Docstoc

09

Document Sample
09 Powered By Docstoc
					    Cyber Journals: Multidisciplinary Journals in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT), May Edition, 2011




             Throughput Performance Enhancement for
             MUDiv/OFDMA using MMSE Equalization
                     without Guard Interval
                      Yuta Ida†, Chang-Jun Ahn, Takeshi Kamio, Hisato Fujisaka, and Kazuhisa Haeiwa


                                                                                   many users in the same channel at the same time [7], [8].
Abstract—Recently, to achieve a high data rate and a high quality                   WiMAX is one of the next generation mobile networks
multimedia service, orthogonal frequency division multiplexing                      designed to support a high capacity and a high data rate.
(OFDM) and orthogonal frequency division multiplexing access                           In a wireless network, the transmitted signal of each user has
(OFDMA)       are widely studied. In a wireless network, the
                                                                                    independent channel fluctuation characteristics. By using such
transmitted signal of each user has independent channel
fluctuation characteristics. By using such characteristic, a
                                                                                    characteristic, the diversity that exists between users is called
multiuser diversity (MUDiv) for OFDMA has been proposed. In a                       multiuser diversity (MUDiv) and can be exploited by the sender
multipath fading environment, inter-symbol interference (ISI) is                    to enhance the capacity of a wireless network [9]-[11].
caused. In OFDM systems, the ISI is eliminated by inserting the                     Therefore, the MUDiv technique achieves dramatically
guard interval (GI). On the other hand, this operation is degraded                  increased the system throughput and the spectral efficiency [12].
the maximum throughput. In this paper, we propose to enhance                        In a MUDiv for OFDMA, the exploiting channel fluctuation
the throughput performance for a MUDiv/OFDMA without GI.                            diversity is in essence done by selecting the user with the strong
                                                                                    subcarrier channels.
 Index Terms—OFDMA, multiuser diversity, guard interval,                               In a multipath fading environment, inter-symbol interference
MMSE, throughput
                                                                                    (ISI) and inter-carrier interference (ICI) are caused due to the
                                                                                    previous symbol and different subcarrier, respectively. In
                                                                                    OFDM systems, the ISI is eliminated by inserting the guard
                          I. INTRODUCTION
                                                                                    interval (GI) longer than the delay spread channel. However,

O     rthogonal frequency division multiplexing (OFDM)
      systems have recently attracted considerable attention as a
      fourth generation mobile communication system due to the
                                                                                    this operation is degraded the maximum throughput due to the
                                                                                    extended packet length. To mitigate this problem, the several
                                                                                    methods without GI have been proposed [13]-[15]. For example,
parallel signal transmission using many subcarriers that are                        [13] improves the performance in the short GI. However, the
mutually orthogonal [1]. Moreover, since the frequency spacing                      system performance without GI is significantly degraded. [14]
of each subcarrier minimum, OFDM can treat a frequency                              mitigates the complexity by using the time domain equalizer
selective fading as a flat fading for each subcarrier [2], [3].                     (TDE). However, TDE is degraded the system performance in
Furthermore, OFDM has been chosen for several broadband                             the rugged environment as a frequency selective fading. [15]
WLAN standard like IEEE802.11a, IEEE802.11g, and                                    adapts the overlap-frequency domain equalization (FDE).
European HIPERLAN/2. In addition, terrestrial digital audio                         However, the noise enhancement due to the residual ISI is not
broadcasting (DAB) and digital video broadcasting (DVB) have                        considered by using the perfect channel estimation. Previously,
also proposed for broadband wireless multiple access system                         we have proposed the ISI and ICI compensation methods for
such as IEEE802.16 wireless MAN standard and interactive                            multiple-input multiple-output (MIMO) systems [16]. To
DVB-T [4]-[6]. OFDM allows only one user on the channel at                          mitigate above-mentioned problems, in this paper, we propose a
any given time. To accommodate multiple users, orthogonal                           MUDiv/OFDMA without GI using the ISI and ICI cancellation
frequency division multiple access (OFDMA) has been                                 to enhance the throughput performance. In Section II, we
proposed [6]. OFDMA combines OFDM and frequency                                     present a MUDiv/OFDMA system. The configuration of the
division multiple access (FDMA), and provides each user with a                      proposed system is described in Section III. In Section IV, we
fraction of the available number of subcarriers. Worldwide                          show the computer simulation results. Finally, the conclusion is
interoperability for microwave access (WiMAX) as                                    given in Section V.
IEEE802.16 standard is applied an OFDMA to accommodate

