Performance Analysis of a Novel OFDM System Based on Dual .pdf

Document Sample
Performance Analysis of a Novel OFDM System Based on Dual .pdf Powered By Docstoc
					         Performance Analysis of a Novel OFDM System Based on Dual – Tree
                      Complex Wavelet Transform (DT-CWT)

                            Mohamed H. M. Nerma1, Nidal S. Kamel2 and Varun Jeoti3
                                   Electrical & Electronic Engineering Department,
                                   University Technology PETRONAS, Malayisa
          E-mail: , , and

                      As demand for higher data rates is continuously rising, there is always a need
                      to develop more efficient wireless communication systems. The work
                      described in this paper is an effort in this direction. We have proposed a
                      novel OFDM system base on DT-CWT. In the proposed scheme, DT-CWT
                      is used in the place of FFT. The proposed scheme offers the best PAPR
                      performance than the conventional OFDM and wavelet packet modulation
                      (WPM) systems at the expense of acceptable computational complexity
                      without using any pruning techniques. The complementary cumulative
                      distribution function (CCDF) of PAPR for the proposed scheme signal
                      achieves about 3 dB improvement in PAPR over the traditional OFDM and
                      WPM signals at 0.1% of CCDF. Also the proposed scheme achieves
                      excellent improvements in BER over conventional OFDM and WPM
                      systems. The need for CP is eliminated in the system design due to the good
                      orthogonality and time frequency localization proprieties of the wavelet.

                      Keywords: OFDM, WPT, CWT, DT-CWT, FFT, MCM, BER, PAPR.

                                                                1.1 Wavelet Modulation
    1     INTRODUCTION                                              Wavelet Transform (WT) is a relatively new
                                                                transform compared to the discrete Fourier transform
        Orthogonal frequency division multiplexing              (DFT).      WT      provides    the   time-frequency
    (OFDM) and wavelet packet modulation (WPM)                  representation of signals, whereas DFT gives only
    have emerged as an efficient multicarrier modulation        the frequency representation. The properties of
    schemes for wireless, frequency selective,                  wavelet, such as localization in time and frequency,
    communication channels. Ease of implementation,             orthogonality across scale and translation presents to
    high spectral efficiency, resilience to impulse noise       a new perspective in digital communication. It is
    and multipath are a few advantages of OFDM and              used as a modulation technique in many
    WPM systems. However, a major drawback in the               communication fields including multicarrier
    signals of these two systems are their large envelope       modulation (MCM) and wireless communication [6].
    function, which limits the efficiency of the non-           The conventional multicarrier systems are DFT
    linear power amplifiers specific to wireless                based systems. In MCM, the broadband channel
    communications by forcing them to operate at lower          splits into larger number of sub-channels. The high
    average power. This problem is quantified by the            data rate bit-streams are divided into parallel sub
    Peak – to – average power ratio (PAPR) and results          streams with lower data rate [7].
    from the superposition of a large number of usually             However, DFT based MCM systems suffer from
    statistically independent sub-channels that can             the high side lobes due to rectangular shaped DFT
    constructively sum up to high peaks. Also from the          window and also these systems waste precious
    central limit theorem (CLT) [1], this causes the            bandwidth due to the redundant cyclic prefix (CP).
    OFDM and WPM signals to have complex Gaussian               Moreover, the pulse shaping function used to
    process behavior and the instantaneous power is chi-        modulate each subcarrier extends to infinity in the
    square distributed. Various schemes have been               frequency domain this leads to high interference and
    developed to reduce high PAPR in OFDM [2] [3],              lower performance levels. WT based MCM systems,
    and WPM signals [4] [5].                                    namely wavelet based OFDM systems (WOFDM)
                                                                can help to mitigate these problems. Therefore,
                                                                conventional DFT based orthogonal systems are
                                                                being replaced by wavelet based transceivers. The

