Design and Implementation of an Optical Code ... - CiteSeerX

Design and Implementation of an

Optical Code Division Multiple Access System

Combined with Dense Wavelength Division Multiplexing

Von der Fakultät für Informatik, Elektrotechnik und Informationstechnik

der Universität Stuttgart zur Erlangung der Würde

eines Doktor - Ingenieurs (Dr.-Ing.) genehmigte Abhandlung

vorgelegt von

Yongjun Zhang aus Beijing, China

Hauptberichter: Prof. Dr.-Ing. Joachim Speidel

Mitberichter: Prof. Dr.-Ing. Manfred Berroth

Tag der mündlichen Prüfung: 05. Dezember 2003

Institut für Nachrichtenübertragung der Universität Stuttgart

2003

1

Acknowledgements

The work presented in this thesis was carried out at the Institut für Nachrichtenübertragung,

University of Stuttgart under the supervision of Professor Joachim Speidel. I would like to

express my gratitude to him for providing a conducive environment for performing this

research at his institute. His advice and encouragement have been an invaluable guidance

throughout this project. I also would like to thank Professor Manfred Berroth for acting as a

co-examiner. I am also grateful to my colleagues at the institute. Their knowledge and

expertise have been very valuable to me and I like to thank them for sharing those with me. I

also want to thank Friedrich-Ebert foundation for their scholarship at the first two years of my

studies. Finally, I am very indebted to my wife for her full support during my studies in

Germany.

Stuttgart, June 2003

Yongjun Zhang

2

Contents

Acknowledgements ................................................................................................................... 1

Contents..................................................................................................................................... 2

Abstract..................................................................................................................................... 5

Zusammenfassung .................................................................................................................... 5

Notation..................................................................................................................................... 6

Abbreviations ............................................................................................................................ 8

Chapter 1 Introduction......................................................................................................... 10

Chapter 2 Fundamentals of OCDMA................................................................................. 13

2.1 Using OCDMA in Optical Access Network............................................................... 13

2.2 Classification of OCDMA .......................................................................................... 14

2.3 Model of DS-OCDMA System .................................................................................. 15 2.3.1 Transmitter:....................................................................................................... 16

2.3.2 Receiver ............................................................................................................ 17

2.4 Code of DS-OCDMA System .................................................................................... 19

2.5 Parameters of OCDMA .............................................................................................. 21

Chapter 3 Experimental Results for OCDMA................................................................... 24

3.1 Incoherent Direct Sequence OCDMA ........................................................................ 24

3.2 Coherent DS-OCDMA ............................................................................................... 25

3.3 Frequency Encoded OCDMA..................................................................................... 28

3.4 Fast Frequency Hopping (FFH) OCDMA.................................................................. 31

3.5 PSK-OCDMA............................................................................................................. 32

3.6 The Other Methods ..................................................................................................... 36

3.7 Some Methods for Improving the Performance of OCDMA..................................... 40

3.8 Conclusion.................................................................................................................. 42

Chapter 4 OCDMA with Combined Electrical and Optical En/Decoding ...................... 44

3

4.1 The System Model...................................................................................................... 44 4.1.1 The Impulse Shape of Chips ............................................................................. 45

4.1.2 Electronic Encoder ............................................................................................ 46

4.1.3 Optical Modulator............................................................................................. 47

4.1.4 Optical Encoder and Decoder ........................................................................... 50

4.1.5 Optical Channel (Optical Fiber)........................................................................ 52

4.1.6 O/E Converter with Electronic Amplifiers ....................................................... 53

4.1.7 Electronic Decoder............................................................................................ 53

4.2 Interferometric Noise .................................................................................................. 54 4.2.1 Characteristics of Interferometric Noise ........................................................... 54

4.2.2 The Source of the Phase Noise ......................................................................... 56

4.3 The Results of the Simulation..................................................................................... 57 4.3.1 The Noise in the Optical Encoder..................................................................... 57

4.3.2 The Noise Performance of Optical Decoder..................................................... 61

4.3.3 The Performance of Electronic Decoder........................................................... 63

4.3.4 The Effect of Width of Chip Impulse ............................................................... 64

4.3.5 Summary........................................................................................................... 65

4.4 The Channel Coding ................................................................................................... 65 4.4.1 Reed-Solomon Code ......................................................................................... 66

4.4.2 Convolutional Code .......................................................................................... 68

Chapter 5 Implementation of a Prototype .......................................................................... 69

5.1 Experimental System.................................................................................................. 69

5.2 Signal Source and Scrambler ...................................................................................... 72

5.3 Impulse Generator ...................................................................................................... 75

5.4 Encoder/Decoder ........................................................................................................ 78 5.4.1 The Codeword................................................................................................... 78

5.4.2 Electronic Encoder ............................................................................................ 80

5.4.3 Optical Encoder................................................................................................. 82

5.4.4 Electronic Decoder............................................................................................ 83

5.4.5 Optical Decoder ................................................................................................ 84

5.5 Optical Transmitter and Receive r ............................................................................... 85 5.5.1 Electronic/Optical Converter ............................................................................ 85

5.5.2 Optical Receiver................................................................................................ 90

4

5.6 Data Recovery ............................................................................................................ 91 5.6.1 Clock Recovery................................................................................................. 91

5.6.2 Phase Tracking.................................................................................................. 92

5.7 Channel Code.............................................................................................................. 95

5.8 Measurement Results .................................................................................................. 96

Chapter 6 OCDMA Combined with DWDM..................................................................... 98

6.1 The Fundamentals of WDM ....................................................................................... 98

6.2 The Feasibility of the CDMA/WDM System........................................................... 100

6.3 Network Architecture and Management................................................................... 103

Chapter 7 Conclusions ........................................................................................................ 105

Appendix ............................................................................................................................... 107

Appendix A1 An Algorithm of BER with Hard- limiter ................................................ 107

Appendix A2 Optical Splitter and Adder ...................................................................... 111

Appendix B Results of the Hardware Experiments ....................................................... 112

Bibliography ......................................................................................................................... 125

5

Abstract

To improve the flexibility and the capability of optical access networks, the investigations of this dissertation have been focused on applying optical code division multiple access (OCDMA) in the upstream of the access network. The noise caused by the interference between the same optical sources and different optical sources with the same and different wavelengths has been analyzed for the optical en/decoder. For a narrowband optical source, the interferometric noise produced by the variation of temperature of optical components restricts the signal processing within the optical domain. Therefore, a scheme with electronic en/decoding process has been proposed. An OCDMA demonstration system involving three users has been developed in the laboratory. The users can access the passive optical network in an asynchronous mode. The implementation of the OCDMA encoder and decoder has been done with both optical and electrical signal processing, respectively. The combination of OCDMA and dense wavelength multiplexing (DWDM) has been analyzed. To improve the performance of data and clock recovery, a new scheme of a phase tracing circuit has been designed. Some channel codes which could be suitable for OCDMA channels have been also investigated.

Zusammenfassung

Die vorliegende Dissertation befasst sich mit der Erhöhung der Flexibilität optischer Zugangsnetze durch den Einsatz der optischen Codevielfachzugriffstechnik (OCDMA, Optical Code Division Multiple Access). Dabei werden die Rückkanäle hoher Bitrate in einem passiven optischen Teilnehmeranschlussnetz (PON) betrachtet. Das Rauschen, das aus der Interferenz zwischen Lichtwellen gleicher oder unterschiedlicher Quellen entsteht, wird für optische Codierer und Decodierer analysiert. Für schmalbandige optische Quellen wird die Güte der nichtkohärenten optischen Signalverarbeitung durch dieses interferometrische Rauschen, das im wesentlichen durch Temperaturschwankungen in den optischen Komponenten entsteht, beschränkt. Deshalb wird auch ein Verfahren mit elektrischen OCDMA-Codiern und -Decodiern untersucht. Ein OCDMA-Demonstrationssystem mit drei Teilnehmern wird im Labor aufgebaut. Jeder Teilnehmer kann asynchron auf das passive optische Anschlussnetz zugreifen. Die Verfahren mit optischen als auch elektrischen Codierern und Decodierern werden theoretisch und experimentell untersucht. Ferner wird eine Kombination von optischem Code- und Wellenlängenmultiplex (DWDM) vorgestellt. Hierzu wird eine neue Schaltung für die Daten- und Taktrückgewinnung entworfen. Ferner werden geeignete Verfahren der Kanalcodierung vorgestellt, welche die Bitfehlerhäufigkeit bei einer größeren Zahl zugeschalteter Nutzer auf ein zulässiges Maß verringern können.

6

Notation

b – propagation constant

Β – input beam radius

hB – the power equivalent bandwidth of h(t)

rB – the bandwidth of electronic receiver

c – the speed of light in vacuum

lc – specifies a codeword

D – chromatic dispersion coefficient of optical fiber

DM – material dispersion

DW – waveguide dispersion

f(t) – impulse response of the equalized filter of the optical receiver and amplifiers

nfλ – the center frequency of the laser with the wavelength nλ

F(s) – the Laplace transform of f(t)

h(t) – impulse response of the CDMA decoder filter

I(t) – the output of the driver of Laserdiode

bI – the bias current of Laserdiode

thI – the threshold of Laserdiode

L – the distance of the transmission

CL – the length of the codeword

cl – the length of the delay line for one chip duration

N – the number of accessed users

gn – the group index in the core of the fiber

oN – power spectral density of the noise

CCP – bit error probability after convolutional decoding

chP – the loss of optical power in fiber

deP – the loss of optical power at the optical decoder

eP – bit error probability of OCDMA link

7

enP – the loss of optical power of the optical encoder

LDP – optical power emitted by the Laserdiode

mP – optical margin of a optical link

maxP – the maximum optical power

markP – the optical power of mark ( “1” )

NullP – the optical power of Null ( “0” )

PONP – the loss of optical power at PON

rP – probability

RSP – bit error probability after RS decoding

SP – symbol error probability

thP – the approximate value of optical power with threshold current

R – reflectivity of the interferometer mirror

ℜ – redundancy of the channel code

chipR – the ratio between number of “1” chips and total number of chips within a codeword

DR – bit rate of user data

maxR – the maximum data rate of transmission through an OCDMA channel

CT – the duration of a chip

DT – the duration of the user bit

∆TD – the impulse degradation from dispersion

Tg – group delay in fiber

w – code weight

η – the efficiency of optical electronic conversion

λ – the wavelength of the light in vacuum

λ∆ – the linewidth of optical source

aλ – auto-correlation

cλ – cross-correlation

θ – the threshold of the decision

ω – the angular frequency of the light

cτ – the coherence-time of the optical source

∗ – denotes a convolution

8

Abbreviations

ASE – amplified spontaneous emission

ATM – asynchronous transfer mode

ATC – automatic temperature control

BER – bit error ratio

CC – convolutional code

CDMA – code division multiple access

CW – continuous wave

DC – dispersion compensation

DD – direct detection

DFB Laser – distributed feedback Laser

DS-CDMA – direct sequence CDMA

DWDM – dense WDM

EAM – electro-absorption modulator

ECL – emitter-coupled logic

EDFA – erbium doped fiber amplifier

EXT – extinction ratio of the optical transmitter

FDMA – frequency division multiple access

FEC – forward error correction

FIR – finite impulse response

FTTH – fiber to the home

FWHM – full-width half-maximum

FWM – four-wave mixing

GVD – group velocity dispersion

HFC – hybrid fiber-coax

IMD – inter-modulation distortion

IM-DD – intensity modulation direct-detection

IMP – inter-modulation product

ISI – intersymbol interference

ITU – International Telecommunications Union

9

LED – light emitting diode

MAI – multi-access interference

MFD – matched-filtered detection

OCDMA – optical CDMA

ONU – optical network unit

OLT – optical line terminal

OTDMA – optical time division multiple access

PLC – planer lightwave circuit

PMD – polarisation mode dispersion

PON – passive optical network

PRBS – pseudo random bit sequence

PSK – phase shift keying

RF – radio frequency

SBS – stimulated Brillouin scattering

SCM – subcarrier multiplexing

SDH – synchronous digital hierarchy

SPM – self-phase modulation

SRS – stimulated Raman scattering

SSFBG – superstructured fiber Bragg gratings

SSMF – standard single-mode fiber

TEC – thermoelectric cooler

TDMA – Time Division Multiple Access

TR – thermistor resistance

WDM – Wavelength Division Multiplexing

XPM – cross-phase modulation

10

Chapter 1 Introduction

For backbone networks, in recent years there has been a rapid increase in the capability of

transmitting high rate data with the usage of both time division multiplexing (TDM) and

wavelength division multiplexing (WDM) technologies within the optical domain. Also

optical fibers will inevitably become the transmission medium of metropolitan area networks

(MANs). While the capacity of the rest of the network has been increased, the access network

constructed with twisted wires and coaxial cables has become the bottleneck in providing

larger bandwidth to users. The expansion of the usage of the Internet has already caused an

increase in the demand for bandwidth in the access networks. Although current broadband

access technologies such as digital subscriber lines (DSL) and cable modems are sufficient for

some applications such as Web surfing and email, one can foresee an increasing demand for

bandwidth outstripping the nowadays supply. The Internet enables bandwidth-intensive

applications such as streaming media, video on demand and video telephony applications. In

turn, these applications drive up the demand for larger bandwidth [1, 2]. As far as

transmission capacity is concerned, optical fiber has no competition, and now it is

increasingly extending to access networks. Access networks face more challenges than other

pure transmission networks. Protocols, multi-service capability and cost are the dominant

issues. Optics has a number of advantages concerning the above issues of networking if a

suitable transmission approach is used in the optical networks. Optical code division multiple

access (OCDMA) is one suitable technology for this aim [3, 4].

Code division multiple access (CDMA) technology was developed for wireless military

telecommunications at the end of the 1970 in order to increase the robustness security of the

information transmission [3]. It is also called spread spectrum technique, and has been used

for a long time in military communications to resist intentional jamming and to achieve low

probability of detection. Now CDMA is one of the most useful technologies in the wireless

area, especially for the third generation wireless system [3, 5]. OCDMA scheme is a kind of

CDMA scheme in which code multiplexing and transmission are performed in the optical

domain.

11

OCDMA supports multiple concurrent accesses to the same time slots and the same frequency

range. It is seen as another kind of multiplexing technique besides TDMA and WDM, and it is

a potential candidate for the future optical network, especially for the access network due to

easy access and flexible network architectures [4, 7, 10, 11]. Compared to TDMA, OCDMA

has the advantages to be compatible with the bursty and asynchronous network traffic,

especially in upstream links. The asynchronous multiplexing by OCDMA systems is a very

desirable feature compared to any other digital multiplexing techniques at high data rates.

Optical time division multiple access (OTDMA) has the inherent problem of accreting phase

of each sub-signal whereas with OCDMA transmission capacity is sacrificed. OCDMA has

been considered as a desirable choice for local area networks which are usually implemented

in passive optical network (PON) topology. In the mid 80’s, researches began to focused on

the selection of codewords and network architectures, and since then OCDMA has become a

branch of optical communications.

It is generally believed that the OCDMA enables fast processing of signals in the optical

domain which could help to reduce the complexity of the electronics hardware and software

existing in current communications networks, and could also allow more flexible networks

[7]. However, though CDMA is successful as a communication technology in the radio,

microwave and millimeter wave bands [12], a commercial optical communication system

using OCDMA technology has not yet been realized and research is still in laboratory status.

The few available complete OCDMA demonstration systems actually have not yet

demonstrated the expected advantages of OCDMA. A number of components still needs to be

developed, such as optical sources, en/decoder and broadband optical amplifiers etc. [2, 4].

In this thesis, a new approach is developed by using an electronic en/decoding process in

OCDMA for optical access networks. Based on the study of the results of the current

experimental research and the analysis of the optical beat interference noise, a prototype with

the chiprate of up to 3Gchips/s has been implemented. It offers 3 OCDMA channels for a

variety of data sources with total throughput of 450Mbit/s. The scheme is oriented for

practical usage in the upstream channels of PON access networks. This system has the

following advantages:

a) use of electronic encoder and decoder to avoid the optical interferometric noise produced

by optical encoder and decoder,

b) the bit error rate is reduced by using an electronic logic decoder which has the function of a

hard-limiter,

12

c) easy to be integrated into DWDM system with narrowband optical sources.

This thesis is organized as follows:

Chapter 2 introduces the basic concepts of access networks, the mathematical model of the

traditional OCDMA scheme, and the comparison of OCDMA with wireless CDMA.

Chapter 3 presents a survey of recent experimental results in the area of OCDMA. It was

found that none of existed OCDMA demonstration system is suitable for practical usage and

the OCDMA technology is currently not mature.

In Chapter 4 the noise characteristics of OCDMA are analyzed. The simulation of optical

carrier is carried out taking into account the coherent noise source caused by temperature

wandering. The limitation of OCDMA is then evaluated.

Chapter 5 describes the implementation of an OCDMA system with both optical and

electrical en/decoders for comparison. An experimental 3-channel OCDMA demonstration

system is developed with a capacity of 150Mbit/s at BER 10 and 100Mbit/s at BER 10

for each channel. With simple error correction coding, BER can be improved much beyond

.

6− 12−

1210−

In Chapter 6 the feasibility and the performance of OCDMA combined with DWDM system

are analyzed.

Finally, conclusions are drawn in Chapter 7.

13

Chapter 2 Fundamentals of OCDMA

Traditional ways of separating signals in time (i.e. TDMA) or in frequency (i.e. FDMA) are

simple to make sure that the signals are orthogonal and non-interfering. However, in CDMA,

different users occupy the same bandwidth at the same time, but they are separated from each

other by the use of a set of orthogonal waveforms, sequences, or codes. The purpose of this

chapter is to provide the fundamentals of optical code-division multiple access technology.

2.1 Using OCDMA in Optical Access Network

The access network provides transport bearer capabilities for the provision of

telecommunications services between a service node interface to a service node and each of

the associated interfaces towards the customer premises networks [8]. With the highly

employed technology in the long haul backbone network and MAN, it becomes more and

more important to employ suitable technology in the access network in order to access more

users and offer high bit rate link to end users. Passive optical network (PON), in which

passive optical components are located between a hub and an optical network unit (ONU), is a

recommended architecture for optical access network nowadays [2]. Bus/tree topologies are

envisages in this architecture due to their expected lower maintenance costs [13]. Before the

achievement of the fiber to the home (FTTH) solution, the access network is typically built as

hybrid fiber-coax (HFC) network [1]. In a PON the fiber outgoing from the optical line

terminal (OLN) in the service node is split by a passive optical splitter, where all ONUs are

connected with separate fibers. In this way, the bandwidth and facilities can be shared by all

ONUs to reduce the access costs. In this case, the use of OCDMA in the upstream (from

ONUs to OLT) is superior to the now widely used TDMA system, in the sense that OCDMA

can simultaneously offer many asynchronous channels [15]. If CDMA is combined with

DWDM, the access number would be further increased.

14

PassiveOptical

Network(PON)

Data Source

#1

Data Source

#N

CDMAEncoder

CDMAEncoder

CDMADecoder

CDMADecoder

DataRecovery

#N

DataRecovery

#1Optical

Figure 2.1 A block diagram of an OCDMA system used in PON

Figure 2.1 shows a block diagram of an OCDMA system used in PON. A typical transmitter

for OCDMA is composed of a date source and an encoder. If the data source is electrical, the

electrical CDMA encoder is followed by a laser to provide an optical signal. In case the data

sequence is optical, the CDMA encoder is an optical device implemented with optical signal

processing. The CDMA encoder maps each output bit of the data source into a high rate

sequence of chip impulses. Thus, on the fiber we have an OCDMA signal. At the receiver the

optical impulse sequence is processed for data recovery. There are N transmitter and receiver

pairs (users). Each transmitter has a different CDMA code word, also called signature or

signature sequences. Thus, a common optical fiber can be shared by N users using their own

optical sequence.

2.2 Classification of OCDMA

In OCDMA schemes there is no simple rule of thumb to relate process gain and signal

bandwidth. Any combination of spatial, temporal, frequency and polarization information can,

in principle, be used for OCDMA coding [11]. There are several different approaches to

design encoders for optical schemes. Two commonly used methods will be discussed here.

Spreading: This is a common method to implement OCDMA. At the transmitter, a short

optical impulse, with chip duration T , which is much smaller than the data bit duration T ,

is fed into an FIR filter composed of optical delay lines. The output is a high-rate impulse

train that constitutes the transmitter code sequence. At the receiver, a similar filtering

C D

15

operation is performed on the intensity (incoherent) of the incident signals by an optical delay

line filter acting as a decoder. Only a signal that is matched to the decoder is despread in time,

with a concomitant increase in peak level compared to unmatched signals. The decoder output

signal can be discriminated by a threshold device. This method is also called direct sequence

OCDMA (DS-OCDMA) or on-off keying (OOK) OCDMA. In order to enhance effectivity of

the decoder, coherent matched filtering can be used for restructuring the original impulse. The

principle advantage of the coherent approach is that for an encoder that produces w chip

impulses, the coherently decoded signal has an intensity w times greater than an incoherent

superposition. For coherent decoder, this intensity is a fixed fraction of the total energy

entering the decoder which is independent of w.

Spectral coding: It uses broadband continuous wave (CW) sources and performs

coding directly in the optical frequency domain. In this approach the spectrum is divided into

W slices and encoded by a W-element users codeword. The mapping of the physical mask

pattern (e.g. using a liquid crystal array) to an optical spectral code allows a wide range of

codes to be employed. This approach is also called frequency encoded OCDMA.

The various fundamental approaches can be combined among themselves or combined with

other modulation schemes to form new solutions. For example, fast frequency hopping (FFH)

is a combination of spreading in time and spectral coding [32], and PSK-en/decoding

OCDMA results from the combination of time spreading and PSK scheme [40]. The

commonly called OCDMA system mostly refers to incoherent DS-CDMA.

2.3 Model of DS-OCDMA System

To simplify a concept of a traditional incoherent DS-OCDMA scheme, three assumptions are

made in the following mathematical model:

• Only a IM-DD optical system will be considered, i.e. optical converters use the optical

intensity modulation and direct intensity detection;

• The chip in the codeword (signature sequence) is represented by a rectangular

impulse;

• The optical sources are incoherent light sources.

