Emerging fiber-optic technologies battle customer skepticism

Dec. 1, 1997

Emerging fiber-optic technologies battle customer skepticism

While new technologies are being introduced at ever-increasing rates, customer acceptance is frequently slow to arrive.

Jeff D. Montgomery and Stephen Montgomery

ElectroniCast Corp.

After decades of relative stability in a regulated environment, the nature of network service providers and their offerings has now entered an era of highly volatile change. While this activity is most dynamic in North America, it is advancing globally. The shift to relatively unlimited competition in an unregulated market is leading to extensive mergers, partnerships, acquisitions, and new ventures on scales ranging from local to global.

The conception and development of new technologies are also climbing exponentially around the world. Yet dozens of world-class laboratories are demonstrating interesting new optoelectronic concepts much more rapidly than customers are willing to accept them. Designers of telecommunications equipment and networks are much more comfortable with the gradual evolution of established technologies and strongly resist adoption of a new technology that represents a major shift. For example, the steps for acceptance for dense wavelength-division multiplexing (dwdm) include:

laboratory demonstration, supported by the presentation of technical papers, product introduction and demonstration, such as at trade shows,

"paper" analysis, by carriers, to evaluate the cost and performance attractiveness of this alternative versus others (this been retarded by unavailability of firm prices for some alternatives),

carrier field trial of the product in an operational network,

carrier laboratory evaluation of the product (and, typically, alternatives),

the writing of procurement specs, issuing of requests for quotes, reviewing responses, and eventual procurement,

delivery and deployment.

Each of these steps requires several months. The major interexchange carriers are now at the end of this cycle, into procurement and deployment. Most local exchange carriers were somewhere between the paper analysis and field trial stages at the end of 1996.

Carrier skepticism is increasing

Such elaborate steps for technology acceptance are the result of the fact that communications carrier network planners are becoming increasingly cautious about "vaporware"--new, advanced technology products that are promoted as being essentially commercially available. With increasing frequency, planners are finding after extensive investigation that many "new" products are only laboratory models that may change significantly before production-level shipments or that may never be delivered. So they are depending less on vendor veracity and more on independent laboratory tests, their own laboratory tests and field trials, and on firm, fully specified quotations. At the equipment level, achieving industry-approved open standards also is a "pacing item," i.gif., an item approved by an appropriate standards committee and operational with competing vendors` products.

The result of this customer caution is a two- to five-year delay between the time the first vendor introduces a new technology and the time it is deployed in operational networks. The speed of adoption depends on the urgency of the problems the new technology might solve, the complexity of the new product, the extent it differs from the current technology, the nature of the vendors, and other elements. The time between new product introductions, however, is decreasing rapidly (see Fig. 1).

Amplifier growth supported by economics and technology

Even with this background of skepticism, some technologies should near acceptance shortly--either because they have weathered the long acceptance cycle or because they meet a critical need. For example, the principles of the optical-fiber amplifier have been understood since the early 1980s. Commercial deployment of these amplifiers, however, did not emerge until the early 1990s. The early adoption of optical-fiber amplifiers was delayed by

relatively high cost, compared to other solutions,

reliability concerns, especially with pump laser diodes,

the general reluctance to use a new technical concept.

These impediments began to dissolve as a result of the accelerating deployment of very expensive submarine fiber-optic cable systems and the continuing development of pump laser diodes.

The early optical-fiber amplifiers used erbium-doped silica fiber, pumped by 1480-nm laser diodes. This was a significant mismatch to nearly all of the fiber deployed by the early 1990s; the fiber had its minimum chromatic dispersion in the 1310-nm band. The erbium-doped fiber amplifier, however, was the most straightforward combination of available pump diodes and doped fiber, and was consistent with the evolution toward long- distance transport in the 1550-nm band. Continuing concerns with flat broadband gain, higher power amplification, lower noise, and servicing the legacy 1310-nm fiber have led to commercial availability of amplifiers with multiple dopants, doped fluoride fiber, 980-nm pumps, 980/1480-nm combination pumping, higher power pumps, and other variations.

The worldwide consumption of optical-fiber amplifiers will expand moderately, from $1.03 billion in 1996 to $5.13 billion by 2006. This includes both captive production by vertically integrated operations and merchant market sales. These values include optical-fiber amplifiers for 1550- and 1310-nm, and other wavelength bands. Optical-fiber amplifier usage will be led by North America, with a 62% share in 1996, dropping in market share slightly to 53% ($2.71 billion) in 2006, as shown in the table.

Components advance

Meanwhile, significant work has resulted in semiconductor and component advances. For example, semiconductor optical amplifiers (soas) are in early commercial application, especially by Alcatel Optronics of France. The development of the soa has lagged behind that of the optical-fiber amplifier (OFA) by three to four years, but is now accelerating. The soa has lower gain than the OFA, and early research devices had excessive noise and other problems. Impressive progress has been made in the development of these devices between 1994 and 1997, however.

