There have been significant improvements in the performance of fiber-optic links over the last four years. We concentrate here on the improvements that DFB lasers have made to directly modulated links and that non-quadrature modulator bias brings to externally modulated links.

In this paper, we theoretically and experimentally investigate the spurious-free dynamic range of an externally modulated coherent AM optical link. The coherent link's performance is also compared to a similar direct detection link. Semiconductor lasers are used as the optical transmitter and local oscillator lasers. However, the wide linewidth of semiconductor lasers can cause substantial performance degradation of coherent links. We show that by using a coherent receiver structure consisting of a wideband filter, rectifier and narrowband lowpass filter, and by selecting a wide enough bandpass filter, the coherent AM link can be made linewidth-insensitive. We also show that for low optical powers, the coherent AM link has a better spurious-free dynamic range than a similar direct detection link, and that laser RIN is not a significant effect.

We propose and analyze a new analog optical link which we refer to as the heterodyne interferometric phase modulated (HIPM) link. The HIPM link phase modulation uses a novel three leg external modulator, has high linearity and suppresses laser relative intensity noise. The link increases the spurious-free dynamic range by 21 dB as compared to a conventional externally modulated AM link for 10 mW of received optical power, relative intensity noise of -130 dB/Hz, and a 15 GHz intermediate frequency.

We report results of an improved hybrid optical transmitter suitable for use in microwave communication systems. Based on a double up-conversion technique, a 1.3 micrometers wavelength semiconductor diode laser is actively mode-locked, and its output is externally modulated by a lithium niobate Mach-Zehnder amplitude modulator to generate a transmitter output from 19 GHz to 21 GHz. Following a brief review of recent table top system measurements and motivation, we show a second generation engineering scheme for packaging the mode-locked source components. The initial performance results of this source are -34.2 dBm(e) mode-locked output power at 13.1 GHz, with an input rf power of 14.5 dBm(e); RIN is measured at -106.5 dBc(e)/Hz.

An organic dye (Rhodamine B) doped polymer optical fiber amplifier (POFA) of the graded- index (GI) type was successfully prepared for the first time. The GI-POFA of only 500 mm in length gave 27 dB in gain at 591 nm of signal wavelength. Additionally, absorption cross section and emission cross section of Rhodamine B in PMMA matrix were estimated, which were required to analyze amplification mechanism in the POFA.

Recently, noise phenomenon, so called Gordon-Haus effect, caused by nonlinearity of fiber and spontaneous emission from optical amplifier has received much attention. In general, it is very difficult to eliminate such an effect, because the Kerr effect of fibers and the spontaneous emission noise of a conventional optical amplifier are inherent in those physics. In this paper, we show that the Gordon-Haus effect can be perfectly eliminated by applying the squeezer which is studied in the research field of squeezed state.

A compact packaged receiver containing an optical pre-amplifier has been assembled and tested as a component for multi-wavelength applications. The increase in the number of channels that can be supported in a broadcast star network using such a component is estimated from experimental results, and directions for further improvements are proposed.

An analysis is presented for an optical fiber n-ary pulse position modulation (PPM) system employing optical preamplification. Calculations, performed at a bit-rate of 622 Mbit/s and a wavelength of 1.53 micrometers , demonstrate that the incorporation of an optical preamplifier leads to a sensitivity benefit of 13.3 dB compared to the non-preamplified case and a sensitivity improvement of 7.5 dB over an equivalent PCM system. Further, the work also shows that sensitivities approaching that of homodyne PSK PCM are potentially achievable.

The effect of frequency-guiding filters on solitons collision in a wavelength division multiplexing system with lumped amplifiers and periodically varying dispersion is investigated. The evolution equations for the soliton frequency shift and time shift are derived. The numerical results indicate that the filters make the frequency shift resulting from solitons collision tend toward zero, thus improving system performance considerably.

In this paper we overview a generalized packet switching solution termed SDL (switched delay lines), which provides end-to-end `optical' bandwidth by maintaining the transmitted information in the optical domain. Compared to circuit switching, packet switching supports flexible sharing of bandwidth among a large number of users and diverse applications. To implement packet switching users need to be able to share various optical resources, such as wavelengths, switches, receivers and transmitters on packet transmission time scale. SDL provides these features, leading to optical packet switching, without the need for O/E and E/O conversions, or electronic buffering and processing of the data packets. It thus allows optical packet switched networks to be realized without the need to wait for the technological realization of optical processing devices to replace the relatively slow nodal electronic processing in existing networks. The design and operation of SDL are presented and demonstrated for a MAN system.

