Get ready for the optical revolution

Get ready for the optical revolution

Thomas G. Hazelton Optical Switch Corp.

There are now three guarantees in life: death, taxes, and the demand for more bandwidth. The increased consumption of bandwidth by businesses and individuals is not a new revelation. But the rate at which demand for this commodity is growing is unprecedented and changing the entire complexion of the telecommunications infrastructure.

It`s amazing to think that less than 20 years ago, the majority of communications traffic was voice-related and carried over analog transmission facilities. Electro-mechanical switching systems routed the majority of analog voice traffic. As the telecommunications industry grew and evolved in the 1980s, digital telephony and lightwave systems began to play significant roles in the advancing transmission and switching systems.

More and more wavelengths

By the mid-1980s, the standards for Synchronous Optical Network (sonet) were beginning to emerge. Network planners and equipment manufacturers alike predicted sonet`s evolution, and they also predicted that sonet`s ability to evolve with advancing laser technology would keep pace with the rate of bandwidth. As we now know, on-demand data services and the predominance of PC applications ignited the desire for even more bandwidth. Interactive Web access, e-mail, and file transfer services began outpacing the ability of both transport and access networks to keep up.

By the mid-1990s, 8- and 16-channel wavelength-division multiplexing (wdm) emerged as a "knight in shinning armor," providing cost-effective relief for transport networks and a clear path for increasing bandwidth. We have now reached a point where optical (or photonic) switching will play a significant role in the evolution of the network. Thus, the next wave of technological advancement will occur within the switching fabric of the optical network.

wdm has proven to be a cost-effective means of increasing the bandwidth of installed fiber plant without the significant investment of new fiber routes. What was once considered "pair gain" technology for fiber spans is quickly becoming the foundation for services that will offer customers a new class of high-bandwidth and broadband capabilities.

Market analysts predict sales of wdm and dense wdm systems will reach $4.3 billion in North America by 2001. This amount translates into thousands of wavelengths deployed as either point-to-point systems or ring topologies within optical networks. Futhermore, upwards of 1.5 million wavelengths are projected to be deployed through the enterprise, local loop, and interoffice network environments by 2007 in the United States alone. It is envisioned that the majority of these wavelengths will require protection, add/drop, and routing capabilities.

Today, wdm networks are managed and protected within the digital domain using sonet and its associated support systems. To leverage the full potential of wdm deployment, service providers, especially in the emerging competitive access networks, will require ways to offer dynamic wavelength services and keep pace with the growing demand for bandwidth. Specifically, they will look for network elements that cost-effectively provide provisioned or switched virtual broadband services on a demand and capacity-enhancement basis. The desire for this type of network flexibility will drive the need for faster, larger N ¥ N matrix optical switches.

To effectively manage the growing number of traffic-bearing wavelengths, a new breed of optical networking element will be required for adding, dropping, routing, protecting, and restoring optical services cost-effectively. The realization of these capabilities is linked to the availability of a reliable, performance-oriented optical-switching technology that enables N ¥ N matrix growth beyond a 4 ¥ 4 or 8 ¥ 8 matrix.

Optical switching has been evolving for many years, and it seems like new technologies surface each month. To date, no single technology provides all the necessary "bells and whistles" to truly address the stringent requirements of the optical networking element. Let`s face it, lithium niobate optical-switching technology was first introduced in 1988 and is only addressing a small segment of this market opportunity.

The FTIR alternative

However, frustrated total internal reflection (ftir) provides a very promising alternative to mechanical, solid-state, and even micro-mirror optical switching. It has the potential to address many of the requirements of future network elements.

ftir applies the displacement of a tri-metallic substrate (bimorph) to position a switch plate in one of two switch states. An incoming beam of light from the fiber is collimated by a grin lens and reflected through a prism to a focusing grin lens. The movement of this extremely low-mass switch plate (only 2.5 microsec) enables an optical-switching junction to be subjected to extreme environmental conditions without degradation of the signal. The actual switch transition time of the ftir switch is less than 2 microsec. Its optical performance is maximized, yielding less than 1-dB loss, excellent optical channel isolation, no polarization, and perfect optical reflectance characteristics with negligible polarization components.

Perhaps the most significant attribute of the ftir switch is that it can be used as a building block (optical junctions) to form large nonblocking optical matrices (potentially 1024 ¥ 1024) without compromising fundamental optical performance. With its symmetry and scaleability, ftir optical-switching technology offers the "fabric" necessary for applications such as optical add/drop multiplexers, optical protection switching, and optical crossconnects.

ftir has been documented for well over 30 years. The techniques employed to get the residual reflection, or "virtual contact," down to less than -17 dB have been researched and documented; however, there appeared little hope of achieving better than -20-dB residual reflection using these methods. Yet, a new technique to get the contact spacing, and thus the residual reflection, down to less than -50 dB has now been developed. In fact, a 1 ¥ 2 optical switch using this technique that operated at better than -50-dB crosstalk was demonstrated earlier this year.

