Click for next page ( 85


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 84
Appendix D Technology Status of Optical Telecommunications LASERS A key element of all long-haul optical communication systems is the semi- conductor laser. The desirable characteristics of the semiconductor laser are determined in large measure by the characteristics of the optical fiber and the lightwave system architecture. Future lightwave systems are likely to contain a large number of closely spaced channels operating in the 1.5- to 1.6-micron wavelength band of low loss. In addition, high-bit-rate systems (> 1 Gbit/s) require narrow-line, single-frequency lasers to offset the chromatic dispersion of the fiber. Two types of lasers have been extensively investigated for obtaining single- wavelength emission. They are the external cavity laser and the distributed feedback (DFI3) laser. External cavity lasers that have been extensively inves- tigated include the cleaved-coupled cavity laser (C3 laser); graded-index, external cavity laser; fiber external cavity laser; the silicon chip Bragg reflector (SCBR) laser; and InP-based, external Bragg reflector lasers. The linewidth of the semiconductor laser is determined by the fluctuations in the phase and intensity of the photon field in the laser cavity. These depend markedly on the cavity length. Continuous wave linewidths on the order of a few kilohertz (kHz) (necessary for coherent applications) have been obtained for external cavity lasers compared to those of many megahertz (MHz) for DEB lasers. It is likely that some form of external cavity laser will be the laser of choice for the next generation of high-data-rate, coherent transmission systems. 84

OCR for page 84
APPENDIX D 85 Among key desirable features of such external cavity lasers would be the fabrication of a truly compact, robust laser. Multielectrode lasers capable of wavelength tuning over the gain spectrum of the laser will become important for closely spaced wavelength WDM applications. Even though tuning features have been demonstrated in the laboratory, we do not to date have a stable, compact, widely tunable, single-frequency semiconductor laser suitable for practical commercial use. Further research in this area is warranted. Another important problem is the behavior of the laser under modulation conditions. "Chirping" (change of frequency during a pulse interval) of the laser under direct modulation is an important limitation as one goes beyond about 2 Gbits/s. The "chirping" behavior of semiconductor lasers is determined in part by the internal structure of the laser. Buried heterostructure (BH) lasers have generally low chirp and are favored, despite their complexity in manufac- ture, for high-bit-rate systems. As one goes to the 10-Gbits/s regime, it is likely that external modulation willbe necessary for high-bit-rate, error-free transmis- sion. The use of external modulators, however, results in additional power loss, and the system designers would need high-power (~50 mW) lasers for practical high-bit-rate systems. In applications where optical interconnection distances are relatively short (e.g., internal equipment interconnects), and laser output power is therefore not critical, lasers with very low-threshold currents ~ < 1 mA) based on single-quantum well structures represent an important emerging technology. FIBER AND CABLES The two main characteristics of a fiber waveguide are its attenuation and bandwidth. Attenuation in present silica materials has been reduced to almost the theoretical lower limit. Research is currently under way to fabricate fibers using fluoride glasses, which may have attenuations a hundred times lower than present fibers. There are enormous problems to overcome in making practical waveguides from these new materials, but the research warrants continuation because the potential benefits are immense. Conceivably, transoceanic cables could be constructed without underwater repeaters, as an example of one benefit. The maximum data rate (bandwidth) of an optical signal that can be sup- ported in a fiber is presently limited by fiber material and waveguide dispersion, interacting with the spectral width of the optical sources. With emerging single- frequency sources, this shows no sign of becoming a limiting factor. Further, carefully tailored waveguide designs have been used to make fibers that can provide for flexibility in the choice of light source (the dispersion-shifted and the dispersion-flattened fibers, for example). This fiber design work is an important

