wimax lte

INTRODUCTION

The IEEE 802.16e (WiMAX Profile 1.0) andThird Generation Partnership Project (3GPP)Evolved Universal Terrestrial Radio Access(E-UTRA) Long Term Evolution (LTE)(Releases 8 and 9) standards have been devel-oped and are part of the IMT-2000 third gen-eration (3G) technologies [1]. IEEE 802.16m(WiMAX Profile 2.0) [2] and 3GPP E-UTRALTE-Advanced (LTE-A) (Release 10) [3] aresti l l being developed primarily to meet orexceed the requirements of the InternationalTelecommunication Union (ITU) for IMT-Advanced fourth generation (4G) technologies.In this article we use 802.16m for IEEE802.16m, LTE for 3GPP releases 8 and 9, LTE-A for 3GPP release 10, and E-UTRA forreleases 8 to 10. With l imited spectrumresources, multiple-input multiple-output(MIMO) techniques are paramount for achiev-ing the minimum target cell spectral efficiency,peak spectral efficiency, and cell edge userspectral efficiency defined by the ITU [4].

This article provides an overview of thedesign challenges and the specific MIMO solu-

tions developed in these standards, in terms ofMIMO configuration and reference signals (RSs)along with enhancements of open-loop (OL) andclosed-loop (CL) MIMO operation. Key tech-niques in MIMO downlink (DL) and the MIMOsystem capabilities are summarized in Tables 1and 2, respectively, and introduced in moredetail later in this article. Space-frequency blockcoding (SFBC), frequency-switched transmitdiversity (FSTD), and cyclic delay diversity(CDD) are OL transmit diversity techniques.Multicell MIMO and uplink MIMO techniquesare defined in 802.16m and under discussion inLTE-A. 3GPP release 8 features are supportedin 3GPP release 9, and all of them are supportedin 3GPP release 10.

MIMO CONFIGURATIONS802.16m and E-UTRA target MIMO schemesfor the same sets of antenna configurations: 2,4, or 8 transmit antennas and a minimum of 2receive antennas in the DL, and 1, 2, or 4transmit antennas in the uplink with a mini-mum of 2 receive antennas. Terminologies in802.16m and E-UTRA are not matched, andas such could be confusing to the reader.Table 3 gives the equivalence between the ter-minologies used in both standards. TheMIMO systems can be configured as singleuser MIMO (SU-MIMO), multi-user MIMO(MU-MIMO), and multicell MIMO; their DLsare illustrated in Fig. 1.

SU-MIMOSU-MIMO transmissions occur in time-frequen-cy resources dedicated to a single-terminaladvanced mobile station/user equipment(AMS/UE), and allow achieving the peak userspectral efficiency. They encompass techniquesranging from transmit diversity to spatial multi-plexing and beamforming. These techniques aresupported in 802.16e/m and LTE, with a mostnoticeable difference in the approach taken forspatial multiplexing.

Spatial multiplexing (SM) is a recognizedtechnique for increasing the peak user through-put by sending multiple spatial streams through

IEEE Communications Magazine • May 201086 0163-6804/10/$25.00 © 2010 IEEE

ABSTRACT

IEEE 802.16m and 3GPP LTE-Advancedare the two evolving standards targeting 4Gwireless systems. In both standards, multiple-input multiple-output antenna technologiesplay an essential role in meeting the 4Grequirements. The application of MIMO tech-nologies is one of the most crucial distinctionsbetween 3G and 4G. It not only enhances theconventional point-to-point l ink, but alsoenables new types of links such as downlinkmultiuser MIMO. A large family of MIMOtechniques has been developed for variouslinks and with various amounts of availablechannel state information in both IEEE802.16e/m and 3GPP LTE/LTE-Advanced. Inthis article we provide a survey of the MIMOtechniques in the two standards. The MIMOfeatures of the two are compared, and theengineering considerations are depicted.

