CHAPTER 1 INTRODUCTION

1.1 Introduction Next-generation wireless communication systems must fulfill the high data rate demand of intensive applications like multimedia services, data transfer, audio, streaming video etc. This has lead to the demand for future wireless terminals of being capable of connecting to various networks to support services like switched traffic, Internet Protocol (IP) data packets and broadband streaming services. Additionally, due to the growth of Internet applications and wireless users, many Wireless Local Area Network (WLAN) standards have been proposed including IEEE802.11 which permit mobile connectivity to the Internet. With a surging demand for wireless Internet connectivity, new WLAN standards have been developed including IEEE802.11b, popularly known as Wi-Fi, which provides up to 11 Mb/s raw data rate, and more recently IEEE802.11g that provides wireless connectivity with speeds up to 54 Mb/s. As a solution to their requirements for high data rates, all these standards use multicarrier communications, and in most cases, Orthogonal Frequency Division Multiplexing (OFDM) is used. In a single carrier communication system, the symbol period must be much greater than the delay time in order to avoid Inter-Symbol Interference (ISI). Since data rate is inversely proportional to symbol period, having long symbol periods means low data rate and communication inefficiency. A multicarrier system, such as Frequency Division Multiplexing (FDM), divides the total available bandwidth in the spectrum into subbands for multiple carriers to transmit in parallel. An overall high data rate can be achieved by placing carriers closely in the spectrum. However, Inter-Carrier Interference (ICI) will occur due to lack of spacing to separate the carriers. To avoid ICI, guard bands will need to be placed in between any adjacent carriers, which results in lowered data rate. OFDM is a multicarrier digital communication scheme to solve both issues. It combines a large number of low data rate carriers to construct a composite high data rate 1

communication system. Orthogonality gives the carriers a valid reason to be closely spaced, even overlapped, without ICI. Low data rate of each carrier implies long symbol periods, which greatly diminishes ISI. In addition, this was achieved with reduced complexity using an efficient Fast Fourier transform (FFT) based implementation. In OFDM a square root raised cosine pulse shape is typically used for the data. But the Use of a rectangular window in FFT causes creation of sidelobes which leads to higher interference and lower performance levels. Orthonormal wavelets are a recent advance in the representation of finite energy signals in (R). This contrasts the traditional Fourier expansion which can only represent signals in (0, 2) with complex exponentials as basis functions. The basis functions in the wavelet expansion are obtained by scaling and translating a single function called the wavelet function. The wavelet function is necessarily bandpass. The scaling feature represents the signal of interest as a superposition of long-duration signals with narrow bandwidth and short-duration signals with wide bandwidth. This also contrasts a Fourier-based approach which is restricted by the uncertainty principle to achieving resolution in time or frequency but not both domains. The development of Multi Resolution Approximations (MRAs), a recursive algorithm surfaced for computing the series expansion coefficients has provided a significant boost to the wavelet analysis as an alternative to fourier transform for OFDM systems. MRA leads to a generic filter bank which performs the filtering of the expansion coefficients of an approximation at a particular resolution with digital lowpass and highpass filters and decimating their outputs to produce the coefficients of the approximation and detail signals, respectively, at a coarser resolution. If the recursion is continued by operating on the decimated lowpass output, the algorithm came to be known as the Discrete Wavelet Transform (DWT). The inverse was readily obtained using expanders and related digital lowpass and highpass filters. This made wavelet processing practical and produced a fundamental connection to a major branch of signal processing called multirate systems. The DWT is a nonuniform analysis filter bank and the Inverse Discrete Fourier Transform (IDWT) is the corresponding nonuniform synthesis filter bank. Point of

interest in the theory of filter banks is the notion of perfect reconstruction. That is, if a digital signal is applied to an analysis bank which creates aliasing, amplitude and phase distortions, there exists a synthesis bank which will cancel these distortions and produce the original signal to within possibly a fixed integer delay. Research in filter banks has also produced M-channel uniform banks where the bandwidth in each channel is equal. Since some applications require the use of uniform banks, wavelet theory has been extended to include these conditions. The wavelets associated with these banks are called M-band wavelets and the modulation using wavelet packets as basis is called Wavelet Packet Modulation (WPM). Wavelet packets are sometimes referred to as multiplicity-M wavelets. In contrast to the single wavelet corresponding to the scaling function in the dyadic case, for M-band wavelets there are M-l wavelet functions. 1.2 Literature Review Alan R. Lindsey and Jeffery C. Dill [14] : Present a multirate wavelet based modulation format by elaborating the recently introduced multidimensional signaling technique (OFDM) which can utilize existing channels designed for conventional Quadrature Amplitude Modulation(QAM). Discusses customizable wavelet packet basis functions being employed as novel pulse shapes upon which independent QAM data at lower rates are placed and performs Time-Frequency comparison of modulation methods in a tone/impulse interference environment using Tiling Diagrams. The advantages include dimensionality in both time and frequency for flexible channel exploitation and an efficient all-digital filter bank implementation. Matthieu Gautier, Marylin Arndt, Joel Lienard [20] : Introduce a wavelet packet based multicarrier modulation scheme and compare it to the useful OFDM modulation in a wireless environment with narrowband interferences and multipath channel interferences. Simulations show that the use of wavelet is more robust to narrowband interferences than the OFDM modulation. Compared to [14], the major improvement of this paper is in the use of complex wavelet to reduce time and frequency dispersive

channel interferences. In a NBI environment, WPM gives satisfactory results in Bit Error Rate compared with OFDM modulation. C. Van Bouwel, J. Potemans, S . Schepers, B. Nauwelaers, A. Van de Capelle [15] : Studies the implementation of wavelet packets, and the effects this implementation has on the requirements imposed in the design of useable wavelets. It shows that the restrictions imposed by the perfect reconstruction requirement necessitate the use of biorthogonal wavelets. This however influences the performance. When the performance under narrowband interference is investigated, the results show that the advantage of higher sidelobe suppression is present but the frequency behavior of the wavelets is however very complicated, and makes optimization using the time-frequency tiling a rather unrealistic task. D. Karamehmedovic, M.K. Lakshmanan, H. Nikookar [13] : Establishes Wavelet Packet based Multi-Carrier Modulation (WPMCM) as an efficient multicarrier modulation technique alternative to the well-established OFDM. It shows that WPMCM has strong advantage of being a generic transmission scheme whose actual characteristics can be widely customized to fulfill several requirements and constraints of an advanced communication systems. This paper studies the performance of WPMCM transceivers in the presence of analogue radio frequency front-end imperfections. This paper analyzes the impact of interference in WPMCM transmission caused by phase noise. The sensitivity of WPMCM transceivers to these errors using standard wavelets is evaluated through simulation studies and their performances are compared and contrasted with OFDM.. Phase noise leads to the loss of orthogonality and consequently subcarriers begin to interfere with each other. In OFDM the performance degradation due to phase noise is limited to the interference among the subcarriers within one OFDM symbol (ICI), while in WPMCM subcarriers from multiple symbols interfere with each other (causing ICI + ISI). Mahesh Kumar Gupta, Sarika Shrivastava, A.S. Raghuvanshi and S. Tiwari [6]: In this paper the performance of Least Square (LS) and Linear Minimum Mean Square

Error (LMMSE) channel estimation technique for Wavelet based OFDM system was compared. The channel estimation based on block type pilot arrangement is performed by sending pilots at every sub-channel and using this estimation for a specific number of following symbols. In this paper wavelet transform is used instead of FFT, which have more bandwidth efficiency, less prone to Doppler shift and also complexity of order N as opposed to the FFTs N. In this paper use of wavelet transform in the place of FFT for the OFDM system elevates the bandwidth efficiency equal to or more than 20% due to no need of cyclic prefix. On considering Rayleigh Fading channel for simulation it was found that LMMSE estimator has 15-20 dB better performance than LS estimator and symlet wavelet has better performance. Khaizuran Abdullah and Zahir M. Hussain [23]: Present a comparative study on DWT based OFDM (DWT-OFDM) and FFT based OFDM (FFT-OFDM) systems. This paper proves that a DWT has to satisfy the orthonormal bases and the perfect reconstruction properties to be considered as wavelet transform. Different wavelet families are used and compared with the conventional FFT-OFDM system. It is found that the DWT-OFDM platform has less mean amplitude for transmitting the signal as compared to conventional FFT-OFDM system. Results show that DWT-OFDM is superior as compared to FFTOFDM with regards to the Bit Error Rate (BER) performance.