16

2.3.1 Transmitter:

Here, an OCDMA system with users and on-off keying optical intensity modulation is

considered. Each user transmits binary symbols at a bit rate of

N

DD TR 1= . The binary data of

nth user can be described by

∑∞

−∞=

−=i

DTnin iTtpbtbD

)()( , (2.1)

where is the i th data bit in the bit stream, denotes a rectangular impulse in 1,0, ∈nib )(tpτ

[ )τ,0 with unit amplitude and duration T . To differentiate the signals of the users, each

user is given a unique signature sequence composed of chips with chip duration

D N

CL

CD LC TT = . Specifically, the n th signature sequence is expressed as

∑∞

−∞=

−=l

cTnln lTtpctcc

)()( , (2.2)

where is the th chip in the signature sequence with c 1,0, ∈nlc l nlnLl cc ,, =+ . Let denote

the intensity signal of the th user, then

)(tsn

n

)()()( tctbts nnnn ⋅Ρ= (2.3)

where Ρ denotes the signal intensity. Figure 2.2 shows the generation of the [15]. n )(tsn

DT

'1' '1''0' '0' t

t

tnP

nb ,0 nb ,1 nb ,2

)(tcn

)(tbn

)(tsn

1

1

0

0

0

CT

Figure 2.2 Binary data signal , signature signal and OCDMA output signal

for nth user

)(tbn

s

)(tcn

)(tn

17

2.3.2 Receiver

The signals from all users are multiplexed into a single optical channel and transmitted to the

receivers. By assuming incoherent light sources and an ideal fiber channel with impulse

response )(tδ , the received signal is a direct superposition of the signals, which is

given by

)(tr N

)()()()(1

tptctbtrN

nNnnnnn∑

=

+−−Ρ= ττ (2.4)

The relative transmission delay nτ of the th user is modeled as an random variable that is

uniformly distributed on [ and accounting for the noise from the transmission

channel. For the sake of simplicity, the demodulation of bit b for the first user, during the

time interval [ , is considered. With the assumption of perfect synchronization at

receiver for the first user, we take

n

)t)DT,0 (pN

0

1,0

)DT,0

1 =τ and Dn T<≤ τ0 for Nn ≤≤2 . The received signal

can then be written as

∑=

−− +−−+−+Ρ=N

nNnnnTnnnnnT tptctpbtctpbPtctpbtr

nDnD2

,0,111,01 )()()()()()()()( τττττ

(2.5)

)(txn ny *nd

)(thn

0t

)(tr

Correlator Sampling Decision

Figure 2.3 The block diagram of the receiver for the nth transmitter in OCDMA

Generally, at the receiver of the th user, see Figure 2.3, the n th signature sequence is

correlated with the received signal. The correlator implemented by a filter with time limited

impulse response which equals to zero except for

n

)(thn DTt <<0 . The output signal of the

correlator is a convolution of input signal and impulse response . Hence the

integrand only for

)(tr )(thn

[ DT,0) ∈ ]t( −τ is non zero. So the output signal of the correlator for n th

user is given by

18

∫−

−⋅=∗=t

Ttnnn

D

dthrthtrtx τττ )()()()()( (2.6)

Because of the time limited impulse response of the correlator, only depends on the

received signal in time interval

)(txn

)(tr [ ]tTt D ,− . will be sampled at time instant t and

decided by the threshold device. The sampling occurs at the end of each bit, and the sampled

value at the time point t is

)(txn 0

DT=0

∫

∫

−⋅=

−⋅==−

D

D

T

Dn

t

Ttnnn

dttThtr

dttthtrtxty

0

000

)()(

)()()()(0

0 (2.7)

Now we only pay attention to the first user, and insert (2.5) into (2.7) and get for : 1=n

[ ] )()()()()( '

2,0,1011,001 tpbbPtbPty N

N

nnnnnnnnn +++= ∑

=

+−−

+ τρτρρ (2.8)

where denotes the noise of the decoder output, and are the cross-

correlations between the first user and th user which can be represented by

'Np )(τρ −

n )(τρ +n

n

.)()()()(

)()()()(

1

01

dttThtctp

dttThTtctp

D

T

nn

DDnTn

D

D

−−=

−+−=

∫

∫

+

−−

ττρ

ττρ

ττ

τ

τ

(2.9)

In the Eq.(2.8), the first term is the useful information of the sent signal, the second term is

the interference from the other sent signals (multi access interference, MAI) and the third term

is the result of correlation process of the noise from the transmission channel and other

sources.

The output of the decision device is

<≥

=θθ

)(,0)(,1

)(0

00

*

tyty

tdn

nn (2.10)

where θ is the threshold of the decision, its value depends on the codeword and other

parameters of the system and will be discussed later.

19

Because the correlation filter in the receiver is a matched filter for the corresponding

codeword , the impulse response of the filter can be expressed by the time inverse of

codeword, from Equation (2.2):

)(tcn

)()(1

0,

'cT

L

lnln lTtpctc

C

C

−⋅= ∑−

=

(2.11)

one can obtain:

))1(()(1

, cT

L

lnlLn Tltpcth

C

C

C−−⋅= ∑

=− (2.12)

For the traditional OCDMA-system only unipolar codewords are used which will be

discussed in the following .

2.4 Code of DS-OCDMA System

The choice of the codewords, the signature sequence, is a key problem to any successful

CDMA scheme, either electrical or optical. For an OCDMA system, in which multiple users

access the same medium using the same time interval and frequency band, the transmitted

concurrent data streams will inevitably produce multi access interference (MAI). In order to

reduce the effects of this interference, the code should meet specific conditions for auto- and

cross-correlation. The proper selection of codewords is more important for incoherent system

than for coherent system, because incoherent system has a higher MAI power level.

The bipolar codewords used in wireless communication play a little roll in non-coherent

OCDMA. This is because the optical signal is equivalent to the instant power which is

nonnegative. This means, that codes based on +1/−1 signals, which are used in wireless

CDMA system, cannot be applied in optical system. Thus for OCDMA, a new code criterion

needs to be designed. A good optical CDMA code has much more 0’s than 1’s in each

codeword, while a well-correlated (+1/−1) sequence typically has about the same number of

+1’s and –1’s. Up to now various code families for OCDMA have been suggested, from one-

dimensional (1-D) codes encoded in time-domain to the two-dimensional (2-D) codes

encoded both in time and wavelength [21] or time and space [30] domain. The code families

include Gold codes [20], symmetric quasi-prime codes [19], quadratic congruence code [17,

18], block-multiplex-code [21] and optical orthogonal code (OOC) [16]. The use of OOCs

enables a large number of asynchronous users to transmit information. Thus the lack of the

20

stringent requirement of network synchronization enhances the flexibility of the system. Now

the OOC is an important candidate for OCDMA used in access networks.

Optical Orthogonal Code: An optical orthogonal code (OOC) is a family of (0,1)

sequences with relatively good properties of auto- and cross-correlation. The auto-correlation

of each OOC codeword exhibits the “thumbtack” shape and the cross-correlation between any

two OOC codewords remains low at all time. The thumbtack shape of the auto-correlation

facilitates the detection of the desired signal, and the low cross-correlation reduces MAI in the

network. Here we give the definition and some fundamental properties of OOC. An optical

orthogonal code ),,,( caC wL λλ is a family of (0,1) sequences with length and weight w ,

which satisfies the following two properties [16]:

CL

1) The auto-correlation property:

∑−

=+ ≤

1

0

CL

tatt xx λτ (2.13)

for any and any integer 1,0∈tx τ , CL<< τ0 .

2) The cross-correlation property:

∑−

=+ ≤

1

0

CL

tctt yx λτ (2.14)

for any and any integer 1,0∈≠ tt yx τ .

In general, for , i.e. for a given integer code length and weight w, where

, and

)1,1,,( wLC

1 =a

CL

)1( −≤− CLww 1=cλλ , one can construct at most N codewords of OOC, the value

of N is upper bounded by

1,)1(

1≠

−−

≤ www

LN C (2.15)

where the symbol [ denotes the integer portion of the real value x. In the following, we only

consider codes with .

]x

( )1,1,, wLC

The Bit Error Ratio of OOC: The BER of OOC can be exactly calculated if we

assume that the chips are synchronous in the system between the users (for asynchronous

chips see [22]). At the receiver, we have multiple overlap. The output of the decision device is

given by Eq. (2.10). If the number of the impulses is equal to or higher than threshold θ , the

decision device outputs a “1” (Mark), otherwise a “0” (Null). If we assume that no random

noise is present in the system, then the false detection of a Mark as a Null is not possible

21

because, when a Mark is transmitted, the threshold is then definitely exceeded (θ must be

smaller than code weight w) and therefore a Mark is detected. But if the information bit is a

Null, it may be mistaken and detected as a Mark when MAI is present in the system.

Let us consider the detection of an individual user n. An impulse from another user sensed by

detector n is called a hit. When user n is sending Null, a false detection occurs when there are

θ or more hits. The cross-correlation property ( 1=cλ ) ensures that each fortuitous user

contributes at most with one hit. The probability of a hit from another user equals to CLw2

[22]. The bit error probability is given by [16]

ji

C

j

C

i

j

N

i

N

r

N

ire

Lw

Lw

ji

iN

sendingareusersotherietcectiondfalsePsendingareusersotheriPP

−

=

−

=

−

−

=

−

⋅

−=

⋅=

∑∑

∑221

1

11

2

1|121

θθ

θ

(2.16)

Algebraic manipulation gives the following alternative formula:

∑−

=

−−

−

−=

1 122

21

21

21 N

i

iN

C

i

Ce L

wL

wi

NP

θ

(2.17)

Obviously, when ( and wN <− )1 w=θ , 0=eP . The BER can be reduced by using the hard-

limiter before the decoder, this will be discussed in detail in Chapter 4 and Appendix A1.

The drawback, however, is that the number of different codewords, i.e., the number of users

in the system, is limited. The number of users that can be accommodated in an OOC system is

upper bounded by Eq. (2.15), given that the upper bound of to support required data rate.

This puts an upper bound on the number of impulses w that can be used for a given number of

users. So there is a tradeoff between the number of users and BER in system design.

CL

2.5 Parameters of OCDMA

A comparison between wireless CDMA and OCDMA is given in the Table 2.1. The main

difference lies in frequency of the carrier, i.e. the frequency of lightwave in OCDMA is much

higher than that of RF in wireless CDMA, and the phase of lightwave can not be well

controlled, and furthermore there is no ±1 for OCDMA. Though bandwidth in the optical

domain is much wider than that in wireless telecommunication, the spectral efficiency

22

remains nevertheless one of a number of important measures of system performance for high

speed network.

Table 2.1 Comparison of OCDMA with wireless CDMA

OCDMA Wireless CDMA

Carrier Lightwave

Phase can not be well controlled

Micro- and millimeter wave Limited availability

Spreading/despreading Time/Wavelength domain Time/Frequency domain

Direct sequence SS, Wavelength hopping

Direct sequence SS Frequency hopping

Code

Fewer Mark in codeword Balanced Mark and NullEn/decoding Optical domain RF domain or baseband Capacity Soft (on demand). Interference limited

Transmission medium Closed-space (optical fiber) Dispersive Low attenuation Nonlinear

Free-space (air) Multipath Large attenuation Linear

Interference suppression Open question Multisectored antenna Multibeam reception Synchronization Power control

Problems in propagation Dispersion effect Nonlinear effect Interferometric effect

Near-far effect Multipath effect Fading

In principle, OCDMA has many advantages like its counterpart in wireless. Compared with

TDM and WDM scheme, OCDMA also has the following advantages:

• Using optical processing to fulfill certain network applications such as addressing and

routing;

• Low delay access suitable to bursty local-area network traffic, effective allocation of

bandwidth responding to requirement, variable bit rate;

• High capability of data throughput and accessed users;

• Flexibility of the network architecture;

• High speed coding and decoding process;

• Asynchronous data transmission simplifies network planning, management and

control;

23

• High security of data and network.

The advantages of asynchronous transmission and the capability of multiple-access in a bursty

environment make OCDMA attractive for LAN applications. Hitherto, the capabilities of

OCDMA are not completely demonstrated, because science and technology are not yet fully

developed. The purpose of this thesis is to make a significant contribution.

24

Chapter 3 Experimental Results for OCDMA

There are many theoretical investigations on OCDMA, such as the selection of the

codewords, the study of various OCDMA structures and networks etc., but only a few

experimental studies can be found. The experimental research in the area of OCDMA began

in the mid 80’s with incoherent DS-CDMA system utilizing incoherent superposition of

optical intensity. In the mid 90’s, focus was on the coherent DS-CDMA which requires that

the optical delay has to be controlled within the scale of the optical wavelength in order to

achieve the correct phases of superimposed fields, and on frequency encoders (FE) which

operate in wavelength domain. With the development of the material and mechanical

technique, some new OCDMA schemes emerged, such as fast frequency hopping (FFH),

phase shift keying (PSK) OCDMA, etc.. Meanwhile non-digital OCDMA schemes have also

been proposed and have been experimentally investigated, and the results show that the non-

digital scheme is even better than others. There are many parameters to describe the

performance of an OCDMA system, such as bandwidth utility, chip period, bit rate and

number of users, etc.. In this chapter we briefly describe the current developments of

OCDMA technology through the experimental results in this area.

3.1 Incoherent Direct Sequence OCDMA

For incoherent DS-OCDMA, the optical processing at the receiver operates with the optical

intensity. According to the structure of the optical encoder/decoder, the early implementation

of OCDMA systems can be divided into two types, i.e. tapped delay lines and ladder networks

(the latter only for same special codewords).

In 1986, P. R. Prucnal et al. [6] conducted the first experiment of OCDMA. The experimental

set-up is shown in Figure 3.1. The data is encoded by the code generator with the appropriate

prime code sequence (in fact, electronic processing was used in the system). The prime

codeword is composed of P=5 (i.e. the number of “1” in codeword is 5) and 32 chips per bit.

The chip rate is 100Mchip/s, the data rate is 3.125Mbit/s. The encoded data is injected into a

multimode fiber using a Laserdiode. At the receiver, the incoming signal is split into 5 signals

25

at first. Then the optical correlator selectively delays the signals before recombining them.

This processed signal impinges on an avalanche photodiode. In 1987, Salehi et al. also

demonstrated a two-transmitter OCDMA-system with optical tapped delay lines decoder [48].

Holmes and Syms implemented an optical processing using ladder networks and measured the

auto-correlation and cross-correlation [19]. At the Institute of Telecommunication, University

of Stuttgart, a demonstration system with electronic encoder and optical decoder has been

implemented. It has four users with a chip rate of 720Mchip/s [15].

D ata G enerator

&B ER Tester

C odeG enerator

LaserDriver

PulseG enerator

ThresholdD etector

A PD andPreA m p

C om biner

O ptical D elay Line

SplitterStar

C oupler

R eceived D ata

Transm itted D ata

C oded D ata

O ptical Signal

Receiver

Figure 3.1 The first experiment for the OCDMA system (see Figure 7 in [6])

Most of the incoherent DS-OCDMA systems were used to demonstrate the feasibility of the

conception of OCDMA. In these early systems, broadband optical sources were used, and in

some cases, multimode fibers were used. To avoid large optical loss with optical processing,

the electronic processing was used only either at encoder or at decoder.

3.2 Coherent DS-OCDMA

With coherent DS-OCDMA, the optical impulse sequence from a low-coherence source can

be overlapped coherently at the decoder, offering further possibilities for system design [37,

38]. The decoder employs a fiber-optic ladder network. Figure 3.2. shows a general n-stage

encoder and a matched decoder. Both the encoder and decoder consist of n delay stages with

delay time iτ . In each stage a single elliptical retarder )( iiR Ω describes the differential

26

birefringence between the two arms. With delays chosen such that ∑ −

=> 1

1

n

i in ττ , an n-stage

encoder transforms an input impulse into a sequence of n2 impulses with equal intensity.

Each impulse represents a chip of the codeword. When the impulse sequence is decoded by an

identical network, impulses which have traversed alternative paths through the system with

the same total delay will be superimposed. Coherent summation can be achieved if delay time

i'τ of each decoder is matched to the corresponding encoder delay iτ such that cii τττ <− ' ,

where cτ is the coherence-time of the source. It is also necessary to ensure that all output

impulses have the same state of polarization, which can be achieved by controlli ng the

birefringence of one arm of each delay stage. Consider birefringence in the delay networks as

shown in Fig 3.2, where the differential retardation of stage i is denoted by an operator iR of

rotation magnitude iΩ on the Poincare sphere.

)( 1Ω1R

1τ

)( 22 ΩR

2τ

)( nnR Ω

nτ

)'(' 1Ω1R

1'τ

)'(' 22 ΩR

2'τ

)'(' nnR Ω

n'τ

LD

PD

Fiber downlead

Encoder

Matched Decoder

pulsesn2

Figure 3.2 The DS-OCDMA system with optical ladder networks (see Figure 1 in [37])

Without external control, the optical phase difference between interfering terms is random and

slowly drifts as optical path lengths fluctuate with ambient temperature changes. Hence

decoder must include the equipment for phase adjustment, although optical phase of

lightwave is diff icult to follow with the disturbance of temperature. Only two approaches

were proposed to solve this problem.

27

In [38] a mechanic equipment is introduced, that can statically adjust the optical phase. A

specially designed fiber stretching device with fine adjustment is used by winding part of one

of the 22m delay lines on two sets of pulley fixtures, pulling one of the one-axis movable

fixture by means of a micrometer head and a long arm lever while the other fixture is fixed in

the experiment (see Figure 3.3). In [37] an active phase controller is demonstrated. Thermal

phase modulators consisting of several 10cm of metal-coated fiber were used to provide the

phase control. These devices utilize temperature-induced refractive index changes in the fiber

core to achieve large phase excursions. The feedback circuit for servo-control of the

differential optical phase is shown in Figure 3.4. Here only two stages in decoder are

demonstrated. Only the outputs of the decoder were given in these two reports without giving

the results of BER of data transmission for the systems.

pivot

Lever

Pulley Fixture

MicrometerHead

Optical Fiber

Figure 3.3 The structure of the mechanic phase adjusting equipment (see Fig. 11 in [38])

28

OpticalSignal

PM PMPD

LIA

+

+

LIA

1ω

1ω

2ω

2ω

Optical Fiber

Electrical Cable

PD: photodiode LIA: lock-in amplifier PM: phase modulator

Figure 3.4 The structure of phase following with feedback circuit (see Figure 2 in [37])

Nowadays the phase control technology in optical decoder is still a problem for coherent

decoding method. The frequency of lightwave is so high that large phase noise will be

produced by the thermal disturbance in optical fiber or optical waveguide. It is very difficult

to use these methods in a practical system, especially when the en/decoder have more than

two stages, where the feedback signal is difficult to be found for each stage.

3.3 Frequency Encoded OCDMA

The coding of the frequency-encoded CDMA (FE-CDMA) system is done in the frequency

domain. A frequency encoder for optical source is shown in Figure 3.5. It consists of a pair of

diffraction gratings placed at the focal planes of a unit magnification, confocal lens pair. The

first grating spatially decomposes the spectral components present in the incoming optical

signal with a certain resolution. A spatially patterned mask is inserted midway between the

lenses at the point where the optical spectral components experience maximal spatial

separation. After the mask, the spectral components are re-assembled by the second lens and

second grating into a single optical beam. The mask can modify the spectral components in

phase and/or in amplitude, depending on the coherence property of the incident optical

source. The number of frequency bands that can be resolved by the encoder will dictate the

code length, and the number of subscribers in the system. That number is given approximately

by

29

)cos(2 rdN

θπ

λλ

⋅Β⋅

⋅∆

≈ (3.1)

where λ is the center wavelength of the optical source, λ∆ is the spectral width being

encoded, is the input beam radius, is the grating period and Β d rθ is the diffracted angle

for the central wavelength. For the receiver, considering complete asynchronous operation

between the users, only the periodic correlation parameters are of interest, since the frequency

slots of the different users will always be aligned.

Source

Data

ToNetwork

DiffractionGrating

Mask (Code)

Figure 3.5 A frequency encoder for optical source used in FE OCDMA (see Fig.1 in [31])

Let and ( be two (0,1) OCDMA codewords, which

are used by two users. The cross-correlation is

),,,()( 21 NxxxX L= ),,,() 21 NyyyY L=

∑=

+=N

ikiiXY yxkR

1)( (3.2)

Define the complement of codewords of by )(X )(X whose elements are obtained from (

by

)X

ii xx −= 1 . The periodic cross-correlation sequence between )(X and ( is similarly )Y

∑=

+=N

ikiiYX yxkR

1)( (3.3)

When the sequences that suit for )()( kRkR YXXY = are used, a receiver that computes

)()( kRkR YXXY − will reject the interference coming from user sequence ( . Normally, m-

sequences or Hademard codes are used.

)Y

30

Mask

LED

ASKData

K x KStar

Coupler3dB

coupler

A(w)

A(w)PD1

PD2

LPF +_

ASKData

A(w)

Figure 3.6 The block diagram of experimental set-up for FEOCDMA (see Figure 2 in

[31])

M. Kavenhrad and D. Zaccarin demonstrated an OCDMA system based on spectral encoding

of noncoherent sources [31]. The block diagram of the experimental set-up is shown in Figure

3.6. The m-sequences are used in this system. There are two photodetectors for detecting

and ∑ +i

kii xx ∑ +i

kii xx , respectively, and the outputs are then subtracted to reduce noise. The

mask employs liquid crystal display (LCD) technology, which functions as a programmable

encoder. However, only the simulation results of the BER for a number of active users were

presented, without giving the experimental result for real data transmission. Sardesai et al.

[39] implemented a test bed using the same method, however, femtosecond impulses were

used for encoding-decoding. Since the signal operates at femtosecond level, this system

requires a transmission fiber with very low dispersion, and it is not practical for the optical

network.

One advantage of using FE-OCDMA is the capability of suppression of the interference from

the other users. However, there are some fatal disadvantages involved: a) requiring a very

large bandwidth of optical source (e.g. 60nm), b) the data rate and transmission distance are

interrelated, allowing no flexibility for distributed accessing, c) the encoder/decoder is pure

optical equipment, notable insertion losses are resulted by coupling of optical signals into

fiber system.

31

3.4 Fast Frequency Hopping (FFH) OCDMA

Fast frequency hopping (FFH) codes in both time and frequency domain, which offers a great

flexibility in the choice of optical codewords. The design of the encoders/decoders is critical

to the performance of the FFH-OCDMA. Time and frequency spreading using pass-band

filters and delay lines were first studied by Tanceveski and Andronovich in 1996 [32]. Since

then, many hybrid (time and frequency) encoders have been considered including arrayed

waveguide grating (AWG) cascaded by parallel optical delay lines [33] and fiber Bragg

grating (FBG) [34, 35].

Lithium NiobateMach-Zender

10 Gbit/s patterngeneratorOptical spectrum

80 kmSMF-28

EDFA

EDFA

EDFA

EDFA

CommunicationsSignal Analyser

DCF

Tunable decoder

Translation stage

EDFA

EDFA

EDFA/ASE

(3.5 dBm)

1547 1551 1555 1559

)(nmλ

-65

-55

-45

-35

Pow

er (d

Bm

)

1 x 2

1 x 2

1 x 8

1 x 8

511362941 λλλλλλλλ

8146951274 λλλλλλλλ

91571061385 λλλλλλλλ

12181013916118 λλλλλλλλ

131911141017129 λλλλλλλλ

1622141713201512 λλλλλλλλ

2127191218252017 λλλλλλλλ

2430221521282320 λλλλλλλλ

1821261913202822 λλλλλλλλ

Figure 3.7 The FFH demonstration system using Bragg grating arrays (see Figure 1 in

[36])

Figure 3.7 shows a new FFH demonstration system using Bragg grating arrays (BGA) given

by H. Ben Jaafer et al. [36]. In this design each codeword is an ordered selection of eight

frequencies chosen among a set of 30 frequency bands spaced by 50GHz. The individual

grating length is 14mm with 1mm spacing and the total length of the BGA en/decoders is

11.9cm. The chip duration is 140ps determined by the Bragg grating spacing. The reflectivity

and band width of each individual grating are typically 13dB and 20GHz, respectively. For

32

the transmission experiment, amplified spontaneous emission (ASE) of an erbium-doped fiber

amplifier (EDFA) is used as an incoherent broadband source. An electrooptical modulator is

used to generate the data stream forming a pattern generator with 100ps impulse width.

Because only one modulator and one data generator are used, the modulated impulses are split

into 16 encoders by passive components. The encoded signals then propagate through 80km

single mode fiber (SMF) and dispersion compensating fiber (DCF). At the receiver end, the

signal is fed into a strain tunable decoder. The decoded signal, consisting of the sum of the

desired user signal and the interference contribution, is evaluated in a communication signal

analyzer.

