This work presents the results of the study on the longitudinal modes behavior in an erbium doped fiber ring laser that uses two optical fibers--one mechanic and another electric. The operation of both filters is combined to cover the whole spectral range corresponding to the erbium fluorescence. The influence of the cavity length, host matrix and combination of two optical filters on the mode suppression and performance of the laser is discussed. We obtained the match conditions for operation of the optical filters which better improve the laser performance. An estimate of the relative intensity noise due to the introduction of the laser in a communication system is also presented.

We report on the development of a compact and fully optical fiber coupled memory unit based on time-domain holography in Er³⁺:Y₂SiO₃ for the storage of large bandwidth, bit-serial optical data at 1.536 and 1.538 micrometers . Using this demonstration unit, we have successfully stored 1020 bits of data in a 40-MHz wide spectral channel at a single spatial location in the sample. The raw bit error rate of the unit was measured to be approximately 10^-7 with the use of a simple peak detection scheme. We further show that the power of the retrieved data can be restored completely to the level of input data by a single erbium-doped fiber amplifier. Implications of the results for an all-optical dynamic optical memory are discussed.

The effect produced by Four-Wave Mixing (FWM) in a Wavelength Division Multiplexing (WDM) transmission is complex to analyze due to its many dependencies. When the optical channels are equally spaced in a WDM transmission, new optical generated signals will fall entirely onto other channels thus causing inband crosstalk. The channel that will suffer from more crosstalk components is that in the middle of the signal band. In this paper we present a theoretical study of the total crosstalk power generated onto each optical channel, equally spaced, using Dispersion Shifted Fiber. Our results show that when the WDM signal range is larger than a certain value, the most degraded channel is not the one that suffers from more contributions but the one that suffers from contributions with higher FWM efficiency. We have studied the limitation that FWM imposes on the number of channels and channel spacing regarding the most degraded channel.

We develop a perturbation theory to analytically calculate effects of interchannel collisions on Gaussian pulses in a wavelength-division-multiplexed system with strong dispersion management. We show that, for complete collisions, the collision-induced frequency shift of a Gaussian pulse is negligible, while for incomplete collisions, this shift is significant. We also show that, as the dispersion management strength increases, the collision- induced position shift becomes more important than the frequency shift. An interesting result of our analysis is that the collision-induced frequency and position shifts depend on the dispersion management strength and the path- average dispersion, but not on the lengths of the two opposite-dispersion fiber segments in the dispersion map. We check the analytical predictions against direct PDE simulations, finding satisfactory agreement between them. We also give an estimate of the limit imposed on the transmission distance by such collisions.

Erbium doped fiber amplifiers (EDFAs) are often used in multiwavelength networks. When these networks are reconfigured or when wavelength-channel failures occur, cross-gain modulation in the amplifiers causes power- fluctuations of the remaining channels. Up to now only result of dynamic models of a single EDFA and chains of EDFAs were presented in literature. We investigated the dynamic gain-behavior of EDFAs placed in a ring-network with respect to the effect of recirculating fluctuating crosstalk. For this purpose we implemented the electrical equivalent of an analytical dynamic EDFA-model into SPICE. Also optical add-drop multiplexer (OADM) functionality was transformed into an electrical equivalent. Rings and chains based on 1 to 9 OADMs, with an EDFA in each node were modelled. In the simulations two wavelength-channels were added and dropped, and the effect of a failure of one channel was investigated by monitoring the power of the remaining channel. We found that the rising-speed of the remaining channel power increases with the pass to drop power-ratio (which is actually the crosstalk to signal power-ratio, XSR). This means that a gain control to compensate for the power-fluctuations needs to be faster when EDFAs are configured in a ring-structure instead of a chain with the same number of EDFAs. The dynamic behavior of EDFAs in a ring-structure converges to that of EDFAs in a chain when the XSR reaches zero.

In self-routing packet networks, the state of intermediate nodes (switches) is set or reset on the basis of the information present in the packet header. Subcarrier multiplexing (SCM) modulates a number of frequency-separated RF sub-carriers onto a common laser at a single wavelength. SCM has the advantage of high data throughput. It also requires fewer opto-electronic components and avoids walk- off between header and payload due to fiber dispersion. In this paper we describe a novel use of sub-carrier multiplexing for self-routing of data packet within the switching fabric of a high performance system area network. Using SCM data packets are routed optically to the destination without being converted to the electrical domain at the intermediate stages within the network.

Optical CDMA offers an alternate solution for video transport/switching to WDMA. Optical CDMA potentially provides a large number of virtual optical circuits for video distribution and channel selection. In a video network, it provides asynchronous, multi-rate, multi-channel communication with network scalability, reconfigurability (channel on demand), and network security (provided by inherent CDMA coding). We have demonstrated a video transport/switching system over a distance of 40 Km using discrete optical components in our laboratory. We are currently pursuing Photonic Integrated Circuit implementation. In this paper, we will describe the optical CDMA video transport/switching system concept/features, the demonstration system, and the network applications to Hybrid Fiber/Coaxial, Fiber-To-The-Curb and Fiber-To-The-Home.