1.3 Motivation and Background The OFDM implemented by using IFFTs and FFTs have some problems. The OFDM suffers from ISI This is usually taken care of by adding a cyclic prefix equal to one fourth of the channel length but this may not always be possible. This occurs due to loss of orthogonality due to channel effects. Less sidelobe suppression- As Fourier transform uses a rectangular window, there is significant amount of inter channel interference.

 Carrier Frequency Offset- The offset between the carrier frequency and the frequency of the local oscillator also causes a large bit error rate.

Wavelet theory provides new ways to see and represent a signal. They are alternatives to Fourier transform. Wavelet transforms are much more local, an event stays connected to the time it occurs as opposed to transforming a pure time description into a pure frequency description. The method of Wavelet Transform finds a good compromise, a time frequency description. Wavelet transforms have been considered as alternative platforms for replacing IFFT and FFT [14]. By using the transform, the spectral containment of the channels is better and there is no need to use a Cyclic Prefix [20].

1.4 Dissertation Objective In this research, an orthonormal wavelet basis functions is proposed to be used as the orthogonal signals on which QAM/QPSK data symbols are to be placed. Specifically, the purpose of this thesis is the development of a general, unified approach to orthogonally multiplexed communications using wavelet bases and the implementation of these systems using perfect reconstruction digital filter banks. Later in the dissertation channel estimation techniques, namely LS and LMMSE have been implemented over wavelet based OFDM and an improvement in the performance is shown by simulation. Objectives of this dissertation are: To analyze the performance of DWT-OFDM as a new method to achieve a nonuniform partitioning of the data bandwidth in orthogonally multiplexed communications over Rayleigh and Nakagami channels. To analyze the performance of WPM-OFDM as a general approach to orthogonally multiplexed communications requiring a uniform partitioning of the data bandwidth in orthogonally multiplexed communications over Rayleigh and Nakagami channels. To implement and evaluate performance of channel estimation algorithms in wavelet based OFDM.

1.5 Thesis Organization This dissertation is organized as follows: Chapter 2 includes an introduction to the basic concepts of Wavelet OFDM, channels. Chapter 3 discusses simulation parameters, setup and the techniques. Modified simulation models for DWT-OFDM, WPM-OFDM and Channel estimation are given and explained. Chapter 4 presents the simulation results and discusses their implications. Chapter 5 concludes the dissertation with conclusion and future scope of the work. 1.6 Conclusion This chapter is an introduction to the domain of my dissertation which searches for alternatives to conventional OFDM to meet the future demands of high data rates. This chapter gives a literature review of the research that has been done in this field and lists the research articles that have been studied in the process of this dissertation. This chapter gives the background of my dissertation and lists the points which have motivated this dissertation. Dissertation objectives have been mentioned and an outline of the dissertation is also presented. discusses DWT and WPM based OFDM systems and presents a brief review of multipath fading

CHAPTER 2 INTRODUCTION TO WAVELET BASED OFDM

2.1 Introduction

Wavelet based OFDM has emerged as a major research area in digital communications as it improves the bandwidth efficiency and mitigates the problems of ICI and ISI. This chapter begins with a brief introduction to conventional OFDM. Basic theory related to wavelet OFDM like MRA and digital filter bank implementations are dealt later. DWT and WPM based OFDM systems are discussed separately. A brief review of multipath fading channels used in simulation is also presented.

2.2 Introduction to OFDM Various multicarrier modulation techniques have evolved in order to meet the demand of high data rates, few notable among them being Code Division Multiple Access (CDMA) and OFDM. OFDM is a frequency division multiplexing (FDM) scheme utilized as a digital multi carrier modulation method. A large number of closely spaced orthogonal sub carriers are used to carry data as shown in figure 2.1. The data is divided into several parallel streams of channels, one for each sub carriers. Each sub carrier is modulated with a conventional modulation scheme (such as QPSK or QAM) at a low symbol rate, maintaining total data rates similar to the conventional single carrier modulation schemes

in the same bandwidth. Advantages of OFDM are delayspread tolerance and spectrum efficiency that encourage their use in broadband communications.

Principle of OFDM

In older multichannel systems using FDM, the total available bandwidth is divided into

N nonoverlapping frequency subchannels. Each subchannel is modulated with a separate symbol stream and the N subchannels are frequency multiplexed, this leads to an inefficient use of spectrum. To overcome the problem of bandwidth wastage, we can instead use N overlapping (but orthogonal) sub carriers, each carrying a baud rate of 1/T

and spaced 1/T apart. Because of the frequency spacing selected, the sub carriers are all mathematically orthogonal to each other as shown in figure 2.2. This permits the proper demodulation of the symbol streams without the requirement of non overlapping spectra.

Another critreia of specifying the subcarrier orthogonality condition is to require that

each subcarrier have exactly integer number of cycles in the interval T. The idea in OFDM is to define a symbol sequence in the frequency domain, transmit it in the time domain, and map the received samples back into the frequency domain.

Fig. 2.1 Comparison of bandwidth between FDM and OFDM [3]

In high speed data transfer, Quality of service is an important criterion. OFDM is a promising candidate that eliminates a need for the complex equalizers. It is an important feature of the OFDM system design that the bandwidth occupied by each subcarrier is less than the coherence bandwidth of the fading channel. The frequency response over each individual subband is relatively flat. Orthogonal signals can be separated at the receiver by correlation techniques; hence, inter-symbol interference among channels can be eliminated. Orthogonality can be achieved by carefully selecting carrier spacing, such as letting the carrier spacing be equal to the reciprocal of the useful symbol period.

Fig. 2.2 Spectrum of OFDM [3]

10

OFDM System Model As shown in figure 2.3, the method for generating an ODFM symbol is as follows. First, the N input complex symbols are padded with zeros to get Ns symbols that are used to calculate the IFFT. The output of the IFFT is the basic OFDM symbol.

Based on the delay spread of the multi path channel, a specific guardtime must be chosen ( ). Number of samples corresponding to this guard time must be taken from the beginning of the OFDM symbol and appended at the end of the symbol. Likewise, the same number of samples must be taken from the end of the OFDM symbol and must be inserted at the beginning. The OFDM symbol must be multiplied with the raised cosine window to remove

the power of the outofband subcarriers.