A significant early application of the soa is in high input/output port count optical crossconnect switches, in which the soa functions as both a switching element and an amplifier to compensate for branching loss. Other soa applications are presently emerging. British Telecom was a leading early developer of the soa, leading to commercial availability via bt&d (a joint venture between British Telecom and Dupont; later acquired by Hewlett-Packard). Alcatel, Ericsson, and sdl are leading soa developers in 1997, with impressive results being achieved in several other global laboratories.

On another front, while there is currently high-volume production of simple, low data rate, low-cost optoelectronic integrated circuits (oeics), and major laboratory efforts toward more-complex, high-performance oeics were pursued through the late 1980s, economic tradeoffs did not justify movement of high-performance oeics to the commercial market. However, with the evolution to dwdm, low-cost parallel optical interconnect links, and highly complex optical crossconnect switches, a major resurrection of development efforts in this field is now proceeding. Lucent Technologies, Alcatel, the Canadian National Research Council, Texas Instruments, and numerous other world-class laboratories have recently demonstrated impressive oeics. The soa can be integrated into some of these complex circuits.

Greater component integration

As photonic and optoelectronic component technologies mature and as consumption quantities accelerate and price pressures increase, there is a trend to greater integration of component assemblies (see Fig. 2). Early component supply was in the form of individual components such as fiber splitters, isolators, and laser diode modules. Over the past few years, companies such as jds fitel and e-tek Dynamics have introduced "super-components," such as the input of an optical-fiber gain block that combines several components in fused-fiber, packaged assemblies.

A later element of this evolution has been the emergence of hybrid assemblies (chips on a substrate). Several developers, for example, including Motorola, Vixel Corp., and several consortia, have developed parallel optical channel transmitters and receivers that incorporate 10 to 32 laser diodes or photodiodes, with all required integrated circuits and fiber interfaces, in a smaller volume than previously needed for a single channel.

The final stage of this evolution will be the monolithic integration of a number of passive and active photonic/optoelectronic functions in a single III-V compound chip. Extensive government-funded development of this technology, aimed mainly at long-term future military/aerospace equipment, proceeded from 1980 through 1990 at Honeywell, McDonnell-Douglas, Hughes, Texas Instruments, and others. There were also significant developments in commercial laboratories, such as Nippon Telegraph & Telephone, ibm, Sumitomo, and Bell Laboratories. The small demand for these functions, however, did not justify the cost of commercializing these designs.

With numerous applications now pushing toward consumption of millions of units per year of III-V compound chips for photonic/optoelectronic monolithic integration functions (for fiber-to-the-home, local-area and premises networks, intra-enclosure, etc.), monolithic integration now looks economically attractive. For example, Lucent Technologies` Bell Laboratories, Crawford Hill, NJ, is developing a variety of monolithic integrated multichannel telecommunication re-ceivers. The integrated circuit accommodates the dwdm demultiplex filter, the photodiodes, and the related receiver integrated circuits. This receiver chip can detect 20 Gbits/sec per channel.

Fiber is forever

Unlike time-division multiplexing and wavelength-division multiplexing (wdm), fiber deployment will not collide significantly with technical performance limitations in the foreseeable future. The dominant singlemode telecommunications optical-fiber producers are Lucent Technologies, Corning, and their licensees. Although these two companies have achieved im-pressive production capacity expansion over the past five years, and further expansion is in progress, a significant shortage of singlemode optical fiber has developed. This, combined with the delay in the availability of OC-192 (10-Gbit/sec) optoelectronic transmitters, has pushed many network planners into accelerated de-ployment of dwdm systems.

Competition in the supply of conventional zero dispersion-shifted (zds) fiber is increasing, particularly from Al-catel, which is ex-panding its capacity aggressively. zds fiber will meet the needs of many current and future networks.

The fiber supply prospect is complicated by the introduction of non-zero dispersion-shifted (nzds) fiber by Corning and Lucent. nzds fiber patents held by these companies are reportedly complete enough to prevent competitive supply. nzds fiber currently is the optimal fiber for 1520- to 1570-nm dwdm. Its zero dispersion point is just below the band edge (Lucent) or just above the band edge (Corning), as shown in Fig. 3, making it feasible to use dwdm continuously across the entire band. It is feasible, however, to achieve substantial dwdm deployment with conventional zds fiber (which is 80% to 85% of the current globally deployed singlemode fiber) or the newer dispersion-shifted fiber.

Having witnessed the emergence over the past 20 years of three successive "ultimate" singlemode fibers, it is logical to assume that a fourth generation will be introduced within the next decade. Research toward this goal is now proceeding. Whether this fourth generation will be offered only by Lucent or Corning is not yet clear. Also, the supply of conventional zds fiber will continue its rapid growth.