Coherent optical pulse CDMA systems based on noncoherent demodulation of M-ary orthogonal signals are proposed. Each of the J information bits is encoded and transmitted as a M (equals2^J) symbol word of a family of Hadamard-Walsh orthogonal sequences. A pulsed laser is employed at the transmitter, modulated by the symbol sequence and encoded by an optical tapped delay-line encoder to generate a unique optical pseudo-random sequence (bipolar code). At the receiver, a pulsed local oscillator followed by a tapped delay-line encoder is employed to produce the optical code sequence of the intended user. Correlation between the received signal and the local code is executed through a coherent optical correlator comprising a 3 dB coupler and balanced detector. Noncoherent demodulation of the M-ary orthogonal signals based on the maximum-likelihood criterion is used to recover the information bits. After a description of the network implementation, the performance of the system is theoretically analyzed and its numerical evaluation given. It is shown that the ratio of the number of available users to the practical code gain at a BER equals 10^-9 is approximately 3.7% - 5%. Although a fewer number of users can be supported compared to full coherent reception (10%), the system no longer requires phase locking whilst keeping the advantages of the coherent approach. Hence the linewidth requirement of the laser sources is relaxed. The performance and the implementation of the system are still significantly better than conventional incoherent optical pulse CDMA systems utilizing unipolar codes.

We report both experimental and theoretical results for a physical implementation of a bipolar encoding scheme suitable for fiber optic networks. The power spectrum of an erbium-doped superfluorescent fiber source is encoded, the bipolar correlations of the codes are verified and rejection of multiple-access interference is demonstrated in a fiber- based testbed. Simulations of the correlation process identify key optical parameters and physical characteristics important to the design of future systems.

Optical wavelength domain code-division multiplexing access (WD-CDMA) using an AOM-based ultrafast optical pulse shaping approach is proposed and demonstrated experimentally at 1550 nm. This new multiplexing technique utilizes wavelength domain codes that are essentially different optical spectral patterns in order to achieve CDMA. In addition to the advantages of the conventional CDMA technique, WD-CDMA can make full use of the entire optical bandwidth without requiring faster optical switches or modulators. This approach also drastically reduces sensitivity to fiber dispersion. Experimentally, we demonstrate an optical spectral encoder using ultrafast optical pulse shaping with 16 wavelength bits over an optical bandwidth of 5 THz. The spectrally-encoded optical pulse generated with the spectral encoder is then decoded with different WD-CDMA codes in the spectral domain. Different code-division channels can thus extract their own bit information while sharing the same spectral-encoded laser pulse as their common carrier. These spectral-encoded pulses are shown using the cross- correlation technique to be confined within a time slot of 15 ps. A larger number of WD bits is also achievable with our system.

We describe a WDM-based optical access network architecture for providing broadband Internet services. The architecture uses a passive collection and distribution network and a configurable Feeder network. Unlike earlier papers that concentrate on the physical layer design of the network, we focus on higher layer architectural considerations. In particular we discuss the joint design of the electronic and optical layers including: WDM Medium Access Control protocols; the choice of electronic multiplexing and switching between the IP and WDM layers: joint optical and electronic protection mechanisms; network reconfiguration algorithms that alter the logical topology of the network in response to changes in traffic; and traffic grooming algorithms to minimize the cost of electronic multiplexing.

The MAC (Medium Access Control) protocol controls B-NT1s' (Optical Network Unit) access to the shared capacity on the PON, this protocol is very important if TDMA (Time Division Multiple Access) multiplexing is used on the upstream. To control the upstream traffic some kind of access protocol has to be implemented. There are roughly two different approaches to use request cells: in a collision free way or such that collisions in a request slot are allowed. It is the objective of this paper to describe a MAC-protocol structure that supports both approaches and hybrids of it. In our paper we grantee the QoS (Quality of Service) of each B-NT1 through LOC, LOV, LOA field that are the length field of the transmitted cell at each B-NT1. Each B-NT1 transmits its status of request on request cell.

In this paper the design of a chip-set for QoS provisioning in ATM-based Passive Optical Networks is discussed. The implementation of a general-purpose switch chip on the Optical Network Unit is presented, with focus on the design of the cell scheduling and buffer management logic. The cell scheduling logic supports `colored' grants, priority jumping and weighted round-robin scheduling. The switch chip offers powerful buffer management capabilities enabling the efficient support of GFR and UBR services. Multicast forwarding is also supported. In addition, the architecture of a MAC controller chip developed for a SuperPON access network is introduced. In particular, the permit scheduling logic and its implementation on the Optical Line Termination will be discussed. The chip-set enables the efficient support of services with different service requirements on the SuperPON. The permit scheduling logic built into the MAC controller chip in combination with the cell scheduling and buffer management capabilities of the switch chip can be used by network operators to offer guaranteed service performance to delay sensitive services, and to efficiently and fairly distribute any spare capacity to delay insensitive services.