FHH-OCDMA system utilizes multiple frequency points, the chip impulses with different

frequencies of optical carriers will travel with different velocities due to the fiber

characteristic, and the relative positions among them will be changed. Consequently, the

decoder cannot correctly recover the desired bit without dispersion compensation, which is

still unusual in LAN. In addition, the utility-rate of the bandwidth of demonstration systems is

much lower than that by using DWDM technology. Furthermore the demonstrated systems

are synchronous, and a narrow frequency band used in FHH-OCDMA codeword is normally

located in a different position for different codewords. When the number of users increases,

the interferometric noise increases which is caused by the overlap of two beams having the

same frequency, similarly as in DS-OCDMA system.

3.5 PSK-OCDMA

With the developments of the fiber Bragg grating technology and planar lightwave circuit

(PLC), it is possible to use the optical phase for coding. A series of experiments of

BPSK/QPSK OCDMA have been carried out at University of Southampton in UK [40, 41].

Strain-tunable fiber Bragg gratings (called superstructured fiber Bragg gratings (SSFBG)

were used as phase encoder/decoders. SSFBG is a single-grating structure having a slowly

varying amplitude and/or a phase pattern (superstructure) imposed upon a uniform

background refractive index modulation. In the weak grating limit (reflectivity < 20%), the

shape of the impulse responses are directly following the shape of the spatial superstructure.

Thus, OCDMA codes can be written directly into SSFBG, such that short impulses reflected

from the structure are shaped into the codewords. Quaternary phase coding with code length

of 255 chips and chip rate of 320Gchip/s is demonstrated [40]. A typical experimental set-up

is shown in Figure 3.8. Here the signals for CDMA is from a single source, therefore it is not

33

sure if the result is the same as for two or more signal sources. Because when the signal

source is the same, the signal pattern of each channel is also the same, therefore there is no

noise from other channels at the same wavelength if a Null is transmitted.

Wavelength Converter

EDFAMOD

1.25 Gbit/s Data in 255-chip QPSK SSFBGs,1λ

Q1

Q2

B3

B4

EDFA

Q1*

Diagnostics

255-chip QPSK SSFBGs,1λ

EDFAEDFA

EDFA

Filter

63-chip BPSK SSFBGs

CW Laser1556.5 nm

1λ

2λ

EFRL1552.5nm10 GHz

,2λ

Figure 3.8 The system using superstructured fiber Bragg gratings (see Figure 3 in [40])

Figure 3.9 shows the experimental set-up of this research group, given in [41]. There are four

optical sources modulated by 10GHz signal which generate continued optical impulses with

20ps width. After the optical amplifier, the impulse trains are modulated by using a single

LiNbO3 intensity modulator to obtain a data sequence at a rate of 311Mbit/s. The modulated

impulses are then reflected from an array of 16 coding gratings to generate 16 simultaneously

coded data signals. The four level phase shift keyed OCDMA codes are derived by reflection.

The use of the continuous grating writing technique within a single uniform phase mask

allows for the excellent control of the amplitude and phase of the refractive index modulation

along the grating length. In this experimental system, besides the use of the same signal

source for all channels as mentioned before, the transmission data rate is so low that the four

channels with the same optical wavelength could be combined in a TDM scheme, i.e. the

optical impulses might superimpose without overlapping. In this case, the arrangement is not

a typical OCDMA scheme.

34

DFB FiberLaser

1567.32 nm

DFB FiberLaser

1548.11 nm

TunableLaser

1548.91 nm

DFB FiberLaser

1548.11 nm

10 GHzClock

EAM

EDFA EDFAMOD

311 Mbit/SData

Q1

Q2

Q3

Q4

A

C

B

Wavelength (nm)

Wavelength (nm)1546 1547 1548 1549 1550 1551

155115501549154815471546 1546 1547 1548 15491549 1550 1551

Wavelength (nm)

B

C

A

Strain tunable Q1*

BERT

Strain tunableQ2*

BERT

BERT

BERT

StraintunableQ3*

Strain tunable Q4*

EDFA EDFA

50km SMF-28

fiber

Pow

er (1

0 dB

/div

)Po

wer

(10

dB/d

iv)

Pow

er (1

0 dB

/div

)

4321 λλλλ

4321 λλλλ

4321 λλλλ

4321 λλλλ 41 ~ λλ

41 ~ λλ

41 ~ λλ

41 ~ λλ

Figure 3.9 The 16 channels OCDMA system using SSFBG (see Figure 1 in [41])

A monolithically integrated all-optical PSK encoder/decoders consisting of tunable tapped

delay-lines with thermo-controlled optical phase shifters was fabricated by the planar

lightwave circuit (PLC) technology at Communication Research Laboratory, Japan. The

schematic structure of the all-optical encoder/decoder is shown in Figure 3.10. In the PLC of

en/decoder, the tap ratio and the phase are both tunable through controlling the electric current

fed into each heater of phase shifters. Electric currents of all heaters are monitored by GPIB

controlled PC. However, there is no feedback control to adjust the temperature. In order to

keep the phase stable against the environmental fluctuation, PLC is encapsulated in a

temperature-controlled box. In order to measure the optical phase for each chip, optical

interferometer and spectrum analyzer are used. The interferogram and spectrum of encoded

signal are simulated and compared with measured value. If the measured value is not correct,

the phase shifter is controlled to fit with the simulated value.

35

Figure 3.10 The schematic structure of the all-optical en/decoder (see Fig. 6 in [51])

The typical scheme using this component (or equipment) in OCDMA system is presented in

[53]. The system setup is show in Figure 3.11. A 10GHz repetition rate, 2.0ps impulse stream

at 1555nm is generated by a hybrid mode locked semiconductor Laserdiode (MLLD). It is

divided into three impulse streams and each of them is externally modulated using a 10Gbit/s

pseudo-random bit stream. The taps of the encoder are tuned so that they split i nto an eight-

impulse sequence with equal amplitude. The carrier phase of each tapped impulse is binary

changed by 0 or π . Then an optical code of 8-chip BPSK impulse sequence with a chip

interval of 5ps is generated. After dispersion compensated transmission, the optical decoder

with the same structured matched filters receives the signals in the time domain, followed by

time-gating. In [52] the camera traces of the three combined signals have not been given,

therefore, whether the decoded signal is disturbed by other two signals remains unknown.

Since the duration of one data bit is 100ps, and an encoded bit only occupies 40ps, it is

possible that the desired signal isn’ t disturbed by other signals. Thus the system did not

explicitl y demonstrate the MAI in PSK-OCDMA. It is claimed by the same researchers in

[42] and [43] that a 160Gbit/s OCDMA/WDM (10Gbit/s 4CDMA 4WDM) transmission

with a novel side-lobes suppression detection transmission and a 6.4Tbit/s OCDM/WDM

(4OCDM 40WDM 40Gbit/s) transmission were achieved. However, by using a system

with only one signal source for all CDMA users and WDM channels, the above mentioned

results are unlikely.

36

EDFA

1 st span73.5 km

10 Gbits/s3 OCDM

Optical encoder/decoder

OPCgenerator

2nd span76.5 km

Opticaldecoder

Opticalencoder

Timing-gating PD BERT

phase shiftertunable tap 5 ps delay

OpticalencoderOpticalencoder

Opticalencoder

10 Gbits/s

EOM

EOM

EOM

10 Gbits/s

10 Gbits/s

10 GHz

Filter3 nm

MLLD1555 nm

OPC gnerator

10 Gbits/S 3OCDM

Pump LD1550 nm

MLLD1540 nm

Filter3 nm

EDFA Filter3 nm

Filter3 nm

Filter3 nm

EDFA

Filter15 nm

EDFA

EDFA Filter 3 nm10 GHz

Time-gating

SOA

SA

Figure 3.11 The system structure of the 3 channels with 10Gbit/s (see Figure 1 in [52])

The PLC en/decoder is much more stable than the fiber-optic interferometer, because the

device is temperature-controlled, and each optical path length is always kept constant.

However, phase noise was observed during experiments which severely deteriorates the

system performance. This OCDMA scheme with super high transmission bit rate combining

TDM and WDM could be used in point to point links for backbone networks. Hitherto,

whether optical PSK can tolerate the interference noise from the multiple optical beams

overlapping is still a question, and the research still continues.

3.6 The Other Methods

In general, for an OCDMA system the codeword is digital, which can be expressed by a

digital model and handled with digital processing. However, there are some experiments in

OCDMA research field using “non-digital codes”, where the differentiation of the users is

achieved by adjusting some continuous variables. It is interesting to note that the best

experimental results of OCDMA come from the non-digital methods.

37

Using Filter with FSR: The research group of Alcatel has implemented a series of

demonstration systems of OCDMA using Mach-Zehnder (MZ) filters and Fabry-Perot (FP)

filters with Free Spectral Ranges (FSR) in the range of 10–20GHz [44]. The applied

technique is a specific form of coherence multiplexing or spectral slicing using optical filters

with periodic power transfer functions. The magnitude of the source spectra (using LED,

several 10nm) are modulated by the data and subsequently coded by such filters. According to

the difference of the period of the power transmission characteristics of filters used in each

channel, the signal of an accessed OCDMA channel is obtained. Figure 3.12 shows the

experimental set-up.

combiner transmissionlink

EDFASMF

data in

data inTx #1

Tx #8 LED

LEDFSR#1

FSR#8

#1

#2

#3

#4

#5

#6

#7

#8

DCF

isolator 3dBFabry-Perot

TIAdB

data out

tunable balanced receiver

transmitters

Figure 3.12 The experimental set-up using FSR filters (see Figure 1 in [44])

The periodicity of the transfer function for these filters is given in terms of the free spectral

range (FSR) τ/1 , where τ is round trip time of the filter. The round trip time is determined

by the filter geometry and given as cLngrMZ )( ∆=τ for MZ filters and cLngrFP )2(=τ for

FP filters, where and are the differential delay and the cavity length of the filters,

respectively, is the group index of the cavity material and is the vacuum speed of light.

In the network different FSR are allocated to different optical transmitters and define the

L∆ L

grn c

38

codes for the system. At the receiver an optical filter with periodic transfer function is tuned

by matching its FSR to the desired channel. The filter types at transmitter and receiver do not

have to be identical, only the FSR or the round trip times Txτ and Rxτ , must be matched (see

Figure 3.13).

ckτ2

Txτfilter tuning,

0/1 v

transmitter filter

receiver filter

+source spectrum

optical frequency0v

ckv τ/=∆

Rxτ

rel.

optic

al p

ower

Figure 3.13 Spectra used in system (see Figure 2 in [44])

In practice the gain spectra of the optical amplifiers define the coherence time, since usually

they are narrower (30nm) than the source spectra (50nm). The filter FSR in this system (MZ

at the transmitters, FP with finesse 4 at the receiver) are chosen in the range of 10–20GHz, the

power extinction of the transmitter filters is 10–13dB and slightly polarization dependent, the

FP receiver filter extinction ratio is only 9dB. Nevertheless the system operated with bit error

rate . The second benefit of periodic encoding is that only the difference between

filter roundtrip times

1010−≤BER)()( k

Txj

Tx ττ − for any two transmitters needs to be larger than some

picoseconds. The absolute value of Txτ is not important as long as it is much larger than cτ .

This demonstration system showed best performance using OCDMA technology in terms of

stability and practice, for which a field test with a transmission distance of 20km has been

done. One of the disadvantages of this scheme is the low bandwidth efficiency. To reduce the

source coherence time cτ , all of the bandwidth which EDFA offers is used. And such a broad

bandwidth normally requires dispersion compensation.

Path-difference multiplexing: This method is known as coherence multiplexing,

and is generally not categorized as an OCDMA scheme [4]. However, since it makes use of a

coherent matched filter and has the features of CDMA (e.g. sharing signal bandwidth,

asynchronous operation), it will also be introduced here.

39

τ =239ps

τ =228ps

ENC 3

τ =216ps

ENC 4

ENC 2

τ =245ps

ENC 3

τ =239ps

DEC 3

MOD

Data

4x1 starcoupler

Bandwidth of opticalsource is 17nm at1550nm band.

8 km DSF

ENC: encoder

DEC: decoder

MOD: modulator

: polarization controller

Figure 3.14 The implementation of path-difference multiplexing (see Figure 1 in [49])

Figure 3.14 shows a path-difference multiplexing system used in a CDMA broadcast network.

Encoder and decoder consist of interferometers with path delay imbalances substantially

greater than the coherence-time cτ of the sources. The differential delay in each

interferometer is used to address the channels. Digital data is applied to an encoder arm by

2π± phase shift keying (PSK) using a phase modulator. This does not produce an intensity

modulation at the output of the encoder since the fields from the two paths combine

incoherently. Decoding is achieved by matching the differential delays of the decoder, T ,

and of the encoder, T , to significantly less than the source coherence-time, i.e.

de

en

cende TT τ<<− . Fields from two of the four available paths from the input to the output are

then correlated. These two fields, one of which has been phase modulated with the data,

interfere coherently and the data is recovered as an intensity modulation. When the delays of

the decoder and encoder are unmatched by more than several coherence-times, i.e. when

ende TT − is greater than a few times cτ , the fields remain substantially uncorrected and only

little coherence is observed. One can use differential detection of both decoder outputs as

balance detection to reduce the beat noise.

The differences between all encoder delays must significantly exceed the source coherence-

time to prevent crosstalk between channels. It is desirable to use a source with a broad

linewidth since this allows short interferometer delays to be used and also reduces the

40

detected optical beat noise. Apart from practical constraints placed on the power budget by

available source power, receiver sensitivity, and splitting losses, the number of users that can

be supported at a given bit rate is severely limited by the optical beat noise generated in the

network. The highest capacity demonstrated is 4Gbit/s in a four-channel coherence-

multiplexed system, with each channel having a capacity of 1Gbit/s by G. Pendock and D.

Sampson [49].

For this method the transmission with high data rate and multi-users can be achieved only

with larger bandwidth of the optical source i.e. shorter coherence-time. The SNR limited by

the optical beat noise is given by [27]

( ) 22 41

1241

NBNNBSNR

rcrcOBN ⋅⋅

≈++⋅⋅

=ττ

(3.4)

where N is the number of users and is the receiver bandwidth. By considering the distance

of transmission and dispersion of fiber, an example is given in [4]: for a system operating at

1550nm on standard single mode fiber with

rB

kmnmpsD //16= , a 60nm linewidth of optical

source will be requested to support 50 users with 40Mbit/s over 12km distance. This

linewidth is broader than the width of WDM system at 1550nm wavelength, which have more

than hundred DWDM channels with 50GHz bandwidth.

Conclusion: The signatures of users are analog compared to digital codewords used by

other methods, and this is unique in the research field of OCDMA. Those systems can

transmit asynchronous data. This is important for LAN or MAN accessing. Because of the use

of broadband optical sources (several 10nm) which have very small coherence time, the

coherence noise is thus reduced. On the other hand, as broadband optical sources are required,

the bandwidth is lower and thus cannot compete with DWDM.

3.7 Some Methods for Improving the Performance of OCDMA

By the experimental researches, the performance of the demonstrated OCDMA system is not

yet good enough as it is expected from theoretical analysis. In order to reduce the interference

noise in OCDMA system, some methods were proposed such as the use of time-gating

detection, hard-limiter and special decoding with balanced detection. These methods

effectively enhance the transmission performance of the most kinds of afore discussed

OCDMA systems.

41

Hard-limiter: The hard-limiter is an effective component to mitigate the channel

interference caused by other users, so that the probability of bit errors is reduced or the

number of simultaneous users can be increased. The hard-limiter can be placed at the front

end of the desired receiver to reduce the effects of the MAI. An optical hard-limiter functions

like a flip-flop. An ideal optical hard limiter is defined as [23]:

≤≤≥

=10,0

1,1)(

xx

xg (3.5)

where 1 is the normalized optical light intensity. The hard-limiter can also be used in

OCDMA system by using time spreading/wavelength hopping (2-D) codes. The optical hard-

limiter is expensive nowadays, a cheaper way is using the logic electronic decoder which also

has the function of the hard-limiter (see section 4.3.3).

Timing Gate: In many OCDMA schemes, data is transmitted in form of a high rate chip

sequence which consists of impulses much shorter than a bit duration. The matched decoder

at the receiver despreads the signal by reconstructing the original bit impulse, whereas

unmatched signals remain spread in time. In a given bit interval, information from the desired

transmitter is contained ideally in a single chip duration T containing the reconstructed

impulse, whereas multiple-access noise is spread randomly over the full bit interval T . A

time gate samples only the signal-bearing fraction of each bit, and rejects signal parts that fall

outside the gated interval. The SNR improvement under time gating is given by the ratio of

the bit interval to the time window

C

D

CD TT β , where β is the number of chips for which the

gate is open. Time gating is effective in reducing any noise in the received optical signal

including multiple-user shot noise, crosstalk and optical beat noise. This method has the strict

condition that the exact gating time slot must be given. As the gating clock can not be

recovered from data stream or from synchronous transmission, this method is not feasible for

access networks.

Balanced Detector: For some systems, the optical decoder is composed of 22×

optical couplers which provides two inverse output signals in case of coherent decoding. This

property can be used for balanced (differential) detection to improve the performance of the

discrimination. The two signals for the desired user have a large difference, while the

difference is small for MAI. Thus, the differential detector has a common mode rejection

ratio. If the two inverse signals are subtracted, this improves the signal-to-noise ratio (SNR)

42

by canceling optical beat noise. The level of achieved cancellation depends on the common-

mode rejection ratio of the balanced receiver. This method is often sufficient to remove the

error rate floor, allowing operation at low BER. For the optical method described in sections

3.2 and 3.6, a balanced detector can be used to enhance the performance of systems [49, 46,

38].

3.8 Conclusion

The investigations carried out so far indicate that the technologies for constructing a practical

OCDMA system are far from mature and are eventually more complex than those required for

TDM or WDM systems. Some barriers limiting the realization of OCDMA cannot be

completely solved in the near future. The motivation of introducing OCDMA in optical access

network is that it would allow a more flexible networks structure and introduce the high-speed

optical signal processing in the optical network [7]. For these reasons several researchers have

been exploring the possibility to utilize fiber bandwidth beyond the limits of electronics speed

for purposes of increased network flexibility and decreased access cost.

Table 3.1 The characteristics of various OCDMA proposals

Incoherent

Direct

Sequence

Coherent

Direct

Sequence

Frequency

Coded

OCDMA

Fast

Frequency

Hopping

Phase

Shift

Keying

FSR

Filter

OCDMA

Coherence

Multiplex

OCDMA

Bandwidth requirement Wide Wide Wide Wide

Dispersion impact √ √ √ √

Bandwidth efficiency Low Low low Low

Channels Multi Single Single Multi 2 5 4 Results of

Experiment Data rate

per

channel

10Gbit/s 155

Mbit/s

1Gbit/s

But as shown in Table 3.1, the goals have been only partially achieved up to now. After more

than a decade of research on OCDMA, it is still in the theoretical level. The development in

this area depends more on the evolution of other relative technology. Except the early lower

43

bit rate demonstration system [6], the problem of the interferometric noise was evaded for all

the experiments of CDMA with optical en/decoding process.

Obviously, the use of OCDMA with all optical en/decoding process in access network is still

difficult up to now. However, the method using electronic en/decoding process to achieve

CDMA in optical channel is a promising choice which can provide a good performance with

narrow bandwidth in optical access network. In chapter 4, the feasibility of OCDMA with the

electronic en/decoding process will be discussed, and the experiments with such a system will

be presented in chapter 5.

44

Chapter 4 OCDMA with Combined Electrical and

Optical En/Decoding

Based on the experiments described in the former chapter, an experimental setup has been

designed to verify the performance of an DS-OCDMA system using optical en/decoding

process and narrowband optical sources (refer Chapter 5). The results show that when two or

more optical beams which come from the same optical source or have the same frequency are

added, interferometric noise will seriously affect the performance of OCDMA system (refer

Figure 5.12.a in Chapter 5 and Figure B.13 in Appendix B). The noise has a very broad range

from under 0.01 Hz to several GHz, and has a very high intensity with the amplitude being

more than two times the signal amplitude. When electronic encoder and decoder are used, this

noise can be reduced or even avoided as will be shown in the following. In this chapter, first

the model of each part of DS-OCDMA system will be given, the origin of interferometric

noise will be analyzed, then the effect of this noise will be simulated. Finally, an optimized

scheme using channel codes is discussed.

4.1 The System Model

Figure 4.1 shows the block diagram of various schemes with optical and electronic

en/decoding. The transmitters a) and b) and the receivers c) and d) are independent, i.e. they

can be freely combined with each other. It is assumed that the data recovery model is ideal

and the clock is perfectly recovered.

In a practical system the chips in OCDMA do not exhibit rectangular shape, because the

bandwidth of components and of the transmission channel is limited. Therefore the equations

for the ideal case given in chapter 2 will be rewritten.

45

a) transmitter with electronic encoder

b) transmitter with optical encoder

Signalgenerator

Impulsegenerator

Electronicencoder

Laserdriver

E/O modulewith ATC

Signalgenerator

Impulsegenerator

Opticalencoder

Laserdriver

E/O modulewith ATC

Amplifier Datarecovery

Opticaldecoder

O/Econverter

c) receiver with optical decoder

Amplifier Datarecovery

Electronicdecoder

O/Econverter

d) receiver with electronic decoder

Figure 4.1 The block diagram of transmitters and receivers

4.1.1 The Impulse Shape of Chips

In order to use the encoder with delay line structure, the users’ data will be preprocessed by

an impulse generator before CDMA encoding. For each Mark of the users’ data, the impulse

generator emits an impulse (i.e. a chip). A series of chips is represented by an electric field

∑∞

−∞=

⋅−=i

iDpulsepulse biTtSEtE )()( 0 (4.1)

where t is time, T is the duration of a user bit, is the shape of a single chip impulse

and the factor b specifies whether the i th bit is a Null or Mark (“0” or “1”). In general

the Gaussian impulse will be selected as the chip shape:

D

=

pulseS

1,0i

⋅−=

m

FWHMpulse T

tS2

222lnexp (4.2)

where T represents the effective duration of the chip impulse, called as time of full width

at half maximum (FWHM), usual by T

FWHM

CFWHM T< (will be discussed later). For t , 2/FWHMT±=

46

the power is half of its maximum value. m denotes the order of the super-Gaussian function.

Here we set m=1, so that the chip shape is similar to the practical chip shape in our

experimental system. Then Eq. (4.2) is simplified as

⋅−=

22

22lnexp

FWHMpulse T

tS (4.3)

The chip shape with T is given in Figure 4.2. Note that the duration of is

unlimited.

nsFWHM 3.0= pulseS

-1 -0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

S pul

se(t)

Time ns

Figure 4.2 The shape of the impulse in Eq. (4.3) with T =0.3ns FWHM

4.1.2 Electronic Encoder

We assume that in an OCDMA encoder with delay line structure the shape of each chip is the

same. Therefore when input is , the output of the electronic encoder is pulseE

( ) li

L

l iFWHMCD

L

llCpulseecode

cbTlTiTtE

clTtEtE

C

C

⋅⋅

−−⋅−=

⋅−=

∑ ∑

∑−

=

∞

−∞=

−

=

1

0

20

1

0

/)(222lnexp

)()( (4.4)

where T is the chip duration in OCDMA system, TC CDC LT /= , where is the length of the

codeword. The factor c specifies whether the l th chip is a Null or Mark in a

codeword.