The concept behind the ftir optical switch is rooted in Snell`s Law, the same principle of optical physics used in fiber optics (see Fig. 1). When light passes from one medium (n1) to another medium (n2), the light is refracted based on the numerical indices (n1, n2) of the media and the angle (q1) at which the light enters.

There is a certain value qc where

This is called the critical angle. For angles greater than qc, total internal reflection is achiev ed--a perfect reflector with virtually no polarization components. This perfect reflection, or total internal reflection, may be "frustrated" by bringing a second refractor into the evanescent field (contact) with the reflecting surface of the prism. The light traveling in the prism will pass through the reflecting surface and travel into the switch plate, altering the angle of the beam.

So, a prism and switch plate compose the centerpiece of the basic ftir switch (see Fig. 2). To couple incoming light from the fiber into the prism, a grin lens is used to convert the light from the fiber into a parallel beam that is transmitted into the prism and reflected off the reflective surface (hypotenuse).

In practice, a switch plate is positioned on top of the reflecting surface of the prism (see Fig. 3). The switch plate is a circular plate of glass with the back surface (the reflecting surface) ground at an angle (angle q) to the front (the contacting surface). A piezoelectric bimorph transducer actuates this switch plate. When the switch plate is not engaged, a total internal reflection of the input light to the primary output occurs. When the switch plate is engaged, the circular switch plate makes intimate contact with the surface of the prism, thus frustrating the beam of light to the secondary output. Likewise, a focusing gradient lens is used to focus the collimated beams of light back into either "primary" or "secondary" output fibers.

This mechanism composes the basic optical-switching junction of the ftir switch. It`s important to note that the output of the ftir switch produces either a primary output beam (when the switch plate is open) or secondary output beam (when the switch plate is closed). Each switch state is controlled as a binary state of "1" or "0," respectively.

Unlike other optical-switching technologies incorporating piezoelectric devices, the ftir switch utilizes a piezoelectric bimorph transducer to achieve an open or closed position of the switch plate. Alignment of the bimorph transducer is never required during the life of the ftir switch. In the closed position, electrostatic forces retain intimate contact with the switch plate, thus surviving optical transmittance degradation when subjected to mechanical shock or vibration. The piezoelectric bimorph transducer only has to deflect the switch plate >1.5 microns to revert total internal reflection back to the "primary" fiber output. The ftir optical switch is thus classified as a "virtual solid-state" optical device.

Increasing optical outputs

Based on the design of the ftir optical-switching junction, it is possible to expand the number of optical outputs by cascading prisms and their associated switch plates together, creating larger 1 ¥ N junctions and N ¥ N optical matrices. The operation of a 1 ¥ N switch is similar to that described for the 1 ¥ 2 switch. As an example of how switch-plate cascading works, assume the architecture for a 1 ¥ 4 matrix. In this example, two prisms are incorporated and placed back-to-back, with each associated switch plate being labeled as (1) or (2) (see Fig. 4).

By bringing switch plate 1 into contact with the surface of prism 1, the optical beam traveling through the prisms is shifted in angle. Similarly, bringing switch plate 2 into contact with the reflective surface of prism 2 shifts the angle of the optical beam twice the amount of switch plate 1. The actuation of this switch plate transfers the optical energy from the first fiber to the second fiber. Switch plate 2 is set at an angle that is two times the angle of switch plate 1. Thus the actuation of switch plate 2 transfers the energy from the second fiber to the third fiber. When both switch plates are actuated, they transfer the energy to the fourth fiber. In this manner, each switch plate provides two switch states, each yielding four possible outputs or a total port count of N=[number of switch plates]n, allowing the input beam from the input fiber to be switched to four output optical fibers (see Fig. 5).

Larger optical junctions

Applying our 1 ¥ 4 optical switch as the example, the optical performance of each output fiber is consistent and repeatable. That is, the insertion loss, crosstalk, polarization-dependent loss, and backreflection performance characteristics are fairly uniform across all four input-to-output conditions. Larger 1 ¥ N junctions are achieved by cascading additional prisms and their associated switch plates and developing N-fiber grin lens fiber arrays.

Take for example the construction of a 1 ¥ 16 optical-switching junction, which is a replication and combination of the same components used in the 1 ¥ 4 example (see Fig. 6). The 1 ¥ 16 optical-switching junction accommodates 16 output fibers through a customized grin lens fiber array arranged with four columns (Y-axis) and four rows (X-axis) of output fibers. The 1 ¥ 16 optical-switching junction therefore is composed of four prisms, each having its own switch plate.

