CHAPTER 5

SYNCHRONIZATION SCHEMES FOR MC CDMA SYSTEM

5.1 INTRODUCTION

Information about the communication channel. such as the channel phase

response, is necessary for the conshuction of various receivers. In many practical

situations, this information is not known apriori and the relevant channel parameters

have to be estimated from the received signal. The three maln channel parameters

required by most receivers are the camer frequency, the carrier phase, and the symbol

timing of the received signal. The camer frequency of the received signal may be

different from that of the nominal value of the transmitter carrier frequency. This

discrepancy can be the results of the deviation of the transmitter oscillator from the

nominal frequency and more importantly, the Doppler effect when the transmitter is

in motion relative to the receiver. In reality, it takes a finite amount of time for the

information-bearing electromagnetic wave to travel from the transmitter to the

receiver. This transmission delay introduces a mismatch between the symbol t~ming at

the transmitter and that at the receiver. The output of the matched filter is to be

sampled at an exact time to optimize the error performance. The symbol timlng at the

receiver is to be known (or equivalently, the transmission delay) in order to eliminate

the performance degradation due to the timing mismatch. The canier phase of the

received signal is the sum of three major components, namely, the random phase of

the transmitter oscillator, the channel phase response, and the phase due to the

transmission delay. In order to model all the three channel defects, a simple non-

dispersive channel model is to be augmented. The received signal in the augmented

model is given by

where r(t) is the received signal and the new parameter fd and T are employed to

model the deviation of the received carrier frequency from the nominal camer

frequency and the transmission delay, respectively. Very often, the phase terms in

(5.1) are combined to a single phase term ( = 0- 2n(f, + f,)r modeling the received

carrier phase.

The process of estimating this parameter is called synchronization. The

process of estimating the carrier phase is known as carrier phase synchronization and

can be accomplished by a Phase Locked Loop (PLL) circuit. The process of

estimating the transmission delay is known as symbol timing synchronization, which

can be accomplished by a delay locked loop (DLL) circuit. It turns out that the same

PLL circuit used for carrier phase synchronization can also be employed to track the

canier frequency mismatch when it is significant.

Consider a BPSK system in which the received signal is given by

where n(t) is an AWGN process with noise spectral density N,J2. Using

synchronization process, the estimates of the parameters fd, T and gS are obtained.

Therefore, the correlator demodulator based on these estimated parameters as shown

in Figure 5.1 is formed.

Fig.S.1 Correlator receiver for BPSK with estimated parameters

It follows that when r -r < T , the average bit error probability (assuming the I r l bit values '0' and '1' are sent with equal probabilities) given by

where a = - E cos[2x( f, - j,)r +( ) - ; ) ]d l T ' "7':' mu(r.r)

It is noted that a5 l with equalityonly if f, = f,, r = r and #=/(mod 2x)

Hence, error in the estimation of any of the parameters will cause the error

probability to go higher than the optimal value Q,( - ). From the above result, it E- is found that if the data rate is much higher than the estimat~on error In fd. then

where A s = s -1 and A$= $ -$ . Therefore, errors in estimating both transmission

delay and the camer phase increases the bit error rate.

Receiver synchronization is a major problem in a mobile radio env~ronment

where the communication channel is subject to rapid changes. Communication In

spread spectrum systems is impossible unless the received spreading waveform and

receiver generated replica of the spreading waveform are initially synchronized in

both phase and frequency. Phase and frequency synchronization is usually

accomplished by performing a two dimensional search in the timelfrequmcy

ambiguity area. Generally, this process must be accomplished at very low SNR, as

quickly as possible, using the minimum amount of hardware [174]. There are four

levels of synchronization

Carrier Synchronization . Symbol Synchronization . Frame Synchronization

Network Synchronization

In case of coherent phase modulation, the receiver is assumed to be able to

generate a set of reference signals whose phases are identical (expect perhaps for a

constant offset) to the phases of the signalling alphabet in use at the transmitter. These

reference signals are compared with the incoming signals in the pmess of making

maximum likelihood symbol decisions. In order to be able to generate these reference

signals, the receiver has to be in synchronization with the received carrier as shown in

the Figure 5.2. This means that there has to be phase concurrence between the

incoming carrier sinusoid and a replica of it in the receiver. In other words. if there

were no information modulated, the incoming sinusoid and the replica In the recelver

would pass through zero simultaneously. This IS what is known as being in phase lock

and is a condition that must be closely approximated ~f coherently modulated signals

are going to be accurately demodulated at the receiver.