   Manuscript received May 10, 2011.                                                                II. MUDIV/OFDMA SYSTEM
   Y. Ida, T. Kamio, H. Fujisaka, and K. Haeiwa are with the Graduate School
of Information Sciences, Hiroshima City University, 3-4-1 Ozukahigashi,
Asaminami-ku, Hiroshima, 731-3194 Japan.
                                                                                    A. Channel Model
   C. Ahn is with the Graduate School of Engineering, Chiba University, 1-33          We assume that a propagation channel consists of L discrete
Yayoi-cho, Inage-ku, Chiba, 263-8522 Japan.                                         paths with different time delays. The impulse response hm(τ, t)
   e-mail: y.ida@chiba-u.jp†

                                                                               74
for user m is represented as                                                                                10

                                          L 1
                   hm  , t    hm.l t     m,l ,                            (1)                  5
                                          l 0


where hm,l, τm,l are the complex channel gain and the time delay                                            0

of the lth propagation path for user m, and L 1 E hm,l  1 ,
                                                         2
                                                                         l 0
                                                                                                            -5
where E         denotes the ensemble average operation. The
channel transfer function Hm(f, t) is the Fourier transform of                                             -10

hm(τ, t) and is given by
                                                                                                           -15
                                  
             H m  f , t    hm  , t  exp  j 2f d
                                  0
                                                                                       (2)                 -20
                                 L 1
                                 hm,l t  exp  j 2f m,l .                                                         User 1
                                                                                                                          User 2
                                 l 0                                                                      -25
                                                                                                                 0                  5            10            15              20
                                                                                                                                 Frequency bandwidth [MHz]
                                                                                                  Fig. 1. Magnitude of the channel transfer function for a radio channel with
B. Subcarrier Selection                                                                           multipath.
   Figure 1 shows the magnitude of the channel transfer function
for different users in a single cell. Subcarriers fade differently
from user to user in OFDMA systems. The diversity that exists                                     C. MUDiv/OFDMA
between users is called multiuser diversity (MUDiv) and can be                                       The transmitter block diagram of the proposed system is
exploited by the transmitter to enhance the capacity of a wireless                                shown in Fig. 2(a). Firstly, the coded data is modulated and Np
network. In a MUDiv/OFDMA, exploiting channel fluctuation                                         pilot symbols are appended at the beginning of the sequence.
diversity is in essence done by selecting the user with the strong                                The MUDiv/OFDMA transmitted signal for user m can be
subcarrier channels. In this case, the selection scheme is very                                   expressed in its equivalent baseband representation as
important to select subcarriers with the highest SNR and to                                                                   N p  N d 1
                                                                                                                                                       2S N c 1
                                                                                                                                                      
guarantee all users the same quality of service (QoS). The                                                       s m t              g t  iT            um k , i         (6)
                                                                                                                                 i 0                  N c k 0
                                                                                                                                                      
subcarrier selection criteria for each user is given by
                                                                                                                                              exp  j 2 t  iT k Ts ,
           Nu 1 Nc 1
                                                           1        allocation
     Z      H k                       m ,k ,  m ,k  
                                      2
                                                                                       (3)
                           m                                                                      where Nd and Np are the number of data and pilot symbols, Ts is
           m 0 k 0                                       0        no allocation ,
                                                                                                  the effective symbol length, S is the average transmission power,
                                                                                                  and T is the OFDM symbol length, respectively. The frequency
where αm,k is the selection parameter, Nc is the number of
                                                                                                  separation between adjacent orthogonal subcarriers is 1/Ts and
subcarriers, and Nu is the number of users, respectively. From
                                                                                                  can be expressed, by using the kth subcarrier of the ith
Eq. (3), the subcarrier is selected for the maximization of Z. In
                                                                                                  modulation symbol dm(k, i) with |dm(k, i)|=1 for Np ≤ i ≤ Np+Nd-1,
this case, one subcarrier is selected at once and users do not
                                                                                                  as
share the same subcarriers. Therefore, the MUDiv technique
promises dramatically increased the system throughput and the                                                                 um k , i   cPN k  d m k , i ,                 (7)
spectral efficiency. On the other hand, Eq. (3) requires large
complexity for calculating subcarrier assignment. To mitigate                                     where cPN is a long pseudo-noise (PN) sequence as a scrambling
the complexity, the block selection method is considered. In the                                  code to reduce the peak average power ratio (PAPR). Moreover,
block selection, the block for each user is given by                                              the kth subcarrier dm(k, i) is given by
                                                                                                                   Nb 1 x N b  q, i  for   k N b     1
                                                 H m q  k                                      d m k , i   q0 m
                                           1                   2
                                                                                                                  