CL

1,0=l

47

4.1.3 Optical Modulator

The electronic signal will be converted into an optical signal. Because the electric/optical

converter requires a large current to drive the Laserdiode, a driver must be used to obtain a

suitable powerful current. In order to set up the operating point for the Laserdiode, the

maximum and minimum current must be adjustable. Normally, the transfer function is linear

and the driver output current is

bi ItEKtI +⋅= )()( (4.5)

where is the bias current for suppressing the transient phenomena of the Laserdiode, K is

the conversion coefficient and is the input signal of the driver.

bI

)(tEi

DFB laser modulator, electro-absorption modulator (EAM) and the Mach-Zehnder modulator

are available to be used in such a high speed optical transmission system. In the experiment

we used the first two types of optical modulator which will be discussed here.

DFB laser: The scheme of the laser model is shown in Figure 4.3 [71]. It contains a white

noise source ω∆ with a deviation of f∆π2 corresponding to the optical laser linewidth f∆ .

The output of the DFB Laserdiode is a time dependent field . The electrical field of the

optical output is therefore determined by

)(tEM

( ) ( )( )( ) ( )( ) ( )( ) ( )

+−++⋅

−= ∫ ττω

τπ

τπδτ

τδαωδ

dIPK

IPKItjek

ktIPtEt

sppjM

0

'2

2ln2

exp1)()(

(4.6)

where α is the linewidth enhancement factor and δ is the additional phase. is the

conversion function given by the I-P curve of the DFB laser and it depends on the character of

the laser. The simplified function can be given as

))(( tIP

( )( )th

th

th ItIifItIif

IatIbtIa

tIP≥<

⋅+⋅⋅

=)()(

,)(),(

(4.7)

where a, b are the coefficients of the converter efficiency before and after the stimulated

emission of the Laserdiode, respectively, ba <<<0 for high speed laser. is the threshold

current of the Laserdiode, and is the power at laser threshold. The first term of the

integrand in Eq. (4.6),

thI

thIa ⋅

ttI

δδ )))((ln( , specifies the dynamic chirp, the other two terms specify

48

the adiabatic laser chirp, in which is the Chirp factor due to nonlinear gain, is the

Chirp factor due to spontaneous emission, which are determined by

pK spK

totnK

K

mα

WHz /

−))(tI

−jekδ

php

kKτ

π =2 (4.8)

and

( )ph

sptotspsmm

gsp S

nKnchvK

ταααπ +

=

2

22 (4.9)

with the photon lifetime phτ , the nonlinear gain , the spontaneous emission factor , the

inversion factor n , the photon number S in the laser cavity, the output coupling

k tot

sp and the

internal loss coefficient sα . Normally, for DFB-Laserdiode and

.

K p 1045.477 9×=

WK sp ⋅×= 0301.2 Hz410 )(' tω corresponds to the line width f∆ of the Laserdiode.

frequencymodulator

whitenoise

I(t)

CurrentEM(t)

Optical

ω

( )

⋅+

))((2

(2)(ln2 tIP

KPK

ttI sp

p

ππ

δδα

−δjekktIP 1))((

)(' tω

fDeviation ∆= π2

Figure 4.3 The model of a directly modulated DFB Laserdiode

DFB laser with EAM: This is a model of the DFB laser with monolithically integrated

electro-absorption modulator (EAM), which is developed for optical transmission system

having a data rate above 10Gbit/s, and for DWDM system. This modulator is set on the basis

of Indiumphosphide (InP) which is integrated in an InP WDM-laser-chip. The electrical field

of the output of the DFB with EAM is

( )

+⋅

⋅= ∫

t

iamM dtjk

PtEhtE0

max )('exp1)()( ττωω (4.10)

where is the electrical input field strength of the EAM and is the transfer

function of EAM. Normally the amplitude modulator has a nonlinear characteristic which

)(tEi )(amh

49

influences the extinction ratio (EXT). An important advantage of this modular is that no chirp

noise occurs in the output spectrum.

Amplitudemodulator

Signal sourceEi(t)

whitenoise

frequencymodulator

ω

CW-Laser

EM(t)

−δjekk

P1

))(( tEh iam

)(' tω

fDeviation ∆= π2

EAM

Figure 4.4 The model of DFB Laser with EAM

))(( tIP in Eq. (4.6) and amh in Eq. (4.10) are in reality nonlinear. Thus, the output of the

modulator is not an ideal impulse as given by Eq. (4.3), because when Null is transmitted, the

laser of the optical transmitter will not be off completely. For a modulator with DFB laser, the

driver has to drive a bias current biasI when signal is Null, in order to reduce the effect of the

transient phenomenon (e.g. the pattern effect) of the Laserdiode to achieve a good waveform

of the signal. For DFB laser with EAM, the optical power can not be completely absorbed

when the signal is Null, which is due to the absorption characteristic of EAM. The result is

that after the optical modulation the signal will be added to an optical bias at Null. This

characteristic is measured by the parameter extinction ratio (EXT) which is important for the

receivers.

In on-off keying optical transmission system EXT is defined as

dBPP

EXTNull

Marklg10= (4.11)

where MarkP and NullP denote the optical power of Mark and Null, respectively. In our

OCDMA system the chips are transmitted instead of bits. We use the peak power ( MaxP ) of

the chips instead of the maximum power of Mark ( MarkP ), and redefine the EXT as:

dBPP

EXTNull

Maxlg10= (4.12)

The EXT is defined in optical power domain, but the signal expressed above is operated in

electronic field domain, their relationship has to be given. When PulseA and NullA denote the

amplitude of the impulse and the absolute value of the base, respectively (Figure 4.5), then

50

(4.13) 2

2)(

NullNull

NullPulseMax

AP

AAP

=

+=

When EXT and are given, Eq (4.13) changes to MaxP

10/

10/

10

10EXT

MaxMaxPulse

EXTMaxNull

PPA

PA−

−

⋅−=

⋅= (4.14)

MaxP can be normalized without influencing the result of the simulation, thus:

10/10 EXTNullA −=

10/101 EXTpulseA −−= (4.15)

EXT can be set by the driving current for DFB laser, and by the base level of the amplifier for

the EAM. An example of the impulse shape is shown in Figure 4.5 with EXT=10

( =0.3162, =0.6838). NullA MaxP

-1 -0.5 0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

ns

E(t)/

E(0) APulse

ANull

Time

Figure 4.5 The optical modulated impulse with EXT=10 and T =0.3ns FWHM

4.1.4 Optical Encoder and Decoder

The optical encoder and decoder operate in the optical domain and are composed of optical

couplers and optical delay lines, as shown in Figure 4.6. The optical impulse produced by the

optical modulator is split into w branches which are coupled at the end coupler after a certain

51

delay for each branch set by the optical delay lines. The performance of the optical couplers is

given in appendix A2. The electric field of the lightwave at the output of an optical encoder is

given by

∑=

−=CL

llCMoenc clTtE

wtE

1)(1)( (4.16)

where the factor w/1 results from splitting which is an ideal value, is an uncoded

signal from the output of the optical modulator and,

ME

1,0∈lc denotes the weight of the chip,

as given in Eq. (4.4).

t1

t2

tw

1:w w:1

)(tEM )(tEoenc )(tEodec)(tER/ /

Figure 4.6 The structure of the optical en/decoder

The structure of an optical decoder is the same as that of an optical encoder, but the delay

time is different. The electric field of the output of an optical decoder is given by

lLcc

L

lRodec c

C

cTlLtEw

E −

−

=

−−= ∑ ))((1 1

0(4.17)

where is the electric field of received optical signal and is the length of the codeword. RE cL

The length of a fiber section to obtain the delay time of a chip, T , is given by c

g

CC n

cTl = (4.18)

where c is the speed of light in vacuum and is the group index in the core of the optical

delay line.

gn

52

4.1.5 Optical Channel (Optical Fiber)

In a PON the optical channel is composed of optical fibers and optical couplers. Optical

couplers consume a great deal of optical power which must be considered in the calculation of

the optical power budget. For the system adopted in the present study, the optical couplers

have little nonlinear characteristics, therefore, the effect on S/N ratio from the optical couplers

will be ignored in the analysis. However, the influences caused by dispersion, PMD and other

linear impairments of the optical fiber on the quality of transmission will be considered.

The chromatic Dispersion: The propagation characteristics for each wavelength

depend on the refractive index of the fiber and on the nonlinearity of the propagation constant.

The optical chips are subject to impulse broadening as they propagate in the optical fiber. For

the single-mode fiber impulse broadening is solely caused by the so-called intramodal

dispersion. The dispersion parameter D is obtained by deriving the group-delay Tg with

respect to wavelength λ [77]:

WMg DD

ddT

LD +==

λ1 (4.19)

where

2

21λ

λλ d

ndcd

dnc

D gM −==

( ) ( )

−

⋅

∆= 2

22

dVVbdV

nn

dVVbd

ddn

cD gg

W λλ

in which ng is the group index, b is propagation constant, V is characteristic frequency and ∆

is the index difference of the core and the cladding. D has two components: one arising from

the material dispersion, DM, the other from the waveguide dispersion, DW. DW can be seen to

be proportional to the small parameter ∆. It is generally smaller than DM. DM crosses the λ–

axis near 1.3µm for the standard fiber. In this case, DW can be tailored through an

optimization of the waveguide profile and plays a crucial role in defining the exact location of

the overall zero dispersion wavelength.

PMD: Polarization-mode dispersion (PMD) is caused by asymmetries and stress

distribution in the fiber core, which locally leads to birefringence, i.e. a polarization-

dependent refractive index. It is another reason which influences the signals transmitted in the

optical fiber. In OCDMA system there are multiple lightwaves, that are emitted from different

53

lasers in local ONU. After transmission through the fiber, the lightwaves have different states

of polarization at the optical receiver. The optical signal to noise ratio decreases due to the

effect of the interference.

4.1.6 O/E Converter with Electronic Amplifiers

Unlike the normal on-off keying, in optical communication systems with OCDMA the main

noise comes from interference by other users. The performance is not sensitive to the noise of

the O/E converter and pre-amplifiers at receiver. So, they can be ignored here, and the O/E

converter can be assumed as ideal. The output of an optical receiver including O/E converter

and amplifiers is given by

2/ )()()( tEtfAtU REO ⋅⋅⋅= η (4.20)

where η is the efficiency of opto-electronic conversion, A is the amplification coefficient,

both of which can be normalized to 1 without affecting the value of S/N. is electrical

field strength of the received optical signal. is the impulse response of the equalizer

filter of the optical receiver and amplifiers, which can be expressed by the fifth-order Bessel

filter. The Laplace transform F(s) is given by [60]

)(tER

)(tf

94594542010515945)( 2345 +++++

=sssss

sF (4.21)

4.1.7 Electronic Decoder

The electronic decoder is composed of electronic delay lines (coaxial cable), the signal splitter

and signal coupler. There are two types of the signal couplers: additive coupler and logic

AND coupler. The electronic decoder with additive coupler has a same function as an optical

decoder. The electronic decoder with logic coupler is composed of logic circuits. Firstly, the

signal is digitized by a decision at the input of the decoder, then the binary signal is split into

w branches which are then coupled into coaxial cable delay lines. After transmission through

the coaxial cables with different delay times suited for the codeword, the signals will be

processed by AND logic. This processing can be express as

( )( )[ ]∏−

=−−−=

1

0/

C

C

L

llLCCEOedec cTlLtUgE (4.22)

where Π denote multiply and is the decision function, given by [ ]xg

54

≤≤>

=θ

θx

xxg

0,0,1

][ (4.23)

An additional advantage of this decoder is that it has the function of a hard-limiter for the

undecoded signal, hence the BER can be reduced without further operation.

4.2 Interferometric Noise

OCDMA system is an interference-limited-system. Though the interferometric noise has

attracted much attentions in the areas of WDM and OTDM networks [78, 79, 69], only

several papers can be found concerning OCDMA system. Tancevski and Rusch [80] analyzed

impact of the beat noise on the performance of 2-D OCDMA system, focus was on the impact

between different frequency points. The analysis of the influence of optical noise on

OCDMA, especially for narrow band optical source was rarely addressed up to now.

4.2.1 Characteristics of Interferometric Noise

The electric field of a CW optical source can be given as

)))((exp()( 00 ttjEtE φω += (4.24)

where 0ω is the optical frequency, )(tφ is the phase noise which is a random time varying

component resulting from spontaneous emission events [64]. Considering the addition of two

CWs and their crosstalk, the interferometric noise may be classified as follows [65]:

Case a) If the two CWs are from the same source and suffer a differential delay τ between

source and detector, their relative phase arrived at the detector is

),()()()(

0

0

ττωτφφτωφ

tttt

Φ+=−−+=∆

(4.25)

The coherence time, cτ , of the laser source provides the characterization of phase

fluctuations. For the 1550nm band DFB laser, cτ ranges from several nanoseconds to a few

100ns. ( )τ,tΦ has a close relationship with cτ when τ is fixed. For a chirp free impulse

source, ( )τ,tΦ can be approximated as a zero-mean Gaussian process of Wiener form [78].

Coherent Crosstalk: τ is much smaller than cτ , resulting in an interference very close to the

coherence limit ( 0=τ ), which is characterized by an absence of laser phase noise (variance

55

of ),( τtΦ <<2π ). However, significant interferometric noise may arise from slow fluctuations

in τ induced by temperature fluctuations and phonon excitations in optical fiber-

environmental phase noise. If τ is constant, the interferometric noise is static and may be

determined.

(exp(0 j φ+ exp(

[ )(t(φ∆ ω

Incoherent Beat Noise Crosstalk (also called phase induced intensity noise, PIIN): τ is

much greater than cτ , resulting in an interference at the so-called “incoherent limit” [65],

such that the variance ),( τtΦ >>2π . Here the interferometric noise is driven by the laser

phase noise which masks any influence of the (lower bandwidth) environmental phase noise.

Partially Coherent Crosstalk: this class falls between the two extremes of coherent and

incoherent beat noise crosstalk.

Case b) Crosstalk arise from distinct laser sources, their electric fields are given as

)))(ttE ddω and )))((0 ttjE xx φω + . The relative phase is given as

( ) ])() ttt xdxd φφω −+−= (4.26)

Incoherent Noise-Free Crosstalk: if the beat frequency of the two laser, πωω 2/)( xd − ,

exceeds the receiver bandwidth, then the interferometric noise is removed by electrical

filtering following detection, and only additive crosstalk components remain.

Incoherent Beat Noise Crosstalk: if the beat frequency of the two lasers, πωω 2/)( xd −

txd )(

, is

smaller than the receiver bandwidth, then the cyclic variation in phase due to ωω − , and

the random variation due to the phase noise elements, generate interferometric noise. This is

also called incoherent beat noise crosstalk because the mixing is incoherent.

The case a) would occur at the optical en/decoder: at optical encoder partially coherent

crosstalk occurs and at optical decoder coherent crosstalk occurs. And case b) would occur

due to the interference by other OCDMA channels with different optical sources. The

intensity of the interferometric noise mainly depends on the phase noise of optical source and

phase noise introduced by environment.

56

4.2.2 The Source of the Phase Noise

The Phase Noise of DFB Laser: The phase fluctuations of lasers determine the

emission spectra. The laser linewidth f∆ is determined by the magnitude of the phase noise:

f∆= πδφ 22 (4.27)

In general it can be assumed that the phase undergoes a random walk where the steps are

individually spontaneous emission events which instantaneously change the phase by a small

amount in a random way [73]. At the optical encoder, the lights in the branches achieved from

same optical source with different delay time will be added, and phase noise will introduce

the coherent noise.

Chirp: When the driver current of the laser is changed, the central emission frequency will

drift, and this effect is named chirp, which is given in Eq. (4.6). Because in access network

the chip rate is relative low, and the distance is short, the dispersion introduced by chirp noise

is small. But in the optical encoder the influence of the coherent noise can not be ignored. For

external optical modulator there is no chirp effect on the modulated signal.

Phase Noise by Thermal Effect: At the optical en/decoder characteristics of the

delay line change with changing temperature, for example both the length l and the reflection

index (see Eq. (4.18)) are depending on temperature. We rewrite the Eq. (4.18) considering

the disturbance of temperature as

gn

cTnTl

T gcC

)1()1(' ∆+⋅∆+=

νψ (4.28)

where dTdlc=ψ is the thermal expansion coefficient and dTdng=ν is the ratio of index

change with temperature change. The change of T will introduce the wandering of the phase

at the output. Because the length of each branch is not the same and temperature is not even

distributed, the variation of the parameters introduced by temperature is different. Since the

phases of the optical beams in each branch at the input point of optical coupler are variable,

amplitude noise will be introduced at the output of optical en/decoders.

'C

57

4.3 The Results of the Simulation

The simulation has been carried out by using Matlab. In this simulation, thermal and shot

noise in photo-detector are neglected with very little influence on the results [28]. The

OCDMA system has the ability of accessing the variable bit rate stream with minimum bit

duration T . Here, the worst case, viz. all of the users’ data streams are at maximum bit rate,

is considered.

min

In general (ratio of index change to temperature change),

(thermal expansion coefficient) [44]. A random

C°= − /10 5ν C°×= − /105 7ψ

T∆ with the variance of 10 °C is chosen as

the wandering of the temperature of the environment. We take a DFB Laser generating an

optical impulse with

4−

nm56.1540=λ , cτ =2ns and FWHM=0.25ns as input. The optical delay

line used is a SSMF with . 45.1=gn

4.3.1 The Noise in the Optical Encoder

The input signal of the optical encoder comes from the optical modulator. The output is given

by Eq. (4.16). Because of EXT, the output of the optical modulator is not ideal, and at the

Null there is little optical power. At the output of optical encoder, the interferometric noise

occurs, due to the phase shifts between each branch caused by changes in the reflective index

and changes in the optical delay lines.

In Figure 4.7 the simulated output of the optical encoder (1, 2, 5) is shown for various

extinction ratios (EXT) of optical source. When EXT is above about 17dB, the output signal

of the optical encoder is still useful.

58

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

the output of the optical encoder by EXT=unlimited

the

optic

al p

ower

ns

Time

Rel

ativ

e op

tical

out

put p

ower

a) EXT= ∞

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=30ns

Time

Rel

ativ

e op

tical

out

put p

ower

b) EXT=30dB

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=20ns

Time

Rel

ativ

e op

tical

out

put p

ower

c) EXT=20dB

59

0 0.5 1.0 1.5 2.0 2.5 3.00

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=18 ns Time

Rel

ativ

e op

tical

out

put p

ower

d) EXT=18dB

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=17ns

Time

Rel

ativ

e op

tical

out

put p

ower

e) EXT=17dB

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=16ns

Time

Rel

ativ

e op

tical

out

put p

ower

f) EXT=16dB

60

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of the encoder by EXT=13ns

Time

Rel

ativ

e op

tical

out

put p

ower

g) EXT=13dB

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

The

opt

ical

pow

er

the output of the encoder by EXT=10ns

Time

Rel

ativ

e op

tical

out

put p

ower

h) EXT=10dB

Figure 4.7 Output signal of optical encoder if optical input signals with various

extinction rations EXT are applied

The output signal of E/O converter of an electronic encoder is show in Figure 4.8. Compared

to the optical encoder output signal in Figure 4.7, the noise is extremely low.

61

0 0.5 1 1.5 2 2.5 30

0.2

0.4

0.6

0.8

1

1.2

1.4

The

opt

ical

pow

er

the output of electronic encoder by EXT=10ns

Time

Rel

ativ

e op

tical

out

put p

ower

Figure 4.8 Optical signal of an electronic encoder (EXT of the E/O converter is 10dB)

4.3.2 The Noise Performance of Optical Decoder

When all CDMA encoding and decoding are implemented by optical elements, the

interferometric noise will also occur at the output of the optical decoder. By the auto-

correction point of the decoded signal, the impulses of chip “1”, which have the same total

delay times included in optical encoder and decoder, would overlap. In this case, the phase

noise, which comes from the optical source such as chirp and linewidth, couldn’t directly

introduced the noise, as it is in optical encoder. So the interferometric noise hasn’t

relationship with the coherent time cτ . The interferometric noise is mostly from the phase

noise introduced by the temperature variation of the optical delay lines. Without external

control, the optical phase difference between interfering terms is random and slowly drifts as

optical path lengths fluctuate with ambient temperature changes.

For the standard single-mode fiber, the dispersion coefficient D is about 18ps/nm/km when

the center wavelength of optical source is 1.5µm. For a Laserdiode which is used in DWDM

system, the width of spectrum, ∆λ, is narrower than 0.1nm. Normally, the range of the local

access network is smaller than 20km. Therefore, the optical impulse is broadened by [88]:

psps

LDTD

36201.018 =××=

⋅∆⋅=∆ λ

For the demonstration system, the width of optical chips is circa 300ps. Thus the dispersion

degradation by can be neglected.DT∆

62

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

0.5

1

1.5

2

2.5

3

The

opt

ical

pow

er

the ideal output of the optical encoderns

Time

Rel

ativ

e op

tical

out

put p

ower

a) without interferometric noise, e.g. ideal situation

Time

Rel

ativ

e op

tical

out

put p

ower

b) with interferometric noise

Figure 4.9 Simulation results of output signal of optical decoder

When the optical source has enough bandwidth ( cτ is relative short) and the chiprate is low

( is relative large), coherent crosstalk can be almost avoided at the optical decoder. The

ideal case can be seen in Figure 4.9 a). Then the reduced interferometric noise will not impact

the decoder function. But for narrowband optical source (

DT

cτ is relative long) and high chiprate

( is short), the coherent crosstalk is significant, and interferometric noise will impact the

output of the optical decoder, when the transmitted signal is encoded by an optical encoder.

This can be seen in Figure 4.9 b).

DT

63

4.3.3 The Performance of Electronic Decoder

When the optical receiver and amplifiers are ideal, the simulated output signals of an

electronic decoder with logic AND coupler and of an electronic decoder with additive coupler

are shown in Figure 4.10 and Figure 4.11, separately. The output of the electronic decoder

with logic AND coupler is purer than of an electronic decoder with additive coupler. We

select electronic decoders with logic AND coupler in the demonstration system.

Time

Rel

ativ

e op

tical

out

put p

ower

Figure 4.10 The result of the output signal of electronic decoder with logic AND coupler

Time

Rel

ativ

e op

tical

out

put p

ower

Figure 4.11 The result of the output signal of electronic decoder with additive coupler

In addition, from above simulated result of electronic decoder with logic AND coupler, we

find that it also has the function of a hard-limiter as in Eq. (4.23) defined. Therefore the BER

could be reduced through the use of such electronic decoder. When the effects of the quantum

noise and the thermal noise are neglected, it has been shown that using the hard-limiter will,

64

in general, improve the system performance and make it possible to accommodate more users

[28]. When using the optical orthogonal codes (OOC), the BER without a hard-limiter is

given in Eq. (2.17) and an algorithm for exact calculation of BER with hard-limiter is for the

first time given in Appendix A1. Figure 4.12 shows the comparison of the performance of

BER with and without hard-limiter.

