The explosion in the telecom marketplace means good news for connector and cable-assembly manufacturers.
This increasing bandwidth use at the network's edge is exponentially increasing bandwidth requirements in the core. We're seeing ultra-high-speed Synchronous Optical Network transceivers for transport, Asynchronous Transfer Mode (ATM) switching in Internet service provider points of presence and in the core network, and faster server uplinks--i.e., Fast Ethernet and ATM. And we're seeing the advent of Gigabit Ethernet and next-generation ATM, which together are bringing more-powerful capabilities to the LAN, LAN backbone, and wide area network (WAN).
With its roots in LAN fiber backbones and server farms, Gigabit Ethernet technology now offers exciting possibilities for configurable 10/100/1000-Mbit/sec desktop adapters. As for ATM, even though it will probably share a LAN backbone role with Gigabit Ethernet, it is clearly becoming increasingly dominant in the WAN.
Since Ethernet and Fast Ethernet protocols are already deployed in 83% of corporate networks, Gigabit Ethernet has emerged as the next logical industry-standard solution for high-speed local-area networking. Similar to the emergence and acceptance of Fast (100-Mbit/sec) Ethernet, Gigabit Ethernet, although in its infancy, is most applicable for aggregating network backbones and server farms. It may also be deployed directly to the desktop for certain power users.
There are several options for implementing the physical medium that will carry Gigabit Ethernet signals. These options include singlemode and multimode fiber as well as variations of standard unshielded twisted-pair (UTP) copper wires. The IEEE standard for Gigabit Ethernet over fiber, 802.3z, was finalized in June, and fiber cable is already being used in network backbones. Depending on the type, fiber can carry high-speed Gigabit Ethernet signals across distances of up to 2500 m.
How does Gigabit Ethernet affect ATM for LAN applications? To begin with, Gigabit Ethernet is not a replacement technology for ATM; rather, it is an alternative to ATM in the Ethernet environment. Actually, the two technologies can coexist-LAN packets can be passed over an ATM network using LAN emulation, MOPA (master oscillator power amplifier), or conventional routing. In terms of availability, ATM has a head start over Gigabit Ethernet, which is in the early stages of product development.
ATM also has established a track record with wide deployment. ATM technology is superior to Gigabit Ethernet in many ways. However, it has been more expensive to implement. ATM handles all forms of traffic, including data, graphics, imaging, video, and voice. Ethernet technology, on the other hand, is designed for data transfer. It is not designed to support multimedia applications, especially voice and video, which are sensitive to delay. To migrate to an ATM-LAN environment, users must replace Ethernet switches with ATM switches.
But what about Fiber Distributed Data Interface (FDDI) campus and building backbones? Is there a Gigabit Ethernet solution for them? Yes, because conversion from FDDI to Gigabit Ethernet can be achieved by replacing the FDDI concentrator or hub, or an Ethernet-to-FDDI router, with a Gigabit Ethernet switch or repeater.
Like initial Gigabit Ethernet implementations, Fibre Channel is aimed at server-to-server connections. The lower levels of the Gigabit Ethernet standard are also based on Fibre Channel. Over the next couple of years, the big question will be how Fibre Channel will be differentiated from Gigabit Ethernet. From a market perspective, Gigabit Ethernet has a large advantage since users are already familiar with Ethernet at slower speeds. Pricing for Gigabit Ethernet and Fibre Channel is likely to be similar.
Realistically, Fibre Channel vendors do not expect to overtake Gigabit Ethernet--but they assert that the technology is well suited for specific applications. The first markets for gigabit Fibre Channel will be very specialized fields such as computer-aided design, medical imaging, and video editing. It will excel in remote storage applications, since it has a range of 6 mi.
Currently, data transmission in premises networks, short-reach (<100-m) telecommunications, and computer interconnects with data rates of 1 Gbit/sec or lower are dominated by copper. However, an increase in data rates demands the use of optical fibers. Data transmission for short distances at data rates up to 12.5 Gbits/sec has been demonstrated previously on multimode glass optical fiber. Recently, low-loss (<100-dB/km) plastic optical fiber has been introduced with very large bandwidths (<1 GHz km) in the data-communication widow. Data transmission over 100 m at up to 11 Gbits/sec using such fiber has already been reported.
Yet, while optical fiber may indeed be capable of incredible bandwidth, it does not mean it will push copper wire out of the communications market, at least not for many years. The inability of fiber-optic-component vendors to close the cost gap with copper and continual improvements in Category 5 cable as well as proposals for yet higher levels have, for the most part, kept fiber confined to the backbone. Fiber-optic interfaces at the desktop won't reach 25% of the market before 2001. That said, even without fiber-to-the-desk happening in a big way anytime soon, the demand for multimode components in backbone installations will continue to rise at about a 20% annual rate.
