Iterative Joint Detection, Decoding, and Channel Estimation in by mercy2beans116

VIEWS: 35 PAGES: 18

									Iterative Joint Detection, Decoding,
  and Channel Estimation in Turbo
        Coded MIMO-OFDM

        GIGA SEMINAR ’08
           Jari Ylioinas
                           Outline

          Introduction
          System Model
          Iterative Receiver
          Soft MIMO Detector
          Channel Estimator
          Conclusions




11/27/08                   CWC | Centre For Wireless Communications   2
                        Introduction


      Orthogonal frequency division multiplexing (OFDM)
         Divides the frequency selective fading channel
          into many parallel flat fading sub-channels.
             Simplifies the receiver design (usually, no need for
            time domain equalization.)
           An attractive air interface for high-rate
          communication systems with large bandwidths.




12/2/08                     CWC | Centre For Wireless Communications   3
                         Introduction


       Multiple-input multiple-output (MIMO) channels
           offer improved capacity and potential for
           improved reliability compared to single-input
           single-output (SISO) channels.

       Combining a MIMO processing with OFDM is
           identified as a promising approach for future
           communication systems.




11/29/08                     CWC | Centre For Wireless Communications   4
                         Introduction

      Iterative joint detection, decoding, and channel
          estimation is considered.
             Iterative joint detection and decoding
              approximates the optimal joint detector/
              decoder.
             Taking channel estimation within the joint
              iterative processing improves spectral
              efficiency since the pilot overhead can be
              reduced.




12/1/08                     CWC | Centre For Wireless Communications   5
                                 System Model


     Source

                                       OFDM
modulator

                                                                                   Add


                                       S/P
        IFFT
           P/S

     Turbo

                                                                                Cyclic
prefix
       Rayleigh


    encoder
      π
     MIMO

                        MAPPER
                                                                       fading

                                                                                   Add


                                       S/P
        IFFT
           P/S

                                                                                Cyclic
prefix

                                                                                                     channel


                                       OFDM
demodulator

                                                                                     Remove


                         Iterative
     P/S
         FFT
            S/P

                                                                                    Cyclic
prefix

                        detection/


               Sink

                         decoding

                       and
channel

                                                                                     Remove


                        estimation
     P/S
         FFT
            S/P

                                                                                    Cyclic
prefix





4 Dec 2008                               CWC | Centre For Wireless Communications                                 6
                                Iterative Receiver

      The optimal joint detector/decoder is
       approximated with iterative detection and
       decoding.
      The detected and decoded data is used in channel
       estimation.
                                                                      Channel
                  Symbol

                                                                      estimator
               estimator

                                                   LD1         LE1
                                                                                     LA2
                      OFDM
                                                                                   LD2
                  OFDM

                      demod.

                  demod.
         Soft
MIMO
                             De-
                    Turbo


              OFDM
                                       +
              demod.

          OFDM
                    detector
             -
                                                                     interleaver
               decoder

          demod.

                                                                                                Decoder

                                                                                               iterations

                                                                                           -
                                                                     Interleaver
     +                        Global


                                                             LA1
                                                                                    LE2                      iterations



12/1/08                                CWC | Centre For Wireless Communications                                            7
                            Iterative Receiver

  Motivation of the receiver
          structure.
             The spectral efficiency can
              be increased.
             If ~0.8 dB higher SNR
              value is allowed, the pilot
              overhead can be
              decreased from 16.7 % to
              0.5 %. (Assuming frame                                           Pilot based

              error rate target (FER) of
              10 %.)




12/1/08                             CWC | Centre For Wireless Communications                 8
                                Soft MIMO Detector




                                                                    Channel
                   Symbol

                                                                    estimator
                estimator

                                                 LD1         LE1
                                                                                    LA2
                      OFDM
                                                                                LD2
                  OFDM

                      demod.

                  demod.
         Soft
MIMO
                           De-
                     Turbo


              OFDM
                                    +
              demod.

          OFDM
                    detector
           -
                                                                   interleaver
                decoder

          demod.



                                                                                          -

                                                           LA1     Interleaver
      +
                                                                                   LE2




12/1/08                                 CWC | Centre For Wireless Communications                                 9
                      Soft MIMO Detector

       A posteriori probability (APP) algorithm is the
           optimal soft MIMO Detector.
              Calculates the Euclidean distance of every
               possible candidate symbol vector and uses
               them in log-likelihood ratio (LLR) calculation.
                Computationally too intensive in many cases.
       List detectors approximate the APP algorithm by
           forming a candidate list which should include the
           most probable candidate symbol vectors.
              In many cases based on the QR decomposition
               (QRD) of the channel matrix and tree search
               algorithms.