                        H m q   
                                                                    ,                 (4)
                                          k 0        Nb                                                          0
                                                                                                                                          otherwise,
                                                                                                                                                                 (8)
where Nb is the block length, β is N c N b  , and y  denotes
                                                                                                  where xm(k, i) is the kth subcarrier of the ith symbol for user m.
the largest integer less than or equal to y, respectively. By using
                                                                                                  In general, the GI is inserted in order to eliminate the ISI due to
Eq. (4), the block selection criteria for each user is given by
                                                                                                  a multipath fading, and hence, we have
           Nu 1  1
                                                           1        allocation
    Z      H  q                 m ,q ,  m ,q  
                                  2
                                                                                       (5)
           m 0 q  0
                          m
                                                           0        no allocation .                                                    T  Ts  Tg ,                               (9)

                                                                                                  where, Tg is the GI length. In Eq. (6), the transmission pulse g(t)
                                                                                                  is given by

                                                                                             75
                                                                                                                                                                         User#Nu-1



                                            Data/FEC              Interleaver           Mod.
                                                                                                      Mux        S/P       Scrambling        IFFT         P/S
                                                                                Pilot generation                                                                       User#0


                                                                                       (a) Transmitter (MS)

                                                                                                                                              Channel
                                                                                                   Side information
                                                                                                                                             estimation
      User#0

                                                            P/S                                                       MMSE detector                           Descrambling      FFT        S/P
          FEC/Data               Deinterleaver
      User#Nu-1

                                                                                              Interference
                                                                    S/P         IFFT                              FFT
                                                                                              cancellation

                                                                                                    (b) Receiver (BS)
Fig. 2. Proposed system.

                                                                                                                                         c PN k  ~
                                                                                                                                           
                                1              for  Tg  t  Ts                                                        r k , i                 r k , i 
                       g t                                                                      (10)
                                                                                                                                        c PN k 
                                                                                                                                                  2
                                                                                                                                                                                                     (14)
                                0              otherwise.
                                                                                                                                                 N u 1

                                                                                                                                                  H k         Ts , iT d m k , i   nk , i ,
                                                                                                                                            2S
                                                                                                                                                                                       ˆ
   The receiver structure is illustrated in Fig. 2(b). By applying                                                                          Nc   m 0
the FFT operation, the received signal r(t) is resolved into Nc
                                                                                                                where (∙)* is a complex conjugate and cPN k  cPN k  is the
                                                                                                                                                                                                2
subcarriers. The received signal r(t) in the equivalent baseband
representation can be expressed as
                                                                                                                descrambling operation, respectively. For Eq. (14), we can see
                          N u 1                                                                                that the received signal has the frequency distortion arising from
                                    
               r t       h , t s t   d  nt ,
                                                       m
                                                                                                    (11)
                                                                                                                a frequency selective fading. To mitigate this frequency
                                   
                          m 0
                                                                                                                distortion, the frequency equalization combining is necessary.
where n(t) is additive white Gaussian noise (AWGN) with a                                                       For the channel estimation scheme using Np pilot symbols, the
single sided power spectral density of N0. The kth subcarrier                                                   channel response of the kth subcarrier is given by
~k , i  is given by
r                                                                                                                                                                      N p 1

                                                                                                                               H k Ts                                r k , i ,
                                                                                                                               ~                          1                                          (15)
                                        r t  exp  j 2 t  iT k Ts dt
                              iT Ts
        ~ k , i   1
        r
                     Ts    iT
                                                                                                                                                 N p 2P N c            i 0


                                 N u 1N c 1                                                                   where P is the transmitted pilot signal power. Here, the
                                    u m e, i                            exp  j 2
                          2S                                   1        Ts
                  
                          Nc      m 0 e 0                    Ts   0
                                                                                                                combining weight for the kth subcarrier is denote by ω(k, i).
                                                                                                                After the frequency equalization combining, the received
                       e  k t Ts          
                                                     
                                                         h , t  iT g t   
                                                                                                    (12)        detected data symbol for user m can be written as
                                                                                                                 d m k , i   r k , i    k , i 
                                                                                                                 ~
                       exp  j 2e Ts d dt  nk , i ,
                                                   ˆ

