JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO.

6, JUNE 2000 765

Coherent Optical CDMA (OCDMA) Systems Used

for High-Capacity Optical Fiber Networks-System
Description, OTDMA Comparison, and
OCDMA/WDMA Networking
Wei Huang, Member, IEEE, Mohamed H. M. Nizam, Member, IEEE,
Ivan Andonovic, Senior Member, IEEE, Member, OSA, and Moshe Tur, Fellow, IEEE, Fellow, OSA

Abstract—As the wavelength resource in mainstream wave- in traffic levels, primarily attributed to the exponential use of
length-division multiple-access (WDMA) systems becomes the Internet together with an increase in the number and range
exhausted, and the bit-rate limitation within a single wavelength of new services. All-optical fiber networking is considered to
bandwidth is reached, alternative approaches to implementing a
high-capacity optical fiber network need to be investigated. Co- be the central solution to overcoming the perceived limitations
herent optical code-division multiple-access (OCDMA) systems, of a totally electronic evolution. At the core of the development
that can access many users simultaneously and asynchronously is the huge inherent bandwidth of single mode optical fiber,
(or synchronously) across the single wavelength and same timeslot exhibiting a low-loss spectrum of up to 25 000 GHz [4], [5].
via spread spectrum techniques, are one alternative. In the
longer term, the advantages of OCDMA in tandem with WDMA Since the speed at which electrical signal can modulate
(OCDMA/WDMA) networks are compelling and worthy of optical carriers is limited, to fully exploit the optical fiber
further investigation in the goal of realising an extensive, flexible, bandwidth, optical multiplexing techniques have to be de-
high throughput and easily managed optical telecommunication ployed. There are three multiplexing alternatives: wavelength
infrastructure. In this paper, coherent OCDMA systems are division multiple access (WDMA) [6]–[8], optical time division
introduced, and the issues of the system implementation within
high-capacity optical fiber networks are discussed. A performance multiple access (OTDMA) [9]–[11] and optical code division
comparison between OCDMA and OTDMA systems is then multiple access (OCDMA) [12]–[16]. The term “division
carried out, both of them using narrow pulse laser sources. An multiple access” is used throughout, rather than “division
optical fiber network utilizing coherent OCDMA techniques as multiplexing,” to define not simply the transmission multiplex
one layer of a multiplexing hierarchy, in tandem with WDMA, but in addition indicates the access, routing and switching func-
is illustrated and a possible hybrid OCDMA/WDMA network
architecture (and its performances and advantages) is described. tionality. Whilst WDMA and OTDMA techniques partition the
available spectrum and time to different users, respectively,
Index Terms—Optical code division multiple access (OCDMA), OCDMA techniques multiplex users simultaneously and
optical fiber communications and networks, optical time division
multiple access (OTDMA), hybrid optical fiber OCDMA/WDMA asynchronously (or synchronously) across the same spectrum
networks, wavelength division multiple access (WDMA). and timeslot through a unique code.
The bulk of recent research activity in optical fiber net-
working has focused on WDMA networks, presently enjoying
I. INTRODUCTION
rapid deployment worldwide as a means of increasing capacity

A T PRESENT, communications networks based on the syn-

chronous digital hierarchy/synchronous optical network
(SDH/SONET) in tandem with asynchronous transfer mode
at minimal cost within the existing optical fiber infrastructure.
Although WDMA can be used as a degree of design freedom
with respect to routing and wavelength selection adding a
(ATM) switching or internet protocol (IP) routing are in place further dimension to network functionality, the fundamental
to satisfy the demand for a range of telecommunication services issue is the limited number of useable wavelengths [17]. This
[1]–[3]. However, the current infrastructure even taking into limitation is due to the nonlinearity introduced within the
account further upgrades, will be unable to supply the higher transmission optical fiber that induces crosstalk when transmit-
transmission capacity required to support the projected growth ting different wavelength signals simultaneously, reducing the
flexibility and capacity of the network. Several demonstrations
of dense WDMA [8] have been reported recently:
Manuscript received December 6, 1999. This work is supported by the
EPSRC and in part carried out at the Tel-Aviv University, Israel, November • wavelengths Gb/s signal transmitted over 8 514 km
1998.
W. Huang, M. H. M. Nizam, and I. Andonovic is with the Department of fiber using an equalization technique via Fourier synthesis
Electrical and Electronic Engineering, University of Strathclyde, Glasgow G1 of the gain characteristic of the system [18];
1XW, U.K. (e-mail: w.huang@eee.strath.ac.uk). • 50 GHz spaced, wavelengths Gb/s signal trans-
M. Tur is with the Faculty of Engineering, Tel Aviv University, Tel Aviv
69978, Israel. mitted over 640 km dispersion-shifted fiber with multi-
Publisher Item Identifier S 0733-8724(00)05083-0. wavelength distributed Raman amplification [19];
0733–8724/00$10.00 © 2000 IEEE
766 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

