Copyright Warning & Restrictions

The copyright law of the United States (Title 17, United

States Code) governs the making of photocopies or other
reproductions of copyrighted material.

Under certain conditions specified in the law, libraries and

archives are authorized to furnish a photocopy or other
reproduction. One of these specified conditions is that the
photocopy or reproduction is not to be “used for any
purpose other than private study, scholarship, or research.”
If a, user makes a request for, or later uses, a photocopy or
reproduction for purposes in excess of “fair use” that user
may be liable for copyright infringement,

This institution reserves the right to refuse to accept a

copying order if, in its judgment, fulfillment of the order
would involve violation of copyright law.

Please Note: The author retains the copyright while the

New Jersey Institute of Technology reserves the right to
distribute this thesis or dissertation

Printing note: If you do not wish to print this page, then select
“Pages from: first page # to: last page #” on the print dialog screen
The Van Houten library has removed some of the
personal information and all signatures from the
approval page and biographical sketches of theses
and dissertations in order to protect the identity of
NJIT graduates and faculty.
 ABSTRACT

PHASE-LOCKED LOOP, DELAY-LOCKED LOOP, AND LINEAR

DECORRELATING DETECTOR FOR ASYNCHRONOUS
MULTIRATE DS-CDMA SYSTEM

by
Sok-kyu Lee

The performance of phase synchronization and code tracking of a digital phase-locked

loop (PLL) and delay-locked loop (DLL), respectively, is investigated in wideband

asynchronous multirate DS-CDMA system. Dynamic Partial Correlation (DPC)

method is proposed to evaluate the autocorrelation and its power spectrum density

(PSD) of the cross-correlated terms in the presence of multirate multiple access inter-

ference (MMAI) under additive white gaussian noise (AWGN) and fading channel

environments. The steady-state probability density function (PDF) and variance

of the phase estimator error and code tracking jitter is evaluated by solving the

first-order Fokker-Planck equation.

Among many linear multiuser detectors which decouple the multiple access

interference from each of the interfering users, one-shot window linear decorrelating

detector (LDD) based on a one bit period to reduce the complexity of the LDD

has attracted wide attention as an implementation scheme. Therefore, we propose

Hybrid Selection Diversity/Maximal Ratio Combining (Hybrid SD/MRC) one-shot

window linear decorrelating detector (LDD) for asynchronous DS-CDMA systems.

The selection diversity scheme at the input of the Hybrid SD/MRC LDD is based

on choosing the branch with the maximum signal-to-noise ratio (SNR) of all filter

outputs. The MR Combining scheme at the output of the Hybrid SD/MRC LDD

adopts to maximize the output SNR and thus compensates for the enhanced output

noise. The Hybrid SD/MRC one-shot LDD with PLL is introduced to track its

phase error and to improve the demodulation performance. The probability density
functions of the maximum SNR of the SD combiner , the near-far resistance (NFR)

of one-shot LDD by Gaussian approximation, and the maximum SNR of the MR

combiner for Hybrid SD/MRC LDD are evaluated, and the bit error probability is
obtained from these pdfs. The performance of Hybrid SD/MRC one-shot LDD is

assessed in a Rayleigh fading channel.

PHASE-LOCKED LOOP, DELAY-LOCKED LOOP, AND LINEAR
DECORRELATING DETECTOR FOR ASYNCHRONOUS
MULTIRATE DS-CDMA SYSTEM

by
Sok-kyu Lee

A Dissertation
Submitted to the Faculty of
New Jersey Institute of Technology
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Electrical Engineering

Department of Electrical and Computer Engineering

January 2001
 APPROVAL PAGE

Phase-Locked Loop, Delay-Locked Loop, and Linear

Decorrelaing Detector for Asynchronous Multirate
DS-CDMA System
Sok-Kyu Lee

Tapper, Dissertat on Advisor Date

sor, Department of Electrical and Computer Engineering, NJIT

Dr. Nirwan Ansari, Committee Member Date

Pr sor, Department of Electrical and Computer Engineering, NJIT

Dr. Joseph Frank, Committee Member Date

Associate Professor, Department of Electrical and Computer Engineering, NJIT

Dr. Yun-Qing Shi, Committee Member Date

Associate Professor, Department of Electrical and Computer Engineering, NJIT

Dr.Eliza(Z.-H.) Michalopoulou , Committee Member Date

Associate Professor, Department of Mathematical Sciences, NJIT
 BIOGRAPHICAL SKETCH

Author : Sok-kyu Lee

Degree : Doctor of Philosophy

Date : January 2001

Undergraduate and Graduate Education :

• Doctor of Philosophy in Electrical Engineering,

New Jersey Institute of Technology (NJIT), Newark, NJ, USA, 2001

• Master of Science in Electrical Engineering,

Polytechnic University, Brooklyn, New York, USA, 1996

• Bachelor of Science in Electronics Engineering,

Kwangwoon University, Seoul, Korea, 1985

Major : Electrical Engineering

iv
 To my father and mother
To my father-in-law and mother-in-law
To my wife, Eunhwa Roh
To my son, Brian Taehyun Lee
 ACKNOWLEDGMENT

I would like to express my deepest gratitude to my advisor, Dr. Jacob Klapper.

His advice, guidance and insight helped me enormously throughout this research.

My gratitude is extended to Dr. Nirwan Ansari, Dr. Joseph Frank, Dr. Yun-

Qing Shi, and Dr. Eliza Michalopoulou for serving as members on the dissertation

committee and for their comments.

I had the pleasure of working with my colleagues at Department of Electrical

Engineering, New Jersey Institute of Technology. Their help and suggestions are

appreciated and acknowledged.

Finally, I would like to sincerely thank my father, mother, brother, and sisters

for always being there when I most needed them. Also, I would like to thank my

father-in-law, mother-in-law, brothers-in-law, and my son, Brian Taehyun Lee for

their love and support.

I dedicate this thesis with my love to my wife, Eunhwa Roh.

vi
 TABLE OF CONTENTS

Chapter Page

1 INTRODUCTION 1

1.1 Wideband CDMA 1

1.2 Phase and Code Synchronization 4

1.2.1 Phase-locked Loop 7
1.2.2 Delay-locked Loop 9
1.3 One-shot Linear Decorrelating Detector 12

1.4 Fading Channel 13

1.5 Contribution of Dissertation 17

2 PHASE-LOCKED LOOP 19

2.1 Transmitter Channel Modelling 19

2.2 Phase-Locked Loop Receiver 21

2.3 Dynamic Partial Correlation (DPC) 24

2.3.1 Case of Pilot Interference 27

2.3.2 Case of Low-Rate Data Interference 27

2.3.3 Case of High-Rate Data Interference 29

2.3.4 Case of AWGN 31

3 DELAY-LOCKED LOOP 41

3.1 Coherent Delay-Locked Loop Receiver 41

3.2 Dynamic Partial Correlation (DPC) 44

3.2.1 Case for Pilot Interference R i p (m) 45

3.2.2 Case for Low-rate 74(m) 46

3.2.3 Case for High-rate RIph(m) 48

vii
 TABLE OF CONTENTS
(Continued)

Chapter Page

3.2.4 Case for AWGN Rn e (m) 50

3.3 System Analysis 51
3.4 Numerical Analysis 53
4 ONE-SHOT LINEAR DECORRELATING DETECTOR 61
4.1 Introdiction 61
4.2 Hybrid SD/MRC One-shot LDD 66
4.2.1 System Desciption 71
4.2.2 Carrier Phase Error Tracking by PLL 76
4.2.3 Bit Error Probability 78
4.2.4 Numerical Analysis 84
5 CONCLUSION 91
REFERENCES 93

viii
 LIST OF FIGURES

Figure Page

1.1 The baseband model of a DS-CDMA transceiver 2

1.2 Block Diagram of Conceptual Phase Locked Loop 7

1.3 Block Diagram of Conceptual Delay Locked Loop 10

1.4 Signal envelope of Rayleigh fading 13
1.5 PDF (Histogram) of envelope of Rayleigh fading 14

1.6 Signal envelope of Lognormal fading 15

1.7 PDF (Histogram) of envelope of Lognormal fading with zero mean 16

2.1 Transmitter Structure for u th user 19

2.2 Variable sequence length multirate spreading 20

2.3 Block Diagram of Phase-Locked Loop 21

2.4 An example of the DPC method in [a,,3] : i = 2, j = 6, k = 3 25

2.5 PDF of Phase Estimator Error under no fading channel 33

2.6 Variable of Phase Estimator Error under no fading channel 34

2.7 Variable of Phase Estimator Error without AWGN 35

2.8 PDF of Phase Estimator Error with Comparison of Unfaded and Faded
Channel Environment 36

2.9 Variable of Phase Estimator Error as a function of various users 37

2.10 Variance of Phase Estimator Error with Comparison of unfaded and faded
channel environment 38
2.11 PDF of Phase Estimator Error in imperfect power control (Log-Normal

fading) environment 39

2.12 Variable of Phase Estimator Error in imperfect power control environment 40

3.1 Block diagram of coherent dalay locked loop 41

ix
 LIST OF FIGURES
(Continued)

Figure Page

3.2 The variable of the tracking jitter as a function of the power gain 53
3.3 PDF of the tracking jitter when both data channels are active with pilot 54

3.4 The difference between the PDFs of the tracking jitter and Gaussian .. 55

3.5 The variance comparison of the low and high rate interferences with same
power gain 56

3.6 Bit error probability for chip synchronization in AWGN channel as a

function of SNR 57

3.7 Bit error probability for chip synchronization in AWGN channel as a

function of the number of users 58

3.8 Bit error probability for chip synchronization in Rayleigh fading channel

as a function of the Average SNR per bit 59

3.9 Bit error probability for chip synchronization in Rayleigh fading channel

as a function of the number of users 60

4.1 System model of multiuser detection 62

4.2 An example of noise power increment of LDD output 65

4.3 The comparison of Verdu's LDD and Hybrid SD/MRC LDD in asynchronous
CDMA systems 69

4.4 Hybrid SD/MRC one-shot LDD for first user 71

4.5 One-shot window for asynchronous CDMA 73

4.6 Hybrid SD/MRC One-shot LDD with PLL for first user 77

4.7 BER of LDD without SD/MRC and Hybrid SD/MRC LDD (L = 1) in

AWGN channel 85
 LIST OF FIGURES
(Continued)

Figure Page

4.8 Analytical BER at L = 2, 4, 8, and 16 in Rayleigh fading channel 86

4.9 BER of Conventional matched filter (Cony. MF) with SD (L=2, L=4),
LDD without SD/MRC, LDD with MRC scheme, and Hybrid
SD/MRC LDD with L = 2,4 in Rayleigh fading channel 87

4.10 bit error probability with comparison of analytical and simulated results
as a function of the number of users in Rayleigh fading channel
environment 88

4.11 pdf of the inversed value of diagonal component of R -1 for first user's
left version (N = 128, and 8„ = 0) 89
4.12 pdf of the phase estimator error in AWGN and Rayleigh fading channels 90

4.13 Comparison of BER between LDD with PLL and LDD without PLL in

AWGN and Rayleigh fading channels 90

xi
 CHAPTER 1

INTRODUCTION

1.1 Wideband CDMA

The origins of spread spectrum are in the military field and navigation systems.

Techniques developed to counteract intentional jamming have also proved suitable for

communication through dispersive channels in cellular applications. In this section

we highlight the milestones for CDMA development starting from the 1950s after

the invention of the Shannon theorem.

In 1949, John Pierce wrote a technical memorandum where he described a

multiplexing system in which a common medium carries coded signals that need not

be synchronized. This system can be classified as a time hopping spread spectrum

multiple access system. Claude Shannon and Robert Pierce introduced the basic

ideas of CDMA in 1949 by describing the interference averaging effect and the

graceful degradation of CDMA. In 1959, De Rosa-Rogoff proposed a direct sequence

spread spectrum system and introduced the processing gain equation and noise multi-

plexing ideas. In 1956, Price and Green filed for the antimultipath 'RAKE' patent.

Signals arriving over different propagation paths can be resolved by a wideband

spread spectrum signal and combined by the RAKE receiver. The near-far problem

(i.e, a high interference overwhelming a weaker spread spectrum signal) was first

mentioned in 1961 by Magnuski.

For cellular applications spread spectrum was suggested by Cooper and

Nettleton in 1978. During the 1980s Qualcomm investigated DS-CDMA techniques,

which finally led to the commercialization of cellular spread spectrum commu-

nications in the form of the narrowband CDMA IS-95 standard in July 1993.