TOPICS IN WIRELESS COMMUNICATIONS

Qinghua Li and Guangjie Li, Intel

Wookbong Lee and Moon-il Lee, LG Electronics

David Mazzarese and Bruno Clerckx, Samsung

Zexian Li, Nokia

MIMO Techniques in WiMAX and LTE:A Feature Overview

LI LAYOUT 4/22/10 11:39 AM Page 86

IEEE Communications Magazine • May 2010 87

multiple antennas, and separating thesestreams at the receiver by spatial processing[5]. The design of SM techniques in 802.16mand LTE emphasizes trade-offs determined inpart by backward compatibility constraints anddifferent assumptions on advanced receivercomplexity. While a linear minimum meansquare error (LMMSE) receiver is the baselinefor performance evaluation, the design shouldaccount for the availability of more complexterminals as technology evolves. Design choicesfor SM also have effects on forward error cor-rection (FEC) encoding coupled with hybridautomatic repeat request (HARQ), feedbackmechanisms, and DL control. One of the fun-damental choices is the transmission of one ormultiple FEC codewords through multiple spa-tial streams.

802.16m has evolved from VE transmissionadopted in the WiMAX profile Release 1.0.Behind this choice is the assumption thatadvanced receivers would be better implement-ed with an optimal two-stream maximum likeli-hood detector (MLD) than with a successiveinterference cancellation (MMSE-SIC) detec-tor. This view was continued in 802.16m, rely-ing on promising advances in near MLDtechniques such as QRM-MLD [6] or spheredetectors [7] for more than two spatial streams.The usage of VE also facilitates the design andimplementation of HARQ processes andrequires only a single report of channel qualityindicator (CQI) for all the multiplexed layers.

Uplink SU-MIMO in 802.16m follows the samedesign as the DL.

On the other hand, LTE has opted for mul-t iple codewords (MCW) on the DL, whileuplink SU-MIMO is still being discussed inLTE-A. MCW allows link adaptation for eachFEC codeword in CL SU-MIMO. An MMSE-SIC receiver may cope with the interferencebetween FEC codewords that are spatiallymultiplexed. CL SU-MIMO with MCWrequires one CQI report and one HARQ pro-cess for each FEC codeword. By contrast ,layer permutation at the transmitter in OLSU-MIMO with MCW has the effect of aver-aging the signal-to-noise ratio (SNR) experi-enced by the two codewords so that a singleCQI can be reported, although both code-words still need separate HARQ processes.Note that each FEC codeword in the layerpermutation experiences the same channelquality as VE when an LMMSE receiver isused. Therefore, the layer permutation inMCW is equivalent to single codeword (SCW)in some sense. Challenges for accurate model-ing of the effective SNR per codeword at theoutput of an MLD also played a role in thedecision to favor MCW with an LMMSE orMMSE-SIC receiver in LTE. Extensive studiesduring both standardization processes on SCWvs. MCW taken as a part of the whole systemunder various operation conditions haveemphasized the merits of each scheme and ledthe two standards to develop in their own way.

Table 1. Key techniques in MIMO downlink.

Key downlink MIMO techniques 802.16m LTE LTE-A

Open-loop transmit diversity SFBC with precoder cycling SFBC, SFBC+FSTD Inherited from LTE

Open-loop spatial multiplexing Single codeword with pre-coder cycling

Multiple codewords with largedelay CDD Inherited from LTE

Closed-loop spatial multiplexing Advanced beamformingand precoding

Codebook-based precoding, UE-specific RS based beamforming

Advanced beamforming andprecoding (under development)

Multi-user MIMO Closed-loop and open-loopMU-MIMO Closed-loop MU-MIMO Closed-loop MU-MIMO (under

development)

Table 2. MIMO capabilities.

Key downlinkMIMOtechniques

802.16m

3GPP E-UTRA

LTE LTE-A

Release 8 Release 9 Release 10

DL

SU-MIMO Up to 8 streams Up to 4 streams Up to 4 streams Up to 8 streams

MU-MIMO Up to 4 users (non-unitary precoding)

Up to 2 users(unitary precoding)

Up to 4 users (non-unitary precoding)*

Underdevelopment

ULSU-MIMO Up to 4 streams 1 stream 1 stream Up to 4 streams

MU-MIMO Up to 4 users Up to 8 users Up to 8 users Under development

*Release 8 unitary precoding for up to 2 users is still supported in Releases 9 and 10.