FSAN GX has discussed on the cost-effective way to deploy the ATM-PON (Passive Optical Network) system that is suitable for providing broadband services. A common broadband PON applicable to many operators increases the world-wide market for the product. The MAC (Medium Access Control) protocol controls ONUs' (Optical Network Unit) access to the shared capacity on the PON, this product is very important if TDMA (Time Division Multiple Access) multiplexing is used on the upstream. In fact there is a relatively large number of MAC proposals mentioning IEEE802.14. The tree algorithm presents the main advantage of being stable, predictable and performant. In this paper, we proposed a new contention resolution algorithm called DQT (Double Queue Tree based) which can support multi level service under practical environments. Simulations result show that channel utility and delay time are improved than other algorithms under practical environments. Proposed DQT algorithm that using double queue consider priority consists of OLT control part and station part show us good simulation result compare to existing MAC algorithm.

For the future development of residential broadband telecommunication services the present access infrastructure must be upgraded. Fiber-To-The-Home (FTTH) optical networks, though still expensive today, appear to be a future-proof solution. FTTH networks based on passive optical Remote Nodes (RN) and supporting ATM switching (ATM-PONs) have been demonstrated and are going to be standardized. In these networks the optical hardware in the RN is very simple, but a multiple access protocol is needed for upstream traffic control. In most protocols proposed in literature the control is performed electronically in a centralized way by the Central Office. In this paper we study multiple access schemes for FTTH networks which perform upstream traffic control in a distributed way with the intervention of the user Optical Network Terminations (ONTs) and of the RN only. We will also consider the case, based on recent research in photonic packet switching, in which the RN is able to perform all-optical ATM-multiplexing. We will employ ONT-RN signalling to make the relatively expensive RN optical buffer co-operate with the electronic buffers located in the ONTs. The different solutions presented will be compared by simulating the network behavior under various traffic conditions. In conclusion, in this work we evaluate performance advantages achievable in FTTH networks by increasing the optical complexity of the RN and by introducing control schemes which allow a distributed upstream traffic management and, in the case of an RN able to perform all-optical ATM-multiplexing, the co-operation of the RN buffer with the ONT buffers.

We study the issues arising when considering the problem of reconfiguring broadcast optical networks in response to changes in the traffic patterns. Although the ability to dynamically optimize the network under changing traffic conditions has been recognized as one of the key features of multiwavelength optical networks, this is the first in-depth study of the tradeoffs involved in carrying out the reconfiguration process. We first identify the degree of local balancing and the number of retunings as two important, albeit conflicting, objectives in the design of reconfiguration policies. We then formulate the problem as a Markovian Decision Process and we develop a systematic and flexible framework in which to view and contrast reconfiguration policies. We also apply results from Markov Decision Process theory to obtain optimal reconfiguration policies even for networks of large size. The advantages of optimal policies over a class of threshold-based policies are also illustrated through numerical results.

The paper presents optical LAN topologies which are made possible using an Arrayed Waveguide Grating Multiplexer (AWGM) instead of a passive star coupler to interconnect stations in an all-optical LAN. Due to the collision-free nature of an AWGM it offers the n-fold bandwidth compared to the star coupler. Virtual ring topologies appear (one ring on each wavelength) if the number of stations attached to the AWGM is a prime number. A method to construct larger networks using Cayley graphs is shown. An access protocol to avoid collisions on the proposed network is outlined.

This paper presents a new physical layer demonstrator for Optical Local Area Network. It offers a large total capacity to a large number of connected nodes without requiring optical amplifiers and achieves high modularity and scalability. A new multichannel Ethernet protocol tailored to the physical layer is also proposed. The physical layer configuration is based on a passive star coupler and combines the advantages of Frequency Division Multiplexing techniques and coherent heterodyne detection. Our experimental demonstrator is based on a 9*9 star coupler and a shared Local Oscillator. Two nodes are implemented and data are ASK modulated at 140 Mbit/s. Emission and detection schemes are simple: only one laser and one photodetector per node are required. Degradation caused by crosstalk has been measured and found to be negligible when channel separation is four times the bitrate. Experimental results show that our system has a potential total capacity of 6.5 Gbit/s and that more than 100 users can be connected to the network. The new multichannel Ethernet protocol simulations indicate considerable improvement of the performance when compared to a single Fast Ethernet channel protocol with the same total capacity.