5.1.2 Symbol Synchronization

All digital receivers need to have demodulators synchronized to the incoming

digital symbol transitions in order to achieve optimum demodulation as shown in the

Figure 5.3. The symbol synchronization can be classified into two basic groups. The

first group consists of the open-loop synchronizers, where circuits recover a replica of

the transmitter data clock output directly from operations on the incoming data

stream. The second group comprises the closed loop synchronizers. where closed-loop

data synchronizers attempt to lock a local data clock to the incoming signal by use of

comparative measurements on the local and incoming signals. Closed-loop methods

tend to be more accurate, but are much more costly and complex.

5.1.3 Frame Synchronization

For a receiver to make sense of the incoming data stream, the receiver needs to

be synchronized with the data stream's frame structure. Frame synchronization is

usually accomplished with the aid of some special signalling procedures from the

w t t e r . This procedure may be very simple, or fairly involved, depending on the

environment in which the system is required to operate.

Out of Synchronization Out of Synchronization Synchronized

Fig. 5.2 Carrier synchronization

Error Detection Error Detection Correct Detection

Fig. 5.3 Symbol synchroniution

74

nk bits Nd bits nr bits Nd bits nr bits

Data Stream

n* bits N~ bits nir bits Nd bits nk bits

Receiver generated frame marker replica

Fig.5.4 Frame synchronization

The simplest frame synchronization aid is the frame marker, illustrated In the

Figure 5.4. The frame marker is a single bit, or a short pattem of bits that thc

transmitter injects periodically into the data stream. The receiver must know the

pattem and the injection interval. The receiver, having achieved data synchronization,

correlates the known pattern with the incoming data stream at the known injection

interval. If the receiver is not in synchronization with the framing pattern, the

accumulated correlation will be low. Hence the receiver comes into frame

sqnchronization; however, the correlation should be nearly perfect, blemished only by

an occasional detection error.

The advantage of the frame marker is its simplicity. Even a single bit can

suffice as a frame marker if a sufficient number of correlations are accumulated

before deciding whether or not the system has achieved synchronization. The major

drawback is that the sufficient number may be very large, and thus the expected time

required to acquire synchronization would be long. Therefore, frame markers are most

useful in systems that transmit data continuously, and would be inappropriate for

systems that transmit in isolated bursts or systems that require rapid frame acquisition.

For systems with inconsistent or burst transmission or systems with rapid quisition

'eguirements [143], a synchronization codeword would typically be sent as pM of a

message header. The rtceiver must know the codeword and be constantly searching

for it in the data stream. Detection of the codeword would indicate a known position

in the data kame. The advantage of this system is that frame acquisition can be

essentially immediate. The only delay would be that required to process the incoming

codeword. The disadvantage is that the codeword must be relatively long, relative to

the frame marker, to keep the probability of false detection low. The complexity of

the correlation operation is proportional to the length of the sequence, so the

correlator must be relatively complicated.

5.1.4 Network Synchronization

For systems using coherent modulation techniques, one-direction

communication such as broadcast channels, or single link communication, such as

most microwave links, land link, or fibre optic links, the synchronization architecture

that makes the most sense is to make the synchronization totally a receiver function.

For communications systems using non-coherent modulation techniques, or that

involve many user accessing a central communication node, such as satellite

communication systems, it often makes sense for synchronization to be mostly or

entirely a terminal function. This means that the terminal transmitter parameters are

modified to achieve synchronization, rather than modify the central nodes receiver

parameters. This must be the approach if the system uses TDMA. In TDMA each user

is allocated a segment of time to transmit information. The terminal transmitter must

he synchronized with the system in order for its transmitted burst of data to arrive at

the central node at the time when the node is prepared to receive the data.

Synchronization of the terminal transmitter also makes sense with systems that

combine signal processing at the central node with FDMA. If the terminals precorrect

their transmission to be synchronized with the central node, the node can use a fixed

set of channel filters and a single timing reference for the processing of all channels.

Otherwise, the node would require a separate time and frequency acquisition and

tracking capability for each incoming channel, and would need to deal with the

possibility of varying amounts of adjacent channel interference. It is quite obvious

that transmitter synchronization is often the better and more reasonable system

Vproach to synchronize a network. For CDMA based systems carrier and s)?nbol

Spchronization is important.

