An active star topology implementation of an Ethernet/IEEE 802.3 Carrier Sense Multiple Access/Collision Detect (CSMA/CD) local area network is presented. A hierarchical star-to-star interconnect topology is analyzed from a network delay standpoint. This analysis leads to an estimation of maximum network extent based on the IEEE 802.3 local network specification. Network diagnostic techniques and methods of network problem isolation are also presented. These analyses indicate that a high degree of network flexibility is achievable when an active star topology is employed. Additionally, maximum flexibility is provided by a modular central star design that supports interchangeable wire or optical fiber channels.

The widespread use of the IEEE 802.3 CSMA/CD (Ethernet) Local Area Network (LAN) has created demand for a fiber optic physical layer implementation to address security issues, hostile electromagnetic environments, modern structured wiring requirements and distance limitations of coaxial based implementations. Active Star CSMA/CD LANs will be described in this paper which consist of a central wiring Concentrator which supports point to point fiber links to Media Access Units (Transceivers) located at the Host computers. The fiber optic Active Star configured CSMA/CD LAN implementation provides a robust network which meets all the requirements imposed on an Ethernet Physical Layer. Collision detection is reliably performed in the electrical domain of the Concentrator. Network requirements included guaranteed collision detection, network reliability and easy addition and rearrangement of host connections. In addition, the Active Star implementation can provide an increased network diameter to 4.2 km and can support the four basic multimode fiber types, simultaneously, with substantial system margins.

Digital transmission for voice and data signals are already employed in most fiber-optic communications. However, due to the requirement of high bit rate for each video signal and the extremely high added cost of digital encoding and decoding, it is still not practical to transmit digital video signals. In the forseeable future, video signals are very likely going to be transmitted in analog form through fiber links. This paper describes the architecture and the design of Short Bus Integrated Service Local Area Network (SBILAN), a local area network which provides voice, data, and video services. The network topology is a star dual short-bus with optical fiber distribution to all terminals. For economic reason, data and voice signals are transmitted in digital form through the parallel packet bus and video is transmitted in analog form through a circuit-switched bus. Digital and analog signals are multiplexed on the optic fiber transmission line by using frequency division multiplexing (FDM) technique. The parallel packet bus can transmit as fast as 160 Mbps and the bandwidth of the circuit switched bus is 4.5 MHz for one video channel. The fiber data link transmits the digital signals at a speed of 10 Mbps. An effective priority channel access scheme is employed that provides fast contention resolution and is collision-free on the packet bus. Details of the network design, the fiber data link design, and the integration methodology are provided.

This paper describes the implementation of a fiber optic inter-repeater link (FOIRL), used for connecting two remote copper segments of an IEEE 802.3 local area network. The rationale for the design, the signalling used and the collision detection mechanism is discussed. The evolution of the draft international standard for the FOIRL and the concurrence amongst various manufacturers is also presented. Finally some examples of typical applications, highlighting the ease of installation, are given.

The application of optical fibers to bus-type local area networks (LANs) has led to a proliferation of network architecture approaches. This has occurred due to the fact that it is not straightforward to implement a bus using passive fiber-optic technology. Nevertheless, a passive approach is one that best mimics coaxial or wireline bus designs, therefore workable designs have emerged, one of which is the so-called passive star approach. This paper discusses the nature of the passive star implementation. The architecture is described and design trade-offs are discussed. Applications of the passive star to CSMA/CD and Token Bus LANs are discussed. In particular, the problem of collision detection in passive star CSMA/CD networks is addressed

This paper reviews the operational characteristics, advantages, and disadvantages of implementing ring architectures for fiber optic local area networks. Different types of single and dual ring approaches are considered, as well as ring networks with novel connection architectures. Methods for the economical growth and upgrading of fiber optic ring systems are presented.

Over the past few years, a great deal of interest in local area network has rapidly been generated. At the same, fiber optic technology has ripened to the point that high reliable optical fibers, cables and devices are now available at moderate cost. Under these conditions, the development of FOLANs which has been made in a steel mill in Brazil.

Several different methods of implementing a fiber optic version of IEEE 802.3 10BASE LANs have been proposed as a candidate for standardization by IEEE. There have been extensive discussions as to the relative merits and features of the several systems. This paper will discuss the merits of each for this particular application on a comparative basis.

