of 19/19
Downlink-Assisted Uplink Zero-Forcing for Uplink Zero-Forcing for TDD Multiuser MIMO Systems Petri Komulainen GIGA Seminar 4.12.2008

Downlink-Assisted Uplink Zero-Forcing for TDD · PDF fileTDD Multiuser MIMO Systems Petri Komulainen GIGA Seminar 4.12 ... “Downlink Assisted Uplink Zero-Forcing for TDD Multiuser

  • View
    218

  • Download
    3

Embed Size (px)

Text of Downlink-Assisted Uplink Zero-Forcing for TDD · PDF fileTDD Multiuser MIMO Systems Petri...

  • Downlink-Assisted Uplink Zero-Forcing for Uplink Zero-Forcing for TDD Multiuser MIMO Systems

    Petri Komulainen

    GIGA Seminar 4.12.2008

  • Contents

    Introduction to MU-MIMOIntroduction to MU MIMOTDD systemGreedy beam selectionTx-Rx UL-DL zero-forcingIterative ZF solutionPil d iPilot designNumerical resultsConclusionConclusion

    4 Dec 2008 CWC | Centre For Wireless Communications 2

  • Introduction to Multiuser MIMO

    Multiple antennas in base antennas in base station and mobilesM lti l ti l Multiple spatial beams (modes) per user allowedRich scattering environment, non-line-of-sightnon line of sightGeographic direction not b iobvious

    4 Dec 2008 CWC | Centre For Wireless Communications 3

  • TDD system: Assumptions

    Uplink and downlink channel reciprocityUL d DL f f b dUL and DL frames on same frequency bandIn OFDMA, same subcarriers used for UL and DLTX uses the knowledge of CSI obtained during RX

    Space division multiple access (SDMA)A b f i t i th ti /f fA number of users co-exist in the same time/frequency frameBeam/user selection (scheduling)

    Opportunistic rate allocationConstant (sum) TX power with per frame variable rates

    Relatively small cells and low mobilitySimultaneous users are symbol synchronousSimultaneous users are symbol-synchronousNo inter-symbol-interference (cyclic prefix)

    4 Dec 2008 CWC | Centre For Wireless Communications 4

  • TDD system: Signal modely gDL and UL RX signals:

    TX-precoder

    Ter-l 1

    Channel

    TX-amplitude (diagonal)

    Dataminal 1

    Base node

    Ter-minal 2Nb = 3b

    Nu = 2

    4 Dec 2008 CWC | Centre For Wireless Communications 5

  • Greedy beam selectiony

    Greedy [4,5] user/beam selection by BSselection by BSThe same set of beams are active in both directions

    Beams spatially compatible in DL Beams spatially compatible in DL are compatible also in UL

    Selection is based on user-Selection is based on user-specific eigenmodes

    Select consecutively, one at a timetimeAt each step, select eigenbeam with largest component orthogonal to previously selected

    Beamforming gain when SNR lowMultiplexing gain

    gbeamsStop selecting if calculated capacity reduces or when all D F d

    Multiplexing gain when SNR high

    DoFs are used

    4 Dec 2008 CWC | Centre For Wireless Communications 6

  • Tx-Rx UL-DL zero-forcing ZF criterion:g

    No restrictions for the

    ZF criterion:

    number of antennas in active terminals [1]

    UL is based on reversal of DL signal processing chain

    D li kDL receiver orthonormalized into UL precoder

    Downlink

    p

    Tx power + rate ll ti d allocation, and

    scheduling/beam selection decoupled from Tx

    Uplink

    precoder design

    4 Dec 2008 CWC | Centre For Wireless Communications 7

    SU-MIMO channel

  • Iterative ZF solution

    Closed form Closed form solution here [2]

    Iteration starts from user-specific eigenmodes [3]Ideally DL receiver is matched filter; active interference suppression not mandatory => single user processing by suppression not mandatory => single-user processing by terminalsIn practice multi-user interference exists in DL due to imperfect CSI > interference suppression is beneficialimperfect CSI => interference suppression is beneficial

    4 Dec 2008 CWC | Centre For Wireless Communications 8

  • Pilot signalsg

    The same precoded pilots used for the purposes of Rx a p od d p o u d o pu po oand TxIn DL, precoded pilots associated with the active data beams are sufficientbeams are sufficientIn UL, append the pilot precoder matrix to be unitary => UL pilot overhead grows with the aggregate number of user antennas (as with all MU-MIMO schemes)

    pilot precoder matrixdata precoders

    additional precoders

    4 Dec 2008 CWC | Centre For Wireless Communications 9

  • Pilot signals: Rx responsesg p

    Pilot responses are the only reference for spatial Pilot responses are the only reference for spatial multiuser processing

    Modified equivalent ZF criterion becomes

    unitary

    4 Dec 2008 CWC | Centre For Wireless Communications 10

  • Summary of UL-DL procedurey p

    User Terminal (UL TX):U1

    User Terminal (UL TX):Estimate DL pilot RX responses (Lk pilot streams per user)SVD for the personal N xL channel or

    U2BASE NODE

    SVD for the personal NuxLk channel or receiverWaterfill (WF) over each users Lk eigenmodes for data

    U3

    for dataUL pilots TX via Lk data precoders and additional Nu-Lk pilot precoders with equal powers => unitary pilot precoder matrix