In the first part of this chapter, a novel method to estimate the frequency offset

for the desired user is proposed to be used in conjunction with the MU1 reduction and

suppression techniques discussed in chapters 2 to 4. In the second part of this chapter

a Non data aided frame synchronization i.e OFDM symbol synchronization algorithm

is proposed.

5.2 FREQUENCY SYNCHRONIZATION SCHEME FOR MC CDMA

SYSTEM

Multicanier transmission and, notably. MC CDMA system, is known to

provide remarkable resilience to multipath fading and impulsive noise. In order to

operate correctly, an MC CDMA receiver calls for accurate compensation of the

camer frequency offset in the input signal. In MC CDMA system, available

bandwidth is decomposed in to a set of disjoint equal bandwidth frequency. Since the

bandwidth of each sub band is less than the coherence bandwidth of the channel, MC

CDMA systems are more robust to the distortion induced by time-disperse channels

than DS CDMA systems. At the front-end of the receiver MC CDMA signals are

subject to synchronization errors due to oscillator impairments and sample clock

d~fferences. The demodulation of the received radio signal to baseband, possibly via

an intermediate frequency, involves oscillators whose frequencies may not be

perfectly aligned with the transmitter frequencies. This results in a carrier frequency

offset.

The most important effect of a frequency offset between a transmitter and a

receiver is the loss of orthogonality between the subcaniers resulting in inter canier

interference (ICI). The characteristics of this ICI are similar to white Gaussian noise

and lead to a degradation of the SNR. For both AWGN and fading channels, this

degradation increases with the square of the number of subcarriers. Like frequency

offsets, phase noise and sample clock offsets cause ICI and thus result in a

degradation of the SNR. In DS CDMA systems, many detection schemes are

proposed to eliminate the MUI. However if these schemes are used for MC CDMA

Systems, their performance deteriorates even in the presence of a small frequency

offset Therefore, the frequency offset and the MU1 should be jointly wnsidered.

Many schemes are proposed to estimate the frequency offget through the use of PLL,

frequency offset estimation and guard interval based hquency offset estimation

algorithms.

5.2.1 PLL Algorithm

PLL Algorithm [I751 operates in two states, ( I ) initial lock acquisition and (2)

tracking, as shown in Figure 5.5. The initial state is the Initial Lock Acquisition state,

where the switch SW is open and the incoming signal symbols rotate with normalized

frequency offset AtTs. The carrier locks detection, obtained by using the bank of

PLDs, work in a window of size N, symbols. The PLDs output is either "I" or "0"

meaning "Lock" or "Non-lock, respectively.

N,, = N,,t I End 1

Rg. 5.5 General logic flow chart of carrier recovery by a typical PLD

The PLL parameters, which are PED and LPF, are reconfigured to adapt the

modulation format and SW is closed so that the PLL can start to work. If every PLD

outputs "O'', the state does not change, and the bank of PLD's work for initial

acquisition in the next window. If lock is lost in tracking state the state is changed

back to the initial lock acquisition state and SW is open. The acquisition time depends

on two parameters: the frequency offset and the phase difference between the adjacent

lock areas. The value of frequency offset and phase difference between the adjacent

lock area affect the rotation speed of the received signals and the maximum h m the

received signal position to the lock area, respectively and the signal is lost. This

increases the acquisition time and hence it is difficult to achieve synchronization.

5.2.2 Frequency Offset Estimation Algorithm

The estimation of frequency offset is addressed in a number of contributions in

this method. As a key part of the synchronization scheme, an estimator is proposed

based on the concept that works without the aid of pilot symbols and is independent

of the modulation of the carriers. In this method statistical redundancy in the received

signal, introduced by the cyclic prefix, provides the information about the offset. A

single OFDM symbol received by the base station is considered. It is assumed that the

N subcarriers constituting this symbol are subdivided in M bands of subcarriers, the

indices of which are collected in the set Mm. One transmitted OFDM symbol in the

mth band of subcaniers is

where NT is the duration of the OFDM symbol without the cyclic prefix and T, is the

length of the cyclic prefix. The model focuses on the frequency sub band division

property of the multiple-access scheme as this significantly affects the offset

estimation. The receiver offset estimator addresses the time-division property of the

target system by applying one estimator to every time slot. Perfect separation of the

users typically is accomplished by the removal of the cyclic prefix and the

demodulation by the FFT. Such separation, however, removes the redundancy which

is needed by the offset estimation in the synchronization scheme.

