A novel geometry for all-optical clock recovery using a semiconductor amplifier yields a strong locking mechanism in a short cavity. The clock recovery laser is characterized with regard to its performance in a system environment. The clock circuit is operated at data rates from 1 to 40 GHz and exhibits a large locking bandwidth and low timing jitter.

In this paper we investigate the design and realization of broadband tee-tuned front-end optical receivers. The receivers will be implemented as a GaAs MMIC using the GEC/Marconi F20 foundry, the active devices are 0.5 micrometers gate-length GaAs MESFETs. Simple design equations are introduced that show the importance of the intrinsic noise parameters P, R and C, which are used to characterize the noise performance of such devices, as well as enabling initial values of tuning elements to be obtained. Finally measured results for a MMIC receiver designed are presented that confirm the expected noise performance as well as the validity of the design process.

The sensitivity performance of an optically preamplified PPM receiver is evaluated, including the effect of timing jitter, through the use of an improved Gaussian approximation. A potential sensitivity of 23.6 photons/bit is demonstrated in the unjittered case, for the parameters considered, with the jittered optimum being at worst only 1 dB less sensitive. These results both surpass the fundamental sensitivity limit of an on-off keyed non-return to zero (OOK NRZ) system. The results predict that wrong slot and jitter errors first become dominant at a PPM coding level higher than that suggested by an earlier model.

A possible way of realizing large optical space switches is by interconnecting multiple 2 X 2 switching elements based on integrated semiconductor optical amplifiers (SOAs). In this paper, the switch size limitations are examined, giving a special emphasis to the phase noise, originated by SOAs. Using a DPSK direct detection scheme and multiple optical filters within the switching network, it is shown that the phase noise becomes the most restrictive factor, preventing to connect more than 200 switching elements, for a bit rate in the gigabit/s range.

We report on a passively mode-locked stretched pulse APM fiber laser that generates subpicosecond pulses and is tunable over a 58 nm range. A 5 nm interference filter external to the laser provides the tunability.

We describe an all-optical network testbed deployed in the Boston metropolitan area, and some of the experimental applications running over the network. The network was developed by a consortium of AT&T Bell Laboratories, Digital Equipment Corporation, and Massachusetts Institute of Technology under a grant from ARPA. The network is an optical WDM system organized as a hierarchy consisting of local, metropolitan, and wide area nodes that support optical broadcast and routing modes. Frequencies are shared and reused to enhance network scalability. Electronic access is provided through optical terminals that support multiple services having data rates between 10 Mbps/user and 10 Gbps/user. Novel components used to implement the network include fast-tuning 1.5 micrometers distributed Bragg reflector lasers, passive wavelength routers, and broadband optical frequency converters. An overlay control network implemented at 1.3 micrometers allows reliable out-of-band control and standardized network management of all network nodes. We have created interfaces between the AON and commercially available electronic circuit-switched and packet-switched networks. We will report on network applications that can dynamically allocate optical bandwidth between electronic packet-switches based on the offered load presented by users, without requiring interfaces between users and the AON control system. We will also describe video and telemedicine applications running over the network. We have demonstrated an audio/video codec that is directly interfaced to the optical network, and is capable of transmitting high-rate digitized video signals for broadcast or videoconferencing applications. We have also demonstrated a state-of-the-art radiological workstation that uses the AON to transport 2000 X 2000 X 16 bit images from a remote image server.

A new form of pulse time modulation, termed digital pulse interval modulation (DPIM) is introduced in this paper, exhibiting improved performance and relaxed timing constraints compared to digital pulse position modulation (DPPM). Theoretical expressions are presented for transmission capacity, code characteristics and power spectral density, along with measurements from a simple experimental prototype, illustrating the advantages of the new scheme.

This paper investigates the implementation of a new digital pulse time modulation technique based on digital pulse interval and width modulation scheme. Original expressions are presented for power spectral density, code characterization and channel capacity, illustrating the advantages of this technique compared with conventional pulse code modulation. Both theoretical and practical results are given showing close agreement.