The windowed OFDM symbol is then added to the output of the previous OFDM symbol with a delay of , so that there is an overlap region of r symbol. The model of receiver is shown in figure 2.4, it is just the reverse of the process at the transmitter. between each

11

Fig. 2.3 OFDM Transmitter [5]

Fig. 2.4 OFDM Receiver [5]

Drawback of OFDM PAPR OFDM on the other hand suffers from noise such as amplitude with a very large dynamic range; therefore, it requires RF power amplifiers with a high peak to average ratio. FEC reduce PAPR as compared to uncoded OFDM. Frequency offset It is also more sensitive to carrier frequency offset and drift than single carrier systems are due to leakage of the FFT. Lesser Sidelobe suppression Use of rectangular window in conventional OFDM leads to presence of sidelobes in frequency domain which causes inter channel interference Inefficient Data transmission The problem of ISI in OFDM symbols, is solved by the cyclic prefix insertion in the symbol frame but this technique causes overhead and makes OFDM inefficient. Synchronization

12

In OFDM system tight synchronization is needed. Often pilot tones are served in the subcarrier space. These are used to lock on phase and to equalization of channel.

2.3 Introduction to Wavelet based OFDM The purpose of this section is to provide the basic knowledge of the concepts that have made possible the implementation of orthogonally multiplexed communication using orthonormal wavelet basis functions as the orthogonal signals on which the QPSK/QAM sequences are placed. This development includes the review of basic concepts of wavelet transforms and their characteristics followed by derivation of the waveform, power spectral density, tiling diagram, bandwidth efficiency and performance in additive white Gaussian noise. As will be seen, dyadic wavelets (DWT) provide a non-uniform decomposition of the time-frequency plane whereas wavelet packets perform uniform decomposition based on M band wavelet basis functions. This allows for greater effectiveness against channel impairments [14]. Efficient digital filter banks exist for the implementation of these systems.

2.3.1 Wavelets, Wavelet Filters and Wavelet Transforms This subsection is about wavelets and filter banks. These are new ways to see and represent a signal. They are alternatives to Fourier transform. Wavelet transforms are much more local, an event stays connected to the time it occurs as opposed to transforming a pure time description into a pure frequency description. The method of Wavelet Transform finds a good compromise, a time frequency description [10]. The most well known tool for signal analysis is fourier analysis, which breaks down a signal into constituent sinusoids of different frequencies. To overcome the deficiencies of fourier analysis for non-stationary signals, STFT was developed. In STFT fourier transform is adapted to analyze only a small section of the signal at a time-a technique called windowing the signal Wavelet analysis represents the next logical step: a windowing technique with variable- sized regions. Wavelet analysis allows the use of

13

long time intervals where we want more precise low frequency information, and shorter regions where we want high frequency information [16]. Wavelets A wavelet is a waveform of effectively limited duration that has an average value of zero. Compare wavelets with sine waves, which are the basis of Fourier analysis. As seen in fig. 2.5 Sinusoids do not have limited duration they extend from minus to plus infinity. And where sinusoids are smooth and predictable, wavelets tend to be irregular and asymmetric.

Fig. 2.5 Comparison between a sinusoidal wave and a wavelet [11]

Fourier analysis consists of breaking up a signal into sine waves of various frequencies. Similarly, wavelet analysis is the breaking up of a signal into shifted and scaled versions of the original (or mother) wavelet.

Properties of Wavelets
A wavelet is a real or complex valued continuous time function properties [18] 1. The function integrates to zero [18]: with the following

(2.1)
2. It is square integrable, or equivalently has finite energy [18]:

14

(2.2)
A function is a wavelet or mother wavelet if it satisfies above two properties and also the admissibility condition [18]

0<

<

(2.3)

The admissibility is sufficient to formulate the simple inverse wavelet transform. Wavelets are basis function in continuous time. A basis is a set of linearly independent functions that can be used to produce all admissible functions f(t) [18]. f(t) = combination of basis functions =

(2.4)

The special feature of wavelet basis is that all functions are constructed from a single mother wavelet . Normally it starts at time t=0 and ends at time t=N. The shifted wavelets start at time t = k and end at time t = k + N . The rescaled wavelets start at time t=0 and end at time t = N/. Their graphs are compressed by the factor . Whereas the graphs of are translated (shifted to the right) by k [11]. Compressed: = Shifted: =

A typical wavelet is compressed j times and shifted k times. The formula is [10]

(2.5)
The remarkable property that is achieved by many wavelets is orthogonality. The wavelets are orthogonal when their inner products are zero [10].

=0

(2.6)

15

In this case the wavelets form an orthogonal basis for the space of admissible functions. This basis corresponds to a set of axes that meet at angles as most good axes do. Orthogonality leads to a simple formula for each coefficients in the expansion for f(t). Multiply the expansion displayed in equation (2.4) by and integrate.

(2.7)

All other terms disappear because of orthogonality. eqn. (2.6) eliminates all integrals of times , except one term that has j=J and k=K. That term produces (. Then is the ratio of the two integrals in eqn. (2.7). As we describe the connection between filter banks and wavelets, you will see that it is the high pass filter that leads . The low pass filter leads to a scaling function . Wavelet Transforms Wavelet Transforms can be classified into following classes on the basis of the function being transformed and the type of filter bank being used for synthesis or analysis of the signal. 1. Continuous Wavelet Transform. 2. Discrete Wavelet Transform 3. Wavelet Packet Transform. 4. Dual Tree Wavelet Packet Transform.

Multi-Resolution Approximations The MRA is a framework for constructing orthonormal wavelet bases. For dyadic scaling and integer translations a Multi Resolution Approximation is a sequence of embedded subspaces of , denoted .Elements of are approximations at a given resolution to signals in . Fine Resolution implies wide bandwidth and coarse resolution implies narrow bandwidth. Details lost when changing from an approximation of fine resolution to a coarse resolution forms orthogonal complement of the coarse resolution subspace.

16

Essential Results on Orthonormal Dyadic Wavelets Definition 1 An MRA is a sequence of closed subspaces of , denoted satisfying [9] 1. 2. f(x) then f(2x) Theorem 1 Let ( be an MRA, then is an orthonormal basis function for , is called the scaling function. Since , the scaling function can be written as a series expansion in the basis functions for , namely [9],

(2.8)
From (2.14), is seen to be the inner product of with . This sequence plays an important role in the discrete wavelet transform (DWT). The Fourier transform of (2.14) is [9]

(2.9)
Where H is the discrete-time Fourier transform of and must satisfy , This constraint on H is called the admissibility condition. Definition 2 Let be the orthogonal complement of in that is and Theorem 2 Let be an MRA with scaling function , then , is an orthonormal basis for

where satisfies [9]

(2.10)

With = Moreover, is an orthonormal basis for . is called the wavelet function.

17

From Definition 2 and Theorems 1 and 2, we see that the scaling and wavelet functions are orthogonal at different coarse scales. Additionally, the wavelet function is orthogonal to itself at different scales by Theorem 2. But, the scaling function is not orthogonal to itself at different scales due to the embedded vector space property of Definition 1. The expression (2.16) can be viewed as a series expansion in the basis functions for since DWT. The Fourier transform of (2.10) is [9] Thus, is the inner product of with, .). This sequence also plays an important role in the

(2.11)
where G is the discrete-time Fourier transform of . Theorem 3 Let andbe the series coefficients of the scaling and wavelet functions in an MRA then their transforms satisfy [9]

(2.12)

The relation in (2.12) is commonly referred to as the power complementary property. Theorem 4 Let and be the series coefficients of the scaling and wavelet functions in an MRA. Let f then [9]

(2.13)
Moreover

(2.14) (2.15)
Equation (2.13) defines one stage of the inverse DWT while equations (2.14) and (2.15) define one stage of the DWT. The signal processing operations associated with these equations are illustrated in Figures 2.6 and 2.7, respectively.