IEEE Communications Magazine • May 201088

MU-MIMO

MU-MIMO [8] has become the key technique tofulfill IMT-Advanced requirements. MU-MIMOallocates multiple users in one time-frequencyresource to exploit multi-user diversity in thespatial domain, which results in significant gainsover SU-MIMO, especially in spatially correlat-ed channels. In configurations such as DL 4 × 2(four transmit antennas and two receive anten-nas) and uplink 2 × 4, single-user transmissiononly allows spatially multiplexing a maximum oftwo streams. On the other hand, linear MU-MIMO schemes allow sending as many as fourspatial streams from four transmit antennas, orreceiving as many as four spatial streams withfour receive antennas, by multiplexing four spa-tial streams to or from multiple users. MU-MIMO techniques provide large sectorthroughputs in areas experiencing heavy datatraffic.

The different nature of DL demodulationpilots/reference signal (RS) has induced the useof different precoding techniques in 802.16mand E-UTRA. With non-precoded common RSsin LTE, the enhanced NodeB (eNB) needs tosignal the index of the precoder to the terminalvia a DL control channel. This constrains theprecoder to belong to the codebook used toreport the precoding matrix indicator or pre-ferred matrix index (PMI). Even though thechoice of the actual precoder eventually belongsto the eNB, a simple way of building a precoderis to form a matrix from orthogonal PMIs report-ed by different users, which leads to unitary pre-coding.

On the other hand, 802.16m, LTE release 9,and LTE-A have chosen to use DL precoded

dedicated pilots (UE-specific RS) even for mul-tiple streams per terminal so that the advancedbase station (ABS)/eNB can employ any pre-coder as long as the same precoder is applied toboth RS and data symbols. Both linear and non-linear MU-MIMO schemes [8] have been con-sidered in the early phase of the two standards.Nonlinear MU-MIMO with dirty paper codingtheoretically offers the best performance, butthere are many practical limitations to its imple-mentation, so linear MU-MIMO has been adopt-ed in both standards for its simplicity and goodperformance. Zero-forcing MU-MIMO [9] isone linear MU-MIMO technique commonlyassumed in both standards, where the precodingmatrix at the transmitter is not unitary. Non-uni-tary precoding techniques offer significant per-formance enhancements over unitary precoding,especially in asymmetric antenna configurations(e.g., DL 4 × 2). A scheduler selects severalusers with good spatial separation and performspseudo inversion of the combined channel matrixin order to obtain the precoding matrix. TheCQI reported by each user is then adjusted atthe ABS/eNB to fit the channel quality after pre-coding. The rank-1 PMI that best approachesthe principal eigenvector of the channel matrixand the corresponding CQI need to be reportedby the terminal.

When the terminal estimates the rank-1 PMIand CQI, it does not know which other terminalit will be paired with by the scheduler. In802.16m the AMS estimates the rank-1 PMI andCQI assuming that the paired AMS reports anorthogonal vector. With this approach, inter-userinterference is somewhat taken into considera-tion for MU-MIMO scheduling. From a differ-ent perspective, LTE emphasizes the transparentUE operation between SU-MIMO and MU-MIMO in terms of CQI feedback, thereby adopt-ing SU-MIMO feedback for MU-MIMOscheduling.

In the DL 802.16m also introduced OL MU-MIMO, where a unitary precoding matrix is pre-set for each frequency-domain resource. EachAMS selects the preferred stream (column of thematrix) for each resource, and reports the corre-sponding CQI. This technique shows promisingperformance with limited feedback in uncorrelat-ed and semi-correlated channels. It is suitable forurban areas where user density is high and thechannel is typically non-line-of-sight.

For uplink MU-MIMO, both standards allowmultiple users to transmit simultaneously in thesame uplink resource. The ABS/eNB distinguish-es the signals from these terminals through thepilots/RSs allocated to each terminal, and sepa-rates them with an advanced receiver, which canbe an MLD receiver in the 802.16m orthogonalfrequency-division multiple access (OFDMA)uplink, or a turbo MMSE receiver in the LTEsingle-carrier FDMA uplink.

REFERENCE SIGNAL FORMULTI-ANTENNA OPERATION

An RS, or pilot, is defined in 802.16m and E-UTRA to allow measurements of the spatialchannel properties and facilitate coherent

Table 3. IEEE 802.16m and 3GPP LTE terminologies.