UbiCC Journal, Volume 4, Number 3, August 2009                                                                           813
    wavelet based transceiver uses quadrature mirror
    filters (QMF) in the synthesis and analysis filter                                                               2
    banks (FBs). The wavelet filters posses the
    advantages of having greater side-lobe attenuation              To reduce the PAPR in OFDM and WPM
    and requires no CP [8]. WPT modulation in wireless          systems, several techniques have been proposed,
    communication has been proposed in [9]. The                 which basically can be divided in three categories.
    characteristics of multicarrier modulated signal are        First, there are signal distortion techniques, which
    dependent on the basis functions being used in a            reduce the peak amplitudes simply by nonlinearly
    modulation scheme. Therefore, using WPT as basis            distorting the OFDM signal at or around the peaks.
    functions, the sensitivity to multipath channel             Examples of distortion techniques are clipping, peak
    distortion, inter symbol interference (ISI), inter          windowing and peak cancellation. The second
    carrier interference (ICI), and synchronization can be      category is coding techniques that use a special
    reduced as compared to traditional OFDM and the             forward-error correcting code set that excludes
    system performance of WOFDM with reference to               OFDM symbols with a large PAPR. The third
    ISI, ICI and signal-to-noise ratio (SNR) is shown to        technique is based on scrambling each OFDM
    be far better than the conventional OFDM (DFT-              symbol with different scrambling sequences and
    OFDM) [9], [10].                                            selecting that sequence that gives the smallest PAPR
    1.2 Peak – to – Average Power Ratio (PAPR)                      This paper is organized as follows: In section II
        The PAPR of the baseband transmitted signal             we discuss the dual-tree complex wavelet transform
          is defined as the ratio of the peak power             (DT-CWT); in section III we discuss the half-sample
    (          max |      |2 ); i.e., the maximum power of      delay condition; in section IV we make a
                                                                comparisons between traditional OFDM and WPM
    the transmitted signal over the average power
                                                                systems; in section V we discuss the OFDM based
    (        E |     |2 ). In digital implementations of
                                                                on DT-CWT; in section VI we discuss the PAPR in
    communications transceivers, rather than using the
                                                                OFDM based on DT-CWT; the BER results are
    continuous time signal             in PAPR computation,
                                                                presented in section VII; and we conclude this paper
    we instead work with          , the discrete time samples
                                                                in section VIII by discussion the simulation results.
    of x(t), provided that an oversampling factor of at
    least 4 is used. PAPR is then expressed as [11]:
                                                                2 THE DUAL – TREE COMPLEX WAVELET
                                                                  TRANSFORM (DT-CWT)
                    max |          |
                     E |       |
                                                                   Since the early 1990s the WT and WPT have
                                                                received more and more attention in modern
    where      . denotes ensemble average calculated
                                                                communications and have been widely used in
    over the duration of the OFDM or WPM symbols.               wireless communication [14]. A number of
        In both OFDM and WPM systems, the signal                modulation schemes based on wavelets have been
    going into the channel         is a sum of random           proposed [15 - 23]. In fact, complex wavelet
    symbols modulating orthogonal basis functions.
                                                                transform (CWT) is applied perfectly to digital
    Based on the CLT, it is claimed that    is complex          image processing. Kingsbury [24 - 28] introduced
    Gaussian and its envelope follows a Rayleigh                and made a concrete description of DT-CWT. The
    distribution. This implies a large PAPR. A high             DT-CWT employs two real discrete WT (DWT); the
    PAPR of the transmitted signals demands a very              upper part of the FB gives the real part of the
    linear transmission path and limits the practical           transform while the lower one gives the imaginary
    deployment of low-cost non-linear power amplifiers          part. This transform uses the pair of the filters (   ,
    forcing them to operate with reduced power                         the low-pass/high-pass filter pair for the upper
    efficiency. Driving amplifiers operating close to           FB respectively) and (            ,           the low-
    saturation with signals of high PAPR results in the         pass/high-pass filter pair for the lower FB
    generation of unwanted spectral energy both in-band         respectively) that are used to define the sequence of
    and out-of-band, which in turn reduces the systems          wavelet function        and scaling function         as
    performance and gives rise to adjacent channel              follows
    interference (ACI).
        In this work, the performance of the OFDM
                                                                               √2 ∑           2                      3
    based on DT-CWT in PAPR reduction is
    demonstrated through the CCDF of PAPR, which is
                                                                               √2 ∑            2                     4
    a performance metric independent of the transmitter
    amplifier. Given the reference level PAPR0 > 0, the
                                                                Where 1            1 0         ,      the      wavelet
    probability of a PAPR being higher than the
                                                                function       , the scaling function         and the
    reference value is the CCDF and is expressed as
    follows [12]:                                               high-pass filter for the imaginary part           are

UbiCC Journal, Volume 4, Number 3, August 2009                                                                            814
    defined similarly. The two real wavelets associated
    with each of the two real transform are
    and      . To satisfying the perfect reconstruction
    (PR) conditions, the filters are designed so that the
    complex      wavelet                                is
    approximately analytic. Equivalently, they are
    designed so that        is approximately the Hilbert
    transform of      .


    The analysis (decomposition or demodulation) and
                                                             Figure 2: The Inverse dual tree discrete CWT (IDT-
    the synthesis (reconstruction or modulation) FBs
                                                             DCWT) Synthesis (modulation) FB.
    used to implement the DT-CWT and their inverses
    are illustrated in fig. 1 and fig. 2 respectively. The
    inverse of DT-CWT is as simple as the forward
    transform. To invert the transform, the real part and
                                                                                             0.5                    8
    the imaginary part are each inverted.
                                                             In practical implementation of the DT-CWT, the
                                                             delay condition (7) and (8) will be satisfied only
                                                             approximately; the wavelets            and      will
       The two low pass filters should satisfy a very
    simple property: one of them should be                   form only an approximate Hilbert pair; and the
    approximately a half-sample shift of the other [29]      complex wavelet                        will be only
                                                             approximately analytic. While the FT is based on
                            0.5                         6    complex valued oscillating cosine and sine
                                                             components form a complete Hilbert transform pairs;
    Since          and      are defined only on the          i.e., they are 90 out of phase with each other.
    integers, this statement is somewhat informal.           Together they constitute an analytic signal   that is
    However, we can make the statement rigorous using        supported on only one-half of the frequency axis
    FT.     In    [5]   it  is    shown    that,    if             0 [24].
                              ,   then                   .
    The converse has been proven in [30] [31], making        3.1 Filter Design for the DT-CWT
    the condition necessary and sufficient. The necessary         There are various approaches to the design of
    and sufficient conditions for biorthogonal case were     filters for the DT-CWT, such as linear-phase
    proven in [32].                                          biorthogonal method, quarter shift method, and
                                                             common factor method. These filters are satisfied the
                                                             following desired properties:
                                                                1- Approximately half-sample delay property.
                                                                2- PR (orthogonal or biorthogonal).
                                                                3- Finite support (finite impulse response (FIR)
                                                                4- Vanishing moments/good stop-band.
                                                                5- Linear phase filters.
                                                             It turns out that the implementation of the DT-CWT
                                                             requires that the first stage of the dual-tree (DT) FB
                                                             requires a different set of filters from the succeeding
                                                             stages, and the succeeding stages can be used same
                                                             sets of filters. See fig. (1). If the same PR filters are
                                                             used for each stage, then the first several stages of
                                                             the FB will not be approximately analytic [24].
    Figure 1: The dual tree discrete CWT (DT-DCWT)
    Analysis (demodulation) FB.                              4    OFDM AND WPM SYSTEMS

