Jagdish Rebello, Arden Olson, and Nan Zhang
While large-scale matrix, all-optical switches are vital to overall network performance, low-port-count switches are an economical and reliable means of providing many switching functions. Microelectromechanical 1 x 2 and 2 x 2 switches have advantages in terms of cost, size, and power consumption.The first microelectromechanical systems (MEMS) applications in optical communications have been for fiber-switching mirrors and V grooves used for multiple-fiber alignment. As MEMS technology and optical telecommunications evolved, large-matrix switch fabrics (so-called free-space, 3-D micromirrors) have drawn the most attention. These high-port-count optical switches in nonblocking configurations are potentially scalable to more than 1000 ports. A scheme to do this using conventional scanning mirrors was first proposed in 1982. In this concept, arrays of moving mirrors and collimators are juxtaposed to make nonblocking interconnects. Either electrostatic or electromagnetic drive mechanisms can be used to move the mirrors.The use of MEMS-based 3-D mirrors is the leading proposed solution for large-matrix, all-optical switch fabrics. However, many technical hurdles must still be overcome before such 3-D switching fabrics will be widely deployed in commercial networks. Although fabricated with many of the same processes used for integrated circuit manufacture, MEMS 3-D switching fabrics are mechanical devices with a large number of moving mirrors. The issues of low manufacturing yields for complete switch fabrics and unproven reliability for telecommunications applications remain major stumbling blocks to large-scale commercial deployment. Other technical challenges include mirror-surface finish, beam divergence, and performance repeatability. Nevertheless, as the technical challenges are surmounted, optical switches will be deployed in increasing numbers in optical networks.
LOW-PORT-COUNT APPLICATIONS
In addition to large-scale 3-D switches, many applications in long-haul and metropolitan-area networks require a large number of small-port-count devices (see Fig. 1). For example, in many network architectures redundant paths are designed to serve as backups in case of failures in the primary paths. To divert the data transmission between the primary and the backup paths, reliable 1 x 2 switches are required.
In several metro-area networks, especially in last-mile deployments, there are only a small number of input and output ports, once again requiring the use of small-port-count devices—for example, a four-channel variable optical attenuator (VOA). And MEMS-based small-port switches are well-adapted for application in reconfigurable optical add/drop multiplexers. In all of these applications, MEMS-based technology is ideal because it offers compelling capacity, cost, and power savings, coupled with tremendous improvements in package density.
In addition, these optical devices offer bit-rate and protocol transparency—a key factor in metro-area networks where different architectures such as SONET, ATM, Gigabit Ethernet, and dense wavelength-division multiplexing (DWDM) are widely deployed. The availability of high-performance, low-cost optical-MEMS components in the near future will thus be key to the widespread deployment of all-optical metro-area networks.
MEMS IN METRO NETWORKSThe metro core, edge, and access networks are evolving rapidly into all-optical DWDM-based networks designed to facilitate last-mile applications. High-speed fiberoptic links are increasingly being used to connect network nodes located only a few kilometers apart. To ensure the continued rapid and widespread deployment of fiber in metro-area networks, it is imperative that the optical components used in these networks be available in low-power-consuming compact packages and at economical prices.An important component for metro-area networks will be low-port-count optical switches for protection switching and network configuration for demand management. Using MEMS technology, 1 x 2 and 2 x 2 switches can be fabricated from silicon chips smaller than 0.5 cm2. When used with optical fibers in metro-area networks, these switches can be packaged in compact, rugged housings that maintain minimum fiber-bend radii, essential for minimizing insertion losses and maximizing fiber life. And using MEMS technologies, multiple switches can be packaged compactly together on a single board. This compactness is especially important in metro-area networks where space in a central office or in an outside plant distribution frame is often at a premium.
The small size of a MEMS switch comes with no penalty on the optical and mechanical performance when compared with much larger and bulky electromechanical switches. In addition, switching speeds are fast (see table).
Small-port-count MEMS-based switches are relatively easy to fabricate using bulk silicon micromachining techniques. For 1 x 2 and 2 x 2 switches, most of the commercial designs available use a single translating mirror (with only one degree of freedom) to divert the incoming signal to the appropriate output fiber. This design implies that the switches can be made by well-established, high-volume fabrication techniques that result in very high device yields and can be price-competitive when compared with other switch designs, which is especially important in metro-area networks where the costs are recovered from a relatively small number of users. In addition, MEMS technologies allow for the design of parts with very good optical alignment and potentially very high coupling efficiencies.
Power consumption is another important factor, especially in a metro-area network. In MEMS-based devices, the mechanically moving parts such as the mirrors in the optical switch are extremely small, often on the order of a few microns, and the distances that the parts move are small. Consequently, the power required to actuate these parts can be as low as a few milliwatts, which in turn implies very low heat dissipation and lower-cost packages. Low power consumption also allows equipment manufacturers potentially to design small backup power supplies into their systems to allow for temporary device functionality in the event of a main power failure.
RELIABILITY AND INTEGRATIONA key argument against MEMS deploy- ment is the lack of long-term performance reliability data. However, MEMS technology is well-established for device manufacture in the automobile, aeronautical, and medical industries. In these applications, where device performance reliability is critical, MEMS technology has emerged successfully—an encouraging fact for optical MEMS devices where long-term performance testing is still in its early stages.Finally, MEMS technology is ideal for metro-area networks because it allows for multifunctional device integration into a common package such a tunable laser integrated with a 1 x 2 MEMS switch in an optical add/drop multiplexer (see Fig. 2). This integration offers excellent savings in terms of space and costs.
Because of these advantages, MEMS technology offers service providers in metro-area networks a cost-effective way of migrating from legacy copper networks to all-optical-fiber networks that will extend right up to the home. This potential in turn can extend the DWDM architecture from the long-haul backbone deep into the metro environment.
Jagdish Rebello is a product manager, and Arden Olson and Nan Zhang are senior project engineers with the Broadband Infrastructure Access Group at ADC, P.O. Box 1101, Minneapolis, MN 55440-1101. They can be reached at 952-233-9030 or contact Jagdish Rebello at [email protected].