802.16m E-UTRA

Advanced base station (ABS) Enhanced NodeB (eNB)

Advanced mobile station (AMS) User equipment (UE)

Transmit antenna Antenna port

Layer Codeword

Stream (i.e., spatial stream) Layer (i.e., spatial layer)

Pilot Reference signal (RS)

Preferred matrix index (PMI) Precoding matrix indicator (PMI)

Dedicated pilots UE-specific reference signals

Vertical encoding (VE) Single codeword (SCW)

Multilayer encoding Multiple codewords (MCW)

Resource unit (RU) Resource block (RB)

Uplink collaborative spatial multiplexing Uplink MU-MIMO

Uplink sounding Sounding reference signal



demodulation at the terminal. An RS can be adedicated RS (DRS), which is targeted for aspecific terminal, or a common RS (CRS), whichis shared among a group of terminals. The RScan be further classified into precoded or non-precoded RS. DRSs are transmitted via virtualantenna ports with a spatial precoding weight toexploit beamforming gain while keeping low RSoverhead when the number of virtual antennaports is smaller than the number of physicalantenna ports. CRSs are transmitted via physicalantenna ports without a spatial precoder toallow for channel measurements of the non-pre-coded MIMO channel.

A lot of commonalities exist in 802.16m andE-UTRA in terms of RS usage on the uplink,although exact details are different. For instance,the sounding RS is employed as a non-precodedDRS in both standards for the purpose of uplinkspatial channel adaptation, including beam selec-tion and scheduling, as well as for measuring theDL channel by exploiting channel reciprocity intime-division duplex systems. In addition, a pre-coded DRS is also employed for coherentdemodulation in uplink.

In contrast, different designs have beenadopted for the DL pilots. Non-precoded com-mon pilots, or midamble, and precoded dedi-cated pilots are both used in 802.16m forchannel measurements and coherent demodula-tion supporting up to eight transmit antennas.Although the midamble can be used for demod-ulation in a specific transmission mode such asOL SU-MIMO, it mainly provides fine channelmeasurements with low pilot overhead becauseit is transmitted in a wideband manner toenable measurements of the whole frequencyband with a small duty cycle. In addition to themidamble, precoded dedicated pilots up toeight streams, which allow flexible beam gener-ation at the ABS side, have been defined forcontiguous resource units (RUs) in 802.16mirrespective of the transmission mode. For dis-tributed RUs supporting only two streams,common pilots precoded by predefined matri-ces are utilized.

On the other hand, non-precoded CRSssupporting up to four antenna ports have beendefined for channel measurements and coher-ent demodulation purposes in LTE Release 8.With the use of non-precoded CRS, the spatialprecoder used in the DL signal should be indi-cated to a terminal in each transmission assign-ment. LTE Release 8 also supports single-layerbeamforming using rank-1 precoded DRS inaddition to the CRS, which has been extendedto two UE-specific RSs to support dual-layerbeamforming in LTE Release 9, and for whichthe spatial precoder does not need to be indi-cated to the terminal. To further enhance thepeak and average throughputs, eight antennaport transmissions supporting up to eight lay-ers have been adopted in LTE-A. PrecodedDRSs supporting up to eight layers and non-precoded CRSs (i.e., LTE-A channel stateinformation [CSI]-RS) with a low duty cyclefor eight antenna ports’ measurement are uti-lized on top of the LTE CRS, which are keptin LTE-A for the continuing support of LTEterminals.

OPEN LOOP TECHNIQUES

In general, OL techniques are designed for highmobility or limited feedback capability. OL tech-niques mainly provide robustness for the linkadaptation considering terminal mobility withinfrequent CSI feedback exploiting long-termchannel statistics rather than short-term fadinginformation. Therefore, CQI may representshort-term or long-term channel information. Asa special case, opportunistic beamforming is alsoclassified as an OL technique since no informa-tion relative to the spatial transmit weights isreported, while short-term CQI is used for theadaptation of modulation and coding rate. Twodifferent types of OL techniques have been con-sidered, space-time coding and random beam-forming, and these techniques are optimizeddifferently in each standard.

