Laser advances mark optical transceiver trends
The development of VCSELs in long wavelengths is but one promising technology that may aid the development of high-speed, low-cost optical transceivers.
Tran Muoi Optical Communication Products (OCP) Inc.
Optical transceivers perform both the signal transmitter and receiver functions in a fiber-optic transmission system. The transceiver uses a light-emitting diode (LED) or a laser to convert an incoming electrical signal to an optical signal, then transmits the light through the fiber. At the other end, another transceiver receives the optical signal and converts it to an outgoing electrical signal. Today, optical transceivers are widely used in fiber-optic transmission systems for telecommunications, data-communications, and networking applications. In the last decade, the operating speeds of transceivers have steadily increased as networks have evolved, from the Fiber Distributed Data Interface (FDDI) rate of 125 Mbits/sec to the OC-48/STM-16 rate of 2.5 Gbits/sec. Over the same period, the transceiver`s size has diminished from a 4ٹ (4 rows of 7 pins) and 2䁿 package to the current 1ٻ and small-form-factor (SFF) package.
Multimode fibers have traditionally been used in local or premises network applications, where transmission distance is relatively short, 2 km or less (see Fig. 1).
The required transmission capacity dictates the fiber and optical source technology being used. Early applications used 850-nm light-emitting diodes (LEDs) at 10-Mbit/sec Ethernet speed. With the introduction of FDDI standards, operating speeds jumped to 125 Mbits/sec. To satisfy the transmission distance of 2 km at 125 Mbits/sec, 1300-nm LED technology over 62.5/125-micron multimode fiber was optimal, because it offered higher coupling efficiency with LED sources than 50/125-micron fiber. The combination of 1300-nm LED and 62.5/125-micron fiber was also used when operating speeds reached 622 Mbits/sec with the introduction of Asynchronous Transfer Mode (ATM) premises networks that supported transmission distances of 500 m or less.
As a result of the early penetration of FDDI, Fast Ethernet, and ATM applications, the installed multimode fiber base in North America is predominately 62.5/125-micron fiber. However, 50/125-micron fiber is more prevalent in the Far East--particularly Japan--and in Europe.
The technology landscape has changed with the recent introduction of Gigabit Ethernet applications. The basic problem is that 1300-nm LEDs are not fast enough to support 1.25-Gbit/sec transmission rates. Strictly speaking, 1300-nm LEDs can support 1.25-Gbit/sec transmission. However, output optical power would have to be compromised, resulting in only short transmission distances. The cost of producing these high-speed LEDs is also a concern. The 1300-nm lasers can, of course, do the job, but they are also too expensive.
Therefore, 850-nm vertical-cavity surface-emitting lasers (VCSELs), introduced commercially only recently, have become ideal optical sources for 1000Base-SX Gigabit Ethernet applications. These devices are cost-effective because they offer the performance characteristics of a laser (high speed) but the manufacturing process of an LED (low cost). The commercial introduction of 850-nm VCSELs really injects new life into the short-wavelength window for gigabit/sec requirements.
There is also a re-emergence of 50/125-micron fiber with the penetration of 1000Base-SX Gigabit Ethernet applications. Depending on its modal bandwidth (160 or 200 MHz/km), the 62.5/125-micron fiber can cover a distance of only 220 to 275 m maximum with 850-nm VCSELs. The 50/125-micron fiber offers a wider modal bandwidth (400 or 500 MHz/km) and can therefore extend the distance to 500 or 550 m. Consequently, for new installations, the 50/125-micron fiber may be a better alternative.
The other option for extending the distance to 550 m on 62.5/125-micron fiber is using the 1300-nm Fabry-Perot (FP) laser instead of the 850-nm VCSEL. This solution is costly, however, because of two factors. For starters, the 1300-nm FP laser is more expensive than the 850-nm VCSEL. Secondly, to avoid the differential mode delay (DMD) problem associated with the coupling of light from an FP laser to a multimode fiber, a special offset launch-mode conditioning patch cord is required between the transceiver and the multimode transmission fiber, thus adding extra complexity and cost into the system.
Herein lies the attractiveness of 1300-nm VCSELs. Even though 1300-nm VCSELs are not commercially available yet, many laboratories and companies are actively developing these devices (see "Low-cost, singlemode transmission with long-wavelength VSCELs" on page 62). The 1300-nm VCSEL offers an option to extend the transmission distance of Gigabit Ethernet on 62.5/125-micron fiber to 550 m (which covers building backbone applications) without the high cost of the 1300-nm FP laser and the need for the special offset launch-mode conditioning patch cord.
For metro and long-haul networks, singlemode fiber is the optimal transmission choice. Because of the desire to have only one fiber type for the whole network, singlemode fiber is also widely used in distribution and local access networks. The 1300- and 1550-nm laser technology that supports transmission distances of up to 100 km for OC-3/STM-1 (155.52-Mbit/sec) to OC-48/STM-16 applications over singlemode fibers is well established. But in the early stages, most of these lasers had to be cooled or temperature-controlled. The need for a thermo-electric cooler results in a transmitter package that is large in size as well as high in power consumption.
With the commercial introduction of FP and distributed-feedback (DFB) lasers using multiple quantum well structure, laser performance is greatly improved. As a result, these lasers can be operated without cooling or temperature control, yet still satisfy the required performance specified in the Synchronous Optical Network/Synchronous Digital Hierarchy (sonet/sdh) standards over the entire temperature range of -40 to -85C. This capability allows laser transceivers to be packaged in the same 1ٻ footprint or small form factor (SFF) as LED transceivers (developed for multimode-fiber applications). It also opens the door to low-cost manufacturing of laser transceivers for singlemode-fiber applications.