The MC CDMA signal as shown in the Figure 5.6 which is received by the

base station, is assumed to have N subcarrim and are subdivided in M bands of

subcarrim, the indices of which are collected in the set Mm. It associates with the mh

transmitted signal a time offset relative to the receiver symbol clock and a frequency

offset relative to the receiver demodulation frequency. In this system the requirement

of a maximum frequency offset of 1-2% of the intercarrier spacing becomes about

50 Hz. Each symbol is cyclically extended by 48 sec (205 samples), such that the total

symbol length becomes 288 sec. The length of the cyclic prefix is suficiently long to

provide immunity against channel dispersion. The system will perform better only if

the channel impulse response does not exceed the cyclic prefix duration.

Corrshtion Averaging

Fig. 5.6 Frequency offset estimation

5.2.3 Guard Interval Based Frequency Olfset Estimation

This section explains a new algorithm for the carrier frequency offset

estimation in an MC CDMA receiver. The samples n (-Ng S n SN,T -1) of the

received baseband signal r(t) belonging to the ifi block are shown in Figure 5.7. The

first Ng samples of the block (-Ng Sn S-1) belongs to the guard interval, while the

other NBT samples (0 9 -1) represent the useful block. This algorithm uses the

redundancy introduced by the guard interval.

Fig.S.7 Guard interval based frequency offset estimation

When the frequency error Af is null, rNs7 .,r*., is a real number. However, in

the presence of a frequency error the two samples r ~ N ~ T ., and r*., are affected by a

different rotation and the imaginary part of their product contains some information

about the frequency offset. The basic idea is to use this information for frequency

offset correction. The error signal €(I) corresponding to the lth block is given by

€(I)= (llL) C I r n [ r ~ ~ ~ ., r*.,] with 1 a SNg

The main disadvantage of this method is that noise due to multipath Rayleigh fading

channel cannot be eliminated very effectively since the subcaniers will be affected by

the AWGN noise also.

These schemes are good for OFDM but are quite troublesome with MC

CDMA due to the MU1 and the full correlation between the different subcaniers.

A novel method is proposed here for frequency synchronization in MC CDMA

system.

5.2.4 Proposed Virtual Subcarrier Based Synchronhtlon Algorithm

This method capitalizes on the fact that the system is not fully loaded, i.e., not

all available subcarriers are used for transmitting symbols. The proposed method

exploits these null, or virtual, subcaniers to enforce time and f q u e n c y

synchronization conversely, the method herein, guarantees unambiguous carrier offset

estimates with performance independent of the channel zeros location. Moreover, it is

specifically devised for multi user systems and is simpler to implement than the

algorithms explained earlier because it does not require any eigen-decomposition of

the received signals. Even though the presence of virtual subcarriers can be seen as

the transmission of known (null) symbols, there is a basic practical difference as the

transmission of null symbols does not imply any waste of power. Furthermore, null

sub carriers are intrinsically present in any non fully loaded MC system (a situation

which occurs most of the time). Figures 5.8 and 5.9 depict a typical MC CDMA

transmission and reception system with proposed synchronization algorithm. To

verify the algorithm steps given in Figures 5.8 and 5.9 is followed and channel

simulation is similar to the one described in section 2.4.

The novelty lies in the algorithm for estimating and compensating both time

and frequency offsets. The basic idea for synchronization is fairly simple. The average

energy falling across the virtual subcarriers is measured and the oscillator frequency is

updated until a minimum of the measured energy is reached. In case of perfect

synchronization and no noise situation, null energy is observed. When some energy is

detected, the receiver oscillator is updated in order to minimize such energy.

Correspondingly, the synchronization algorithm is rather simple. An iterative search is

run which yields the frequency offset that minimizes the energy falling in the band

corresponding to the virtual subcarriers. Perfect synchronization leads to zero

MWISI (inter symbol interference) among MC CDMA symbols and null energy in

the band not occupied by information symbols. Conversely, non perfect

synchronization is revealed by the presence of MUIDS1 and of a non null energy in

correspondence with the null subcarriers.