The capacity of fiber optic systems is generally limited by electronics rather than the fiber itself. Recent advances in electronics technologies, notably Gallium Arsenide, has increased the potential bandwidth for LANs to several Gigabits per second. This high speed together with new protocols and architectures will allow true integration in a packet switched network, of video and high-speed data (e.g. LAN interconnection) with lower rate data and voice traffic. We discuss research efforts on Gigabit/second packet network prototypes and briefly mention the emerging technologies that will impact such experiments.

The ANSI Fiber Distributed Data Interface (FDDI) is intended as a local area network standard for high-speed data transfer between computers and computers/peripherals. Three major applications are foreseen for FDDI: (1) Backbone network to tie other lower speed lower speed LANs together, (2) Back-end network for processor to I/O interconnections, and (3) ISDN applications. The paper will discuss these applications and outline the advantages of FDDI for these high-speed applications.

Theoretically based equations are obtained that relate the zero dispersion wavelength (λ₀) and dispersion slope (S0) of a graded-index multimode fiber to the fiber's normalized refractive index difference (Δ). Comparison of these curves with empirical data on 32 spools of fiber shows good agreement between λ0 and Δ, but large scatter in S_o. These relations have proven useful in establishing typical ranges of chromatic dispersion performance for commercially available multimode fibers.

A new parameter for characterizing the time response of a fiber optic channel is proposed to replace the traditional step response rise and fall times. The proposed parameter is shown to have significantly higher correlation with data dependent jitter for a set of measurements on seven different LED transmitters and two types of multimode fibers.

This paper describes the use of Surface Acoustic Wave (SAW) technology for clock recovery, and the performance of this technology in the Fiber Distributed Data Interface (FDDI) draft standard. FDDI, because it operates at 125 MBd over the fiber optic media, requires tight control of the jitter accumulated in each point-to-point fiber link. The clock recovery function must be able to recover the clock and correctly sample the bit stream, given a relatively narrow "eye opening" at its input. Furthermore, the clock must be maintained during the FDDI "line states," which can have a very low transition density. This paper will first describe the particular implementation of SAW technology used for clock recovery, the SAW filter, and will define it purpose within the clock recovery function. Then, the jitter characteristics of the FDDI signal at the input to clock recovery, as well as the performance of the SAW-based clock recovery function under these input conditions, will be discussed. Experimental results obtained using a typical, Commercially available, SAW filter-based module will be presented. The various "detuning" sources of the SAW filter, which detract from the capability of the SAW-based module to perform accurate sampling, will be discussed. The performance of the module under FDDI line state conditions, particularly Master Line State (MLS) and Quit Line State (QLS), will be analyzed. The QLS, which indicated a disabled upstream transmitter or a cable break, contains no transitions and therefore no information for clock recovery. A circuit will be presented which uses the station's local oscillator and the "signal detect" function of the fiber optic receiver to derive the recovered clock in the event of QLS.

In order to regenerate pulses in a transmission system for pulse coded signals, a clock signal should be available. In self timing repeaters, this clock signal is derived from the baseband information signal itself. This includes filtering by means of a resonant circuit (SAW filter) or by a phase lock loop (PLL).

Fiber optics is slowly, but steadily, moving into the telephone loop plant environment. Initially the key to movement of fiber optics into the telephone loop plant is cost and reliability. Trends toward low cost and high reliability appear in the declining prices of single-mode fiber and in an emerging interest in light emitting diodes as the optical source in short-distance single-mode fiber links. The attractiveness of an LED used with single-mode fiber is based on its potentially low cost, its potentially high reliability, and the reduced transmitter complexity it makes possible. By using present LED technology, it is possible to design single-mode lightwave systems that can transmit data at the DS3 rate over more than 10 km and over more than 20 km at the DS1 rate. It is expected that with future improvements in LED technology, it may be possible to achieve transmission spans of over 5 km at 565 Mb/s and of over 10 km at 140 Mb/s. Such system flexibility addresses nearly every emerging requirement for transport in the telephone loop plant.

This paper describes the procedures for the deployment of fiber hubs and fiber optic feeder systems in the subscriber loop environment. The term fiber hub, as used here, refers to a digital hub fed by a fiber feeder system. A detailed description is provided for the planning, design, and administration of the fiber feeder cable and the fiber hub. Based on current technology, the concepts presented will provide a reliable, flexible, customer responsive, and cost effective fiber optic system for the subscriber loop.

