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
133
Embed
Design and Implementation of an Optical Code ... - CiteSeerX
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
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
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
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
4.2.2 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
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
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
yjzhang
These coding scheme is used to introduce randomness (and redundancy) into the digital information stream to ensure efficient timing recovery (and to facilitate error monitoring) at the receiver.
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.