18

Fig. 2.6 One Stage of Reconstruction [9]

Fig. 2.7 One Stage of Decomposition [9]

2.3.2 System Architecture for DWT-OFDM The purpose of this subsection is to develop the fundamental theory of orthogonally multiplexed communication using orthonormal dyadic wavelet basis functions as the orthogonal signals on which the QPSK/QAM sequences are placed. This development includes the derivation of the waveform, power spectral density and tiling diagram. Efficient digital filter banks exist for the implementation of these systems which are introduced in this chapter.

19

Waveform Development for DWT based OFDM In general, a quadrature amplitude modulated (QAM) digital signal is defined as [1]

(2.16)
Where: is the pulse shape E is the average symbol energy T is the symbol duration are complex-valued QAM symbols The data symbols are assumed to be identically distributed, zero mean and unit variance. Let be a scaling function in an MRA, then the signal defined by (2.16) is free of inter symbol interference (lSI) by Theorem 2 and the orthonormality of [14]. Hence, MRAs provide a broad new class of pulse shapes for use in conventional data communications. Assuming the data symbols are independent, the power spectral density of this modulation format is [14]

(2.17)
By Definitions 2.1 and 2.3, the vector space can be decomposed into the direct sum [9]

(2.18)
Where J is a positive integer. It follows that can be expanded in the basis functions for the subspaces on the right hand side of (2.18). We now rewrite (2.16) as the multi dimensional signal [14]

(2.19)

Where: are complex-valued QAM symbols for n=0,..,J-l

20

Since data has been placed at different scales, (2.19) can also be referred to as multiscale modulation (MSM). Orthonormality of the individual basis functions prevents lSI. Their mutual orthogonality prevents interference across scales. Power Spectral Density To derive the power spectral density of DWT, we need the following lemma. Lemma- Let and be scaling and wavelet functions in an MRA, then their Fourier transforms satisfy [9]

(2.20)
Where f Proof: Using (2.9) and (2.11), we have

(2.21)
Factoring out the term yields

Applying Theorem 3 to (2.21) prove the lemma. Now, assuming the symbols at different scales are independent and identically distributed with zero mean and unit variance, the power spectral density for MSM is given by [14]

(2.22)
To simplify (2.28), apply Lemma to the term involving the scaling function and the n=1 term of the series. This produces a twice rate term involving the scaling function. We can repeat the application of Lemma with this new term and the n=2 term of the series. Since the series is finite, we can continually repeat application of Lemma yielding [14] 21

(2.23)
which is identical to the power spectral density of QAM as given by (2.23). Consequently, this waveform can be used on existing channels which have been designed around conventional QAM signals [9]. This allows interoperability with systems where redesign would not be cost effective as in the case of existing satellite systems. We also note that (2.23) is only dependent on the scaling function and not on the wavelet function. System Architecture of DWT based OFDM As noted, the optimum receiver in AWGN for DWT consists of a bank of filters matched to each pulse shape in eqn. 2.19. The transceiver processing is illustrated in Figure 2.8. Throughout this dissertation, we will assume carrier and symbol synchronization which allows us to consider low pass equivalent signals [12]. At the transmitter, a data source is demultiplexed into J symbol streams operating at dyadic sub-rates of the source. An IDWT is computed generating a sequence at the source rate. This sequence is transmitted across the AWGN channel, The received signal is match filtered and sampled. Taking the DWT of this sequence produces soft decisions of the transmitted symbols which are forwarded to the appropriate data sink possibly for further processing.

Fig. 2.8 System Architecture for DWT OFDM [15]

To see in further detail how the transforms are able to replace much of the analog processing, begin with the transmitter and note because of (2.23), (2.16) can be written as

22

(2.24)
Where is a complex sequence determined by the data symbols. Further note, this

sequence will not in general be elements of a traditional QAM constellation. To obtain the relationship between and the data symbols we apply Theorem 2 in a recursive manner. The signal processing is illustrated in Figure 2.9 and corresponds to a nonuniform synthesis filter bank.

Fig. 2.9 Detailed Illustration of the Transmitter IDWT [23]

Figure 2.10 Pulse Generator [24]

The pulse generator can be modeled as shown in figure 2.10 and its output can be written as [24]

(2.25)
Where is the Dirac delta function. The pulse generator output can now be written as [24]

(2.26)

23

Note that the filter is scaled by

to ensure orthonormality. Making the variable

substitution t = uT, (2.29) becomes [24]

(2.27)
Applying the distribution property of Dirac delta functions, (2.27) reduces to (2.25). At the receiver, the output of the match filter is given by [24]

(2.28)
Where is the filtered AWGN. Sampling (2.28) at kT yields [24]

(2.29)
Where is the sampled noise sequence. Making the variable substitution and using the orthonormal property of , (2.29) reduces to [24]

(2.30)
We can now recursively apply equations of Theorem 4 to the sequence producing estimates of the data symbol [16]. The signal processing is illustrated in Figure 2.9 and corresponds to a non-uniform analysis filter bank. In the absence of noise, we will recover the transmitted data symbols precisely. It is interesting to note from Figures 2.9 and 2.11 that the digital filters in the DWT are matched to the digital filters in the IDWT.

24

Fig. 2.11 Detailed Illustration of the Receiver DWT [23]

2.3.3 System Architecture for WPM-OFDM Since some applications require a uniform partitioning of the bandwidth, for example the mitigation of a stationary narrowband interference [19], we develop in this subsection a new modulation format which achieves a uniform decomposition based on M-band wavelet basis functions. Thus, this development includes the derivation of the waveform and its power spectral density. In addition, an efficient digital filter bank implementation will be introduced with a detailed study so that uniform and non-uniform banks can be treated in a unified manner. In some ways, M-band wavelets are a natural generalization of dyadic wavelets.

Waveform Development for WPM based OFDM In general, quadrature amplitude modulated (QAM) digital signal is defined as [1] ) Where: is the pulse shape E is the average symbol energy T is the symbol duration are complex-valued QAM symbols Let be an M-band scaling function for some M>2. By Definition 1, the vector space can be decomposed into the direct sum [9] (2.32) It follows that can be expanded in the basis functions for the subspaces on the right hand side of (2.35). We now rewrite (2.34) as the multidimensional signal [14] (2.33) 25 (2.31)

Where: are complex-valued QAM symbols for n=0,..,M-l

The communication signal defined by (2.33) will be referred to as M-band Wavelet Modulation (MWM). By Theorems 1 and 2, the waveform defined in (2.36) is free of lSI. Also using Definition 1, the waveform is free of cross channel interference. In contrast to MSM, the symbol period on each pulse in (2.33) is equal and has a value of MT. Implementation of WPM based OFDM As noted, the optimum receiver in AWGN for WPM is a bank of filters matched to each pulse shape in (2.33). Also, as was seen with DWT, a much more efficient realization is obtained with digital filter banks. The transceiver processing is illustrated in Figure 2.12. At the transmitter, a data source is demultiplexed into M symbol streams operating at one the source rate. A uniform synthesis filter bank operates on these subsequences producing a sequence at the source rate. This sequence is transmitted across the channel using a pulse shape defined by the M-band scaling function. The received signal is match filtered and sampled. The sampled sequence is then operated on by a uniform analysis filter bank producing soft decisions of the transmitted symbols.