SPACE TIME/FREQUENCY CODETransmit diversity techniques provide spatialdiversity gain, which translates into higher linkmargin than single-transmit-antenna tech-niques. For a configuration with two transmitantennas, both standards have adopted the fre-quency domain version of the Alamouti code[10] as the basic transmit diversity MIMOmode, where pairs of adjacent subcarriers arecoded together instead of two adjacent timeslots. Because fast changing channel in thetime domain would destroy the orthogonalityof the code, an SFBC outperforms an STBC inhigh-speed scenarios. Application of the SFBCwith more than two transmit antennas necessi-tates the application of a precoder to convertfour or eight physical antennas into two virtualantennas. Both techniques adopted in 802.16mand LTE limit the transmission of SFBC to apair of subcarriers while making effective useof all spatial degrees of freedom over a set ofsubcarriers in order to provide robustnessagainst spatial correlations in the channel. Thereceiver can use the same decoding processindependent of the number of physical trans-mit antennas.

Figure 1. MIMO configurations.

Joint scheduler or processor

null

ABS

/eN

B

AM

S/U

E

(a)

IFFT

SU-MIMO precoder

null

ABS

/eN

B A

MS/

UE1

(b)

(c)

IFFT

MU-MIMO precoder

AM

S/U

E1

AM

S/U

E2

ABS

/eN

B1

ABS

/eN

B2

AM

S/U

E2



The usages of the precoders in 802.16m andLTE are slightly different, due to the differencein the design of DL demodulation pilots, butthey target the same objective. While 802.16mchose a combination of precoder cycling andSFBC with precoded pilots, LTE opted for acombination of FSTD and SFBC with non-pre-coded CRS. The precoder cycling creates a fixedset of two virtual antennas across all subcarrierswithin an RU and changes the virtual antennasby using different precoder weights across RUs,while FSTD cycles transmissions over pairs oftransmit antennas across subcarriers within anRU. These designs also incur different con-straints on the channel estimation at the receiv-er, with a trade-off between the reducedoverhead offered by precoded pilots in 802.16mvs. the wider range of interpolation available inthe frequency domain with non-precoded pilotsfor finer channel estimation in 3GPP LTE.

RANDOM BEAMFORMING/PRECODER CYCLINGAs described above, space-frequency codesexploit spatial diversity so that the variance ofthe CQI is reduced as the diversity order increas-es, which allows for robust transmission withinfrequent CSI feedback. However, although therobustness of space-frequency codes is superiorto other OL schemes, their limited design flexi-bility led both standards to additionally employrandom beamforming. Random beamformingartificially increases the channel selectivity bychanging beams within allocated time/frequencyresources, and strong FEC codes (e.g., turbocodes) enjoy this artificial frequency diversitygain.

Precoder cycling, in which predefined pre-coders are cyclically allocated to a group of con-tiguous subcarriers, is utilized in both standardsas a random beamforming technique. The prede-fined set of precoders is selected from the pre-coding codebook as a subset, which has a goodChordal distance property so that diversity ordercan be maximized. Precoder cycling is used forproviding beam diversity gain and beam selec-tion gain in 802.16m. To obtain the beam diver-sity gain, resources are distributed within a widefrequency band where predefined precoders ineach localized frequency band form differentbeams; hence, the aggregated resources at thereceiver may enjoy the beam diversity gain. Toenable the beam selection gain, on the otherhand, a localized resource is allocated to a ter-minal based on the CQI feedback for its pre-ferred subbands. Since the predefined precodersare cyclically changed according to the subbands,opportunistic beamforming gain can be achievedby allocating the preferred subbands as reportedby a terminal. Since non-precoded CRSs areemployed in LTE, the predefined precoders canbe changed every few subcarriers within eachRB so that the beam diversity gains are fullyexploited even in a single RB allocation. Layerpermutation is performed along with precodercycling in E-UTRA to further increase diversitygain from virtual antennas with MCW transmis-sions. The combination of precoder cycling andlayer permutation is called large-delay CDD andwas adopted as an OL SM technique in E-UTRA.