    We can rewrite the half-sample delay condition in            Traditionally, OFDM is implemented using FFT.
    terms of magnitude and phase function separately as      This transform however has the drawback that is
    shown in (7) and (8) as follows.                         uses a rectangular window, which creates rather high
                                                             sidelobes. Moreover, the pulse shaping function used
                                                             to modulate each subcarrier extends to infinity in the

UbiCC Journal, Volume 4, Number 3, August 2009                                                                           815
    frequency domain (FD) this leads to high                 receiver side a DT-CWT is used in place of FFT
    interference and lower performance levels. ISI and       block in conventional OFDM system or in place of
    ICI can be avoided by adding a CP to the head of         DWPT block in WPM system. Data to be transmitted
    OFDM symbol. Adding CP can largely reduce the            are typically in the form of a serial data stream. PSK
    spectrum efficiency. The WPM system has a higher         or QAM modulations can be implemented in the
    spectral efficiency and providing robustness with        proposed system the choice depends on various
    regard to inter-channel interference than the            factors like the bit rate and sensitivity to errors. The
    conventional OFDM system, because of the out-of-         transmitter accepts modulated data (in this paper we
    band energy (low sidelobes). Moreover WPM is able        use 16 QAM). This stream is passed through a serial
    to decompose T-F plane flexibly by arranging FB          to parallel (S/P) converter, giving N lower bit rate
    constructions. In addition, using FFT in the             data stream, and then this stream is modulated
    traditional OFDM gives resolution only in the FD         through an IDT-CWT matrix realized by an N-band
    while using WPT in WPM system gives resolution in        synthesis FB. Before the receiver can demodulate the
    both FD and time domain (TD) [16].                       subcarriers, it has to perform the synchronization.
        WPM system do not required CP, thereby               For the proposed system, known data interleaved
    enhancing the spectrum efficiency. According to the      among unknown data are used for channel
    IEEE broadband wireless standard 802.16.3,               estimation. Then, the signal is down sampled by N
    avoiding CP gives wavelet OFDM an advantage of           and demodulated using elements of the DT-CWT
    roughly 20% in bandwidth (BW) efficiency.                matrix realized by an N-band analysis FB. The signal
    Moreover as pilot tones are not necessary for wavelet    is equalized after DT-CWT stage.
    based OFDM system, they perform better in
    comparison to existing OFDM systems like 802.11a
    or HiperLAN, where 4 out of 52 sub-bands are used
    for pilots. This gives wavelet based OFDM system
    another 8% advantage over typical OFDM
    implementations [39]. We expected the OFDM based
    on DT-CWT will take all the advantages of WPM
        However, a major problem of the common
    discrete WPT (DWPT) is its lack of shift invariance;
    this means that on shifts of the input signal, the
    wavelet coefficients vary substantially. The signal
                                                             Figure 3: DT-CWT modulation (DT-CWTM)
    information may even not be stationary in the sub-       functional block diagram.
    bands so that the energy distribution across the sub-
    bands may change [15]. To overcome the problem of            IDT-CWT works in a similar fashion to an IFFT
    shift dependence, one possible approach is to simply     or IDWPT. It takes as the input QAM symbols and
    omit the sub-sampling causing the shift dependence.      outputs      them     in    parallel   time-frequency
    Techniques that omit or partially omit sub-sampling      “subcarriers”. In fig. (2) as the synthesis process, it
    are also known as cycle spinning, oversampled FBs        can be shown that the transmitted signal,            is
    or undecimated WT. However, these transforms are         constructed as the sum of M waveform individually
    redundant [33], which is not desirable in multicarrier   modulated with the QAM or PSK symbols [34], φ t
    modulation. As an alternative, we used a non-            is the scaling function, ψ t is the wavelet function,
    redundant WT that achieves approximately shift
                                                             and a is kth symbol, i 1, 2, … , N as follows:
    invariance [34], this transform yields to complex
    wavelet coefficients that modulate the data stream in                                                .                     9
    the same way that WPM do.
    5    OFDM BASED ON DT-CWT                                                                                            .    10
                                                                                     ,       ,       ,           ,