We present here a WDM experiment using dual-wavelength LEDs of 1300/1500 nm with gum emitter spacing between the active facets modulated at 140/560 Mb/s over 10 km SMF. The launched power into SMF was -22.5 dBm and -19.2 dBm for the 1300 nm and 1500 nm channel respectively. Chromatic dispersion penalty for the 1500 nm channel at 560 Mb/s was about 6.5 dB and the electrical crosstalk penalty between channels was 2.5 dB. The experiment demonstrated enhanced channel capacity and the feasibility of a local subscriber loop network without employing active light source at the customer premises.

Recent experiments have demonstrated the feasibility of both low- and high-speed 800 nm transmission on 1300 nm optimized single-mode fiber. Such a system is attractive for the local distribution network, because it combines the high performance of single-mode fiber with the low cost of short-wavelength opto-electronic devices. This paper provides an overview of this technology. System design considerations, including splice and connector requirements to minimize mode coupling and modal noise at misaligned joints, and the choice of sources to optimize the trade-off between modal noise and material dispersion, are discussed.

At the present time the copper twisted pair is limited in its transmission capabilities. In the long term it seems likely that the local loop will make greater use of fiber optic technology with the anticipated increase in the user demand for simultaneous transmission of high quality video, voice and high speed data. The current copper loop network is based on the star and double star configuration. This particular structure is familiar to the operating companies, and consequently deploying fiber in this configuration will allow for easier engineering, installation, operations, maintenance, records and administration. This paper reports on our preliminary analysis of providing fiber access to the subscriber directly from a central office or from an optically fed remote terminal. The focus is on the selection of the optimum location for the remote terminal based on the service requirements and economics under various scenarios.

This paper reviews the design considerations for optical fiber, cable, and associated components for a broadband fiber optic subscriber loop network in Pacific Bell. The most optimum fiber and cable designs for the feeder, distribution, drop and customer in-house portion of the subscriber loop are identified. Furthermore, the proper interconnecting hardware such as splices, connectors and splice enclosures for the subscriber loop applications is also briefly discussed.

In response to the Department of Defense's need to enhance the survivability of command and control systems, The MITRE Corporation developed SURVNET, a survivable fiber optic local area network. The network supports data communications with a fault-tolerant, distributed architecture capable of continued communication despite media failure and node outages. SURVNET is configured as a modified fiber-optic broadcast bus. The physical and data link layers are implemented with a combination of IEEE 802.3 (Ethernet) and an augmented version of IEEE 802.4 token passing bus protocols. Special nodes in the network, incorporating fault-tolerant software, are doubly connected to the fiber bus. Periodically, these nodes broadcast a self-addressed test message to determine if continuity exists on the network segment between the node's two physically separate connections. If a discontinuity is detected, the node utilizes its two connections to bridge between the isolated bus segments.

With the recent emergence of fiber optic technology, optical fibers have been examined as an alternative to coaxial cable in a number of local area network designs. To support this transmission medium, MITRE has developed an active star node that permits unique network topologies to be employed. With the introduction of low cost components, the active star node has been redesigned to employ this new technology. This document describes the low cost fiber optic network node design and examines current applications and limitations of this implementation.

A statistical design methodology for fiber optic long haul systems is presented. Based on actual data, the methodology improves the repeater spacings by 27% at 1300 nm and over 28% at 1550 nm. Our analysis is applied to 810 Mb/s transmission systems. It is expected that the repeater spacings for the new gigabit systems that MCI is currently installing will also increase in the same proportion as data becomes available. In addition, new generation LID splicers have been introduced and are expected to yield further repeater spacings.

The average (X) and standard deviation (s) of losses in single-mode fused silica fibres are determined, by a Monte-Carlo technique, taking into account: (11) the tolerance in the operating-wavelength of multi-longitudinal mode lasers; (2) the operating wavelength-regions according to CCITT and EIA; and (3) the OH-ion fibre content. As manin result is noted that while the effect of the laser-wavelength distribution is important, a significant OH-ion concentration (up to 2 dB/Km in the 1.3789 μm overtone) does not degrade the attenuation spread (3 s/X) neither the maximum attenuation (X + 3s) in the operating wavelength-regions.

It it hypothesized a new approach to describe the chromatic dispersion of dispersion-shifted single-mode fibres (DS-SMF). Measurements on a commercial available DS- SMF support this hypothesis. This new approach has interest because it reveals that for both conventional and dispersion-shifted fibres, the same curve shape could be employed to specify the dispersion.