Commercial operation of IS-95 systems started in 1996. Multiuser detection (MUD)

has been subject of extensive research since 1986 when Verdu formulated an optimum

1
 2

multiuser detection scheme for the additive white gaussian noise (AWGN) channel,

maximum likelihood sequence estimator (MLSE).

During the 1990s wideband CDMA techniques with a bandwidth of 5 MHz or

more have been studied intensively throughout the world, and several trial systems

have been built and tested. These include FRAMES Multiple Access (FRAMES

FMA2) in Europe, Core-A in Japan, the European/Japanese harmonized WCDMA

scheme, cdma2000 in the United States, and the Telecommunication Technology

Association I and II (TTA I and TTA II) schemes in Korea. Introduction of third-

generation wireless communication systems using wideband CDMA is expected

around the year 2000.

Figure 1.1 The baseband model of a DS-CDMA transceiver

A spread spectrum CDMA scheme is one in which the transmitted signal is

spread over a wide frequency band, much wider than the minimum bandwidth

required to transmit the information being sent. It employs a waveform that for

all purposes appears random to anyone but the intended receiver of the transmitter

waveform. Actually, for ease of both generation and synchronization by the receiver,

the waveform is pseudorandom, meaning that it can be generated by mathematically

 3

precise rules, but statistically it nearly satisfies the requirements of a truly random

sequence. In spread spectrum CDMA all users use the same bandwidth, but each

transmitter is assigned a distinct code. A block diagram of the baseband model of a

DS-CDMA modulator and demodulator is shown in Fig.1.1.

For more than a decade, research has been ongoing to find enabling techniques

to introduce multimedia capabilites into mobile communications. Research efforts

have been aligned with efforts in the International Telecommunication Union (ITU)

and other bodies to find standards and recommendations which ensure that mobile

communications of the future have access to multimedia capabilities and service

quality similar to the fixed network.

Compared to the second-generation CDMA, the following new main capabilities

characterize the third-generation CDMA:

• Wider bandwidth and chip rate

• Provision of multirate services

• Packet data

• Complex spreading

• A coherent Uplink using a user-dedicated pilot

• Multiuser detection

The third-generation mobile communication systems, called International

Mobile Telecommunications 2000 (IMT-2000) in the International Telecommuni-

cation Union (ITU) have been under intense research and development recently.

One of the main objectives for the IMT-2000 air interface is that the system will

be designed to support the multimedia services at data rates as high as 2Mb/s [5].

The most promising candidate is the code division multiple access (CDMA) system.

As is well known, each user in a CDMA system is assigned a unique code, called

pseudo-random (PN) codes. The PN codes must be generated at the receiver as well
 4

and must be synchronized to coincide perfectly with the timing of the received trans-

mission. The tracking process, which should be refined after synchronization, is the

focus and thus, we approach the code jitter tracking process with the delay-locked

loop (DLL) in this paper.

1.2 Phase and Code Synchronization

CDMA based on spread spectrum communication is an access method where all

the users are permitted to transmit simultaneously, operate at the same nominal

frequency, and use the entire system bandwidth. Recently, extensive investigations

for effective operation of multirate signals have been carried out into the application

of a code division multiple access (CDMA) system as an air interface multiple access

scheme for IMT-2000.

Wideband CDMA systems are designed to accommodate a higher data rate for

better multimedia service. Since the spreaded signal bandwidth is the same for all

users, the transmitter needs multiple spreading factors (SF) to support a variable

rate code in the physical channel. Variable Length Orthogonality (VLO) presented

in [8] is necessary for channelization and the Gold code is used as a PN code for

scrambling in the asynchronous system.

Fine synchronization is becoming increasingly important for the development

of advanced receivers for wideband DS-CDMA systems in which multiuser inter-

ference cancellation is employed to overcome the reduction of system capacity due

to this fundamental source of interference. Carrier phase synchronization of wireless

communication systems is an already well-developed technique after the considerable

research efforts provided in the past[3] [4]. It is evident that the performance of the

PLL is degraded in the presence of MMAI and AWGN. Performance is dependent on

the signal-to-noise and interference power ratio (SNIR) and the frequency difference

between the desired signal and the interference signal. Among the many behaviors
 5

which can explain the performance of the PLL, our main concern is the steady-state

PDF of the phase estimate error as a function of the number of the simultaneous

users.

Although various aspects of Multirate CDMA communication were discussed

in a number of publications, there were few analytical results on asynchronous PLL

systems and their applications. The need for considering the aperiodic crosscorre-

lation properties of the code sequence is shown in papers [10] [11] [12]. However,

they did not treat asynchronous multirate code sequences nor evaluate the

auto-correlations of cross-correlations and PSDs of the MMAI terms.

Our approach is first to derive the carrier phase estimate of the pilot signal

that allows the receiver to extract and to synchronize its local oscillator to the

carrier frequency and phase of the received signal. Therefore, we configure an

integrated structure of DLL and PLL for receiving the desired signal effectively and

simply since noncoherent code tracking performance will not be influenced directly

by the performance of the carrier phase synchronization. After receiving, we evaluate

gaussian noise and MMAI components using the Dynamic Partial Correlation (DPC)

method to solve the cross-correlation terms and the corresponding power spectral

densities of various interference terms in detail in an asynchronous multirate

DS-CDMA communications environment.

Since a linear analysis yields a large deviation between the analytical results and

actual performance in a low SNR environment [1], [2], here the effect of these sources

of noise on the PLL performance is performed based on a nonlinear model. We also

adopt a pilot signal for coherent demodulation. Thus, the steady-state PDF and

variance of the phase estimator error of the PLL under various fading environments

is performed by solving the first-order Fokker-Planck equation[4]. Finally, we will

compare the performance in faded environment with the performance in unfaded

environment.
 6

Delay lock tracking of PN codes for CDMA systems is an already well-developed

technique after the considerable research efforts provided in the past[3] [4]. The

general DLL process is performed by the so-called early-late gate(ELG) device.

Although its output S-curve is very useful to track the PN code phase of the desired

user, it is not perfect since several noise factors distort it. The noncoherent DLL

is the most popular technique. It suffers from tracking jitter due to the noise

enhancement arising from the square-law detector (squaring loss). The coherent

DLL overcomes this problem and improves the transmission performance. However,

it requires accurate channel estimation in the receiver, which is generally difficult to

do in fast fading. Fortunately, the chip synchronization process can take advantage

of the pilot channel in both the forward and the reverse link of the third-generation

systems.

In this paper we model the system that can be characterized by new capabilities

such as asynchronous operation, multirate service and pilot aided operation. In the

asynchronous system each user or cell site is assigned to a distinct Gold or Kasami

code for scrambling. To meet the requirment of multimedia services, variable length

orthogonal (VLO) code is necessary for channelization [7]. For coherent operation,

the channel estimator supports DLL. Therefore, various effects of noise and inter-

ference in our system will be distinguished from the result shown in publications as

[6], [10]-[12]. The scrambing PN code of different users or cell sites occurs to the DLL

as an interference. The amount of interference is represented by the cross-correlation

of codes. Since Pursley and Sarwate showed the cross-correlation properties of PN

codes for CDMA in [12], the result was extended to multirate CDMA [34]. As

they examined only the mean and the variance, we can't extend to evaluate the

performance of DLL and it will be mentioned in detail later. The interferences of the

output of the ELG device are the time function consisting of the cross-correlations.
 7

After we determine the autocorrelation and its PSD of the time functions, we evaluate

the performance via the PDF of the jitter.

1.2.1 Phase-locked Loop

Phase-Locked Loops are used widely in modern communication systems. There

are several kinds of phase-locked receivers on the market for a variety of appli-

cations. These applications include demodulation of information carrying signals,

synchronization, tracking, and ranging. Its application is more useful in low SNR

environments because of the improved performance in the threshold region. Because

of the importance of the application of the phase-locked loop (PLL), there has been

a tremendous amount of work done in this area. There are several books [1],[2] which

look into the analysis and design of the PLL in one way or the other.

Figure 1.2 Block Diagram of Conceptual Phase Locked Loop

A PLL is a device which continuously tries to track the phase of the incoming

signal. It is realized by a phase detector, a loop filter, and a voltage-controlled

oscillator (VCO). The conceptual configuration is shown in Fig.1.2.

 8

The basic operation requirement for the loop is to track the phase of the input

signal. This is accomplished by forming in the phase detector a voltage proportional

to the phase difference between the input signal and VCO signal. This error signal

is then amplified, filtered to remove excess noise and other undesirable components,

and applied to the VCO to effect input signal phase tracking. Consider an input

signal of form

where A is the peak signal amplitude (a constant), w i is the input signal center

frequency, and øi is the input signal phase. Assume that the VCO output signal is

of the form

where ø(t) is the VCO phase estimate. A particular form of VCO signal is chosen,

with the main feature being the quadrature relationship between the input and the

VCO output. A secondary feature is the 2/A amplitude, chosen solely as a matter

of convenience. Both characteristics stem from the multiplier form of the phase

detector, as will be apparent presently.

Let the phase detector be characterized as an ideal multiplier which, in turn,

is an idealization of the doubly-balanced mixer often used as a phase detector in

practice. The result is an output of

If we neglect the upper sideband component (the first term in (1.3)), or assume

that it is subsquently rejected by the loop filter, the filter output becomes

where K2 is the baseband amplifier gain, f(t) is the impulsive response of the filter,
 9

and ® denotes convolution. This baseband voltage is applied to the VCO, and

results in an FM signal whose frequency deviation is directly propotional to the

control voltage,

with K = K2 K3 , and K3 being the VCO sensitivity.

1.2.2 Delay-Locked Loop

Once receiver timing has been synchronized to within a fraction of a chip time, the

estimate should be further refined to approach zero. Furthermore, because of the

relative motion of transmitter and receiver and the instability of clocks, corrections

must be made continuously. This process, which is called tracking, is performed by

the so-called "early-late" gate device[15]. It is an elabolation of the basic demodulator

and acquisition device.

The function of a baseband DLL is to track the time-varying phase of the

received spreading waveform C(t — . The function will denote the receiver

estimate of τ(t) and τ(t) and τ(t) are always functions of time, whether or not

this dependence is written explicity. The received signal consists of the spreading

waveform C(t — 'T) and additive white gaussian noise. That is,

Fig.1.3 is a conceptual block diagram of the tracking loop. It consists of a

phase discriminator, a loop filter, a voltage controlled oscillator, and a code waveform

generator

Input signal r(t) is cross-correlated with the advanced version

and retarded version C(t — T — (0/2)T) of the local PN code generator sequence.

Parameter A is the total normalized time difference between the advanced and

retarded discriminator channels.

 Figure 1.3 Block Diagram of Conceptual Delay Locked Loop

Consider the operation of the delay-lock discriminator as a static phase-

measuring device in a noiseless environment. That is, let T and ii be fixed and
-

determine the output of the discriminator. This output will contain a component

which is a function of 6 = (7 - — f )/ T, and is suitable for driving the VCC just as

-

the phase-locked-loop multiplier output contained a component sin(0 — 0 0 ). In the

static case, it is convenient to write x 1 , x 2 , and ε(t, 5) as explicit functions of 7 , f ,

- -

and t. Thus, the advanced-correlator output is

and the retarded-correlator output is

 11

The delay-locked discriminator output is the difference of x 1 (t) and x 2 (t) and

is

The dc component of this signal is used for code tracking. The dc component

of E(t, 7, 'T) is denoted D o (T, f) and the time average of E(t, 'r, f ). Thus
-

where NT, is the period of C(t). The autocorrelation function of C(t) is

Assuming that code self-noise can be ignored, discriminator output can be

written

where

The filter output is the convolution of the input signal and the impulse response.

That is,
 12

Finally, considering the VCC, the nonlinear integral equation representing the

operation of the tracking loop of Fig. 1.3 is

1.3 One-shot Linear Decorrelating Detector

There has been great interest in improving DS-CDMA detection through the use of

multiuser detectors. In multiuser detection, code amplitude and phase information

of multiple users are jointly used to better detect each indivisual user. The important

assumption is that the codes of multiple users are known to the receiver a priori.

Verdú's seminal work [30] proposed and analyzed the optimum multiuser

detector, otherwise known as the maximum likelihood sequence detector. Unfor-

tunately, this detector is much too complex for practical DS-CDMA systems.

Therefore, over the last decade or so, most of the research has focused on finding

suboptimal multiuser detector solutions which are more feasible to implement.

Most of the proposed detectors can be classified in one of two categories: linear

multiuser detectors and subtractive interference cancellation detectors. In linear

multiuser detection, a linear mapping (transformation) is applied to the soft outputs,

which hopefully provides better performance.