We describe a low-latency distributed control architecture for a high-bandwidth, high performance TDM-WDM based local area network developed at the National Research Council of Canada. The control architecture is designed with the aim of low average connection latency and communication wait time amongs network nodes. The protocol supported in the network is lightweight and reliable as control signaling takes place over a dedicated channel requiring no acknowledgement from the destination devices. We present network performance results obtained from modeling the control architecture, medium access control protocol, number of wavelength channels, and transmit and receive component specifications. The results indicate that the control architecture keeps the network latency and communication wait time close to the base value over an impressive range of bandwidth demands.

Most existing or proposed WDM networks employ circuit switching, typically with one session having exclusive use of one entire wavelength. Consequently they are not suitable for data applications involving bursty traffic patterns. The MIT AON Consortium has developed an all-optical LAN/MAN testbed which provides time-slotted WDM service and employs fast-tunable transceivers in each optical terminal. In this paper, we explore extensions of this service to achieve fine-grained statistical multiplexing with different virtual circuits time-sharing the wavelengths in a fair manner. In particular, we develop a real-time distributed protocol for best-effort traffic over this time-slotted WDM service with near-optical fairness and throughput characteristics. As an additional design feature, our protocol supports the allocation of guaranteed bandwidths to selected connections. This feature acts as a first step towards supporting integrated services and quality-of-service guarantees over WDM networks. To achieve high throughput, our approach is based on scheduling transmissions, as opposed to collision- based schemes. Our distributed protocol involves one MAN scheduler and several LAN schedulers (one per LAN) in a master-slave arrangement. Because of propagation delays and limits on control channel capacities, all schedulers are designed to work with partial, delayed traffic information. Our distributed protocol is of the `greedy' type to ensure fast execution in real-time in response to dynamic traffic changes. It employs a hybrid form of rate and credit control for resource allocation. We have performed extensive simulations, which show that our protocol allocates resources (transmitters, receivers, wavelengths) fairly with high throughput, and supports bandwidth guarantees.

This paper considers multicast routing and wavelength assignment (MCRWA) in a wide area all-optical network. We present static and dynamic MCRWA algorithms, and evaluate their blocking performance under various fanout splitting policies. Through simulations and comparisons with a lower bound, we demonstrate that dynamic MCRWA can perform significantly better than its static counterparts. We then quantify the throughput gain due to link sharing in multicasting. We also present an analytical model for estimating the blocking performance of the static MCRWA algorithms. The performance improvement with multifiber networks and with wavelength conversion are also investigated.

All-optical WDM networks are fast becoming the natural choice for future backbone. In this paper, we establish the efficiency of multicasting over unicasting in all-optical WDM networks, assess the usefulness of wavelength conversion for multicasting, and explore the issues related to the splitting (or copying) capability of the nodes. The comparison between multicasting and unicasting is based on the number of wavelengths as well as the amount of bandwidth required for a given set of multicasting sessions. For each multicasting session, a source-specific multicasting forest (or trees) is constructed first, taking into account the sparse splitting capability of the nodes in the network. Then, each multicasting tree is partitioned into segments according to the sparse wavelength conversion capability of the nodes on the tree such that each segment needs to be assigned the same wavelength. Simulation results obtained for a practical network such as NSFNET and randomly generated networks show that multicasting can reduce both the bandwidth consumed and the number of wavelengths required by as much as 50% or more when the size (i.e. the number of destinations) of each multicasting session is reasonably large. Such a reduction due to multicasting is not affected much by the wavelength conversion capability, the number of multicasting sessions and the size of the networks whose topology is more or less random. The results have also shown that sparse splitting can be nearly as effective as full splitting for multicasting.

In high-speed SONET/WDM rings, the dominant system cost may be the cost of electronic components, or in other words, the cost of SONET ADMs (S-ADMs). The number of S-ADMs can be dramatically reduced by using wavelength ADMs (WADMs) and grooming traffic (i.e., multiplexing lower rate signals) appropriately. In this paper, we proposed optimal or near- optimal algorithms for traffic grooming and wavelength assignment to reduce both the number of wavelengths and the number of S-ADMs. The algorithms proposed are generic in that they can be applied to both unidirectional and bidirectional rings with an arbitrary number of nodes under both uniform and non-uniform traffic with an arbitrary grooming (multiplexing) factor. Lower bounds on the number of wavelengths and S-ADMs required for a given traffic pattern are derived, and used to determine the optimality of the proposed algorithms. Our study shows that these lower bounds can be closely approached in most cases or even achieved in some cases using the proposed algorithms. In addition, even when using the minimum number of wavelengths, the savings in S-ADMs due to traffic grooming (and the use of WADMs) are significant, especially for large networks.