CLOSED LOOP TECHNIQUES

CL-MIMO exploits CSI at the transmitter(CSIT) for increasing coverage or throughput. InCL-MIMO the transmitter acquires the CSIfrom feedback or channel sounding, and thenforms a beamforming or precoding matrix. Themajor challenges lie in efficiently obtaining theCSI. For accurate CSIT, frequent update isrequired for mobile terminals. However, over-head and delay limit CSIT accuracy. First, theframe structure inherently sets a delay betweenthe channel measurement and the actual beam-forming transmission. Any channel variation dur-ing the delay degrades the performance. Second,the overhead for acquiring CSIT becomes bur-densome as the number of reporting terminalsand mobility increase. On one hand, the CSI ofmultiple terminals is collected, but only a fewfavorable terminals are scheduled for transmis-sion. Unfortunately, this selection gain increaseslogarithmically with the number of reporting ter-minals. On the other hand, the feedback/sound-ing overhead increases linearly with the numberof reporting terminals. Therefore, efficient feed-back/sounding techniques are essential.

FEEDBACK MECHANISMSFeedback is required when channel reciprocity isunavailable (e.g., in frequency-division duplexsystems). The major challenge lies in how toreport the preferred beamforming matrix (direc-tions), which is used for the transmitter to com-pute the actual precoder over a limited feedbackbandwidth. For overhead reduction, the wholebeamforming matrix is quantized by a matrix orvector codebook. The index of the selected quan-tization codeword (the PMI defined in Table 3)is fed back. An L bit codebook consists of 2L

codewords, where L is the required number ofbits for indexing each codeword. In this sectionthe term codeword means the quantization code-word in the quantization codebook. In the DL,after measuring the channel, the terminal search-es for the best codeword in the codebook foroptimizing the performance and reports the PMIto the ABS/eNB. After receiving the PMI, theABS/eNB looks up the codeword and computes aprecoder. Three feedback types are devised,called base codebook, adaptive (or transformed)codebook, and differential codebook, respectively.The base codebook has the least signaling over-head, and the other two have better performancewith additional signaling overhead.

The engineering considerations of the basecodebook design comprise performance gain,overhead, robustness, complexity, and poweramplifier imbalance. First, 802.16m defines 3-bitfor 2-transmit antennas (2-Tx) as well as 4-bitand 6-bit feedbacks for 4-transmit antennas (4-Tx), while LTE defines 2-bit and 4-bit feedbacksfor 2-Tx and 4-Tx, respectively. Besides the pre-ferred beamforming matrix, an indication of thepreferred number of spatial streams is alsodefined. In addition, 802.16m has 4-bit feedbackfor 8-transmit antennas (8-Tx) and an enhanced6-bit feedback for 4-Tx. Second, codewords witha rotated block diagonal structure are explicitlyemployed in 802.16m and LTE for dual-poleantennas. Third, base codebooks can be dynami-

CL-MIMO exploits

CSI at the transmitter

(CSIT) for increasing

coverage or

throughput. In

CL-MIMO, the

transmitter acquires

the CSI from

feedback or channel

sounding, and then

forms a beamform-

ing or precoding

matrix. The major

challenges lie in

efficiently obtaining

the CSI.



cally generated from a few parameters for reduc-ing storage complexity. In addition, the code-word entries of all LTE base codebooks andmost of the 802.16m codebooks are selectedfrom quaternary phase shift keying (QPSK) and8-PSK constellations for reducing the storagerequirement and computational complexity. Fur-thermore, the high-rank codewords with morecolumns include the low rank codewords with afew columns as a subset. This reduces the com-plexity of searching for the best number of spa-tial streams. Finally, each LTE codeword andmost of the 802.16m codewords load transmis-sion power evenly on each antenna for loweringthe power amplifier cost.

Since the optimal codebook varies with thedeployment scenario, adaptive codebook isdefined in 802.16m. The adaptive codebookchanges its codeword distribution according tolong-term channel statistics captured in thetransmit covariance matrix, which characterizesthe spatial correlations across transit-side anten-nas. Using that matrix, each vector codeword ofthe rank-1 base codebook is linearly transformedand normalized for generating a codeword in theadaptive codebook. Effectively, more codewordsare steered around the directions where theideal beamforming direction likely appears. As aresult, the overall quantization error is reduced.This gain increases with antenna correlation,which increases as the antenna spacing and theangle spread of departing signals at theABS/eNB decrease. Since MU-MIMO has high-er gains in highly correlated channels, the adap-tive codebook is most beneficial for MU-MIMO.Antenna configurations, inaccurately calibratedtransceiver chains, and propagation channelproperties are inherently captured in the mea-sured covariance matrix, making the adaptivecodebook robust in a wide range of scenarios.The adaptive codebook, however, requires addi-tional signaling and feedback overhead forreporting the covariance matrix that is neededonce for the whole frequency band and for aperiod greater than 20 ms.