Among many suboptimal detectors proposed, Lupas and Verdu's linear decor-

relating detector (LDD) [28],[29] has attracted wide attention. The LDD achieves the

same near-far resistance as the optimum detector while its complexity is linear in the

number of users. Although the LDD is much simpler than Verdu's optimum detector,

it is still too complicated and will lead to unacceptable detection delay while dealing

with asynchronous CDMA. Therefore, another scheme, also proposed by Lupas [33],

called one-shot LDD, where the detection of data is based on the signal observed in
 13

one bit length period, the detection of an asynchronous CDMA is transformed into

that of synchronous CDMA. One-shot LDD is synchronized with one of the users,

so for a U-user asynchronous CDMA, a (2U-1)-user synchronous CDMA problem

must be solved for the detection of one user, and a total of U different (2U-1)-user

synchronous CDMA problems must be solved for the detection of the total U users.

1.4 Fading Channel

The mobile radio channel places fundamental limitations on the performance of

wireless communication systems. The transmission path between the transmitter

and the receiver can vary from simple line-of-sight to one that is severely obstructed

by building, mountains, and foliage.

Figure 1.4 Signal envelope of Rayleigh fading

Unlike wired channels that are stationary and predictable, radio channels are

extremely random and do not offer easy analysis. Even the speed of motion impacts
 14

how rapidly the signal level fades as a mobile terminal moves in space. Modeling the

radio channel has historically been one of the most difficult parts on measurements

made specifically for an intended communication system.

Figure 1.5 PDF (Histogram) of envelope of Rayleigh fading

Propagation models that predict the mean signal strength for an arbitrary

transmitter-receiver separation distance are useful in estimating the radio coverage

area of a transmitter and are called large-scale propagation models, since they charac-

terize signal strength over large Tx-Rx separation distances (several hundreds or

thousands of meters). On the other hand, propagation models that characterize the

rapid fluctuations of the received signal strength over very short travel distances

(a few wavelengths) or short time durations (on the order of seconds) are called

small-scale or fading models.

From now on, we will consider only small-scale fading models ignoring the

distance. In the fading model, depending on how rapidly the transmitted baseband
 15

signal changes as compared to the rate of change of the channel, a channel may be

classified either as a fast fading or slow fading channel. In the fast fading channel, the

channel impulse response changes rapidly within the symbol duration. That is, the

coherence time of the channel is smaller than the symbol period of the transmitted

signal. On the other hand, a slow fading channel is one whose impulse response

changes at a rate much slower than the transmitted baseband signal.

Figure 1.6 Signal envelope of Lognormal fading

In mobile radio channels, the Rayleigh distribution is commonly used to

describe the statistical time varying nature of the received envelope of a flat fading

signal, or the envelope of an individual multipath component. When the received

signal is made up of multiple reflective rays plus a significant line-of-sight (no faded)

component, the envelope amplitude due to small-scale fading has a Rician pdf

containing a specular component. As the amplitude of the specular component

approaches zero, the Rician pdf approaches a Rayleigh pdf, expressed as

 16

where r is the envelope amplitude of the received signal, and σ2 is the predetection

mean power of the multipath signal. The Rayleigh faded component is sometimes

called the random, or scatter, or diffuse component. The Rayleigh pdf results from

having no specular component of the signal; thus, for a single link it represents the

pdf associated with the worst case of fading per mean received signal power.

Figure 1.7 PDF (Histogram) of envelope of Lognormal fading with zero mean

Clarke developed a model where the statistical characteristics of the electro-

magnetic fields of the received signal at the mobile are deduced from scattering [9].

Fig.1.4 shows the envelope of Rayleigh faded signal as a function of time from the

Clarke simulation method.

 17

Fig.1.5 illustrates the probability density function (PDF) of the Rayleigh faded

envelope as a function of the envelopes in multipath channel environment.

In power control environment, if the received signals are imperfectly power

controlled, the signal powers are log-normally distributed [3]. Fig.1.6 shows the

envelope of Lognormal faded signal as a function of time and Fig.1.7 shows the pdf

of the Lognormal faded envelope as a function of the envelopes.

1.5 Contribution of Dissertation

• Dynamic Partial Correlation (DPC)

- Evaluation of autocorrelation and its power spectrum density function of

cross-correlated terms in asynchronous multirate DS-CDMA systems.

- Application for asynchronous multirate multiple access interference.

• Phase-Locked Loop (PLL)

- Integration of PLL and DLL to operate efficiently.

- Performance analysis via the PDF and variance of phase estimator error

using DPC method in asynchronous multirate DS-CDMA system under

AWGN, Log-Normal, and Rayleigh fading channel environments.

- Combination of PLL and one-shot CLDD to improve the performance of

multiuser detection.

• Delay-Locked Loop (DLL)

- Integration of PLL and DLL to operate efficiently.

- Performance analysis by the PDF and variance of the delay jitter

of PN code signature waveform using DPC method in asynchronous

multirate DS-CDMA system under AWGN and Rayleigh fading channel

environments.
 18

• Linear Decorrelating Detector (LDD)

- Propose the Hybrid SD/MRC one-shot linear decorrelating detector.

- Adopt the selection diversity (SD) combining scheme to compensate the

deep faded signal by the selection of the largest instantaneous SNR from

the SD combiner.

- Adopt the maximal ratio (MR) Combining scheme to maximize the output

SNR for compensating the reduction of signature waveform energy of

conventional LDD.

- Apply the Phase-Locked Loop to the Hybrid SD/MRC one-shot LDD for

fine synchronization of carrier phase error in asynchronous DS-CDMA

system.

- Evaluate the pdfs of the maximum SNR at the SD combiner output, the

near-far resistance (NFR) of LDD, and the maximum SNR at the Hybrid

SD/MRC one-shot LDD output.

- Obtain the bit error probability by using the pdfs, analyze the performance

of Hybrid SD/MRC LDD, and compare with conventional LDD in

Rayleigh fading channel.

 CHAPTER 2

PHASE-LOCKED LOOP

2.1 Transmitter Channel Modelling

Figure 2.1 Transmitter Structure for uth user

The transmitted signal of the u th user is modelled by multiplying s u (t) with a

carrier cos w e t. The signal s u (t) with the pilot and multirate data is expressed by

°
where the pilot signal and its channelization code sequence W is assumed to be all

-1 [5] and Gr is the power gain and r = l, h denote low-rate and high-rate signals,

respectively. Defining a rectangular pulse by h T (t) = 1 for 0 <= t <= T and hT (t) =

0, otherwise, the data sequence with r rate, dru (t), can be written as

with consecutive data bits {dru (n)} taking on values of +1 with equal probability.

Since the two different data channels should share the same channel bandwidth,

multiple spreading factors (SF) are required [13].

19
 20

For variable sequence length synchronization, we shall denote the bit interval of

the low-rate data as T1 and the number of bits of channelization code corresponding

to T1 as N1 . The high-rate data will be transmitted data with bit intervals of duration

Th and channelization code length Nh. The channelization code waveform, Wr(t),

for the r channel can be written as

where T„, = Tl/Nl = Th /Nh and Wr(n) is a periodic sequence with length N, chosen

in VLO code set.

Figure 2.2 Variable sequence length multirate spreading

As shown, for example, in Fig.2.2, each bit of the low rate data dul is spread by a

code of length N1 = 2'2 , where n is an abitrary integer. Since the bit duration of high

rate data Th is half the duration of a bit in the low rate T1 , we need a spreading code

of length Nh = N1 /2 = 2 n-1 for spreading. Orthogonality between different rates

and SFs is necessary for multirate multiuser detection to be accomplished. Thus the

channelization code waveform WV) should be orthogonal to

where the ratio is an integer.

 21

The scrambling PN code waveform, Cu (t), for u th user can be written as

where Cu (n)is a periodic sequence with length N chosen in Gold or Kasami code

set. If we assume that the length of the PN code is the same as TI , then T1 = NT,.

From now on, we assume that

the ratios N/NI and N/NhΩ are integers.

2.2 Phase-Locked Loop Receiver

Fig.2.3 illustrates a receiver configuration for reception of the pilot signal in

asynchronous multirate DS-CDMA systems.

Figure 2.3 Block Diagram of Phase-Locked Loop

 22

We assume that a channel delays the signals transmitted through it and

corrupts them by adding gaussian noise. Hence, the incoming signal r(t) to the

integrated receiver of DLL and PLL in Fig.2.3 can be expressed as

where A(t) and ø u(t) are the received envelope and the carrier phase delay due to

the channel, respectively, U indicates the number of users, 'n u is the code phase delay,

and n(t) is additive white gaussian noise having a zero mean and a one-sided PSD

Without loss of generality, we will assume that the signal to be tracked belongs

to the pilot signal of the first user. The locally generated code and phase sequence

of user one with pilot signal is represented by

where Tl and .. 1 (t) are the estimates of the code and phase delay of the local oscillator

for user one determined by the DLL and PLL estimator signal, respectively. When

the one-sided loop bandwidth B L is much less than the PN code chip rate 1/Tc , the

effect of the PN code self-noise on the loop performance can be neglected [15].

Assuming that the PN code of the desired user was tracked well, T i = 0 and

the output of the Integrate-and-Dump operations after multiplying the input r(t) by

the estimate of the local oscillator, VLO(t), X can be evaluated as

 23

terms are as follows :

The error signal to the loop filter is multiplied by the term of the estimated

pilot envelope with all l's as follows:

From Fig.2.3, the instantaneous phase delay estimate of the PN code generator

output, ø1, is related to e(0) by

where Kp is an overall closed loop gain including the gain of the voltage controlled

clock which drives the carrier generator. The coherent integro-differential equation

of the operation of Fig.2.3 with p = d/dt becomes

where
 24

2.3 Dynamic Partial Correlation (DPC)

We will now find the probability density function (PDF) as a function of the number

of simultaneous users for investigating the behavior of the phase estimate error. We

have to evaluate the power spectrum density (PSD) of the corresponding MMAI and

AWGN terms prior to finding the PDF, p(0).

However, the conventional continuous partial correlation method cannot

evaluate the autocorrelations and corresponding PSDs of the cross-correlation

terms with asynchronous multirate signals clearly since the method cannot treat

with a multirate signal. Therefore, we will introduce a new method called Dynamic

Partial Correlation (DPC) to solve the autocorrelations and corresponding PSDs

of cross-correlation terms in any specific time duration in NT, for asynchronous

multirate signals.'

The crosscorrelation between two codes in a specific interval [α,β] is defined as

represented by two rectangular pulse trains as

where the magnitudes pru(l) and qur (l) taking on values of +1 can be found by multi-

plying a code bit of the desired user and a delayed code bit of the 0 1 user and

the channelization code as

vhere zr is an integer which satisfies the equation (N — 1—

'Sections 2 and 3 with DPC method did with Hunkee Kim and D. S. Phatak at Dept.
of Electrical Engineerong of Binghamton, SUNY.
 25

and rectangular pulse defined as Π(a,b) = 1

for a < t < b and 11(a, b) = 0, otherwise.

Figure 2.4 An example of the DPC method in [a, /3] : i = 2, j = 6, k = 3

 26

Thus, the summations of

where B(a,b) = 1 for a> b and B(a,b) = 0, otherwise. Therefore, equation (2.19)

can be solved in terms of a and 3 as

where the three constant terms are as follows :

Eqns. (2.25)-(2.27) have the constant terms and first order terms for a

and O. Therefore, the PSD is simplified by using the definition of

 27

2.3.1 Case of Pilot Interference

The sequences of the channelization code with pilot signal are all l's. Therefore,

is the average received signal power.

The PSD of R IP (m) can be evaluated as

where

with the constant term Jp = Mu(0,171,G).

2.3.2 Case of Low-Rate Data Interference

Next, we will find the PSD of R. /67; (m) due to multiple access interferences with

different data rate. The autocorrelation function of the low-rate MAI is represented

by DPC terms as

where
 28

The low-rate continuous partial

cross-correlation can be represented as

subscript, -1,0, and 1 denote the last previous data bit, the present data bit, and the

first next data bit, respectively.