Regular meshed topologies are very attractive for the implementation of large all-optical self- routing communication networks (or interconnection systems) that provide end-to-end transparent channels to all source-destination pairs in the optical domain, and can operate with either packet or circuit switching in a WDM environment. In this paper we study a regular meshed topology obtained with the superposition of several identical de Bruijn graphs, and compare its characteristics with those of other proposed regular meshes, considering their relative merits in terms of the maximum and average distance between source and destination, the degree of connectivity for each node, and the minimum number of wavelengths in the WDM comb necessary to discriminate all source-destination pairs.

In this paper, we propose Comb-Net, a tree PON architecture in which several laser combs at the root of the tree supply wavelengths to all stations in the PON. Namely, each station receives several wavelengths from comb laser sources located at the root of the PON, via a Comb Distribution network. It then selects (by fixed or tunable filters) some predefined or agreed upon wavelengths, and uses these as sources; i.e., it modulates them (in amplitude or phase) using an external modulator. The modulated wavelengths are then transmitted on the Multiaccess Communications network, which is separate from the Comb Distribution network. The novelty of Comb-Net is to replace the individual (fixed or tunable) lasers at the station with a set of centralized comb generators. The advantages are better stability control, lower noise and potentially lower cost since each station does not require a dedicated source. In the paper, we describe an example of Comb-Net architecture in which the Multiaccess Data network is based on a physical tree topology and on two separate virtual topologies embedded within the physical topology using WDM: a single hop topology used for circuit switched connections; and a multihop topology used for packet switched traffic. We illustrate the Comb-Net architecture with a simple case study.

Tunable-channel multi-access (TCMA) networks are a new class of multi-channel networks that employ local channel tuning to reduce the node complexity. These networks are hybrids of switching networks and multiaccess networks and are vary suitable for high-speed multimedia integrated networking. Various distributed network protocols for these TCMA networks based on star, bus and ring topologies are discussed, including two particularly promising protocols, ACTA and EQEB. Both are based on bus/ring topologies, compatible to ATM, have simple design, high throughput, low delay and a performance that is independent of the round-trip delay. Various multiplexing strategies (space-division, wavelength-division, time-division and subcarrier) employing regenerative and non-regenerative implementations are also discussed.

Selfrouting in multistage interconnection networks (MINs) with nearest-neighbor (NN) interstage interconnects is applied by mapping the system onto an equivalent shuffle net. For this purpose, additional interconnects between the switches are introduced which provide a 1:1 mapping of shuffle MINs onto NN-MINs (and vice versa). Both homogeneous and inhomogeneous lay-outs of the interconnection system are applied. Examples are presented and discussed for the interconnection of 8 X 8 and 16 X 16 data arrays, respectively.

The use of optical processing to perform switch-routing functions in photonic packet-switching networks prevents electronic data-flow bottlenecks and permits real-time routing of packets at very high speed. The architecture of a 2 X 2 photonic packet-switching node with optical packet generation, optical packet synchronization, optically-processed routing control, contention resolution (using deflection routing), and optical packet demultiplexing is presented. Terahertz optical asymmetric demultiplexers (TOADs), which allow ultrafast optical demultiplexing for terabit per second pulses with less than one picojoule switching energy, are utilized in the switch architecture. A two-channel optical demultiplexing experiment using the TOAD is reported, where two 2 ps, 100 Mbit/s pulse trains are extracted successfully from two time-multiplexed channels with a channel spacing of 40 ps.

STARNET is an optical broadband local area network based on a physical passive star topology intended for backbone applications. Over a single physical network, STARNET offers all users both a moderate-speed packet switched subnetwork and a reconfigurable high- speed (up to 2.5 Gbps) Wavelength Division Multiplexed (WDM) circuit switched subnetwork. Based on these two data transport facilities, several topological and protocol solutions are available to the users. As a result, STARNET supports traffic of widely different speed and continuity characteristics. This paper describes an ongoing experimental effort aimed at the realization of a STARNET prototype network at the Optical Communication Research Laboratory of Stanford University. The prototype nodes are FDDI data rate and format compatible in the packet switched subnetwork and currently use 8B/10B encoded data at a rate of 1.25 Gbps in the circuit switched subnetwork. The STARNET network serves high performance workstations running distributed multimedia applications, including video- conferencing. In this paper, the electronic hardware which interfaces the workstation to the STARNET optics is described. The moderate-speed interface uses an off-the-shelf FDDI-on- copper module to interface to the STARNET optics; a custom high-speed interface module provides fast packet switching (electronic) over the circuit switched (wavelength) subnetwork.