                                                                                                                                     H k Ts , iT d m k , i   k , i   nk , i    k , i 
                                                                                                                                  2S
                                                                                                                                                                             ˆ
                                                                                                                                  Nc
where nk , i  is AWGN noise with zero-mean and a variance of
        ˆ
                                                                                                                                                       for   k N b     1. (16)
2N0/Ts. After abbreviating, Eq. (12) can be rewritten as
                                        N u 1N c 1                                                            From Eq. (16), the decision variable of the kth subcarrier and
        ~ k , i   1
                                            u m e, i   exp  j 2
                                 2S                                     Ts
        r                                                                                                       the ith data symbol for user m is obtained by
                     Ts          Nc       m 0 e0
                                                                     0


                                                
                                                                                                                                Nb 1
                                                                                                                 ~ k , i  
                                                                                                                                 d N               q, i  for   k N b     1. (17)
                                                                                                                                  ~
                       e  k t Ts 
                                                     
                                                           h , t  iT g t                    (13)         xm                     m        b
                                                                                                                              q 0
                       exp  j 2e Ts d dt  nt 
                                                   ˆ
                                  N u 1

                                    H k        Ts , iT u m k , i   nk , i .
                          2S
                                                                        ˆ
                          Nc       m 0


After descrambling, the output signal r(k, i) is given by
                                                                                                           76
                          III. PROPOSED SYSTEM                                                                                   ~     N u 1
                                                                                                                                                     
                                                                                                                            R i  R i   λ isi,ici FDi , m
                                                                                                                                               ˆ
                                                                                                                                                m 0                                          (22)
A. MUDiv/OFDMA without GI
                                                                                                                                      N u 1
                                                                                                                                                            
  In the proposed system, we have Eq. (9) as                                                                                           λFD
                                                                                                                                       m 0
                                                                                                                                                    i,m    Ni ,
                                         T  Ts .                               (18)                                                         
                                                                                            where λ is the time domain channel matrix and N i is the noise
In this case, the received signal r(k, i) after the pilot signal                            term with the residual ISI and residual ICI, respectively. After
separation contains the ISI and ICI. In this paper, we eliminate                            the FFT operation, the ICI equalized signal Di, m is generated
the ISI and ICI in the time domain. Firstly, the received signal
r(k, i) after the IFFT operation is rewritten the time domain                               by using Eq. (15). Observing Eqs. (21) and (22), the proposed
matrix form as                                                                              method using the orthogonality reconstruction with inserting the
                                                                                            subtracted signal due to the ISI compensation can mitigate the
                 N u 1
          Ri     λ
                 m 0
                           isi, i 1   FDi 1, m  λ ici ,i FDi , m   N i ,   (19)        enhancement of the noise term. Therefore, the detected signal
                                                                                                                                           
                                                                                             Di, m is accurately detected compared with Di, m . Next, we

where λisi,i-1 and λici,i are the Nc×Nc ISI and ICI channel matrices                        explain that the frequency equalization combining using Eq.
                                                                                            (15).
for the (i - 1)th and the ith symbols, F is the IFFT operation, and
Ni is the Nc×1 noise matrix, respectively. In next subsection,
we explain the ISI and ICI compensation methods.                                            D. Zero Forcing (ZF)
                                                                                              The ZF weight ωzf(k, i) is given by
B. ISI and ICI Equalization
                                                                                                                                 zf k , i   ~
                                                                                                                                       1               (23)
                                                                                                                                             .
   In general, the first data symbol of the received signal has no                                                                  H k T 
the ISI [16]. Hence, the received signal of Eq. (19) for i = 0 is                                                                  
expressed as                                                                                Here, the time domain matrix R i after the FFT operation is
                                                                                                                                 
                                                                                            rewritten the kth subcarrier r k , i  . By using Eq (23), the
                                N u 1
                    R0          λ
                                 m0
                                           ici , 0   FD0, m  N 0 .             (20)        detectd signal d zf , m k , i  can be written as
                                                                                                                       
For i > 0, we eliminate the ISI and ICI. The ISI equalization is                                    d zf ,m k , i   r k , i    zf k , i 
performed by using the previous detected signal Di,m .   ~                                                                            N u 1

                                                                                                                                        H k          T , iT d m k , i    zf k , i 
                                                                                                                              2S
                                     ~                                                                                
Therefore, the time domain signal R i to eliminate the ISI is                                                                 Nc       m 0

obtained by                                                                                                                  n k , i    zf k , i 
                                           N u 1                                                                                     N u 1
                                                                                                                                                               n k , i 
                                                                                                                                        k , i d k , i   H k T 
                    ~                                    ~                                                                    2S
                    Ri  Ri                 λ isi,ici FDi 1,m
                                              ˆ
                                            m0                                 (21)
                                                                                                                      