Therefore, the PSD of Rldi (m), SiL(f) can be evaluated as

where
 29

2.3.3 Case of High-Rate Data Interference

The autocorrelation function of the high-rate MAI is represented by DPC terms as

where

where 6 constant terms, Js, not depending on m are following with 71 = 1, 2, 3 and

Then the high-rate continuous partial cross-correlation is

 30

The PSD of the high-rate term can be represented as

where

with
2.3.4 Case of AWGN

Assuming that the AWGN has zero mean and that the in-phase term and quadrature-

terms are independent, and ignoring parameters(, k, u), the autocorrelation of

AWGN is represented by

2.4 System Analysis

The solution of a nonlinear stochastic differential equations such as (2.17) for the

steady-state probability density function (PDF), p(0), may be accomplished via the

Fokker-Planck method [4], [11], [14], [17]. From here on, we will use 0 = 0 1 for
convenience and simplicity. To illustrate the theory, we shall consider the case of

a first-order loop (i.e., F(s) = 1) and also assume d(ø1)/dt = 0. The PDF, p(0),
 32

satisfies the stationary equation :

where

From (4.28), using the well-known method of [1],[4], [22], we get the PDF of

the phase estimator error as

where /0 0 is the modified Bessel function of 0th-order. For v, it can be shown by

using the aymptotic formula I0 (v) ,c:-., exp(v) / V2R v that p(0) approaches a Gaussian
-

density function with zero mean and variance o ?.. Parameter v represents the loop
-

signal-to-noise and interference power ratio (SNIR), defined by

where the loop bandwidth B L = KpT l / 4 for the loop transfer function

By using the PDF of the phase estimator error given by expression (3.42), the

variance of the phase estimator error can be evaluated using

The PDF of the phase estimator error can be approximated by a Gaussian

distribution for a large equivalent loop SNIR, and the variance becomes
 33

2.5 Numerical Analysis

We have evaluated in detail the PDF of the steady-state phase estimator error of a

phase-locked loop in a multirate multiple access interference (MMAI) and AWGN

environment. The transmitter has three different channels which can be separated

by channelization codes. The transmitter signal for each user is despreaded by Gold

code with length N=128. At first, the PDF of the steady-state phase estimator

error of a phase-locked loop in a multirate multiple access interference (MMAI) and

AWGN environment without fading channel is investigated. Next, the PDF and

variance under no power controlled slow fading channel environment is shown in

various aspects.

Figure 2.5 PDF of Phase Estimator Error under no fading channel

 34

Fig.2.5 illustrates the behavior of the phase estimator error PDF, p(0), as a

function of the number of simultaneous users U, power gain of low-rate data, and

power gain of high-rate data. The peak height of the PDF clearly increases with

a reduction of the power gain of low-rate and high-rate data (in other words, the

tracking ability of the phase-locked loop is degraded), and decreases as the number of

simutaneous users increases. As shown in Fig.2.5, if the power gain of the multirate

data signal increases by approximately the power of the pilot signal, the variance of

the phase estimator error decreases rapidly; therefore, a tradeoff exists between the

power limitations imposed by the pilot signal, low-rate data signal, and high-rate

data signal.

Figure 2.6 Variance of Phase Estimator Error under no fading channel

 35

Fig.2.6 shows the variance of the phase estimator error as a function of the

number of simultaneous users with the power gain of multirate signals as a parameter.

Note that the variance increases with a reduction of the power gain of multirate data

signals. The signal-to-AWGN ratio is fixed at 6dB. For example, when the number

of users is U = 60, the variance of phase tracking error for G I = 0.3 and Gh = 0.5 is
required about 0.125 while the variance for G 1 = 0.1 and Gh = 0.2 is required 0.09.
This means that the larger the power of multirate signal, the larger the variance and

thus the PLL can not track the phase error of the pilot signal exactly.

Figure 2.7 PDF of Phase Estimator Error without AWGN

Fig.2.7 shows the behavior of the phase estimator error PDF, p(0), as a function

of the number of simutaneous users U under no power controlled fading environment.

 36

The peak height of pdf decreases as the number of users increases. For example, the

peak height for 20 active users is around o.6 much higher than that for 60 active

users at the fixed multirate signal powers G 1 = Gh = 0.1. The tracking ability of
PLL may be degraded quickly with the increase of the number of active users.

Figure 2.8 PDF of Phase Estimator Error with Comparison of Unfaded and Faded
Channel Environment

Fig.2.8 shows the PDFs of phase estimator error in unfaded channel and faded

channel environments. As expected, it shows that the tracking performance of the

phase-locked loop in faded channel is worse than that in no fading channel for no

power controlled environment.

Fig.2.9 illustrates the variance of phase estimator error as a function of number

of simultaneous users and AWGN. The variance increases as the number of the users

increases, but is not influenced greatly by AWGN. When the variance for 5 active
 37

users is 0.035 at SNR 10 dB, the variance for 80 active users is 0.35 at the same SNR

and for 40 active users is 0.75. The number of users effects the tracking ability of

PLL since the multiple access interference increases as the number of users increases.

Figure 2.9 Variance of Phase Estimator Error as a function of various users

Fig.2.10 shows that the difference of the tracking performance of phase

estimator error between fading and unfading channel environment increases with an

increase of the number of simultaneous users. And the variance of phase estimator

error decreases as signal-to-AWGN+MAI ratio(SNIR) increases with a fixed power

gain of multi-rate signal.

Since the performance of Rayleigh faded received signals are degraded rapidly,

adaptation of power control system can mitigate near-far effect and improve the SNIR

performance. Assuming the power control of real system is imperfect, the power

controlled received signal with fading channel has log-normally distributed envelope
 38

under the variance 2-3[dB] in log scale. As see in the Fig.2.10, the variances of all

the channels do not change above the SNR of 6 dB. This means that the multiple

access interference and faded signal effect much more than the AWGN channel. For

example, the variances for all the cases are 0.77, 0.38, 0.16, and 0.09, respectively,

at the SNR 6 dB, the variances did not more decrease with the reduction of SNR.

Figure 2.10 Variance of Phase Estimator Error with comparison of unfaded and
faded channel environments

Fig.2.11 illustrates the behavior of the phase estimator error PDF, p(0), as

a function of the number of simultaneous users U. The performances of PDFs for

Rayleigh and Log-normal faded environment cannot be compared directly. But, as

seen in the simulated results, the Log-normal faded environment with variance 2-3

[dB] almost equals to Rayleigh faded environment while the received signal average

powers are same. If the power of multirate data increases, the peak height of pdf will
 39

be decreases deeply since the multirate data channels are also invoking the multiple

access interference with oche user.

Figure 2.11 PDF of Phase Estimator Error in imperfect power control (Log-Normal
fading) environment

Finally, Fig.2.12 shows the variance of phase estimator error in imperfect power

controlled environment as a function of the number of simultaneous users and the

variance of the imperfect power controlled received envelope. The performance of

variance does not effect greatly the performance, especially, for a lot of active users.

For example, when the variances for 10 active users with the variance of the fading

envelopes of 2 and 3 dB at an SNR 10 dB, are 0.22 and 0.35, respectively, the

variances for 40 active users with the fading envelopes 2 and 3 dB, are 0.16 and 0.17,

respectively. These results are almost the same for 80 active users.
 40

Fine synchronization is becoming increasingly important for the development

of advanced receivers for wideband DS-CDMA systems in which multiuser inter-

ference cancellation is employed to overcome the reduction of system capacity due to

this fundamental source of interference. In the numerical analysis, the performance

was, as expected, degraded when there is a phase delay error of the received signal

in wideband CDMA systems, the PLL was used to mitigate the degradation of

performance of receiver in AWGN and Rayleigh fading channels, and we confirmed

that the performance of receiver with PLL improved the pdf and variance of phase

estimator error.

Figure 2.12 Variance of Phase Estimator Error in imperfect power control

environment
 CHAPTER 3

DELAY-LOCKED LOOP

3.1 Coherent Delay-Locked Loop Receiver

The coherent DLL receiver consists of the ELG for spread PN code tracking and the

channel estimator for coherent tracking and several filters [5]. The block diagram of

the coherent delay locked loop is shown in Fig.3.1.

Figure 3.1 Block diagram of coherent delay locked loop

Without loss of generality, we will assume that the signal to be tracked belongs

to the first user, represented as u = 1. Therefore the function of a baseband DLL is

to track the time-varying delay of the untracked PN sequence C 1(t-τ 1) of a desired

41
 42

user. Since the tracking begins after successful code acquisition, the code delay of a

desired user is assumed as —Tc/2 < 7 -1 < Tc/2.

The incoming baseband signal r(t) to the coherent DLL in Fig.3.1, can be

modelled as

where a u is the attenuation factor due to the fading, τu is the overall time delay, and

øu is the path phase delay of the u th user received signal. Because of the asynchronous

system, u th user's time delays u except u = 1 can be assumed as 0 <= u < NT,. We

now assume n(t) in (3.1) as a white Gaussian process with one sided PSD N 0 /2.

We consider a slow Rayleigh fading channel. If the channel impulse response

changes at a rate much slower than the transmitted baseband signal, the channel

may be assumed to be static over one or several reciprocal bandwidth intervals. In a

slow fading channel, the factor a and corresponding phase delay ø can be considered

as zero-mean complex gaussian random variables having a variance o -2 each so that

the a is rayleigh distributed with the variance o -2 = (2 — 7/2)o-2 and ø is uniformly

distributed. Channel estimator estimates a and ø by using the demodulated pilot

sequence.

The locally generated scrambling PN code waveform of user one is C 1 (t —

where Tl is the estimate of the code delay for user one. The received signal is the

input to the ELG where it is correlated with an early code waveform

and a late code waveform C 1 (t — Tl — A). When the single-sided loop bandwidth BL

is much less than the PN code chip rate 1/T,, then the effect of the PN code self-noise

on loop performance can be neglected [15]. The output of the Integrate-and-Dump

operations, X± can be expressed as

where ±L denotes one half of the length (in seconds) of a PN code chip,
 43

The error signal to the loop filter is multiplied by the estimated fading envelope

'd i and the estimated conjugate phase e - j(7'' in the following way:

where D1^P(0, γ1) is the normalized desired output of the ELG S-curve and it is

multiplied by ξ1=αuα1 and it also distorted by the total interference /(t), and

we assume perfect channel estimation. Total interference N(t) is

where the interference terms are introduced in detail in (3.5)-(3.11).

The term due to the pilot channel interference is

where U is the number of total active users and

and the code cross-correlation

Next I^l(t) and I^h(t), terms due to the data channels, are expressed by

where Ur is the number of the active users of the r rate data channel in total users

and
 44

with the multirate code cross-correlation having the data and channelization

waveforms as

From Fig.3.1, the instantaneous delay estimate of the PN code generator

output, VCC, is related to e(t) by

where f (t) is the loop filter transfer function in the time domain and K v is the gain

of the voltage controlled clock in Hz/V which drives the PN code generator. From

the definition E u Tc =τu-τ1, the normalized code tracking jitter E l for the desired

user satisfies 1611 < 1 and is expressed by

By using (3.3),(3.12), equation (3.13) can be rewritten

where * denotes the convolution operation.

3.2 Dynamic Partial Correlation (DPC)

Next, we will evaluate the PSD in (3.14) by its autocorrelation functions. Since all

noise sources are pairwise uncorrelated with respect to one another, the autocorre-

lation of N(t) is represented separately and its PSD is

 45

with

We have assumed that the phase difference 0„ = is uniformly distributed

on the interval [0, 271], τu is uniformly distributed on the interval [0, NTC ], and a u is

Rayleigh distributed. Also, the data symbols dru are asumed to take on values +1 or

-1 with equal probability.

3.2.1 Case for pilot interference R i p (m)

Because the combination of data sequence and channelization code of pilot channel

is all l's, partial correlation between desired code and the other code for the pilot

channel is not needed. Therefore, R i p (m) does not depend on m, i.e, the shifted

versions by m are same since all PN codes are periodic with Tl. Φu^P is a special case

of (2.19) with a = 0, 13 = T1 , and d(t)W(t) = 1.

Let the phase difference be = kTc pu(t) and qu (t) taking on values of

±1, are the same as (2.20) and (2.21) except that the definition

Therefore the crosscorrelation between two pilot PN codes in a region [0, Ti ] is

defined as
 46

Thus, the autocorrelation of I^ P(t) is

in the next cases since it is one. The PDF and properties

of u will be discussed in the numerical analysis.

The PSD of R i p is

3.2.2 Case for low-rate RI^l (m)

From now, for simplicity, let the DPC function

common factor and we define the operation

Mark + means that the denoted terms are including +A. Therefore the +

term and the — term are distinghished from each other.

The autocorrelation function of the low-rate MAI is represented by the DPC

terms as

where
 47

with 4 constant terms. Js. not depending on m given by

In (3.22), the continuous partial crosscorrelation derived from (3.10)

and (2.19) as

the subscripts, -1,0, and 1 indicate the turn of entry. As shown in (3.22) and (3.24),

data bits at that time are decided by the position of m. That is, if m is less than ' y1 ,
-

then dlu(-1) and du^l(0) are used and if m is greater than '-y, then d u^l(0) and d u/ (1)

are used. To find these two regions of m is easy for the low-rate case since one low

data bit matches to one PN code with length Ti. In the case of high-rate, the regions

of m are increased relative to ratio Q.