Fiber optic capabilities can be exploited to greatly simplify communication networks while at the same time greatly expanding their capabilities. With fiber optic technology, it becomes feasible to construct a broadband integrated services digital network (B-ISDN) that is scaled to cover the United States and can provide an STM-1 (155 Mb/s) interface for each subscriber. This paper describes an all-optical backbone for this national B-ISDN network. A switching technique based on spatial grating and shifting of optical beams is presented. With this technique, signals are directed along optically transparent paths through the backbone switches, and signal characteristics remain unaltered as signals flow through the backbone.

This paper describes the development of an optical packet transport network, known as WAvelength Switched Photonic NETwork (WASPNET)--a collaboration between Strathclyde, Essex and Bristol Universities as well as BT, Fujitsu and GPT. One of its main objectives is to reduce packet contention at each node. Normally, this is resolved using node deflection routing or optical delay loops (i.e. the solution is focused at the node design strategy). However in WASPNET, this problem is considered not only as a node design problem but also as a network control and management issue. Although suitable node design can reduce packet loss performance, an appropriate network control can reduce the probability of contentions, hence, improve the network throughput and node cost. This suggests that the network management strategy also influences the node design. A possible network control methodology, the SCattered- Wavelength-Path (SCWP), has been identified to support WASPNET implementation. The paper presents some of the comparison studies that were carried out. These include comparing its limitations, control complexity, packet loss performance and buffer requirements against another technique--the Shared-Wavelength-Path. It highlights solutions to problems encountered by the SCWP. Although the studies performed were intended for WASPNET transport system, the findings are invaluable for those involved in WDM network design.

Designing optical communication networks that support large numbers of processors, while maintaining acceptable performance under traffic patterns relevant to parallel- distributed computing, remains an open problem in optical networks research. Central obstacles to its solution appears to hinge on the inadequacy of the standard design methods to overcome the current limitations on the number of available wavelength channels per fiber and the cost and the technologically difficulty of fully-optical switching of wavelength channels. This paper proposes a WDM optical network architecture design within the limits of current technology. The design goals are to support a large number of nodes and to parametrically configure the network to perform with acceptable packet loss under a variety of offered traffic patterns consistent with communication primitives often used in parallel-distributed computing. Scalability problems due to the limitations of multiplexing in a single dimension (e.g., space, time, and wavelength) are alleviated by an orchestration of the communication traffic according to three design strategies: multidimensional space-time-wavelength multiplexing, spatial reuse of wavelengths, and parametrically driven network reconfiguration as collection groups each made up of processor-clusters where the number of processors per cluster is limited by the number of available wavelengths. The design does not require tunable components nor wavelength sensitive switching. A group-localized traffic model is proposed as a motivation for network design and used for its experimental evaluation.

To achieve optical transparency in Wavelength Division Multiplexing (WDM) networks of increasing size, in terms of both the number of nodes and network span, optical amplifiers are necessary to compensate for propagation and splitting losses encountered by an optical signal from any source to any destination. A recent study showed that in small and medium size broadcast-and-select star/tree networks the necessary number of amplifiers can be minimized by allowing wavelengths to operate at different power levels. Finding this minimum number of amplifiers and their locations requires solving a Mixed Integer Non-Linear Problem. This paper presents a generalization of the above problem, called Generalized optimal Placement of optical Amplifiers in broadcast-and-select WDM networks, or GPA problem. In the GPA problem the number of optical amplifiers is minimized taking into account that the fiber layout may include star/tree and ring topologies, and the installation and maintenance costs of the amplifier may depend on its location, i.e., solutions with equal number of amplifiers may not be equally appealing. The GPA problem is solved using a Simulated Annealing (SA) approach whose flexibility makes it an ideal heuristic to cope with the various fiber layouts and the location dependent cost issue. In addition, the SA approach offers the advantage of escaping local minima of the cost function, thus providing satisfactory solutions independent of the chosen initial solution. The optimal solution of the GPA problem provides a versatile technique to design cost effective WDM networks based on various fiber layouts, taking into account other practical constraints, such as the installation and management costs of the necessary optical amplifiers.

This paper presents an approach that introduces the formal Specification and Description Language (SDL) in the development of TMN systems. Using the management of optical networks as an example it will be outlined how SDL can be used to ease implementation and evaluation of management systems. Owing to the formal nature of SDL system behavior can be specified thoroughly and prototypes can be generated semiautomatically. Applying this approach, dynamic behavior can be evaluated by integrating the generated prototype in an SDL test and simulation environment. In this environment trace tools support description of management interactions in Message Sequence Charts at runtime. Such traces permit to draw conclusions concerning the correctness and proper design of information models. The described methodology is being applied to the development of prototype management applications in the research project MOON. It will be described how existing draft information models (e.g. information models for the management of optical networks) can be evaluated by mapping them to SDL.