OCR for page 84
86 APPENDIX D research effort that should proceed simultaneously with other activities to facilitate alternative system designs. Beyond attenuation and bandwidth are a number of environmental requirements defining fiber degradation in use. These requirements have become increasingly important as the desire has evolved to both reduce fiber protection and submit fibers to more hostile environments. This has led to complex materials studies aimed at the improvement of materials properties. One such large effort is in hermetic coatings for strength. A second involves studies of atomic defects generated (or activated) by hydrogen diffusion into the glass, leading to optical absorption at system wavelengths being used. While this phenomenon is generally understood and empirical results show that the attenuation increase is usually negligible, there remains concern about adverse environments with high hydrogen content or high temperature. Continuing research to understand this problem at the atomic level will help provide reassurance to fiber users and will assist manufacturers in expanding the environmental durability. The function of cabling is primarily to protect the fiber from mechanical stress, both lateral and longitudinal. The latter is conceptually the most straightforward and is accomplished by adding strength members to the cable, as well as by incorporating excess fiber length so that some cable elongation can occur before the fibers are strained. Progress in this area has been adequate and continues as higher-specific-strength materials become available. Lateral stresses on the fibers are more complex and actually are generated during the cabling process. These minute cabling stresses result in "micro- bends" that increase the fiber attenuation. Advances in cabling techniques and the increased use of single-mode fibers have diminished the importance of this problem at present attenuation levels. However, increased understanding is helpful in advancing present technology and will become essential if the lower- attenuation fluoride fibers are to become commercially available. Microbends from cabling have an added dimension of complexity when multimode fibers are used, since the mode stripping (removal of higher-order propagating modes) they cause also increases the fiber bandwidth (reduces pulse spreading because of the smaller spread of propagation speeds amongst the modes that are not stripped). System designers desire a length-bandwidth relation that optimizes fiber performance. Since pulse spreading in multimode fiber depends on a large number of fiber and excitation parameters, so far it has not been possible to provide a satisfactory analytical expression for the length-bandwidth relation. While the importance of the multimode fiber length-bandwidth relationship has diminished with increasing single-mode fiber usage, multimode fibers are being considered for future applications, which would again increase the importance of research in this area.

OCR for page 84
APPENDIX D 87 Aside from these specific technical areas, research and development in cable technology is gradually advancing with corresponding progress in cabled fiber cost reduction. This cost-performance improvement work should continue to be supported. PASSIVE COMPONENTS The cost and performance of passive components are major impediments to the advance of a number of applications of optical communications, e.g., local area networks. An exception to this may be fibers and cables that have benefited from a great deal of research and development over the years. The other components (discussed here) have not received as much attention and are not improving at the same rate as fibers and cables. For example, almost all the commercial passive components of today were designed and built in the laboratory over 10 years ago--the incorporation of new ideas has been rare. The obstacles these components present can be illustrated by the performance of Remountable connectors, perhaps the most commonly used device, which typically can have insertion losses of up to about 1 dB. The power loss a system can tolerate is typically 20 to 30 dB. Therefore at the 1-dB-per-connector loss level, no more than about 20 connectors can be incorporated between a source and detector. Contrasting this with the negligible loss of present coaxial con- nectors shows that major design compromises have to be made in optical fiber system design. The basic reason for this difficulty is the small cross-sectional size of the optical beam--which is, at the same time, a major advantage in terms of miniaturizing complex systems. There is a lack of sufficiently precise, three- dimensional forming techniques for manufacture of optical fiber system components. The two-dimensional solution, photolithography, has been used to some advantage when the third dimension is small; but it is generally inadequate. The forming problem has led to cost-performance trade-offs in component insertion loss. This problem becomes particularly vexing with the increasing incorporation of single-mode (i.e., small-core) fibers that require tighter tolerances. It is generally conceded that systems of the future will utilize single-mode fibers, so that this problem will assume increasing importance. These general considerations, and some more specific ones, will become more evident in the following discussion of individual components. Connectors Single-f~ber connectors are classified into two types, contact or buttjoint and expanded beam. The expanded-beam connector is more stringent in angular alignment tolerances, whereas the contact method is more stringent in