        In this study, simulations are focusing on using
    DT-CWT in the OFDM system as a non-redundant
    WT that can achieves approximately shift invariance
    [35 – 38]. Similar to the conventional OFDM and                     /   ,    /       ,                   ,       ,       . 11
    WPM systems, a functional block diagram of OFDM                                              /
    based on DT-CWT is shown in fig. (3). At the
    transmitter an inverse DT-CWT (IDT-CWT) block is         Where , is a constellation encoded            data
    used in place of inverse FFT (IFFT) block in             symbol modulating the         DT-CWT function.
    conventional OFDM system or in place of inverse          The IDT-CWT synthesis a discrete representation of
    DWPT (IDWPT) block in WPM system. At the                 the transmitted signal as sum of M waveforms

UbiCC Journal, Volume 4, Number 3, August 2009                                                                                      816
    shifted in time that embed information about data              sequence x n has high PAPR. Furthermore, to
    symbols. Those waveforms are built by successive               demonstrate the similarities between power spectrum
    iterations of ^ , ^ and ^ , ^ . The DT-CWT at the              density (PSD) characteristic of conventional OFDM,
    receiver recovers the transmitted symbols           ,          WPM and the proposed scheme, the simulated PSD
    through the analysis formula exploiting orthogonality          characteristics are presented in fig. 4. This fig. was
    properties of DT-CWT and schematically                         shows that the proposed scheme performs better than
    represented in fig. (1).                                       the other two systems.
    In the baseband equivalent OFDM transmitter with                   The simulation parameters are documented as
              frame of N QAM symbols,              ,               follows: Modulation type is 16-QAM; the number of
     0,1, … ,     1, the OFDM frame is given by:                   subcarriers is 64 subcarriers; a wavelet packet base is
                                                                   Daubechies-1 (DAUB-1); maximum tree depth (D =
                                                                   7); PAPR threshold is 2dB; shaping filter is Raised
                                                              12   Cosine (rolloff factor      0.001, upsampler = 4); DT-
                                                                   CWT using different filters (LeGall 5,3 tap filters
                                                                   (leg), Antonini 9,7 tap filters (anto), Near Symmetric
    While for WPM system, the transmitted signal                   5,7 tap filters (n-sym-a), and Near Symmetric 13,19
    is constructed as the sum of M wavelet packet                  tap filters (n-sym-b)) for the first stage of the FB and
    function          individually modulated with the QAM          (Quarter Sample Shift Orthogonal 10,10 tap filters
    symbols as the case in this paper‡.                            (q-sh-06) only 6,6 non-zero taps, Quarter Sample
                                                                   Shift Orthogonal 10,10 tap filters (q-sh-a) with 10,10
                                                                   non-zero taps, unlike q-sh-06, Quarter Sample Shift
                               ,                              13   Orthogonal 14,14 tap filters (q-sh-b), Quarter Sample
                                                                   Shift Orthogonal 16,16 tap filters (q-sh-c), and
                                                                   Quarter Sample Shift Orthogonal 18,18 tap filters (q-
       The construction of discrete versions of                    sh-d)) for the succeeding stages of the FB.
    transmitted waveforms for the conventional OFDM
    and WPM systems using (12) and (13) is quite                   6    PAPR ANALYSIS RESULTS
    similar. For any time index n, both waveforms are
    sum of random symbols a or a , .                                    Consider the simple case where all the sub-
                                                                   symbols are independently and identically distributed
                                                                   (i.i.d). then, by the CLT, the real and imaginary parts
                                                                   of the N-point IFFT output have mutually
                                                                   independent Gaussian probability           distribution
                                                                   function with zero mean and standard deviation to σ .
                                                                   The instantaneous power of baseband signal          , is
                                                                   defined by


                                                                       We can characterize the instantaneous power as a
                                                                   chi-square distribution with two degrees of freedom
    Figure 4: PSD of the Conventional OFDM, WPM
    and OFDM based on DT-CWT.                                                        1
                                                                                                ,       0               15

      In order to achieve fair comparisons, same
    simulation parameters are used. The simulations                If E |    | is normalized to unity, then the CCDF
    were carried out for conventional OFDM with a 64               of the PAPR is given by:
    subcarriers using a 16 QAM modulation. The
    Daubechies-1 (Daub-1) wavelet packet bases were                          Pr             1       1                   16
    used to construct the wavelet packet trees in WPM
    system. For the proposed scheme, near symmetric
    13,19 tap filters and quarter sample shift orthogonal          Where N is the number of subcarriers.
    14 tap filters were used to construct the real and the             In order to analyze PAPR, we generate the
    imaginary part of DT-CWT respectively. From the                transmitted waveforms using 16 QAM modulation
    CLT, x n is complex Gaussian distributed, and the              with 64 subcarriers for all these systems. Fig. 5
                                                                   shows, in the TD, the envelope of the proposed
                                                                   system. For the comparison, we also plotted the
     We use φ n for the wavelet packet function and ψ n for the    envelope of the conventional OFDM and WPM
    DT-CWT function to avoid any confusion.