The concept of asymmetry in free space optical links is discussed. Simple equations describing the minimum carrier sense distance, the minimum and maximum distance of reliable link, and maximum interference distance are derived. Examples from the recently drafted IRDA specification for IR links are given. A solution is presented to the parity variation problem by controlling the field of view of such transceivers.

The optimum presynchronization filter for a pulse position modulation (PPM) signal is composed of a matched filter in cascade with a derivative network. This paper addresses the problem of realizing such a filter for detecting the temporal position of a pulse used in the optical fiber PPM scheme. Firstly, a matched filter is realized from a time domain specification on the received pulse shape. Secondly, a derivative network is used to shape the pulse. The effect of the derivative network on the pulse shape is investigated by varying the value of the differentiating capacitor C_d and observing changes in the timing variance and jitter wrong slot error (JWSE) probability. The paper shows that there exist an optimum value of C_d for which the pulse shape is as close as possible to the true derivative and one which offers significant improvement both in the timing variance and JWSE probability compared to that of the matched filter.

A routing all-optical switch in which the signal, consisting of a fundamental soliton, is directed into either of the output ports of a 2 X 2 nonlinear directional coupler made of two fibers is investigated by means of computer simulation. The routing is controlled by copropagating a controlling pulse along with the signal pulse. Four different schemes for doing this are explored where the polarization of the controlling pulse with respect to the signal soliton may be varied, or the controlling pulse may be launched into either of the input ports, or it may have a wavelength different from that of the signal. It is shown that only one of the schemes allows a high switching efficiency and enables distortionless propagation of the signal pulse through the coupler.

The concept of optically integrated networking is experimentally verified. The operational procedures of the integrated system and methods for signal separation and detection are explained. The degree of degradation of each traffic as a result of the spectral overlapping is analyzed and power penalties evaluated. It is verified that the communication traffic may be recovered with negligible power penalty. The optimum solution in terms of the signal to noise ratio for the sensor traffic is shown to be the whitened matched filter detection. However, this solution results in an unacceptable degree of intersymbol interference (ISI). Thus, a zero forcing ISI solution is also presented and the system performance is re-evaluated taking both the noise and ISI into account.

This paper shows the state of the art in fiber optical code-division multiple-access (CDMA). Recent work in this area for both, systems and sequences is reviewed and analyzed. For that purpose a classification of systems, corresponding to the manner of signal processing and a classification of known (0,1)-sequences are presented. It is shown that due to the limits by currently available device technology especially two techniques are promising for implementation in broadband telecommunication networks: spectral encoding with integrated optical filters and CDMA in combination with wavelength multiple access schemes. Further an overview about some important experiments in this field is given.

The aim of this work is a proof-of-concept demonstration of an optical CDMA system based on spectral amplitude encoding of incoherent broadband sources. A Light Emitting Diode is employed as a source in a multi-mode fiber-optics system, and simple pseudorandom sequences such as m-sequences are used for optical encoding. We propose a multi-pass configuration with a Littrow grating arrangement for the optical encoder and decoder. Since this is a power-limited system, the multi-pass arrangement allows us to greatly reduce the number of lossy optical components and to increase compactness and power throughput. A differential receiver with a simplified design is employed. Measurement results are presented and discussed.

A hybrid pulse position modulation code division multiple access (PPM-CDMA) scheme is proposed and studied in the context of optical fiber video transmission and distribution. With compressed HDTV video rates of 5 - 20 Mbit/s and with the current optical fiber technology, the strategy is shown to be capable of accommodating up to 50 video channels. Original results are presented for the bit rate demonstrating that there exists an optimum PPM-CDMA order that maximizes the achievable bit rate at a given number of channels. Results are also presented for the digital video channels crosstalk which takes the form of false alarm errors caused by the OOC crosscorrelations. The optimum system operating parameters that minimize these errors are specified.