Digital transmission through optical fibres and 'packet-switching' are two independent technologies that have gained wide acceptance during the past decade. This paper considers the possibility of a fusion between the two technologies. The network envisaged is a long-haul, packet-switched network, with optical fibre links between the nodes, for the integrated transmission of voice and data. The advantages of implementing packet-switching on an optical fibre network are brought out. The proposed network effectively utilizes the bandwidth of 140 Mbit/s fibre optic links to transport voice and data packets in a store-and-forward manner from the source node to the destination node. The issues involved in the selection of routes for voice and data packets are discussed.

Very active research and development efforts are being carried out by many laboratories around the world for the transportation and distribution of high speed data towards the realization of a broadband communication network of the future. Various names have been used to describe this network, such as Broadband Integrated Service Digital Networks (B-ISDN), Integrated Broadband Communication Network (IBCN), and Information Network Systems (INS). The various types of services offered include video telephony, video teleconferencing, broadband videotex, high speed data and document transmission, retrieval of high definition graphics, text and data...Single mode fibers have become the ideal transmission medium for this application because of their extremely high information capacity. The technology for single mode fiber transmission using direct detection technology has been well developed and widely used for high capacity terrestrial as well as undersea trunking applications. Field trials as well as detailed studies of broadband distribution networks using direct detection technology over optical fibers have been reported by many organizations worldwide.

Gigabit/s optical transmission systems operating on single-mode fibre, which upgrade the transmission capacity of 565 Mbit/s systems now being installed, will become available in the next years. To relieve the burden on electronic circuitry operating at the highest speed the system architecture is chosen such that these circuits can be kept simple. All complicated signal processing functions are carried out at tributary speed and can be implemented with Si IC technology, while for the highest speed circuits GaAs technology may be appropriate. Excellent linearity of optical sources allows multilevel modulation of the optical signal. Feasibility of a 1.13 Gbit/s system with 8 plesiochronous in- and outputs at 140 Mbit/s, realized with todays industrially available electronics, sources and detectors, and operating over 33 km of single-mode fibre, has been demonstrated. Multigigabit/s systems are preferably operated in the 1550 nm wavelength region, taking advantage of the low fibre attenuation. Ultimately the fibre dispersion characteristic should be shaped to have low dispersion in this low attenuation region.

There will be an evolution in fiber optic transmission equipment used in the local loop between the present and the mid-1990s. Current applications in fiber transmission is, and has been, focused on interoffice and long-haul trunking. Presently, the local loop is dominated by wire and digital loop carrier systems, with some use of fiber. Fiber optic digital multiplexers are now penetrating the local loop, to be followed by synchronous multiplexers and standardized fiber optic interfaces in the near future.

The basic premise of this paper is that at some point in the future all the transmission facilities of the public (telephone) network will be fiber-optic, i.e., end-to-end fiber connectivity will be provided to each subscriber. With the advent of coherent transmission systems the available bandwidth will become enormous, and thus the question we address is: given the availability of "infinite" bandwidth to every subscriber in the relatively near future, what does this imply for the network architecture? In particular, which switching technology is best suited to providing all foreseeable voice, data and video services? Having discussed advantages and disadvantages of architectures based on different switching technologies, we conclude that an architecture based on circuit-switched, fixed-bandwidth channels for the transport of user information is the most appropriate for a network in which there are no constraints on the transmission bandwidth. The standard channel capacity should be determined by the service having the greatest bandwidth requirement.

The issues involved in the design of a broadband integrated services distribution network which utilizes optical fibers in the subscriber loop are considered in this paper. A "hybrid" scheme is proposed which provides integrated access to ISDN and wideband services by delivering voice/data signals in a digital format, while using fan analog format for full motion video services to subscribers. The paper discusses implementation details of the network and provides a cost comparison to alternative options.

The performance of lithium niobate external modulators is presented. Two classes of devices are examined: directional couplers and Mach Zehnder interferometers. Some considerations concerning the use of external modulators in high-bit-rate transmission systems will be given.

