sub- the active spatial lay- plexed by code-division multiplexing (CDM), ject to substantial inter-user interference. Alloca- while for ranks 3 and 4, DM-RS patterns for the tion of four UE devices would be viable in a ers where data is first and second layers and for the third and situation where the cell contains a large number present. Thus, the fourth layers are multiplexed by frequency-divi- of UE devices with high SNRs, and the rate gain sion multiplexing (FDM). Hybrid CDM+FDM due to the spatial multiplexing and multiuser DM-RS is advanta- DM-RS patterns are adopted for ranks 5–8 with diversity outweighs the per-UE rate loss caused geous in terms of two CDM groups. by the power splitting and increased interference. the pilot overhead if Since scheduling more than four UE devices in MU-MIMO DIMENSIONING MU-MIMO results in overly complicated the rank of the data Considering the trade-off between performance scheduling and UE implementation, and further- transmission is less and signaling overhead, Rel. 10 has made the more such a scenario is uncommon in real following decisions on the dimensioning of MU- deployments, this case will be rare in practice. than the number of MIMO systems: For example, four layers may be able to be transmit antennas, • No more than four UE devices are co- supported in single-pole ULA systems, where which is typically the scheduled. spatial separation between UE devices is easier. • No more than two layers are allocated per However, in a realistic deployment with dual- case in 8 Tx antenna UE. polarized arrays, it is very hard to control the deployments. RE (resource element) Time 8 DM-RS ports Rank 8 Port 7 Port 8 Port 11 Port 12 (1111) (1-11-1) (11-1-1) (-1-111) Layer 1/2/5/6 12 subcarrier Port 9 Port 10 Port 13 Port 14 Layer 3/4/7/8 (1111) (1-11-1) (1-1-11) (-111-1) OCC (orthogonal cover code) mapping for DM-RS 7 OFDM symbol (0.5 ms) 1 RB (resource block) (a) 0 1 4 5 0 1 0 1 0 1 4 5 4 5 4 5 0 1 4 5 2 3 6 7 2 3 2 3 2 3 6 7 6 7 6 7 2 3 6 7 (b) CRS port #1,2 DMRS(rel.9/10) DRS (dedicated RS, rel.8) port #5, if configured CRS port #3,4 DMRS(rel.10) PDCCH PDSCH Figure 1. DM-RS and CSI-RS patterns. 130 IEEE Communications Magazine • March 2013C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  5. 5. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® CQI/PMI/RI To maximize performance using SU-MIMO the advanced UE 1 Packet data eNB receiver, information about DM-RS ports, number of Dynamic UE 2 switching Precoding co-scheduled UEs, Scheduler between W MCS (modulation SU/MU coding scheme) and UE K rank of each UE are MU-MIMO recommended in the K UEs index {1,2,⋅⋅⋅,K} MUEs downlink control signalings. DCI for dynamic switching Figure 2. SU/MU-MIMO dynamic switching. MU-MIMO interference at the eNB with the possible. This limits the scheduler flexibility and feedback mechanism defined in Rel. 10. Hence, may compromise the performance of MU- for systems with dual-polarized arrays, the num- MIMO systems. ber of co-scheduled users will be two for most Control-signaling overhead: While transpar- cases. ent MU-MIMO requires no additional control With a highly correlated ULA, the eNB may signaling regarding co-scheduled UE devices, direct two streams separately to two UE devices non-transparent MU-MIMO requires more and thereby obtain the spatial multiplexing bene- PDCCH resources to support user multiplexing fits. This is particularly true in a highly loaded and an advanced receiver operation of the UE. scenario where the increase in the spatial degrees In transparent MU-MIMO, UE may have to of freedom can be achieved by multi-user resort to blind detection techniques to use scheduling. Hence, a rank-1 transmission per UE advanced receiver algorithms. Also, UE com- may be the most common scenario for this case. plexity might be substantial to support func- The gain of having more than rank 1 is negligible tions such as modulation classification and in ULA but beneficial in dual-polarized systems. enhanced channel estimation. However, if its For example, in dual-polarized 8-Tx deploy- own DM-RS ports and co-scheduled DM-RS ments, each polarization array can form a beam ports are delivered to the UE, channel informa- and direct a data stream to a UE device, result- tion for the desired layers and the interfering ing in a rank-2 transmission per user. layers