Figure 4.12 Comparison of BER with and without hard-limiter

4.3.4 The Effect of Width of Chip Impulse

If the rate of the transmission data and the codeword are fixed, the effect of shape of the chip

impulse will affect the result of the CDMA decoder, such as ISI in on-off keying systems. By

using the Gaussian impulse, given by Eq. (4.3), the relationship between the BER and FWHM

of chip impulses is simulated. Since the data of users are asynchronous in OCDMA, the

requirement of the width of impulse is strict. In the simulation, the code words (19; 1, 2, 9)

and (19; 1, 5, 11) are used for the interference data source, and codeword (19; 1, 3, 6) for

desired date source. When ignoring the other noise except interferometric noise, the simulated

result of BER is shown in Figure 4.13. If the width of the impulse increases, the BER gets

bigger. When the FWHM of chip impulse equals to about 80% of the chip duration T (i.e.

), the BER is approximately zero.

C

CFWHM TT 8.0≈

65

BER

nsTonsT

C

C

50.0:33.0:

==∗

FWHMT

Figure 4.13 Relationship between the BER and the FWHM of Gaussian chips impulse

4.3.5 Summary

From the results of the simulation we can find that the output signals of the optical encoder or

decoder has large noise and the decoder even loses the function as an auto-correlator. The

noise locates in a broadband, from under 0.01 Hz to several GHz. Because of the

implementation, the phase noise of the optical carrier is converted into the amplitude noise at

the optical receiver. The energy swings at the output of the optical decoder. Up to now, this

energy swing cannot be avoided. With large S/N ratio, the results of electronic en/decoders

are satisfying. Furthermore, using logic electronic decoder would reduce the BER in OCDMA

system. Otherwise, the width of the impulses of chip “1” must be limited to a small value for

given chiprate and BER.

4.4 The Channel Coding

In order to reduce the BER introduced by noise from CDMA processing and transmission

channel, the use of a channel code for data transmission is a feasible approach to optimize

system performance. Here we discuss the possible improvement by combining the CDMA

scheme with external error correcting codes. The advantages of channel coding are well

known for the classical Gaussian noise channel, the use of channel coding for the OCDMA

crosstalk channel is somewhat different, but this difference can be ignored in the calculation

[60].

66

PassiveOptical

Network

ChannelEncoder

OpticalCDMAEncoder

OpticalCDMADecoder

ChannelDecoder

OCDMA Channel

InformationBits

ReceivedBitsOther

SendersOther

Receivers

Figure 4.14 Block diagram for CDMA system with channel coding

Channel coding can be applied externally to the OCDMA link, as shown in Figure 4.14. At

the transmitter, channel encoder converts the information bit stream into binary channel

symbols, which are then sent through the fiber link with the OCDMA processing. After

OCDMA detection at the receiver, channel decoding converts the detected link symbols to

information data stream. All of the channel coding processes operate in the electronic domain.

Generally, two types of channel codes are considered to be used in OCDMA system. The first

type is the standard forward error correction coding such as Reed-Solomon (RS) block codes

and convolutional codes. The second type involves impulse position modulation (PPM). Here

the first type will be used because PPM code has very high code redundancy [68].

4.4.1 Reed-Solomon Code

In this approach, data bits are coded into Reed-Solomon (RS) codewords. A codeword of an

(n, k) RS code contains n symbols, in which k symbols are original data. Each symbol is

composed of m bits. n and m are related by

12 −≤ mn (4.19)

The code can correct errors in t symbols and t is given by

2)( knt −= (4.30)

The probability of decoded bit error for a RS code is given by [70]

( ) inS

iS

n

tjm

m

RS PPjn

ntjP −

+=

−

−

+

−

≤ ∑ 112

21

1

(4.31)

where is the probability of RS symbol error which is given by. SP

meS PP )1(1 −−= (4.32)

where is the error probability of the OCDMA link. eP

67

The channel code introduces the redundancy which will reduce transmission capability of the

system. Two methods can be used to retain the bit rate and the number of users: reducing the

length of OCDMA codeword and/or code weight with same chip rate, or increasing the chip

rate with same codeword. In [68] the first method has been examined theoretically, the results

show that, for a system with a large number of users and codewords with large code weight,

the BER can be obviously reduced. Practically, it is difficult to use this method, since the

OCDMA codewords must be reselected and the OCDMA en/decoders rebuilt, so we will

focus on the second method.

The redundancy of RS channel code is

kkn −

=ℜ (4.33)

Normally, one symbol includes 8 bits, i.e. m=8, so n ≤ 255. Figure 4.15 a) shows the results of

the RS code with k=249 and n=255 (ℜ =2.41%). In this case, only when <10 , the RS

channel code improves the system performance very well. Figure 4.15 b) shows the results

when m=3, n=12, t=3 (ℜ =50%). In this case, only when <10 , the RS code operates

well.

eP 5−

eP 2−

a) RS with n=255, k=249, m=8: redundancy=2.41%

68

b) RS with n=12, k=6, m=3: redundancy=50%

Figure 4.15 The BER of OCDMA with and without RS code

4.4.2 Convolutional Code

In this case the channel coder uses a convolutional coding (CC) to convert the data bit steam

into a coded bit stream for OCDMA system. The error probability at the decoded for the

convolutional code is computed in [70], and approximated for OCDMA in [68]. In addition,

turbo codes are concatenated convolutional codes in which the information bits are first

encoded by a convolutional encoder, and then after passing through an interleaver, are

encoded by a second convolutional encoder. From the simulation results given in [94], turbo

coding offers considerable coding gain over other methods, however with considerable

en/decoding complexity. It is difficult to use convolutional code in our demonstration system

with bit rate more than 100Mbit/s, because the normal digital signal processor can’t decode

complex convolutional codes at this frequency. Therefore we will not discuss this method in

detail.

It is clear that channel coding can be highly effective for enhancing the performance of the

system. Therefore we have designed an RS en/decoder operating at 155Mbit/s for the

demonstration system described in Section 5.7.

69

Chapter 5 Implementation of a Prototype

In this project an OCDMA demonstration system was implemented, in which the

characteristics of OCDMA processing with both optical and electronic encoder and decoder

have been experimentally studied. The system demonstrates the upstream data transmission

between optical network unit (ONU) and optical line termination (OLT) in a PON. For each

user (or ONU), a data rate of 155.52Mbit/s can be achieved to suit the STM-1 in SDH or OC-

3 level for SONET network. To the best knowledge of the author, this demonstration system

is the first implementation with the aim of using the OCDMA technology for distributed users

with such high data rate for access network.

In the experiments, many interrelated factors have been carefully considered, such as

operating characteristics of fibers, optical sources and photodetectors. In this chapter, the

implementation will be described, results of the demonstration system will be given, and

finally some limitations revealed by the experiments will also be discussed.

5.1 Experimental System

The demonstration system is divided into three parts: transmitters, channels (in PON) and

receivers. Each transmitter includes impulse generator, electronic (or optical encoder) and

electronic/optical converter. The channels are represented by optical couplers and the fiber.

The receiver is composed of optical (or electronic converter), amplifier, signal decision, signal

power splitter, electronic CDMA decoder and data recovery. The scheme of the

demonstration system is shown in Figure 5.1.

CDMATX 1

CDMATX 2

CDMATX 3

CDMADecoder n

Data 1

Data 2

Data 3

O/EData nPON

Figure 5.1 The structure of the demonstration system

70

Bit Rate: For upstream links in the access network, the transmission bit rate in the near

future is estimated to be in the level of STM-1. Based on this, components operating at the

frequency corresponding to a data rate of 2.5Gbit/s were chosen. In practice, the chip rate can

be up to 3.0Gchip/s and each of three accessed data rates can be up to 160Mbit/s.

Wavelength: The OCDMA system can actually operate at any wavelength. However, here

the 1550nm band was chosen, in order to easily combine with DWDM technology in future,

which is developed very well in this band [88]. To avoid the incoherent beat noise crosstalk

introduced in section 4.2, the wavelength of each Laserdiode is not exactly the same.

Power Budget of the optical Link: The optical power budget is an important

factor for the design of the optical system. Nowadays, some kinds of Laserdiodes for optical

communication can launch a peak power up to +20dBm, but normally it isn’t used at the

extreme case considering its life time. Here the output power was set at +10dBm (denotes as

=10dBm). Each optical encoder or decoder with four taps (code weight less than 5)

consumes 12dB optical power as a minimum, in which 6dB is consumed by 1:4 optical

splitter and the other 6dB by 4:1 optical coupler (in principle, 4:1 optical couplers could be

constructed with 0dB optical loss, but here they are implemented with normal optical

couplers). If 1dB is taken as the extra loss in practice, then the losses of optical power in

optical encoder and decoder are =13dB and =13dB, respectively (electronic

en/decoders don’t consume optical power). The PON network is represented by a 4:1 optical

coupler which also needs 6dB in principle. Here =6.5dB is taken in practice. When there

are 10 km standard optical fiber used for the transmission link between the transmitter and the

receiver in access network, the loss of the optical fiber is only 2dB. Considering the loss of

the moveable connectors, we take 4dB for the whole optical link, i.e. =4dB. For optical

receiver, in ITU standard the sensitivity of 2.5Gbit/s optical transmission system can be

smaller than –27dBm. Here our optical receiver has a signal to noise ratio high enough to

reach this standard, thus here the sensitivity of the optical receiver is chosen 27dBm in

system design.

LDP

enP deP

PONP

chP

RP

In this system, we demonstrate all combinations of the optical and electronic en/decoder.

When both optical encoder and optical decoder are used in an OCDMA channel, the optical

margin is

71

dBPPPPPPP RdeenchPONLDm

5.0=−−−−−=

(5.1)

When optical encoder and electronic decoder are used, or electronic encoder and optical

decoder are used in an OCDMA channel, the optical margin is

dBPPorPPPPP RdeenchPONLDm

5.13)(

=−−−−=

(5.2)

For OCDMA channel, in which electronic encoder and decoder are used, the optical margin is

dBPPPPP RchPONLDm

5.26=−−−=

(5.3)

At all cases, the optical power budget is enough, hence no optical EDFA needs to be used in

this demonstration system. However in a practical access network, the number of branches of

PON is large, the optical power loss would also be large, therefore, the use of EDFA is

necessary. At the optical receiver we adjust the optical power of transmitters with different

encoding process at the same level by using the variable optical attenuators, in order to

balance the interferometric noise for each OCDMA channel.

The Interface: The interface in this system operates with standard ECL level. Here we

do not build the clock recovery and data reshaper at the signal input, so the data sources have

to offer the clock. But it has no effect on the demonstration of OCDMA processing. The

system can access the data stream with variable bit rate lower than 250Mbit/s.

Some Improving of the System: In this demonstration system several new methods

will be employed in the implementation. Firstly, it is noticed that all of the receivers are at the

same location (at optical line termination), when the electronic decoders are used instead of

the optical decoder in conventional OCDMA system. Thus, the users can share the same

signal from a single O/E converter. In the final scheme, electronic decoders will be located

behind the O/E converter. Secondly, the signal from the electrical amplifier will be decided

between Mark and Null, then the electronic decoders will operate in logic domain instead of

analog domain. This will reduce the bit error rate and simplify the implementation. At last, the

electronic and optical delay line will be employed in the implementation of encoder, decoder

and impulse generator, which has benefit of simplifying the circuit design, as can be seen in

the following sections.

Is summary, the system is realized under the following rules:

72

• Modularized structure;

• Easy to be expanded and upgraded;

• Easy to measure the parameters ;

• Reduced cost.

In the following sections several building blocks of the experimental system will be

described, and the experimental results will be given.

5.2 Signal Source and Scrambler

A scrambler is used for users’ data in this demonstration system in order to easily recover data

clock at the receiver, as it is used in other on-off keying optical transmission systems. The

scrambling circuit generates a random data pattern by modulo-2 addition of a known bit

sequence and it is modulo-2 added with the data stream. At the receiver the same known bit

sequence is again modulo-2 added to the received data, and the original bit sequence is

recovered. Although the randomness of scrambled NRZ data ensures an adequate amount of

timing information, the penalty for its use is an increase in the complexity of the NRZ

encoding and decoding circuitry. Without inputting of user’s data, a scrambler can be seen as

a data generator for system testing.

The pseudo-random binary sequence (PRBS) is the best bit pattern for scrambling and the

system test. It has random bits with almost equal probability of 0 and 1. The period of PRBS

is

bitsN r 12 −= (5.4)

where r is the order of the PRBS. Normally r equals 7, 15, 21 or 23 for system test. The

primitive generation polynomials of order r are:

1,231,21

1,151,7

523

221

15

7

++=

++=

++=

++=

xxrxxrxxr

xxr

(5.5)

Every primitive polynomial of order r defines a recurrence relation for obtaining a new

random bit from the r preceding ones. The recurrence relation guarantees to produce a

sequence for maximal length, i.e., cycle through all possible sequences of r bits (except the

sequence of all zero) and get N random bits before the sequence repeats. Only a r bit long

shift register with feedback and a few XOR (“exclusive or” or bit addition mod 2) gates are

73

required for the implementation in hardware, as shown in Figure 5.2. The contents of selected

taps are combined by exclusive-or operation and the result is added to the input.

1 2 3 4 r-1 r

+

Output+

Input

Figure 5.2 Scrambler for generation of pseudo random bits

In order to better test the characteristics of the system and allow for various data sources, the scramblers in this system are built with variable, adjustable sequences types at the output by controlled switches. The scramblers can also operate as a signal source with internal or external clock. The output sequences include square waveforms with the clock divided by 2, 4, 10 from the internal or external clock, and PRBS sequences with order of 7, 15, 21, 23. The block diagram of the variable signal source is shown in the Figure 5.3.

InternalClock

TTL toECL

Clockdriver

Against all 0and Reset

1 2 5 7 15 21 23 24

S1

+S2 S3

+

To SET pin ofthe d-flip-flop

External clock

Clock output

Data input

Data Output

Figure 5.3 The block diagram of the variable signal source

The internal clock employs a quartz oscillator with TTL level (its jitter is smaller than 50 ppm).

Because the logic gate circuits operate with ECL level, the internal clock signal must be

converted to ECL by TTL-ECL converter. Due to the small fan-out coefficient of an ECL-IC

output at high frequency, the clock driver circuit is required. For the same reason, all feedback

74

signals of XOR are obtained from negative output of D-flip-plop instead of positive output,

because when QBA =⊕ ,

QABBABABABABA =⊕=⋅+⋅=⋅+⋅=⊕ (5.6)

If a scrambler is used as a signal generator and when all D-flip-flop outputs are low (“0”), the

circuit is in the all 0 state. In this case the scrambler generates an output sequence of “0” bits.

To avoid that, we have designed a circuit “against all 0” to prevent this state. The circuit

monitors the output of a D-flip-flop to sample the state information. When the output is

always “0” for a certain time (for example 1ms), it is sure that all 0 state exists in the

scrambler. Then an enable impulse will be sent to reverse the output of one D-flip-flop. Then

the output sequence of the scrambler will break away form the all 0 state.

The various combinations of switches S1, S2 and S3 control the output pattern. When S1 is

switched to external clock and user’s data and clock are given, the circuit will work as a

scrambler. When S1 is switched to internal clock without data input, the output pattern is

given by S2 and S3. The scrambler can operate with the clock from 10 to 250MHz. As an

example the measured result of order 7 PRBS is shown in Figure B.14 in the annex. A photo

of the hardware of the scrambler and signal source is shown in Figure 5.4.

Figure 5.4 The hardware for the scrambler and signal source

75

5.3 Impulse Generator

Because OCDMA processing operates at chiprate, at first the users’ data has to be converted

from data rate to chiprate for CDMA coding, we design an impulse generator at transmitter to

achieve this function. For each Mark of data, this impulse generator produces a slender

impulse having a width equal to the duration of one chip, irregarding whether the Marks of

the data are continued or not. The output is the signal source of the CDMA encoders. The

impulse generator circuit needs the data and clock as inputs simultaneously. As an interface,

the line receivers for receiving of the data and the clock can be established by differential

receiver MC10EL16. This device with rise and fall times typically at 225ps is suitable for

interfacing with high frequency sources.

For the traditional method to receive data and operate it from the data rate to chiprate, the

logic diagram of impulse generator and its logic waveforms are shown in Figure 5.5. The chip

clock can be achieved from the data of the downstream in network, from the data’s clock or

from a local quartz oscillator, depending on the structure of the network. In Figure 5.5 it

comes from user’s data, and the advantage of this method is synchronism between chiprate

and user’s data rate. The “× ” is an N frequency-multiplying circuit. Here N is the length of

the codeword, that will be in detail discussed in next subchapter. D-flip-flop, AND gate and

OR gate are implemented by the IC of MC10EL series which are the fastest series for silicon

devices nowadays. The driver of the chip clock is indispensable for the D-flip-flops that work

at 2.5GHz for chiprate which is close to the limit rate for MC10EL series. Hence the chip

frequency is high, therefore the phase delay of signals in circuit design has to be considered,

that includes the delay time of logic gate and the delay time of connection line in printed

circuit board. For example, the AND gate before point A in Figure 5.5 is used as a delay gate,

in order to match the phase between the clock and the data.

N

For above mentioned scheme, the driving of two D-flip-flop with 2.5GHz and the use of

frequency-multiplier N are still not satisfying in the hardware implementation. Even if the

chip clock is generated by a local oscillator which is independent on data clock, the data

stream and the chip stream are slightly asynchronous which introduces a jitter at CDMA

processing. This jitter will affect the quality of the clock recovery at the receiver and increases

the BER.

76

Input of clock at 50MHz

A

B

C

D

E

F

G

H

Linereceiver

DLine

receiver & D

D

XOR &

1

Driver

Data in50Mb/s

Clock of chiprate

Output of Impulse

N: length of code wordD: D-flip-flopXOR: exclusive OR&: logic AND gate

A

B

C E

F

G

H

rN

Figure 5.5 The logic diagram of impulse generator and its waveform

The above mentioned two problems can be avoided by using time delay line instead of the

two high frequency D-flip-flops. Figure 5.6 shows such an improved circuit of impulse

generator and its logic waveform. The electrical delay line generates the impulse at chiprate.

Thus the generation of the chiprate clock can be avoided. The delay time of the delay line is

the duration of a chip, T . The length l of the delay line is: C

TSl ⋅= (5.7)

where S is the speed of electromagnetic wave in a coaxial line, normally it is 20cm/ns. When

is 400ps, l equals to 8cm. In practice the length is chosen to 7cm due to other phase delays

from the printed circuit board and devices.

CT

77

Linereceiver

DLine

receiver & XOR &

1

Data in50Mb/s

Input of clock at 50MHz

Output of ImpulseD: D-flip-flop XOR: exclusive OR &: logic AND gate

A

B

C

G

H

D

E

Delay line

A

B

C

D

E

G

H

Figure 5.6 The improved logic diagram of impulse generator and waveform

Data ClockData Clock

GeneratedImpulse

Figure 5.7 The output impulse of the implemented CDMA impulse generator

78

This method not only simplifies the circuit, but also has benefits for the data access. With the

circuit the impulse is generated at the start point of the data, thus the OCDMA processing will

be synchronous with the data without other conditions. This means it is possible to access

variable bit rate data into OCDMA system without increasing the phase jitter of the recovered

data clock at the receiver. The waveform of composed impulse can be seen in Figure 5.7.

5.4 Encoder/Decoder

The encoder and decoder are important parts for the CDMA processing. We have

implemented both electronic and optical encoder/decoder, respectively, in order to

demonstrate their characteristics in various schemes of OCDMA system. The structure of the

encoder and decoders are the same, i.e. using the optical or electrical delay line to compose

the codeword.

5.4.1 The Codeword

OOC code widely used in OCDMA [16] is chosen as signature code for this demonstration

system. This code can be employed to support both equal-rate and variable-rate/multi-rate

users without changing CDMA encoders and decoders.

Table 5.1 Some shortest OOC [14]

Code weight ( w)

number of users (N)

OOC of equal length ( ) CL

OOC of variable length

3 3 19; 1,2,9 19; 1,3,6 19, 1,5,11

16; 1,2,6 19; 1,3,9 21; 1,4,11

4 4 49; 1,2,21,24 49; 1,3,14,18 49; 1,6,13,22 49; 1,7,15,25

42; 1,3,14,18 46; 1,6,13,22 48; 1,2,21,24 49; 1,7,15,25

The Table 5.1 lists some codewords of shortest OOC with auto-correlation 1=aλ and cross-

correlation [14], where w is code weight, N is the maximum number of users for this

codeword family. The code length is the minimal number of chip positions for the

codewords. The numbers after the code length are the positions of “1” chips within the

CL

1=cλ

79

codeword. The encoders send codewords only for Marks of the users’ data, and don’t send

any chip for Nulls of users’ data. In this system, the codeword with weight w=3 and number

of users N=3 is used (Table 5.1). As an example, Figure 5.8 shows the coding with codeword

family “19; 1,2,9” , “19; 1,3,6 ” and “19, 1,5,11”. Some encoder/decoders with codewords of

weigh w=4 are also tested.

1 0 1

1 2 3 4 5 6 7 8 9 10 11 12 13 1415 16 17 1819 1 2 3 4 5 6 7 8 9 10 11 12 13 1415 1617 18 19 1 2 3 4 5 6 7 8 9 10 11 12 13 1415

A

B

C

D

E

A: Data of user B: chip clock C: code word ( 1,2,9 )D: code word (1,3,6 ) E: code word (1,5,11)

Figure 5.8 Chip patterns of a code family (with n=3, w=3), here data and chip clock are

synchronous

In order to maintain the orthogonal character, the data rate is limited. In Figure 5.8 the chip

clock is synchronous with the users’ data rate and in this case the data rate is the maximum

data rate which CDMA processing can operate with. When the data rate is smaller than this

maximal rate, the system can also well access this lower rate or variable bit rate without

increasing system complexity. In this case the time slip between the two codewords will be

occupied by null with any duration. The maximum data rate is calculated by

sbitL

ChiprateRC

/max = (5.8)

In this system, the maximal chiprate equals 3.0Gchip/s and equals 19, then is

157.89Mbit/s for an OCDMA user. The total throughput of the system, , is

473.68Mbit/s.

CL maxR

N×Rmax

80

5.4.2 Electronic Encoder

t1

.

.

.

Signal driver Signal adder

t2

tw

Figure 5.9 The structure of the encoder

In this demonstration system, the impulse generators are established as interfaces for user’s

data. Hence the electronic encoder of OCDMA directly operates with the stimulated signal

from impulse generators. The principle structure of the electronic or optical encoder which is

built with delay lines is shown in Figure 5.9. The left part of the structure is signal power

splitter or signal driver and the right part is signal power adder. The two parts are connected

by the electronic delay lines. Each delay line has an individual delay time, which contributes

an impulse (chip) for the codeword at output of electronic encoders. The difference between

the delay times decides the pattern of the codeword. For example, the codeword E (1,5,11) in

Figure 5.8 has the differences between delay times

C

C

TttTtt

104

13

12

=−=−

(5.9)

where T is the duration of one chip and is the inevitable time delay in the first branch

at the implementation. When chip clock is 2.5GHz, = 0.4 ns, t

C 1t

CT 12 t− = 1.6 ns and t 13 t− =

4.0 ns. Using Eq. (5.7), the length of the delay line can be calculated. In practice the real delay

times have to be fixed by the measurement, because the electrical circuit and PLC also

influence the delay time. The measurement of the delay time isn’t difficult by using a high

frequency oscilloscope with 10GHz bandwidth. Then the difference between the delay times

can be accurately adjusted. The block diagram of an electronic encoder is shown in Figure

5.10 a). By using this scheme we can implement all codewords with through

adjusting the length of the delay line. For example, Figure 5.10 a) shows the delay times of

4≤w

81

the circuit for a codeword (49; 1,3,14,18) with code weigh 4, and b) shows its measured

waveform.

