Anita S. Becker
The higher port density and lower per-port cost of the small-form-factor (SFF) platform make an industry-wide migration away from traditional, one-inch connector systems a natural progression. In this evolution, SFF transceiver manufacturers are playing a critical role in determining which connector systems will survive the hyper-competitive new market.
SFF optical transceivers measure 0.535 inches wide and take up half the printed-circuit-board space of conventional SC duplex transceivers in fiber-optic data and telecommunications systems. These smaller-sized transceivers allow system designers to fit twice the number of transceivers onto the same board-doubling port counts currently available in fiber-optic hubs, routers, and switches from a maximum of 12 ports to 24. The SFF components themselves enable port spacing similar to RJ-45 copper interfaces.
By freeing board real estate, SFF optical transceivers can be used to create more-effective designs. Overall, less components and mounting hardware are needed with this technology. All of these factors together lead to lower cost.
In the year since SFF transceivers began their entry into the marketplace, they can now be found in next-generation systems on the drawing boards of leading networking-equipment vendors. SFF technology currently enables line rates up to 2.5 Gbits/sec for applications that include Fast Ethernet, Gigabit Ethernet, Asynchronous Transfer Mode (ATM), Fiber Distributed Data Interface (FDDI), Fibre Channel, and Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH). Some companies, such as Lucent Technologies Microelectronics Group, are working on developing capacity for 10-Gbit/sec applications.
Early in 1998, AMP Inc. (Harrisburg, PA); Hewlett Packard Co. (HP-Palo Alto, CA); Lucent Technologies Microelectronics (Allentown, PA); Nortel Networks (Brampton, ON, Canada); Siemens AG-Fiber Optics (New York City); and Sumitomo Electric Lightwave Corp. (Research Triangle Park, NC) signed an SFF multisource agreement (MSA) that describes the smaller fiber-optic transceiver package outline and pin definition. The agreement specifies 10-pin and 20-pin configurations, all of which support today's leading connector styles. Connectors were deliberately excluded from the specifications, and the marketplace was left to sort out which of the many styles would become predominant. The MSA also does not specify the speed of transceiver operation, optical performance, or specifics of multimode- or singlemode-fiber operation.
The lack of a connector specification makes the SFF transceiver agreement unique, says Stan Swirhun, CEO for optical transceiver manufacturer Picolight (Boulder, CO). "I believe this is the first fiber-transceiver industry agreement that simply describes the package outline and therefore allows for multiple optical receptacles...and this is why there are products with LC, MT-RJ, and VF-45 optical interfacesellipseand why there is competition among them."
The leading optical-connector options today include the LC, developed by Lucent Technologies; the MT-RJ by AMP, HP, US Conec, Siecor, and Fujikura Ltd.; the SC/DC by IBM and Siecor; and the VF-45 by 3M. Most suppliers have announced active and passive components for one or more of these four connector systems, even though other systems are also available, in clud ing Opti-Jack from Panduit Corp., the LX.5 from ADC Telecommunications, and the MU from a variety of Japanese manufacturers. So far, the leading connector systems seem to be the LC, MT-RJ, and VF-45, with the MT-RJ being the smallest of the four options.
The LC and MT-RJ are the only systems that can currently support both singlemode and multimode applications. However, developments are being made to support faster speeds and longer distances with the VF-45 system, which currently is used primarily in premises and campus backbone net work ing applications.
Although there are major benefits to working within the SFF, designers face a few challenges as well. The two most prominent are heat dissipation and elimination of electromagnetic interference (EMI), as more components running at higher speeds are squeezed into tighter spaces. The intrinsic small optical apertures of the connector ferrules (except for the ferruleless VF-45) and the manufacturer design features such as metal transceiver housings reduce these problems. Also, SFF transceiver manufacturers are offering 3.3-V products that consume less power than their 5-V predecessors.
HP offers a special option that will help designers deal with the radiation problem in high-density cases. "For those really high-port-density applications, we've put in a nose shield that sits on top of the MT-RJ connector on the transceiver," says Adam Carter of HP's business-development manager, fiber-optics communications di vision. "This [feature] basically stops any radiation coming out of the holes in the panel."
Because of the VF-45's ferruleless design, Infineon Technologies, a sub sidiary of Siemens (Munich, Ger many), designed an efficient metal transceiver housing. "We now have a metal package that sinks the heat away from the packet element," explains Jerry Sheridan, product marketing manager for small form factor, North America.
Designed by Bell Labs, the LC is a high-density connector for fiber-optic applications used in both public and private networks and is similar in style to the RJ-45 telephone jack. Used for singlemode- and multimode-fiber applications, the LC links the transceiver to the optical network through an LC duplex receptacle. Lucent Technologies, Methode Electronics, Molex, and Sumitomo, are manufacturing their respective SFF transceivers to fit with the LC duplex receptacle.
Startup Picolight also chose the LC receptacle for its new 1.25-Gbit/sec and 2.5-Gbit/sec transceivers. "Pico light is generally 'connector agnostic,'" says Swirhun. "Instead, we respond to our customers' desires, and we saw enthusiastic customer interest to both the LC and MT-RJ."
This year, Picolight introduced its first multimode 1.25-Gbit/sec fiber-optic transceivers with the LC receptacle. "[We did this] partly because there are other companies, Lucent and Sumitomo among them, with compatible singlemode transceivers or plans for them. And partly because we saw a positive response to this receptacle from two of our three market segments: telecom switch providers and storage-area-network providers," explains Swirhun. "We believe that, ultimately, MT-RJ will also be supported at the gigabit-per-second data rate by multimode and singlemode solutions. We also believe that VF-45 may remain a multimode solution only."