With the high interest of network operators and manufacturers for wavelength division multiplexing (WDM) networking technology, the need for management systems adapted to this new technology keeps increasing. We investigated this topic and produced outputs through the specification of the functional architecture, network layered model, and through the development of new, TMN- based, information models for the management of optical networks and network elements. Based on these first outputs, defects in each layer together with parameters for performance management/monitoring have been identified for each type of optical network element, and each atomic function describing the element, including functions for both the transport of payload signals and of overhead information. The list of probable causes has been established for the identified defects. A second aspect consists in the definition of network-level parameters, if such photonic technology-related parameters are to be considered at this level. It is our conviction that some parameters can be taken into account at the network level for performance management, based on physical measurements within the network. Some parameters could possibly be used as criteria for configuration management, in the route calculation processes, including protection. The outputs of these specification activities are taken into account in the development of a manageable WDM network prototype which will be used as a test platform to demonstrate configuration, fault, protection and performance management in a real network, in the scope of the ACTS-MEPHISTO project. This network prototype will also be used in a larger size experiment in the context of the ACTS-PELICAN field trial (Pan-European Lightwave Core and Access Network).

Optical packet switches offer high speed, fine granularity, flexibility and transparency to data rate and format. There has been much work on the design of optical packet switches each having distinct advantages and disadvantages. Nevertheless, their common limitation is optical splitting loss, which is compensated by optical amplifiers, further degrading performance because of the induced amplifier noise. Hence, it is desirable to design an optical packet switch with a low optical splitting loss. This study has focused on the ALCATEL broadcast-and-select switch, which has significant optical splitting and combining losses for large switches. Arrayed-waveguide gratings (AWG) have been chosen to reduce the switch splitting loss replacing the demultiplexers and Semiconductor Optical Amplifier gates (SOA gates) in the ALCATEL switch. The switch still has the same functionality with an AWG which can be used as an interconnect, and has been demonstrated with insignificant crosstalk of approximately 30 dB. In this paper, three optical packet switches using AWGs are studied; the broadcast-and-select switch, the feed-forward delay switch and the feed-back delay switch. An additional novel feature is their use of wavelength division multiplexed inputs and outputs. Here, their optical performance is investigated with respect to bit error rate and power penalty, and compared with the ALCATEL broadcast-and-select switch using SOA gates.

The future telecommunication network will have to face the dramatic increase of subscribers as well as the increase of the user bandwidth through new services. All-optical packet switching techniques can become a strategic objective to offer on an unique technology a service-transparent network. In this paper, we will describe in detail the structure of an optical packet switching node developed in the framework of the ACTS 043 KEOPS project. An analysis of the key functions will be reported to fulfill system requirements including cascadability. In particular the input synchronization, the Broadcast-and-select switching matrix and the output regenerative interface will be described and physical performance will be assessed through theoretical analysis: quality of the signal, packet jitter and packet power fluctuation. The electronic circuitry for the control of the components of each sub-block will be described. Finally, experimental validations of a 160 Gbit/s throughput node will be reported. In order to complete the analysis, the logical performance in a Bernoulli-type traffic will be regarded. In particular an optimized buffer including a recirculation loop will be studied. Logical performance exhibiting a packet loss rate lower than 10^-9 for a 0.8 load and mean packet delay as low as 3 packet slots will be illustrated, thereby demonstrating full compatibility with ATM constraints. Finally, new perspectives in terms of throughput potential through cascading will be drawn.

Imperfect isolation of switching elements inside optical space switches gives rise to leakage signals that originates multisource crosstalk. The effects of this crosstalk are examined as a function of channel frequency spacing in an 8 X 8 Benes architecture. The statistics of this disturbance are also discussed and power penalties evaluated. In particular, we have found that the Gaussian approximation is not appropriated for this switch size, when the switch crosstalk is larger than -35 dB. Our results also show that, if input signals are all of nominally the same wavelength, crosstalk levels less than -32 dB are required to achieve a power penalty below 1 dB. However, when the channel frequency spacing increases power penalty decreases and in the limit, when the interferometric noise falls completely outside the receiver electrical bandwidth the multisource crosstalk becomes negligible.

Effect of nonlinear polarization rotation in birefringent optical fiber was used to distinguish the different levels of incident optical beams. This effect was employed to design all-optical switching and routing devices for processing of bit word streams and optical polarization transistor to amplify optical pulses. These devices can be used to perform summation and multiplication of binary words. Serial and parallel representation of the bit words discussed from the point of view bit flow architecture of optical computers. A method of time distributed parallel representation of the bit words is proposed. It is shown that this representation permits to obtain such form of the result that the beginning of the result may be used in further computations before the previous computation is completed. Thus, in spite of the fact that the whole process of multiplication of two N-bite words required 2N time steps of computation, the delay between receiving the beginning of the input words and sending the beginning of the result to the next processor is only one time step. The performance of all-optical computer may be improved by the use of extremely short femtosecond optical pulses. In this case, nonlinear effects can compensate the dispersion broadening of the optical pulses and soliton regime of propagation is achieved. Nonlinear interaction between the optical solitons with orthogonal polarization permits to perform high- performance computations.