Fan-out

Buffer

Fan-out

Buffer

Fan-out

Buffer

ORFromimpulse generator

Output of code pattern

t1

t2=t1+2TC

t3=t1+13TC

t4=t1+17TC

Figure 5.10 a) Block diagram of electronic encoder with codeword (49; 1, 3, 14, 18)

Codeword

Data Clock

Figure 5.10 b) The waveform of codeword at the output of the electronic encoder

82

5.4.3 Optical Encoder

In principle the structure of the optical encoder in Figure 5.11 is similar to the electronic

encoder, except that the former uses optical instead of electrical delay lines and the stimulated

optical signal instead of electrical one. There are equivalent circuit structures from signal

processing which reduce the total loss [15]. As in our experimental system the overall loss is

not a problem, we take the structure in Figure 5.11. For codewords with code weight less than

4, the main components, 1:4 optical couplers, are available on the market. The insertion loss

of optical power is about 6.5dB with a little deviation. After the optical splitting and adding,

each optical chip impulse in the codeword has about 13dB loss compared to the input

impulse. For implementation the deviations could be reduced a little by carefully choosing the

connected pairs of parts of splitter and adder with the aim of balancing the optical output

between the impulses.

t1

1:4(6dB)

4:1(6dB)

OpticalCoupler

OpticalCoupler

t2

t3

t4

Figure 5.11 The structure of the optical encoder

Due to the physical structure of the optical couplers, the delays of each port are different, and

this difference is unknown. Therefore, for implementation of our prototype optical encoders,

the delay times of each branch have been measured, and a broadband optical transmitter and

receiver are required. Depending on the parameters of optical fiber at the connection, the

accuracy of the delay time is within 10ps (corresponding to about 2mm fiber). As it is

analyzed in chapter 4, the output codeword is accompanied by the interferometric noise,

whose intensity depends on the EXT of the optical source. An optical encoder with codeword

(19; 1, 5, 11) is implemented. Figure 5.12 shows the output signal with different EXT of the

optical source. Obviously EXT of only 10dB is not sufficient.

83

a) with EXT=10dB b) with EXT=17dB

Optical Output

Bit Clock

Optical Output

Figure 5.12 Output of the optical encoder with codeword (19; 1, 5, 9)

5.4.4 Electronic Decoder

Using electronic decoder has many advantages such as requiring only one optical/electronic

converter for the whole system, avoiding the optical interferometric noise, and reducing the

BER when its input signal is processed by an electrical decision. Logic electronic decoders

are implemented by using the same structure of the electronic encoder as shown in Figure 5.

10 a). In addition, because only when the input signal matches the pattern of codeword, the

decoder outputs an impulse, the logic decoder can restrain the output of the sidelobe. This has

the benefit for clock recovery. Figure 5.13 shows the output of such a logic electronic decoder

with matched encoded input, and one can see that the output is very pure.

84

Output of Decoder

Data Input

Figure 5.13 The output of the electronic encoder

5.4.5 Optical Decoder

The optical decoder is implemented with the similar structure as the optical encoder, the

difference is the length of delay line. The output of the optical decoder is disappointing, when

the data is encoded by optical encoder. The interferometric noise is so large that the decoded

signal in Figure 5.14 can hardly be decided by a threshold detector. The cause for that noise

has been analyzed in chapter 4.

Figure 5.14 The output of the optical decoder with input signal coded by optical encoder

85

5.5 Optical Transmitter and Receiver

5.5.1 Electronic/Optical Converter

The E/O converter is an important functional block of this demonstration system. Currently

high speed optical transceivers are widely used in commercial optical systems. But we cannot

directly use such sophisticated transceivers of a normal on-off-keying system, because in

OCDMA system chip stream isn’t the same as the data stream in the normal optical systems:

the total duration of chips is much lower than Null duration. As a consequence, APC

(automatic power control) circuit in optical transmitters would alarm that the optical power is

too low, and would push more current into the Laserdiode. Hence we have to establish

dedicated E/O converters by ourselves.

At the beginning of the project, the Laser module Alcatel 1915 LMM was chosen for this

system after consideration of the parameters, the cost and the requirement of upgrade. This

Laser module contains a 10Gbit/s digital WDM-LD Laserdiode of DFB (distributed feed

back) type with monolithically integrated electro-absorption modulator (EAM). This outer

modulator provides much lower dispersion penalties than a directly modulated DFB Laser.

Before the end of 1988, this kind of Laser modules were composed of a DFB Laser and an

expensive complex external modulator based on Lithium-Niobat technology. Instead of an

external modulator, an EAM is set on the basis of Indiumphosphide (InP), which is integrated

in an InP WDM-laser-chip. The inner structure and the outlook are shown in Figure 5.15.

S o u r c e : A l c a t e l O p t r o n i c s

Isola tor

R m a x . R m a x .

R m i n .

P a s s i v a t e d F i lm

M o d u l a t o r

D F B L a s e rC o n n e c t e d S u r f a c e

f rom Elec t ro ly t i c Go ld

Isolat ionbe tween E lec t rodes

Figure 5.15 The structure of DFB Laser with EAM

86

Besides the Laserdiode and modulator, the module Alcatel 1915 LMM also contains a

thermoelectric cooler (TEC), a thermistor resistance (TR) and a monitor photodiode (PD). We

design all of the circuits in order to build a safe operation environment and monitor the

working state. Figure 5.16 shows the block diagram of the E/O converter using this Laser

module.

DC Driver

Laser modul1915 LMM Optical Fiber

Optical powermonitor circuit

ATCcircuit

Data

Input

Alarmsignal

Driver for EAM

Figure 5.16 Block diagram of the E/O converter

ATC circuit: An automatic temperature control circuit of Laser module is important for

the safety of the Laserdiode and for reducing the wavelength drift, which is especially

important for a DWDM system. By using the TR and TEC a closed-loop of ATC is designed,

the block diagram is shown in Figure 5.17. The TR, which has a negative temperature

coefficient, acts as a sensor for temperature in the Laser module. It detects the variation of the

temperature within the Laser module and converts it into a voltage signal. The low pass filter

will drop the higher frequency components within the signal, in order to give time for TEC to

response and avoid the self-excitation in the closed-loop, because the process of cooling by

TEC is relatively slow. After the filter the signal will be amplified for TEC. Normally the

capacity of the driving current is up to 800 mA. In practice by adjusting the driving current

through the TEC, the operating temperature of the Laser module can be controlled within the

desired values.

Detection oftemperature

variation

Low pass filter

Signal amplifier and driver

for TECTEC

TR

Figure 5.17 The block diagram of ATC circuit

87

Optical power monitor: In order to monitor the operational state, we designed an

optical power monitor circuit. Figure 5.18 is the block digram of the monitor circuit. The PD

detects the optical signal that is proportional to the optical power output from the pigtail fiber.

After the preamplification the signal is filtered by the low pass filter, in order to obtain the

mean power signal. This signal will be processed by the main-amplifier and failure decision

circuits, then the resulting voltage is displayed at the front panel by a warning LED. When the

optical power is much smaller than normal, the warning LED should be turned on.

PreamplifierLow pass

filter

Main-amplifierand

Failure decisionPD

Alarmsignal

Figure 5.18 The block diagram of the optical power monitor circuit

We have to pay attention to the value of the optical power at the monitoring diode and

especially at the measurement of the optical power in OCDMA system. Here the input signal

of the E/O converter is different from normal digital optical transmission systems, because the

signal at output of the OCDMA encoder is not balanced between Nulls and chips (Marks or

impluses). The ratio of the “1” chips in an OCDMA channel is:

1PLwR

Cchip ⋅= (5.10)

where w is code weight and is the length of the codeword and is the probability of the

Mark of users’ data (in general it is about 0.5). For example, when w equals to 4, equals to

49 (codeword is 49; 1, 3, 14, 18) and equals to 0.5, from Eq. (5.10) is only 4.08%.

The output of the optical power monitor circuit is only the mean optical power of the output

of the Laser module. Because is much smaller than in a normal system, it has to be

considered here, that when the mean optical power is the same, the peak value of optical

power of an OCDMA encoded signal is many times higher than that in a normal system. In

the above example, the peak optical power of the chips is 11dB bigger than that in a normal

stream when the measured mean optical power is the same.

CL 1P

CL

1P chipR

chipR

DC Driver: The Laser module needs a DC power driver for the DFB to emit continues

lightwave. Due to safety consideration, the DC power has to have a good performance for the

88

on and off operation, i.e. it has very slow rise time and fall time without impulse and tide to

avoid harm the LD. Consider that the maximum forward voltage of the Laserdiode is 2V and

the maximum forward current is 150mA, we designed a circuit to fulfill this function as

shown in Figure 5.19. LM317 is a 3-terminal adjustable regulator that is capable of supplying

a 1.2V to 37V output range. They are very easy to use and require only two external resistors

to set the output voltage. Here the power supply is 10V, then the voltage at point A is set to be

6 V by adjusting R1. T1, R5 and C1 compose the slow starting circuit. When R5 is big

enough, it generates a relative long time to charge the big capacitor C1. IC2 and R2 compose

a precision current limiter, the current is limited to:

2lim

45.1R

VVVI BA −−

= (5.11)

IC3, R3 and R6 compose the adjustable regulator, which limit the voltage of the output. D1 is

a diode for protecting the Laserdiode against reverse current in case of accident. R7 is a

resistor for limiting the current, in practice to 60mA. When the Laserdiode operates with a

current of 33mA, the optical output power is +0.7dBm.

LM317 LM317 LM317input output

R1T1

C1

R2

R3D1

R4

R5

R6

IC1 IC2 IC3

R1=1KΩ R2=100Ω R3=500ΩR4,R6 =240Ω R5=50kΩ C1=100ηF

A B

10V

C R7

Figure 5.19 The circuit for DC power adjustment of DFB Laser

Driver for EAM: We used FMM311 GaAs Laser driver by Fujitsu as the high frequency

driver for electro-absorption modulator. FMM311 is a high-data rate driver circuit designed

for fiber optic transmitters operating at data rates up to 2.7Gbit/s. The device is capable of

driving high-power Laserdiodes at peak current up to 80mA, typically. The block diagram of

laser driver circuit is shown in Figure 5.20 (a), and the high frequency equivalent is shown in

Figure 5.20 (b).

89

EA

LD

(a) (b)

Driver

IRLR IR

LR

50ΩCoaxial line ecodeE

DI oVoV

DI

Figure 5. 20 (a) Block diagram of driver and Laser module; (b) the high frequency

equivalent

Note that the output of the FMM311 is a current source instead of a voltage source required

by the EAM in Laser module. The output voltage V is given by 0

DLI

LI IRRRRV

+⋅

≈0 (5.12)

where is the load resistor of coaxial cable which equals to 50Ω, and is a resistor at the

driver side. When is also selected as 50Ω, the result is:

LR IR

IR

DIV ⋅Ω= 250 (5.13)

For example, when is limited to 60mA, the maximum V is 1.5V. DI 0

0 0.5 1 1.5 2 2.5 3−30

−25

−20

−15

−10

−5

0

5

10dB

V

The input voltage at the EAM

The

optic

al a

bsor

ptio

n

Figure 5.21 The optical absorption capacity versus the input voltage at the EAM of

Alcatel 1915 LMM

90

From the analysis in chapter 4, for optical encoder the EXT has to be larger than 17dB. For

A1915LMM Laser module, it is difficult to reach this value. This is related to the

characteristic of EAM, the function curve of which is shown in Figure 5.21. The maximum

EXT of A1915LMM is only 14dB, so this optical modulator can not be used in OCDMA

transmitter with optical encoder, because large interferometric noise will occur at the output

of the optical encoder. In order to demonstrate the OCDMA processing with both optical

encoder and optical decoder, we also implement the O/E converters with directly modulated

DFB-Laser module, A1915 LMI. The block diagram of its structure is shown in Figure 5.22.

The function blocks are almost the same as those in A1915LMM. Here EXT can reach 30dB.

Laser driver Laser moduleOptical Fiber

Optical powermonitor circuit

ATCcircuit

Input

Alarmsignal

Figure 5.22 The structure of the O/E converter with direct modulation Laser module

After optimizing and testing, we have implemented the E/O converter with good performance

and safety.

5.5.2 Optical Receiver

The optical receiver of OCDMA system consists of O/E converter, electronic amplifier and

decision as in usual systems.

O/E Converter: In this system we use an Agilent 11982A amplified lightwave converter

as the optical/electronic converter. It combines a PIN photodetector with a low noise

preamplifier and is a sensitive O/E converter with wide bandwidth. Agilent 11982A covers

the wavelengths from 1.200nm to 1.600nm and bandwidths from DC to 12GHz. Its 300

volts/watt conversion gain and 0.05% input optical reflection significantly improve sensitivity

for the system.

Amplifiers: The electronic signal from O/E converter will be amplified by broadband

amplifiers. The Model 5850 ultra-broadband amplifiers produced by Picosecond are selected

due to the attractive price/performance ratio. They are AC coupled and are extremely

broadband covering the range of 5 decades from 80kHz to 5GHz. They have clean transient

91

responses and smooth gain vs. frequency responses. In practice, due to the AC coupling of the

amplifiers and unbalance of the signal waveforms, the gains and output range of the

amplifiers are not fully used. The amplifiers must be connected in cascade for higher gains

and in this case the noise is increased. Because the Model 5850 Amplifier is a static-sensitive

device, static discharge must be avoided.

Decision: After being amplified, the signal of receiver will be decided in order to convert

analog signal into digital signal for electronic decoders. Nowadays we can not find on the

market a suitable decision device that operates at 2.5G chips/s. Hence we use a logic gate with

a short rise/fall time of 150ps as a decision element. The level of decision is important for the

correct transmission. We use a high speed ECL IC with two inverse inputs, one is the input of

the analog received signal, and the other is the DC level for offsetting the threshold which

could be reset when the structure of the system is changed.

5.6 Data Recovery

Data recovery is composed of clock recovery and the data sampling. The decoded signal is

despreaded into a series of impulses, in which not all of them means Marks of desired user’s

data. There are many noise impulses produced by the other users’ interference, which mostly

locate at the fail position (when the position is right, it could be the source of the bit errors).

Hence there are two difficult points at the data recovery of the OCDMA system: one is

restoring clock from signal with fail information and the jitter of the clock must be much

lower than the chip duration in order to correctly sample the very slim auto-correction

impulse; the other is adjusting the sampling time at the center of the auto-correction impulse

in order to avoid the phase swing of the restored clock by the temperature.

5.6.1 Clock Recovery

The clock recovery is achieved by using a phase locked loop (PLL) circuit. The basic block

diagram of the structure is shown in Figure 5.23 a). During the implementation of the PLL

circuit, in order to achieve a clock with low jitter, three measures are taken in the design.

Firstly, an impulse processing circuit is used to shape and broaden the impulses. After

broadening of the impulse, the bandwidth of the signal is reduced, so the phase detector can

operate at relatively low frequency. Secondly, the voltage controlled oscillator (VCO)

operates at multiple frequencies of the clock, therefore PLL is more sensitive for the

92

controlling of the VCO. Thirdly, an active filter is built as the loop filter in PCB, so PLL has a

larger searching range than the passive filter [97]. The parameters of the filter have been

optimized in order to achieve a restored clock with lower jitter. Figure 5.23 b) shows the

result of the recovered clock.

Impulseprocessing PD LF VCO

Clockoutput

1/N

a) the block diagram of the PLL

b) the result of the clock recovery

Input signal

Recoveredclock

Figure 5.23 The block diagram of the PLL and the measurement result

5.6.2 Phase Tracking

Using the general PPL clock recovery, the phase of recovered clock might vary around the

correct position, and the variance might change with temperature. In order to sample the clock

signal at a very accurate position, the data recovery must include a special circuit to find the

very accurate time point for sampling the data correctly. Here we design a data recovery with

automatic phase tracking circuit for OCDMA receiver.

The auto-corrected impulse from the decoder has a very small width, only about 400ps, and

the range of correct sampling time must be as small as that. In order to avoid the bit error at

the sampling, the range of correct sampling time is much smaller. So when the circuit directly

searches and follows the changeable clock phase, the circuit has to operate with high

93

frequency, e.g. about 5GHz, which is too high for our demonstration system. In order to

reduce the operation frequency, we design a signal sampling array, in which the signal is

sampled by n D-flip-flops, whose clocks have equal added delay time (see Figure 5.24).

When the delay time of clocks of D-flip-plops exceeds a bit duration, i.e. DTn ≥τ , and the

delay unit, τ , is small enough, for example CT31=τ , then from the outputs of the sample

array, at least there is a sampled signal which is the correctly sampled signal of the

decoded signal. can easily be found. Because for the decoded signal the noise impulse is

far less than the auto-correction impulse, thus only correctly sampled signal has the

balance of Marks and Nulls (assumed that user’s data is balanced with Marks and Nulls).

Within a relative long check time (for example 1000 ), we can easily find the correctly

sampled signal from the output signals of the sampling array. Meanwhile the signal

and also have more Marks than other sampled signals. From the change of the and

, one can find the tendency of the drift of clock phase and select a new correct sampled

signal or as . This process of statistic selection operates at the user’s data rate, and

can be implemented by the components with relatively low operation frequency.

iS

iS

iS

1+

iS

DT

1−iS

1+iS

iS

1−iS

1+iS

1− iS iS

Signal

Splitter

D

D

D

Clock driver : delay unit,Clock

DecodedSignal

S1

S2

S n

for example equal to the1/3 width of impulse.

CK1

CK2

CKn

τ2

τ1

τn

τn

τ)1( −n

τ

τ

Figure 5.24 The sample array

The above scheme needs too many sampling units ( τDT> ), and it is complicated to realize

the signal splitter and clock delay. Therefore, we have designed a scheme operating actively

with a variable delay circuit shown in Figure 5.25. The blocks Sample Array and the Clock

94

Recovery from Figure 5.24 have been introduced. After tradeoff between reduction of the

number of sampling units and simplification of the following signal processing, the range of

sampling time of sampling array must be over one duration of chip, CT . The block Variable

Delay is composed of digital programmable delay chips (MC100E196) with a delay step

resolution of 20ps and delay range of 2ns (when data rate is 150Mbits/s, then DT is 6.67ns,

the delay range of only 4 cascaded chips can be over DT ). The block Logic Processing can be

implemented by using field programmable gate arrays (FPGA).

SampleArray

Logic processingat data rate

Clock Recovery

variable delay

Decoded Signal

RecoveredData

Delay-Control-address

Sn

Figure 5.25 The block diagram of the data recovery

SearchingData

TracingData

Find data / Adjust low-address-bits

Do not find data / Adjust high-address-bits

Find data / Adjust low address-bitsDo not find data / Adjust high-address-bits

Figure 5.26 The state diagram of the logic processing

95

The state diagram of the block Logic Processing is shown in Figure 5.26. Because we pay the

main attention to the OCDMA processing and optical noise, the basic circuit of phase tracking

was implemented.

5.7 Channel Code

In order to reduce the BER for this demonstration system, a suitable RS channel code is

selected, which is fairly well developed in hardware. The insertion of RS code into a CDMA

network should have little cost impact on the overall network implementation. But all of the

products available in the market are designed for asynchronous operation mode or

synchronous operation mode with relative low bit rate (10Mbit/s), in other words they are

designed for use at the information source and information sink, not for transmission link. We

have designed a scheme with the chips of RS channel code. The main problem of this scheme

isn’t the RS code itself, but the design of the interface for the chips with the asynchronous

operation mode at data rate 155Mbit/s. Converting clocks and searching the header of the

code frame are the main parts of the scheme. In Figure 5.27 we give the block diagram of this

scheme. More time is needed before the hardware can be completely implemented.

F.249/255

Series/Parallel

R-SEncoder

Parallel/Series Buffer

Clock

f=155.52MHzClock Convertor and

Logic Controller

DataEncoded

Data

50MHz Initialization

b) Decoder of R-S code for (255/249)

Clock

f=155.52.255/249 MHz

DataEncoded

Data

f=155.52MHz

Series/Parallel

R-SEncoder

Parallel/Series Buffer

Clock Convertor andLogic Controller

50MHz Initialization

a) Encoder of R-S code for (255/249)

Figure 5.27 The block diagram of the RS channel encoder and decoder. The RS code is

255/249, the redundancy is 2.41%

96

5.8 Measurement Results

In the last step, we combined the functional blocks and built a demonstration system with 3

OCDMA transmitters and an OCDMA receiver, which can receive the signals of all

transmitters through changing the decoders. The experimental setup is shown in Figure 5.28.

The codeword for the 3 transmitters are (19; 1, 2, 9), (19; 1, 3, 6) and (19; 1, 5, 11) for

19=CL and 3=w . The waveform at each point in Figure 5.28 is given in Appendix B. The

actual FWHM of the impulse is smaller than 300ps (see Figure B.3, B.5 and B.8).

Bit Error Measuring Set

Wandel&GoltermannPF-4

PRBS27-1

Clock155.52MHz

PRBS 223-1

TX 1Impulse

Generator

Codeword:(19; 1,2,9)

Alcatel 1915LMM?=1540.35nm

ElectronicEncoder

EAMdriver

E/Owith EAMScrambler

PRBS221-1

Clock155.52MHz

TX 2Impulse

Generator

Codeword:(19; 1,5,11)

Alcatel 1915LMI?=1549.62nm

ElectronicEncoder Driver E/O

DFB-LDScrambler

TX 3Impulse

Generator

Codeword:(19; 1,3,6)

OpticalEncoderDriver E/O

DFB-LDScrambler

Alcatel 1915LMI?=1550.12nm

OpticalCoupler

(4:1)

Clock155.352MHz

RX

DataRecovery

Codeword:(19; 1,3,6)

O/EConverter

ElectronicDecoder Amplifiers

Agilent 11982A

OpticalDecoder

Figure 5.28 The experimental setup of the OCDMA demonstration system

When the encoders and decoders are implemented with chip duration CT =500ps, the

maximum data rate can be up to 105.2Mbit/s, it can transmit 100Mbit/s data with maximum

5% redundancy introduced by the channel code. The BER of the system with electronic

decoder is at the level of 1110− with the interference of other two transmitters. When only one

or two transmitters are active, the BER is zero. This result is satisfying for access network.

Based on above experiment, when we change CT to 330ps for all encoders and decoders, the

maximum date rate can be up to 159.5Mbits/s, it can transmit 155.52Mbits/s data with

97

maximum 2.55% redundancy of channel code. The BER at this data rate is only 10 . The

reason of high BER is that the impulse is not slim enough for T . This verifies that

FWHM of the impulse must be smaller than the theoretical chip duration for BER free

system.

5−

psC 330=

For this demonstration system, because the number of users is only 3, and the code weight is

3, the interference of the other users can not produce severe noise impulses. A failure impulse

only exists when the desired data is Mark, i.e. the noise impulses depend of the data signal. So

the requirements for the clock and its phase are not strict.

From the measurements, it is noted that the OCDMA system is not as sensitive to the

transmission of long strings of NRZ “1” bits of users’ data as in normal optical system. The

BER is almost the same when the transmission data pattern is all “1”, PRBS or 010101

sequence, respectively. The reason is that the transmission band is independent on the pattern

of data, but it depends on the CDMA processing.

Through this experimental system, the possibility of using OCDMA with electronic

en/decoding process in access network has been demonstrated.