After changing all DPC functions M,/, in (3.22), if a Fourier transform is taken

by m, the PSD of I^l (t) is

with
 48

(3.27)

3.2.3 Case for high-rate RI^h(m)

The autocorrelation function of the high-rate MAI is represented as

As is mentioned in the low-rate case, while the number of region of m for high-

rate is relative to the ratio Ω, the region is decided by γu where

Thus, the region varies according to 7„ and 'a' in (3.29) is a indicator

with {0, 1, . ..,QTtyhe—n,1'a}is<d=c regon.Fxampl,if'
-

zero, if Th < γu= < 2Th then 'a' is one, and so forth. Additionally we need 'b' to
represent middle regions except for the first and last region. It is a integer with

values {0, 1, .. , Q — 2}.

 49

In (3.29) 9 constant terms. Js. not depending on m are the following with

where the high-rate continuous partial crosscorrelation is :

 50

Similar to the low-rate case, the PSD of the high-rate term can be represented

as

with

where the constant terms are the following with n = 1, 2,3 and

3.2.4 Case for AWGN Rne (m)

As we assumed, since n (t) is an AWGN process with zero mean and PSD N0 /2 ,the

autocorrelation of AWGN is represented by

ignored. Since we assumed A = T c/2, the phase difference between C1(t—T— A) and

moment of the a defined by two complex gaussian random variables.

Hence, it's PSD is

3.3 System Analysis

The solution of nonlinear stochastic differential equations such as (3.14) for the

steady-state probability density function (PDF) p(E) may be accomplished via the

Fokker-Planck method [4], [11], [14], [17]. To illustrate the theory, we shall consider

the case of a first-order loop (i.e., F(s) = 1) and also assume d(τ1)/dt = 0 and a

negligible recovered carrier phase error = 0. Notice DP (0, 7 1 ) in (3.3) can be repre-

The PDF p(E) satisfies the stationary equation:

 52

where A(E) and B(6) can be found from (3.14) [4].

where SN (f) is the PSD of the total noise defined in (3.4).

The solution to (4.28) takes the well-known form [11], [17] of

where the normalization constant is

Finally, the PDF and the mean-squared tracking jitter variance over the code

can be found as
 53

3.4 Numerical Analysis

We now evaluate in detail the PDF of the steady-state tracking jitter of the delay-

locked loop in a Multirate Multiple Access Interference(MMAI) environment. The

transmitter has three distinct channels that can be seperated by channelization codes.

Since the high rate data is set to be 4 times faster than the low one, W ^l is both 4

times longer than W h and orthogonal to the set {W h }. Each transmitter for an

individual user (or cell site) is spread by Gold codes with length 128.

Figure 3.2 The variance of the tracking jitter as a function of the power gain.

Figure 3.2 shows the variance of the tracking jitter as a function of power gain

of the data channels. Power gain of the pilot channel is set to 1 and all other power

gains are normalized to it. Though the DLL tracks the PN code, actually it is the

pilot channel that is tracked since the pilot channel does not have any signal except

the PN waveform. Therefore, the two data channels affect the DLL performance
 54

as interference. The variance of the tracking error for faded channel is higher than

that for unfaded channel. This means that the error tracking ability of DLL for

faded channel is degraded in contrast to that for the unfaded channel. The variance

difference bewteen the faded and the unfaded channel is very close since the power

gain of the data channel is much higher than that of pilot and thus the unfaded

channel has the effect of much more active users.

Figure 3.3 PDF of the tracking jitter when both data channels are active with pilot.

Seeing Figure 3.2 again, when the two data channels are not active i.e., G=0
and Gh = 0, the variance is about 0.045. If we set G ^l and Gh to 0.5 and 1 respectively,
the amount of variance is increased about 0.025. It means that two data channels

cause the performance to be decreased by 1.8dB. Thus, the choice of the power

gains is important since a tradeoff exists between the noise effect of the data channel

and power efficiency. Since the high-rate signal should be transmitted with a larger
 55

power than the low-rate signal to maintain their performance generally in a practical

system, we used a Gh that is two times G ^l. The calculated PDF of the code phase

jitter given by (3.42) is now compared for different situation.

Figure 3.4 The difference between the PDFs of the tracking jitter and Gaussian

Figure 3.3 shows the PDFs for different number of users when all data channels

are active. The peak height of the PDF is decreased exponentially as the number of

active channels increases. It means that the capacity of a channel decreases rapidly

and thus the performances between 40 and 80 channels are closer than 10 and 20

channels. The shape of the PDF looks Gaussian but is not. The absolute sum of

the difference between PDFs of the tracking jitter and gaussian is shown with the

variance in Figure 3.4. The difference decreases rapidly but it does not go to zero

since we assumed the code phase delay was limited in Tc and the jitters happen either
 56

Figure 3.5 The variance comparison of the low and high rate interferences with
same power gain

very rarely or not outside of 3T/2. Consquently, the variance of tracking error of

DLL increases as the number of users increases and thus, the channel capacity is

decreased and the performance of demodulation is degraded.

Next we fixed power gains to 1 and we varied the number of active channels.

Two cases are compared : when the high data channel is not active and when the

low data channel is not active, we seek to determine which data channel effects the

performance as a worse interference. Figure 3.5 shows the worse interference is a low

data channel. The reason can be found in the channelization code since the two data

channels use the same scrambling code. While the channelization code is short and
 57

periodic for the high rate, the one for the low rate is long as much as NT,. Thus, it
makes the scrambling PN code to more distorted.

Figure 3.6 Bit error probability for chip synchronization in AWGN channel as a
function of SNR

Figure 3.6 shows the bit error probability for chip synchronization in AWGN
channel as a function of the SNR. As explained for Figure 3.2, when we demodulate
only the desired user's pilot signal, the other users' pilot signals and corresponding
multirate data signals interfere with the desired user's pilot signal. Therefore, the
total of interfering users (or signals) is 3(U-1). For example, if the interfering users
are 10, then the total interfering channels are 30. When the bit error probability is
10 -2 for G1=Gh=0.5, the required SNR is 3 and 12 dB, but it is much larger than
that for G1=Gh=1 in synchronous and asynchronous chip, respectively. This means
that the more channels of multirate data signal for each user, the worse the channel
 58

capacity, and thus the demodulation performance will be degraded quickly with the

increase of the data channel.

Figure 3.7 Bit error probability for chip synchronization in AWGN channel as a
function of the number of users

Figure 3.7 shows the bit error probability for chip synchronization in AWGN

channel as a function of the number of users. The bit error probability is degraded

when the DLL does not track the code delay error exactly. For example, when the

bit error probability is 10 -2 for asynchronous case with G1=Gh=0.5, the required

number of users is 9 (or 27 interfering channels) more than 11 (or 33 interfering

channels) that for synchronous case. This means that the chip delay jitter effects

greatly the performance of receiver.

Figures 3.8 and 3.9 show the bit error probabilities for chip synchronization in

Rayleigh fading channel as a function of the Average SNR per bit and the number
 59

of users, respectively. The bit error probability decreases quickly when the power

of data channel is small, but the probability decreases slowly when the power of

data channel is big since the multiple access interference increases much more for the

bigger power of data channel as seen in Figure 3.9.

Figure 3.8 Bit error probability for chip synchronization in Rayleigh fading channel
as a function of the Average SNR per bit

This chapter has focused on code tracking performance of coherent DLL in

an asynchronous multirate CDMA environment with MMAI over additive white

gaussian noise and Rayleigh fading channels. We proposed the DPC method to

analyze the autocorrelation and PSD of the time function having MMAI crosscorre-

lation. The PDF of the code tracking jitter and its variance were derived in a first

order loop and several results showed the effect of MMAI under various conditions.

The bit error probability of the code tracking jitter shown through a comparison of
 60

code synchronous case and asynchronous case as the functions of the average SNR

per bit and the number of users over AWGN and Rayleigh fading channel.

Figure 3.9 Bit error probability for chip synchronization in Rayleigh fading channel
as a function of the number of users
 CHAPTER 4

ONE-SHOT LINEAR DECORRELATING DETECTOR

We propose Hybrid Selection Diversity/Maximal Ratio Combining (Hybrid SD/MRC)

one-shot window linear decorrelating detector (LDD) for asynchronous DS-CDMA

systems. The selection diversity scheme at the input of the Hybrid SD/MRC LDD

is based on choosing the branch with the maximum signal-to-noise ratio (SNR) of

all filter outputs. The MR Combining scheme at the output of the Hybrid SD/MRC

LDD adopts to maximize the output SNR and thus compensates for the enhanced

output noise. The Hybrid SD/MRC one-shot LDD with PLL is introduced to track

its phase error and to improve the demodulation performance. The probability

density functions of the maximum SNR of the SD combiner , the near-far resistance

(NFR) of one-shot LDD by Gaussian approximation, and the maximum SNR of the

MR combiner for Hybrid SD/MRC LDD are evaluated, and the bit error probability

is obtained from these pdfs. The performance of Hybrid SD/MRC one-shot LDD is

assessed in a Rayleigh fading channel.

4.1 Introduction

Multiuser detection is used to improve performance by canceling the intra-cell

interference and thus increasing the system capacity. The actual capacity increase

depends on the efficiency of the algorithm, radio environment, and the system load.

In addition to capacity improvement, multiuser detection alleviates the near-far

problem typical to DS-CDMA mobile systems.

The idea of multiuser detection was first mentioned in 1979 by Schneider [35].

In 1983, Kohno et al. published a study on multiple access interference cancel-

lation receivers [36]. In 1984, Verdu proposed and analyzed the optimal multiuser

detector and the maximum likelihood sequence detector, which, unfortunately, is too

61
 62

complex for practical implementation since its complexity grows exponentially as a

function of the number of users [37]. Consequently, Verdu's work inspired a number

of researchers to find suboptimal multiuser detectors, which could achieve a close to

optimal performance with reasonable implementation complexity.

Figure 4.1 System model of multiuser detection

Fig.4.1 depicts a system for multiuser detection. Each user is transmitting data

bits, which are spread by the spreading codes. The signals are transmitted over an

additive white gaussian noise (AWGN) channel. In the receiver, the received signal

is correlated with replicas of the user spreading codes. The correlator consists of

a multiplier. A matched filter can also be used. Multiuser detection processes the

signals from the correlators jointly to remove the unwanted multiple access inter-

ference from the desired signal. The output of a multiuser detection block are the

estimated data bits.

A CDMA channel with U users sharing the same bandwidth is shown in Fig.4.1.

The signaling interval of each user is T seconds, and the input alphabet is antipodal

binary +1,-1. The objective is to detect those polarities, which contain the trans-
 63

mitted information. During the n th signaling interval, the input vector is

where d u (n) is the input symbol of the u th user. User u (u

is assigned a signature waveform (or code, or spreading chip sequence)

cu (t) which is zero outside [0, T] and is normalized

The baseband signal of the u th user is

where τu is the transmission delay, and α u(n ) is the channel attenuation. According

to (4.2), each user's signal travels along a single path, so this model does not illustrate

multipath propagation. For synchronous CDMA, the delay τu = 0 for all users. For

asynchronous CDMA, the delays can be different. The channel attenuation is

are the received power and phase of the u th user, respec-

,
tively. The received signal at baseband is the noisy sum of all the users' signals

where n(t) is additive white gaussian noise (AWGN). The first step in the detection

process is to pass the received signal r(t) through a matched filter bank (or a set

of correlators). It consists of U filters matched to individual signature waveforms

followed by samplers at instances nT + τu , n = 1,2, • • •. The outputs of the matched

filter bank form a set of sufficient statistics about the input sequence d(n) given r(t).

Thus, we will consider the equivalent discrete time channel model which arises at
 64

the output of the matched filter bank. The output vector

be expressed as

where R is cross-correlation matrix and a are U x U diagonal matrices, and n is

a colored Gaussian noise vector. The components of the matrix R are given by

cross-correlation between signature waveforms

is the attenuation vector of the received

signal.

For example, in a two-user system, the matrix

where p is the cross-correlation between the signature waveforms of the users in

Eq.(4.6).

Inspection of Eq.(4.5) immediately suggests a method to solve for d, whose

components du contain the bit information sought. If n was identically zero, we have

a linear system of equations, v = RPd, the solution of which can be obtained by

inverting R. With a non-zero noise vector n, the result is

where it is seen that the information vector d is recovered but contaminated by a

new noise term. From Eq.(4.8), the signal of the u th user is

 65

Note that the decorrelating detector completely eliminates MAI. However, the

which is greater than the noise power No at

the output of the matched filter. For an example, in Eq.(4.7), the noise power at

the output of the decorrelating detector is N o — ρ1,2ρ2,1) and is larger than 1

since ρ1,2ρ2,1 is always smaller than 1. Fig.4.2 shows an example of AWGN power

increment of conventional LDD output. The value of the inverse matrix of user 1

increases as the number of users increases for a fixed signal-to-AWGN ratio of 10 dB.

