Communications convergence requires network service providers to consider higher levels of provisioning with more bandwidth as well as how to reduce network overhead costs.
One of the most recent technology advancements in the area of optical networking is the small form factor (SFF) transceiver, available from Agilent Technologies and other major manufacturers. By combining this device with next-generation serializer-deserializer (SerDes) ICs, design engineers are able to increase board density, alleviate board layout issues, gain confidence in their EMI (electromagnetic interference) performance and reduce their cost and time-to-market.
The SFF transceiver is rapidly becoming the platform of choice for next-generation network connections where high port density is an issue.
More than 1 million MT-RJ SFF transceivers have sold into Fast Ethernet and OC-3 ATM/Sonet applications, and transceivers capable of higher data rates are continually being introduced to the market. In sum, the SFF transceiver is ramping faster than any other optical transceiver platform to date.
SFF transceivers are now available for Ethernet applications ranging from 10 Mbits/second to 1 Gbit/s, and for ATM/Sonet applications ranging from OC-3 to OC-48. While the SFF transceiver has solved the port density issue at OC-3 and Fast Ethernet speeds, the benefits are even more compelling at Gigabit Ethernet data rates and above.
In low-port-count applications, the SFF transceiver paired up with a single-channel SerDes device gives the design engineer added flexibility in a board layout. With the smaller package size, the design engineer has more room to place support circuitry and other components on the card.
When higher port densities are required, the SFF transceiver with a quad-channel SerDes IC allows the engineer to fit twice as many transceivers on a card than its predecessors-the 1X9 SC duplex form factor and the single-channel SerDes IC. What was a multiple-board solution can be a single-board solution, which reduces end-product costs by eliminating duplicate supporting circuitry. Since cost-per-port is an important metric in these markets, reducing overall board cost is a major advantage.
If a network element such as a router or switch has Gigabit Ethernet or OC-48 ports, for example, the ASICs and processors inside the switch are typically running much faster. Since the SFF transceiver is the I/O interface to the switch, and was designed to be a panel-mount solution, it was developed not just to run quieter by minimizing EMI itself but also to have the ability to shield other noisy parts from the outside world.
As port densities and data rate both increase, the ability to minimize EMI becomes a very valuable feature that enables engineers to quickly design and lay out physical-layer interfaces.
Lowering the number of printed-circuit boards also reduces end product cost, and, in some cases, an entire board can be eliminated. When the SFF transceiver is used with the quad SerDes IC, two eight-port cards can be consolidated onto a single 16-port card design. The OEM can now provide the equivalent port density of two SC duplex pc boards on a single SFF transceiver board.
This means that OEMs reduce their product cost by eliminating pc-board material, connectors, power converters and various other components that are required to simply support the connection of the board to the chassis. Increased port density not only reduces costs, but it also opens slots for additional fabric switch cards, router cards, or interface cards.
Depending on the manufacturer's design, SFF transceivers can offer more robust EMI performance than any other fiber optic transceiver in the market.
Component manufacturers have focused heavily on this aspect of the transceiver design and are bringing robust EMI solutions to the market.
The MT-RJ connector, available from connector suppliers like AMP, offers EMI advantages itself. The MT-RJ connector uses a plastic ferrule, which allows the component manufacturer to EMI-isolate the front of the package from the back of the package. This feature provides a definite advantage over SC duplex and some LC SFF transceiver designs.
The introduction of silicon microbench technology into SFF transceiver design has also improved EMI performance. Silicon microbench allows the manufacturer to move away from the traditional TO-can optical subassembly that is found in both LC and SC duplex connectors that employ a ceramic ferrule.
Almost all ceramic-ferrule connectors used in transceivers have a TO-can packaged optical subassembly. While these cans are typically plastic for multimode and low-data-rate single-mode parts, they are usually metal for higher data-rate single-mode parts. The TO cans used in both SC duplex and SC manufacturing typically protrude through the EMI barrier and into the front of the part and encapsulate the ferrules. As a result, the TO cans create a conduit for EMI to travel from the inside of the package to the outside world.
The silicon microbench process employs a fiber to connect the ferrule to the laser diode. Since neither the glass nor the ferrule is conductive, no EMI is transferred. This offers a major advantage over most SC duplex and some LC connectorized transceiver designs.
In addition to being able to remove the TO cans, the silicon microbench process allows the component manufacturer to completely separate the transceiver body from the snout (the nose of a transceiver that fits through the panel). The only required opening through the barrier is used for two fibers, which connect to the MT-RJ plastic ferrule on the other side. By isolating the front of the package completely from the back, the design engineer can connect the snout to the chassis ground while connecting the back of the transceiver body and the electronics inside the signal ground.
The snout on the SFF transceiver is also more robust than that of the SC duplex. It has been designed to make chassis (panel) contact on all four sides. This helps to provide a robust shield, which keeps the EMI inside both the transceiver and the entire network element from leaking out. The SC-duplex part does come in metalized variations, but the shields are not industry-standard. Additionally, the outside of the package cannot be separated between the back and the front. Finally, most SC-duplex shields do not make contact on all four sides of the box.
The SFF transceiver offers a smaller aperture as well. One of the SFF connectors (the MT-RJ connector) allows for the smallest panel opening, which enables the spring clips around the entire snout to make chassis contact on all sides. The four-sided contact allows for a mechanical septum to be placed between parts on the panel, even when they are placed side-by-side. This guarantees connection to the chassis on all four sides.
When stacked side-by-side in high-port density applications, other SFF connector transceivers, such as the LC, only make connector-to-chassis contact on two sides of the interface-the top and the bottom. The other two sides provide only connector-to-connector contact.
By developing new ICs and optics that can operate at 3.3 V, component suppliers have enabled design engineers to eliminate the 5-V rail from boards all together. This reduces system cost in several ways: (1) the power supply is simpler; (2) fewer power planes are required, which reduces the number of layers in a circuit board; and (3) lower power consumption of 3.3-V components makes thermal management easier.
The power reduction created by migrating from a 5-V GBE transceiver to a 3.3-V transceiver can be as great as 40 to 50 percent. This is power that can be put to greater use for higher-performing ASICs, added features and/or services, or increasing the number of optical interfaces that the network element can support. This enables the switch, hub or router to provide added functionality at higher data rates.
A complete physical layer solution is an important contribution from component suppliers. It provides component compatibility and simplification of design, which allows the design engineer to spend more time on the core of the network element. Design engineers can spend more time focusing their efforts on developing features and services that differentiate their solutions in a competitive market.