Planar lightwave circuits (PLCs) of all-optical multi-layer switches are presented where the cycle structure of the spatial switches is modeled by rearrangeable nonblocking pruned rectangular cellular arrays (PRCAs). The chosen network model is the coupled PRCA. However, refined modeling of the proposed M X N-gon prism switches causes, beside coupling, additionally, the embedding of small PRCAs into larger ones and their intersection. Dependent on the implementation of PLCs of PRCAs (ordinary, reversed and combinations) the (1) moving location of switches is needed and the (2) multiple use of directional couplers arises, respectively. The PLCs of wide-sense nonblocking dilated multi-layer switches with simple switch setting are also presented.

Recent developments in high density Wave Division Multiplexing fiber systems allows for the deployment of a dedicated optical Internet network for large volume backbone pipes that does not require an underlying multi-service SONET/SDH and ATM transport protocol. Some intrinsic characteristics of Internet traffic such as its self similar nature, server bound congestion, routing and data asymmetry allow for highly optimized traffic engineered networks using individual wavelengths. By transmitting GigaBit Ethernet or SONET/SDH frames natively over WDM wavelengths that directly interconnect high performance routers the original concept of the Internet as an intrinsically survivable datagram network is possible. Traffic engineering, restoral, protection and bandwidth management of the network must now be carried out at the IP layer and so new routing or switching protocols such as MPLS that allow for uni- directional paths with fast restoral and protection at the IP layer become essential for a reliable production network. The deployment of high density WDM municipal and campus networks also gives carriers and ISPs the flexibility to offer customers as integrated and seamless set of optical Internet services.

One of the objectives of the European ACTS 043 KEOPS project, was to assess the feasibility of a high capacity all-optical packet switching network to face the dramatic increase of traffic needs. The initial objective was to cascade a maximum of 16 network sections (involving transmission links up to 100 km and one optical packet switching node) at 10 Gbit/s to validate the concept. In this paper we present both the experimental validation and a logical analysis. The physical performance has been assessed through a loop cascade of 40 network sections including 160 Gbit/s throughput switching nodes and 100 km of transmission. Recent experimental results have shown that such a network could be extended to a world scale. The limits of operation have been checked by regarding interferometric noise influence in the cascade and evolution of power discrepancies through the network. For the first time, these results really indicate that it is possible to provide high capacity, full flexibility and total expandability at the network level without any opto- electronic conversion. Finally, we will give simulation results exhibiting the packet loss rate, the packet delay and the occupation rate in the buffer. In particular, we demonstrate that the packet loss rate was preserved during the cascade. Results are comparable with ATM constraints as well as with other data transmission formats. This set of results demonstrates the feasibility of an all-optical packet switched network while providing both high quality of signal and high traffic performance.

A novel synchronization architecture is presented for packet based optical networks, exhibiting low insertion loss and reduced crosstalk, taking advantage of the characteristics of AWG (arrayed waveguide grating) filters. Thus far, only logarithmic delay lines have been investigated rigorously, the generic structures either based on 2 X 2 switches or on a 1-to-m splitter combined with semiconductor optical amplifier gates. The first scheme introduces not only high loss but also crosstalk, producing a high amount of interferometric noise. The latter has a lot of splitting loss to accommodate but overall, fewer stages are necessary to achieve the same delay as more than two paths per stage can be set up. In this paper, AWGs in combination with wavelength converters replace the splitter/SOA-gate geometry minimizing the optical loss, ensuring that higher levels of optical power remain to traverse the adjacent switching matrix. This novel set-up ensures that only low levels of amplification are needed minimizing concomitant noise accumulation. An incoming cell or data stream will be converted to a distinct wavelength determined by an evaluation circuit in the electronic domain. The chosen wavelength maps the input to a length of fiber which in turn represents the necessary delay, effectively executing path length (and hence time alignment) equalization of different incoming packets. Once the wavelength conversion is executed the cell is fed into an AWG, governed by a `hardwired' translation-table (input/output), ensuring cells propagate to the correct output. Finer delays are realized by cascading the principle stage.