In this paper we show the application of optical CDMA based on spectral encoding with integrated optical devices and discuss the use of cyclic shifted m-sequences to realize full orthogonal transmission. Degradations of the system performance due to non-ideal source spectra, crosstalk and losses in the optical coder devices are examined. Some bounds for different kinds of realization are shown. Possible sources are broadband semiconductor diodes like SLD, optical semiconductor amplifiers or multiwavelength lasers. For optical coding solutions with acoustical tunable optical filters, arrayed waveguide structures and fiber grating filters are investigated. The total network capacity will be analyzed under the above mentioned conditions and it is shown that such a network can accommodate some hundred users with a network throughput of some Gb/s.

We describe the design and performance of a wide-area all-optical network. An improved model of the benefits of wavelength changers in all-optical mesh networks is presented and comparisons are made to simulations. The model is shown to be accurate for a moderate number of wavelengths and a small blocking probability for a variety of networks.

Kautz directed graphs (digraphs) arise from larger de Bruijn digraphs by the deletion of nodes or from smaller de Bruijn digraphs by additional nodes, respectively. Thus all-optical self- routing networks with Kautz topologies have different properties compared with de Bruijn. Kautz digraphs may be extended by their sum and product graphs and contain planar (crossover-free) embeddings.

In many networks a nodes throughput or deliverable data rate does not match the nodes network access rate. This is particularly true in fiber-optic LANs where many users must share a single communications channel and a nodes network throughput is a fraction of the rate at which its transmitter may access the fiber. Even in broadband networks which offer transport rates in excess of the access rate, the throughput still may be network topology limited by congestion, queuing and other transport delays. This paper will demonstrate how an optical cross-connect can be used in the LAN environment to segment a network into new and separate broadcast domains, which improve the overall network capacity. It also demonstrates how an optical cross-connect can be used to groom traffic in a multihop network to better utilize the network's existing bandwidth capacity.

Crosstalk is a major concern in optical switching networks made of 2 X 2 electro-optical switches such as Dilated Benes Networks (DBNs). One of the important performance parameters of these optical architectures is SXR (Signal-to-Crosstalk Ratio). The paper proposes ways to automate the process of obtaining crosstalk ratios and their related characteristics. This paper contains two parts. The first part is an algorithm to obtain the crosstalk ratios of each and every switch in a DBN. These results can be used to calculate the SXR of the network or to detect and locate switches with excessive crosstalk. In the second part, we present a feasible way to configure a network to obtain an approximate value of its worst-case SXR. The difference between the actual and the approximate worst-case SXR values is almost negligible, while the complexity involved in obtaining the actual worst-case SXR can be prohibitively high.

First-order crosstalk in Multistage Interconnection Network (MINs) can be avoided by ensuring only one input of each switch is active at a given time, in other words, no two connections can use the same switch simultaneously. A space domain approach dilates an N- by-N network into one that is essentially equivalent to a 2N-by-2N network. The time domain approach proposed extends the `dilation' concept from space to time to avoid crosstalk by establishing the connections sharing the same switch in different time slots. This paper studies the tradeoffs of these two approaches in two different MINs, namely, Banyan and Benes, under both individual and stage switch control. Theoretical analysis showed that in a Banyan with stage control, the set of N-by-N permutations whose connections can be established crosstalk free in two time slots using the time domain approach is more than the set of permutations that can be realized in one time slot using the space domain approach. In addition, simulation results showed that given the same set of one-to-one or one-to-many connections, the number of time slots needed by the time-domain approach is less than twice of that needed by the space-domain approach.

The use of subcarrier multiplexing/multiple access (SCM/SCMA) is a simple and cost effective alternative to time division multiplexing/multiple access (TDM/TDMA) for creating multipoint connections in lightwave networks. The SCM/SCMA technique can be used in all- optical networks to produce a significant increase in throughput and connectivity. However, when the signals from two or more lasers operating in a WDM/SCM system are combined at a detector, optical beat interference (OBI) occurring at a frequency equal to the difference in optical frequencies of the lasers can potentially severely interfere with the information being transmitted. Thus, to achieve satisfactory performance, the optical carriers must have the correct spacing, their individual wavelengths must be precisely controlled and the subcarrier frequencies carefully selected to minimize OBI. Our experimental work focuses in studying the OBI and its effect as the optical carriers are moved closer in frequency. Our results provide the minimum spacing needed to achieve a 10^-9 Bit Error Rate at both 3 Mbps and 200 Mbps. We also discuss an interesting application for the above scheme on the Optical Network Technology Consortium's all-optical testbed.