Figure 2.12 Transceiver Processing For M-band Wavelet Packet Modulation [14]

The implementation of the uniform filter banks of Figure 2.12 can be achieved in a number of ways. We will adopt in this dissertation the tree structure. The advantages of using trees is in its ability to implement uniform filter banks, discrete wavelet transforms, transforms associated with wavelet packets and time-varying tiling diagrams. The tree structured uniform synthesis bank is shown in Figure 2.13 while the analysis filter bank is shown in Figure 2.14.

26

Figure 2.13 Detailed Illustration of the Transmitter Uniform Synthesis Filter Bank [15]

A particularly useful feature of the M-band scaling function is that we can obtain a full digital implementation while maintaining the theoretical power spectral density.

Figure 2.14 Detailed Illustration of the Receiver Uniform Analysis Filter Bank [15]

2.4 Fading Channels Radio-wave propagation through wireless channels is a complicated phenomenon characterized by various effects, such as multipath and shadowing. Fading is caused by interference between two or more versions of the transmitted signal which arrive at the receiver at slightly different times [4]. These waves combine at the receiver antenna to give a resultant signal which can vary widely in amplitude and phase, depending on the distribution of intensity and relative propagation time of the waves and bandwidth of the 27

transmitted signal. Over the years several studies and measurements have been undertaken in different locations for mobile channels. Types of Fading In mobile communication the signal power drops off at the receiver due to I. II. III. Mean path loss Macroscopic fading, also called shadowing Microscopic fading, also referred to as small scale fading.

Mean path loss arises from inverse square law of power loss and depends on the distance of the traveling wave. Macroscopic fading or shadowing results from a blocking effect by obstacles such as buildings, large trees and mountains as seen in figure 2.15. Microscopic or small scale fading arises due to the multipath propagation where the received signal consists of an infinite sum of attenuated, delayed and phase-shifted replicas, caused due to the scattering of the transmitted signal by obstructions. Multipath propagation and the mobility of the receiver result in the spreading of the signal in different dimensions. These are mainly delay spread due to the presence of resolvable multipath components in time and Doppler spread in frequency due to the mobility of the terminal.

Fig. 2.15 Illustration of a multipath channel [5]

Flat and Frequency selective fading

28

Small-scale fading can be further classified as flat or frequency selective and slow or fast. A received signal is said to undergo flat fading, if the mobile radio channel has a constant gain and a linear phase response over a bandwidth larger than the bandwidth of the transmitted signal. Under these conditions, the received signal has amplitude fluctuations due to the variations in the channel gain over time caused by multipath. However, the spectral characteristics of the transmitted signal remain intact at the receiver. On the other hand, if the mobile radio channel has a constant gain and linear phase response over a bandwidth smaller than that of the transmitted signal, the transmitted signal is said to undergo frequency selective fading. In this case, the received signal is distorted and dispersed, because it consists of multiple versions of the transmitted signal, attenuated and delayed in time. This leads to time dispersion of the transmitted symbols within the channel arising from these different time delays resulting in Inter-Symbol Interference (ISI). When there is relative motion between the transmitter and the receiver, Doppler spread is introduced in the received signal spectrum, causing frequency dispersion. Statistical Modeling of Flat Fading Channels MATLAB provides a simple and easy way to demonstrate fading taking place in wireless systems. The RF (radio frequency) signals with appropriate statistical properties can readily be simulated. Statistical testing can subsequently be used to establish the validity of the fading models frequently used in wireless systems. The different fading models and MATLAB based simulation approaches will now be described. 2.4.1 Rayleigh Channel The mobile antenna, instead of receiving the signal over one line-of-sight path, receives a number of reflected and scattered waves, as shown in Figure 2.9. Because of the varying path lengths, the phases are random, and consequently, the instantaneous received power becomes a random variable. In the case of an unmodulated carrier, the transmitted signal at frequency c reaches the receiver via a number of paths, the path having an amplitude , and a phase i. If we assume that there is no direct path or line-of sight (LOS) component, the received signal s(t) can be expressed as [2]

29

(2.34) where N is the number of paths. The phase i depends on the varying path lengths, changing by 2 when the path length changes by a wavelength. Therefore, the phases are uniformly distributed over [0, 2]. When there is relative motion between the transmitter and the receiver, equation (2.34) must be modified to include the effects of motion induced frequency and phase shifts. Let the ith reflected wave with amplitude and phase i arrives at the receiver from an angle relative to the direction of motion of the antenna. The Doppler shift of this wave is given by [2] (2.35) Where v is the velocity of the mobile, c is the speed of light (3xm/s), and the s are uniformly distributed over [0, 2]. The received signal s(t) can now be written as

Expressing the signal in in-phase and quadrature form, eqn. (2.37) can be written as [2] (2.36) Where the in phase and quadrature components are respectively given as [2] (2.37) (2.38) The envelope R is given by When N is large, the in-phase and quadrature components will be Gaussian. The Probability Density Function (PDF) of the received signal envelope, f(r), can be shown to be Rayleigh given by [2]

30

,r 2.4.2 Nakagami-m distribution

(2.39)

It is possible to describe both Rayleigh and Rician fading with the help of a single model using the Nakagami distribution. The fading model for the received signal envelope, proposed by Nakagami, has the PDF given by [2] (2.40) Where (m) is the Gamma function, and m is the shape factor (with the constraint that m ) given by [2] (2.41) The parameter controls the spread of the distribution and is given by [2] (2.42) the corresponding cumulative distribution function can be expressed as [2] (2.42) Where P(.) is the incomplete Gamma function. In the special case m = 1, Nakagami reduces to Rayleigh distribution. For m > 1, the fluctuations of the signal strength reduce compared to Rayleigh fading, and Nakagami tends to Rician. 2.5 Conclusion This chapter has given a review of the developments that have taken place in the use of wavelets for communication, especially OFDM. This chapter also presents the digital filter bank implementation of DWT and WPM OFDM systems. Filter bank implementation has been adopted for simulation purposes but with a few changes that

31

would be advent in the later part of this dissertation. As we can see from this chapter, wavelets present an efficient solution to the problems that are common in conventional OFDM.

CHAPTER 3 SIMULATION SETUP

3.1 Introduction In this dissertation performance analysis of wavelet OFDM over Rayleigh and Nakagami fading channels is performed with comparison made on the basis of the BER performance of the models that have been simulated using MATLAB. It is assumed that the system is fully synchronized. BER analysis of Channel estimation techniques; namely LS and LMMSE over the Rayleigh fading channel has also been performed. Monte Carlo technique of BER calculation for DWT-OFDM and Semi-Analytic technique for WPMOFDM are followed for simulation.