                                                                                                                              Nc       m 0
                                                                                                                                               zf            m ~
                                N u 1
                                                            ~                                                                                          for   k N b     1, (24)
                               λ         ici   FD i , m  N i ,
                                m 0
                                                                                            where  zf k , i   H k T , i  H k T  and n k , i  is the noise
                                                                                                                               ~
        ˆ
where λ isi,ici is the estimated ISI and ICI channel matrix and                             term, respectively. From Eq. (24), the decision variable is
~                                                          ˆ                                obtained by
N i is the noise term with the residual ISI, respectively. λ isi,ici
                                                                                                                   N b 1
is consisted the estimated channel impulse responses for Eq.                                 x zf , m k , i     d       zf , m   N b  q, i          for   k N b     1.
(15) after the IFFT operation. After the FFT operation, the ISI
                                                                                                                  q 0
equalized signal Di, m is generated by using Eq. (15).                                                                                                                                        (25)
Observing Eq. (21), the ISI is eliminated. However, since the                               The ZF scheme can restore the orthogonality, but it enhances the
ICI is remained, the noise component is enhanced. Therefore,                                noise term due to the residual ISI and ICI.
we mitigate the noise enhancement due to the ICI.

                                                                                            E. Minimum Mean Square Error (MMSE)
C. Replica Signal Insertion based on ICI Equalization
                                                                                              The MMSE weight is given by
   To avoid the noise enhancement due to the ICI, we consider
                                                                                                              k , i   ~ H k 2T  ,
                                                                                                                            ~
the orthogonality reconstruction with inserting the eliminated                                                                                                                                (26)
                                                                                                          mm
signal due to the ISI compensation. The ICI equalized signal
                                                                                                                        H k T    2
                                                                                                                                    ~
 R i with inserting eliminated the part of the signal using
                                                                                                   ~
previous detected signal Di, m is given by                                                  where  2 is the estimated noise power. In this paper, by using
                                                                                            the detected signal d zf , m k , i  ,  2 is obtained by
                                                                                                                                    ~

                                                                                       77
                             Nu 1 N d 1
                                 
                               r k , i   H k T d k , i 
                                                                                                                 0
                     1                        ~                               2                             10
           2 
           ~
                                                                  zf ,m            .   (27)                                                   With GI (Nb=16, ZF)
                     Nd      m  0 i 0                                                                                                       Without GI (Nb=16, ZF)
                                                                                                                                              With GI (Nb=16, MMSE)

By using Eq. (26), the detected signal d mm, m k , i  can be
                                                                                                                 -1
                                                                                                            10                                Without GI (Nb=16, MMSE)


written as                                                                                                       -2
                                                                                                            10
                    
 d mm ,m k , i   r k , i   mm k , i 
                                 N u 1

                                  H k      T , iT d m k , i   mm k , i 
                                                                                                                 -3
                           2S                                                                               10
                 
                           Nc    m 0

                      n k , i   mm k , i                                                             10
                                                                                                                 -4




                                                                n k , i   H  k T 
                                 N u 1                                      ~
                                  mm k , i d m k , i   ~
                           2S
                 
                                                                 H k T    2
                                                                              2                                  -5
                           Nc    m 0                                              ~                        10



                                                 for   k N b     1, (28)                                  -6
                                                                                                            10


                                                              H k T                
                                                                                                                      0   5         10        15           20             25
where  mm k , i   H k T , T   H  k T 
                                     ~                           ~            2
                                                                                    2 .
                                                                                    ~                                               Eb/N0 [dB]
                                                                                                   Fig. 3. The BER of the conventional MUDiv/OFDMA with the GI and without
Observing the noise term of Eqs. (24) and (28), the MMSE                                           the GI at Doppler frequency of 10 Hz.
scheme can mitigate the noise enhancement due to the residual
ISI and ICI. Therefore, the MMSE scheme is accurately                                                            0
                                                                                                            10
detected to compare with the ZF scheme. Finally, the decision                                                                            With GI (Nb=16, ZF)
variable is obtained by                                                                                                                  With GI (Nb=16, MMSE)
                                                                                                                                         Without GI (Nb=16, MMSE)
                                                                                                             -1
                                                                                                           10                            ISI cancellation (Nb=16, MMSE)
                    N b 1
xmm, m k , i     d                N b  q, i 
                                                                                                                                         Proposed method (Nb=16, MMSE)
                             mm , m                     for   k N b     1.                                                        Proposed method (Nb=8, MMSE)
                    q 0
                                                                                                             -2
                                                                                                           10
                                                                                       (29)