UbiCC Journal, Volume 4, Number 3, August 2009                                                                                817
    waveforms corresponding to the same information           stage of the FB and quarter sample shift orthogonal
    symbol pattern. The transmitted envelopes for the         (q-sh) 14 tap filters in the succeeding stages, OFDM
    conventional OFDM and WPM systems illustrate              DT-CWT_2 using (n-sym 13,19 with q-sha 10 (10
    approximately similar behavior, where the peak is         non zero taps) filters), OFDM DT-CWT_3 using
    about 2.25, while the transmitted envelope for the        (antonini (anto) 9,7 tap filters with q-sh0 10 (only 6
    proposed system demonstrates better behavior than         non zero taps) filters), OFDM DT-CWT_4 using
    the other two systems, where the peak is only about       (anto 9,7 with q-sh 14 filters), OFDM DT-CWT_5
    1.25 and this is the reason for that the proposed         using (n-sym 5,7 with q-sh 14 filters), OFDM DT-
    system gives better result for PAPR than the other        CWT_6 using (LeGall (leg) 5,3 tap filters with q-sh
    two systems.                                              14 filters), OFDM DT-CWT_7 using (n-sym 5,7
                                                              with q-sh 16 filters) and OFDM DT-CWT_8 using
                                                              (leg 5,3 with q-sh 18 filters).
                                                                  The results in fig. 7 show that there is no
                                                              observed degradation as a result of using different set
                                                              of mismatching filters in the design of the proposed

    Figure 5. The Envelope of the Conventional OFDM,
    WPM and OFDM based on DT-CWT.

                                                              Figure 7: The effect of using different set of filters
                                                              in design of the OFDM based on DT-CWT.

    Figure 6: CCDF the Conventional OFDM, WPM
    and OFDM based on DT-CWT.

        The results for PAPR are best quantified using
    CCDF. In fig. 6, for 64 subcarriers with 16 QAM
    modulation, the CCDF plots for the proposed system,
    conventional OFDM and WPM system are shown.               Figure 8: CCDF of PAPR for 16-QAM modulated
                                                              conventional OFDM symbol with various values of
    This figure shows that the OFDM based on DT-
                                                              subcarriers (N).
    CWT offers the best PAPR performance without
    using any reduction techniques. The proposed
                                                                  Again, the results for PAPR were repeated in
    scheme signal achieves about 3 dB improvement in
                                                              both fig. 8 (for the conventional OFDM system) and
    PAPR over the traditional OFDM and WPM signals
                                                              fig. 9 (for the proposed system) using different
    at 0.1% of CCDF while the other two systems, are
                                                              numbers of subcarriers (64, 128, 256, 512, and 1024)
    approximately given same results of PAPR.
                                                              with 16QAM modulation. We observe from these
        The simulation results for PAPR were also
                                                              figures that the PAPR increases as the number of
    repeated with 16 QAM modulation and 64
                                                              subcarriers numbers (N) increases. As shown in fig.
    subcarriers (in fig. 7) using different set of filters.
                                                              6, 7 and in fig. 8, 9 the PAPR performance of DT-
    OFDM DT-CWT_1 illustrate the PAPR when using
                                                              CWT based OFDM systems is better than the
    near-symmetric (n-sym) 13,19 tap filters in the first

UbiCC Journal, Volume 4, Number 3, August 2009                                                                          818
    conventional OFDM system or even WPM systems.
    As we saw in figure 5 showing the time-domain
    signal envelop of all the systems, this improvement
    in PAPR performance is explained from lower peaks
    of DTCWT systems observed in the fig. 5.

                                                            Figure 11: The effect of using different type of
                                                            filters in BER for OFDM based on DT-CWT.

                                                            8    CONCLUSION
    Figure 9: CCDF of PAPR for 16-QAM modulated
    OFDM based on DT-CWT symbol with various                    In this paper a new OFDM scheme that is based
    values of subcarriers (N).                              on DT-CWT is proposed. Comparing the proposed
                                                            scheme in terms of PAPR and BER with the
    7    BER ANALYSIS RESULTS                               traditional OFDM and WPM systems we see that the
                                                            proposed scheme offers 3dB better PAPR
         The results given in this section compare the      performance over the conventional OFDM and
    BER in the OFDM based on DT-CWT, with that for          WPM systems at 0.1% of CCDF. While the
    traditional OFDM, and WPM. Also, same simulation        conventional OFDM and WPM systems shows
    parameters are used to achieve a fair comparison.       similar behavior. Simulation results shows that there
    The results of BER in OFDM based on DT-CWT              is no observed PAPR and BER degradation as a
    using different set of filters are also shown in this   result of using different set of mismatching filters in
    section.                                                DT-CWT based system. The proposed scheme
         The results for BER are shown that the proposed    outperforms the traditional OFDM and WPM
    scheme gives excellent improvements in BER over         systems in term of BER. Also we found that the
    conventional OFDM and WPM systems. At the same          conventional OFDM system gives better results of
    time the conventional OFDM outperform the WPM           BER than WPM system.
    system in term of BER. As shown in fig. 10.
         The simulation results for BER were repeated
    using different set of filters (OFDM DT-CWT-1,          ACKNOWLEDGMENTS
    OFDM DT-CWT-2, …, OFDM DT-CWT-8) and is                     Authors wish to thanks Prof. Nick G. Kingsbury
    shown in fig. 11. We see there is degradation of BER    (University of Cambridge, United Kingdom) and Dr.
    as a result of using different set of mismatching       Mohan Baro (Dalhousie University, Canada) whose
    filters in the design of the proposed scheme.           comments and suggestions have largely contributed
                                                            to improve this paper.