• wavelengths Gb/s, 1 Tb/s throughput experiment the best trade-off regarding capacity, performance, flexibility
demonstrated over 342 km in nonzero dispersion fiber and complexity of a variety of approaches to network provision.
[20]. An alternative to providing additional multiplexing at a single
However, upgrading these systems through the introduction of wavelength is through CDMA techniques. Direct-sequence (DS)
additional wavelengths will become more difficult due to the CDMA is a technique originating from spread spectrum (SS)
limitations governed by device characteristics. With respect to communications [30]. DS-CDMA techniques play an important
channel switching, wavelength routing is the next switching role in current wireless communication systems such as cellular
dimension for dense WDMA networks, with interferometric [31], microcellular [32], indoor [33] and satellite communica-
crosstalk being an essential issue in the implementation of tions [34]. Efficient multiple access capacity, effective reuse
cross-connects based on space and wavelength. This kind of of the frequency resource, asynchronous access capacity and
crosstalk, increasing with the number of ports and wavelengths, immunity to the multipath fading characteristic of wireless
cannot be filtered at the receiver since it predominately occurs transmission are important characteristics driving the use of this
at the same wavelength as the signal [21]. Hence, the extent of approach. DS-CDMA is also the likeliest candidate for the next
the wavelength routing that is realisable which in turn deter- generation of broad-band mobile communications [35].
mines switch size and flexibility, places fundamental limits on CDMA techniques can be implemented directly in the op-
network flexibility [22], [23]. Both practical and fundamental tical domain based on continuous-wave (CW) [12], [13] or ultra-
limits will restrict the number of wavelengths, depending on short optical pulse laser sources [14]–[16]. Coherent OCDMA
the specific application [24]. techniques governed by similar principals to radio DS-CDMA,
can multiplex multiple users onto the huge optical spectrum
Even when considering the partitioning of the huge band-
via spread spectrum techniques, offering the same transmis-
width of optical fiber into around 100 sub-bands, additional
sion capacity as OTDMA systems (via interference cancella-
layers of multiplexing are required to fully exploit each subband
tion in an asynchronous transmission mode [36]). The tech-
(at a bandwidth of approximately 250 GHz). It is clear that cur-
niques can be used as an additional optical layer of a multi-
rently, due to a more straightforward implementation, WDMA
plexing hierarchy based on WDMA and allow electrical user
techniques that operate at a small number of wavelengths with
signals access to the optical network in a more natural way
each sub-band supporting data rates defined by the modulation
[37], [38]. The asynchronous operation of OCDMA systems
and detection speed limitation, is the preferred option. However
is a very desirable characteristic of any multiplexing technique
as the wavelength resource becomes exhausted, a further stage
especially at high data rates, and the accurate time of arrival
of system upgrade will be required.
measurement capability relaxes the problem of synchronization
Electrical TDMA is utilized to multiplex data from inde-
inherent in OTDMA implementations [39], [40]. OTDMA and
pendent WDMA users into single wavelength channels before
OCDMA systems both utilize narrow pulsed laser sources, but
then transmission across the network [25], [26]. To overcome
the latter has the advantage of de/multiplexing and detection
the bit-rate limitation imposed by electronics within a single
relying on commercial components rather than new generation
wavelength bandwidth, OTDMA techniques using narrow
devices required by OTDMA [9]–[11], [16]. Hence, the attrac-
optical pulse laser sources have been developed [9]–[11]. Since
tion of the approach is compelling despite the added imple-
the approach uses extremely narrow optical timeslots with
mentation complexity brought about at the receiver due to need
correspondingly large bandwidths, a natural increase in the
for carrier phase locking and the phase stability requirements
data rates (40 Gb/s–100 Gb/s or higher) at a single wavelength
(the linewidth) of the laser sources [41]. A network employing
bandwidth, is obtained by multiplexing low bit rate tributary
OCDMA/WDMA is capable of establishing more traffic con-
data streams directly in the optical domain. To further increase
nections than a network using WDMA only. These results also
the capacity, a combination of OTDMA with WDMA have
suggest that OCDMA/WDMA reduces connection setup failure
been developed as follows:
in any network condition and hence, reduces connection setup
• wavelengths Gb/s (1.4 Tb/s) transmitted through time as well as node buffering requirements [42]. In the longer
50 km of dispersion shifted fiber [27]; term, OCDMA in tandem with WDMA is thus worthy of further
• wavelengths Gb/s (1 Tb/s) transmitted over 342 investigation in the goal of realising an extensive, high capacity
km [28]; and easily managed optical telecommunication infrastructure.
• wavelengths Gb/s (3 Tb/s) transmitted over 40 km In the following sections, coherent OCDMA systems
of dispersion shifted fiber using supercontinuum WDMA are introduced. A performance comparison between
sources [29]. OCDMA-OTDMA is then carried out. An optical fiber
However, the technology involved in the implementation of network utilising the OCDMA technique as one layer of the
OTDMA is very different from that of WDMA; whereas the multiplexing hierarchy, in tandem with WDMA, is developed,
latter systems use components that are available commercially, and a possible hybrid OCDMA/WDMA network architecture
most of the devices required to implement OTDMA are still (and its performances and advantages) is described.
confined to the laboratory. In addition, OTDMA has still a
number of difficult hurdles to solve such as synchronization
II. OCDMA SYSTEM DESCRIPTION
and the immature technology of selecting one channel out of
multiplexed picosecond optical pulse data streams. Hence, scope In the optical domain, CDMA implementations can use both
for evolving new innovative methods exists in order to determine CW and narrow pulse laser sources. In this section, OCDMA
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 767

Fig. 2. Incoherent OCDMA systems using pulsed laser sources with passive
matched filter detection.
Fig. 1. Coherent OCDMA systems using CW laser sources with full coherent
detection.

2) OCDMA Using Narrow Pulse Laser Sources: Both inco-

systems with particular application in high-capacity hybrid herent and coherent OCDMA systems using narrow pulse laser
OCDMA/WDMA network implementations are illustrated and sources have been proposed, and most of the research activity
the key application issues are investigated. has centred on incoherent implementations, driving initially the
development of unipolar pseudo-orthogonal (0, 1) codes [14],
[15]. For the purposes of comparison, both incoherent and co-
A. Generalized System herent implementations are introduced.
1) OCDMA Using CW Laser Sources: OCDMA system 3) Incoherent Implementation: The procedure for an
using CW laser sources (Fig. 1), similar to radio DS CDMA approach using unipolar codes and incoherent passive
principals, can be based on coherent implementation [12], [13]. matched-filtered detection (MFD) directly in the optical
As shown in Fig. 1(a), the data bits at the transmitter are spread domain is shown in Fig. 2(a) [14]. Differing from radio DS
by high-speed electrical bipolar sequences that then CDMA principles, the system relies on a simple, intensity
modulate a CW laser source via an electrooptic modulator based, pulse time addressing process. A pulse laser source is
(EOM) for transmission. The transmitted optical signals for intensity-modulated by electrical (0, 1) data bits and encoded
different users are multiplexed on one channel of an optical by an optical delay-line encoder [see Fig. 2(b)] to produce an
fiber network through the assignment of unique codes. unipolar signature code.
At the receiver, the received optical signal beats with a At the receiver, an optical MFD (an optical delay-line de-
CW local oscillator via a 3-dB coupler and a dual balanced coder matched to the encoder at the transmitter) produces a
detector, effectively performing the function of a mixer. The peak in the correlation output for the intended user. Data bits
laser frequency of the CW local oscillator is locked to the are discriminated in the chip duration using a photodiode fol-
laser of the transmitter via an external absolute reference. The lowed by a thresholding process. An example that illustrates
electrical output signal from the dual balanced detector then the correlation process is given in Fig. 2(b) in which the de-
undergoes a radio frequency (RF) correlation and detection sired code (01011) and the nondesired code (01100) are split
process [Fig. 1(b)]. The signal is correlated with the electrical to the different branches of the optical decoder matched to the
code sequence of the desired user, integrated and discriminated desired code. On the branches the signals go through different
to recover the data bit. delay lines, and are then re-combined. At the output of the op-
An example of the correlation detection process is given in tical decoder the desired correlation and the interference signals
Fig. 1(b) in which the desired code and the nondesired (inter- are (011131110) and (012111000), respectively. The following
ference) code are (-1-11-11) and (-111-1-1), respectively. The detection process has to determine an autocorrelation peak of
incoming codes are correlated to the local code (-1-11-11), chip 3 in the presence of an interference chip of level 1 and other
by chip, through a multiplier, and the resulting correlation se- side-lobe interference.
quence is (11111) for the desired signal and (1-111-1) for the Compared to conventional electronic bipolar
interference signal. Both are integrated, producing an autocor- codes such as Gold codes [43], the cross-correlation function
relation peak 5 and a cross-correlation interference of 1, respec- of unipolar codes is high and the number of codes in the
tively. The integrator output is then used to recover the data bit family is very low. Thus, long, sparse codes comprising very
via an estimation process. Since CDMA systems have to sup- few ones and narrow pulses have to be employed to support
port many users, the codes used should satisfy specific correla- a large number of users and higher transmission capacities,
tion conditions: both the time-shifted autocorrelation function respectively. The implementation is thus limited by the losses
of each code and the cross-correlation function between codes of the complex hardware and the requirement for chip duration
must be low when compared to the autocorrelation peak value. detection, preventing the use of ultrashort pulses.
768 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

Fig. 3. Coherent OCDMA systems using pulse laser sources with full coherent
detection.