Figure 4.2 An example of noise power increment of LDD output

Despite this drawback, the decorrelating detector generally provides significant

improvements over the conventional detector. A more significant disadvantage of the

decorrelating detector is that the computations needed to invert the matrix R are

difficult to perform in real time. For a synchronous system, the problem is somewhat

simplified : we can decorrelate one bit at a time. In other words, we can apply the
 66

inverse of a UxU correlation matrix. For an asynchronous system, however, R is of

order NU, which is quite large for a typical message length N.

4.2 Hybrid SD/MRC One-shot LDD

As one of the main multiple-access techniques, code-division multiple access (CDMA)

has been intensively investigated over the late decade. It has been found that the

main obstacle to delivering CDMA's attractive promise in spectral efficiency is the

near-far (NF) effect, especially in the direct sequence spread spectrum multiple access

(DS/SSMA) systems [14], [38]. The NF effect exists when a weak received signal is

interfered with by one or more strong signal such that the effective cross-correlations

among the signals degrade the performance of the conventional CDMA receiver.

Multiuser detection is one of solutions to combat this problem. The interest

in the problem of multiuser detection was motivated by the work of Verdú where

he investigated the optimum maximum-likelihood multiuser receiver [30]. Verdú's

receiver is near-far resistant, but its complexity is exponential in the number of active

users. Recently much attention has been given to suboptimal detectors. Lupas and

Verdú's proposed linear decorrelating detector has attracted wide attention [38].
Since the decorrelator is a sequence detector, the detection process cannot be started

until the whole transmitted sequence is received at the receiver. In practice this is

not feasible and would result in a very long delay. Therefore, several finite delay

decorrelator schemes have been proposed. One of these schemes is the one-shot

detection scheme [29]. Suboptimum decorrelating detectors, which are based on

linear transformation of the sampled matched filter outputs, were considered in [38]

for the synchronous channel. For the asynchronous case, Verdú proposed to use a

one-shot linear decorrelating detector, in which 1+2(U-1) = 2U-1, (U is the number

of users) filters are matched to the spreading code with one bit duration of a specific

user [39].
 67

Wireless communication systems are subject to severe multipath fading that

can seriously degrade their performance. Thus, fading compensation is typically

required to mitigate the effect of multipath. Diversity combining scheme, which

combines multiple replicas of the received signal, is a classical and powerful technique

to combat multipath impairment. Two of many diversity combining schemes : the

maximum SNR selection combining and maximum ratio combining schemes have

attracted wide attention for compensating the Rayleigh faded envelope of received

signals [47],[48].

The maximum SNR selection diversity and maximal ratio combining schemes

have been analyzed for the last fifty years. The probability density function (pdf)

and symbol error probalibity of maximum-selection diversity reception schemes over

a Rayleigh fading channel was obtained in [45],[46],[48]. The maximum SNR at the

output of matched filter of the selection diversity combiner is random, and its pdf

was already well evaluated by many researchers [44],[48]. The pdf and symbol error

probalibity for conventional linear decorrelating detector was analyzed with Gaussian

approximation in [51].

The power control problem in uplink of DS-CDMA systems arises because of the

multiple access interference. Due to the propagation mechanism, the signal received

by the base station from a user terminal close to the base station will be stronger

than the signal received from another terminal located at the cell boundary. Hence,

the distant users will be dominated by the close user. This is called the near-far

effect. One solution to combat this near-far problem is multiuser detection where the

receiver exploits the knowledge of the sprea spectrum signals of the interfering users

to eliminate the near-far effect. The interest in the problem of multiuser detection

was motivated by the work of Verdu [30]. Verdu proposed the asynchronous LDD

with a packetized signal model [54], one of suboptimal detectors to combat the near-

far problem.
 68

In the asynchronous LDD, however, the detection should be delayed, at least, to

the end of the packet. Of course, the shorter the packet is, the smaller the detection

delay. However, the shorter packet means low channel efficiency since there should be

a no-signal period between each packet transmission. Without this no-signal period,

the asynchronous LDD proposed in [38] might be disturbed and all the interference

can not be decorrelated.

Another asynchronous LDD scheme has also been proposed by Verdu [54],

named as truncated window decorrelator or one-shot window decorrelator. Fig.4.3

(a) shows the bit period of the desired user and the truncated window for the decor-

relator. There are three independent signal within the window for 2 users case :

and

In Fig.4.3 Verdu's one-shot window decorrelator has the size : [2(U — 1) +

1]x [2(U — 1) + 1]. Verdu's decorrelator is to demodulate only the symbol of desired

user while the number of the decorrelator's matrix operation is (2U — 1) 3 . To

demodulate another user's symbol at the same time period, the matrix operation

must be repeated with the same complexity as that in the previous. Thus, to detect

simultaneously U-multiuser's symbols in one-shot window, the total number of the

matrix operation is U x (2U — 1) 3 .

On the other hands, In (b) of Fig.4.3, Hybrid SD/MRC one-shot LDD is to

demodulate U-symbols of U-multiusers in one-shot window simultaneously while

the decorrelator matrix size is 2U x2U. Thus, our decorrelator needs only one

matrix operation and each user needs two pieces at the input of the decorrelator

for asynchronous CDMA systems. The total number of the matrix operation with
 69

Figure 4.3 The comparison of Verdu's LDD and Hybrid SD/MRC LDD in
asynchronous CDMA systems
 70

post-decorrelation MR combining is (2U) 3 for combining of the left piece of the

present window and the right piece of the next window.

In this paper, we consider the NFR which Verdu defined as the multiuser

asymptotic efficiency [54] [29] [33] , and the compensation of the deep faded signal

in multiple access communications over a Rayleigh fading channel, and thus present

the Hybrid SD/MRC one-shot LDD which consists of the maximum SNR selection

diversity scheme at the input of one-shot LDD and the maximal ratio combining

scheme at the output of the LDD. The maximum SNR of the SD combiner output,

each diagonal component of R -1 of LDD, and the maximum SNR of MR combiner are

random variables, and thus each pdf of the random variables has been found by many

researchers [47],[48],[50]. However, they did not treat the pdfs for the asynchronous

one-shot LDD. Thus, we evaluated the joint pdf of the random variables, the BER,

and assessed the performance of the LDD with the analytical and simulated results

for the asynchronous one-shot LDD.

Fine synchronization of carrier phase error is also becoming increasingly

important for the development of advanced receivers of DS-CDMA systems in which

the NFR exists to overcome the reduction of system capacity. Thus, we adopt

the conventional phase-locked loop (PLL) to track the carrier phase error, derive

the steady-state pdf of the phase estimator error, and analyze the performance of

the PLL in the Hybrid SD/MRC LDD. By applying a single-user phase synchro-

nization scheme, phase locked loop (PLL), at each output of the LDD, it is shown

that the Hybrid SD/MRC one-shot LDD can be used for the coherent detection of

asynchronous CDMA without the preknowledge of user phase.

4.2.1 System Description

Consider an asynchronous CDMA system. Defining a rectangular pulse by hTα (t) = 1

for 0 < t < Tα and 4,, (t) = 0, otherwise, the spreading code waveform c(t) and the
 71

information wavefom d(t) can be expressed as

where consecutive signature and data bits {c(i)} and {d(i)} take on values of ±1

with equal probabilities, and T, and Tb are the chip duration of the spreading code

and the data duration of the information, respectively.

Figure 4.4 Hybrid SD/MRC one-shot LDD for first user

Consider there are L diversity channels, carrying the same information-bearing

signal. Each channel is assumed to be slowly fading with Rayleigh distributed

envelope statistics. The fading processes among the L diversity channels are assumed

to be mutually statistically independent. The signal in each channel is corrupted by

an additive zero-mean white gaussian noise. Thus, the equivalent low-pass received

signals for the lth diversity channel of uth user can be expressed as
 72

where

the number of total active users

the n th data bit of the uth user

the received power of u th user

the delay offset in each information bit

the carrier frequency

the initial carrier phase of uth user

the AWGN with one side PSD No

and the u th user's spreading code waveform

satisfies :

and

the left and right versions of the spreading waveform in one-shot window as shown

in Fig.4.5, respectively.

Diversity reception schemes are used to reduce the effects of fading. The basic

idea of a diversity reception is that if two or more independent samples of a signal are

taken, then these samples will fade in an uncorrelated manner. This means that the

probability of all the samples being simultaneously below a given level is much less
 73

than the probability of any individual sample being below that level. The selection

diversity combiner in Fig.4.4 is the simplest of all the diversity schemes. An ideal

selection combiner chooses the signal with the highest instantaneous SNR, so the

output SNR is equal to that of the best incoming signal.

In Fig.4.4, each user at the input of the SD combiner has two matched filters

for correlating of the corresponding code. One filter is matched to the left part of

its code in one-shot window, corresponding to the interval [0, τu] and other one is

matched to the right part of the code, corresponding to the interval [τu, Tb ]. The

SD combiner with each L diversity reception for each user has two outputs (v uL and

v uR ), each with maximum SNR.

Figure 4.5 One-shot window for asynchronous CDMA

The idea of a one-shot LDD is to transform the detection of asynchronous

CDMA into the detection of synchronous CDMA. Consider the one-shot window

as shown in Fig.4.5. Here we take a one-shot window approach where in order

to demodulate every bit we discard all information outside its interval. Defining

the received signal for uth user from Eq. (4.14) is rewritten as
 74

Assuming knowledge of the code delay waveform 6,„(t — nTb ) and carrier delay
-

phase øu from the channel estimator, the SD combiner output can be expressed as

The output of the bank of matched filters in matrix notation is given by

1,2,• • ,U, the correlation matrix R with the relative phase delay On ( = øu - øv)

between users can be expressed as

The linear decorrelating detector output gives

where the decorrelating detector applies the inverse of the correlation matrix,

to the conventional detector output in order to decouple the data.

Selection diversity is easy to implement, but it is not an optimal diversity

scheme because it does not use all of the possible branches simultaneously. Maximal

Ratio combining uses each of the LDD output pieces in a co-phased and weighted

manner such that the highest achievable SNR is available at the receiver at all times
 75

[48]. In Fig.4.4, the output of the LDD for first user has just two branches with each

other's data bit.

The MR combining scheme weights each branch output proportional to the

SNR for each branch to maximize the output SNR of the combiner. And, of course

this combining is not for the multipath fading but for the compensation of the

different noise enhancement by the inverse matrix decorrelating. Therefore, assuming

perfect fine synchronization of the carrier phase error, the soft estimate of LDD with

MR Combining scheme, yu(n), can be found by the sum of LDD output x u as follows:

where Mu' L and Ma are weighting coefficients to be determined to minimize the

output SNR at y. (n) and the signs ' and " denote the present matrix output and

next matrix output of the LDD, respectively. Note that after decorrelation, there are

only signal and enhanced background noise due to removing of interferences. The

input SNR (signal-to-background noise ratio) at v 1L and v1R are exactly the same.

However, the output SNR at x 1L and x1R are different from each other since each

output branch for the desired user (user 1) experiences different noise enhancement

by the inverse matrix filtering. The output SNR at each branch is enhanced by a

factor of λ1L and )1R, respectively, where

Since the input SNR is exactly the same, the factors )' 1L and λ1R are proportional to

the output SNR of each branch. And, thus, we can use these to post-decorrelation

MR combining weights, i.e.,

 76

Then, Eq.(4.22) can be rewritten as

Thus, we see that the decorrelating detector completely eliminates the multiple

access interference (MAI) and the Maximum Ratio Combining scheme does maximize

the output SNR.

4.2.2 Carrier Phase Error Tracking by PLL

Tracking synchronization is becoming increasingly important for the development

of advanced receivers for wideband DS-CDMA systems in which multiuser inter-

ference cancellation is employed to overcome the reduction of system capacity due

to this fundamental source of interference. Carrier phase synchronization of wireless

communication system is an already well-developed technique after the considerable

research efforts provided in the past [12].