The increased data traffic experienced today and the projected increase in the data traffic in the future demand exploration of novel approaches to IP transport such as transport of IP traffic over optics. The bimodal nature of the IP traffic short packets which are typical of transactional-style flows and large packets or bursts which are encountered in the transport of large data blocks requires, design of routers that are capable of routing packets with variable lengths efficiently. In this paper, we discuss the design aspects of such all-optical IP-switches. The broadcast and select architecture is a prime candidate for implementing optical IP routers. Construction of optical routers with buffering, wavelength conversion and multipath routing are considered. The merits and demerits of all these cases and the effect of buffer size, wavelength conversion and multiple-path routing on the blocking probability and probability of packet loss are discussed.

Optical burst switching (OBS) is a new paradigm proposed to efficiently support the ever-growing broadband multimedia traffic either directly or indirectly (e.g., via IP) over all optical WDM networks. In this paper, we propose a new prioritized OBS protocol based on Just-Enough-Time which can provide Quality of Service in buffer-less WDM optical networks. Specifically, we apply OBS to support two traffic classes: real-time and non-real-time, such that each burst belonging to the former is assigned a higher priority by simply using an additional offset time between the burst and its corresponding control (set-up) packet. We analyze the lower and upper bounds on the blocking probability of each traffic class, and evaluate the performance of the proposed prioritized OBS protocol through analysis and simulation. We show that real-time traffic can achieve a significantly reduced blocking probability by using a reasonable amount of additional offset time. In addition, the overall blocking probability and throughput can be maintained regardless of the additional offset time used.

A single failure in a communication network may trigger many alarms. When the communication network uses optical fibers as a transmission medium and increases its capacity by using Wavelength Division Multiplexing and Space Division Multiplexing, the number of alarms and the difficulty to locate the failure are considerably higher. In this case, a single failure may interrupt several channels so that the quality of lost information is larger. We propose an alarm filtering algorithm for the fault management of an optical network that supports multiple failures and works in the presence of passive elements, that is, network elements which may fail but never generate an alarm (e.g. optical fibers). Our algorithm avoids the use of failure probabilities because they are difficult to estimate and it does not need a global knowledge of the network topology. Moreover it also tolerates alarm losses. The algorithm results in the presentation to the human manager of a list of faults which may have caused the observed alarms. The alarm filtering problem is defined and formalized at the level of abstraction for devising the Alarm Filtering Algorithm. The algorithm is then applied to the optical network of the ACTS COBNET project.

Fault-tolerant routing for WDM routed all-optical networks has not been studied in details in the literature except for single fault occurrence in a simple topology, like a ring. Conventional routing schemes, being static in nature, are difficult to be modified to handle faults as fault occurrences need dynamic rerouting. In this paper, we present a fault-tolerant routing scheme for WDM routed all- optical networks. This paper considers modifications to a dynamic routing scheme to handle rerouting in case of fault occurrence. The performance degradations of the network because of fault occurrence are then studied in terms of blocking probability and set-up time through simulations.

To enable the SuperPON systems developed within the ACTS- PLANET (Photonic Local Access Network) project to meet the availability targets of the offered services redundant elements are provided to give protection in the event of failures. Using extensive reliability data a semiconductor optical amplifier based SuperPON architecture is evaluated in terms of availability. An amount of redundancy (mainly of optical amplifiers and of cable plant) is required to provide the availability for the system to support telephony in addition to the broadband service basket. Cable cut is seen to be a dominant factor. The telephony availability requirement can be achieved for the Fiber To The Home scenario with redundancy most importantly in the first distribution section and in the amplified splitter. The possible use of dual homing to counter catastrophic failures and the issues surrounding protection switching are also discussed.

Developing an efficient restoration strategy assumes a greater significance in a WDM backbone than a conventional WAN backbone due to the fact that a single link carries an enormous amount of bandwidth. This paper provides a summary of results of our spare capacity planning and restoration schemes on Pan-American and Pan-European networks. The spare capacity planning for both path and link based restoration are considered and their costs compared. Distributed and centralized restoration schemes, which have no preplanning and no a priori knowledge of the network capacities, are simulated on these networks for single link failures. Both link and path restorations are considered. We compare the restoration times of the various scenarios (centralized- link, distributed-link, centralized-path, distributed-path restorations) that arise in these two networks and show that our algorithms were able to achieve 100% restoration under a second for the networks considered. Our results show that centralized restoration can be faster than distributed restoration. This is against the common belief that distributed restoration is faster than the centralized restoration. We also analyze why this happens and the conditions (traffic matrix) under which the distributed restoration may be faster than the centralized. Additionally, we show that a centralized link restoration is always faster than a distributed link restoration.

We present a new survivable, reconfigurable wavelength division multiplexing (WDM) ring architecture for the metropolitan area or regional networks. It uses bi- directional WDM and space division multiplexing with simple switches for 1 + 1 protection at optical channel layer. The ring can be realized with available technology at relatively low cost. This architecture has advantages in scalability and reconfigurability. Achievable ring sizes are studied, and adaptability to varying traffic needs, utilizing multi-hop connections, is evaluated.