4) Coherent Implementation: Coherent OCDMA systems

using narrow optical pulse lasers adopt a similar principal as
radio DS CDMA [Fig. 3(a)] [16]. A pulsed laser emits a narrow
optical pulse sequence at an interval of the data bit duration. Fig. 4. Data streams in asynchronous CDMA users.
The sequence is then phase-modulated by electrical data bits
via an EOM and encoded by an optical tapped delay line
encoder with different delays and predetermined phase shifts herent OCDMA systems using full coherent detection since syn-
on each branch, producing an optical bipolar code [Fig. 3(b)]. chronism has to be maintained between the incoming and the
At the receiver, a pulsed LO in tandem with the same optical local codes. Code synchronization can be realized through a
encoder as that of the user transmitter is used to generate the two-stage process: a coarse alignment referred to as code acqui-
code of the intended user. Through proper synchronization of sition [39] and a subsequent fine alignment referred to as code
the code [39], [40] and optical carrier phase [41], the local code tracking [40].
is multiplied by the received signal, chip by chip via a 3 dB cou- 1) Code Acquisition: As shown in Fig. 4(a), the difference
pler followed by a balanced detector (optic-electromultiplier). in time between the desired code (first user) and the local code
The correlation process is similar to that introduced in Fig. 1(b). is smaller than one data bit duration . The code acquisi-
The output of the balanced detector is an electrical signal, com- tion technique relies on a variable optical delay device config-
prising the autocorrelation and cross-correlation components, ured in the optical path of the local code, before the correla-
subsequently integrated over a bit duration and discriminated tion process. Before transmitting data, a training sequence (un-
to recover the data bit. modulated code sequence) is sent to the receiver to aid in the
The issues pertinent to the implementation of the coherent coarse acquisition of the incoming code to within a fraction
OCDMA systems within high-capacity optical networks are dis- (usually one-half or one-fourth) of the chip duration. The ac-
cussed in the following subsections. quisition process is progressed through the adjustment of the
optical delay, taking correlations between the two codes sequen-
B. Near-Far Issue tially until the code phase difference is compensated by the op-
tical delay [Fig. 4(b)], thereby locating the peak of the autocor-
The “near-far” problem is an essential issue in most CDMA relation function.
systems, especially in wireless applications. Fortunately, the The performance of a noncoherent code acquisition process
transceivers in an optical fiber networks are fixed, and to a in the presence of multiuser interference and optical device
certain extent, the optical path loss between the transmitter noise has been investigated in [39]. The acquisition is per-
and the receiver can be predicted. Hence, a gain-clamped formed in a noncoherent manner, the simple reason being
preamplifier [Figs. 1(a) and 3(a)] can be used to compensate that the optical carrier phase is still unknown. Once the codes
for optical power loss, so that all the transmitted optical signals are acquired, the process is stopped and the system moves to
originating from different locations can be received at similar transmitting data.
optical power levels. 2) Code Tracking: The receiver configuration for full co-
herent OCDMA detection [the same concept can be used for
C. Code Synchronization the system in Fig. 1(a)] is depicted in Fig. 5. The receiver exe-
Code synchronization is of major importance for efficient op- cutes three processes: a noncoherent delay-locked code process,
eration of any CDMA system. It is even more crucial in co- a decision-directed phase locking process and a demodulation
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 769

Fig. 6. Multistage interference cancellation (IC) for coherent OCDMA

systems.

linewidth to bit-rate ratio is about at a phase locking error

Fig. 5. Receiver configuration with full coherent reception for use in coherent
OCDMA systems.
variance of 0.03 and about at 0.01 [41].

E. Polarization Control
process. The key to simplifying the configuration is to effec- Polarization control is necessary in coherent systems to align
tively combine the phase locked loop and the demodulator with the polarization between the received and local optical signals.
the code-tracking loop. Fine polarization control ensures and maintains the maximum
Code tracking maintains the incoming and local codes in photocurrent produced by the correlation. In the extreme case,
synchronism for the duration of the transmission. Noncoherent no photocurrent is produced at the output of the balanced de-
operation is also used for the code tracking because, at low tector if the polarization directions of the two signals are or-
signal-to-noise ratios (SNR’s), it is difficult to lock the optical thogonal. Alignment of the polarization may be realized based
carrier phase prior to code tracking. Noncoherent code tracking on polarization control techniques such as installing special po-
thus offers an important advantage since the tracking perfor- larization-maintaining fibers, polarization tracking, polarization
mance is not directly influenced by the carrier phase locking scrambling, polarization shift keying and polarization diversity
process. [44]. Many demonstrations of coherent optical communication
In the tracking process, the local code and its one code systems have managed the issue of polarization [45], [46].
chip duration delayed version of the in-phase and the
quadrature-phase components are generated by a optical F. Multiuser Interference
code phase delay device and optical carrier phase shifter. Although the coherent OCDMA systems in Figs. 1(a) and
The local signals are correlated with the received optical signal 3(a) using bipolar orthogonal Walsh codes [16] in synchronous
by a 3-dB coupler and a dual balanced detector. The resultant transmission have no cochannel interference. Multiuser
correlation output is used to drive a loop filter and voltage interference is a major issue causing severe performance degra-
controlled clock (VCC). Since the transceivers within OCDMA dation in asynchronous transmission due to the nonorthogonal
systems are fixed, the code delay of all users is arbitrary but properties of the spreading codes. Whilst the poor bit error
constant within the duration of the transmission. Furthermore, rate (BER) ( – ) of current radio DS-CDMA systems
the code tracking performance can be improved by increasing is acceptable for voice transmission in current wireless com-
the optical power and reducing the loop bandwidth [40]. munications [31], this BER performance is unacceptable in
an optical fiber network (where a generally accepted BER is
D. Carrier Phase Synchronization in excess of ). However, this fundamental interference
Carrier phase synchronization is a prerequisite for successful is not completely random and is a function of carrier phase,
demodulation. The phase locking process (Fig. 5) is based on access delay and data bits of nondesired users. If the random
a digital decision-directed loop. In this loop, the data recovered ingredients of all users are estimated by the use of multiuser
by the demodulator is used to multiply a one-bit delayed output and multistage detection, multiuser interference can be rebuilt
of the integrator, subsequently of use for the phase locking op- and removed from the desired signal [47], [48].
eration, removing the information data component. The dis- The interference cancellation approach proposed for coherent
criminator output is then fed into a loop filter, followed by a OCDMA systems of Fig. 3(a) is shown in Fig. 6 [the principal
VCC, used to drive a continuously variable optical carrier phase is also suitable for the system in Fig. 1(a)]. Since full coherent
modulator in order to align the carrier phase. The factors lim- reception is used, the carrier phase and the code access delay
iting the performance of the phase locking are multiuser inter- can be derived from the code tracking and carrier phase locking
ference, phase noise and optical shot noise [41]. The required processes. Data symbols are known from the pre-stage detec-
locking performance has to be obtained through selection of a tion. The interference can be rebuilt via the estimated ingredi-
pulsed laser source with specified linewidth. The required laser ents and removed from the desired threshold decision signal.
770 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

The threshold decision is then repeated. The key advantage de-

rived from the proposed system implementation, is that the can-
cellation processes can be performed in the electrical domain.
It is assumed that the first user is desired by the receiver, and
the other users are nondesired users. In Fig. 6,
the signal from integrator prior to the threshold decision for the
first user can be represented as

(1)

where “0” indicates decisions without multiuser interference

cancellation. is the data bit of the first user [Fig. 4(b)].
is the data bit duration, related to the code chip duration as
where is the code period. is the optical-electro
conversion factor, and is the optical shot noise. The second
and third terms in (1) are multiuser interference, represented by

(2a) Fig. 7. BER as a function of received optical power with interference

cancellation.