Subcarrier multiplexing (SCM) is an attractive technique to deliver broadband services over optical fibers. The existing SCM systems employ an analogous signal for optical transmission and their performance is limited. Pulse time modulation (PTM) is an intermediate technique which offers high performance at a low cost in optical transmission systems. In this paper we present two popular PTM techniques suitable for SCM systems. Theoretical predictions and practical measurements have been used to evaluate system performance. The technique employed has improved the dynamic range and the optical power budget, compared to conventional systems.

In this paper we use small signal analysis to predict the intrinsic non-linear distortion of an injection-locked semiconductor laser diode directly driven by two subcarriers. We show that the second-order harmonic distortion may be reduced by between 12 and 27 dB over the free- running laser bandwidth of 3.4 GHz and that the third-order intermodulation distortion may be reduced by between 12 and 22 dB over a bandwidth of 5 GHz, for an injection ratio of 0.5.

We have developed a novel dispersion-reduction technique using subcarrier multiplexing which permits the transmission of multiple 2.5 Gbit/s data channels over hundreds of kilometers of conventional fiber-optic cable with negligible dispersion. Using a lithium niobate external modulator having a modulation bandwidth of 20 GHz, we are able to multiplex several high-speed data channels at a single wavelength. At the receiving end, we demultiplex the data and detect each channel using a 2-GHz bandwidth optical detector. All of the hardware in our system consists of off-the-shelf components and can be integrated to reduce the overall cost. We demonstrated our dispersion-reduction technique in a recent field trial by transmitting two 2.5 Gbit/s data channels over 90 km of commercially-installed single-mode fiber, followed by 210 km of spooled fiber. For comparison, we substituted the 300 km of fiber with equivalent optical attenuation. We also ran computer simulations to evaluate link behavior. Technical details and field trial results will be presented.

Technology that debuted in the public switched network has been moving, over recent years, into private networks. While there is no need to justify Wavelength Division Multiplexing (WDM) in the public switched network, which makes good use of higher capacity equipment, the cost effectiveness of private WDM networks is not at all clear. In this paper, we investigate of deploying WDM with Asynchronous Transfer Mode (ATM) networks. The WDM networks provide the physical layer; the ATM networks form a logical layer, using light paths from the WDM network. There are 4 different benefits to be derived from such an arrangement, including the possibility of reduced network cost. An algorithm for the design of logical networks is developed. This algorithm has been incorporated in the Intrepid network design tool.

An experimental all-optical, wavelength-routed network testbed has been constructed in the Boston metropolitan area. The network has 20 optical channels, space by 50 GHz and provides dedicated circuit-switched wide-band service at user defined modulation formats and rates up to 10 Gbps, and time-slotted WDM services for medium and low-rate users. We are now characterizing the deployed network which spans over 87 km interconnecting four all-optical local-area networks in Littleton, Lexington, and Cambridge Massachusetts. We discuss wavelength sharing and reuse, local broadcast, routing, multi-cast and multi-hop connections at 1.244, 2.488, and 10 Gbps. We present the system design and the performance (e.g. BER and cross-talk) of local-broadcast, metropolitan-area-routed and broadcast transmission modes.

This paper studies the use of wavelength translation in different Clear Channel setup strategies in a scalable all-optical network. the network considered used a recursive grid architecture representing the 100 largest cities and constituting 561 access stations. As the network grows, the grid is subdivided in a recursive fashion to accommodate more nodes, making the network modular. The algorithm used for routing of the calls is distributed, dynamic, and scalable in the sense that the complexity of the algorithm grows slower than linear with the network size. The performance of wavelength translation with static and dynamic routing strategies is studied using simulation in terms of the blocking of incoming requests for Clear Channels. Optical wavelength translation is provided in the network and it is shown that for the network considered, the improvement in performance is marginal. The Clear Channel blocking, with and without wavelength translation, is studied for different traffic patterns ranging from fully uniform traffic to a pattern exhibiting very strong communities of interest. The impact of wavelength translation is considered for these different traffic patters. The Clear Channel setup strategies along with the architecture are all scalable and hence are applicable to very large networks.