For overhead reduction, the correlationbetween consecutive beamforming reports is alsoexploited by differential codebooks in 802.16m.Instead of depicting the preferred beamformingmatrices in full, each differential feedback onlyspecifies the incremental change between thecurrent and previous matrices. Because thechange is usually within a small range, fewercodewords are needed to cover the small rangethan the whole beamforming space covered bythe base codebook. The reduction of codebooksize not only saves feedback overhead but alsoreduces the quantization complexity. 2-bit, 4-bit,and 4-bit codebooks are defined in 802.16m for2-Tx, 4-Tx, and 8-Tx, respectively. The down sideof differential codebook is the error propagationeffect. That is, once an error occurs, that errorcorrupts the subsequent feedback reports untilthe differential process is reset.

UPLINK SOUNDINGIn time-division duplex systems, the transmittercan learn about the DL channel from soundingon the uplink channel by exploiting thereciprocity of the propagation channel. To

achieve full reciprocity, calibration of thetransceiver RF chains is needed at theABS/eNB. Both 802.16m and LTE providesounding mechanisms for estimating the uplinkchannel on a subband or wideband scale. In802.16m, the uplink sounding channel is inher-ited from 802.16e with some enhancements. InLTE the ABS/eNB assigns different trainingsequences to multiple terminals for sharing thesame training resources simultaneously. Finally,channel estimation error due to noise and inter-cell interference limits the performance ofuplink sounding.

LONG-TERM BEAMFORMINGLong term beamforming applies a rough pre-coder for the whole frequency band and over aperiod of several frames. This reduces the densi-ty of feedback and sounding multiple times. Therough beamforming direction corresponds to adominant multipath component. Furthermore, inline-of-sight scenarios with closely spaced anten-nas, the beam pattern is wide in space, and thusallows the beam to cover a high-speed terminalfor several time frames. Both 802.16m and LTEsupport PMI feedback for long-term beamform-ing. In addition, 802.16m also allows using thereported transmit covariance matrix to derivethe precoder.

CONCLUDING REMARKSIn this article we have reviewed the MIMO tech-niques adopted or under development in IEEE802.16m and 3GPP LTE/LTE-Advanced. Wehave emphasized the design trade-offs consid-ered during both standardization processes andoutlined some of the implementation challenges.

REFERENCES[1] ITU-R Rec. M.1457-8, “Detailed Specifications of the

Radio Interfaces of International Mobile Telecommuni-cations-2000 (IMT-2000),” May 2009.

[2] ITU-R SG WP 5D, “Acknowledgment of Candidate Sub-mission from IEEE under Step 3 of the IMT-AdvancedProcess (IEEE Technology),” Doc. IMT-ADV/4-E, Oct. 23,2009.

[3] ITU-R SG WP 5D, “Acknowledgment of Candidate Sub-mission from 3GPP Proponent (3GPP Organization Part-ners of ARIB, ATIS, CCSA, ETSI, TTA AND TTC) underStep 3 of the IMT-Advanced Process (3GPP Technolo-gy),” Doc. IMT-ADV/8-E, Oct. 23, 2009.

[4] ITU-R Rep. M.2134, “Requirements Related to TechnicalSystem Performance for IMT-Advanced Radio Inter-face(s),” Nov. 2008; http://www.itu.int/publ/R-REP-M.2134-2008/en

[5] G. J. Foschini and M. J. Gans, “On Limits of WirelessCommunications in a Fading Environment When UsingMultiple Antennas,” Wireless Personal Commun., vol. 6,no. 3, Mar. 1998, p. 311.

[6] K. Higuchi et al., “Adaptive Selection of Surviving Sym-bol Replica Candidates Based on Maximum Reliability inQRM-MLD for OFCDM MIMO Multiplexing,” IEEEGLOBECOM ‘04, vol. 4, Nov. 29–Dec. 3, 2004, pp.2480–86.

[7] M. O. Damen, H. El Gamal, and G. Caire, “On Maxi-mum-Likelihood Detection and the Search for the Clos-est Lattice Point,” IEEE Trans. Info. Theory, vol. 49, no.10, Oct. 2003, pp. 2389–2402.