Using and , the threshold decision signal for the second

(2b) stage cancellation is now given by

where and are the continuous time partial (5)

cross-correlation functions defined in [49]. is the carrier
phase, is the access delay and are data where
symbols [illustrated in Fig. 4(b)].
1) First-Stage Cancellation: It can be seen from (1) that the (6a)
desired data bit is interfered by two bits ( and )
from the users in the channel. As illustrated
in Fig. 4(b), is estimated at point 1 as at the receiver
for the second user; the code timing and the carrier phase
(6b)
are known from the code tracking and carrier phase locking,
respectively. Hence, the interference due to with duration
can be removed from the decision making process of user Fig. 7 illustrates the BER as a function of the received op-
one. A similar process is applied to all other nondesired users tical power for 60 users where two-stage interference cancel-
at point 2 where the first stage cancellation is carried out. The lation has been employed [36]. One stage cancellation partly
threshold decision signal for the first stage cancellation is cancels multiuser interference; significantly improved system
performance is obtained through the two-stage cancellation. It
(3) is clear that two-stage cancellation is very powerful with almost
all of the multiuser interference cancelled, achieving the perfor-
where “1” represents the threshold decision of the first stage mance equivalent to that of a single user.
cancellation, and the estimated interference is Since carrier phase synchronization is also influenced by
multiuser interference, using the outputs from the two-stage
cancellation process, the carrier phase can be reestimated.
(4) Using the reestimated carrier phase information, third and
fourth stage cancellation can be performed (Fig. 6), a direct
repeat of the first and second stage cancellation, respectively.
2) Second-Stage Cancellation: In the second-stage cancel-
The results of system performance and the requirements placed
lation, after a waiting period of , the data symbols, ,
on the laser linewidth are given in Table I. The variance of the
have been detected as for all users . Since all trans-
phase estimation and the number of simultaneous users
mitted data experience similar processes, at point 4 the users
at an error probability of and received optical power
will have already obtained second-stage estimates for data bits
of 30 dBm for different code length are given. By generating
as using the first-stage cancellation [Fig. 4(b)]. phase reestimation through interference cancellation, the
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 771

TABLE I chromatic dispersion are minimized [10]. Another very impor-

VARIANCE OF PHASE ERROR AND NUMBER OF USERS AT
BER = 10 AT A RECEIVED POWER OF 30 DBM 0 tant issue with laser sources is the jitter on the output pulses,
which should be minimized for successful recovery of the data.
Low jitter permits narrow time separations between adjacent
timeslots or chips. Since OCDMA systems employ full coherent
reception, narrow laser linewidths and high carrier phase sta-
bility is also required.
Multiplexing: As depicted schematically in Figs. 3(b) and 8,
OTDMA multiplexers are simpler than OCDMA encoder ge-
ometries. The added complexity of the encoder in OCDMA re-
sults in an increased optical power loss, since relatively long
spreading codes ( chips) may be required in extensive asyn-
chronous OCDMA implementations. However, the issue will
not be a major obstacle to implementation given the strides made
in component development, especially in opto-hybrid integrated
circuits featuring a combination of passive and active devices.
The signals are multiplexed by an interferometer-type multi-
plexer realized in a planar lightwave circuit geometry [52].
Transmission and Dispersion: In both OCDMA and OTDMA
signal transmission, optical amplification is used routinely to
compensate for component insertion losses, avoiding SNR
degradation and maintaining acceptable BER’s. Chromatic
Fig. 8. Schematic of a general OTDMA systems using optical pulse laser
sources. dispersion is another major issue, and techniques such as
dispersion compensation [55], spectral inversion and dispersion
management [56], [57] are required to reduce its effects on
requirement on the laser linewidth to bit rate ratio
transmission performance. For step-index fiber, dispersion
can be reduced from to .
compensation techniques using opposite dispersion optical
fiber at the transmitter have been used to provide unrepeatered
III. SYSTEM COMPARISON: OTDMA VERSUS OCDMA transmission. For dispersion shifted fiber systems, nonlinear
OTDMA utilising narrow optical pulse laser sources is a well- transmission (solitons) can be used over a wide wavelength
documented method of multiplexing for very high rate trans- range or minimization of the dispersion can be achieved by op-
mission (40–100 Gb/s or higher) at a single wavelength. Since erating near the dispersion zero. OCDMA is robust to crosstalk
many approaches to implementing an OTDMA system can be from adjacent chips because the crosstalk is asynchronous to
found in the literature together with experimental proof-of-prin- the intended chips, and can be rejected by the despreading
ciple demonstrations [9]–[11], [50]–[54], Fig. 8 simply shows process. In contrast, the crosstalk from an adjacent timeslot
the basic principles of operation. A pulsed laser is used to pro- in OTDMA directly influences the decision making process.
duce a regular train of narrow optical pulses (pulse width con- Since CDMA channel signals are completely random, i.e., no
sidered to be the same as the timeslot) at a repetition rate that “distinct” pulses can be observed in the channel, OCDMA
corresponds to the bit rate of the (electrical) data signals. The techniques are more resistant to crosstalk from adjacent chips
optical pulse train is split into several separate paths, and each of and are thus more robust to dispersion than OTDMA.
them is individually modulated by an electrical data channel, de- Synchronization and Clock Recovery: Both OCDMA and
layed by a fraction of optical delay (equivalent to the timeslot), OTDMA demultiplexers are driven at the electrical data clock
and combined to form a single multiplexed optical data stream. rate and not at the code chip or full optical timeslot data
At the receiver, the optical data stream is demultiplexed, then rates. Hence, both require an accurate electrical clock at the
detected to recover the information. Both electrooptic [50] and transmitter and receiver. The ability to extract clock signals
all-optical gating [51], [52] demultiplexing methods have been from OTDMA data is vital to a successful demultiplexing
demonstrated. Some means of clock recovery at the bit rate is process. Either phase locked loops (electro-optic or optical)
required to drive and synchronise all demultiplexers [53], [54]. [53], [58], [59] or all-optical mode-locking [54] clock recovery
A comparison between OCDMA and OTDMA, from the techniques have been demonstrated. All optical clock recovery
system implementation viewpoint with particular regard to techniques utilize a component nonlinearity either in optical
technology issues, is summarized in Table II and discussed. fiber and semiconductor laser amplifiers for their operation.
Laser Sources: Both OCDMA and OTDMA systems employ In comparison, time synchronization in OCDMA systems is
narrow optical pulsed laser sources. To enable multiplexing with much simpler, as described in the previous section. However,
low crosstalk between adjacent channels, the duration of the op- fine carrier phase synchronization is required to lock the carrier
tical pulse must be significantly less than the duration of the al- phase of the incoming signal to the local code.
lowed timeslot or chip duration. In addition, the pulses have to Demultiplexing: Both electro-optic [50] and all-optical [51],
be as spectrally pure as possible so that pulse broadening prob- [52] nonlinear or gating demultiplexing methods have been
lems due to the interaction of source characteristics with fiber demonstrated in OTDMA systems, most of the devices required
772 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

