To strike a balance between meeting current needs to cut operating costs versus the need for next-generation networks, service providers should consider all-optical networking.
By Tom Albertson, Agere Systems
Network managers are betting increasingly on all-optical networking for their fibre infrastructures. The development of protocol- and rate-independent switching fabrics that avoid problems associated with optical/electronic/optical (OEO) devices will largely determine the outcome of this gamble.
While several all-optical techniques are available now, the emerging choice for forward-looking switching arrays is three-dimensional micro-electromechanical systems (3D MEMS).
Such systems provide the immediate benefits of protocol- and rate-independence, low overall signal loss, small form factor, and competitive switching speeds. 3D MEMS also have the potential for improvements in all performance parameters. In addition, the silicon-based fabrication techniques used in making 3D MEMS promise higher levels of integration of control and management electronics on MEMS arrays, which can greatly enhance network management capabilities and operating efficiencies.
Despite the current debate about the "glut" of fibre in the ground, there has never been any serious question about the need for extra bandwidth. In every era, with every new technological breakthrough, carriers and their customers have called for more.
But capacity is only one dimension of the challenge carriers face in satisfying users' needs. They need to manage their current assets to maximum advantage and to lay the technological groundwork for future systems and services. This is especially critical in an era in which DWDM systems are becoming ubiquitous.
Fibre networks must deal with many service and transport protocols and speeds, as well as dynamic service and configuration demands. Carriers must operate optical networks that provide transport flexibility.
Optical networks must include bit-rate-independent, protocol-transparent photonic switches and cross-connects to provide reliable optical channel routing and transport for the services layers (Internet Protocol, TDM, ATM).
Optical switching accelerates
The compound annual growth rate of optical cross-connect system sales has been forecast by KMI Corp to rise 124% through to 2006, while per-port prices are forecast to decline 31% per year. KMI pegs the 2000 market at USD50m, forecast to rise to USD300m in 2001 and USD6.3bn by 2006, due to higher-port-count products, penetration into smaller offices, and expanding regional sales. KMI also predicts that optical cross-connects will be the choice for the future, saving up to 70% of network costs compared to SONET equipment while allowing more speed and capacity per unit.
Three O-O techniques
Currently, network designers and systems engineers have three basic technology choices for all-optical switching fabrics: bubble, liquid crystal, and MEMS.
The first is based on technology used in bubble-jet printers. The device has a dual-layered architecture: a top layer of silicon that holds the ink-pen technology and a glass layer on top through which multiple light signals travel. The bottom layer has 32 parallel waveguides that have a higher refractive index than the silica substrate. The waveguides hold a liquid with the same refractive index as the silica.
Light passes straight through each waveguide. However, if a bubble is injected at a junction point, light is redirected to an intersecting waveguide. Bubbles are formed by tiny electrodes in the upper silicon layer. The electrodes heat the liquid to form a gas, just as in a bubble jet printer. Advocates claim that bubble technology is potentially low cost because it is based on mass-produced ink-jet technology and can be highly reliable because it has no moving parts.
Another approach based on existing technology is that of liquid crystals, used ubiquitously in displays. The operative principal of liquid crystals - bi-stable polarisation of light passing through the device by varying the voltage applied across the crystal - can be harnessed to provide a switching function. Like bubbles, liquid-crystal switches have the potential to be quite inexpensive, use little power, and be highly reliable because they have no moving parts. However, switching speed is a major concern.
The third approach is MEMS, which are arrays of minuscule mechanical mirrors and components batch-fabricated on silicon substrates. By manipulating magnetic and electrostatic forces between the substrate and the components, tiny mirrors can be made to pop up and tilt to switch light beams. Light from input ports can be focused on a set of tilting mirrors, which can direct the light to other mirrors and then to output ports. The light travels through free space.
By using silicon processing technology, MEMS arrays can achieve high density, precision fabrication, and potentially low cost in volume production. The small mechanical structures can be moved in milliseconds and, importantly, require much less power than traditional OEO switches.
Underpinning the choice of technologies is an equally important consideration: the use of two-dimensional (2D) versus 3D architecture in the switching fabric. Bubble and liquid crystal switch and cross-point arrays use a fabric. In this approach, N input fibres are connected to N output fibres - a configuration that simplifies control electronics.
MEMS can be fabricated in either 2D or 3D arrays. 2D arrays typically are configured in a "digital" approach: mirrors and fibres are arranged in a planar fashion, with pop-up mirrors moving between two positions. Controls are simplified, but scalability - as in bubbles and liquid crystals - is limited by the need for N2 components to accommodate a large number of ports. In a 2D MEMS array, an 8x8 switch would require 64 mirrors.
As with bubbles and liquid crystals, the largest practical 2D MEMS single array is 32x32. Moreover, losses in 2D devices vary with the length and course of the light path (Figure 1).
In 3D MEMS arrays, mirrors can tilt in a variety of positions. Using a 3D MEMS beam-steering architecture, signals are directed to an array of dual-axis mirrors that tilt in multiple positions, bouncing an incoming signal off one of the mirrors to another dual-axis mirror and finally to the desired output port (Figure 2). The continuous-tilting feature of 3D MEMS enables port counts to grow on a 2N scale, instead of N2: two arrays of N mirrors each are used to connect N inputs to N outputs.
Scalability - the key to flexibility
3D MEMS arrays are inherently scalable. Scalability is of paramount importance to carriers, who need to invest in a technology that has a future, enabling them to use the same supplier and integrate the same architecture. The 2N design approach results in linear growth of crosspoints and components, making 1,024x1,024 and larger arrays practical. By contrast, an array of 1,000 cross-connects in a 2D system using an N2 architecture would require 1 million mirrors.
Combined with the extremely compact physical dimensions of the micromirror assembly, 3D MEMS design enables the fabric to be scalable to very large cross-connect applications, using a small form factor with relatively low power dissipation. Moreover, 3D MEMS arrays are inherently non-blocking, achieving "any/any" connectivity between all input and output ports, and reducing path-dependent insertion losses that occur at each cross-point in 2D systems (Table 1).
One of the most important forward-looking elements in 3D MEMS is its synergy with silicon IC fabrication. The ability to tap the immense experience of IC fabrication techniques means that performance parameters will continue to improve, so systems designers will be able to create networks that use lower-powered optics with fewer amplifiers. Use of IC technology enables ever-greater integration of control electronics and other circuitry on the same fabric as the mirrors, enabling greater savings in space, power and overall costs.
Silicon fabrication techniques also facilitate reliability. MEMS arrays have undergone extensive, long-term testing, with mirrors successfully operated through their entire range of movement billions of times. Like 2D MEMS, 3D MEMS arrays will continue to be tested against Telcordia specifications, and will prove to meet the most stringent requirements of the communications and information industries.
By creating a switching fabric that operates quickly and efficiently and is highly cost-effective and scalable, 3D MEMS could vastly extend the reach of fibre communications beyond core networks - even to the "last mile" of end-customers in their homes and offices. The result will help service providers generate more revenue through offering a host of new services and capabilities, to more people, more economically than before.
Tom Albertson is a product marketing manager for Agere Systems in Allentown, PA, USA.
E-mail: Albertson@agere.com Tel: +1 484 397 2000