                                                                                                             -3
                                                                                                           10
                IV. COMPUTER SIMULATED RESULTS
   In this section, we show the performance of the proposed                                                10
                                                                                                             -4


method. Figure 2 shows a simulation model of the proposed
system. On the transmitter, the data stream is encoded. Here,                                                -5
                                                                                                           10
convolutional codes (rate R = 1/2, constrain length K = 7) with
interleaving used. These have been found to be efficient for
transmission of an OFDM signal over a frequency selective                                                  10
                                                                                                             -6

                                                                                                                      0   5         10        15           20             25
fading channel. The coded bits are QPSK modulated, and then
                                                                                                                                    Eb/N0 [dB]
the pilot signal and the data signal are multiplexed. After serial
                                                                                                   Fig. 4. The BER of the conventional methods and the proposed method for Nb
to parallel (S/P) converted, the OFDM signal is allocated based                                    = 8, 16 at Doppler frequency of 10 Hz.
on Eq. (5). The scrambling operation is adapted to reduce the
PAPR with a PN code. The OFDM time signal is generated by
an IFFT. The transmitted signal is subject to broadband channel                                    maximum Doppler frequency is 10 Hz. In the receiver, the
propagation. In this simulation, we assume that OFDM symbol                                        received signal is S/P converted. The parallel sequences are
period is 8.96 s, and L = 5 path Rayleigh fadings have                                            passed to the FFT operator and convert the signal back to the
exponential shapes and a path separation Tpath = 140 ns. The                                       frequency domain. The frequency domain data signal is
                                                                                                   detected and demodulated. Since the detected signal contains
                                   TABLE I                                                         the ISI and ICI, it is necessary to eliminate them. The ISI
                          SIMULATION PARAMETERS.                                                   equalization is performed with the previous detected signal as
                  Data modulation            QPSK                                                  Eq. (21). Moreover, to restrict the orthogonality, the ICI
                   Data detection           Coherent
                                                                                                   equalization is performed to insert the replica signal as Eq. (22).
                  Symbol duration            8.96 s
                    Frame size          Np = 2, Nd = 20
                                                                                                   Finally, the MMSE equalization is performed to mitigate the
                     FFT size                  64                                                  noise enhancement due to the residual ISI and ICI as Eq. (28).
                 Number of carriers            64                                                  After the detection, bits are decoded by using the Vitebi soft
                  Number of users               4                                                  decoding algorithm. The packet consists of Np = 2 and Nd = 20
                   Guard interval       16 sample times                                            data symbols. Table I shows the simulation parameters.
                      Fading         5 path Rayleigh fading                                           Fig. 3 shows the BER of the conventional MUDiv/OFDMA
                 Doppler frequency            10 Hz
                       FEC            Convolutional code
                                                                                                   with and without GI at Doppler frequency of 10 Hz. For
                                        (R = 1/2, K = 7)                                           inserting GI, MMSE shows about 5 dB gain compared with ZF.

                                                                                              78
              0                                                                                -4
         10                                                                                10
                                      With GI (Nb=16, MMSE)
                                      Without GI (Nb=16, MMSE)
                                      ISI cancellation (Nb=16, MMSE)
           -1
         10                           Proposed method (Nb=16, MMSE)

                                                                                               -5
                                                                                           10
           -2
         10



           -3                                                                                  -6
         10                                                                                10



           -4
         10

                                                                                               -7
                                                                                           10

         10
           -5                                                                                                                       With GI (ZF)
                                                                                                                                    With GI (MMSE)
                                                                                                                                    Without GI (MMSE)
                                                                                                                                    ISI cancellation (MMSE)
           -6                                                                                                                       Proposed method (MMSE)
         10                                                                                10
                                                                                               -8

                  0    5         10        15           20             25                           0   2   4    6       8     10        12     14      16
                                 Eb/N0 [dB]                                                                      Block length Nb
Fig. 5. The BER of the conventional methods and the proposed method for Nb        Fig. 7. The BER versus the block length Nb for the conventional methods and
= 16 in the perfect channel at Doppler frequency of 10 Hz.                        the proposed method with Eb/N0 = 18 dB at Doppler frequency of 10 Hz.