TABLE II
COMPARISON OF OCDMA AND OTDMA TECHNIQUES USING NARROW OPTICAL PULSE SOURCES

however are still confined to the laboratory. The complexity

of the demultiplexing process in OTDMA, and the relative
immaturity of the hardware required for its implementation,
coupled with the difficulties of clock recovery and maintaining
synchronization between nodes at ultra high bit rates, is a major
obstacle. In contrast, notwithstanding carrier phase synchro-
nization and the concomitant requirement with respect to the
laser linewidth, OCDMA receivers rely on a well-understood
process using commercially available, mature components. The
simplicity of the demultiplexing, given practically realisable
code phase synchronization, is one of the major advantages
of OCDMA. With the use of the interference cancellation, Fig. 9. Wavelength and code allocation within optical fiber networks based on
the OCDMA system matches the transmission capacity of OCDMA/WDMA.
OTDMA.
Integration with WDMA: To increase transmission capacity, However, through integration with WDMA, the application
OTDMA can be combined with WDMA. Several systems com- sector for OCDMA can potentially extend to the transport
bining the OTDMA with WDMA have been developed, such as layer of the telecommunication hierarchy. The partition of
reported in [27]–[29]. The componentry derived form the above the available optical bandwidth for the use within a hybrid
developments can be harnessed to develop OCDMA/WDMA OCDMA/WDMA network is illustrated in Fig. 9. The total
approaches to system realization. spectrum is split into a number of the subbands, and each of the
subbands is further utilized as optical spread spectrum. Low
IV. OCDMA/WDMA BASED OPTICAL FIBRE NETWORKS bit rate users or user grouptraffic are assigned codes and thus
multiplexed onto the subbands.
WDMA is a simple, natural approach to harnessing the band-
width of optical fibers. However with ever-increasing demand B. Integrating OCDMA and WDMA
to support higher levels of traffic, a pure WDMA stance to pro- There are a number of ways to implement an
viding network capacity and functionality will not be sufficient, OCDMA/WDMA network. A simple example is shown in
fundamentally limited by the number of useable wavelengths. Fig. 10. To develop a flexible OCDMA/WDMA network,
Therefore another layer of multiplexing is required for future the entire network area is divided into several subzones,
network expansion; OCDMA and OTDMA are both candidates deploying a subnetwork (subnet) in each subzone. It is a
as additional multiplexing strategies on each WDMA subband. three-layered network, consisting of a number of subnetworks
Here coherent OCDMA approaches are developed taking into interconnected by access nodes. Each of the access nodes
consideration the implementation arguments presented in the also serves as a gateway for the user and transport tiers.
previous sections. Traffic from the user subnet nodes is offered into the transport
network via the access nodes. As the traffic routed along
A. Partition of Optical Bandwidth the optical path from access node A to access node C and
Implementing OCDMA in large networks has posed a diffi- via access node B increases, the capacity across the optical
cult challenge for network designers. This is because OCDMA path will then have to increase. In a pure WDMA network
by itself is a broadcast technology, with all information going to implementation, adding more wavelength channels within each
all parts of the network, most suitable for local area networks. fiber normally accommodates traffic growth. However, in the
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 773

Fig. 10. A Simplified three layered network.

Fig. 11. OCDMA/WDMA network operating strategy.

proposed OCDMA/WDMA network, this can be achieved by all the photonic provisioning and restoration functions within
introducing coded channels at each wavelength. the network. Traffic entering the WDMA layer is multiplexed
Fig. 11 shows the various stages experienced by the traffic onto the appropriate wavelengths. Traffic is routed across the
routed from access node A to access node C. Traffic from the network according to its wavelength and not its header address.
user subnet is first multiplexed onto a coded channel or a set of Hence, it provides a course routing function. It also provides an
coded channels. It is then routed to the transport node adjacent optical path protection and configuration function by reconfig-
to the access node. At the transport node, the traffic is multi- uring and rerouting the wavelength channels. All traffic carried
plexed onto an appropriate wavelength according to its destina- by the affected channels will be reconfigured or rerouted ac-
tion. At each intermediate transport node, traffic is routed to the cording to the new optical configurations and routes defined by
appropriate output port according to its wavelength. The traffic wavelength.
may undergo wavelength conversion, performed directly in the The OCDMA layer builds on the optical topology provided
optical domain using optical wavelength converters. A generic by the WDMA layer. It provides codes for each individual
transport node architecture is shown in Fig. 12. traffic stream offered by the user subnet layer. Effectively
the function of the OCDMA layer is to increase the capacity
C. The Management Hierarchy offered within the wavelength channels. Each traffic stream
Fig. 13 shows the overall management hierarchy for an can be multiplexed into a single or a set of coded channels
OCDMA/WDMA network. The physical layer transmits at each wavelength. This allows multiple class traffic types
bits over a communication channel. It deals with the optical to be carried within a single wavelength channel. Hence, the
interfaces over the physical medium and also provides network OCDMA layer provides all the necessary management and
security functions such as network failure protection. operating functions for each coded channel and indirectly
The WDMA layer builds on the transmission capability of the for each traffic stream. These include routing of each traffic
physical layer. It provides the services to the OCDMA layer and stream from one user subnet to another user subnet across the
774 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

Fig. 12. Schematic architecture of the transport node.

the network condition. It will evaluate the shortest path given

the free wavelength channel status. This information about the
node condition is shared among the transport nodes within the
same transport subnet and is regularly updated. Once the path
has been identified, the access node will reserve the wavelength
channel at the selected output fiber of the access node. The ac-
cess node will then forward the REQ to the transport node con-
nected to the output fiber. This intermediate transport node will
then decide whether to forward the REQ message to the next in-
termediate transport node. If no wavelength channels along the
selected path are free, then the REQ message will be removed
and a failure acknowledgment (NACK) message is sent back to
the user subnet node. If there is a free wavelength channel then
Fig. 13. General network management functions.
that wavelength is reserved and the REQ will be forwarded to
the next intermediate transport node. The procedure is repeated
hybrid network, restoring failed traffic connection by assigning at each intermediate node until it reaches the access node con-
alternative coded channels or a set of coded channels and nected to the destination subnet.
resolving the channels contention using code converters. The If the destination user subnet node is connected to a different
user subnet layer manages the way the traffic is formatted transport subnet, then the access node connected to the source
or presented such as SDH, ATM, or Internet protocol (IP). It user subnet will try to establish connection to the access node
provides the monitoring and control functions on traffic offered connecting the destination transport subnet. That access node to
into the transport tier. the destination user subnet node then uses the same REQ mes-
In the network example shown in Fig. 10, each transport node sage in order to establish connection within the transport subnet.
does not have a direct add-drop capability. In order to introduce Whenever a connection is successfully reserved, than the ac-
this capability, an access node functionality is added into each of cess node connecting the destination user subnet will send an
the transport layers. This reduces the network layers to two: the acknowledgment back to the access node connecting the source
user and the transport tiers. Hence, traffic from the user tier can user subnet. Once the source user subnet node receives the ACK
be directly offered into the transport tier as shown in Fig. 14. The message, it will release the traffic into the transport network.
schematic architecture of the transport node is shown in Fig. 15. The network shown in Fig. 14 however requires different con-
nection protocols. Unlike the protocols described above, each
D. Connection Protocol code channel can hop onto a wavelength channel occupied by
Assume that each user subnet node is allocated a set of dis- some other connections. This is because of the add–drop ca-
tinct codes. In the network shown in Fig. 10 connections can pability introduced within each transport node. The transport
established using the existing network wavelength assignment node connecting the source user subnet node will first seek suit-
and connection protocols [60], [61]. For example, whenever a able paths with high probability that one of the wavelengths
user subnet node wishes to communicate with other user subnet within the fiber along the path is not occupied by traffic using
nodes, it sends a request message (REQ) to the access node. the same codes. Once identified, the access node will reserve the
The access node will then evaluate the suitable path to estab- codes within the wavelength channel at the selected output port
lish the connection based on the information it receives about and forward the REQ to the intermediate transport node. The
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 775

Fig. 14. Upgraded transport layer function.

Fig. 16. Code assignment protocol.

