What to expect in next-generation gigabit interface converters
Bryan Gregory
Molex Fiber Optics Inc.
The GBIC industry standard allows users to cut component costs and plug in an array of optical interfaces from multiple vendors.
The data-communications market is evolving from electrical interfaces to technology that is increasingly dependent upon fiber optics. The industry`s migration to fiber optics is driven by the need for faster transmission and greater link distances. As data rates increase, fiber optics become essential to support adequate link distances. In gigabit-speed networks, such as Gigabit Ethernet and Fibre Channel, standards support link distances of about 25 m over coaxial cable, compared with up to 10 km over optical fiber. Proprietary optical interfaces can extend link distances to more than 70 km.
Next-generation networks will operate at speeds of 2 to 10 Gbits/sec. At these speeds, optical interfaces will probably replace electrical lines as the preferred connection for even the shortest links. As the optical data-communications market grows, the cost of optics will continue to drop, making this medium extremely attractive.
Optical interfaces today
The current generation of optical interfaces consists of a wide range of physical-layer implementations. Common optical transceivers use different light sources: 10- or 100-Mbit light-emitting diodes (LEDs), 1-Gbit vertical-cavity surface-emitting lasers (VCSELs), long-wavelength Fabry-Perot (FP) lasers, or distributed-feedback (DFB) lasers used for trunking. Each of these optical interfaces is incompatible with the others. Similarly, each addresses a separate need and requires a different installation model.
Aside from variations in the optics, there are also distinct package styles from a variety of fixed pin-in-hole mounted transceivers to pluggable optical "cards." Knowing which optical interface to choose requires an extensive knowledge of the end user`s basic optics, production needs, and physical layout requirements.
Because optical transceivers are one of the most expensive components in a data-communications system, managing the design and production planning is a significant undertaking. Pin-in-hole 1ٻ transceivers are currently the most common package style. These devices are inexpensive and easy to design into a system. If a system needs only one type of optical interface, the device of choice is usually a 1ٻ transceiver. Mounting these modules usually requires wave soldering to the printed circuit board.
A typical data-communications system, such as a large Gigabit Ethernet switch, may use several different types of optical transceivers. Even though all these transceivers might be contained in a "standard" package, such as a 1ٻ, each type usually requires modifications to the interface circuit and hence different bills of material. If multiple sources are used, it is sometimes necessary to modify the circuit for each supplier. For example, a Gigabit Ethernet card using four types of 1ٻ transceivers (1000Base-SX, 1000Base-LX, 100Base, and 10Base), with two suppliers for each device, may require a number of bills of material.
Managing inventory and lead times for a system with multiple interfaces is a complex undertaking, especially when the optical transceivers must be ordered from different vendors. If standard systems with fixed numbers of optical ports are produced rather than user-specified port counts, the customer may have to buy more optical interfaces than the application actually needs, which significantly increases the overall system cost.
Pluggable modules
The Gigabit Interface Converter (GBIC) industry standard was introduced to respond to these issues. The GBIC (pronounced "gee-bic") is a module that plugs into a system similar to the way a PCMCIA card plugs into a computer (see Photo 1). GBICs are available in several varieties, from electrical units with DB9 and HSSDC coaxial interfaces to short-wavelength modules with VCSELs, long-wavelength devices with FP lasers, and even long-reach 1550-nm DFB-based units. These modules support transmission distances in the range of 25 m to about 70 km.
The GBIC interface offers many advantages to the end user. Just as a PCMCIA card enables notebooks to have custom-configurable ports, a GBIC interface allows a Gigabit Ethernet or Fibre Channel system to be configured to specific requirements in the installed environment. If the end user needs to set up a combination of electrical and optical links, it is easy to plug in the appropriate modules and attach the cables. If a customer only needs one optical interface and the system has multiple GBIC ports, then he or she only needs to install--and pay for--one device. As the network grows, users can easily add GBICs without upgrades or even a short-term system shutdown. Once a system is set up in the field, if a device should fail, the user can replace it on the spot.
GBICs also offer advantages to the system providers. As an industry standard, a circuit using GBIC input/output ports should be able to accept any type of GBIC module from any vendor. Unlike 1ٻ optical transceivers, which only comply with a very loose industry standard, GBIC modules are carefully defined. This standardization offers equipment providers greater control over inventory and system costs, while at the same time providing future business opportunities for system upgrades.
Even though most of the industry is already treating the GBIC specification as a finished document, the GBIC standard group is still finishing the final draft (Revision 5.1a). Most of the market, however, is already producing optical transceivers and systems based on the standard, despite the few minor issues that are delaying the completion of this document.
The next generation
Current GBIC devices are available with a duplex SC interface. This interface is robust and affordable, but it is also large. The next generation of pin-in-hole, small-form-factor (SFF) optical transceivers is beginning to adopt optical interfaces such as the MT-RJ and LC (see Photo 2). These modules are similar to the 1ٻ in functionality and mounting requirements. Like the 1ٻ, they require pin-in-hole mounting and wave soldering. However, these devices allow much greater port densities. The bezel opening required to accommodate an SFF module is similar to that of an electrical RJ-45- type connector.
There is no standard for pluggable SFF transceivers. But a number of optical transceiver suppliers are discussing "mini-GBIC" style modules. These mini-modules will be the same size as current SFF transceivers, but will mate with a surface-mountable rail mechanism that allows full GBIC-style functionality in a very small size. Beyond the basic GBIC functionality, the next generation of pluggable modules is likely to offer a much wider variety of formats. As data-communications standards evolve to include wavelength-division multiplexing and single-fiber bidirectional functionality (several wavelengths transmitted in different directions on a single fiber), mini-GBIC modules may offer upgrade paths that will help futureproof systems. A "mini-GBIC" multisource agreement is expected before the end of this year.
In addition to supporting data-communications standards such as Fibre Channel and Gigabit Ethernet, pluggable modules should support Synchronous Optical Network and Asynchronous Transfer Mode applications. Whereas mini-GBICs will offer serial ID functions, telephony variants may include functions such as clock recovery and additional alarms. These modules will offer unique flexibility to optical access and enterprise telephony systems.
Optoelectronic convergence
Traditionally, optical transceivers were designed by semiconductor or electronics companies, and passive optical connectors were conceived by vendors specializing in connector technology. There was not a lot of synergy between the two groups.
The push toward smaller transceivers and tighter integration of the electronics behind the optical interface necessitates a deep understanding of both fields. Next-generation transceivers and connectors, such as products based on the MT-RJ and LC interfaces, will have to be designed in conjunction with each other. As a result, suppliers that offer both passive and active components will have a distinct advantage. Data-communications system manufacturers that manage both the active and passive qualification under the same group will likewise have an advantage. Decisions affecting one area will affect both. It is even possible that future cabling will include a built-in optical or electrical interface. At that point, it will be impossible to separate the passive and active technologies. This level of integration may be the next logical step beyond today`s transceiver cards for the most common and least expensive transceivers. q
Bryan Gregory is the marketing manager of optoelectronics at Molex Fiber Optics Inc. (Downers Grove, IL).