98

Chapter 6 OCDMA Combined with DWDM

As it has been shown from the results of the analysis and demonstration system, OCDMA

technology with electronic encoder and decoder can be easily implemented within a narrow

optical spectrum, hence it is possible to insert an OCDMA group into a WDM channel. When

OCDM and DWDM are combined in access network, by using todays mature technology, the

number of accessed users and the throughput can satisfy the demand easily. In this chapter we

will discuss the feasibility of the OCDMA/DWDM system. At first the development of

DWDM will be reviewed, then the noise in OCDMA/DWDM system will be calculated and

its performances will be evaluated. Finally an example of system structure will be given.

6.1 The Fundamentals of WDM

By using WDM two or more optical signals having different wavelengths may be combined

and simultaneously transmitted in the same direction over one fiber. These signals are then

separated by wavelength at the distant end. Distinguishing from the coarse WDM (CWDM),

dense WDM (DWDM) is a technology that can increase capacity of a fiber span by many

folds. The wavelengths implemented by DWDM are located in two different bands, i.e. C

band (1529–1541nm) and L band (1549–1577nm). Current DWDM systems can carry 80

channels [86] and systems supporting more than 250 channels have been reported [85]. Each

channel could operate at up to 10Gbit/s or even 40Gbit/s.

Since WDM is a protocol-independent technology, it could be used in almost all network

architectures, such as ATM over DWDM and IP over DWDM. Certainly OCDMA could be

also easily combined with DWDM, when spectrum of each OCDMA group is within a

DWDM transmission band. Because of the high cost, early implementations of WDM systems

are limited to backbone network. However, new developments in materials and techniques

allows manufacturers to reduce the cost of WDM systems significantly. As a result, WDM

system begins to appear in MANs and LANs.

In DWDM system, optical filters are the critical components, the characteristics of which

determine the performance of DWDM. Different types of optical filters can be used in

99

DWDM systems, such as Fabry-Perot filters, grating filters, interference filters, etc. We will

not go into details of the structure on these devices, since they can be simulated according to

their characteristics. We only pay attention to the results of the filtering.

Generally this kind of devices can be simulated by means of transfer function H(ν). For

instance the transfer function of a Fabry-Perot filter can be written as [88]

∆−

−

−=

νννπ

ν)(2exp1

1)(0iR

RH FP (6.1)

where is the reflectivity of the interferometer mirrors and ∆ν is the spacing between

adjacent resonant frequencies. To obtain a good quality filter, the interferometer loss has to be

small, this implies 1-

0>R

R <<1. In this approximation, the 3-dB bandwidth of the Fabry-Perot

filter, Bo, is given by

νπ

∆−

=RRBo

1 (6.2)

Generally, the factor )/()1( RR π−=ℑ , that depends only on the reflectivity of the mirrors,

is called filter finesse. So the filter is characterized by the free spectral range ∆ν and the

finesse . ℑ

Around a given resonance peak the transfer function of a Fabry-Perot filter can be

approximated by a Lorentzian function:

0

0 )(1

1)(

Bi

H FP ννν

−−

≈ (6.3)

The demand for bandwidth has accelerated the deployment of DWDM filters. Besides the

Fabry-Perot filter, the Fiber Bragg Grating filter which features flat top response with steep

spectral skirt also represents a key element in fiber optical communication systems. Due to

their fabrication process, they have very low insertion loss, immunity against electromagnetic

interference, electrical isolation and light weight. DWDM gratings can be designed for

channel separations down to 25GHz and used to select the different wavelengths of the grid.

However, for the application in access network with large access number, the use of WDM

would require special steady optical source and narrowband optical filters. The smaller the

bandwidth, the stricter the parameters of components have to be.

100

6.2 The Feasibility of the CDMA/WDM System

When an OCDMA system uses optical source with narrow bandwidth, it is compatible with

components, devices and strategies used in the wavelength division multiplexing area. For

this reason, it appears to be a valid solution for quick and potentially low cost solution. A

network employing OCDMA/DWDM is capable of establishing more traffic connections than

a network using WDMA only. In OCDMA/DWDM systems the noise impact is greater than

in OCDMA system.

FWM Distortion: Four wave mixing (FWM) is one of a broad class of harmonic

mixing or harmonic generation processes. This is a strong nonlinear effect that mixes the

frequencies of optical channels in the 1550nm band to generate distortions in that band.

However FWM is still a weak effect, in the sense that it only accumulates when the signals on

the optical channels remain in phase with each other over a long distance i.e. in long distance

optical systems.

The Interferometric Noise: From the analysis in chapter 4, when two beams of the

different optical sources are overlapping, the interferometric noise occurs. Figure 6.1 shows

the spectrum of the interferometric noise from the overlap of two CWs, which have the

bandwidth λ∆ . If the two CWs have the same central wavelength, the spectrum of noise is in

base band with the bandwidth 2λ∆ as shown in Figure 6.1 a). If the two CWs have different

wavelength, the spectrum of noise with the bandwidth λ∆ locates at the central frequency

12 λλ − as shown in Figure 6.1 b). )/( 0λλF is the Laplace transform of the impulse response

of the equalized filter of the optical receiver and amplifiers, given in Equation 4.21, where 0λ

is equal to the effective bandwidth of the optical receiver. The interferometric noise between

two DWDM channels is very small because of the differences between the center frequencies

of each channel.

101

E

E

Spectrum of the CW1

Spectrum of the CW2

E

Spectrum of the interferometric noise of theoverlap of CW1 and CW2

E

E

Spectrum of the CW1

Spectrum of the CW2

E

Spectrum of the interferometric noise of theoverlap of CW1 and CW2

λ∆

λ∆

λ∆

λ∆

2λ∆

λ

λ

λ λ

λ

λ1λ

2λ

1λ

2λ

12 λλ −

)/( 0λλFλ∆

a) in the case of 21 λλ = b) in the case of 21 λλ ≠

)/( 0λλF

Figure 6.1 The spectrum of the interferometric noise from overlapping of two CWs

For the users which are within an OCDMA group located in a DWDM channel, the

interferometric noise produced by the interference from other users, have to be calculated. By

using Eq. (4.10), two Laser modules, which are composed of the DFB laser and EAM, were

simulated. Since external modulators are used, their optical products have no chirp effect.

Therefore the spectrums of the modulated optical lightwave could be simplified and it is

possible using CWs instead at the modulated lightwave in simulation to find the noise

characteristics. Because the distance in access network is short, we don’t consider the FWM

as a significant noise here. Figure 6.2 shows the simulated result of the interferometric noise

of the overlapping of two CWs with same wavelength. The center wavelength of two Laser

modules is 1.550nm, and the bandwidth is 1GHz. For the overlapping of the optical source

with a narrow bandwidth at same wavelength, the interferometric noise is so strong in low

band range, that the impulses in a codeword could be counteracted and this interference

would restrict the usage of coded multiplexing.

102

time

norm

aliz

ed a

mpl

itude

ns

Figure 6.2 The simulated interferometric noise of the overlapping of two CWs with same

wavelength

Figure 6.3 The simulated interferometric noise of the overlapping of two CWs with

different wavelength ( GHzff 1012=− λλ )

When the difference of wavelength between laser sources ( 12 λλ − in Figure 6.1) is larger than

the sum of the bandwidths of chips in the OCDMA and the bandwidth of optical sources, the

interferometric noise could be well reduced after the filter for receiving the chips in electronic

103

domain. In OCDMA/DWDM system, the wavelength of optical sources for one OCDMA

group have to be well chosen to suit the DWDM channel and reduce the interferometric noise.

Figure 6.3 shows the simulation results of the noise performance of two overlapped

lightwaves with GHzff 1012=− λλ

D

3=N

. In this case, after the filter which suits for the

transmission of baseband binary data with 2.5Gbit/s, the interferometric noise has very little

effect on the transmission of the chips for OCDMA.

6.3 Network Architecture and Management

A network architecture of the proposed OCDMA/DWDM technology is shown in Figure 6.4.

N users will be supported within an OCDMA coding group which transmits in a single

DWDM channel. The total bit rate of one OCDMA group through an optical channel is

, where is the bit rate of the users’ data and it is implemented up to 155Mbit/s in

the demonstration system using electronic en/decoder. The choice of the number of optical

DWDM channels m, depends on the tradeoff between the requirement and cost. The

frequency grids of channels are given by ITU-T Rec. G.692. The filters for grids with

200GHz, 100GHz, 50GHz and even 25GHz are commercially available. We have measured a

fiber Bragg grating filter with 100GHz grid, its 3dB pass band is about 30GHz. For an

OCDMA system with a chiprate of 2.5Gchip/s, the results of the simulation show that this

bandwidth can allow accessed users per wavelength. In our system, the users will be

divided into some CDMA groups by the location. The users in the same CDMA group will

operate with the same wavelength of a DWDM channel, and the CDMA codes differ from

each other. When the users are located close together, they can directly use electronic CDMA

through electrical cable. Then the electronic signal is converted to an optical signal, e.g. group

2 in Figure 6.4.

DRN × R

In the longer term, OCDMA in combination with WDMA is thus worthy for further

investigation with the goal of implementing a wide spread, high capacity and easy to be

managed optical telecommunication infrastructure, especially for access network. In access

network bit rate is not so high that it is limited by electronic components and the distance in

access network is shorter then in backbone network. But the requirement of a cost effective

network architecture and network management is higher than in the backbone network.

104

Electronic Data 1

E/O Optical CDMA-Coder

ImpulseGenerator

PON

CDMAdecoder

DatarecoverDesision

CDMAdecoder

Datarecover

Amplifier

OpticalFilterλ1

Electronic Data 1

E/Oλ1

Optical CDMA-Coder

ImpulseGenerator

E/Oλ1

Optical CDMA-Coder

ImpulseGenerator

O/E

Electronic Data 1

E/Oλ2

ImpulseGenerator

Electronic CDMA-Coder

ImpulseGenerator

Electronic CDMA-Coder

ImpulseGenerator

Electronic Data N

CDMAdecoder

Datarecover

CDMAdecoder

Datarecover

OpticalFilterλ2

O/E

CDMAdecoder

Datarecover

CDMAdecoder

Datarecover

OpticalFilterλn

O/E

Electronic Data 1

E/O Optical CDMA-Coder

ImpulseGenerator

E/Oλn

Optical CDMA-Coder

ImpulseGenerator

E/Oλn

Optical CDMA-Coder

ImpulseGenerator

Electronic Data N

Electronic Data 1

E/O Optical CDMA-Coder

ImpulseGenerator

PON

CDMAdecoder

DatarecoverDesision

CDMAdecoder

Datarecover

Amplifier

OpticalFilterλ1

Electronic Data 1

E/Oλ1

Optical CDMA-Coder

ImpulseGenerator

E/Oλ1

Optical CDMA-Coder

ImpulseGenerator

Electronic Data N

O/E

E/Oλ2

ImpulseGenerator

Electronic CDMA-Coder

ImpulseGenerator

Electronic CDMA-Coder

ImpulseGenerator

CDMAdecoder

Datarecover

CDMAdecoder

Datarecover

OpticalFilterλ2

O/E

CDMAdecoder

Datarecover

CDMAdecoder

Datarecover

OpticalFilterλm

O/E

E/O Optical CDMA-Coder

ImpulseGenerator

E/Oλm

Optical CDMA-Coder

ImpulseGenerator

E/Oλm

Optical CDMA-Coder

ImpulseGenerator

OCDMA group 1

OCDMA group 2

OCDMA group m

Figure 6.4 The structure of the PON access network with OCDMA/DWDM system

105

Chapter 7 Conclusions

Though the technology of CDMA has been already successfully used in wireless data

transmission, OCDMA remains outside the mainstream of the research in optical

communications. However, with the progress in the optical network technology and the

increasing demand for simple access protocol, the research on OCDMA has become active

recently, and it would be a potential candidate technology for future optical networks,

especially for optical access networks. In this thesis, we present the results of our research in

the field of OCDMA communications. Our intention is to explore the possibility of an

OCDMA system using electronic en/decoding and the possibility of its combination with

DWDM. Therefore, using the narrow spectral optical sources ready for insertion into a

DWDM channel, we implemented the OCDMA demonstration system for the comparison of

electronic and optical en/decoding.

We have evaluated the current experimental research in the OCDMA field, the criteria include

the achieved number of access users, the requirement of spectral width, feasibility for a

practical system, etc.. The results show that none of the contemporary real OCDMA systems

can demonstrate the primary advantages of OCDMA technology, such as a large access

number.

We observed the influence of the interferometric noise produced by overlapping light waves

from the same optical source during optical encoding and decoding. The noise produced at the

optical encoder could be reduced to an acceptable level by increasing the extinction ratio of

the output of the electronic/optical converter. But at the optical decoder all chips within a

designed codeword have to be overlaid to despread the signal. By analyzing the optical

encoding and decoding processes, we concluded that in an OCDMA system the

interferometric noise is one of the main sources for system limitation. With a coherent optical

decoder, the interferometric noise can be avoided, as long as complete coherent decoding is

achieved. In an optical decoder the arriving phase of each chip is very sensitive to the small

wander of temperature, thus the temperatures of en/decoders have to be controlled.

Furthermore, from the tiny variance of temperature (<10 ) remaining under the C°−5

106

temperature controlling, it could produce big phase drifting and strong noise. Then a servo-

control for phase drift tracking has to be set for each branch within the decoder. Furthermore,

when code weight is larger than two, the information about the phase in each branch cannot

be recovered from the received signal. We conclude that without a great technical

breakthrough, the OCDM technology uniquely using optical en/decoding processes cannot

compete with gradually matured DWDM in the near future.

Using electronic encoding and decoding in an OCDMA system could avoid the above

mentioned interferometric noise. We have examined the performance of OCDMA systems

with an electronic en/decoding process through the implementation of a demonstration system

with three users, each operating with a data rate up to 155Mbit/s. With the improvement of

clock frequency of electronic components in the future, the scheme with electronic coding for

higher chip rates up to 10Gchips/s, even 40Gchips/s could be possible. In addition, a logic

electronic decoder has the function of the hard-limiter, therefore the BER will be reduced. To

my best knowledge we are the first to give an algorithm for exact calculation of the BER of

the OCDMA system with a hard limiter.

Based on the implemented demonstration system, we have analyzed the possibilities of an

access scheme combining OCDMA and DWDM. When the differences between the center

frequencies of the optical sources of an OCDMA group are bigger than the bandwidth of the

chips, and all spectrum used by OCDMA group is within a DWDM channel, the noise

performance will be good enough to achieve the transmission. The combination of OCDMA

and DWDM might find its market in access networks, due to larger access number and higher

flexibility of the network architecture. The results of our work on the application of OCDMA

over DWDM show that this scheme by using matured technology can be adapted to the

requirements imposed by the access network.

107

Appendix

Appendix A1 An Algorithm of BER with Hard-limiter

Here we give an algorithm of the BER in OCDMA system with hard-limiter at the receiver.

An optical hard limiter converts the channel from an “adder channel” to a “binary OR-

channel” by removing intensity information from the fiber beyond the two levels of “off” and

“on”. In contrast to the published algorithm in [22, 23, 28], which only gives an upper bound

limit, we give a exact calculation of BER with hard-limiter, that is suitable for simulation with

computer programs.

We will model the network as chip synchronous, and T will be the unit of time shift. In

other word, T (the time duration of one bit of information) is divided into segments, i.e.,

. Here we assume that the codeword is , i.e. only one chip overlap

for the auto- and cross-correlation, and N users are addressable by the code family. The a

priori probability of Mark and Null for all users is represented as and . We define

as the probability of corresponding values i and j, in which i is the number of all the other

users except a desired user in the system and j is the number of positions, which locate in the

codeword of the desired user and are disturbed by chip of the others users, and, from

its definition. i means that there are two or more disturbances in the same chip positions

within w. The hard-limiter can limit the signal amplitude with the result that the disturbed

value at one position is only 1, i.e. after hard-limiter, only when all of the w positions are

disturbed, there is a possible bit error.

C

,, w

D CL

P

CCD TLT ⋅=

jiP ,

)1,1(LC

MarkP Null

wj ≤

j>

When only two users are in the system, i.e. i=1,

⋅

−=

−= q

wLwP

wP

CMark 1

11

10,1 (A1.1)

qw

LwP

wP

CMark ⋅

=

=

111,1 (A1.2)

108

where is the probability of the other user, MarkPCL

w is the probability of the disturbance when

one of the w chips of other user is located in one position within desired codeword, and

is the number of the empty position within the codeword. We define

1w

CMark L

wPq ⋅= , in order

to simplify the expressions.

When another user is added to the system, 2=i , based on the above analysis, there are three

cases: all w positions of the desired codeword are empty, ; only one position is disturbed,

; and two positions are disturbed, . They can be represent by

0,2P

1,2P 2,2P

⋅

−⋅= q

wPP

110,10,2 (A1.3)

⋅

−−+⋅

⋅= q

wPq

wPP

11

11 1,10,11,2 (A1.4)

qw

PP ⋅

−=

11

1,12,2 (A1.5)

where means w-1 empty positions in the desired codeword could be disturbed. Now

there are one or two positions of w disturbed and when , there is no bit error.

−1

1w

3≥w

When one more user is added to the system, 3=i ,

⋅

−⋅= q

wPP

110,20,3 (A1.6)

⋅

−−+⋅

⋅= q

wPq

wPP

11

11 1,20,21,3 (A1.7)

⋅

−−+⋅

−⋅= q

wPq

wPP

12

11

12,21,22,3 (A1.8)

qw

PP ⋅

−=

12

2,23,3 (A1.9)

In general, when i , w<

109

⋅

−⋅= − q

wPP ii 1

10,10, (A1.10)

⋅

−−+⋅

⋅= −− q

wPq

wPP iii 1

11

1 1,10,11, (A1.11)

⋅

−−+⋅

−⋅= −− q

wPq

wPP iii 1

21

11

2,11,12, (A1.12)

... ...

ikqkw

Pqkw

PP kikiki <<

⋅

−−+⋅

−−⋅= −−− 2

11

1)1(

,11,1, (A1.13)

......

qiw

PP iiii ⋅

−−= −− 1

)1(1,1, (A1.14)

When i w=

⋅

−⋅= − q

wPP ww 1

10,10, (A1.15)

⋅

−−+⋅

⋅= −− q

wPq

wPP iww 1

11

1 1,10,11, (A1.16)

... ...

wkqkw

Pqkw

PP kwkwkw <<

⋅

−−+⋅

−−⋅= −−− 1

11

1)1(

,11,1, (A1.17)

......

qPP wwww ⋅

= −− 1

11,1, (A1.18)

wwNull PP ,⋅ is the probability of a bit error with w other users. means that when desired

user transmits Null, bit error emerges.

NullP

When 1−<< Niw

⋅

−⋅= − q

wPP ii 1

10,10, (A1.20)

110

⋅

−−+⋅

⋅= −− q

wPq

wPP iii 1

11

1 1,10,11, (A1.21)

⋅

−−+⋅

−⋅= −− q

wPq

wPP iii 1

21

11

2,11,12, (A1.22)

... ...

wkqkw

Pqkw

PP kikiki <<

⋅

−−+⋅

−−⋅= −−− 2

11

1)1(

,11,1, (A1.23)

......

wiwiwi PqPP ,11,1, 11

−−− +⋅

= (A1.25)

wiNull PP ,⋅ is the probability of the bit error with i+1 other users. Here, because is the

probability for the case of emerging bit error, the new other user doesn’t play any roll in this

case, so it is directly added in .

wiP ,1−

wiP ,

At last when , 1−= Ni

... ...

wNwNwN PqPP ,21,2,1 11

−−−− +⋅

= (A1.26)

wNNull PP ,1−⋅ is the exact BER of the decoder with hard-limiter.

111

Appendix A2 Optical Splitter and Adder

In optical encoder the optical splitter consists of N-1 cross couplers (2×2 ). The input signal is

split to N output ports. As shown in Figure A2.1, when N=4, there are 3 optical couplers.

E1, in

E 2, in E2, out

E1, outα−1

αj

2 x 2

Ei

Eout1

Eout2

Eout3

Eout4

2 x 2

2 x 2

2 x 2

Figure A2.1 The structure of the optical couplers of 2x2 and 2x4

The cross coupler (2×2 coupler) has the following transmission matrix:

⋅

−−

=

in

in

out

out

EE

jj

EE

,2

,1

,2

,1

11

αααα

(A2.1)

where the 0>α is the coupler factor, and are the input electrical fields and

and are the output fields at the ports 1 and 2 of the coupler, respectively.

inE ,1 inE ,2 outE ,1

outE ,2

For the output of the 1:4 optical splitter, only one stimulated signal, , is given, then: iE

ii

out

out

out

out

Ejj

E

jjj

j

EEEE

⋅

−−−

−

=⋅

⋅−⋅

⋅−−⋅−

=

ααααα

α

αααααααα

)1()1(

1

11

11

4

3

2

1

(A2.2)

As a result of the cross coupler performance, the output 1 exhibits a phase difference of ,

the outputs 2 and 3 of 90 , and output 4 of 180 , with respect to input. Normally the X

couplers have a coupler factor

°0

° °

α = 0.5, then:

i

out

out

out

out

Ejj

EEEE

⋅

−

=

5.05.05.05.0

,4

,3

,2

,1

(A2.3)

112

Appendix B Results of the Hardware Experiments

Here are some experimental results of the OCDMA demonstration system.