Fig. 15. A schematic architecture of the transport node with add–drop

capability. The advantage of the second network architecture and the
protocols is that it requires less wavelength channels as traffic
transport node will then search for wavelength channels that are can hop to any wavelength channel occupied by other connec-
not carrying any of the codes reserved by the REQ. Once found tions. However, the codes used within the wavelength channel
the REQ is forwarded to the subsequent intermediate node. If must not be the same as the codes assigned to the traffic.
not, a NACK message will be sent back to the user subnet node
and all the reserved codes are released. This process is repeated E. Network Performance
at each intermediate node until the REQ reaches the transport Simulations were carried out on the network example shown
node connecting the destination user subnet node. Each time in Fig. 10. Two different network connections were compared.
a wavelength channel that is not carrying the wanted codes is In the first network connection, each transport node is physi-
identified, the codes within the channel are reserved. The desti- cally connected to an access node and a number of other trans-
nation transport nodes then send an ACK message back to the port nodes, referred to as NET1. Each access node serves as
source user subnet node. Each time the ACK arrives at an inter- a gateway to four different traffic sources or users. Each fiber
mediate node it causes the node to update its routing table and within the optical layer carries four wavelength channels. In
configure its switch connection. Once the user subnet receives the OCDMA/WDMA network each wavelength channel can be
the ACK message, it starts releasing the traffic onto the defined multiplexed onto four different codes. The second network con-
code channels into the transport network. A block diagram de- nection, NET2, is similar to NET1 except it has two different
scribing the reservation algorithm is shown in Fig. 15. access nodes connected to each transport node.
776 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

Fig. 17. Network throughput at maximum load condition comparing

OCDMA/WDMA strategy and WDMA strategy.

Two different types of multiplexing strategies were con-

sidered; a hybrid OCDMA/WDMA and pure WDMA. Their
performance was analyzed by congesting the network. Each
network user node generates a load at maximum capacity equiv-
alent to a single wavelength channel bandwidth. The worst case
condition was considered whereby all nodes tried to establish
connection with a common destination simultaneously. The
results shown in Fig. 17 indicate that a network employing
Fig. 18. Average delay at access node comparing OCDMA/WDMA strategy
OCDMA/WDMA is capable of establishing more connections and WDMA strategy.
than a network using WDM only. These results also suggest
that OCDMA/WDMA reduces connection setup failure under
any network condition and hence, reduces connection set-up component nonlinearities i.e. effectively “spreading” the
time. This is verified by another set of simulations on NET2 limitations introduced by high peak power pulses.
which measure the delay experienced by traffic from one user • Fully asynchronous operation providing a flexible access
destination wishing to be routed to a particular user destination technique.
under two different network conditions. The first considers the • Accurate time of arrival measurement offering a means to
case whereby only 30% of the user subnet nodes within the achieve fine time synchronization.
network are busy communicating with that user destination, • Offering a natural increase in the transmission security, not
the second considers the case whereby 90% of the user subnet exhibited by any other multiplexing technique.
nodes is communicating with that user destination. It is assumed • Since the high throughput can be obtained by multiplexing
that whenever a connection is successfully established, the user many low bit rate OCDMA users, the network has high
will hold on to the connection for an average of s. From the capacity but low implementation complexity. This imple-
simulations it was found that each user in an OCDMA/WDMA mentation allows processing to be carried out at the lower
network encounters less delays at the access node to establish a bit-rate, rather than at the chip rate. Hence, conventional
connection than with pure WDMA (Fig. 18). electronic processing using existing chip sets can support
the network.
F. Advantages of OCDMA/WDMA Networks • Large transmission capacity, harnessing the huge inherent
optical spectrum bandwidth through narrow optical
The hybrid OCDMA/WDMA networks preserve the advan-
pulses.
tages of both CDMA and WDMA, offering a series of attributes
• Network expansion can be realized through increasing the
summarized as follows.
number of subnets.
• A multiplexing hierarchy utilising two degrees of
freedom, wavelength and code, defining a larger number
V. CONCLUSION
optical paths than available with pure WDMA.
• The OCDMA dimension allows access to a larger number A single-mode optical fiber has a total low loss bandwidth of
of addresses, able to support variable bit rate services as 25 000 GHz. In the future, WDMA will not be the sole multi-
well as bursty traffic. Alternatively a fewer number of plexing strategy to fully exploit this bandwidth for high-capacity
wavelengths would be required to support a certain level optical networking. As the available wavelengths resource in
of traffic. mainstream WDMA systems becomes exhausted and given the
• The OCDMA technique within the WDMA subbands can bit rate limitation within a single wavelength bandwidth has to be
effectively reduce the transmission limitations imposed by overcome, another multiplexing layer has to be used. OTDMA
HUANG et al.: COHERENT OPTICAL CDMA (OCDMA) SYSTEMS USED FOR HIGH-CAPACITY OPTICAL FIBER NETWORKS 777