                                                            A. CCDF Approximations
                                                                We can characterize the instantaneous power as a
                                                            chi-square distribution with two degrees of freedom

                                                                                      ,        0                17

                                                            As a result, the cumulative distribution function
                                                            (CDF) is defined as:
    Figure 10: BER performance of conventional
    OFDM, WPM and OFDM based on DT-CWT.

UbiCC Journal, Volume 4, Number 3, August 2009                                                                        819
                                                                             1              1
                                                                                        .                                22
                                              1                 18          √2              √2

    If E |    | is normalized to unity, then the CCDF                Note that the identity matrix on the right-hand side of
    of the PAPR is given by:                                         (22) is twice the size of those on the left-hand side.
                                                                     Therefore if the two real DWTs are orthonormal
                                                                     transforms, then the DT-CWT satisfies .               ,
                          1       1                             19   where * denotes the conjugate transpose.

    Where N is the number of subcarriers, however, this
    approximation is not close to the experimental results           9    REFERENCES
    as the assumption made in deriving CCDF that the
                                                                         [1] Papoulis A., Pillai S. U., “Probability
    samples should be mutually uncorrelated is no longer                      Random         Variables     and    Stovhastic
    valid when oversampling is employed [40]. There                           Processes,” New York: McGraw-Hill Inc.,
    has been several attempts to determine the closed for                     2002.
    approximations for the distribution of PAPR. Some                    [2] M. Breiling, S. H. muller, and J. B. Huber,
    of the approximations are shown in Table 1.                               “SLM peak power reduction without
                                                                              explicit side information”, IEEE Commun.
                                                                              Lett., vol. 5, no. 6, pp. 239-241, 2001.
    Table 1: Approximation to CCDF of PAPR.                              [3] J. Tellado, Ed., Multicarrier Modulation
                                                                              with low PAPR Application to DSL and
                        CCDF                          Remarks                 wireless. New York: Kluwer Academic
                                                                              Publishers, 2002.
                                                       = 2.8 and         [4] M. Baro and J. Ilow, “PAPR reduction in
                                                       N 64                   wavelet packet modulation using tree
                                                        [40]                  pruning”, in 2007 IEEE 65th Vehicular
                                                                              Technology Conference VTC 2007 –
                                                                              spring, Dubin, Ireland, Apr. 2007.
                                                                         [5] M. Baro and J. Ilow, “PAPR reduction in
                                                         [41]                 OFDM using wavelet packet pre-
                                                                              processing”, Consumer Communications
                                                                              and Networking Conference, 2008. CCNC
                                                                              2008. 5th IEEE.
                                                                         [6] M K Lakshmanan, H Nikookar, "A review
                                                         [42]                 of Wavelets for Digital Communication,"
                                                                              Wireless Personal Communication (2006)
                                                  /   N and are               37: 387- OFDM* 420, Springer 2006.
                                      √               large [43]         [7] J. A. C. Bingham, "Multicarrier modulation
                                                                              for data transmission: An Idea Whose Time
                                                                              Has Come," IEEE Communications
                                  /                                           Magazine, vol. 28, no.5, pp. 5-14, 1990.
                                                                         [8] H. M. Newlin, "Developments in the use of
                                                                              wavelets in communication systems," TRW
                                                                              Systems & Information Technology Group,
                                                                              Sunnyvale, California.
                                                                         [9] A. Jamin, P. Mahonen, "Wavelet packet
    B. DT-CWT and IDT-CWT                                                     modulation for wireless communications,"
        If the two real DWTs are represented by the                           Wireless Communications and Mobile
                                                                              Computing 5 (2): 123-137 (2005).
    square matrices     for the upper part and for the                   [10] C. V. Bouwel, J. Potemans, S. Schepers, B.
    lower part, then the DT-CWT can be represented by                         Nauwelaers and A.V. de Capelle, "Wavelet
    the following form.                                                       Packet Based Multicarrier Modulation,"
                                                                              Proc. IEEE Benelux Symposium on
               1                                                              Communications             and       Vehicular
                              .                                 20            Technology, Leuven, Belgium, 19 October
              √2                                                              2000.
                                                                         [11] C. Schurgers and M. B. Srivastava, “A
    And the IDT-CWT is given by                                               systematic approach to peak – to – average
                                                                              power ratio in OFDM,” in SPIE’s 47th
                                                                              Annual Meeting, San Diego, CA, 2001, pp.
                                          .                     21            454-464.
                   √2                                                    [12] S. H. Han and J. H. Lee, “An Overview of
                                                                              to peak – to – average power ratio reduction
    Note that the complex sum/difference matrix in (20)                       techniques for multimedia transmission,”
    is unitary (its conjugate transpose is its inverse).                      IEEE Wireless Communications, vol. 12,
                                                                              no. 2, pp. 56-65, 2005.
                                                                         [13] Richard Van Nee and Ramjee Prasal,
                                                                              “OFDM          for     Wireless    Multimedia