32

3.2 Simulation Techniques Monte Carlo Simulation Monte Carlo simulations are based on games of chance. Monte Carlo estimation is based on the relative frequency interpretation of probability [20]. We count the number of occurrences, NA, corresponding to an event, A, of interest. The probability of the event A is approximated by the relative frequency of the event, which is defined by replicating the random experiment an infinite number of times. This gives (3.1) In using Monte Carlo simulation to estimate a performance parameter of a communications system, such as the symbol error probability, unbiased and consistent estimates are clearly desirable. If an estimator is unbiased we know that, on the average, Monte Carlo simulation provides the correct result. In addition, if an estimator is to be useful it must have small variance so that, with high probability, the estimate lies in the neighborhood of the true value being estimated. Semi-Analytic Simulation Fortunately there are alternatives to the pure Monte Carlo method. One of the most powerful of these alternatives is the Semi-Analytic (SA) method, in which analytical and simulation techniques are used together in a way that yields very rapid estimation of the BER. The SA simulation method, like all rapid simulation techniques, allows analytical knowledge to be traded off against simulation run time. In applying semianalytic simulation, the combined effects of and are determined by an MC simulation, and the effects of noise, represented by , are treated analytically. The SA simulation technique is applicable to any system in which the probability density function of the noise component of Vk can be analytically determined. SA BER Estimation for QPSK The probability of the event A, defined in the relative frequency sense, is obtained by

33

We now consider a SA estimator for the symbol error probability PS in a QPSK system. Since a QPSK signal constellation has four signal points rather than two, and since the signal space has two dimensions rather than one, the SA estimator for QPSK is different from the estimator for PSK in that a dimension must be added for the quadrature channel. Since the simulation will account for the effects of ISI but not the effects of noise. The direct and quadrature components of are denoted and , respectively, where = Re( and = Im(. When noise is considered, by adding nx and ny to and , respectively, a correct

decision is made, conditioned on S1 transmitted, if ( + nx, + ny) D1. An error is made if ( + nx, + ny) D1. Keep in mind that since we are developing a SA estimator, the impact of noise is treated analytically and does not appear in Figure 3.1. Assume that is the kth transmitted symbol in a simulated sequence of N symbols. For each value of k, 1 k N, will be , , or . The bound on the conditional symbol error rate is: (3.2)

Fig. 3.1 SA BER estimation for QPSK [25]

34

The technique used here to develop a semianalytic estimator is easily extended to MPSK and QAM 3.3 System Parameters Table 3.1 lists the parameters that have been taken for the purpose of simulation
Table 3.1 Simulation Parameters

PARAMETERS N(Number of Channels) Size of FFT No. of decomposition levels Ns(number of bits per sample) Modulation Wavelet Channel

FFT-OFDM 30 64 NA 100 QAM/QPSK NA

DWT-OFDM 30 NA 6 100 QAM Sym, Haar, db

WPM-OFDM 30 NA 6 100 QPSK Db

AWGN + Rayleigh, AWGN + Rayleigh AWGN + Nakagami m Number of Pilot symbols(LS & 16 16 MMSE)

16

3.4 Modified Models of wavelet OFDM In this section the simulation models used in this dissertation have been introduced and discussed. System models of DWT OFDM and WPM OFDM have been given, internal diagram of IDWT/DWT block used for MATLAB simulation is given separately. Later in the section simulation setup for Channel Estimation techniques for DWT-OFDM is given. The CE is performed using DFT based LS and LMMSE techniques. 3.4.1 Modified Model of DWT-OFDM

35

From Figure 3.2, it is obvious that the transmitter first uses a QAM digital modulator which maps the serial bits d into the OFDM symbols , within N parallel data stream (i) where 0 i N 1.

Fig. 3.2 Modified Model for Simulation of DWT-OFDM

The main task of the transmitter is to perform the discrete wavelet modulation by constructing orthonormal wavelets. As shown in fig. 3.3 each is first converted to serial representation having a vector xx which will next be transposed into CA. This means that CA not only its imaginary part has inverting signs but also its form is changed to a parallel matrix. Then, the signal is up-sampled and filtered by the LPF coefficients or namely as approximated coefficients. These coefficients are also called scaling coefficients. Since our aim is to have low frequency signals, the modulated signals xx perform circular convolution with LPF filter whereas the HPF filter also perform the convolution with zeroes padding signals CD respectively. Note that the HPF filter contains detailed coefficients or wavelet coefficients. Different wavelet families have different filter length and values of approximated and detailed coefficients. Both of these filters have to satisfy orthonormal bases in order to operate as wavelet transform.

Fig. 3.3 Model for IDWT/DWT Blocks in DWT-OFDM

The number of CA and CD depends on the OFDM subcarriers N. Samples of this processing signals CA and CD that pass through this block model is shown in Figure 3.3. 3.4.2 Modified Model for WPM-OFDM The simplified block diagram of the WPM-based OFDM communication system is shown in Figure 3.4. The number of subcarriers is N . At the transmitter, the binary data stream b is encoded first and then it is QPSK modulated. After modulation frequency interleaving is performed, after that Serial to Parallel conversion is performed and fed to the terminal of a synthesis wavelet packet tree, also called the synthesis filter banks (SFB). Figure 3.4 shows a regular wavelet packet tree with depth J = 2

36

Fig. 3.4 Modified Model for simulation of WPM OFDM

When regular wavelet packet tree with depth J = is used, the corresponding SFB is also known as Inverse Wavelet Packet Transform (IWPT) and the analysis wavelet packet tree or Analysis Filter Banks (AFB) is called Wavelet Packet Transform (WPT) [14]. The output of the SFB is the transmitted signal with length L which can be computed recursively by equation 2.13. The transmitted signal passes through multipath fading channel where it gets multiplied by the fading channels PDF function and AWGN noise is added to it. At the receiver, the received signal r, consisting of the sum of the multipath Rayleigh fading signals and the additive white Gaussian noise, is decomposed by the AFB and the original data blocks can be recovered recursively by equations 2.14 and 2.15. 3.5 Simulation Model for Channel Estimation in DWT-OFDM The OFDM system based on pilot channel estimation is given in Figure 3.5. The input binary information is first grouped and mapped according to the modulation in QPSK modulator. After inserting pilots to all sub-carriers with a specific period data is send to inverse discrete wavelet transform (IDWT), which converts frequency domain signal X(k) into time domain signal x(n) and also provides orthogonality to sub-carriers. Then it sends to the Rayleigh fading channel h(n), [8] h(n) = (m) (n-(m)) (3.3)

Where is complex path gain and is path delay. At the receiver data is converted back to frequency domain X(k) from time domain x(n) by Discrete wavelet transform (DWT). After that pilot signal are extracted. Then channel is estimated He(k) via LS and LMMSE estimator. Then the transmitted data is estimated by [8]: , k = 0,1, N-1 Then the binary information data is obtained back after QPSK demodulation.
Fig. 3.5 Model of DWT-OFDM with Channel Estimation [8]

(3.4)

37

Figure 3.6 shows a block diagram of DFT-based channel estimation, given the LS or LMMSE channel estimation. Note that the maximum channel delay L must be known in advance.

Fig. 3.6 Modified Model of DFT based Channel Estimation for DWT-OFDM

3.6 Conclusion In this chapter simulation models developed for wavelet based OFDM are presented. Modified models for DWT-OFDM and WPM-OFDM are presented and illustrated. Channel estimation techniques for DWT-OFDM are also given and explained. These models have been developed for ideal conditions and there is no need for synchronization. Such digital implementation can bring Wavelet based OFDM at par with conventional OFDM but still a lot of research is required for developing wavelet based OFDM systems suited for practical implementation.

CHAPTER 4
38

SIMULATION RESULTS AND DISCUSSIONS

4.1 Introduction In this chapter, simulation results are presented which show the performance comparison of DWT-OFDM with FFT-OFDM and WPM-OFDM with FFT-OFDM in terms of BER in various wireless channels. The FFT-OFDM, DWT-OFDM and WPM-OFDM systems are developed, analyzed and simulated using MATLAB. Later in the chapter simulation result showing the performance of LS and MMSE channel estimation in wavelet based OFDM is presented, this analysis is also done in terms of BER in various wireless channels. Performances of systems have been compared with different wavelets families like symlet, Haar and daubechies. Wireless channels used in simulation are Rayleigh and Nakagami-m. 4.2 Performance Analysis of Wavelet based OFDM under Rayleigh Fading channel In this section the BER curves for wavelet based OFDM have been compared. Two types of wavelet transformation techniques, namely WPM and DWT have been adopted. WPM-OFDM, FFT-OFDM and DWT-OFDM in Rayleigh fading channel have been compared. SA simulation technique has been used for WPM-OFDM and Monte Carlo simulation technique has been used for DWT-OFDM and FFT-OFDM. Wavelets considered for DWT-OFDM are Haar, Symlet and db while for WPM-OFDM only db wavelet has been considered.