The Microelectronics Group insists the primary reasons for supporting the LC were not because Lucent Technologies developed the connector. "We had several reasons for choosing the LC connector," says Arlen Martin, marketing manager for transceivers, Microelectronics Group. "One is the wider separation of the fiber, the pitch between the fiber ports, that helps with EMI and crosstalk designs between the transmitter and receiver. And the LC is the only connector that is really shown to work for all applications, not just the lower-speed FDDI and ATM, but also for singlemode fiber at the higher-speed ATM, Gigabit Ethernet, Fibre Channel, and OC-48 applications."
SpeedBlaster is the most recent addition to Lucent's family of NetLight SFF transceivers. NetLight transceivers are used in Gigabit Ethernet, ATM, SONET/SDH, and Fibre Channel applications. It provides bandwidth of 155 Mbits/sec and 622 Mbits/sec up to 1.2 and 2.5 Gbits/sec.
SpeedBlaster is the only transceiver in the family to combine the speed of 2.5 Gbits/sec with the SFF footprint. The 20-pin version of the transceiver provides optional features such as clock-recovery functionality and laser-diode monitor and control.
"We're not going to stop at 2.5 Gbits/sec," says Martin, adding that plans include 10-Gbit/sec products. "From a transceiver platform, which includes the connector, we need the capability of going to higher speeds."
"We've supported the MT-RJ connector from day one," says HP's Carter. "We are looking at proliferating this connector style out in the future for certain data rates. At the higher data rates, there are some technical issues that we would have to get over within the module to be able to offer something such as an OC-48 transceiver that meets all the SONET/SDH standards. That's something we're looking at. And we think the MT-RJ transceiver can do it."
"We've concentrated on getting the premises-networking product out first for Fast Ethernet and Gigabit Ethernet and for ATM at OC-3 data rates [155 Mbits/sec]," Carter says. "What we're trying to do is get into the hands of our customers a product that they'll be able to do OC-3 to OC-48 in singlemode fiber. Then, just by changing a laser and maybe the odd integrated circuit, you could also have in the same package style and with similar footprints the OC-3 or a Fast Ethernet part."
HP has MT-RJ products for Fast Ethernet and ATM OC-3 applications and is developing products for Gigabit Ethernet, Fibre Channel, and ATM OC-48. HP's MT-RJ fiber-optic transceivers are designed into network interface cards that are now available from Canary Communications Inc., IMC Networks, and Racore Technology Corp. To support its communications products, HP recently launched Agilent Technologies, a new company formed from its test-and-measurement solutions and com muni cations-com ponents divisions. The MT-RJ gained two more supporters when Infineon/Siemens and Sumitomo joined the ranks of the five original consortium members.
Infineon Technologies is shipping a new SFF multimode fiber-optic PIN-preamplifier trans ceiver suitable for Ethernet, Fast Ethernet, and 4/16 token-ring applications. The new multimode 1x5 transceiver product rounds out the supplier's growing VF-45 product offer ing. "We can supply the full multimode range of products today and will be able to support singlemode applications by first quarter of 2000," says Infineon's Sheridan.
Infineon chose to support the VF-45 con nect or because of price. "We found that the VF-45 would always realize at least a 20% cost advantage in a manufacturing cost over the com peting designs," says Sheridan. "That was primarily due to two reasons: (1) It is a ferruleless design, using a low-cost, injection-molded V-groove fiber interconnect system; (2) the alignment from the fiber to the active transceiver components is done with the V-groove, eliminating some of the high-tolerance, mechanical devices needed to align the fibers in some of these competing technologies."
Two other manufacturers are designing and delivering transceivers with the VF-45 interface. According to 3M, Honeywell Inc. (Minneapolis, MN) and Sumitomo will offer transceiver products ranging from 10 Mbits/sec through gigabit.
The transition from today's connectors to SFF concepts can be expected at differing rates depending upon the application. Lucent's Martin says his company is no longer designing transceivers for the SC duplex: "We haven't developed any new 1-inch-wide versions because we believe that everything over time will migrate to the smaller form factor." Infineon's Sheridan agrees: "The small form factor will just take over as the engine for bringing fiber out of the backbone to the desktop and through the whole business."
Market dominance of one or two connector styles will depend on further active-component manufacturer backing. Stephen Montgomery, president and director of the components group for ElectroniCast, a San Mateo-based market-analysis firm, compares the transceiver manufacturers' role in the so-called "connector fight" to what happened in the videocassette-recorder industry with Sony Beta and VHS technology.
"When it comes right down to it, a lot of customers may feel that connectors are a commodity, which means same price, same size, same function, same attributes-give or take here or there," says Montgomery. "They don't care about the connectors; what they care about is they don't want to buy a Sony Beta. They want to make sure what they have, and the transceiver company that they're buying it from, will be in business and will be able to supply their needs when they need to buy 100,000 units four years from now."
Montgomery adds that customers are also looking for vendors that will be able to incorporate other types of functions such as dense wavelength-division multiplexing into transceivers in a single, integrated package. "What this means is the small form factor is eventually going to get smaller," predicts Montgomery. "These are the issues that a lot of the board-level companies are looking at, as are the OEMs [original equipment manufacturers] who are selling the larger systems. They need to bring the footprint size of the entire systems down much, much smaller."
This article appeared in the October 1999 issue of Integrated Communications Design, Lightwave's sister publication. Anita S. Becker is associate editor of ICD.