With the increasing deployment of optical fibers in transmission system and demand for high bandwidth services, such as video and high definition TV, high-speed switching (throughput and switching speed) on the order of Gb/s is desirable in telecommunications. Most of the optical switching techniques developed to-date are addressing space division switching, in which switching speed is not a major concern. Very few optical switching techniques using time division switching have been reported. In this paper we will describe a high-speed switching concept by using a ring-based structure. This concept can be used as a self-routed circuit or packet switching. The switch has the ability to switch different types of information (voice, data, and video) and also can be applied to baseband channel-to-channel switching. This general purpose optical switch can be used in conjunction with an optical local area network as a distributed switch or can he used as a central office switch to switch trunk lines and individual lines.

The determination of the sensitivity of a single-mode optical fibre to microbending and macrobending is of great importance. Two kinds of spot-size rule such a sensitivity: w_n, the r.m.s. width of the near-field, and W_oo the inverse of the transverse propagation constant in the cladding. While measurement techniques for w_n are already known, this paper proposes an original method for w_oo measurements and concentrates on more plausible algorithms for macrobending and microbending loss determination from those measurements.

We discuss the potential errors associated with the measurement of optical power in a field environment. The potential errors arise because field use is often inconsistent with the calibration method. Errors may be due to the use of connectors of different types and due to variation among vendors for a given connector type. We consider potential errors in power measurements due to the variation in a connector type among vendors.

The recent shift in Fiber Optics from long haul telecommunications down to short links (LAN), such as smart buildings, interoffice networks and even mobile applications requires an adaptation to existing test instruments. The OTDR is probably most affected by these requirements. Natural physical constraints such as the speed of light and the inherently low backscatter levels are challenging factors in the design of such an instrument.

In measurements of the mode field diameter of single-mode fibers, systematic errors can result from mathematical approximations in several published formulae, and also from the use of coarsely-spaced truncated experimental data points. These two error sources have been examined for a number of fiber types, for four different methods of determining the Petermann II definition of mode field diameter. The methods are the simple far-field scan, the variable aperture in the far field, the knife edge in the far field, and the mask in the far field. We present computational formulae and experimental conditions that keep the individual errors below 0.2%.

A reliable test procedure for an optical communication transmission system has been developed and demonstrated over 6GHz range. The frequency response of an optical test system is measured for a short fiber (~5 meters) with a periodically deformed axis, inducing selective spectral loss on the propagating fundamental mode. The results of this measurements are compared with 10km transmission system and found in agreement with the test measurements.

A new optical spectrum analysis method using Babinet compensator is presented in this paper. We show how the combination of optical polarizer, analyzer and Babinet compensator can modulate a light beam to provide the Fourier transform relation between the spectrum to be analyzed and the position on a detection plane. A computer simulation result is given out.

The Optical Time Domain Reflectometer (OTDR) has been used to measure splice loss in fiber optic cables for several years. The OTDR requires bidirectional measurements be made and values averaged to get an accurate result. One theory implies that knowing optical characteristics of the fibers being spliced could allow for a measurement to be made in one direction only. This paper investigates that theory, and our findings indicate that bidirectional measurements are required in order to achieve accurate results.

The calibration of chromatic dispersion measurement equipment is discussed. The need for absolute standards for both wavelength and dispersion coefficient scale factor is paramount. Wavelength calibration may be performed using several interference filters of certified wavelength. Known optical time delays of high absolute accuracy enable calibration of dispersion coefficient. The development and performance of instruments using absolute dispersion and wavelength techniques in order to offer transferable calibration standards is presented.

The chromatic dispersion of four commercially available single-mode fibers was investigated over a temperature range from -40° C to +65°C. Values for zero-dispersion wavelength temperature coefficient were ~ 0.020 nm/° C for two depressed-clad fibers, 0.022 nm/° C for a matched-clad fiber, and 0.030 nm/° C for a dispersion-shifted fiber. Such a small temperature dependence would not pose a serious problem for current installation. However, future high bit-rate systems should be designed to minimize this effect.

There is a widespread misconception that a connector can be completely characterized by measuring its insertion loss and reflection properties at a single wavelength. This paper presents experimental evidence of large connector loss variations with wavelength and a theorectical model of this phenomena. Connectors which have an air gap separating two polished fiber end-faces (such as FC connectors) exhibit loss oscillations as high as 1.1 dB (peak-to-peak) as the wavelength is varied from 1300 nm to 1550 nm. In addition, the optical power reflected from such a connector also fluctuates considerably (from less than -30 dB to as high as -7 dB) over the same wavelength range. In a fiber span terminated with FC connectors at both ends it is possible, therefore, for the total connector loss and reflected power to increase by more than 2 dB and 20 dB, respectively, when operated at a wavelength of 1550 nm rather than 1300 nm. This would be of concern in high speed lightwave systems which have low system gains and whose lasers are very sensitive to reflected light. This effect can also lead to large errors in spectral attenuation measurements (both in the lab and in the field) unless the proper precautions are taken. We observed this phenomenon while constructing a spectral attenuation test set composed of multiple lasers in the 1300 nm and 1550 nm windows for a "window verification" project.