4.2.1 Performance Analysis of WPM-OFDM under Rayleigh Fading channel In this subsection BER performance of WPM-OFDM and FFT-OFDM has been compared, modulation format is QPSK and wavelet used is daubechies N. Simulation technique used is semianalytic and time and frequency synchronization is assumed.

39

B E R v s S N R c u r v e f o r W P M - O F D M u n d e r R a y le ig h F a d in g c h a n n e l
10
0

WPM FFT

BER

10

-1

10

-2

S N R (d B )

10

12

14

16

18

20

Fig. 4.1 BER vs SNR curve for WPM-OFDM and FFT-OFDM under Rayleigh Fading channel Table 4.1 BER vs SNR values for FFT-OFDM and WPM-OFDM system

SNR (dB) FFTOFDM WPMOFDM

2
0.392 2 0.333 2

4
0.3591

6
0.331 7 0.256 9

8
0.2945

10
0.251 2 0.168 9

12
0.2015

14
0.156 1 0.092 2

16
0.1141

18
0.077 1 0.042 8

20
0.0527

0.2960

0.2089

0.1275

0.0642

0.0282

Figure 4.1 show that WPM-OFDM with db 10 wavelet filter is superior then FFT-OFDM. Table 4.1 shows that WPM-OFDM with db 10 wavelet filter outperforms FFT by about 3 dB at 0.06 BER. 4.2.2 Performance Analysis of DWT-OFDM under Rayleigh Fading channel In this subsection BER performance of DWT-OFDM and FFT-OFDM are compared, the modulation format is QAM and the wavelets used are symlet and haar. MC simulation technique is adopted.

40

10

B E R v s S N R c u r v e f o r D W T - O F D M u n d e r R a y le ig h f a d in g c h a n n e l
Sym IF F T H aar

10

-1

10

-2

BER
10
-3

10

-4

10

-5

-5

10

15

20

25

S N R (d b )

Fig. 4.2 BER vs SNR curve for DWT-OFDM and FFT-OFDM under Rayleigh Fading channel

Table 4.2 BER vs SNR values for DWT and FFT-OFDM under Rayleigh channel SNR(db) FFT-OFDM DWT-OFDM (sym) DWT-OFDM (haar) 0 4.078e-01 9.690e-02 8.906e-02 5 3.450e-01 1.901e-02 8.672e-03 10 2.521e-01 7.668e-03 1.563e-05 15 1.334e-01 7.452e-03 6.724e-07 20 5.360e-02 7.452e-03 2.342e-08 25 2.162e-02 7.452e-03 1.894e-08

Fig. 4.2 show that DWT-OFDM with haar wavelet filter is superior then DWT-OFDM with symlet wavelet filter, Table 4.2 shows an SNR improvement of about 2 dB at 0.01 BER for symlet based DWT-OFDM over FFT-OFDM. Haar based DWT-OFDM outperforms FFT-OFDM by about 15 dB at 0.02 BER. The reason for very low BER in case of Haar wavelet is that Haar wavelet offers excellent time and spectral containment but unfortunately it is discontinuous hence it can not be used for practical purposes. 4.3 Performance Analysis of Wavelet OFDM under Nakagami-m Fading channel 41

In this section the BER curves for FFT-OFDM and DWT-OFDM in Nakagami-m fading channel have been compared. Parameter m has been taken 1.5. Wavelets considered for DWT-OFDM are Haar and Symlet. 4.3.1 Performance Analysis of DWT-OFDM under Nakagami-m Fading channel In this subsection BER performance of DWT-OFDM and FFT-OFDM are compared, the modulation format is QAM and the wavelets used are symlet and haar.
10
0

B E R v s S N R c u r v e f o r D W T - O F D M u n d e r N a k g a m i- m F a d in g c h a n n e l
SYM FFT HAAR

10

-1

10

-2

BER
10
-3

10

-4

10

-5

-5

10

15

20

25

S N R (d b )

Fig. 4.3 BER vs SNR curve for DWT-OFDM and FFT-OFDM under Nakagami-m Fading channel Table 4.3 BER vs SNR values for DWT and FFT-OFDM under Nakgami-m channel SNR(db) 0 5 10 15 20 FFT-OFDM DWT-OFDM (sym) DWT-OFDM (haar) 4.075e-01 8.837e-02 8.348e-02 3.437e-01 1.418e-02 6.281e-03 2.308e-01 7.452e-03 1.563e-05 1.082e-01 7.572e-03 7.914e-07 2.572e-02 7.596e-03 2.822e-08 25 4.040e-03 7.452e-03 2.097e-08

Figure 4.3 shows that DWT-OFDM with haar wavelet filter is superior then DWT-OFDM with symlet wavelet filter. Table 4.3 shows an SNR improvement of about 2 dB at 0.01 BER. Symlet based DWT-OFDM outperforms FFT by about 15 dB at 0.02 BER. 4.4 Performance Analysis of LS and LMMSE channel estimation for DWT-OFDM under Rayleigh Channel 42

In this section we compare the BER curves of LS and LMMSE channel estimation in DWT-OFDM with the BER curve of DWT-OFDM without any channel estimation. This channel estimation, based on block type pilot arrangement is performed by sending pilots at every sub-channel and using this estimation for a specific number of following symbols. We have considered multipath Rayleigh fading channel for the simulations. BER calculations are performed by Monte Carlo method.
10
0

B E R v s S N R c u r v e f o r L S c h a n n e l e s t im a t io n D W T b a s e d O F D M

LS CE s y m (N o C E )
10
-1

BER
10
-2

10

-3

-5

10

15

20

25

S N R (d B )

Fig. 4.4 BER vs SNR curve for LS channel estimation on DWT-OFDM Table 4.4 BER vs SNR values for LS channel estimation on DWT-OFDM

SNR (dB) No Channel Estimation LS Channel Estimation

0 8.531e02 1.101e01

5 1.378e-02 1.862e-02

10 7.501e-03 3.219e-03

15 7.562e-03 1.844e-03

20 7.562e03 1.375e03

25 7.562e03 2.344e03

43

10

B E R v s S N R C u r v e f o r L M M S E c h a n n e l e s t im a t io n o n D W T O F D M

MMSE s y m (N o C E )
10
-1

BER
10
-2

10

-3

-5

10

15

20

25

S N R (d B )

Fig. 4.5 BER vs SNR curve for LMMSE channel estimation in DWT-OFDM Table 4.5 BER vs SNR values for LMMSE channel estimation in DWT-OFDM

SNR (dB) Without Channel Estimation LS Channel Estimation

0 8.531e02 8.503e01

5 1.378e-02 1.097e-02

10 7.501e-03 2.531e-03

15 7.562e-03 4.281e-03

20 7.562e03 3.125e03

25 7.562e03 4.531e03

The comparison of figure 4.4 and 4.5 shows that the LMMSE estimator gives a performance improvement over the LS estimator, Table 4.4 and 4.5 show that LMMSE technique gives an SNR improvement of 16 db at BER, but the complexity of LMMSE estimator is more than LS estimator. All the taken wavelet family has nearly similar BER performance but Symlet has better performance because it gives fewer fluctuations BER value compared to others.