is an obvious approach to achieving this goal, increasing the bit [21] F. Liu, R. J. S. Pedersen, and P. Jeppesen, “Very low crosstalk wave-
rate carried by each wavelength. Here, another approach has length router construction using arrayed-waveguide grating multi/de-
multiplexers,” Electron. Lett., vol. 35, no. 10, pp. 839–840, May 1999.
been considered; coherent OCDMA techniques using coherent [22] H. Takahashi, K. Oda, and H. Toba, “Impact of crosstalk in an arrayed-
reception in tandem with WDMA are proposed. The implemen- waveguide multiplexer on N 2N optical interconnection,” J. Lightwave
tation of these OCDMA techniques was introduced and possible Technol., vol. 14, no. 6, pp. 1097–1105, June 1996.
[23] C.-S. Li and F. Tong, “Crosstalk and interference penalty in all optical
optical network architectures based on OCDMA/WDMA were networks using static wavelength routers,” J. Lightwave Technol., vol.
postulated. A comparison between OTDMA and OCDMA, both 14, no. 6, pp. 1120–1126, June 1996.
based on short-pulse laser source, is given in which implemen- [24] C. A. Brachett, “Foreword is there an emerging consensus on WDM
networking?,” J. Lightwave Technol., vol. 14, pp. 936–941, June 1996.
tation issues are central to the arguments. [25] D. Norte and A. E. Willner, “All-optical data format conversions and
reconversions between the wavelength and time domains for dynami-
REFERENCES cally reconfigurable WDM networks,” J. Lightwave Technol., vol. 14,
pp. 1170–1182, June 1996.
[26] J. P. R. Lacey, M. V. Chan, R. S. Tucker, A. J. Lowery, and M. A. Sum-
[1] M. de Prycker, Asynchronous Transfer Mode: Solution for Broadband merfield, “All-optical WDM to TDM transmultiplexer,” Electron. Lett.,
ISDN: Ellis Horwood, 1993. vol. 30, pp. 1612–1613, 1994.
[2] I. V. D. Voorde and G. V. D. Plas, “Full service optical access networks: [27] S. Kawanishi, H. Takara, K. Uchiyama, I. Shake, O. Kamatani, and H.
ATM transport on passive optical networks,” IEEE Commun. Mag., vol.
35, no. 4, pp. 70–75, Apr. 1997.
Takahashi, “1.4 Tbits ( 200 Gbit/s 27 ch) 50 km optical transmission
experiment,” Electron. Lett., vol. 33, pp. 1716–1717, 1997.
[3] “Architechtures of Transport Networks Based on the Synchronous [28] C. D. Chen, I. Kim, T. V. Nguyen, K. Ogawa, R. E. Tench, L. D. Tzeng,
Digital Hierarchy (SDH),” CCITT Recommendations G. 803, CCITT
Meeting Rep. R-106 (WP XVIII), COM XVIII-R 106-E, July 1992.
40
and P. D. Yeates, “ Gbit/s 225 ch (1 Tbit/s aggregate capacity) WDM
transmission over 342 km of fiber,” Electron. Lett., vol. 35, pp. 648–649,
[4] P. E. Green Jr., “Optical networking update,” IEEE J. Select. Areas 1999.
Commun., vol. 14, pp. 761–779, June 1996. [29] S. Kawanishi, H. Takara, K. Uchiyama, I. Shake, and K. Mori, “3 Tbit/s
[5] K. Nosu, H. Toba, K. Inoue, and K. Oda, “100 channel optical FDM tech-
nology and its applications to optical FDM channel-based networks,” J.
(160 Gbit/s 219 channel) optical TDM and WDM transmission exper-
iment,” Electron. Lett., vol. 35, pp. 826–827, 1999.
Lightwave Technol., vol. 11, pp. 764–776, 1993. [30] C. E. Cook, F. W. Ellersick, L. B. Milstein, and D. L. Schilling, Spread-
[6] M. S. Goodman, H. Kobrinski, and K. W. Loh, “Application of wave-
Spectrum Communication. New York: IEEE Press, 1983.
length division multiplexing to communication network architectures,”
[31] W. C. Y. Lee, “Overview of cellular CDMA,” IEEE Trans. Veh. Technol.,
in Proc. ICC’86, 1986.
vol. 40, pp. 291–302, May 1991.
[7] R. A. Barry and P. A. Humblet, “On the number of wavelengths and
[32] W. Huang and M. Nakagawa, “Nonlinear effect of direct-sequence
switches in all-optical networks,” IEEE Trans. Commun., vol. 42, no.
CDMA in optical transmission,” IEICE Trans. Commun., vol. E78-B,
2/3/4, Feb./Mar./Apr. 1994.
no. 5, pp. 702–708, May 1995.
[8] C. A. Brackett, “Dense wavelength division multiplexing networks:
[33] M. Kavehrad and P. J. McLane, “Spread spectrum indoor digital radio,”
Principle and applications,” IEEE J. Select. Areas Commun., vol. 8, pp.
IEEE Commun. Mag., vol. 25, pp. 32–40, June 1987.
948–964, Aug. 1990.
[9] R. A. Barry et al., “All-optical network consortium—Ultrafast TDM net- [34] K. S. Gilhousen, I. M. Jacobs, R. Padovani, and L. A. Weaver, “Increased
works,” IEEE J. Select. Areas Commun., vol. 14, pp. 999–1013, June capacity using CDMA for mobile satellite communication,” IEEE J. Se-
1996. lect. Areas Commun., vol. 8, pp. 503–514, May 1990.
[10] D. M. Spirit, A. D. Ellis, and P. E. Barnsley, “Optical time division [35] N. Morinaga, M. Nakagawa, and R. Kohno, “New concepts and tech-
multiplexing: System and networks,” IEEE Commun. Mag., vol. 32, pp. nologies for achieving highly reliable and high-capacity multimedia
56–62, Dec. 1994. wireless communications systems,” IEEE Commun. Mag., vol. 35, pp.
[11] S.-W. Seo, K. Bergman, and P. R. Prucnal, “Transparent optical networks 34–40, Jan. 1997.
with time-division multiplexing,” IEEE J. Select. Areas Commun., vol. [36] W. Huang and I. Andonovic, “Optimal performance of coherent optical
14, pp. 1039–1051, June 1996. pulse CDMA systems based on code and phase synchronization and
[12] L. A. Rusch and H. V. Poor, “Effects of laser phase drift on coherent interference cancellation,” Proc. Inst. Elect. Eng.—Optoelectron., vol.
optical CDMA,” IEEE J. Select. Areas Commun., vol. 13, pp. 577–591, 145, no. 6, pp. 353–359, Dec. 1998.
Apr. 1995. [37] G. Vannucci, “Combining frequency-division and code-division multi-
[13] G. J. Foschini and G. Vannucci, “Using spread-spectrum in a high-ca- plexing in a high-capacity optical network,” IEEE Commun. Mag., pp.
pacity fiber-optic local network,” J. Lightwave Technol., vol. 3, pp. 21–30, Mar. 1989.
[38] G.-C. Yang, “Performance comparison of multiwavelength CDMA and
+
370–379, Mar. 1988.
[14] P. R. Prucnal, M. Santoro, and T. Fan, “Spread spectrum fiber optical WDMA CDMA for fiber-optic networks,” IEEE Trans. Commun.,
local area network using optical processing,” J. Lightwave Technol., vol. vol. 45, no. 11, pp. 1426–1434, Nov. 1997.
4, pp. 307–314, May 1986. [39] W. Huang and K. Kitayama, “Code acquisition in optical pulse
[15] J. A. Salehi, “Code division multiple-access techniques in optical fiber CDMA utilizing coherent correlation demodulation,” in Proc. IEEE
networks—Parts I: Fundamental principles,” IEEE Trans. Commun., GLOBECOM’97, Apr. 2000, pp. 265–269.
vol. 37, pp. 824–833, Aug. 1989. [40] W. Huang and I. Andonovic, “Code tracking in optical pulse CDMA
[16] W. Huang and I. Andonovic, “Coherent optical pulse CDMA systems through coherent correlation demodulation,” in Proc. IEEE ICC’98, At-
based on coherent correlation detection,” IEEE Trans. Commun., vol. lanta, GA, June 1998, pp. 1264–1268.
47, pp. 261–271, Feb. 1999. [41] W. Huang, I. Andonovic, and M. Tur, “Decision-directed PLL used for
[17] A. R. Chrapllylvy, “Limitations on lightwave communications imposed coherent optical pulse CDMA systems in the presence of multiuser in-
by optical-fiber nonlinearities,” J. Lightwave Technol., vol. 8, pp. terference, laser phase noise and shot noise,” J. Lightwave Technol., vol.
1548–1557, Oct. 1990. 16, pp. 1786–1794, Oct. 1998.
[18] T. Matsuda, M. Murakami, and T. Imai, “340 Gbit/s ( 34 2 10 Gbit/s) [42] W. Huang, M. H. M. Nizam, I. Andonovic, and M. Tur, “OCDMA/WDM
WDM transmission over 8 514 km using broadband gain equalization networks based on coherent OCDMA techniques with interference can-
technique for transoceanic systems,” Electron. Lett., vol. 35, no. 13, pp. cellation,” in Proc. OFC’2000, submitted for publication.
1090–1091, June 1999. [43] D. V. Sarwate and M. B. Pursley, “Crosscorrelation properities of pseu-
[19] H. Suzuki, N. Takachio, H. Masuda, and M. Koga, “50 GHz, 32 2 10 dorandom and related sequences,” Proc. IEEE, vol. 68, pp. 593–619,
Gbit/s dense WDM transmission in zero-dispersion region over 640 km May 1980.
of dispersion-shifted fiber with multiwavelength distributed Raman am- [44] D. J. G. Mestdagh, Fundamentals of Multiaccess Optical Fiber Net-
plification,” Electron. Lett., vol. 35, no. 14, pp. 1175–1176, July 1999. works. Norwood, MA: Artech House, 1995.
[20] C. D. Chen, I. Kim, O. Mizuhara, T. V. Nguyen, K. Ogawa, R. E. Tench, [45] J. M. Kahn, A. H. Gnauch, J. J. Veselka, S. K. Korotky, and B.
40
L. D. Tzeng, and P. D. Yeates, “ Gbit/s 2 25 ch (1 Tbit/s aggregate L. Kasper, “4-Gb/s PSK homodyne transmission system using
capacity) WDM transmission over 342 km of fiber,” Electron. Lett., vol. phase-locked semiconductor lasers,” IEEE Photon. Technol. Lett., vol.
35, no. 8, pp. 648–649, Apr. 1999. 2, no. 4, pp. 285–287, 1990.
778 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 6, JUNE 2000