can easily be estimated by the orthogonal DM-RS ports, and thereby an advanced receiv- TRANSPARANCY OF MU-MIMO er technique (e.g., interference rejection com- In transparent MU-MIMO mode, the UE can bining or nonlinear interference cancellation) perform decoding with only its own control sig- can be employed. In summary, to maximize nal information (e.g., its own rank and DM-RS performance using the advanced receiver, infor- port). In contrast, information on co-scheduled mation about DM-RS ports, number of co- UE devices needs to be provided in the non- scheduled UE devices, MCS (modulation transparent MU-MIMO mode. The co-schedul- coding scheme), and rank of each UE device ing information includes the total rank and the are recommended in the downlink control sig- DM-RS ports of co-scheduled UE devices. Key nalings. Considering the limited budget for distinctions between two modes and their impli- PDCCH, the increase in control information cations on the system design are as follows. overhead is undesirable and may put a more Scheduling flexibility: The transparent MU- stringent limit on the number of co-scheduled MIMO allows for more flexible scheduling. For UE devices per subframe. example, under the transparent MU-MIMO, To address this issue, Rel. 11 is discussing resources (RBs) allocated for UE multiplexed in enhanced PDCCH on the PDSCH region. MU-MIMO do not need to be fully aligned; the After weighing these pros and cons, Rel. 10 eNB may allocate partially overlapping has decided to support transparent MU-MIMO. resources to the UE devices and may even mul- Additional signalling issues will be discussed in tiplex different numbers of users on different the upcoming releases. RBs. Also, DM-RS ports of co-scheduled UE can be determined more dynamically for each SU/MU-MIMO DYNAMIC SWITCHING RB. On the other hand, under the non-transpar- While SU-MIMO systems improve the peak/ ent MU-MIMO, the eNB may have to align average UE throughput by transmitting multiple resource allocations of co-scheduled UE devices streams in the same time/frequency band, MU- so that a single signaling of the total rank and MIMO systems provide higher peak/average sys- the DM-RS ports of co-scheduled UE devices is tem throughput by exploiting multiuser diversity. IEEE Communications Magazine • March 2013 131C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  6. 6. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® implicit feedback indicates the mechanism of the One codeword: Two codewords: Rel. 8 feedback employing PMI, RI, and CQI. Codeword 0 enabled, Codeword 0 enabled, Considering its low feedback overhead, implicit Codeword 1 disabled Codeword 1 enabled feedback has been selected in Rel. 10. Nonethe- less, since explicit feedback provides better Message Message scheduling flexibility, this issue will continue to be discussed in the future LTE-A releases. 0 1 layer, port 7, SCID=0 0 2 layers, ports 7–8, SCID=0 Third, UE feedback design should take real deployment scenarios into account. Note that 1 1 layer, port 7, SCID=1 1 2 layers, ports 7–8, SCID=1 device space limitations do not allow a large spacing at the eNB, not to mention at the UE. 2 1 layer, port 8, SCID=0 2 3 layers, ports 7–9 One commonly used technique for reducing the physical space is to employ dual-polarized anten- 3 1 layer, port 8, SCID=1 3 4 layers, ports 7–10 nas, where the antennas are grouped into pairs to form polarized collocated antennas. 4 2 layers, ports 7–8 4 5 layers, ports 7–11 In the standardization process, many approaches to improve the Rel. 8 implicit CSI 5 3 layers, ports 7–9 5 6 layers, ports 7–12 feedback mechanism have been proposed [8]. We briefly describe key features of these 6 4 layers, ports 7–10 6 7 layers, ports 7–13 schemes. Adaptive codebook: The eNB can apply the 7 Reserved 7 8 layers, ports 7–14 transformed adaptive codebook on top of the fixed baseline codebook. UE may feed back an Table 2. Antenna port(s), scrambling identity, and number of layers indication. estimate of the correlation matrix or eigenvector to the eNB on a long-term basis. Any codebook may be used as the baseline codebook, so the To get the maximal benefit of both, Rel. 10 Rel. 8 codebook could also be reused for this introduced dynamic switching between the SU- approach. MIMO and MU-MIMO modes (Fig. 2). The Differential