W

T

Scr

T

Scr

T

Scr

Figu

3 2

1

E

BER Measuring Set

Figure B.1 a) Experimental setup of th

Bit Error Measuring Setandel&GoltormannPF-4

PRBS27-1

Clock155.52MHz

PRBS 223-1

X 1Impulse

Generator

Codeword:(19; 1,2,9)

Alca=15

ElectronicEncoder

E A Mdriver wi tambler

PRBS221-1

Clock155.52MHz

X 2Impulse

Generator

Codeword:(19; 1,5,11)

Alca=15

ElectronicEncoder

Driver DF

ambler

X 3Impulse

Generator

Co(1

OEn

DriverE/O

DFB-LDambler

Alcatel 1915LMI=1550.12nm

Clock155.352MHz

RX

DataRecovery

Codew(19; 1

ElectDeco

re B.1 b) The block diagram of the experiment

TX

TX

e demonstration syste

tel 1915LMM40.35nm

E/Oh EAM

tel 1915LMI49.62nm

E/OB-LD

deword:9; 1,3,6)

pticalcoder

ord:,3,6)

ronicder Amplifiers

al setup of the demon

Receiver

O/

PON

m

OpticalCoupler

(4:1)

E/OConverter

Agilent 11982A

OpticalDecoder

stration system

113

Figure B.2 a) Printed circuit board of the electronic encoder

Figure B.2 b) The waveform of the output of electronic encoder for codeword (19; 1,2,9)

in TX1 (Point A in Figure B.1)

114

Figure B. 3 a) The circuit of the E/O converter using EAM-Laser module

Figure B.3 b) The waveform of the optical impulse from the E/O converter using EAM-

Laser module at TX1 (Point B in Figure B.1)

115

Figure B.4 a) The circuit of the E/O converter using DFB-Laserdiode

Figure B.4 b) The waveform with single persistence of the optical impulse from the E/O

converter using DFB Laserdiode at the TX2 (Point D in Figure B.1)

116

Figure B.4 c) The waveform with infinite persistence trace of the optical impulse of the

E/O converter using DFB Laserdiode at the TX2 (Point D in Figure B.1)

Figure B.5 The waveform of the output of electronic encoder for codeword (19; 1,5,11)

of the TX1 (Point C in Figure B.1)

117

Figure B. 6 a) Printed circuit board of impulse generator

Figure B.6 b) The waveform of the output from the impulse generator at TX2 (Point E

in Figure 5.1)

118

Figure B. 7 a) The hardware of the optical encoder

Figure B.7 b) The waveform of the output from the optical encoder with codeword (19;

1,3,6) at TX3 (Point F in Figure B. 1)

119

Figure B. 8 a) The implemented PON using a 4:1 optical coupler

TX3 TX1 TX2 TX3

overlap

Figure B.8 b) The waveform of combined three encoded signals (Point G in Figure B.1)

120

Figure B.9 The waveform of the output of the electronic amplifiers at the receiver (Point

H in Figure B. 1)

Output of decoder

Data of transmission

Figure B.10 The waveform of point J: the output of the electronic decoder with

codeword (19; 1,3,6)

121

Variabledelay

Signaldriver

FPGASampling array

Clockdriver

Figure B. 11 a) The hardware of the circuit for data recovery (upside)

Delay line for impulse-broadeningMicrostrips forphase delay of

sampling signal

PLL

Figure B. 11 b) The hardware of the circuit for data recovery (downside)

122

Recovered clock

Output of the decoder

Figure B.11 c) The waveform of the recovered clock from the data recovery

Figure B.12 The waveform of the output of the impulse broadening of the data recovery

(Point K in Figure B.1). The jitter comes from the failure impulse introduced by

interference, and it can be dropped by retiming

123

Figure B.13 The output of the optical decoder with desired encoded signal by optical

encoder (the waveform of point L)

The waveform of PRBS

The clock

Figure B. 14 The waveform of the 7 order PRBS of the output of scrambler in TX1

124

PRBS with 21 order

The clock

Figure B. 15 The waveform of the 21 order PRBS of the output of scrambler in TX1

125

Bibliography

[1] J. Speidel, “Zugangsnetze im Wettbewerb,” Hüthig-Verlag, Heidelberg, 2000.

[2] M. Medard and S. Lumetta, “Architectural Issues for Robust Optical Access,” IEEE

Communications Magazine, July 2001.

[3] R. Prasad, et al., “An Overview of CDMA Evolution Towards Wideband CDMA,” IEEE

Communication Surveys, Vol. 1, No. 1, 1998.

[4] D. D. Sampson, G. J. Pendock and R. A. Griffin, “Photonic Code-Division Multiple-

Access Communications,” Fiber and Integrated Optics, Vol. 16, PP. 129-157, 1997.

[5] Technical Progress Report, “Broadband Wireless Standards,” Institute for

Telecommunication Sciences of National Telecommunications and Information

Administration (NTIA) of the U.S., 2000.

[6] P. Prucnal, M. Santoro and T. Fan, “Spread Spectrum Fiber-Optical Local Area Network

Using Optical Processing,” IEEE Journal of Lightwave Technology, Vol. 4, No. 5, 1986.

[7] J. A. Salehi, “Emerging Optical Code-Division Multiple Access Communications

System,” IEEE Network, March 1989.

[8] D. Buchholz, “Broadband Fiber Access: A Fiber-to-the-Customer Access Architecture,”

Bell Labs Technical Journal, January-March 1999.

[9] F. Khaleghi and M. Kavehred, “A Subcarrier Multiplexed CDM Optical Local Area

Network, Theory and Experiment,” IEEE Transactions on Communications, Vol. 43,

No.1, 1995.

[10] J. A. Salehi, A. Weiner and J. P. Heritage, “Coherent Ultrashort Light Pulse CDMA

Communications Systems,” IEEE Journal of Lightwave Technology, Vol. 8, No. 3,

1990.

[11] K. Iversen, D. Hampicke and F. Schleuß, “Optischer CDMA: Systeme und Sequenzen,”

Nachrichtentech., Elektron., SCIENCE, Berlin 45 (1995) 5.

[12] A. J. Viterbi, “CDMA: Principles of Spread Spectrum Communication,” Reading, Mass

Addison-Wesley, ISBN 0-201-63374-4, 1995.

[13] H. Kim and Y. C. Chung, “Passive Optical Network for CDMA-Based Microcellular

Communications Systems,” IEEE Journal of Lightwave Technology, Vol. 19, No. 3,

2001.

126

[14] Robert Fritsch, “3. Ergebnisbericht zum FMS-Projekt III. 3. Arbeitsbereich: Optisches

CDMA als Mehrfachzugriffsverfahren für Rückkanäle in optischen Verteilnetzen im

Teilnehmerzuangsbereich,” Institut für Nachrichtenübertragung, Universität Stuttgart,

Dezember 1999, http://www.inue.uni-stuttgart.de/FMS/abschluss.

[15] Roman Dischler, “1. Ergebnisbericht zum FMS-Projekt III. 3. Arbeitsbereich: Optisches

CDMA als Mehrfachzugriffsverfahren für Rückkanäle in optischen Verteilnetzen im

Teilnehmerzugangsbereich,” Institut für Nachrichtenübertragung, Universität Stuttgart,

31. März 1998, http://www.inue.uni-stuttgart.de/FMS/abschluss.

[16] F. Chung, J. A. Salehi and V. Wie, “Optical Orthogonal Codes: Design, Analysis and

Application,” IEEE Transactions on Information Theory, Vol. 35, No. 3, 1989.

[17] V. Svetislav, Z. Kostic and E. Tiltebaum, “A New Family of Optical Code Sequences

for Use in Spread-Spectrum Fiber-Optical Local Area Networks,” IEEE Transactions on

Communications, Vol. 41, No. 8, 1993.

[18] Z. Wie, et al., “Modified Quadratic Congruence Codes for Fiber Bragg-Grating-Based

Spectral Amplitude Coding Optical CDMA Systems,” IEEE Journal of Lightwave

Technology, Vol. 19, No. 9, 2001.

[19] A. S. Holmes and R. A. Syms, “All-Optical CDMA Using ‘Quasi-Prime’ Codes,” IEEE

Journal of Lightwave Technology, Vol. 10, No. 2, 1992.

[20] S. Tamura and S. Nakano, “Optical Code-Multiplex Transmission by Gold Sequences,”

IEEE Journal of Lightwave Technology, Vol. 3, No. 1, 1985.

[21] L. Tancevski and I. Adonovic, “Block Multiplexing Codes for Optical Ladder Network

Correlators,” IEEE Photonics Technology Letters, Vol. 6, No. 2, 1994.

[22] J. A. Salehi, “Code Division Multiple-Access Techniques in Optical Fiber Networks -

Part I: Fundamental Principles,” IEEE Transactions on Communications, Vol. 37, No. 8,

1989.

[23] J. A. Salehi and C. A. Brackett, “Code Division Multiple Access Techniques in Optical

Fiber Networks - Part 2: Systems Performance Analysis,” IEEE Transactions on

Communications, Vol. 37, No. 8, 1989.

[24] M. Azizoglu, J. A. Salehi and Y. Li, “Optical CDMA via Temporal Codes,” IEEE

Transactions on Communications, Vol. 40, No.7, 1992.

[25] L. Tancevski, et al., “Massive Optical LAN’s Using Wavelength Hopping/Time

Spreading with Increased Security.” IEEE Photonics Technology Letters, Vol. 8, No. 7,

1996.

127

[26] T. W. Mossberg and M. G. Raymer, “Optical Code-Division Multiplexing: The

Intelligent Optical Solution,” Optics&Photonics News, March 2001.

[27] G. Pendock and D. Sampson, “Capacity of Coherence Multiplexed CDMA Networks,”

Conference on Lasers and Electrooptics (CLEO’96), Anaheim, 1996.

[28] T. Ohtsuki, “Performance Analysis of Direct-Detection Optical Asynchronous CDMA

Systems with Double Optical Hard-Limiters”, IEEE Journal of Lightwave Technology,

Vol. 15, No. 3, 1997.

[29] G. Foschini and G. Vannucci, “Using Spread-Spectrum in a High-Capacity Fiber-

Optical Local Network,” IEEE Journal of Lightwave Technology, Vol. 6, No. 3, 1988.

[30] E. Shivaleela, K. Sivarajan and A. Selvarajan, “Design of a New Family of Two-

Dimensional Codes for Fiber-Optic CDMA Networks,” IEEE Journal of Lightwave

Technology, Vol. 16, No. 4, 1998.

[31] M. Kavehrad and D. Zaccarin, “Optical Code-Division-Multiplexed System Based on

Spectral Encoding of Noncoherent Sources,” IEEE Journal of Lightwave Technology,

Vol. 13, No. 3, 1995.

[32] L. Tanceveski and I. Andonovic, “Hybrid Wavelength Hopping/Time Spreading

Schemes for Use in Massive Optical Networks with Increased Security,” IEEE Journal

of Lightwave Technology, Vol. 14, No. 12, 1996.

[33] S. Yegnanarayanan and A. S. Bhushan, “Fast Wavelength-Hopping Time-Spreading

Encoding/Decoding for Optical CDMA,” IEEE Photonics Technology Letters, Vol. 12,

No. 5, 2000.

[34] N. Wada, H. Sotobayashi and K. Kitayama, “2.5Gbit/s Time-Spread/Wavelength-Hop

Optical Code Division Multiplexing Using Fiber Bragg Grating with Supercontinuum

Light Source,” IEE Electronics Letters, Vol. 36, No. 9, 2000.

[35] H. Fathallah, L. A. Rusch, et al., “Passive Optical Fast Frequency-Hop CDMA

Communications System,” IEEE Journal of Lightwave Technology, Vol. 17, No. 3,

1999.

[36] H. B. Jaafar, S. LaRochelle and P. Y. Cortes, “1.25Gbit/s Transmission of Optical FFH-

OCDMA Signals over 80Km with 16 Users,” Optical Fiber Communications

Conference and Exhibition, 2001.

[37] R. A. Griffin, D. D. Sampson and D. A. Jackson, “Demonstration of Data Transmission

Using Coherent Correlation to Reconstruct a Coded Pulse Sequence,” IEEE, Photonics

Technology Letters, Vol. 4, No. 5, 1992.

128

[38] Y. L. Chang and M. E. Marhic, “Fiber-Optic Ladder Networks for Inverse Decoding

Coherent CDMA,” IEEE Journal of Lightwave Technology, Vol. 10, No. 12, 1992.

[39] H. P. Sardesai, C.-C. Chang and A. M. Weiner, “A Femtosecond Code-Division

Multiple-Access Communications System Test Bed,” IEEE Journal of Lightwave

Technology, Vol. 16, No. 11, 1998.

[40] P. C. Teh, M. Ibsen, et al., “Demonstration of a Four-Channel WDM/OCDMA System

Using 255-Chip 320-Gchip/s Quarternary Phase Coding Gratings,” IEEE, Photonics

Technology Letters, Vol. 14, No. 2, 2002.

[41] P. C. Teh, M. Ibsen, et al., “A 16-Channel OCDMA System (4 OCDMA x 4 WDM)

Based on 16 Chip, 20Gchip/s Superstructure Fiber Bragg Gratings and DFB Fiber Laser

Transmitters” OCIS codes: 060.4250, 2002.

[42] N. Wada and W. Chujo, “160Gbit/s OCDMA/WDM (10Gbit/s x 4CDMA x 4WDM)

Transmission with a Novel Side-Lobes Suppression Detection,” Optical Fibre

Communication Conference, Anaheim, 2001.

[43] H. Sotobayashi and W. Chujo, “1.6Bit/s/Hz, 6.4Tbit/s OCDM/WDM (4 OCDM x 40

WDM x 40 Gbit/s) Transmission Experiment,” ECOC, 2001.

[44] T. Pfeiffer, J. Kissing, et al., “Coarse WDM/CDM/TDM Concept for Optical Packet

Transmission in Metropolitan and Access Networks Supporting 400 Channels at

2.5Gbit/s Peak Rate,” IEEE Journal of Lightwave Technology, Vol. 18, No. 12, 2000.

[45] T. Pfeiffer, B. Deppisch, et al., “Optical CDMA Transmission for Robust Realization of

Complex and Flexible Multiple Access Networks,” Optical Fibre Communication

Conference, Paper WM-51, 1999.

[46] T. Pfeiffer, B. Deppisch, et al., “High Speed Optical Network For Asynchronous Multi-

user Access Applying Periodic Spectral Coding of Broadband Sources,” IEEE

Electronics Letters, Vol. 33, No. 25, 1997.

[47] T. Pfeiffer, B. Deppisch, et al., “Temperature in Sensitive Spectrally Encoded Optical

Multichannel System for Simultaneous Transmission of Digital Baseband and Subcarrier

signals,” IEEE Electronics Letters, Vol. 36, No. 5, 2000.

[48] J. A. Salehi and C. A. Brackett, “Fundamental Principles of Fiber Optics Code Division

Multiple Access (FO-CDMA),” IEEE International Conference On Communications,

June 1987.

[49] G. J. Pendock and D. D. Sampson, “Increasing the Transmission Capacity of Coherence

Multiplexed Communication Systems by Using Differential Detection,” IEEE Photonics

Technology Letters, Vol. 7, No. 12, 1995.

129

[50] G. Vannucci and K. Kikuchi. “Experimental Spreading and Despreading of the Optical

Spectrum,” IEEE Transactions on Communications, Vol. 37, No. 7, 1989.

[51] N. Wada and K. Kitayama, “Fundamental Properties of 10Gbit/s Coherent Optical Code

Division Multiplexing: Experiment,” Optical Review (Japan), Vol. 5, No. 1, 1998.

[52] H. Sotobayashi, W. Chujo and K. Kitayama, “3 x 10Gbit/s OCDM Transmission at

1550nm Band over 150km Standard Fiber Using Midspan Optical Phase Conjugation,”

OECC2000, 14A2-4, pp. 456-457, 2000.

[53] M. Razavi and J. A. Salehi, “Statistical Analysis of Fiber-Optical CDMA

Communication Systems – Part I: Device Modeling,” IEEE Journal of Lightwave

Technology, Vol. 20, No. 8, 2002.

[54] M. Razavi and J. A. Salehi, “Statistical Analysis of Fiber-optical CDMA

Communication Systems – Part II: Incorporating multiple optical amplifiers,” IEEE

Journal of Lightwave Technology, Vol. 20, No. 8, 2002.

[55] PTDS product information, www.virtualphotionics.com.

[56] F. Forghieri, R. W. Tkach and A. R. Chraplyvy, “Fibre Nonlinearities and Their Impact

on Transmission Systems: In Optical Fibre Telecommunications,” Vol. IIIa, Academic

Press, 1997.

[57] S. Shen and A. M. Weiner, “Suppression of WDM Interference for Error-Free Detection

of Ultrashort-Pulse CDMA Signals in Spectrally Overlaid Hybrid WDM-CDMA

Operation,” IEEE Photonics Technology Letters, Vol. 13, No. 1, 2001.

[58] J. Chen and G. Yang, “CDMA Fiber-Optic Systems with Optical Hard Limiters,” IEEE

Journal of Lightwave Technology, Vol. 19, No. 7, 2001.

[59] M. Pauer, P. J. Winzer and W. R. Leeb, “Bit Error Probability Reduction in Direct

Detection Optical Receivers Using RZ Coding,” IEEE Journal of Lightwave

Technology, Vol. 19, No. 9, 2001.

[60] P. Winzer and A. Kalmar, “Sensitivity Enhancement of Optical Receivers by Impulsive

Coding,” IEEE Journal of Lightwave Technology, Vol. 17, No. 2, 1999.

[61] S. Kim, J. Lee and J. Jeong, “Transmission Performance of 10-Gbit/s Optical Duobinary

Transmission Systems Considering Adjustable Chirp of Nonideal LiNbO3 Mach-

Zehnder Modulators Due to Applied Voltage Ratio and Filter Bandwidth,” IEEE Journal

of Lightwave Technology, Vol. 19, No. 4, 2001.

[62] H. Sunnerud, “Polarization-Mode Dispersion in Optical Fibers: Characterization,

Transmission Impairments and Compensation,” Technical Report No. 399, Photonics

Laboratory, Chalmers University of Technology, Göteborg, March 2001.

130

[63] J. Hansryd, H. Sunnerud, et al., “Impact of PMD on Four-Wave-Mixing-Induced

Crosstalk in WDM Systems,” IEEE Photonics Technology Letters Vol. 12, No. 9, 2000.

[64] K. Petermann, “Laser Diode Modulation and Noise,” New York, Kluwer-Academic,

1993.

[65] P. J. Legg, M. Tur and I. Andonovic, “Solution Paths to Limit Interferometric Noise

Induced Performance Degradation in ASK/Direct Detection Lightwave Networks,”

IEEE Journal of Lightwave Technology, Vol. 14, No. 9, 1996.

[66] A. C. Kersey, A. Dandridge and A. B. Tevten, “Dependence of Visibility on Input

Polarization in Interferometric Fiber-Optic Sensors,” Optical Letters, Vol. 13, pp. 288-

290, 1988.

[67] I. H. Malitson, “Interspecimen Comparison of the Refractive Index of Fused Silica,”

Journal of Optical Society American, Vol. 55, pp. 1205-1209, 1965.

[68] M. R. Dale and R. M. Gagliardi, “Channel Coding for Asynchronous Fiberoptic CDMA

Communications,” IEEE Transactions on Communications, Vol. 43, No. 9, 1995.

[69] J. Zhang, et al., “Interferometric Noise in Optical Time Division Multiplexing

Transmission System,” IEEE Journal of Lightwave Technology, Vol. 20, No. 8, 2002.

[70] G. Clark and J. Cain, “Error-Correction Coding for Digital Communications,” New

York, Plenuim, March 1988.

[71] VPI transmission Maker™- VPI component Maker™, “Photonic Modules Reference

Manual.”

[72] R. H. Wentworth, “Theoretical Noise Performance of Coherence-Multiplexed

Interferometric Sensors,” IEEE Journal of Lightwave Technology, Vol. 7, No. 6, 1989.

[73] R. W. Tkach and A. R. Chraplyvy, “Phase Noise and Linewidth in an InGaAsP DFB

Laser,” IEEE Journal of Lightwave Technology, Vol. 4, No. 11, 1986.

[74] M. Gustavsson, L. Gillner and C. Larsen, “Statistical Analysis of Interferometric

Crosstalk: Theory and Optical Network Examples,” IEEE Journal of Lightwave

Technology, Vol. 15, No. 11, 1997.

[75] J. Müchenheim, “Optischer Codevielfachzugriff in lokalen Netzen,” Dissertation an der

Fakultät für Elektrotechnik und Informationstechnik der Technischen Universität

Ilmenau, 1998.

[76] B. E. Briley, “An Introduction to Fiber Optics System Design,” Amsterdam, North-

Holland, ISBN 0-444-70498-1, 1988.

[77] N. Grote, “Fiber Optic Communication Devices,” New York, Springer, ISBN 3-540-

66977-9P, 2001.

131

[78] P. Pepeljugoski and K. Lau, “Interferometric Noise Reduction on Fiber-Optical Links by

Superposition of High Frequency Modulation,” IEEE Journal of Lightwave Technology,

Vol. 10, No. 7, 1992.

[79] K. Jepsen, et al., “Investigation of Cascadability of Add-Drop Multiplexers in ITDM

Systems,” ECOC, pp. 619-620, 1998.

[80] L. Tancevski and L. Rusch, “Impact of the Beat Noise on the Performance of 2-D

Optical CDMA Systems,” IEEE Communication Letters, Vol. 4, No. 8, 2000.

[81] K. Yehia, “Treiberschaltung für ein Lasermodul bei 1550 nm” 1. Semesterarbeit am

Institut für Nachrichtenübertragung, Universität Stuttgart.

[82] B. E. Briley, “An Introduction to Fiber Optics System Design”, Amsterdam, North-

Holland, ISBN 0-444-70498-1, 1988.

[83] A. Mendez, et al., “Strategies for Realizing Optical CDMA for Dense, High Speed,

Long Span, Optical Network Applications,” IEEE Journal of Lightwave Technology,

Vol. 18, Dec. 12, 2000.

[84] J. Elbers and C. Glingener, “Efficient Design of High-Capacity Dense Wavelength-

Division Multiplex Systems,” International Journal of Electronics and Communications,

(AEÜ) 55, No. 5, 2001.

[85] K. Fukuchi, M. Kasanatsu, et al., “10.92-Tb/s Triple-Band/Ultra-Dense WDM Optical-

Repeatered Transmission Experiment,” Optical Fibre Communication Conference,

PD24, Anaheim, 2001.

[86] D. Chen, et al., “3.2Tb/s Field Trial (80 x 40Gb/s) over 3 x 82km SSMF Using FEC,

Raman and Tunable Dispersion Compensation.” Optical Fibre Communication

Conference, Anaheim, 2001.

[87] Y. Wong, D. Muehlner, et al., “Technology Development of a High-Density 32-Channel

16Gb/s Optical Data Link for Optical Interconnection Applications for the

Optoelectronic Technology Consortium(OETC),” IEEE Journal of Lightwave

Technology, Vol. 13, No. 6, 1995.

[88] S. V. Kartalopoulos, “Introduction to DWDM Technology: Data in a Rainbow,” IEEE

Press, ISDN 0-7803-5399-4, 1999.

[89] D. Cotter, J. K. Lucek, M. Shabeer, K. Smith, D. C. Rogers, D. Nesset and P. Gunning,

“Self-Routing of 100Gbits/s Packets Using 6 Bit Keyword Address Recognition,” IEEE

Electronics Letters, Vol. 31, No. 17, 1995.

132

[90] A. Ishwar and S. Bhardwaj, “All Optical Logic Circuits Based on the Polarization

Properties of Non-Degenerate Four-Wave Mixing,” California Institute of Technology

Pasadena, California, 2001.

[91] Y. Chonbodeechalermroong and P. L. Chu, “Simple Star Multihop Optical Network,”

IEEE Journal of Lightwave Technology, Vol. 19, No. 4, 2001.

[92] A. A. Sawchuk and T. C. Strand, “Digital Optical Computing,” IEEE, Special Issue on

Optical Computing, pp. 758-779, July 1984.

[93] H. Sotobayashi, et al., “1.6-b/s/Hz 6.4-Tb/s QPSK-OCDM/WDM (4 OCDMA x 40

WDM x 40Gb/s) Transmission Experiment Using Optical Hard Thresholding,” IEEE

Photonics Technology Letters, Vol. 14, No. 4, 2002.

[94] J. Kim and H. V. Poor, “Turbo-Coded Optical Direct Detection CDMA System with

PPM Modulation,” IEEE Journal of Lightwave Technology, Vol. 19, No. 3, 2001.

[95] P. Legg, et al., “Solution Paths to Limit Interferometric Noise Induced Performance

Degradation in ASK/Direct Detection Lightwave Networks,” IEEE Journal of

Lightwave Technology, Vol. 14, No. 9, 1996.

[96] B. Moslehi, “Noise Power Spectra of Optical Two-Beam Interferometers Induced by the

Laser Phase Noise,” IEEE Journal of Lightwave Technology, Vol. LT-4, No. 11, 1986.

[97] Zhigong Wang, “Monolithisch-integrierte Silizium-Bipolar-Schaltungen zur

Taktrückgewinnung bei Datenraten von mehreren Gbit/s,” Dissertation Heft Nr. 902/1,

Universität Bochum, 1990.