Extensive research and considerable progress in the dense optical wavelength division multiplexing (WDM) technology has made it the most likely candidate for implementation in next-generation high performance switches, LANs and MANs. The first key challenge is the continuous development of cost-effective optical components in order to realize practical networks. The next main challenge is to design an effective channel access protocol that will: (a) take full advantage of the existing and emerging optical technologies, (b) fairly and successfully coordinate transmissions between the networks nodes, and (c) efficiently manage the enormous fiber bandwidth. There have been many access protocols proposed for WDM networks including quite a few proposals that require many more channels than the number of nodes. The practical device characteristics impose an upper limit on the number of available WDM channels. Our focus in this paper is on systems with many nodes and small number of channels. We study the use of in-band signaling mechanism compared to the separate control channel (out-band signaling) approach. Signaling is used to achieve coordination between source and destination with respect the communication channel. The objective of this paper is to understand the key trade-offs in the protocol design for such networks through close examinations of the two approaches. Our intent is not to claim the superiority of either approach, rather to stimulate further studies on such systems.

Currently, the developed bi-directional broadband Passive Optical Networks (PONs) have a maximum splitting factor around 32 and a range less than 20 km. Due to the expected switching node consolidation the access network will have to cover a wider range. An access network based on a cascade of PONs is proposed. Optical transmission is supported by the introduction of optical amplifiers, offering the advantage of transparency. In this paper, the technological issues to realize a high split, wide range PON are discussed. Main focus is on the realization of a node amplifying a burst mode signal, and the reliability and upgradeability of such networks.

This paper describes an object-oriented software system that can be used to predict the spectral performance of optical networks. The system can be used by the planars and designers of such networks to assess the spectral performance of existing or proposed optical networks. The software package combines the theoretical models developed in the course of a research program with experimental data obtained from the devices used in the UK field trials of fiber in the local access network. Using this package, a planner or designer is able to construct networks on actual location maps and then obtain forward and return loss spectral characteristics on any paths throughout the network. Since the development of the system is based on object oriented technology, it is future-proofed to the extent that any changes in the technology of the local access network can readily be incorporated into the existing modelling process.

The future widescale deployment of broadband multiservice networks is dependent on an acceptable commercial relationship being achieved between new service revenues, operating cost savings and the network capital investment required. A new generation of broadband access networks is envisaged which is capable of both broadband and narrowband transmission, and can provide both broadcast and interactive services over a much larger coverage area than copper pair yet makes good use of existing copper pair infrastructure. To ensure that such networks emerge practically and are economically viable, requirements must be agreed and a wide market established. The aims of this paper are to draw attention to the benefits of networks with broadband and narrowband capability, to discuss the benefits of co- operation on requirements and to describe what has so far been achieved in the standards bodies.

This paper describes the results of an investigation into the behavior of two medium access control (MAC) protocols for the provision of asynchronous transfer mode (ATM) over passive optical networks (PONs). The two paradigms of APON provision--cell-based MACs and frame-based MACs--are investigated. The protocols have been modelled using object-oriented cell-level discrete-time simulation techniques. The investigation compares the two protocols for their cell delay variation (CDV) performance in response to varying terminal peak transmission rate, traffic model and traffic load distribution asymmetry. The protocols are also examined in terms of their effect on emergent traffic shape. Computer simulation characteristics indicate that a trade-off exists between CDV and traffic shape since the frame- based MAC is superior in CDV performance and the cell-based MAC produces superior traffic shape. Also presented is a hybrid approach combining the multiple cell-allotment strategy approach of the frame-based MAC protocol with the permit-allotment strategy of the cell-based MAC protocol. Simulation results demonstrate that mean delay performance of the hybrid is superior to both standard approaches.