    K=3Nb=4Nu=2

    L1 = 1L2 = 0L3 = 2

    Base Node (DL TX):Estimate all UL pilot RX responses (Nu pilot streams per user)Estimate all UL pilot RX responses (Nu pilot streams per user)Perform beam selection and iterative multiuser ZF processingWaterfill (WF) over the chosen beams => Transmit Lk data streams per user

    4 Dec 2008 CWC | Centre For Wireless Communications 11

    DL pilots TX via data precoders but with equal powers

  • Numerical results: Simulation setupNumerical results: Simulation setup

    Frequency flat fading, independent between antennasf fWaterfilling power allocation, assuming knowledge of N0

    Common UL-DL power allocation rule:

    Acronym Note Explanation

    ZF Iterative proposed schemes

    UL and DL co-exist, greedy beam selection

    ZF Closed form

    Best user SVDbench-marks

    Single-user MIMO by SVD precoding for the strongest user, UL and DLmarks ,

    Non-precoded UL only, no precoding or power allocation, stream per TX antenna for the selected

    4 Dec 2008 CWC | Centre For Wireless Communications 12

    (strongest) users => all DoFs used

  • Numerical results: Achievable rate

    The achievable data rate per stream s for user k can be evaluated from the SINR seen after transmitter-channel-receiver chain:

    Receiver TX precoder

    TX power

    Inter-stream interference exists due to non-ideal CSI in both TX and RX

    4 Dec 2008 CWC | Centre For Wireless Communications 13

  • Performance: Ideal CSI

    SNRP

    kk

    =0

    SNRN

    =

    Nb = 4, Nu = 2, greedy beam selectionIdeally UL and DL are equal in performance given same power Ideally UL and DL are equal in performance, given same power constraints

    4 Dec 2008 CWC | Centre For Wireless Communications 14

  • Performance: Estimated CSI and ZF RX

    {K=4, Nb=4, Nu=2}, noisy CSI estimated from pilot onlyEstimated CSI in RX vs Estimated CSI in RX and TX PN kpilotEstimated CSI in RX vs. Estimated CSI in RX and TXOnly ZF processing used here

    Zero-forcing RX in BS Least-norm RX in UE (is ideally ZF when precoding was perfect)

    0pilotSNR N

    kp

    =

    Least norm RX in UE (is ideally ZF when precoding was perfect)

    UL RX suffers from UL pilot power distribution

    4 Dec 2008 CWC | Centre For Wireless Communications 15

  • Performance: Estimated CSI and LMMSE RX

    {K=4, Nb=4, Nu=2}, noisy CSI estimated from pilot onlyEstimated CSI in RX vs. Estimated CSI in RX and TXLMMSE RX used here 0

    pilotSNR N

    PNk

    kpilot=

    Only non-precoded UL seems to really benefit

    4 Dec 2008 CWC | Centre For Wireless Communications 16

  • Conclusion

    Low overheadCo existence of UL and DL taken into accountCo-existence of UL and DL taken into accountSingle set of pilots sufficient for both coherent RX and TX precoding

    SimplicityDecoupled MIMO channels accommodate simple power/rate allocation, as well as coding and modulation, gIdeally, terminals perform single-user MIMOLinear TX and RX processing

    PerformancePerformanceBeamforming gain from TX CSI also in uplinkOpportunistically uses all available degrees of freedomZero-forcing means suboptimal use of MU-MIMO channel capacity

    Loss is small due to user/beam selection/

    4 Dec 2008 CWC | Centre For Wireless Communications 17

  • References

    [1] P. Komulainen, A. Tlli, M. Latva-aho, and M. Juntti, Downlink Assisted Uplink Zero-Forcing for TDD Multiuser MIMO Systems, accepted for publication in

    l k fIEEE Wireless Communications & Networking Conferece, WCNC 2009, Budapest, Hungary, April 2009.

    [2] Q. H. Spencer, A. L. Swindlehurst and M. Haardt, Zero-Forcing methods for [2] Q. H. Spencer, A. L. Swindlehurst and M. Haardt, Zero Forcing methods for Downlink Spatial Multiplexing in Multiuser MIMO Channels, IEEE Trans. on Signal Processing, vol. 52, no. 2, pp. 461471, Feb. 2004.

    [3] Q H Spencer and A L Swindlehurst A hybrid approach to spatial [3] Q. H. Spencer and A. L. Swindlehurst, A hybrid approach to spatial multiplexing in multiuser MIMO downlinks, EURASIP J. Wireless Comm. and Netw., vol. 2004, no. 2, pp. 236247, Dec. 2004.

    [4] A. Tlli and M. Juntti, Scheduling for multiuser MIMO downlink with linear processing, in Proc. IEEE Int. Symp. Pers., Indoor, Mobile Radio Commun., vol. 1, Berlin, Germany, Sep. 2005, pp. 156160.

    [5] G. Dimic and N. Sidiropoulos, On downlink beamforming with greedy userselection: Performance analysis and a simple new algorithm, IEEE Transactions on Signal Processing, vol. 53, no. 10, pp. 38573868, 2005.

    4 Dec 2008 CWC | Centre For Wireless Communications 18

  • Questions?Questions?

    Thank You!