UbiCC Journal, Volume 4, Number 3, August 2009                                                                                 820
             Communications”, P. cm Artech House                  Vancouver, BC, Canada, Sept. 10–13, 2000,
             Universal Personal Communications Series,            vol. 2, pp. 375–378.
             Inc. Boston. London, 2000.                      [28] N.G. Kingsbury, “Complex wavelets for
        [14] Panchamkumar D Shukla, “Complex                      shift invariant analysis and filtering of
             wavelet       Transforms       and      Their        signals,” Appl. Comput. Harmon. Anal.,
             Applications” Master Thesis 2003. Signal             vol. 10, no. 3, pp. 234–253, May 2001.
             Processing      Division.    University    of   [29] I W. Selenick, “Hilbert transform pairs of
             Strathclyde Department of Electronic and             wavelet bases,” IEEE Signal Processing
             Electrical Engineering.                              Lett., vol. 8, no. 6, pp. 170-173, June 2001.
        [15] M. Guatier, J. Lienard, and M. Arndt,           [30] H. Ozkaramanli and R. Yu, “on the phase
             “Efficient Wavelet Packet Modulation for             condition and its solution for Hilbert
             Wireless Communication”, AICT’07 IEEE                transform pairs of wavelet bases,” IEEE
             Computer Society, 2007.                              Trans. Signal Processing, vol. 51, no. 12,
        [16] A. Jamin, and P. Mahonen, “Wavelet Packet            pp. 2393–3294, Dec. 2003.
             Modulation for Wireless Communications”,        [31] R. Yu and H. Ozkaramanli, “Hilbert
             Wiley Wireless Communications and                    transform pairs of orthogonal wavelet bases:
             networking, Journal, vol. 5, no. 2, pp. 123-         Necessary and sufficient conditions”, IEEE
             137, Mar. 2005.                                      Trans. Signal Processing IEEE Transactions
        [17] Xiaodong Zhang and Guangguo Bi,                      on Signal Processing, vol. 53, no. 12, Dec.
             “OFDM Scheme Based on Complex                        2005.
             Orthogonal                           Wavelet    [32] R. Yu and H. Ozkaramanli, “Hilbert
             Packet”,        transform pairs of biorthogonal wavelet
             20844/00965270.pdf.                                  bases”, IEEE Transactions on Signal
        [18] M. Guatier, and J. Lienard, “Performance of          Processing, VOL. 54, NO. 6, JUNE 2006.
             Complex        Wavelet      packet      Based   [33] I. W. Selesnick, “the Double Density Dual-
             Multicarrier Transmission through Double             Tree DWT”, IEEE Transactions on Signal
             Dispersive Channel”, NORSIG 06, IEEE                 Processing, 52(5): 1304 – 1315, May 2004.
             Nordic Signal Processing Symposium              [34] J. M. Lina,”Complex Daubechies Wavelets:
             (Iceland), June 2006.
                                                                  Filter Design and Applications”, ISAAC
        [19] C. J. Mtika and R. Nunna, “A wavelet-                Conference, June 1997.
             based multicarrier modulation scheme,” in       [35] Mohamed H. M. Nerma, Nidal S. Kamel,
             Proceedings of the 40th Midwest                      and Varun jeoti, “PAPR Analysis for
             Symposium on Circuits and Systems, vol. 2,           OFDM Based on DT-CWT” Proceeding of
             August 1997, pp. 869–872.
                                                                  2008 Student Conference on Research and
        [20] N. Erdol, F. Bao, and Z. Chen, “Wavelet              Development (SCOReD 2008), 26-27 Nov.
             modulation: a prototype for digital                  2008, Johor, Malaysia.
             communication systems,” in IEEE Southcon
             Conference, 1995, pp. 168–171.                  [36] Mohamed H. M. Nerma, Nidal S. Kamel,
                                                                  and Varun jeoti, “On DT-CWT Based
        [21] A. R. Lindsey and J. C. Dill, “Wavelet               OFDM: PAPR Analysis” Multi-Carrier
             packet modulation: a generalized method              Systems & Solutions 2009, V. 41, pp. 207-
             for         orthogonally         multiplexed         217, SpringerLink April 26, 2009-
             communications,”        in     IEEE      27th        Dordrecht, Netherlands.
             Southeastern Symposium on System
             Theory, 1995, pp. 392–396.                      [37] Mohamed H. M. Nerma, Nidal S. Kamel,
                                                                  and Varun jeoti, “BER Performance
        [22] A. R. Lindsey, “Wavelet packet modulation            Analysis of OFDM System Based on Dual –
             for         orthogonally         multiplexed         Tree Complex Wavelet Transform in
             communication,” IEEE Transaction on                  AWGN Channel” Proceeding of 8th
             Signal Processing, vol. 45, no. 5, pp. 1336–         WSEAS International Conference on
             1339, May 1997.                                      SIGNAL PROCESSING (SIP '09), Istanbul,
        [23] C. V. Bouwel, J. Potemans, S. schepers, B.           Turkey, May 30 - June 1, 2009.
             Nauwelaers, and A. Van Caelle, “wavelet         [38] Mohamed H. M. Nerma, Nidal S. Kamel,
             packet Based Multicarrier Modulation”,               and Varun jeoti, “An OFDM System Based
             IEEE Communication and Vehicular                     on Dual Tree Complex Wavelet Transform
             Technology, SCVT 2000, pp. 131-138,                  (DT-CWT)” Journal: Signal Processing:
             2000.                                                An International Journal, Volume: 3, Issue:
        [24] Ivan W. Selesnick, Richard G. Baraniuk,               2, Pages: 14-21, May 2009.
             and Nick G. Kingsbury, “The Dual-Tree           [39] M. K. Lakshmanan and H. Nikookar, “A
             Complex Wavelet Transform,” IEEE Signal              Review of Wavelets for Digital Wireless
             Processing Mag, pp. 1053-5888, Nov 2005.             Communication”,           Wireless    Personal
        [25] N.G. Kingsbury, “The dual-tree complex               Communications Springer, 37: 387-420,
             wavelet transform: A new technique for               Jan. 2006.
             shift invariance and directional filters,” in   [40] R. Van Nee and A. De Wild, “reduction the
             Proc. 8th IEEE DSP Workshop, Utah, Aug.              peak-to average power ratio of OFDM” 48th
             9–12, 1998, paper no. 86.                            IEEE Vehicular Technology Conference,
        [26] N.G. Kingsbury, “Image processing with               Ottawa, Canada, May 18-21, 1998; 2072-
             complex wavelets,” Philos. Trans. R. Soc.            2076, IEEE New York, USA.
             London A, Math. Phys. Sci., vol. 357, no.       [41] H. Ochiai and H. Imai, “On the disitribution
             1760, pp. 2543–2560, Sept. 1999.                     of the peak-to-average power ratio in
        [27] N.G. Kingsbury, “A dual-tree complex                 OFDM signals”, IEEE Trans. Commun.
             wavelet      transform     with     improved         2001; 49, 282-289.
             orthogonality and symmetry properties,” in      [42] S. wei, D. L. Goeckel, and P. E. Kelly “A
             Proc. IEEE Int. Conf. Image Processing,              modern extreme value theory approach to