44

4.5 Conclusion In this chapter BER performance of DWT-OFDM and WPM-OFDM have been compared with conventional OFDM, performance of Channel estimation Algorithms has also been compared for DWT-OFDM. All above simulations have been performed assuming that there is no need of synchronization in the system. It is assumed that the filter weights are known at the receiver but this is not practical. Results show that the BER performance of wavelet OFDM is better than conventional OFDM. Though Haar wavelet shows best results but it cannot be used practically as it is highly discontinuous. In Channel Estimation, MMSE technique outperformed LS technique but it has higher complexity.

45

CHAPTER 5 CONCLUSIONS AND FUTURE SCOPE

5.1 Conclusion This dissertation presents a unified approach to orthogonally multiplexed communication using wavelets. In particular, we first used orthonormal dyadic wavelet functions or DWT-OFDM. This waveform achieves a non uniform partitioning of the data bandwidth. It was then shown how DWT-OFDM provided an efficient digital implementation. The modified model of DWT-OFDM was simulated using MC technique; comparison was made on the basis of BER vs SNR curves which proved that DWT-OFDM is better than conventional OFDM with Haar based DWT giving the best results showing an improvement of 15 db at BER. Then we used orthonormal wavelet packet functions or WPM-OFDM to define a communication waveform which provides a uniform decomposition of the data bandwidth. Again all fundamental characterization of the waveform was determined. Additionally, a modified digital implementation based on tree structured uniform analysis-synthesis filter banks was introduced, comparison of BER vs SNR curve showed an improvement of 3 db at BER. In the later part of dissertation implementation of LS and MMSE channel estimation techniques over wavelet OFDM was performed. MMSE technique outperformed LS technique by 16 db at BER but it has higher complexity.

5.2 Future Scope

46

This dissertation has developed the foundations of two new modulation formats and has successfully applied them to a pertinent application and different fading channels. Future work can be done in following few directions: Throughout this dissertation synchronization has been assumed. In order to produce a working prototype, approaches to synchronization must be developed and analyzed. This could include both open loop approaches based on parameter estimation techniques and closed loop approaches for highly dynamic conditions. Gain control is a particularly important subject area for the spread spectrum application. That is, the results in this dissertation have shown that very strong interferences can be mitigated. But these strong signals could cause significant problems for RF processing and ND conversion. The analysis presented in this dissertation assumed the filter weights were known by the receiver. Practical approaches to estimating the filter weights are therefore needed. Since the problem formulation resembles the sensor array problem, a starting point would be to consider existing techniques from this field. Adaptability of the estimation procedure should be addressed so that time varying interference can be effectively mitigated. Also a detailed analysis should be performed to understand the effects that the noisy weights have on theoretical performance. For the case of MSM, the multirate nature of the signals invites the possibility of new coding strategies. Using the notion of a supersymbol, one could develop a coding strategy which could compensate for frequency selective distortions in the channel.

47

REFERENCES

[1] J.Proakis, Digital Communications, Tata McGraw Hill, 1983. [2] A. Goldsmith, Wireless Communication, Cambridge University Press, 2005. [3] G. L. Stuber, Principles of Mobile Communication, Second Edition, Springer Science and business media, inc. (USA), 2002. [4] Tutorial Intuitive Guide to Principles of Communications from www.complextoreal.com. [5] Adarsh B. Narasimhamurthy, Mahesh K. Banavar, and Cihan Tepedelenlioglu OFDM Systems for Wireless Communications Pearson Education (USA), 2005 [6] M. Kumar, S. Shrivastava, A. S. Raghuvanishi and S. Tiwari, "Channel estimation for Wavelet based OFDM System", International Conference on Devices and Communications (ICDeCom), pp. 1-4, Feb, 2011. [7] Che-Shen Yeh and Yinyi Lin, Channel Estimation Using Pilot Tones in OFDM Systems IEEE Transactions On Broadcasting, Vol. 45, No. 4, December, 1999 [8] Yong Soo Cho Jaekwon Kim Won Young Yang Chung G. Kang, Mimo-Ofdm Wireless Communications with MATLAB, John Wiley and Sons (Asia) Pte. Ltd., 2010. [9] Stephane G. Mallat, A Theory for Multiresolution Signal Decomposition: The Wavelet Representation IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 11, No. 7, July, 1989 [10] I. Daubechies, Ten Lectures on Wavelets, Academic Press, 1992. [11] Michel Misiti,Yves Misiti, Georges Oppenheim, Jean-Michel Poggi, Wavelet Toolbox-MATLAB R2012a user guide, Mathworks inc., 2012. [12] Raghuveer M Rao and Ajit S Bopardikar, Wavelet Transforms-Introduction to Theory and Applications Addison Wesley Publications, 2006 [13] Hans-Georg Stark, Wavelets and Signal Processing An Application-Based Introduction, Springer Publications, 2007

48

[14] A. Lindsay, Wavelet Packet Modulation for Orthogonally Transmultiplexed Communications, IEEE Transactions on Signal Processing, Vol. 45, pp. 1336-1339, May 1997. [15] C. Van Bouwel, J. Potemans, S. Schepers, B. Nauwelaers, A. Van de Capelle, K.U.Leuven, Wavelet Packet Based Multicarrier Modulation, IEEE Transactions on Signal Processing, Vol. 51, pp. 1263-1288, April 1998. [16] S. Baig, F. U. Rehman, and M. J. Mughal, Performance comparison of DFT, discrete wavelet packet and wavelet transforms, in an OFDM transceiver for multipath fading channel, Proceedings of 9th International Multitopic Conference, INMIC05 , pp. 1-6, Sept. 2005 [17] M.K. Lakshmanan and H. Nikookar, Construction of Optimal Bases for Wavelet Packet Modulation under Phase Offset Errors", 3rd European Wireless Technology Conference, Paris, France, September, 2010. [18] M. K. Lakshmanan, H. Nikookar, "A review of wavelets for digital wireless communication", Wireless Personal Communications, Springer Netherlands, vol. 37, No. 3-4, pp. 387-420, May, 2006. [19] D. Gupta, V. B Vats and K. K. Garg, "Performance Analysis of DFT-OFDM, DCTOFDM, and DWT-OFDM Systems in AWGN Channel", IEEE Fourth International Conference Wireless and Mobile Communications,( ICWMC '08), pp. 214-216, Athens, 2008. [20] M. Gautier, M. Arndt and J. Lienard, Efficient wavelet packet modulation for wireless communication, in AICT Vol. 3, pp. 231-234, Mauritius, May 2007. [21] Mirghani, R. & Ghavami, M., Comparison between Wavelet-based and Fourierbased Multicarrier UWB Systems, IET Communications, Vol. 2, Issue 2, pp. 353-358, Feb 2008 [22] S. Weinstein and P. Ebert, Data transmission by frequency-division multiplexing using the discrete fourier transform, , IEEE Transactions on Signal Processing vol. 19, pp. 628634, 1971. [23] Khaizuran Abdullah and Zahir M. Hussain, "Studies on DWT-OFDM and FFTOFDM Systems", International Conference on Communication, Computer and Power M, Vol. 4, pp 431-435, FEBRUARY 15-18, 2009

49

[24] W.W. Jones, "A Unifed Approach to Orthogonally Mulliplexed Communication Using Wavelet Bases and Digital Filter Banks", Ph.D. Dissertation, Ohio University, Athens, OH, August, 1994. [25] William H. Tranter, K. Sam Shanmugan, Theodore S. Rappaport, Kurt L. Kosbar Principles of Communication Systems Simulation with Wireless Applications, Pearson Education, 2004.

50