OCR for page 84
88 APPENDIX D lateral positional tolerances. The practical compromise between these two needs further investigation, including the necessity for optically polishing fiber ends. Single-fiber connectors typically cost $10 to $100 for multimode fiber and $20 to $200 for single-mode fiber, depending on performance. In general, assembly in the field needs simplification. Multif~ber connectors generally accommodate only a few fibers and employ some type of circular geometry. As in all connectors, draftspersons also have difficulty assembling these in an adverse environment. Array connectivity seeks the advantage of easily joining cables with large numbers of fibers. Presently, a flat grooved plate holding the array of fibers is the basic element. The difficulty of obtaining uniform results for all fibers is apparent, and no attempt is made to fabricate such connectors in the field. Instead, cables are terminated in the factory, with attendant lack of flexibility in deployment. Connectors with low reflection coefficients are increasingly important in high-data-rate and coherent communications systems because reflections returned to the lasers used in those applications can cause instabilities in the laser output that are harmful to system performance. Reflections at connectors placed in series along a fiber can also cause fluctuations in the output power at the end of the fiber due to interference effects. Wavelength-Division Multiplexing (WDM) Components for Filtering, Multiplexing, and Demultiplex~ng Wavelength-division multiplexing (WDM), optical communication in the wavelength multiplex mode, allows modulated radiation from several laser sources of clearly distinct wavelengths to be transmitted simultaneously over a single fiber. Spectrally selective optical multiplexers or de-multiplexers are used at the start and end of the transmission route to ensure low-loss combina- tion and separation of light of the various wavelengths. WDM technology is a key to the utilization of the full bandwidth capability of optical fibers. Commercial communication systems in operation today utilize wavelengths that are widely separated (0.8 microns, 1.3 microns, and 1.5 microns), but future systems are expected to utilize single frequency lasers in the 1.5- to 1.6-micron low-loss band, spaced a nanometer or less apart in wavelength. Integrated optics is expected to play an important role in developing the necessary active and passive WDM components required for high-capacity WDM. These will require arrays of stable, single-frequency lasers of well- defined wavelength, low-loss waveguides, narrow-band gratings, filters, and multi-wavelength photodetectors. Realization of practical WDM components for closely spaced wavelength applications hinges greatly on the development of new materials technology.

OCR for page 84
APPENDIX D 89 Vapor-phase growth techniques for fabrication of arrays of InP-based lasers and detectors of high uniformity need to be developed. New materials combinations with large electrical field effects for switching of optical signals in waveguides need to be explored. The possibility of using the mature silicon materials and processing technology for fabrication of low-loss waveguides and gratings on a silicon chip is particularly attractive. Fiber Backplane and Other Optical Interconnects As data rates increase, interconnections inside equipment become increasingly difficult to implement with conventional copper traces on circuit boards and backplanes, twisted pairs, and coaxial cables. Recently very short- distance f~ber-optic interconnects have been used in place of copper wires and cables. However, there is research in progress to implement both the photonic equivalent of printed-circuit interconnects and also optical free-space intercon- nects. The challenges associated with optical circuit-board and backplane traces are twofold. First, appropriate materials and processing technologies must be created to form light-guiding regions on circuit boards and backplanes of sufficiently low loss, of sufficient dimensional tolerance, and sufficient reliability --all at an acceptable cost. Second, optoelectronic interfaces are needed to economically couple light into these light-guiding traces and to remove the light at the other end (or possibly at several places along a light-guiding trace). It is desirable that these light-launching and light-receiving optoelectronic interfaces be integrated into electronic circuits, so that packaged electronic circuits with these optoelectronic interfaces can be fabricated as components that can be mounted directly on circuit boards containing the optical traces they will access. Free-space interconnects offer the possibility that arrays of optical transmitters can be connected to arrays of optical receivers with relatively simple and rugged lenses to define the optical paths. Such free-space optical interfaces offer certain advantages over fiber and optical traces on circuit boards, such as propagation delays that can be identical for a large number of connections that are not exactly parallel. Free-space interconnects eliminate the need for connectors and may improve reliability of interconnections. Realization of practical free-space interconnects awaits the development of arrays of reliable optoelectronic transmitter and receiver modules and ap- propriate lens and physical design technologies to achieve the desired align- ments. Research in interconnect topologies is also needed to obtain a synergy between the capabilities of free-space interconnects and interconnection applications that can use those capabilities.