To provide very high end-to-end bandwidth, all-optical networks need to maintain the transmitted information in the optical domain throughout the network. Through packet switching, this bandwidth can then be flexibly shared among a large number of users according to the demands of their applications. In this process users need to share the various optical communication-handling network resources, such as wavelengths, receivers and transmitters. Extant packet switching solutions, based on electronic buffering and processing, are not appropriate for the optical environment, as, due to the relatively slower electronic rates, their use leads to the `electronic performance bottleneck.' In this paper we present an `all-optical' ring network architecture, PIPELINE, which exploits both WDM and subcarrier signalling techniques. WDM is used to provide separate channels (one for each node) to exploit the fiber bandwidth, but keeping the node interface at electronic rates. Subcarrier signalling allows for outband packet signalling which does not require the payload O/E conversion at each intermediate node. A simple access protocol to avoid receiver contention is also proposed and studied to determine its fairness. Physical implementation issues and network scalability, i.e., maximum number of nodes, are carefully evaluated to estimate the ultimate potential of this and similar network structures.

Multiple Channel Architecture (MCA) computer systems need high speed optical communication channels that quickly transfer control between multiple masters. Many optical channel arbitration methods have been studied for use in local area networks including the ALOHA, Code Division Multiple Access (CSMA), and Carrier Sense Multiple Access with Collision Detection (CSMA-CD) protocols. Described are the properties required for MCA arbitration methods, details of the CSMA-CD protocol used in this study, and recent results showing message traffic histograms of a MCA computer simulation utilizing CSMA-CD arbitration and a transmission rate of 2 Gbit/sec.

We propose an interchannel parallel coding scheme in the wavelength domain -- the WDM coding system. The system differs from usual serial coding systems and provides many advantages. First, data channels are completely unaltered in the coding process, rendering it very suitable for practical lightwave systems with standard bit rate. Secondly, parallel encoding/decoding are simpler than those of serial coding systems, being easier to be implemented in high-speed optical systems. Thirdly, comparing to serial coding, WDM coding is able to reduce heavily the number of encoding/decoding pairs: an example Hamming coded WDM system reduces the number from 12 X 11 equals 132 to 1 at the line rate of STS-12. Fourthly, WDM coding system could offer infinite coding gain in dispersion-limited lightwave systems. Finally, WDM coding system could correct single channel burst error.

Previously, we determined fundamental performance limitations associated with `all-optical' packet switches, in which the packet buffering is implemented via fiber delay lines. In this work, we propose and analyze an optical packet (ATM) switch architecture that comes close to achieving the optimal performance (i.e., best possible delay-throughput performance and minimal possible buffer requirements) of a random-access, shared-memory design. The proposed Shared-Memory Optical Packet (SMOP) Switch buffers packets in recirculation delay lines of appropriately-selected lengths, and uses a novel control algorithm that: (i) keeps packets in their proper first-in, first-out sequence, (ii) supports multiple levels of priority traffic, (iii) minimizes the needed number of recirculation loops (which reduces the size of the switch fabric), and (iv) ensures that packets pass through the recirculation delay lines only a small number of times (e.g., less than 10).

While fiber optic packet switched networks can make use of the advantages provided by optics for transmission of messages, it is not possible to provide an efficient storage mechanism at intermediate nodes for data that remains in optical format. Hence, it is necessary to avoid conversion to electronic format for temporary storage at intermediate nodes when they face output port contention. As a result, in order to maintain fast packet switching, optical packets must be continuously moved until they reach their destination. In such bufferless networks, if two or more packets coming into a node simultaneously request the same output port, one of the requests can be satisfied and the others must be sent out through other available ports. The addition of time switching, to the usual space switching using this deflection routing technique, allows temporal reordering of packets so as to reduce the contention for spatial channels. In this paper we present the performance enhancement due to space-time switching in bufferless packet switched networks.

Ultra-high-speed optical communication networks will require routing switches which support data rates beyond the capabilities of optical-electronic conversion. This paper describes the preliminary design of an `all-optical' routing switch for a network in which data messages remain in optical form throughout the transmission from source to destination. The basic switch architecture and fundamental design trade-offs affecting switch implementation are discussed. Characteristics of the optical device technology required to implement the routing switch elements are also presented.

This paper compares two optical switching architectures: the feed-forward and the recirculation, based on practical considerations, such as the size of the optical switching fabric. This work was stimulated by several papers in the technical literature claiming to address the above comparison. However, erroneously, these papers in their comparison rely on metrics borrowed from the `electronic' domain. Because the optical and the electronic domains differ so much, the conclusions presented in these papers are meaningless, at best. In this paper, we also discuss some additional properties of the feed-forward architectures, making them so much more attractive for wide-area all-optical connectivity.