feedback: Differential feedback main feature of the dynamic switching is to enables the eNB to track channel variation change the mode per subframe basis, based on between adjacent feedbacks efficiently. Initially, the channel condition and traffic. the UE feeds back a full CSI as a reference. In order to support SU/MU-MIMO dynam- Next, the UE feeds back the difference between ic switching, as well as the rank adaptation in the previous CSI and the current one. The UE the SU-MIMO mode, information on DM-RS quantizes this difference using a differential antenna ports and the number of layers should codebook so that the index of the differential be delivered to the UE. For this purpose, a codebook is reported to the eNB. A well-known new downlink control information (DCI) for- drawback of differential feedback is the sensitivi- mat, DCI format 2C, has been defined. As ty to error propagation. Due to channel correla- shown in Table 2, DCI format 2C contains tion between two adjacent channel blocks, the antenna port information, the scrambling iden- difference can be delivered with improved accu- tity (SCID) of the DM-RS, and the number of racy or lower overhead. layers. In LTE-A, the eNB can transmit up to Multiple description coding (MDC): In every two codewords by spatially multiplexing them PMI report, more than one PMI generated from into up to eight layers. The boldface entries in different codebooks are provided to give differ- the table indicate message formats for MU- ent observations of the channel. By interpolating MIMO, and the rest indicate those for SU- different observations, a more accurate estimate MIMO. The UE dimensioning and layer of the channel can be obtained. While the differ- separation of Rel. 10 MU-MIMO is the same ential codebook exploits temporal correlation as that of Rel. 9 MU-MIMO discussed in the between adjacent CSIs, such is not the case for previous section. MDC since the collection of the PMI from dis- tinct codebooks is used. Therefore, even if one CSI FEEDBACK MECHANISM PMI is dropped, the eNB can still estimate the There are a number of factors that have been or channel by using the rest. will have to be considered in the design of the Best/worst companion PMI: This approach feedback mechanism in Rel. 10 and future performs an additional feedback for better co- releases. First of all, backward compatibility scheduling. Toward this end, both best and worst should be guaranteed; Rel. 10 UE moving into a PMIs are delivered. Moreover, best companion Rel. 8/9 network should be able to be configured report includes a delta CQI representing the to report feedback information (PMI, RI, and SNR loss induced by the addition of another CQI) according to the Rel. 8/9 format for seam- user on the same resource. Although additional less service. Likewise, PMI, RI, and CQI feed- overhead is relatively small, it is still non-negligi- back mechanisms need to be supported for Rel. ble, and the load of the scheduler to find the 8/9 UE in the Rel. 10 network. best UE pairs is considerable. Second, feedback overhead should be investi- Multirank PMI/CQI: This method reports gated carefully. Overall, there are two types of multiple PMI/CQI pairs with different ranks. Each feedback mechanisms — explicit feedback and feedback corresponds to SU-MIMO with any rank implicit feedback. Explicit feedback refers to the or MU-MIMO with low rank. This multiple feed- delivery of the channel matrix H and/or covari- back information expedites dynamic switching ance matrix HHH from the UE to the eNB, while between SU-MIMO and MU-MIMO. 132 IEEE Communications Magazine • March 2013C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  7. 7. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® 5 percent cell edge Average cell spectral ZFBF MU-MIMO 8 ¥ 2 with MAX 4 LAYERS spectral efficiency efficiency (b/s/Hz) (b/s/Hz) Extended Rel. 8 codebook (4-bit wideband) 3.26 0.14 Extended Rel. 8 codebook (4-bit subband) 3.30 0.14 Rel. 10 dual codebook (W1 wideband + W2 subband) 4.38 0.21 Adaptive codebook (unquantized Rt, report every 500 ms) 4.70 0.22 Table 3. 