5.25 Performme of V i a l Subcarrier Algorithm

MC CDMA system is simulated with 16/32/64 Recessing gains and

subcarriers, 214 virtual subcarriers and the channel is assumed to be Rayleigh fading

in AWGN environment. Rgure 5.10 shows the comparison of BER plot for the

MC CDMA system with and without ~j'nchr~nization for 16 subcarriers/processing

gain out of which 4 are virtual subcarriers. Simulation is performed for SNR from 0

dB to 10 dB in steps of 1 dB and the result implies that the system performs better

with synchronization. Figures 5.1 1 and 5.12 depict BER plots for MC CDMA system

for 32 and 64 subcaniers/processing gain respectively with 4 virtual carriers. Figure

5.13 compares the performance of a synchronized system using the proposed

algorithm with 32 SCPG for 214 VC. Figure 5.13 brings out that for the same

processing gain the performance of the system with 4 virtual carriers is better than the

2 virtual carriers. This is due to the fact that average error from the 4 VC can be used

effectively for frequency correction than from 2 VC. Figure 5.14 highlights the

performance comparison of the system with 64/16 PG for 4 VC, which indicates that

the performance is better for higher processing gain.

Fig.5.10 System performance for 16 SU16 PG and 4 VC

84

Fig.S.ll System performance for 32 SU32 PG and 4 VC

ld

16'

ig 10."

10'

SNR in dB

Fig.5.12 System performance for 64 SU64 PG and 4 VC

Fig.5.13 BER comparison of 2 4 VC with 32 SC/PG

10.' :

Bk

SNR in a

Fig.5.14 BER comparison of 64/16 PG with 4 VC

5.3 W E SYNCHRONIZATION FOR MC CDMA SYSTEM

Many systems used in digital communication smd data in synchronous back-

to-back frames. Whm a receiver tunes to such a data stream, it has no knowledge of

the frame boundaries. Frame synchronization is required to ensure that the receiver

reliably interprets the received bit stream on correct boundaries. Frame

synchronization is usually accomplished with the aid of some special signalling

procedure from the transmitter. The data aided technique and the non-data aided

technique are the two types of frame synchronization.

5.3.1 Data Aided Algorithm

In data aided technique some non information bits are stuffed at regular

intervals for frame synchronization purpose. The major drawback in this technique is

that the frame marker size may be very large in order to avoid false detection and the

expected time required to acquire synchronization would be long and also

transmission overhead is present. In the case of non data aided technique, the data

which is to be transmitted will be encoded in such a way that, encoded data itself can

be used for frame synchronization. In the case of CDMA systems, reception is not

possible unless the received spreaded waveform and receiver generated replica of the

spreading waveform are synchronized in both phase and frequency. Generally, this

process must be accomplished at a very low SNR, as quickly as possible, using the

minimum amount of hardware. As the data aided scheme incorporates transmission

overhead, it decreases channel capacity and bandwidth efficiency.

5.3.2 Non Data Aided Algorithm

To overcome this problem, nondata aided technique is being practiced,

wherein the information bit itself is used for frame synchronization. The advantage of

this technique is that it does not require any transmission overhead, it increases

channel capacity and bandwidth efficiency. The proposed frame synchronization

algorithm uses the non data aided technique and also supports variable 6ame size for

all users. The existing non data aided frame synchronization method for a single user

multi carrier system is described below. The basic idea behind this technique is, the

deletion of a subchannel is used to indicate the start of the frame. The subchannel to

be deleted is determined by the encoding block and the information of the deleted

subchannel is also retrieved by the algorithm used. For example consider 32 symbols

are there in a block of a frame. If the first symbol is 3, data in the third position is

placed in fust position and the third symbol is made zero. Therefore both the

information in first and third position is available and in the mean time inserted zeros

ensure frame synchronization. Let x be G random data symbol vector, i.e.

x= {XI x2 . . . XGI and D = { 1, Z... U), where denotes the transpose and D is the

data alphabet of size U. Each data symbol is mapped to a signalling symbol, y,, using

the MAP operation. The mapping is such that each possible data symbol is associated

with one particular value of a set of U distinct values y = [y I y 2. . y G] T, yn= A(x , ), A = (A(1) , A(2), ..., A(U) j, yn E A where n E [],GI, A is the signalling alphabet

with size U. Each y, is allocated a subchannel. The vector y is encoded by the Frame

encoding block only if it is the first symbol of a frame; otherwise, Z = y. Then, Z is

modulated to produce a symbol vector S. The Rayleigh fading channel is described by

a diagonal matrix C(a), with the fading coefficients on the main diagonal. Also the

additive white Gaussian noise, q(t), is combined with the signal. Thus, received signal

is r = C (a) x S + ~ ( t ) . The receiver consists of a demodulation block, DEMOD,

followed by frame detection and decoding block, FDDE, a signalling symbol decision

block, SSD, and a signalling symbol demapper, DMAP [I 761. In the receiver, perfect

OFDM symbol, carrier recovery and perfect channel estimation are assumed. A linear

equalizer is used to offset the effect of fading in the DEMOD. The frame encoder only

encodes the first symbol of every frame. The total number of subchannels, G, in

vector y is divided into q groups of U sub channels for frame encoding

y=[ y(l) y(2) y(3) ...y(q ) y, lT where y(n) is a IxU vector and y, is a vector of

size (G - qU) containing the remaining signalling symbols. Here, q is defined as the

maximum group available for frame encoding and is given by the integer part of the

division of G by U. Note that G > U is required here. The vector y~ is not involved in

the frame encoding process. Each vector y(n) in y is encoded as follows:

21 C y,l,Z ,I f O,& t y,, where p E [2,U], n # xl. The xl and 21 correspond to the

first data symbol and the signalling symbol in the first subchannel in each vector y(n)

respectively. The first data symbol, xl, specifies the i' subchannel which will replace

the first subchannel, Z I

FRAME

G SYMBOLS

NO

SPLIT

SUBGROUPS

ENCODED FRAME

I

I I

I

1

Fig.5.15 Frame encoding block

SUB GROUP

1

SUB GROUP

( G N )

SUB GROUP

2

REMAINING SYMBOLS (G-(q*U))

1 1 P

T C Y ( 1 ) Y ( l ) + Y( 'n Y(T) l 0

T C Y(1) Y ( l ) + Y(T) Y(T) C 0

7

- T C Y(I) Y ( l ) C Y(T Y(T) l 0

The flow chart describing this is given in Figure 5.15.Then. the i"' subchannel

is deleted and no signal energy will be transmitted in i' subchannel. It should be

noted that no information is lost in this process because the information in i*

subchannel is transmitted via the first subchannel, and the information in the first

subchannel is specified by the position of the deleted subchannel. The frame detection

and decoding block must first detect the start of the frame before it can recover the

data from a frame encoded symbol. Like the frame encoder, the received symbol is

divided into q groups of U subchannels: V= [ v(l) v(2) v(3). . .. . ... v(q ) v, lT where v(n) is l x U vector of the received signalling symbols, v(n) and v(r) is a vector

of size (G - qU) containing the remaining received signalling symbols. Again, the

vector v(r) is not involved in the frame detection or frame decoding processes. In each

v(n)vector, a subchannel that has the minimum absolute magnitude among the

subchannels within the group is selected. Then. the selected minimum absolute

1 mapitudo subchannels from all the groups are summed, i.e E = - imin(uh(u(n)))

4 I

where min(.) is the minimum of a group of number n and n E [I,q] and abs(.) is the

absolute value of the argument. E is then compared with a predefined threshold. This

threshold can be chosen according to the required false alarm and miss detection rates,

and considering the prevailing channel conditions: E > T F , HA and E < T F, HF where

TF is the threshold for frame detection, HF and HA are the hypothesis indicating the

frame is present or absent respectively. If all the elements in signalling alphabet set A

have similar energy and the consequences of a false alarm and a miss detection are the

same, then the threshold should be chosen as half the maximum absolute magnitude.

After the frame is detected, the subchannels in the frame encoded symbol have to be

rearranged to recover the actual symbol. This is simply the reverse of the frame

encoding procedure. For each v(n) vector, the position of the minimum absolute

magnitude subchannel among the subchannels within the group is determined by

i = pos (min (abs(v(n) ) ) ) where n E [I,q], and pos(.) is the position of the argument

in a vector. Then, the signalling symbol in i" subchannel is replaced by the signalling

vmbol in the first subchannel, and the signalling symbol for the first subchannel is

the ih element in signalling alphabet set, Z, f V 1 , Z l f A(i), Z,, f Vn where

n E [2,U], n #I. This frame decoding procedure is implemented only when a frame is

detected, otherwise, Z=V.

The existing method shows that the h e information can be embedded in the

normal data stream without the need for a special preamble sequence. Consequently,

h e synchronization can be achieved without transmission overheads. Furthermore,

the method has low complexity and is simple to implement. The major drnwbacks of

this algorithm are that FER offered is not satisfactory and it does not support multiple

services which is required for future generation services.

5.3.3 Proposed Modified Algorithm

In the modified algorithm, the main objective is to reduce the frame error rate.