Long distance data transmission using solitons multiplexed on different wavelengths makes more efficient use of fiber bandwidth than transmission on a single wavelength channel. However, perturbations and nonlinear distortions limit the number of wavelengths which can be multiplexed and detected at the end of the fiber. Perturbations, such as loss, cause permanent frequency shifts if a collision occurs between solitons widely separated in frequency. Densely packing the solitons spectrally, though, results in distortions in spectral intensity which limit the use of standard wavelength demultiplexing techniques. We examine methods by which solitons, densely multiplexed in wavelength, may still be detected even during collisions. The theoretical feasibility of encoding the data on the eigenvalues of the linear evolution equations associated with soliton propagation by the inverse scattering transform is discussed, as are more practical techniques using only the spectral intensity of the waveform.

Techniques for design and implementation of passive broadcast star couplers using holographic devices have been reported. These devices whose refractive index varies in space, also have the capability of multiplexing and demultiplexing optical signals in WDM systems. In this paper, we investigate the wavelength sensitivity of such devices for optical filtering. Specifically, we examine their power transmittance function. We then present the cross-talk performance in multi-channel WDM-OOK systems. The filter performance is compared with that of a Fabry-Perot filter.

A reflection grating wavelength demultiplexer fabricated on planar silica-on-silicon is investigated for use both as a multi-channel demultiplexer and as part of a fast single- wavelength selector device. Design and characterization of a preliminary grating device which resolves 128 channels with 0.85 nm channel spacing are presented in this paper.

Recently, subcarrier multiplexing method combined with the conventional microwave techniques provides high capacity multichannel optical fiber communication. But as the microwave frequency gets higher, the carrier generation and carrier recovery circuits are more difficult to implement by using traditional microwave technology. This paper proposes a way to generate many subcarriers using only a single reference tone by deep phase modulation method. Tunabilities of operating band and carrier spacing are presented in experimental results. It is found that the optimum phase modulation index is 3.1 for minimum power deviation to generate 7 carriers. Those generated subcarriers can operate from dc to K-band that can be used in transmitting end or receiving end.

An optical crossbar switch using semiconductor optical amplifiers (SOAs) as switching elements is presented. The switch is constructed using the strictly non-blocking matrix-vector multiplier (MVM) architecture and is capable of interconnecting 8 single mode fiber optic inputs and outputs. Successful operation of the switch has been demonstrated with the High Performance Parallel Interface (HIPPI) protocol and 1.1 Gb/s digital data streams as well as 1.4 GHz analog video signals from a satellite downlink.

A cadmium Sulphide doped fiber is used to design an all-optical Kerr shutter arranged as a standard Mach-Zehnder interferometer. The high third order susceptibility of cadmium sulphide enabled low power, in milliwatt, operation of the device. At a pump power of 25 mW, in an interferometer of length 15 cm, the switching speed was calculated to be 100 gigabit/second for the switch.

Performance of wavelength division multiplexing (WDM) optical communication system using a F-P demodulator is analyzed. The power penalty expression resulting from LD linewidth, channel spacing, performance of F-P demodulator and crosstalk are proposed.

In this paper, synchronous fiber-optic code division multiple access (S/CDMA) networks using matrix code are presented. The algorithm for constructing the S/CDMA matrix code from the modification of the modified prime sequence code is described, and the error probability is also derived. In comparison with the modified prime sequence code, the matrix code can greatly relax the laser pulse width requirement for the S/CDMA network.

Equalizers based upon a simple length of waveguide or microstrip have been demonstrated for this application. As the required length of the equalizer is dependent on the fiber span, then for very long transmission distances these approaches tend to become rather impractical. We have explored an alternative pseudo-lumped-element approach utilizing GaAs monolithic microwave integrated circuit (MMIC) technology, providing very compact group delay equalizers which may readily be integrated with other receiver signal processing functions. We have previously published preliminary results obtained with single network sections. These provide the requisite dispersion compensation over a restricted bandwidth. Here for the first time, we explore how multi-section GaAs MMIC group delay networks may be employed to achieve dispersion compensation over a much wider frequency range, illustrating the effectiveness of the approach with reference to an 8 Gbit/s 200 km system application.