8 ¥ 2 antenna configuration (XXXX Æ ÔÔ channels, 0.5 l antenna spacing, 8° angle spread). DUAL CODEBOOK OPERATION Release 10 adopted the multigranular feedback 1 structure referred to as dual-stage feedback [8]. The precoder in dual-stage feedback consists of 0.9 two matrices (W 1 and W 2 ), and each of them belongs to a separate codebook. A composite 0.8 precoder is the product of the two and given by [9] 0.7 W = W1W2, (1) 0.6 CDF where W 1 is in charge of wideband and long- 0.5 term channel properties, and W 2 captures fre- quency selectivity and short-term channel fading. 0.4 A major advantage of decoupling the long-term and short-term CSI feedback components is low 0.3 feedback overhead and improved performance over the single codebook [10]. Below we provide 0.2 major principles behind the Rel. 10 codebook Extended Rel. 8 CB (wideband) Extended Rel. 8 CB (subband) design. Readers are referred to [11] for more 0.1 Rel. 10 dual CB details. Adaptive CB • The codebook design puts more emphasis 0 on lower ranks (ranks 1 and 2), where the 0 5 10 15 20 Average UE throughput (Mbp/) precoding gain is more pronounced. • All entries in a precoding matrix need to Figure 3. Average UE throughput for various codebook schemes. have the same magnitude (constant modu- lus) to relieve the burden on the power amplifier design. • Unitary precoding is preferred to maintain dual codebook, we compared the performance a constant average transmitted power. of Rel. 8 and Rel. 10 codebooks. • W 1 (= [X 0; 0 X}]) should be block diago- In our simulation, we assume that each UE nal since this structure is well matched to has a single layer and sends the feedback infor- the spatial covariance of the dual-polarized mation (RI, PMI, and CQI) to the eNB. When antenna setup with any antenna spacing the UE is requesting two ranks (2 PMI and 2 (e.g., l/2 or 4l) CQI), therefore, the eNB regards it as two • X is a 4 ¥ Nb matrix, where Nb is number of independent UE devices with a single layer beams (1~4). For each N 1, adjacent over- each. Using all feedback information from the lapping beams are used to reduce the edge UE, the eNB performs the MU-MIMO opera- effect in the frequency-selective precoding. tion. In our simulations, the following cases are • At least sixteen 8Tx discrete Fourier trans- investigated: form (DFT) vectors are generated from • Extended Rel. 8 codebook (CB): The UE W 1, and a co-phasing via W 2 is performed reports a 4-bit PMI. Since the codebook for in order to match with the spatial covari- 8Tx is not defined in LTE Rel. 8, we instead ance of the ULA antenna setup and to cope use the rank-1 codebook of 802.16m with the phase shift between polarizations. (WiMAX) [13], which is a unitary codebook In Rel. 10, the dual codebook structure is consisting of 16 elements. Note that Rel. 8 applied only on the 8Tx mode, and the 4Tx and employs wideband PMI only in MU-MIMO, 2Tx modes still employ the legacy codebook but we also test the subband PMI for the (Rel.-8-based codebook). According to the sake of comparison. recent study in Rel. 11, however, a dual code- • Release 10 dual codebook: The UE reports book scheme is beneficial for a 4Tx case such as 4-bit wideband PMI W1 and 4-bit subband cross-polarized macro sites and outdoor small PMI W2 [11]. cells with localized antennas [12]. • Adaptive codebook [14]: In this codebook, In order to observe the gain of the Rel. 10 W 1 is replaced by the channel covariance IEEE Communications Magazine • March 2013 133C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  8. 8. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Parameter Value While the differential codebook exploits Duplex method FDD temporal correlation Bandwidth 10 MHz between adjacent CSIs, such is not the Network synchronization Synchronized case for the MDC Cellular Layout Hexagonal grid, 19 cell sites, 3 sectors per site since the collection of the PMI from Users per sector 10 distinct codebooks is Downlink transmission scheme 8 ¥ 2 MU-MIMO ZFBF with rank adaptation with 1 layer per UE used. Therefore, even if one PMI is Downlink scheduler Proportional Fair scheduling in the frequency and time domain. dropped, the eNB CQI and PMI 5ms feedback period can still estimate 1 PMI and 1 CQI feedback per subband (=4 consecutive RBs) the channel by Downlink link adaptation 6ms delay total (measurement in subframe n is used in subframe n+6) using the rest. Unquantized CQI, PMI feedback error: 0% MCSs based on LTE transport formats [36.213] Allocation localized Maximum 3 retransmissions, IR, no error on ACK/NACK, 8 ms delay Downlink HARQ between retransmissions Downlink receiver type MMSE based on DM RS of serving cell Non-ideal channel estimation on CSI-RS and DM-RS vs. CINR curves Channel estimation based on LLS provided as an input to SLS. Antenna configuration XXXX->ÔÔ channels, 0.5l antenna spacing, 8° angle spread LTE: L = 3 symbols for DL CCHs Overhead of DM-RS: RANK 1, 2 : 12 REs/RB/subframe Control Channel overhead, Overhead of CSI-RS: 4/8 sets of CSI-RS every 5 ms and 1RE/port/RB Acknowledgements etc. (This is, in the 8Tx antenna case, 8 REs/RB per 5 ms) Overhead of 2-port CRS BS antenna downtilt Case 1 3GPP 3D: 15 degrees Channel model SCM urban macro high/low spread for 3GPP case 1, 3 km/h 7 intercell interference links are explicitly considered. Intercell interference modeling The remaining links are modeled with Rayleigh distribution. Max. number of layers (UE) 4 Table 4. System simulation assumptions. matrix R t = E[H H H]. That is, adaptive codebook consists of the long-term channel W = Fdiag(p )1/2 = H H (HH H ) 1 diag(P)1/2 ˆ ˆ ˆ (2) covariance matrix R t and W 2 . In order to observe the best possible performance, we where fk is the kth column of F and p = (p1, …, used unquantized matrix Rt. pK) is the vector of power normalization coeffi- Once the codebook cn is determined, we per- cients form zero-forcing beamforming (ZFBF) for each scheme. Let N and K be the number of transmit P 1 pk = antennas at the eNB and selected users, respec- K fk 2 tively; then the transmit signal after the linear precoding becomes x = Wu where W is an N ¥ K ZFBF precoding matrix, and u = {u1, …, uK)T Detailed simulation assumptions are provid- is the user symbol vector. ed in Table 4. In a real implementation, the ^ ^ Using the estimated channel matrix H = [ hT, 1 eNB usually uses cross-polarization considering ^T T ^ H is the channel response ..., h K] where hk = c n space limitation for 8Tx. On the other hand, of the kth UE, W is computed as because it is difficult to implement exact dual 134 IEEE Communications Magazine • March 2013C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  9. 9. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® polarized antennas in a terminal, UE employs [3] F. Boccardi et al., “Multiple-Antenna Techniques in LTE- Advanced,” IEEE Commun. Mag., vol. 50, no. 3, Mar. single polarization while maximizing antenna 2012, pp. 114–21. Given the recent spacing within terminal with fixed area. In a [4] T. Yoo, N. Jindal, and A. Goldsmith, “Finite-Rate Feed- nutshell, we simulate an XXXX(eNB) Æ back MIMO Broadcast Channels with a Large Number deployment of ÔÔ(UE) configuration, where XXXX and ÔÔ of Users,” Proc. IEEE ISIT, July. 2006, pp. 1214–18. heterogeneous [5] 3GPP TS 36.913, “Evolved Universal Terrestrial Radio refer to cross-polarization and single-polariza- Access (E-UTRA); Requirements for further advance- networks, there is a tion, respectively. The simulation results are ments for Evolved Universal Terrestrial Radio Access (E- summarized in Table 3 and Fig. 3. While achiev- UTRA) LTE-Advanced.” need in future ing performance close to the adaptive code- [6] M. Sharif and B. Hassibi, “On the Capacity of MIMO Broad- cast Channels with Partial Side Infonnation,” IEEE Trans. releases to account book, Rel. 10 dual codebook achieves 34.3 Info.Theory, vol. 51, no. 2, Feb. 2005, pp. 506–22. percent gain on average and 50 percent gain in for small cell [7] T. Yoo and A. Goldsmith, “On the Optimality of Multi- the cell edge when compared to the extended Antenna Broadcast Scheduling Using Zero-Forcing deployment when Rel. 8 codebook. Beamforming,” IEEE JSAC, vol. 24, no. 3, Mar. 2006, pp. 528–41. designing the Therefore, we conclude that Rel. 10 dual [8] 3GPP TSG RAN WG1 #60, R1-101129, “On Extensions codebook is a very competitive option in cross- to Rel-8 PMI Feedback,” Feb. 2010. MU-MIMO. polarized antenna setup. [9] 3GPP TSG RAN WG1 #61, R1-103332, “Way Forward MU-MIMO systems on UE Feedback,” May 2010. [10] 3GPP TSG RAN WG1 #61, R1-103378, “Performance will continue to be Evaluations of Rel.10 Feedback Framework,” May 2010. CONCLUDING REMARKS AND [11] 3GPP TSG RAN WG1 #62, R1-105011, “Way Forward improved to provide on 8Tx Codebook for Rel.10 DL MIMO,” Aug. 2010. FUTURE DIRECTION [12] 3GPP TR 36.871 V11.0.0, “Evolved Universal Terrestrial solutions for Radio Access (E-UTRA); Downlink Multiple