A simple interferometer for making accurate outside diameter measurements of optical fiber is described. The apparatus, which measures the diameter in terms of Fizeau fringes is easy to implement, requires minimal alignment and can be constructed from low cost components. The device has a precision of ± 0.03μm and an ultimate accuracy of 1 part in 10⁶.

Basic approaches to fiber optic process control networks will be reviewed. Circuit and packet based switching networks in point-to-point, star and ring configurations will be surveyed. An actual system installation utilizing a ring based fiber optic multiplexer in an oil field will be presented.

The requirements for a Local Area Network (LAN) suitable for use in factory environments are discussed. In many cases these requirements are met by fiber optic Ethernet (IEEE 802.3 compatible) LANs. Implementation of a passive fiber optic star coupler based Ethernet LAN is presented in terms of hardware and system design rules. Application of the passive star coupler topology LAN to a number of factory applications is described.

Manufacturers need to improve their operating methods and increase their productivity so they can compete successfully in the marketplace. This goal can be achieved through factory automation, and the key to this automation is successful data base management and factory integration. However, large scale factory automation and integration requires effective communications, and this has given rise to an interest in various Local Area Networks or LANs. In a completely integrated and automated factory, the entire organization must have access to the data base, and all departments and functions must be able to communicate with each other. Traditionally, these departments and functions use incompatible equipment, and the ability to make such equipment communicate presents numerous problems. ARCNET, a token-passing LAN which has a significant presence in the office environment today, coupled with fiber optic cable, the cable of the future, provide an effective, low-cost solution to a number of these problems.

The current status of standards development activities for fiber optic Local Area Networks (LANs) is reviewed. The discussion is limited to those standards based LANs intended to operate at 20 Mbs or less.

In this talk we will present fundamental theoretical considerations affecting the performance of coherent lightwave communication systems. These systems include a single-mode laser and a modulator at the source and a coherent receiver which is the optical onalog of a superheterodyne radio set. Instead of detecting photons directly, the coherent receiver first converts the incoming signal from the optical regime down to the radio regime, and then uses conventional electronic circuitry to perform various signal processing operations, such as amplification and demodulation. In principle, this technique can yield large increases (~ 20 dB) in receiver sensitivity compared with direct detection using avalanche photodiodes. One of the prime causes of this degradation has been identified as laser phase noise, a phenomenon that is known to be particularly serious in semiconductor devices.

This paper examines possible use of coherent based fiber optic systems in the telephone network. Coherent technology can increase the bandwidth of fiber based systems beyond the limits of current direct detection methods used for today's fiber optic systems by increasing the sensitivity of the receiver. Coherent technology can be used for one operating wavelength or in combination with Wavelength Division Multiplexing (WDM) devices for many operating wavelengths. "Wavelength management" will be necessary for the telephone companies should WDM devices be used with coherent technology. The most likely application for coherent systems is as transport for broadband ISDN when Gigabits per second (Gbps) bit rates and above are needed. High speed transport may be necessary between the remote electronics and broadband ISDN central office, between high speed fiber based Local Area Networks (LANs), and between central offices of a telephone company. System design should hinge on the standards developed for broadband ISDN. First, background information is given to identify the technology basics and application information, then telephone company operations issues are discussed, and finally conclusions are given.

A computer simulation of DPSK and PSK transmission systems has highlighted several parameters that critically affect system performance. The simulations reveal that, in practice, PSK systems can be made to operate closer to their ideal detection sensitivity than DPSK systems. These findings are supported by experimental measurements. A PSK system was assembled using a travelling wave laser amplifier repeater, the addition of which caused no significant degradation of receiver sensitivity. The experiment confirms the suitability of laser amplifiers for coherent systems.