ElectroniCast`s forecast of singlemode optical-fiber consumption in North America is a quantity increase of 23% per year through 2001, followed by a 10% per year increase from 2001 to 2006. It appears likely that there will be adequate supply to meet this demand.

Photonic switch and matrix overview

During the past 15 years, signal transport via optical fiber has progressed from the 45 Mbits/sec of at&t`s Northeast Corridor to equipment now being deployed that carries 10 Gbits/sec per wavelength. Terabit-per-fiber (transmission rate of 1 Tbit/sec over a single fiber) has been demonstrated. Currently, nearly all of this high-bandwidth transmission is switched by systems that convert optical signals to electronic, often demultiplex to parallel streams of much lower data rates, switch in semiconductor integrated circuits, multiplex backup, and reconvert to optical retransmission. This is limiting the real communications bandwidth, and the limitation will become more serious as bandwidth per subscriber multiplies over the next decade.

Although the advancement of semiconductor electronic switching speeds continues to confound forecasters, many applications will have an urgent need for purely optical throughput at nanosecond switch times for a modest cost per switch. This need has been foreseen since the advent of optical-fiber transmission. Tens of millions of dollars have been spent on research and development of optical switches in nonlinear optical crystals (such as lithium niobate) and polymeric film, semiconductor devices, movable mirrors that are micro-machined in silicon wafers, and other technologies. Wavelength-selective switching, using tunable/switchable wavelength transmitters or filters, and also demultiplexing to a single wavelength per fiber for space switching, are now being demonstrated. No technology to date has achieved all goals, but steady progress is being made, and several of these alternatives will phase into commercial availability within the next decade.

The forecasted expansion of network bandwidth capacity and throughput by more than twenty times over the next decade will require a major expansion of central office digital crossconnect and trunk add/drop multiplex capabilities. Selective optical bypass of a part of the signal flow around the digital crossconnect, and wavelength-selectable optical add/drop multiplexing, can substantially reduce the cost of this increased capability. Major photonic switch markets to both equipment vendors and carriers are being created by this trend. These systems are now in the development and early deployment phase.

The shift to signals on many wavelengths in each fiber, and dynamic reconfiguration of the active wavelengths in each fiber, are driving the need to monitor the status of these signals at many points in the network. This, in turn, will drive rapidly increasing use of 1¥N and M¥N photonic switches. Meanwhile, the tangible cost of a fiber cable cut will increase by a factor of about 200 times over the next decade, as shown in Fig. 4 (six-hour disruption of a long-haul trunk). This increase will be due to the dwdm-supported growth in number of channels per fiber, multiplied by the increased average data rates per channel, multiplied by the increased average use time per channel. The probability of cable cuts may be increased by vandalism and sabotage. Network operators will protect themselves against this risk by a major expansion of redundant route diversity cables and photonic matrix switches, ranging from major long-haul trunks to individual subscriber connections.

Integration affects photonic switches

As mentioned previously, there is a strong resurgence of research and development of high-performance oeics, many of which will incorporate optical switching functions. These switching functions will complement, rather than displace, the future applications of photonic space switches and wavelength-selectable switches.

The global consumption of photonic switches and switch matrices will rise dramatically from $94 million in 1996 to $3.2 billion in 2006. This will be driven by the rapid expansion of fiber-optic transport and access networks, mainly in telecommunications. Private data networks will also remain a major switch user. Most of the 1996 switch market value consisted of simple, conventional switches such as 1 ¥ 2, 2 ¥ 2, and 1 ¥ N configurations (see Fig. 5). The future growth of photonic switching, however, will be dominated by complex switch matrix systems. u

Jeff D. Montgomery is chairman and founder and Stephen Montgomery is president of ElectroniCast Corp., San Mateo, CA, usa. They may be reached at tel: (650) 343-1398, fax: (650) 343-1698, e-mail: [email protected]

Sponsored Recommendations

Innovating the network edge with 100ZR QSFP28: The next frontier in coherent optics

Jan. 15, 2025
In this webinar, Juniper Networks, EXFO and Precision Optical Technologies are teaming up to showcase the new 100ZR QSFP28 pluggable coherent technology, exploring its foundational...

On Topic: Metro Network Evolution

Dec. 6, 2024
The metro network continues to evolve. As service providers have built out fiber in metro areas, they have offered Ethernet-based data services to businesses and other providers...

State of the Market: AI is Driving New Thinking in the Optical Industry

Dec. 5, 2024
The year 2024 marked an inflection point for AI. In August, OpenAI’s ChatGPT reached 200 million weekly active users. Meanwhile, McKinsey reported that 72% of ...

Optical Transceivers in the Age of AI: Impacts, Challenges, and Opportunities

Jan. 13, 2025
Join our webinar to explore how AI is transforming optical transceivers, data center networking, and Nvidia's GPU-driven architectures, unlocking new possibilities in speed, performance...