Reflecting Ethernet's large installed base, the fastest-growing market for fiber-optic transceivers today is for Fast Ethernet backbone implementations. Cost remains a major obstacle to fiber's success at the desktop. While estimates vary considerably, fiber installations still run anywhere from 20% to more than 50% higher than comparable copper installations. Anticipation is building for the higher port densities and lower costs that will come with a new generation of transceivers that comply with smaller-footprint packages.
In the next couple of years, three different small-form-factor (SFF) transceiver/connector architectures will vie for industry acceptance as the best way to squeeze more ports onto a printed-circuit-board module. AMP, Hewlett-Packard, US Conec, Siecor, Fujikura, and others are offering the MT-RJ, a compact connector that combines two fibers into a single molded ferrule. The connector doubles the port density of a hub card relative to the traditional 568SC or ST connector, and at only half the height, it allows developers to stack transceivers.
Transceivers based on the MT-RJ are half the size of the current industry-standard 1x9 transceiver with a duplex SC interface. Their key attraction is that they offer a transceiver port density equal to the 0.55-in density achievable with the traditional RJ-45 copper interface. As a result, the optical interface will no longer be a limiting factor on port density in hubs, switches, and I/O cards. Ultimately, increased fiber-optic port density should result in lower cost per port.
More than 30 manufacturers have developed products using the MT-RJ interface; together, these manufacturers offer transceivers running from 10 Mbits/sec to gigabit-per-second rates. Late last year, Hewlett-Packard brought out a broad family of MT-RJ optical transceivers, including 850-nm vertical-cavity surface-emitting laser and 1300-nm Fabry-Perot laser-based devices. All comply with the IEEE802.3z Gigabit Ethernet standard for use in 1000Base-SX and 1000Base-LX applications.
In competition with the MT-RJ, 3M is offering Volition, a cabling alternative that features the VF-45 connector. The VF-45 is a ferrule-less interface that reduces cost by sliding the cable into molded plastic V-groove clamps rather than using precision metallic parts. While the same size as the RJ-45 modular jack, the VF-45 comes in at one-seventh the cost of traditional fiber connectors. VF-45 transceivers are available in 1x5, 1x9, 2x5, simplex, and quad-pack pin configurations.
A dozen vendors joined together late last year to create an independent organization to promote the use and deployment of VF-45 interconnect technology. Among the VF-45 Action Group's members are 3M, BATM Advanced Communications, Corning, Infineon Technologies (formerly Siemens Microelectronics), Phobos Corp., RACORE Technology Corp., and Sumitomo Electric Lightwave. 3M's strategy is to license its technology and continue working with both electronics and non-electronics manufacturers to implement VF-45 technology.
Another SFF connector is the LC, developed by Lucent Technologies. It is ferrule-based and uses an insertion/release mechanism similar to an ordinary telephone plug. The connector's excellent return loss has made it attractive for longer, singlemode applications.
While the overall North American connector, cable-assembly, and backplane industry experienced a decline of 3.3% in 1998 compared to 1997, fiber connectors, couplers, adapters, and cable assemblies increased 14.4%. Total 1998 glass optical-fiber connector sales in North America were $1.2 billion. Of this total, singlemode accounted for $487 million, multimode for $522 million, application-specific for $122 million, and pre-connectorized cables and cable assemblies (pre-wired shelves) for $256 million.
Singlemode-fiber connectors, couplers, adapters, and cable assemblies in these market numbers include, by design type, in descending volume sequence, the SC, FC, ST, biconic, and D4. Other designs with lower volume include Quick-Connect designs the LC, FC-APC, E-2000, SC-APC, DIN-APC, and D3. By design type, the leading singlemode-cable assemblies in descending volume sequence were the SC, FC, ST, and D4. Other designs with lower volume include biconic, application-specific, military, E-2000, LC, and SC-APC.
Multimode-fiber connectors, couplers, adapters, and cable assemblies, in descending 1998 volume sequence, were the ST, SC Simplex, SMA, Quick-Connect, and FC. Other designs with lower volume include biconic, military, E-2000, and mini-BNC. By design type, the leading duplex/multifiber connectors in descending 1998 volume sequence were FDDI, MPO/MPX, SC duplex, and Escon. Other designs with lower volume include the MT, MAC, mini-MT, backplane connectors, MU mother/daughtercard connectors, LC, MT-RJ, VF-45, SC/SD, and OptiJack.