[8] Q. H. Spencer et al., “An Introduction to the Multi-UserMIMO Downlink” IEEE Commun. Mag., vol. 42, no. 10,Oct. 2004, pp. 60–67.

[9] B. C. B. Peel, B. M. Hochwald, and A. L. Swindlehurst,“A Vector-Perturbation Technique for Near-CapacityMultiantenna Multiuser Communication — Part I: Chan-nel Inversion and Regularization,” IEEE Trans. Com-mun., vol. 53, Jan. 2005, pp. 195–202.

Long term beam-

forming applies a

rough precoder for

the whole frequency

band and over a

period of several

frames. This reduces

the density of feed-

back and sounding

by multiple times.

The rough beam-

forming direction

corresponds to a

dominant multipath

component.



[10] S. M. Alamouti, “A Simple Transmit Diversity Tech-nique for Wireless Communications,” IEEE JSAC, vol.16, no. 8, Oct. 1998, pp. 1451–58.

ADDITIONAL READING[1] N. Jindal, “MIMO Broadcast Channels with Finite-Rate

Feedback,” IEEE Trans. Info. Theory, vol. 52, July 2006,pp. 5045–60.

BIOGRAPHIESQINGHUA LI ([email protected]) received his Ph.D. degreein electrical engineering from Texas A&M University, Col-lege Station, in 2001. Since then he has been with thewireless research group of Intel Corporation, Santa Clara,California. His research interests are in the areas of beam-forming, relay, 60 GHz, ultra-wideband, interference miti-gation, multi-user detection, and channel modeling. He hascoauthored more than 100 issued/pending patents mainlyfor WiMAX, WiFi, and 3GPP systems.

GUANGJIE LI is a senior researcher at Intel Lab China focus-ing on wireless physical layer signal processing, multi-antenna technology, the IEEE 802.16m standard, and IAsignal processing since 2005. Before joining Intel, heworked in R&I of Alcatel-Lucent Shanghai on the researchof a next-generation wireless algorithm. He graduatedfrom Beijing University of Posts and Telecommunications in2002 with his Master’s degree.

WOOKBONG LEE ([email protected]) received his M.S.degree in electrical engineering from Korea University,Seoul, in 2005. He joined LG Electronics, Korea, as aresearch engineer in 2005. His research interests includespace-time coding, signal processing, and coding for wire-less communications. He is participating in IEEE 802.16mstandardization, especially advanced antenna techniques.

MOON-IL LEE ([email protected]) received his M.S. degreein electrical engineering from Ajou University, Korea, in2005. He joined LG Electronics to work in mobile commu-nication technology research in 2005. He actively partici-pated in 3GPP standardization for LTE radio accesstechnologies, primarily within the area of advanced anten-na techniques. He is active in concept development and3GPP standardization of LTE-Advanced and future wirelesstechnologies.

DAVID MAZZARESE graduated from ENSEA, France, in1998, and obtained his Ph.D. in electrical engineeringfrom the University of Alberta and TRLabs, Canada, in2005. He has been a senior engineer in the WirelessStandards and Research Group of Samsung Electronics,Suwon, South Korea, from 2005 to 2010. He has heldpositions in IEEE 802.22 and IEEE 802.16m standardsgroups.

BRUNO CLERCKX received his M.S. and Ph.D. degrees inapplied science from the Université Catholique de Lou-vain, Belgium. He held visiting research positions at Stan-ford University and Eurecom Institute, France. He iscurrently with Samsung Advanced Institute of Technolo-gy, Samsung Electronics, Korea. He is the author or coau-thor of one book and more than 30 research papers. Hehas been contributing to 3GPP LTE/LTE-A and IEEE802.16m since 2007.

ZEXIAN LI ([email protected]) received his B.S. andM.S. degrees from Harbin Inst itute of Technology,China, in 1994 and 1996, respectively, and his Ph.D.degree from Beijing University of Posts and Telecommu-nications, China, in 1999. Since 2005, he has been withNokia, Finland. His research interests include futurewireless communication networks, advanced signal pro-cessing, and applications in broadband wireless commu-nications.


Documents

wimax lte