OCR for page 84
9o APPENDIX D Other Passive Components As higher data rates and coherent techniques are employed in f~ber-optic systems, the lasers used to meet the requirements of these applications are increasingly sensitive to reflections that cause various instabilities. Optical isolators are needed to reduce reflection effects. Miniature opto-isolators are being employed in high-performance laser packages today. Further improve- ments in packaging and integration are desirable to reduce cost and increase reliability. Optical directional couplers with low insertion losses and predictable coupling ratios are needed for removing and adding light in passive bus configurations. Although various directional coupler designs have been demonstrated and manufactured (e.g., fused tapered couplers), improvements in cost, performance, and reliability are still needed for many applications. Similar remarks apply to star couplers, which are used in networking configura- tions that are alternatives to passive bus configurations. Some couplers are incorporated in flat substrates and therefore suffer from geometrical mismatch in going from fibers to rectangular waveguides. Planar structures by themselves do not have a bright future but will become very important when combined with active devices in the emerging technology of optoelectronic integrated circuits. PHOTONIC SWITCHES A number of technologies have been demonstrated for switching an optical signal between two or more outgoing paths. These include mechanical devices that physically move fibers or that physically move lenses or mirrors directing an optical beam; optoelectronic devices where an applied voltage across two or more electrodes causes a field within an electro-optic material, which in turn changes the coupling of waveguides within the material or otherwise modifies the optical characteristics of an optical circuit within the material; electrically, acoustically, or optically controlled gratings created within a material to cause diffraction of an optical beam; and electrically or optically controlled non-linear optical devices. Of the variety of optical switching devices demonstrated or proposed, some are more practical than others, and some have near-term applications (e.g., simple mechanical switches for remotely controlled optical cross connects). However, optical switching devices are in general larger and more power consuming than their electronic counterparts; and many of these devices have numerous practical limitations such as temperature sensitivity, polarization dependence, wavelength dependence, requirements for high voltages, and high loss. Materials improvements and device-design improvements are the two key

OCR for page 84
APPENDIX D 91 dimensions of current research on these devices. While these device limitations are being actively addressed, much systems research is also needed to achieve large-scale application of optical switching devices in systems as a replacement for electrical-to-optical conversion accompanied by electronic switching. OPTICAL AMPLIFIERS Optical amplifiers are potentially important building blocks of all optical communication systems. In present optical communication systems the amplification function is accomplished by converting the optical signal to electronic form (detection), amplifying the electronic signal with an electronic amplifier, and then reconverting the amplified electronic signal to optical form. There are two main types of optical amplifiers: (1) f~ber-based amplifiers, and (2) semiconductor laser-based amplifiers. The main uses of optical amplifiers are in (1) pre-amplifier applications where amplification of low-level signals is performed and there is no intentional loss between the output of the amplifier and the receiver and (23 in-line applications where relatively large optical signals are amplified and loss is expected between the output of the amplifier and the receiver. The former are likely to be important in high-bit-rate (>2 Gbits/s) systems if good APDs do not exist. In-line amplifiers are believed to be useful in both long haul (to compensate for fiber losses), in the local loop (to compensate for split-off and coupling losses), and in optical switching to compensate for losses in the switches. Over the last few years, there has been considerable worldwide activity in developing amplifiers with large available gain, low insertion loss, low noise, large bandwidth, and saturation output power. To date, no practical semicon- ductor laser amplifier has been developed. The main potential advantages are the ease of manufacturing, high gain, and the tunability of the bandpass used for noise filtering and channel selection. However, semiconductor laser amplifiers have polarization-dependent gain that needs to be controlled through devel- opment of better optical isolators. Fiber amplifiers, especially Raman amplifiers, suffer from the high pump power required for amplification. Research needs to be performed on special fibers with low loss and high Raman cross-section as well as special dopants for optically pumped fiber amplifiers. This is a promising field that needs increased attention. INTEGRATION AND PACKAGING The interfacing of optical components with electronic ones is a key element for all future information transmission systems where one envisages the merger

OCR for page 84
92 APPENDIX D of optical signal processing with purely electronic media such as high-speed computers. One can imagine that in the interest of low cost and circuit simplicity, the terminal sources and receivers in the optical link may take on electronic processing involved with the communication link. One can fabricate, for example, a heterojunction bipolar transistor driver and a laser on a single chip or a pin photodiode and a field-effect transistor on the same chip. More complex integrated devices involving arrays of lasers, detectors, amplifiers, transistors, and modulators can be imagined. One of the main motivations for optoelectronic integration besides cost is performance. As speed increases, the interconnection of integrated circuits and subsystems becomes more critical and cans ot be easily implemented with technologies available today. Compound semiconductor-based transistors are intrinsically faster than Si ones, and the monolithic approach provides significant additional improvements through reduction of undesirable parasitics associated with packaging discrete devices. A key stumbling block in the exploitation of optoelectronic integrated devices has been the materials and processing technology. With high levels of integration, large-area compound semiconductor substrates of exceptional quality (low defect and dislocation density) and a vapor-phase crystal growth technique for growing uniform, epitaxial layers on the surface are required. Recent progress with hybrid MOCHA and MBE techniques suggests that this may be close at hand. However, because of the many conflicting processing requirements for optical and electronic devices, the ability to grow patterned structures in situ in a multichamber MBE machine up ultimately be extremely important if one is to exploit the full benefits of optoelectronic integrated devices. Major emphasis should be placed on developing further the materials and processing technology based on hybrid MBE/chemical-vapor deposition multiwafer, multichamber machines. In addition to the integration of optoelectronic devices and electronic devices, the integration of the optoelectronic device with a fiber into an appropriate package is an area ripe for development. In many applications what is desired by the system or subsystem manufacturer is an optoelectronic module consisting of some electronics, an optical emitter or detector (or possibly both), and an attached pigtail consisting of an appropriately protected optical fiber. Today, because of relative low-volume production, many of the assembly operations for optoelectronic modules are done by manual procedures or at least require considerable human intervention. In the future, as higher- volume applications for optoelectronic modules emerge, automated assembly and testing technologies will be essential to compete in these markets. This involves careful design of the components (including locational tolerances) within the assembly in order to facilitate automated assembly, and also development of appropriate robotic manufacturing equipment with the necessary tolerances.