Input Multi- those issues. In this article, we have reviewed the main fea- ple Output (MIMO) enhancement for LTE-Advanced tures of MU-MIMO techniques adopted in LTE (Release 11).” and LTE-A standards. The major enhancement [13] IEEE Std 802.16mTM-2011, “Amendment 3: Advanced of the downlink MU-MIMO in Rel. 10 is the Air Interface,” May 2011, pp. 740. [14] 3GPP TSG RAN WG1 #58, R1-093059, “Adaptive Feed- support of eight transmit antennas to allow back: A New Perspective of the Adaptive Codebook,” simultaneous transmission of up to eight layers. Aug. 2009. Such enhancement is made possible by the intro- duction of new measurement and demodulation BIOGRAPHIES reference signals enabling non-codebook-based CHAIMAN LIM is currently working toward his Ph.D. degree precoding and a novel CSI feedback mechanism in the School of Information and Communication, Korea relying on a high-performance and low-overhead University, Seoul. His research interests include communica- double codebook structure. tions and information theory, signal processing for wireless Recently, standardization effort for LTE communications, and advanced MIMO systems. Rel. 11 has started, and the radio access net- T AESANG Y OO received his B.S. degree (Hons.) in electrical work (RAN) working group is discussing engineering from Seoul National University, Korea, in 1998, enhancement of downlink MIMO as a study and his M.S. and Ph.D. degrees in electrical engineering item. Given the importance of MU-MIMO from Stanford University, California, in 2003 and 2007. Since 2006, he has been with Corporate Research and operation for network operators to meet the Development of Qualcomm, where he is currently working increasing user demands of the fourth genera- on 3GPP LTE and LTE-Advanced wireless systems. His tion (4G) and beyond 4G (B4G) wireless sys- research interests include heterogeneous cellular networks, tems, CSI feedback and control signaling will multiple-antenna systems, receiver algorithms, and various design and performance aspects of LTE/LTE-A systems and be further improved in future release. Also, a standards. future release will pay particular attention to enhancing the MU-MIMO for realistic deploy- B RUNO C LERCKX received his Ph.D. degree from Universite ment scenarios. For instance, enhancing the catholique de Louvain, Belgium. He held visiting research positions at Stanford University and EURECOM, and was CSI feedback for 4Tx cross-polarized antenna with Samsung Electronics from 2006 to 2011. He actively arrays will be one priority. Moreover, given contributed to 3GPP LTE/LTE-A and acted as Rapporteur for the recent deployment of heterogeneous net- the 3GPP CoMP Study Item. He is now a lecturer (assistant works, there is a need in future releases to professor) at Imperial College London and an editor for IEEE Transactions on Communications. account for small cell deployment when designing MU-MIMO. MU-MIMO systems B YUNGJU L EE received his B.S. degree from the School of will continue to be improved to provide solu- Information and Communication, Korea University, in tions for those issues. 2008, where he is currently working toward his Ph.D. degree. His research interests include wireless communica- tions and network information theory. ACKNOWLEDGMENTS This work was supported by the KCC R&D pro- B YONGHYO S HIM received B.S. and M.S. degrees in control gram (KCA-12-911-01-110) and the NRF grant and instrumentation engineering from Seoul National Uni- versity in 1995 and 1997, respectively, and an M.S. degree of MEST Korea (No. 2012R1A2A2A01047510). in mathematics and a Ph.D. degree in electrical and com- puter engineering from the University of Illinois at Urbana- REFERENCES Champaign in 2004 and 2005, respectively. From 2005 to 2007, he worked for Qualcomm Incorporated. Since [1] http://network4g.verizonwireless.com/#/4g-network-ver- ______________________________ September 2007 he has been with the School of Informa- izon-wireless _______ tion and Communication, Korea University, where he is [2] E. Dahlman, S. Parkvall, and J. Skold, 4G LTE/LTE-Advanced currently an associate professor. His research interests for Mobile Broadband, Academic Press, 2011. include wireless communications and information theory. IEEE Communications Magazine • March 2013 135C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®