UbiCC Journal, Volume 4, Number 3, August 2009                                                                     821
             calculate the distribution of the peak-to-                         Nidal S. Kamel, received his
             average power ratio in OFDM systems”,                              Ph.D degree (Hons) in
             IEEE     International    Conference   on
             Communications, New York, Apr. 28- May                             telecommunication          and
             2, 2002; vol. 3, 1686-1690; IEEE New                               statistical signal processing
             York, USA.                                                         from        the      Technical
        [43] H. Yu, M. Chen and G. wei “Distribution of                         University      of    Gdansk,
             PAPR in DMT systems”, Electron. Lett.
             2003; 39, 799-801.                                                 Poland, in 1993. His research
        [44] R. Rajbanshi, A. M. Wyglinski and G. J.                            is     focused     on   linear
             Minden, “Multicarrier Transceivers: Peak-                          estimation, noise reduction,
             to-average power ratio reduction”, IEEE      pattern recognition, optimal filtering, and
             Commun.       Society.     WCNC      2008    telecommunications. Currently he is working for
                                                          Universiti Teknologi PETRONAS, Malaysia. He is
                                                          senior member of IEEE.

                                                                                  Varun Jeoti Jagadish,
                         Mohamed H. M. Nerma, was                                 received his Ph.D. degree
                         born in Khartoum, Sudan. He                              from Indian Institute of
                         received the B.Sc. degree in                             Technology Delhi India in
                         Electrical         Engineering                           1992. He worked on several
                         (Control)      from     Sudan                            sponsored R&D projects in
                         University of Science and                                IIT Delhi and IIT Madras
                         Technology (SUST), Sudan in                              during 1980 to 1989
                         1999. He received the M.Sc.                              developing Surface Acoustic
                         degree in Communication          Wave Pulse Compression filters, underwater optical
    Engineering from Karary Academy of Technology,        receivers etc.. He was a Visiting Faculty in
    Sudan in 2002. From 2002 to 2006 he was lecturer in   Electronics department in Madras Institute of
    the Sudan University of Science and Technology. He    Technology for about 1 year during 1989 to 1990
    is currently PhD. Student in University Technology    and joined Delhi Institute of Technology for next 5
    PETRONAS, Malaysia.                                   years till 1995. He moved to Electrical & Electronic
                                                          Engineering (E&E Engg) department of Universiti
                                                          Sains Malaysia in 1995 and joined E&E Engg of
                                                          Universiti Teknologi PETRONAS in 2001. His
                                                          research interests are in Wireless LAN and MAN
                                                          technologies, DSL technology and related signal

UbiCC Journal, Volume 4, Number 3, August 2009                                                                   822

Shared By:
liningnvp liningnvp http://