OCR for page 84
APPENDIX D 93 RECEIVER SUBSYSTEMS The key elements of a receiver subsystem are the optical detector and the low-noise amplifier that couples the optical detector to conventional electronics. In telecommunications applications two types of optical detectors are typically used. These are the PIN photodiode and the avalanche photodiode (APD). Both types of devices can be fabricated from silicon material for 800- to 900-nm wavelength systems. Both types of detectors can also be fabricated from indium-gallium arsenide phosphorus material compositions for 1300- and 1500-nm wavelength systems. At these longer wavelengths, however, the APD is just beginning to emerge from the research laboratory as a practical device. Leakage current (dark current) is often critical to the overall performance of the receiver, and careful material processing is needed to keep leakage current low in the longer-wavelength devices. In the APD, precise control of the composi- tion and thickness of a sequence of sequentially grown layers of material is also critical. To construct a receiver with high sensitivity and large dynamic range (ability to accommodate a wide range of optical signal levels without overload) as well as high bandwidth requires careful circuit design and the use of advanced electronic integrated-circuit technology. The United States is well positioned in terms of both detector technology and low-noise mnplif~er technology, although recently the Japanese have been more aggressive in the development of InGaAsP APDs (which were fast demonstrated in U.S. research laboratories). What is ripe for development is the fabrication of low-cost InGaAsP APDs and the incorporation of these devices into low-cost modules with attached fiber pigtails and with integral low-noise preamplifiers. An area meriting continued research emphasis is the monolithic or hybrid integration of very low-capacitance photodiodes and very low-front-end capacitance preamplifiers in order to achieve the ultimate in low-noise performance, and thus high sensitivity. The United States is very competitive In this area. LONG-WAVELENGTH AVALANCHE PHOTODECTECTORS For high-bit-rate, high-performance, 1.3- to 1.6-micron communication applications, APDs are the detectors of choice because their internal carrier multiplication process allows weaker light signals to be properly detected. APDs currently in development are based on the InP/InGaAs(P) material system and are of the heterojunction type with a separate absorption region (usually the InGaAs layer) and a separate multiplication region. The pn junction

OCR for page 84
94 APPENDIX D is in the wide bandgap InP layer to avoid excessive tunneling dark currents. Hence the name SAM APD is given to this structure. The response speed of SAM APDs is limited by trapping of holes at the heterojunction interface. This can be greatly improved by placing a graded InGaAsP layer between the absorbing and multiplication regions. Such separate absorption and graded multiplication (SAGM) APDs have exceptional response speeds and high sensitivity. For example, with a receiver operating at 8 Gbits/s, at a wavelength of 1.5 microns, and using a GaAs PET front-end amplifier, a sensitivity of -26 dbm has been obtained for a 10-~ bit error rate. TypicalAPDs of this type yield receiver sensitivities that are 5 to 10 db better than those achieved with non-multiplying PIN detectors. The sensitivity of a receiver employing an APD is determined by the relative impact ionization rates of electrons and holes and by dark current. Recently, several advanced APDs have been proposed that rely on superlattice band- structure engineering to modi~the relative impact ionization rates. SuchAPDs are in early stages of research and require exceptional control of materials both in terms of doping as well as composition for their practical realization. They, however, hold the promise for achieving receivers with a performance dictated by the laws of quantum mechanics.