[46] S. Norimatsu, H. Mawatari, Y. Yoshikuni, O. Ishida, and K. Iwashita, “10 Mohamed H. M. Nizam (S’93–M’96) received the
Gbit/s optical BPSK homodyne detection experiment with solitary DFB first class honors B.Eng. degree and the Ph.D. degree
laser diodes,” Electron. Lett., vol. 31, no. 2, pp. 125–127, Jan. 1995. in electronic and electrical engineering at the Univer-
[47] S. Verdu, “Minimum probability of error for asynchronous Gaussian sity of Strathclyde, U.K.
multiple-access channels,” IEEE Trans. Inform. Theory, vol. IT-32, pp. He is currently a Research Staff at the same
85–96, Jan. 1986. university. His research interests include optical
[48] E. A. Sourour and M. Nakagawa, “A modified multi-stage co-channel packet switched networks design, performance
interference cancellation in asynchronous CDMA systems,” in Proc. 5th evaluation (including traffic analysis), network
IEEE Int. Symp. Personal, Indoor Mobile Radio Commun. (PIMRC’94), management, resource allocation, intelligent routing
Hague, The Netherlands, Sept. 1994, pp. 425–429. and restoration/survivability control design.
[49] M. B. Pursley, “Performance evaluation for phase-coded spread-spec-
trum multiple-access communications—Part I: System analysis,” IEEE
Trans. Commun., vol. COM-25, pp. 795–799, Aug. 1977.
[50] A. D. Ellis, T. Widdowson, X. Shan, and D. G. Moodie, “Three-node,
40 Gbits/s OTDM network experiment using electro-optic switches,”
Ivan Andonovic (S’79–SM’97) is a Professor of
Electron. Lett., vol. 30, no. 16, pp. 1333–1334, Aug. 1994.
Broadband Networks, who joined the Electronic and
[51] S. Kawanishi and O. Kamatani, “All-optical time division multiplexing
Electrical Engineering Department at the University
using four-wave mixing,” Electron. Lett., vol. 30, no. 20, pp. 1697–1698,
of Strathclyde, U.K., in 1985, following a three-year
Sept. 1994.
period as Research Scientist (at Barr & Stroud)
[52] K. Suzuki, K. Iwatsuki, S. Nishi, and M. Saruwatari, “Error-free de-
responsible for the design, manufacture, and test of
multiplexing of 160 Gbits pulse signal using optical loop mirror in-
guided wave devices for a variety of applications. His
cluding semiconductor laser amplifier,” Electron. Lett., vol. 30, no. 18,
main interests center on the development of optical
pp. 1501–1502, Sept. 1994.
networks, specifically in guided wave architectures
[53] D. M. Patrick and R. J. Manning, “20 Gbits all-optical clock recovery
for implementing optical signal processing, and
using semiconductor nonlinearity,” Electron. Lett., vol. 30, no. 2, pp.
switching functions for future networks. He was
151–152, Jan. 1994.
holder of a two-year Royal Society Industrial Fellowship in collaboration with
[54] S. Kawanishi and M. Saruwatari, “Ultra-high-speed PLL-type clock re-
BT Labs, during which time he was tasked with investigating novel approaches
covery circuit based on all-optical gain modulation in travelling-wave
to networking. He has edited two books and authored/co-authored five chapters
laser diode amplifier,” J. Lightwave Technol., vol. 11, pp. 2123–2129,
in books and over 170 journal and conference papers. He has been Chairman
Nov. 1993.
on the IEE professional group E13, held a BT Short Term Fellowship and held
[55] A. D. Ellis and D. M. Spirit, “Unrepeatered transmission over 80 km
Visiting Scientist status with the Communications Research Labs. of Japan.
standard fiber at 40 Gb/s,” Electron. Lett., vol. 30, no. 1, pp. 72–74, Jan.
Prof. Andonovic is a Fellow of the IEE and a member of the Optical Society
1994.
of America (OSA).
[56] M. Nakazawa, K. Suzuki, H. Kubota, E. Yamada, and Y. Kimura,
“Straight-line soliton data transmission at 20 Gbit/s beyond
Gordon-Haus limit,” Electron. Lett., vol. 30, no. 16, pp. 1331–1333,
Aug. 1994.
[57] H. Kubota and M. Nakazawa, “Soliton transmission control in time
Moshe Tur (M’98–SM’94–F’98) was born in Tel-Aviv, Israel, in 1948. He
and frequency domain,” IEEE J. Quantum Electron., vol. QE-29, pp.
received the B.Sc., M.Sc., and Ph.D. degrees from the Hebrew University,
2189–2197, 1993.
Jerusalem, Israel, the Weizmann Institute of Science, Rehovot, Israel, and
[58] O. Kamatana, S. Sawanishi, and M. Saruwatari, “Prescaled 6.3 GHz
Tel-Aviv University, Tel-Aviv, Israel , respectively.
clock recovery from 50 Gbit/s TDM optical signal with 50 GHz PLL
He spent the academic years 1981–1983 at Stanford University, Stanford, CA,
using four-wave mixing in a travelling-wave laser diode optical ampli-
where, initially following his Ph.D. work, he studied speckle phenomena, and
fier,” Electron. Lett., vol. 30, no. 10, pp. 807–809, May 1994.
various theories of wave propagation in random media. Later, he participated
[59] O. Kamatani and S. Kawanishi, “Ultrahigh-speed clock recovery with
in the development of new architectures for single-mode fiber optic signal pro-
phase lock loop based on four-wave mixing in a traveling-wave laser
cessing, and investigated the effect of laser phase noise on such processors. He
diode amplifier,” J. Lightwave Technol., vol. 14, pp. 1757–1767, Aug.
is presently the Chairman of the Interdisciplinary Department and Professor of
1996.
Electrical Engineering in the Faculty of Engineering at Tel-Aviv University,
[60] R. Ramaswami and A. Segall, “Distributed network control for wave-
Tel-Aviv, Israel, where he established a fiber-optic sensing laboratory. He au-
length routed optical network,” in Proc. IEEE INFOCOM’96, 1996, pp.
thored or co-authored more than 150 journal and conference technical papers
138–147.
with recent emphasis on fiber-optic bit rate limiters, fiber lasers, fiber-optic
[61] Y. Mei and C. Qiao, “Efficient distributed control protocols for WDM
sensor arrays, the statistics of phase-induced intensity noise in fiber-optic sys-
all-optical networks,” in Proc. Int. Conf. Computer Commun. Networks
tems, fiber sensing in smart structures, and polarization mode dispersion in fiber
(IC3N), Sept. 1997.
optic communication and sensing.
Dr. Tur is also a Fellow of the Optical Society of America (OSA).

Wei Huang (S’94–M’96) received the B.S. degree

in physics from Jinan University, Guangzhou, China,
in 1985 and the M.E. and Ph.D. degrees in electrical
engineering from Keio University, Japan, in 1993 and
1996, respectively.
He was a Visiting Researcher at the Commu-
nications Research Laboratory, Ministry of Posts
and Telecommunications, Japan, from December
1995 to December 1997 under the COE programme.
Since January 1998, he has been with University of
Strathclyde, U.K. His research interests include SS
CDMA techniques for both mobile communications and high-capacity optical
fiber networks, error-correcting coded modulation and optical atmospheric
transmissions. He was a Visiting Researcher at the Tel-Aviv University, Israel,
in November 1998.