Fig.4.6 shows a block diagram of the LDD with a PLL to track the phase delay

error On . Assuming the phase delay error at all the diversity branches of the same

user is the same, the information data recovered by the demodulator multiplies the

soft estimate output of Eq.(4.22) to remove the data information component

The phase error estimate in Eq.(4.25) is fed into a loop filter F(S), followed by

a VCC, Kp /S, where Kp is the overall closed-loop gain. The VCC is used to adjust

the carrier phase. F(S) is the transfer function of the loop filter where F(S) = 1 for

a first-order PLL, and Kp /S is the transfer function of the VCC. The instantaneous

phase estimate ø „ for the local oscillator is then related to the phase discriminator
^
 77

where S is the Heaviside operator S = d/dt. Therefore the change rate of the phase

estimator error can be expressed by

where dø u /dt is the rate of change of the u th user's initial phase, f(t) is the transfer

function of the loop filter in the time domain, N(t) = n yu , (n) + nyu''(n), and *

denotes the convolution operation.

Figure 4.6 Hybrid SD/MRC One-shot LDD with PLL for first user

The solution of a nonlinear stochastic differential equation such as (4.27) for

the steady-state probability density function (PDF), p(0), may be accomplished via

the Fokker-Planck method [4], [11], [14], [17]. From here on, we will use 0 = 0 for

convenience and simplicity. To illustrate the theory, we shall consider the case of
 78

and also assume The PDF, p(0),

satisfies the stationary equation :

where

with the power spectrum density of the noise term, SN.

From (4.28), using the well-known method [1],[4], [22] we get the PDF of the

phase estimator error as

where /0 (•) is the modified Bessel function of 0th-order. For v, it can be shown by

density function with zero mean and variance Parameter v represents the loop

signal-to-noise ratio(SNR).

4.2.3 Bit Error Probability

4.2.3.1 pdf of Maximum SNR of SD Combiner The bit error probability

will be obtained in a traditional way with the branch of the diversity reception, l =

1,2,• • ,L, by averaging the results for a slow Rayleigh fading channel over the distri-

bution of the maximal instantaneous SNR. The pdf of the maximal instantaneous

SNR over L independent identically distributed (iid) diversity paths is well known

[47]. In practice, the system cannot perform on an instantaneous basis; so, to be

successful, it is essential that the internal time constants of a selection diversity are

substantially shorter than the reciprocal of the signal fading rate.

 79

Consider L independent Rayleigh fading channels available at a receiver. We

assume that the instantaneous amplitude received by each diversity branch is statis-

tically independent of other branches but has equal average signal power αl^2. The

probability density function of the signal envelope, on branch l, is given by

with the noise power in lth branch, Not

the probability density function for γl is given by

The probability that all L independent diversity branches receive signals which

are simultaneously less than some specific SNR threshold 'y is

Therefore, the probability that at least one branch will exceed the threshold

SNR value of 'y is given by

To determine the average SNR of the received signal when diversity is used,

it is first necessary to find the pdf of the fading signal. For selection diversity, the

average SNR is found by differentiating Eq.(4.34). Therefore, letting -y 573 be the

 80

4.2.3.2 pdf of NFR of One-shot LDD We found the pdf of the output SNR
of the selection combiner at the input of the linear decorrelating detector. To obtain

the bit error probability of the receiver, it is necessary to evaluate the pdf of NFR

of the linear decorrelaing detector.

In the cross-correlation matrix of Eq.(4.20), each element except the diagonal

element of the matrix is a cross-correlated term due to multiple access inter-

ference. Lehnert and Pursley introduced the Gaussian approximation to analyze

the performance of the conventional detector for time varying CDMA signal [49]

and Yoon and Hong used it for discrete time random signature sequences [50]. We

will use the Gaussian aproximation to analyze the statistical characteristics for the

off-diagonal components in the inverse matrix of R.

Consider the detection of first user and the delay of the uth user with respect

of p1LuL and ρ1RuR conditioned on D u and 6u can be expressed as

where the variances are given by

and, N is the processing gain (PG) and random variable ρ1L uL + ρ1RuR has an

unconditional distribution N(0, N/3) [51].

Considering the diagonal element of R -1 corresponding to the left or right

version of uth user in one-shot LDD, Rug, for Q = L, R, in random CDMA system,

the diagonal elements are random. Using the well-known method [52], [53], the
 81

probability that the mth bit of uth user with the left or right version of one-shot

window is misdetected is given by

The NFR W is a random variable in a random-CDMA and thus we will evaluate

the pdf of the NFR to get the bit error probability of LDD. The pdf of the off-

diagonal element of asynchronous linear decorrelating detector for random CDMA

has been analyzed by approximation in [51] but is covered briefly here for clarity.

Their approach was based on the random signature sequences for one packet, set

all the off-diagonal components in the cross-correlation matrix to zero except for

one row and one column corresponding to the cross-correlation of the desired user's

signature waveform with those of all other interfering users and showed the validity

of the following approximation :

where the processing gain is assumed to be large enough such that (1). < 1.

The approximation in Eq.(4.41) can be used in the one-shot LDD for random

signature sequences since each cross-correlation term for one packet is the same as

that for one bit. Therefore, we can get the approximate pdf of the near-far resistance

for asynchronous LDD as [51]

where
 82

and the chi-square distribution function with M degree of freedom :

4.2.3.3 Output SNR of MR Combiner In the previous subsection, we found

the pdf of the maximum SNR of SD combiner and the pdf of NFR of LDD. The LDD

output has only two branches with the corresponding partial energy for each user,

the left version of the present one-shot window and the right version of the next

one-shot window are weighted summed, and thus the demodulation output with

the full energy of one-shot duration is found. prior to finding the hard decision of

demoduation, we minimized SNR reduction due to the enhancement of LDD output

noise by using the maximal ratio combining scheme.

The output of the MR combiner for uth user in Fig.4.4 is

the variance :
 83

In Eq.(4.46), the λuΩ is a random variable and independent identically

distributed [38],[51], and thus the SNR of MRC output can be expressed as

The partial energy terms E uL and EuR in Eq.(4.47) are not statisically

independent for each other as defined in Eq.(4.17) and thus the normalized partial

energy is EuR = 1 - EuL . It is difficult to evaluate the SNR with the mean of the

partial energy. Therefore, we approximate E[E uΩ ] to 0.5 by normalization since

the time delay τu is uniformly distributed in [0, Tb ] and thus the mean value of

the partial energy of left version is the same as that of right version in one-shot

window.The SNR can be expressed as

are independent identically distributed, and

The pdfs of maximum SNR of SD combiner, NFR of one-shot LDD and the

maximum SNR of the MR combining output are known, and thus the average bit

error probability can be expressed as

 84

where fγSD (7sD) and fw (w) were evaluated in Eq.(4.36) and Eq.(4.42), respectively,

and the Q-function was defined in the previous page.

4.2.4 Numerical Analysis

Using the results of the previous sections, a numerical evaluation of the performance

of linear decorrelating detector with Selection Diversity and MR Combining schemes

is given in this section. The code sequences used in the analysis use only code length

N = 128 of 511 chip Gold codes to keep the property of random-CDMA code. The

PLL performance in a conventional LDD is analyzed using the pdf in AWGN and

Rayleigh fading channel environments. The performance of LDD with maximum

SNR selection diversity and MR Combining scheme is shown through bit-error rate

(BER) in AWGN and Rayleigh fading channel environments. All simulations used

the Matlab tool.

Fading is caused by interference between two or more versions of the trans-

mitted signal which arrive at the receiver at slightly different times. Rayleigh

fading channel is the most severe mobile radio channel. Since third generation

wideband CDMA uses fast closed loop power control, the near-far effect is mitigated

by equalizing the user powers and signal-to-AWGN+MAI ratio(SNIR) performance

is improved by compensating for the channel fading[16]. However, our simulation

considers the uncompensated Rayleigh fading channel. Therefore, we analyze the

performances of the Hybrid SD/MRC LDD in Rayleigh fading with the assumption

of a power controlled environment.

Fig.4.7 shows the bit error probability for Maximal Ratio Combining (MRC)

scheme to maximize the output SNR of LDD without SD scheme since we fixed

AWGN term and thus the term is not random at the output of the SD combiner.

The MRC scheme provides better improvement in bit error performance for moderate

to high average SNR in AWGN channel. For example, when Pe = 10 -5 , the required
 85

average SNR per bit for MRC LDD is around 11dB, which is less than that required
of CLDD. This shows that the MRC scheme improves the demodulation performance
by the maximization of SNR at the LDD output.

Figure 4.7 BER of LDD without SD/MRC and Hybrid SD/MRC LDD (L = 1) in
AWGN channel

Fig.4.8 illustrates the concept of SD scheme in bit error probability. We

compute the bit error probability for Hybrid SD/MRC LDD by using Eq.(4.49) for
L = 2,4,8, and 16. When Pe = 10 -2 , the LDD without SD/MRC needs the average
SNR per bit, around 12 dB, while the required SNR of SD/MRC LDD receiver is
just 8 dB at L = 2 and 6.5 dB at L = 4. The SD receiver can improve much better
the performance of demodulation as the number of branches increases.
Fig.4.9 shows the bit error probability of LDD without SD/MRC, MRC LDD,
and SD/MRC LDD with L=2,4 in Rayleigh fading channel. The instantaneous
 86

maximum SNR SD scheme can compensate the performance degradation due to

the deep faded signal and thus, provide much better performance than the LDD

without SD/MRC. For example, when Pe = 10 -2 , the required average SNR per
bit for SD/MRC with L=2 is around 9 dB less than that for MRC LDD with L=1,

while another 2.5 dB can be saved by an L=4 diversity scheme. This clearly demon-

strates that the transmitter signal power may be reduced when selection diversity

and maximal ratio combining schemes are adopted.

Figure 4.8 Analytical BER at L = 2, 4, 8, and 16 in Rayleigh fading channel

To verify the analysis derived in the previous section, the analytical BER is

compared with the simulated results for asynchronous CDMA as a function of the

number of users in Rayleigh fading channel. In this simulation, we used the base-

band BPSK signal at the signal-to-AWGN ratio 10 dB. The BER is slicely degraded
 87

as the number of users increases due to the enhancement of output noise term of the

LDD.

Figure 4.9 BER of Conventional matched filter (Cony. MF) with SD (L=2, L=4),
LDD without SD/MRC, LDD with MRC scheme, and Hybrid SD/MRC LDD with
L = 2,4 in Rayleigh fading channel

Fig.4.11 depicts the comparison of the simulated pdf and the analytical value

as descrived by Eq.(4.42). For simplicity of simulation, we use the synchronous chip

duration and thus, we changed the variance to 1/2N. The inversed value of diagonal

component value of R -1 increase as a function of the number of users increases and

the BER performance of the LDD is degraded with the increase of the number of

users. For example, the mean value and peak height of the pdf are about 9.3 and

0.2 for 10 active users while the mean and peak height are about 0.85 and 0.06 for

20 active users.
 88

Fig.4.12 illustrates the behavior of the phase estimator error pdf, p(0), in

AWGN and Rayleigh fading channels when the signal-to-AWGN power ratio is fixed

at 10 dB. The height of the pdf was normalized with the reference of only one user's

height. The peak height of the PDF clearly increases in AWGN channel. However,

the peak decreases in Rayleigh fading channels. This means the tracking ability of

the phase-locked loop is degraded as the channel environment is poor.

Figure 4.10 bit error probability with comparison of analytical and simulated results
as a function of the number of users in Rayleigh fading channel environment

Fig.4.13 shows the bit error probability between the LDD with PLL and without

PLL as a function of the average SNR per bit in AWGN and Rayleigh fading channels.

The phase tracking error in a real system is very small after aquisition, but we used

the range of the error, ±10 degree, to investigate plainly the behavior of the error

tracking of PLL. The carrier phase error causes the degardation of demodulation
 89

performance in multiuser detection. For example, when the bit error probability is

Pe 10 -2 for the conventional LDD without PLL in Rayleigh fading channel, the
required average SNR per bit is about 1 dB larger than that with PLL, while another

3 dB can be saved by SD/MRC LDD receiver. In AWGN, the required SNR per bit

of the LDD without SD/MRC and SD/MRC LDDs can be also saved by PLL. This

means that fine synchronization of carrier phase error can mitigate the degradation

of performance of the receiver.

Figure 4.11 pdf of the inversed value of diagonal component of R -1 for first user's
left version (N = 128, and (5,, = 0)
Figure 4.12 pdf of the phase estimator error in AWGN and Rayleigh fading channels

Figure 4.13 Comparison of BER between LDD with PLL and LDD without PLL
in AWGN and Rayleigh fading channels
 CHAPTER 5
CONCLUSION

This dissertation first analyze the phase-estimator error performance of a PLL in an

asynchronous multirate DS-CDMA system with MMAI and AWGN. We constructed

an integrated PLL and DLL receiver to synchronize the phase delay and the code

delay simultaneously. In particular, we proposed the DPC method to evaluate

the autocorrelation and its PSD of the crosscorrelation terms in a multirate signal

environment.