              0
         10                                                                                7
                                      With GI (Nb=1, ZF)
                                      Without GI (Nb=1, ZF)
                                      With GI (Nb=16, MMSE)
           -1
         10                           Without GI (Nb=16, MMSE)                             6
                                      Proposed method (Nb=16, MMSE)
                                      Proposed method (Nb=8, MMSE)

           -2                                                                              5
         10


                                                                                           4
           -3
         10


                                                                                           3
           -4
         10

                                                                                           2
           -5                                                                                                            With GI (Nb=16, ZF)
         10                                                                                                              With GI (Nb=16, MMSE)
                                                                                           1                             Without GI (Nb=16, MMSE)
                                                                                                                         ISI cancellation (Nb=16, MMSE)
                                                                                                                         Proposed method (Nb=16, MMSE)
           -6
         10                                                                                                              Proposed method (Nb=8, MMSE)
                  0    5         10        15           20             25                  0
                                 Eb/N0 [dB]                                                     0           5            10               15                 20
                                                                                                                     Eb/N0 [dB]
Fig. 6. The BER of the conventional methods and the proposed method for Nb
= 1, 8, 16 at Doppler frequency of 10 Hz.                                         Fig. 8. The throughput of the conventional methods and the proposed method
                                                                                  for Nb = 8, 16 at Doppler frequency of 10 Hz.

Moreover, MMSE shows about 8 dB gain compared with ZF                             diversity can enhance the BER performance. In Fig. 5, the BER
without inserting GI. Therefore, the MMSE equalization can                        performance is improved under the perfect channel estimation.
enhance the BER performance both inserting GI and without                         This is because the channel response of Eq. (15) is not contained
GI.                                                                               the noise and the noise enhancement is mitigated. Therefore, the
   Figs. 4, 5, and 6 show the BER of the conventional methods                     proposed method shows the approximately same BER
and the proposed method at Doppler frequency of 10 Hz. In Fig.                    performance compared with inserting the case with GI in the
4, the BER performance of ISI cancellation without GI shows                       same block length. In Fig. 6, the BER of Nb = 1 shows a good
about 6 times improvement compared with not inserting GI                          performance, but large complexity is required. The BER
since ISI is eliminated. Moreover, the ISI cancellation shows                     performance of Nb = 1 with ZF shows about 67 times compared
about 2 times improvement compared with inserting GI with ZF.                     with MMSE of Nb = 16 for not inserting GI. On the other hand,
The BER performance of the proposed method for Nb = 16                            the proposed method of Nb = 8 shows the approximately same
shows about 6 times improvement compared with ISI                                 BER performance compared with ZF for Nb = 1 with inserting
cancellation without GI. However, the proposed method of Nb =                     GI. Therefore, the proposed method of Nb = 8 achieves the same
16 shows about 3 dB penalty compared with inserting GI with                       BER performance compared with ZF for Nb = 1with inserting
MMSE since the noise is enhanced due to the residual ISI and                      GI.
ICI. On the other hand, the proposed method of Nb = 8 shows the                      Fig. 7 shows the BER versus the block length Nb for the
best BER performance. It means that a strong multiuser                            conventional methods and the proposed method with Eb/N0 = 18