This is accomplished by incorporating the barker sequence in the existing algorithm.

The existing algorithm encodes only the fint G symbols and transmits the remaining

data bits of a particular frame which is of a fixed size as such. Wherein here the data

bits are frame encoded as in the existing algorithm to indicate the start of the frame

with a modification. Here instead of zeros, a five bit barker sequence, say 11 101 is

used as a marker. For a bit value of 1, the algorithm is applied for G symbols. For a

bit value of 0, the G symbols are transmitted as such. The frame error rate for the

proposed algorithm is seen to be less when compared to the existing one.

As the existing algorithm provides only a fixed frame size for a particular user,

it is not possible to support integrated services. In the existing frame synchronization

scheme, if the frame synchronization technique is applied to all q groups of U sub-

channels, then the frame size is constrained to be any multiple of G symbols.

However, if it is applied to the upper half of the q groups and to the lower half of the

q groups independently, then the frame size can be any multiple of O.SG. If the upper

half and lower half are fbther split into two subgroups, the frame size can be any

multiple of 0.25G. Nevertheless, this can be done only if the division of G by the

number of subgroups, ns, yields an integer. It should be noted that, as ns increases, the

number of sub-channels that can be used for frame detection in each subgroup, Glns,

decreases. Hence, the flexibility in frame size is inversely dependent on the robustness

of the frame synchronization. The minimum possible frame size, U, is achieved when

ns = q, where U is the signalling symbol alphabet size. Thus the existing algorithm

ensures different frame size for different user. But if each user is assigned a fixed

frame size, one cannot implement integrated service facility. By varying the size of

the frame for each user, integrated senices can be incorporated. If the frame size is

made variable in the existing algorithm the probability of miss (p3 and false alarm

@I) is large resulting in increased h e e m r rate. But in the proposed algorithm with

the help of the barker sequence the data bits are frame encoded as in the existing

algorithm to indicate the start as well as the end of the frame. The e n c d n g is carried

at the start and end of the frame and the intermediate symbols are transmitted as such.

At the receiver side, the probability of a 0 in the marker sequence becoming a 1 is

very less. This property of the algorithm decreases the probability of false detection

and facilitates variable frame sizes at reduced FER.

5.3.4 Performance Comparison for Single Service and Multi Service System

With the MC CDMA system setup, the frame synchronization block is

included to implement the data aided technique. The source alphabets D are randomly

generated which takes a value up to U, say U = 7. In this trial 2000 symbols are

generated. The generated symbols are converted to bits and the frame marker 1s

inserted at the start of each frame. The marker sequence used here is a 15 bit code,

Marker = [ 1 1 1 1 0 1 0 1 1 0 0 1 0 0 01. The resultant bits are spreaded and is

modulated using BPSK. The modulated signal is transmitted through the Rayleigh

fading channel and the performance of the system is analyzed. At the receiver side,

coherent detection is carried out and the bits are retrieved. For various signal to noise

ratios (SNR), the bit error rate (BER) is calculated. Figure 5.1 6 explains the simulated

plot between BER and SNR in Rayleigh channel for a MC CDMA system

incorporating data aided algorithm. In non data aided technique, the source alphabets

x are generated randomly which takes a value from D. x = [XI, XI, x3 .... x,] and

D = [I , 2, 3, U] where x, E D. Here U = 7 and n = 2000. The generated symbols are

mapped using a matrix, map = [7 6 5 4 3 2 11, that is in the reverse order. The frame

size (F) is fixed to be of a particular value (say 100 symbols). A number G (say 25)

which is a factor of F is chosen and the first G symbols in each frame are considered

for the frame e n d i g and the remaining symbols are transmitted as such. The G

symbols are divided into q groups each having U symbols, where q is mod ( G U . The

first symbol in each group is taken and demapped and the position of the s p b o l to be

encoded is given by this value. The symbol at that position is moved to the first

psition of the group and the symbol at that position indicated by demappad value is

deleted (made as zero). Hence the position of the deleted symbol gives the actual first

symbol in the group and the first symbol present in the group gives the actual symbol

at the deleted position. Hence no data is lost in frame encoding and no additional bits

are required for frame synchronization. This is followed for every frame and it is

transmitted through MC CDMA system similar to the data aided technique as

explained above.