A further focus is on code tracking performance of a coherent DLL in an

asynchronous multirate CDMA environment with MMAI and AWGN. We proposed

the DPC method to analyze the autocorrelation and PSD of the time function having

MMAI crosscorrelation. The PDF of the code tracking jitter and its variance was

derived for a first order loop and several results showed the effect of MMAI under

various conditions. Though numerical results may be different in real systems, we

chose parameters similar to those in a commercial system. Therefore, we expect that

our results will facilliate the evaluation of third generation mobile communication

systems

A main issue in multiuser detection is to decouple the MAI of the received signal

of a DS-CDMA system. Among many linear multiuser detectors which decouple the

multiple access interference from each of the interfering users, one-shot window linear

decorrelating detector (LDD) based on a one bit period to reduce the complexity of

the LDD has attracted wide attention as an implementation scheme. Therefore, we

proposed Hybrid Selection Diversity/Maximal Ratio Combining (Hybrid SD/MRC)

one-shot window linear decorrelating detector (LDD) for asynchronous DS-CDMA

systems. The selection diversity scheme at the input of the Hybrid SD/MRC LDD

is based on choosing the branch with the maximum signal-to-noise ratio (SNR) of

all filter outputs. The MR Combining scheme at the output of the Hybrid SD/MRC

91
 92

LDD adopts to maximize the output SNR and thus compensates for the enhanced

output noise. The Hybrid SD/MRC one-shot LDD with PLL is introduced to track

its phase error and to improve the demodulation performance. The probability

density functions of the maximum SNR of the SD combiner , the near-far resistance

(NFR) of one-shot LDD by Gaussian approximation, and the maximum SNR of the

MR combiner for Hybrid SD/MRC LDD are evaluated, and the bit error probability

was obtained from these pdfs. The performance of Hybrid SD/MRC one-shot LDD

was assessed in a Rayleigh fading channel.

 REFERENCES

1. Jacob Klapper, and John T. Frankle, Phase-Locked and Frequency Feedback

Systems, New York: Academic Press, 1972.

2. Floyd M. Gardner, Phaselock Techniques, New York: John Wiley and Sons,
1979.

3. Andrew J. Viterbi, CDMA : Principles of Spread Spectrum Communication,

Massachusetts: Addison-Wesley Publishing, 1995.

4. Andrew J. Viterbi, Principles of Coherent Communications, New York:

McGraw-Hill, chpt.4, 1966.

5. Fumiyuki Adachi, Mamoru Sawahashi, and Hirohito Suda "Wideband DS-CDMA

for Next-Generation Mobile Communications Systems," IEEE Commun.
Mag., vol. 36, pp.56-69, Sept. 1998.

6. Roger L. Peterson, Rodger E. Ziemer, David E. Borth, Introduction to Spread

Spectrum Communications, Prentice Hall 1995.

7. Esmael H. Dinan and Bijan Jabbari, Speading codes for Direct Sequence CDMA
and Wideband CDMA Cellular Networks, IEEE Commun. Mag., vol. 36,
pp.48-54, Sept. 1998.

8. F. Adachi, M. Sawahashi, and K. Okawa, "Tree-Structured Generation of

Orthogonal Spreading Codes with Different Length for Forward Link of
DS-CDMA Mobile Radio," Elect. Lett., vol. 33, pp.27-28, Jan. 1997.

9. Clarke, R.H., "A Statistical Theory of Mobile-radio Reception," Bell Systems

Technical Journal, vol. 47, pp.957-1000, 1968.

10. D.R. Anderson and P.A. Wintz "Analysis of a spread-spectrum multiple-access

system with a hard-limiter," IEEE Trans. Commun. Technology, vol.
COM-17, pp.285-290, Apr. 1969.

11. Wei Huang, Ivan Andonovic, and Masao Nakagawa, "Code Tracking performance
of DS-CDMA schemes in the Presence of multisuer interference and
additive gaussian noise," IEEE ISSSTA'98, Sun City, South Africa,
vol.3/3, pp.843-847, Sept, 1998.

12. Michael B. Pursley, "Performance Evaluation for Phase-Coded Spread-Spectrum

Multiple-Access Communications-Part I: System Analysis" , IEEE Trans.
Commun., vol. COM-25, pp.795-799, Aug. 1977.

13. Urbashi Mitra, "Comparison of Maximum-Likelihood-Based Detection fot Two

Multirate Access Schemes for CDMA Signals," IEEE Trans. Commun.,
vol. 47, pp.64-77, Jan. 1999.

93
 94

14. Jack K. Holmes, Coherent Spread Spectrum Systems, New York, John Wiley
and Sons, 1982.
15. J.J. Spilker, Jr., "Delay lock tracking of binary signals," IEEE Trans. Space
Electron. Telemetry, vol. SET-9, pp.1-9, Mar. 1963.

16. Tero Ojanpera, and Ramjee Prasad, Wideband CDMA for Third Generation
Mobile Communications, Boston: Artech House Publishers, 1998.

17. Marvin K. Simon, "Noncoherent pseudonoise code tracking performance of

spread spectrum receivers," IEEE Trans. Commun., vol. COM-25,
pp.327-345, March 1977.
18. Andreas Polydoros and Charles L. Weber, "Analysis and Optimization of
Correlative Code-Tracking Loop in Spread-Spectrum Systems" IEEE
Trans. Commun., vol. COM-33, pp.30-43, Jan. 1985.

19. James Caffery, Jr. and Gordon L. Stűber "Performance of Non-coherent DLL in
Multiple Access Interference," IEEE ISSSTA'98, Sun City, South Africa,
vol.3/3, pp.823-82'7, Sept, 1998.
20. Riccardo De Gaudenzi and Marco Luise "Decision-Directed Coherent Delay-Lock
Tracking Loop for DS-Spread-Spectrum Signals," IEEE Trans. Commun.,
vol.39, pp.758-765, May 1991.
21. Heinrich Meyr, "Delay-Lock Tracking of Stochastic Signals," IEEE Trans.
Commun., vol. COM-24, pp.331-339, March 1976.

22. Wei Huang, Ivan Andonovic, and Masao Nakagawa, " PLL Performance of DS-
CDMA Systems in the Preence of Phase Noise, Multiuser Interference, and
Additive Gaussian Noise", IEEE Trans. Commun., vol. 46, pp.1507-1515,
Nov. 1998.
23. Chang Y. Yoon, William C. Lindsey, " Phase-Locked Loop Performance in the
Presence of CW Interference and Additive Noise" , IEEE Trans. Commun.,
vol. COM-30, pp.2305-2311, Oct. 1982.
24. Fulvio Gini, Georgios B. Giannakis, "Frequency Offset and Symbol Timing
Recovery in Flat-fading Channel: A Cyclostationary Approach" , IEEE
Trans. Commun., vol. 46, pp.400-411, March 1998.

25. Aaron Weinberg, Bede Liu "Discrete Time Analyses of Nonuniform Sampling
First- and Second-Order Digital Phase Lock Loops " , IEEE Trans.
Commun., vol. COM-22, pp.123-137, Feb. 1974.
26. John E. Ohlson, "Phase-Locked Loop Operation in the Presence of Impulsive
and Gaussian Noise " , IEEE Trans. Commun., vol. COM-21, pp.991-996,
Sep. 1973.
 95

27. Martin T. Hill, Antonio Cantoni, "A Frequency Steered Phase-Locked Loop ",
IEEE Trans. Commun., vol. 45, pp.737-743, June 1997.

28. R. Lupas, S. Verdu, "Linear Multiuser Detectors for Synchronous Code-Division

Multiple Access Channels", IEEE Trans. Info. Theory, vol. IT-35, pp.123-
136, Jan. 1989.

29. R. Lupas, S. Verdu, "Near-Far Resistance of Multiuser Detectors in

Asynchronous Channels", IEEE Trans. Commun., vol. COM-38, pp.496-
508, Apr. 1990.

30. S. Verdu, "Minimum Probability of Error for Asynchronous Gaussian Multiple-

Access Channels", IEEE Trans. Info. Theory, vol. IT-32, pp.85-96, Jan.
1986.

31. Z. Xie, R.T. Short, C.K. Ruthforth, "A Family of Suboptimum Detectors for
Coherent Multiuser Communication", IEEE JSAC, vol. 8, pp.683-690,
May 1990.

32. U. Madhow, M.L. Honig, "MMSA Interference Suppression for Direct-

Sequence Spread-Spectrum CDMA", IEEE Trans. Commun., vol.
COM-42, pp.3178-3188, Dec. 1994.

33. R. Lupas, "Near-far resistance linear multiuser detection", Ph.D. thesis, Dept.
Elec. Eng., Princeton Univ., 1989.

34. T. Ottoson and A. Svensson "Multi-rate Schemes in DS/CDMA Systems," IEEE

VTC'95, Chicago, pp. 1006-1010, July, 1995.

35. Schneider, K.S., Optimum detection of code division multiplexed signals, IEEE
Trans. Aeros. Electron. Syst., vol. AES-15, No.1, pp.181-185, Jan. 1979.

36. Kohno, R., H. Imai, and M. Hatori, Cancellation technique of co-channel inter-
ference in asynchronous spread-spectrum multiple access systems, IEICE
Trans. Commun., vol. 65-A, pp.416-423, May 1983.

37. Verdu, S., Optimum Multiuser Signal Detection, Ph.D. Thesis, University of
Illinois, Urbana-Champaign, 1984.

38. Lipas, R., and Verdu, S., Near-far resistance of multiuser detectors in
asynchronous code-division multiple access communications, IEICE Trans.
Commun., COM-38, pp.496-508, Apr. 1990.

39. Peng, M., Y.J. Guo, S.K. Barton, One-shot Linear Decorrelating Detector for
Asynchronous CDMA, Proceedings of Globecom'96., pp.1301-1305, 1996.

40. Bernard Sklar, Rayleigh Fading Channels in Mobile Digital Communication

Systems, Part I : Characterization, IEEE Communications Magazine,
pp.90-109, July 1997.
 96

41. Shimon Moshavi, Multi-User Detection for DS-CDMA Communications, IEEE

Communications Magazine, pp.124-136, Oct. 1996.

42. Alexandra Duel-Hallen, Jack Holtzman, and Zoran Zvonar, Multiuser Detection
for CDMA Systems, IEEE Personal Communications, pp.46-58, Apr.
1995.
43. Zoran Zvonar and David Brady, Multiuser Detection in Single-Path Fading
Channels, IEEE Trans. on Commun., pp.1729-1739, Feb./Mar./Apr.
1994.
44. L. R. Khan, Radio Squarer, Proc. of IRE, 42 Nov. 1954.
45. Michel D. Yacoub, Foundations of Mobile Radio Engineering, Boca Raton: CRC
Press, 1993.
46. Vijay K. Garg and Joseph E. Wilkes, Wireless and Personal Communications
Systems, ATT Bell Lab.: Prentice Hall, 1996.

47. Elisabeth A. Neasmith and Norman C. Beaulieu, New Results on Selection

Diversity, IEEE Trans. commun., Vol.46, pp.695-704, May 1998.

48. Gwo-Tsuey Chyi, John G. Proakis, and Catherine M. Keller, On the Symbol
Error Probabiliy of Maximum-Selection Diversity Reception Schemes Over
a Rayleigh Fading Channel, IEEE Trans. commun., Vol.37, pp.79-83,
Nov. 1989.
49. J. S. Lehnert and M. B. Pursley, Error Probabiliy for Binary Direc-Sequence
Spread Spectrum Communications with Random Signature Sequences,
IEEE Trans. commun., Vol.COM-35, pp.87-98, Nov. 1987.

50. S. H. Yoon and K. S. hong, Bit Error Probability of Linear Decorrelating Detector
for random Signature Sequences, Proc. of Vehicular Technology Conference
(Fall)., Amsterdam, The netherlands Sep. 1999.

51. S. H. Yoon and Yeheskel Bar-Ness, An Approximate Analysis of Asynchronous

Linear Decorrelating Detector for Dynamic CDMA System, Proc. of
IS S S TA 2000, New Jersey, USA Sep. 2000.

52. Theodore S. Rappaport, Wireless Communications: Principles and Practice,

New Jersey: Prentice Hall, 1996.
53. F. Zheng and S. K. Barton, Near-Far Resistant Detection of CDMA Signals via
Isolation Bit Insertion, IEEE Trans. commun., Vol.COM-43, pp.1313-
1317, Feb./March/April. 1987.
54. S. Verdu, Multiuser Detection, Cambridge University Press, 1998.