                                                                             79
dB at Doppler frequency of 10 Hz. For Nb = 16, the proposed                                   on OFDM access in IEEE802.16”, IEEE Commun. Mag., vol. 40, pp. 96-
                                                                                              103, April 2002.
method shows about 7 times improvement compared with ZF
                                                                                       [7]    WiMax Forum, http://www.wimaxforum.org/
with inserting GI. However, this shows about 9 times penalty                           [8]    S. J. Vaughan-Nichols, “Mobile WiMax: the next wireless battleground?”,
compared with inserting GI of MMSE. For Nb ≤ 8, the BER of                                    IEEE Computer Society, vol. 41, no. 6, pp. 16-18, June 2008.
proposed method approaches the BER performance of MMSE                                 [9]    W. Seo, H. Song, J. Lee, and D. Hong, “A new asymptotic analysis of
with inserting GI. Therefore, a multiuser diversity is effective to                           throughput enhancement from selection diversity using a high SNR
                                                                                              approach in multiuser systems”, IEEE Trans. on Wireless Commun., pp.
mitigate the ISI and ICI for Nb ≤ 8.                                                          55-59, vol. 8, no. 1, Jan. 2009.
   Fig. 8 shows the throughput of the conventional methods and                         [10]   S. Kwack, H. Seo, and B. G. Lee, “Suitability-based subcarrier allocation
the proposed method at Doppler frequency of 10 Hz. The                                        for multicast services employing layered video coding in wireless OFDM
throughput Ttp is defined as                                                                  System”, Proc. of VTC2007, pp. 1752-1756, Oct. 2007.
                                                                                       [11]   J. Hui and Y. Zhou, “Enahanced rate adaptive resource allocation scheme
                           Nd  Nc  C  R                                                    in downlink OFDMA system”, Proc. of VTC2006, vol. 5, pp. 2464-2468,
                   Ttp                             
                           N p  N d  T  1  Pper ,
                                                                         (30)                 May 2006.
                                                                                       [12]   C. Ahn, “Reinforced multiuser diversity using frequency symbol
                                                                                              spreading and adaptive subcarrier block selection for OFDMA”, IEICE
where C is the modulation level and Pper is the packet error rate ,                           Technical Report, vol. 109, no. 266, CS2009-44, pp. 13-18, Nov. 2009.
respectively. In this simulation, we assume that the GI length is                      [13]   S. Trautmann and N. J. Fliege, “Perfect equalization for DMT systems
                                                                                              without guard interval”, IEEE Journal on Selected Areas in Commun., pp.
Ts/4. In this case, the symbol duration T is 11.2 μs, and the                                 987-996, nol 20, no. 5, June 2002.
maximum throughput with GI would be 20 % degraded from Eq.                             [14]   G. Mkrtchyan, K. Naito, K. Mori, and H. Kobayashi, “ML time domain
(30). Thus, the maximum throughput of the proposed method                                     channel estimation and equalization for OFDM without guard interval”,
for Nb = 16 shows the improvement of about 20 % compared                                      Proc. of ECTI-CON2005, pp. 473-476, May 2005.
                                                                                       [15]   T. Kobori and F. Takehata, “An application of frequency diversity to
with inserting GI. Therefore, the proposed method of Nb = 16                                  OFDM transmission based on overlap-FDE”, IEICE Trans. Commun.,
achieves the enhancement of the maximum throughput.                                           vol. J93-B, no. 12, pp. 1651-1659, Dec. 2010.
Moreover, the proposed method of Nb = 8 shows the best                                 [16]   Y. Ida, C. Ahn, K. Kamio, H. Fujisaka, and K. Haeiwa, “Large delay
throughput performance. This means that a strong multiuser                                    spread cancellation based on replica signal for HTRCI-MIMO/OFDM”,
                                                                                              Proc. of WPMC’09, S-11-4, pp. 1-5, Sep. 2009.
diversity also enhances the throughput performance.


                            V. CONCLUSION
   In this paper, we have proposed the ISI and ICI
compensations to enhance the throughput performance for a
MUDiv/OFDMA without GI. When the GI is not inserted, the
system performance is significantly degraded due to the ISI and
ICI. Therefore, we have performed ISI and ICI compensations.
From the simulation results, the proposed method shows the
approximately same BER performance compared with the case
with inserting GI. For the throughput performance, the proposed
method improves about 20 % compared with the case with
inserting GI. As a result, the proposed method achieves the
enhancement of the maximal throughput. Moreover, the
proposed method of Nb = 8 shows the best performance both the
BER and the throughput. Therefore, a strong multiuser diversity
enhances the BER and the throughput performances.


                               REFERENCES
[1]   2010 White paper information and communications in Japan,
      http://www.soumu.go.jp/johotsusintokei/whitepaper/ja/h22/pdf/index.ht
      ml
[2]   L. Cimini, “Analysis and simulation of digital mobile channel using
      OFDM”, IEEE Trans. Commun., vol. 33, no. 7, pp. 666-675, July 1985.
[3]   J.A.C. Bingham, “Multicarrier modulation for data transmission: an idea
      whose time has come”, IEEE Commun. Mag., vol. 28, pp. 5-14, May
      1990.
[4]   ETSI ETS 301 958, “Digital video broadcasting (DVB); interaction
      channel for digital terrestrial television (RCT) incorporating multiple
      access OFDM”, ETSI, Tech. Rep., March 2002.
[5]   “IEEE draft standard for local and metropolitan area network-part 16: Air
      interface for fixed broadband wireless access systems - medium access
      control modifications and additonal physical layer specifications for 2-
      11GHz”, IEEE LAN MAN Standards Committee, 2002.
[6]   I. Koffman and V. Roman, “Broadband wireless access solutions based

                                                                                  80

				
DOCUMENT INFO
Shared By:
Tags:
Stats:
views:47
posted:7/10/2011
language:English
pages:7
Description: Cyber Journals: Multidisciplinary Journals in Science and Technology: May Edition, 2011, Vol. 2, No. 5