We show quantitatively for the first time that the feedback bandwidth is a critical parameter controlling the linewidth in external cavity semiconductor lasers. The external cavity structures can be divided into four categories according to their feedback bandwidth Avf and feedback ratio R₃. The limit of linewidth narrowing depends on feedback bandwidth rather than the length of external cavity. For very narrow linewidth (< 10 kHz), the feedback bandwidth required is < 100 kHz, provided optimum choice of the other three critical parameters R₂, R₃, 1_ex is made as defined below.

Four birefringent elements are shown to provide endless polarisation transformation between two time-varying states. The scheme is demonstrated experimentally using four squeezers under computer control. Further implementations of the algorithm are discussed.

A polarization-insensitive detection technique for coherent heterodyne fiber communications is described. The technique utilizes a dual-IF detection scheme derived by mixing the signal light with orthogonally polarized and frequency offset components of the local oscillator. Experimental results obtained using both bulk-optic and an integrated-optic, polarization selective phase modulator to provide the necessary LO polarization mode differential frequency offset are reported.

The double-balanced mixer has a significant noise advantage over the single type. This paper outlines the reasons, observes the progress in optical implementation, and considers ways forward using currently available devices.

An experimental study of the polarization properties of AT&T's rectangular polarization-maintaining fiber is reported. The extinction ratio is measured at 1320 nm for lengths of 20m, 40m, 100m, 310m, and 804m. The first demonstration the the length dependence of the extinction ratio of a polarization-maintaining fiber obeys random coupling theory is also reported; this is due to the rectangular geometry of the fiber which facilitates winding and packaging the fiber with no perturbations. The polarization-holding parameter is 2.08 ± 0.09 x10^-6m^-1. The state of polarization as a function of the input angle of the linear polarized field is determined for various lengths. For lengths ≥100m, depolarization effects eliminate all other states of polarization except the linear on-axis polarization state. For shorter lengths, other polarization states can propagate in the fiber; however, if the input angle of the field is within 5° of the polarization axis of the fiber only the linear on-axis polarization state will propagate in the fiber. Using an elasto-optic point perturbation technique, a beat length of 6.78 ± 0.36 mm was measured directly at a wavelength of 1320 nm, inferring a phase index difference of 1.95 x10^-4.

A simple, non-modulating elasto-optic point perturbation technique has been developed to non-destructively and directly measure the beat length of polarization-maintaining fibers over a wide wavelength range. The elasto-optic point perturbation technique involves a translatable non-modulating mechanical point perturbor which perturbs the coated PM fiber at a point at some angle with respect to the principal axis, while the state of polarization (i.e., the Stokes vector) was monitored as a function of length (i.e., the relative translation distance of the point perturbator) for a given wavelength. This technique allows, in principle, for the determination of both magnitude and orientation of any arbitrary birefringence. We present data only on the linearly birefreingent polarization-maintaining fibers. The sample investigated was a rectangular polarization-maintaining fiber produced by the preform deformation process. The beat length determined from the period of the Stokes vector was compared with our theoretical model of the technique. The comparison demonstrates the excellent agreement between the non-modulating elasto-optic point perturbation technique and theory, while the precision was to within ± 0.16 mm. In addition, the beat length data from this direct technique was also compared to beat length data from an indirect wavelength scanning technique. The indirect data differed from the direct data by 12% at the 1302 nm wavelength.

Polarization-maintaining fibers preserve a stable state of polarization over long propagation lengths and are ideal candidates for applications such as coherent communication, fiber devices and interconnection of guided wave devices. Connection technology for polarization-maintaining fibers is reviewed, design difficulties are examined, characterization techniques for these splices and connectors are analyzed. Reported results on individual polarization-maintaining fiber splices and connectors are described. In addition, linear arrays of polarization-maintaining fibers suitable for guided wave applications are reviewed. It is concluded that splices and connectors for polarization-maintaining fibers are being studied by many investigators; however, most of these techniques require active alignment of the polarization axes. Passive alignment appears feasible if a suitably accurate, external polarization-axis reference is designed into the fiber.

The use of coherent detection in optical fiber communications offers advantages of increased signal transmission capacity, improved receiver sensitivity, and flexibility of system design. This paper describes the practical limits to coherent digital system performance and design considerations to a 140 Mb/s FSK digital system. Despite the recent advances made in various research laboratories, coherent fiber optic systems are still in the infancy stage, largely attributed to the lack of components needed to realize the ultimate performance which coherent technology can offer. In this paper, the factors affecting performance degradation are identified and the relative significance of each factor is examined.