11/29/08                      CWC | Centre For Wireless Communications   10
                      Soft MIMO Detector

   We derived a new list parallel interference
          cancellation (PIC) detector based on the space-
          alternating generalized expectation-maximization
          (SAGE) detector.
             Uses breadth-first search scheme.
               Good in the implementation point of view.
            Shows good performance in 2 x 2 antenna
      configuration.
   We proposed list re-calculation in iterative detection
    and decoding.
                                                           List Detector
                                     OFDM

                                 OFDM

                                     demod.

                                 demod.

                             OFDM
                                         List
                                                                                    LLR
                             demod.
                                    algorithm
                         OFDM

                         demod.

                                                                              LA1

12/1/08                            CWC | Centre For Wireless Communications               11
                Soft MIMO Detector

      Performance examples.




   MT=MR=2, 64QAM                                   MT=MR=4, QPSK


12/3/08               CWC | Centre For Wireless Communications      12
                                 Channel Estimator




                                                                    Channel
                  Symbol

                                                                    estimator
               estimator

                                                 LD1         LE1
                                                                                   LA2
                       OFDM
                                                                              LD2
                   OFDM

                       demod.

                                  Soft
MIMO
                           De-
                    Turbo


                   demod.

               OFDM
                                    +
               demod.

           OFDM
                   detector
           -
                                                                   interleaver
               decoder

           demod.



                                                                                         -

                                                           LA1     Interleaver
      +



                                                                                   LE2




11/29/08                                CWC | Centre For Wireless Communications                                13
                  Channel Estimator

       The least-squares (LS) estimation is the best
        linear unbiased estimator for Gaussian noise.
           However, in decision directed (DD) mode of
            operation a matrix inversion is required.
       The frequency domain (FD) SAGE algorithm [Xie
        et al. IEEE Trans. Comm.]
           Converts iteratively the LS estimation of MIMO
            channel into multiple SISO channel estimation
            problems (avoids matrix inversion).
           With non-constant envelope constellations, it
            starts to lose to the LS estimation.



11/29/08                 CWC | Centre For Wireless Communications   14
                 Channel Estimator

      We generalized the FD SAGE channel estimator for
       non-constant envelope constellations.
          The drawback with generalized FD SAGE is the
           required matrix inversion.
          However, the size of the matrix to be inverted is
           smaller than that of with the LS estimator.
      We derived the time domain (TD) SAGE channel
       estimator.
          Avoids the matrix inversion without performance
           degradation with non-constant envelope
           constellation.



12/3/08                 CWC | Centre For Wireless Communications   15
                          Channel Estimator

      Complexity and performance examples.


    Algorithm    # complex        # complex
                multiplications    divisions


          LS     21418400           800

    FD SAGE       186368              -
  GFD SAGE        840592            1200

    TD SAGE       245880            120


   MT=MR=2, L=10, K=512,                                          MT=MR=4, 64QAM
   NI=3 (number of iterations)


12/3/08                               CWC | Centre For Wireless Communications     16
                         Conclusions

       Iterative joint detection, decoding, and channel
           estimation was considered in MIMO-OFDM system.
          A new list PIC detector was discussed which gives
           nice performance in 2 x 2 antenna configuration.
          List re-calculation was presented as a way to speed
           up the convergence in iterative detection decoding.
          The time domain SAGE channel estimator was
           shown to solve the problem of the FD SAGE
           channel estimator with non-constant envelope
           constellations.
          The iterative receiver was shown to improve the
           spectral efficiency remarkably.


11/27/08                    CWC | Centre For Wireless Communications   17
Literature:
J. Ylioinas, M.R. Raghavendra, M. Juntti ” Avoiding Matrix Inversion in DD
SAGE Channel Estimation in MIMO-OFDM with M-QAM”, IEEE Signal
Processing Letters, submitted

J. Ylioinas, M. Juntti ”Iterative Joint Detection, Decoding, and Channel
Estimation in Turbo Coded MIMO-OFDM”, IEEE Transactions on Vehicular
Technology, In press.



                   Questions?

            Thank You!

								
To top