The highest growth is occurring in the duplex/multifiber connectors. In 1998, total sales were $86 million and are projected to increase to $287 million within five years. The hot product lines are the MT-RJ, VF-45, and LC. The MT-RJ, which was developed by AMP and the consortium of manufacturers, has been focused on premises equipment applications. For example, AMP focuses on the hub side of the industry (such as Cisco); thus, most of its MT-RJ output is in cable assemblies (usually 3 m in length). On the other hand, Siecor also offers the MT-RJ and focuses on the socket side as well as the cable assemblies. Molex is a licensee and is expected to introduce the product shortly. Currently, about 12 licensees exist, with a projected 15 licensees by the end of 1999.
Meanwhile, the top 10 fiber-optic connector manufacturers in North America, from the top down, are Lucent, AMP, 3M, ADC Telecommunications, Siecor, Molex, Alcoa Fujikura, Fiber Optic Network Solutions, Amphenol, and Methode. On a worldwide basis, there are now three manufacturers that have reached the $100-million shipment level. In 1998, Lucent and AMP (nearly equal in volume) were approaching $200 million in sales. 3M also reached $100 million in shipments. All amounts and rankings exclude transceivers.
In fiber-optic technology, development is proceeding on "fiber-in-board"--the inlay of fiber channels within the plane of either rigid printed circuit boards or flexible boards. One such development program is at the GEC-Marconi Research Center in the United Kingdom, where scientists and engineers at the center have been engaged in advanced development of fiber-in-board technology. GEC offers application in optical backplanes by combining multilayer printed-circuit- board fabrication with fiber-optic techniques. Also, the routing of fibers within the board are adjusted to each application. One backplane available is a 64-way, singlemode-fiber backplane incorporating a 16x16 integrated optic-star coupler. Prototypes have been developed for up to 10 layers of fiber on a board about one-half inch thick.
Other manufacturers have developed fiber-in-board technology. One such example is Lucent, which has developed fiber-in-board (referred to as OptiFlex) for its own use. Nippon Telephone & Telegraph Electronics (Japan) has an optical wiring board, which has a flex structure as opposed to semi-flex or rigid. AMP, likewise, has expended major research efforts and can introduce product when required by an original equipment manufacturer.
Fleck Research forecasts that fiber-in-board optical backpanel sales will reach $87 million worldwide within five years. Within 10 years, sales of optical backpanels will be astounding.
Fiber-optic connector sales may be growing, but price erosion is a major drag on the business. In singlemode ST fiber-connector designs, price erosion continued in 1998 with most suppliers reporting up to a 15% decline. Prices for multimode ST connectors are usually in the $1.75 to $2 range. But on large volume buys, 1998 often saw prices fall into the $1.35 to $1.45 range, and some parts may be tagged even less. Price erosion in Europe was even greater than in the United States in 1998--some 30%. At these low prices, there are no special termination features (enhancement features to make field installation easier).
Gigabit Ethernet is the proposed solution for prolonging the life of Ethernet LAN technology into the next century. Gigabit Ethernet expands Ethernet bandwidth by a magnitude over Fast Ethernet. The Gigabit Ethernet switch market will grow to $1 billion by 2000.
The Gigabit Ethernet market will be driven by user needs for faster response times, more segment capacity, and substantial improvements in backbone bandwidth and server performance. For example, desktop computers are increasing in speed at a rapid rate. PCI is the interface of preference because it runs at high speeds. A 32-bit PCI runs at 1.056 Mbits/sec, while the new 64-bit PCI doubles that speed. Also, workstation and server performances are advancing, enabling these devices to flood multiple Fast Ethernet links with network traffic.
Gigabit Ethernet, like Fast Ethernet, borrows an established physical-layer standard from another technology. While Fast Ethernet has adopted a version of the FDDI physical-layer standard, Gigabit Ethernet will use a modified version of the ANSI X3T11 Fibre Channel standard physical layer (FC-0). Why Fibre Channel? Because Fibre Channel is currently the only technology that supports gigabit speeds. It currently supports rates up to 4.268 Gbits/sec. ATM currently tops out at 622 Mbits/sec (OC-12) but will be extended to 2.5 Gbits/sec (OC-48).
Gigabit Ethernet's initial draft specification specifies the transmission medium as multimode or singlemode optical fiber. Category 5 copper media are expected to be supported when Gigabit Ethernet is used at the desktop-this will probably be 4-pair Category 5 cable limited to 100 m running at 1.25 Gbits/sec. The Electronic Industries Alliance/Telecommunications Industry Association (Arlington, VA), however, recommends fiber for backbone applications.
Gigabit Ethernet technology will be deployed initially in a campus or building environment as a solution to bandwidth performance limitations where volume traffic conditions exist. Gigabit Ethernet to the desktop is not expected in the early phase of deployment. Deployment will take place in backbones, switch-to-switch links, and in switch-to-high-performance server links--users will migrate from Fast Ethernet or Fiber Distributed Date Interface to Gigabit Ethernet.
Ken Fleck can be reached through the Fleck Research Website at www.fleckresearch.com.