For sonet/sdh applications, transmission distance is normally categorized into short reach (2 km), intermediate reach (15 km), long reach 1310-nm (40 km), and long reach 1550-nm (80 km)--(see Fig. 2). All of these distances for OC-3/STM-1 to OC-48/STM-16 applications can be met using 1300-nm FP, 1300-nm DFB, or 1550-nm DFB lasers operating without temperature control.
For Gigabit Ethernet applications, the 1000Base-LX standard specifies a distance of 5 km. But the distance can be easily extended to 10 km by narrowing the center wavelength range and spectral width of the 1300-nm FP laser. By using 1300- and 1550-nm DFB lasers, the distance can be extended further to about 35 and 70 km. Why is the transmission distance for OC-48/STM-16 using 1300- and 1550-nm DFB lasers longer at 40 and 80 km? For OC-48/STM-16 long-reach applications, the detector used is an avalanche photodiode (APD) instead of the simple PIN detector normally used for Gigabit Ethernet applications. The higher sensitivity of the APD allows a larger optical-power budget and hence a longer transmission distance.
Optical transceivers are available from many suppliers for all of the applications discussed: FDDI, Fast Ethernet, Gigabit Ethernet, and ATM OC-12/STM-4. These transceivers use 1300-nm LED, 850-nm VCSEL, or 1300/1550-nm FP and DFB lasers without temperature control, depending on the applications. Optical transceivers are generally offered in the 1ٻ or 2ٻ package with duplex SC connector receptacles. ST- and FC-connector receptacle transceivers are also available from some suppliers.
The penetration of higher speeds (OC-48/STM-16) in local networks is fueling a demand for low-cost transceivers operating at 2.5 Gbits/sec. Transceivers in the 1ٻ or 2ٻ package satisfying short- and intermediate-reach OC-48/STM-16 optical specifications are beginning to appear on the market. The 2.5-Gbit/sec transceiver using 850-nm VCSELs over multimode fibers is likely to be commercially introduced soon.
In addition to the continual push for higher speeds, there is also an ongoing race to reduce the physical size of optical transceivers. Typically, the networking equipment--whether it is a hub, router, or switch--needs to interface with many optical-fiber paths. The equipment backplane or front panel has a limited space to accommodate all the optical connectors for interfacing to these fiber paths (ports). The only way to increase the port density of the equipment is by reducing the size of the optical connectors. A number of SFF connectors have been proposed and introduced commercially (see Lightwave, December 1998, page 55). Consequently, next-generation transceivers will feature these SFF connector interfaces more and more.
As always, the motivation is for higher speed, smaller size, and yet lower cost. Some of the technologies that will allow optical transceivers to meet these requirements include:
High-speed electronic integrated circuits (for 10-Gbit/sec applications or above)
LEDs, VCSELs, lasers, and detectors, capable of operating in a non-hermetic environment
passive alignment of optical sources and detectors with fiber coupling
controlled-wavelength VCSEL and DFB lasers
optical device (LED, VCSEL, laser, and detector) array technology.
The electronic integrated circuits (ICs) required for optical transceivers typically include LED or laser drivers, low-noise transimpedance amplifiers, post amplifiers, clock-recovery circuits, and decision circuits. The clock-recovery circuit and the decision circuit are often combined in a single IC known as a clock- and data-recovery circuit (CDR). There is a good selection of these ICs, up to 2.5 Gbits/sec, from a number of commercial suppliers.
Traditionally, ICs have been offered with a 5V power supply. There is a growing demand, however, for a 3.3V power supply, which would reduce power consumption and be compatible with the other ICs used in network interface cards (NICs). As a result, most ICs are now offered in 3.3V versions as well. Looking forward, we expect 10-Gbit/sec ICs to soon be introduced commercially and perhaps be widely available within a year.
A key optical device to look for in the near future is the 1300-nm VCSEL. Because of the classic tradeoff between performance and price, we will probably see two versions of the 1300-nm VCSELs: one for multimode-fiber applications and one for singlemode-fiber applications. It is important to note that the VCSEL is inherently a single longitudinal-mode laser; it oscillates in only one longitudinal mode (spectral line) like a DFB laser. Thus, the 1300-nm VCSELs have the potential to displace not only FP lasers but also DFB lasers in some applications.
In current optical transceivers, the optical devices are typically contained in a hermetic TO can or similar hermetic package. To achieve low-cost manufacturing, next-generation transceivers will probably be in a totally plastic package where the optical devices must operate reliably in a non-hermetic environment. The alignment of the optical devices with the coupling fiber will also have to be done passively (no X-Y-Z active adjustment).
It is also important to follow developments in wavelength-division multiplexing (WDM) and array technology, which are alternative methods for increasing total transmission capacity throughput. The use of dense WDM for high-capacity metro and long-haul transmission is well publicized. However, WDM is also applicable in the local and premises network environment. Here, the emphasis is on a smaller number of channels (four or eight at most) as well as low cost. Controlled-wavelength VCSEL arrays and FP or DFB laser-diode arrays offer low-cost solutions to these applications.
As optical networks advance, optical transceivers will continue to evolve to higher speeds and possibly even smaller form factors. At the same time, optical technologies such as WDM and array will further help to increase throughput. See "Optical Transceiver Evolution" on this page for more information on possible future scenarios.
Tran Muoi is the president at Optical Communication Products (OCP) Inc. (Chatsworth, CA).