Tuned front-end optical receivers, for use in Coherent or Subcarrier Multiplex Lightwave systems, offer significant improvements in signal to noise performance over conventional baseband receivers. To design these receivers correctly a detailed knowledge of the noise performance of short gate-length GaAs MESFETs and HEMTs is required. The intrinsic noise parameters, P, R, and C completely characterize the noise performance of such devices but closed form analytic expressions are not yet available that enable accurate determination of these parameters. Here we consider a technique that extracts the intrinsic noise parameters, P, R, and C from measured noise data. Novel analytic expressions are presented as well as results that prove the validity of the technique.

Work in the area of digital pulse position modulation (digital PPM) has shown that this type of modulation can yield sensitivities that are typically 4 - 5 dB better than an equivalent PCM system. Recent experimental work has shown that the receiver in a digital PPM system does not need to have a wide bandwidth. Instead, the bandwidth can be very low so that the receiver is effectively impulsed by the digital PPM signal. The advent of very high speed Si digital ICs, and fast lasers, means that digital PPM can now be used to code gigabit PCM signals. This paper presents original theoretical results for a digital PPM system coding 1 Gbit/s PCM signals into 8 Gbit/s digital PPM signals. The paper also addresses the difficulties that the system designer is likely to encounter, and discusses some possible solutions.

In this paper the channel capacity of a fiberoptic transmission system is estimated by upper and lower bounds. It is found that there is only a very small difference between these bounds so that the channel capacity is directly determined by them. The channel capacity is compared with the data rate of a binary transmission system with on-off-keying (OOK) of the optical power. A numerical example shows that the binary OOK-system achieves only about 20 - 25% of the channel capacity. This means that with improved transmission schemes the data rate of existing optical transmission systems can be increased dramatically. The only question left unanswered, is that of which modulation scheme does yield this improvement. It can be shown that a small improvement can be achieved using multilevel pulse amplitude modulation (PAM). For multilevel pulse position modulation (PPM), no improvement is obtained.

Methods that increase the bandwidth of an optical fiber communication link beyond what is achievable with a single high speed laser and detector are of great interest. Time division multiplexing, TDM, is an attractive method of increasing overall system bandwidth, without having to increase the bandwidth of the individual transmitters and receivers. Mode-locked and gain-switched lasers are attractive sources for TDM. However, due to inadequate wavelength control in the laser fabrication process and the relatively wide spectrum resulting from mode-locking and gain-switching, fiber dispersion can degrade TDM systems. We introduce a simulation package that is used to model an experimental TDM fiber link and investigate the limits of transmission distance due to dispersion in a high speed communication link.

A new method is described which enables the propagation of arbitrary optical waveforms through monomode fiber to be studied analytically. The basis of the method is the representation of the electric field of the emitted optical waveform by a set of wavelet transform coefficients. For certain wavelet functions, a closed form solution of the wave equation can be obtained, thereby allowing an analytic description of the propagating optical waveform. Arbitrarily shaped pulses or pulse sequences, with or without frequency chirping of the source, are accommodated with ease. The coefficients of the wavelet expansion can be used to plot the time-resolved frequency spectrum of the waveform, making the method well suited to the analysis of practical laser pulses as encountered, for example, in high bit-rate optical fiber communication systems.

An improved model of an optical phase-locked loop (OPLL) is presented. Based on this model and rate equations of an external cavity semiconductor laser, the effects of OPLL feedback on the FM noise and the AM noise of the external cavity semiconductor laser are discussed. A new linewidth formula of the local laser is derived.

Timing performance of optical fiber pulse position modulation (PPM) utilizing self synchronization is considered when the main impairment is the systematic jitter associated with the extracted slot and frame synchronization. An original spectral model is developed and used to evaluate the slot timing performance in terms of both the timing variance and the probability of PPM wrong slot errors (WSE) caused by the jittered clock. A novel method for frame extraction is used and characterized in terms of both, the timing variance of the extracted clock and the error probability (WSE) caused by the frame clock jitter. The analysis explicitly illustrates the impact of the various PPM and phase lock loop (PLL) parameters on the extracted clocks and system error performance, thus allowing the identification of the optimum parameters for a given required performance.

In this paper we reported design, fabrication, and characterization of the integrated optic devices formed on glass substrate with the thermo-optical operation principle. The results of complex dynamic calculation of the thermally-induced buried channel waveguides have been reported. Switching energy level and delay time have been expressed as a result of complex analysis of integrated optic elements based on thermally-induced waveguides. The 4 X 4 multi-mode waveguide switch array has been elaborated and its characteristics are discussed.

We have measured the performance of a new demultiplexing device, known as a Terahertz Optical Asymmetric Demultiplexer, while operating it in an optical time division multiplexed system with an aggregate bandwidth of 50 Gbits/s. These measurements also illustrate the device's adjustable tolerance to jitter.