Like the 1 ¥ 4 optical matrix, the angle for switch plate 1 and switch plate 2 is at angle q. Also, like the 1 ¥ 4, the angle for switch plate 3 and switch plate 4 is at angle 2q. Switch plates 3 and 4, however, are rotated about their axes 90 with respect to switch plate 1 and switch plate 2. In this relationship, switch plates 1 and 2 translate the spatial position of the output beam to the X-axis of the grin fiber array while switch plates 3 and 4 translate the spatial position of the output beam to the Y-axis of the grin fiber array. Therefore, up to 16 switch states, and thus 16 input/output conditions, can be accommodated.

Nonblocking optical-switching matrix

The ftir 1 ¥ N optical junction provides tremendous flexibility and optimal performance for building larger N ¥ N matrix designs. One of the key benefits of ftir technology is derived from the ability to configure 1 ¥ Ns in a back-to-back sequence, thus creating a true nonblocking optical-switching matrix, providing growth scaleability as well as optical symmetry.

In Figure 7, 32 1 ¥ 16 optical junctions are configured back-to-back to form a 16 ¥ 16 switching matrix. Sixteen fibers from each grin fiber array are fully interconnected (via fusion splicing) to each of the remaining 31 1 ¥ 16 optical-switching junctions. All port-to-port connections are capable of bidirectional transmission, and light can originate on either the "A" or "B" side of the matrix. As previously mentioned, control of the matrix is accomplished through a binary addressing scheme, driven by a high-speed bus control, and controlled digitally given the "0" or "1" control state of each switch plate.

Unlike Benes or Clos multistage architectures, which use many smaller N ¥ N switches to build large matrices, the ftir switch architecture offers uniform and symmetrical performance when any A-side port is connected to any B-side port.

ftir performance

ftir technology provides consistent performance across all matrix sizes. All optical parameters, except for insertion loss, remain uniform across all junction or matrix sizes. System insertion loss varies with size and is determined by a number of factors, including fiber-lens interface, loss through the grin lens, the grin/prism interface, grin blur circle loss, alignment, and path-length loss (number of cascade prisms). Path-length loss is calculated from the number of prisms cascaded in an optical-switching junction. If two junctions are used back-to-back to create a matrix, the path-length loss is multiplied by two to derive the total system loss. Predicted insertion-loss performance derived from system analysis for all matrix sizes is shown in Table 1.

Testing of ftir switch technology is ongoing. Preliminary analysis of the empirical data suggests that the "alpha" units are meeting or exceeding the theatrical models. To date, 10 1 ¥ 2 alpha units have been produced and tested (at 1550 nm), yielding the optical results shown in Table 2.

Testing has been conducted to determine the failure point of the switch itself. The piezoelectric bimorph transducer was cycled "open" and "closed" more than one and a half million times without failure. Furthermore, it has been observed that the optical performance might actually improve with switch-plate wear.

Potential applications

The opportunity for optical-switching matrices will be driven primarily by the rapid deployment of wdm systems throughout both access and transport networks, though many other markets, including cable television, broadcast, private networks, and military, will benefit from high-quality optical switching.

Within the telecommunications arena, the first application of ftir switching likely will occur within the optical add/drop multiplexer (oadm) network elements currently in definition by itu and Bellcore standards committees and in development by telecommunications-equipment manufacturers around the world. These "next-generation" optical networking elements will provide the building blocks necessary to extend wdm from a static point-to-point service into a dynamic platform of enhanced features and services.

oadms will be required to perform add/drop, protection switching, and routing of wavelengths. One ftir 8 ¥ 8 or 16 ¥ 16 optical-switching matrix can perform all of these functions on a per-wavelength basis and also enable optical performance facilities to test and isolate problems (see Fig. 8). Because of the fast switch transition time of each switch plate (<2 microsec), these larger N ¥ N matrices can be coordinated, so all of the switch plates associated with the desired path-to-path connection will complete the crossconnect transaction in under 500 microsec. This capability will be a key issue when mixing optical networking with existing sonet operational support systems.

In addition to add/drop capabilities, oadms will also be required to provide user-network interface (uni) services, which are wavelength tributaries offered by a wdm optical ring. Here, wavelengths are groomed and adapted to other enterprise wdm networks outside the service provider`s network. For these applications, the oadm will be required to provide optical "bridging" as well as "drop-and-continue" wavelength provisioning to meet the various grades of service levels offered in the future. It is also envisioned that optical performance monitoring, service restoration, and fault isolation will be necessary components of the uni service offering (see Fig. 9).

The next step

Photonic switching is the next logical step in a long history of switching technology that started with manual "plug board" operators and eventually evolved to mechanical cross-bar and finally digital switching. The next revolution in the telecommunications industry will occur within the optical domain.

To offer broadband services cost-effectively and ubiquitously in the future, networks will have to migrate toward optical networks that coexist with the existing time-division multiplexed networks of today. Optical switching will play a significant role in this revolution, and ftir optical switching is one of the promising optical-switching technologies of the future. u

Thomas G. Hazelton is market manager at Optical Switch Corp. (Richardson, TX), www.opticalswitch.com.