Figure 5.17 shows the simulated plot between BER and SNR in Rayleigh

fading channel for a MC CDMA system incorporating non data aided algorithm. At

the frame detection block, the fiames are scanned for the G symbols that are frame

encoded. If it identifies that G symbols, then it is indicated as the start of the frame

and a stream of G symbols are added to it, to have a complete frame (100-25 = 75

symbols). Again the receiver scans for the next starting point of the frame. If the

frame start is not identified the frames are left as such. Figure 5.18 compares the BER

for non data aided and data aided techniques in Rayleigh fading channel in a

MC CDMA system. From this figure it is clear that there is not much of a difference

in bit error rate for non data aided and data aided techniques. This is due to the fact

that whatever be the frame synchronization algorithm used the bit error rate depends

only on the channel and the type of modulation used. Further the BER depends on the

multiple access technique used and the system provides a minimum BER due to the

multi canier code division multiple access technique used. Hence the difference is

produced only in the frame error rate of the systems. To analyse the performance of

modified algorithm, a barker sequence of length 'mb' is chosen, which satisfies the

correlation property. The first mb x G symbols in the frame is considered for frame

encoding, where mb is the length of the barker sequence. The barker sequence is

assumed as 11 101 and mb = 5 and G = 25. Hence first 125 symbols are taken and are

kame encoded with respect to the barker sequence. The digit '1' in the barker

sequence indicates that the G symbols are frame encoded as explained above. The

digit '0' indicates that the G symbols have to be transmitted without frame encoding.

Thus the first 125 symbols arc frame encoded based on barker sequence.

Fig.5.16 Performance of data aided algorithm

10.'

P to"

: \\ \

\ !

'0'1 1 s t i i 2 i i 3 b ; SLIR in dB

Fig.5.17 Performance of non data aided algorithm

Fig.5.18 Comparison of non data aided and data aided schemes

Table 5.1 and Figure 5.19 show the comparison of FER in Rayleigh fading

channel for modified algorithm, the existing non data aided and data aided technique

for fixed frame size. The FER of the modified algorithm is very less when compared

to the existing technique for a fixed frame size. This is due to the correlation property

of the barker sequence which provides a reduced probability of false alarm and

probability of miss. In order to make the frame size a variable one and to reduce the

frame error rate further, the frame marker encoding is canied out both at the start and

end of the frame. i s , the last 125 symbols are frame encoded with the same barker

sequence as explained earlier. The comparison of the frame error rate for the different

dgorithms for variable frame sizes are shown in Table 5.2 and Figure 5.20 (S,- start

only encoded, S&& start and end encoded) and it reveals that the modified frame

synchronization algorithm with the frame marker encoding at the start and end of the

frame provides a better FER than any other scheme.

Table 5.1 CompPrison of the existing and d i e d algorithm for Rxtd frame size

- --

Fig. 5.19 FER Comparison of algorithms for fixed frame size

Table 5.2 Comparison of existing and modified algorithm for variable frame size

Fig.5.20 FER Comparison of algorithms (variable frame size)

5.4 CONCLUSION

In future wireless communication system, the main aim is to integrate all the

multimedia services (voice, data and video services) in to a single unit. Hence systems

which incorporate high data rate services are required to provide to the users with all

the facilities. The channel capacity of the system should be improved in order to

accommodate more number of users. Canier synchronization algorithm for MC-

CDMA systems has been proposed which tackles a very important issue in MC-

CDMA transmission in Rayleigh and Rician channels. In this investigation the

simulated results of BER without and with canier frequency synchronization are

being compared, and it is proved that the BER reduces when the synchronization

algorithm is incorporated, which in turn increases the capacity.

The modified frame synchronization algorithm presented in the second pM of

this chapter exploits the advantage of the Barker sequence and it can be seen that FER

reduces to a greater extent. This algorithm provides an added advantage to the Barker

sequence because of the property that the probability of the bit 0 of the frame marker

becoming a 1 is very less. This is due to the fact that at least one of the symbols in all

the q groups should be deleted in order to make a 0 as 1. Thus the probability of

losing a frame is highly reduced in this case. This algorithm also works for a variable

frame size of a particular user. In this case, to reduce the FER further, both the start

and end are indicated for each frame carrying out the same modified algorithm at both

ends of the frame. As capacity of CDMA system is interference limited, for a fixed

FER and BER, more number of users can be accommodated if this modified frame

synchronization algorithm is used. Thus the virtual subcarrier synchronization

algorithm and modified frame synchronization algorithm provide a better

performance than the existing algorithm for MC-CDMA systems and also ensure

efticient bandwidth utilization.