Optical switches enable dynamic optical add/drop modules
John Barthel and Tom Chuh
Add/drop nodes are key elements in metropolitan networks because they enable traffic channels to both exit and enter. The use of all-optical, add/drop modules can improve efficiency and produce revenue for network operators.
Network service providers are seeking a new class of equipment that will expand their reach while reducing provisioning time, increasing accuracy, and improving revenue. Customers want fast turn-up of new capacity and the ability to utilize only the bandwidth they need as their demand varies.
This demand fluctuation, largely found in the metro and regional networks, requires dynamic optical add/drop capabilities to provide flexible data on and off ramps. At the same time, dynamic optical layer switching is also required and can be implemented using microelectromechanical system (MEMS) switches.
Current architectures do provide adequate bandwidth using optical-to-electrical signal conversions and fixed node configurations, but the ability of service provider to effectively upgrade networks and offer revenue-generating services is limited. Provisioning—or the time it takes to turn on new services—with fixed architecture and optical-to-electrical conversion is measured in weeks because of difficulties in changing and testing equipment configurations. This delay can result in lost customers and lost billable revenue. In addition, once the service is established, reserved bandwidth may not be needed all the time, resulting in inefficient use of system capacity and loss of additional revenue.
DYNAMIC NETWORK SERVICES
Carriers must improve system efficiency and use the entire bandwidth available on each wavelength. This new level of service will reduce operational costs, enable true point-and-click service, and result in improved time-to-bandwidth and an expanded revenue stream. This next-generation dynamic network is being built on a mesh structure, which requires new innovations and added complexity at the nodes, specifically in the optical add/drop modules.
The optical add/drop module must be able to dynamically drop and add wavelengths in response to bandwidth, protocols, and services being added at a node in mesh configurations. The optical add/drop module must also be able to instantly restore services over different optical pathways if there is a fiber cut or an equipment failure.
This dynamically configurable network can best be implemented if the optical layer is separated from the communication stack and optical-to-electrical conversions are minimized. Since the dynamic optical layer is independent of protocol and data rate, the carrier can provide ATM, IP, or SONET services over the same channels.
MATRIX OPTICAL SWITCH
With recent technological innovations, larger port-count matrix optical switches have become the enabling component for a dynamic optical add/drop module. True matrix switches, also referred to as N x N switches, allow any input channel to be switched to any output channel. The N x N switch typically uses an N2 type of switching fabric. For example, in an 8 x 8 MEMS-based switch, the core is an array of 64 mirrors that move into and out of the beam paths.
An optical connection is established when a mirror moves into the beam path, coupling the signal between any of the two fibers. These switches are nonblocking, which means that the signals on one channel do not interfere with any other signal, channel configuration, or protocol. So they can be used in many optical switching applications, from switching a complete fiber output to redirecting individual wavelengths (see Fig. 1).
By using a matrix optical switch, a dynamic optical add/drop module becomes simple to implement and inherently very flexible. At each optical add/drop module, the system architect can decide whether the wavelengths will be used for express channels, add/drop channels, or simply not used (see Fig. 2).
For example, a 32-channel wavelength-division-multiplexing (WDM) network may not require all of the wavelengths, enabling a customer to implement a "pay as you grow" architecture. Eight channels can be designated as not used. These unused channels can be activated in the future by simply adding the connections or the add/drop functions as needed.
The network also may not need unlimited add/drop functionality at every node. So another eight channels can be designated as fixed, or express traffic, which will pass through without access to the add/drop capabilities at this node. These channels bypass the switch and are ported directly back to the multiplexer. For a system upgrade, fixed express channels can be easily transformed to dynamic add/drop channels by installing additional add/drop functions and changing port connections. Channels designated for full wavelength add/drop capability are directed into the 8 x 8 matrix optical switches in groups of four.
So the resulting optical add/drop module is configured with 16 dynamic channels, 8 fixed express channels, and unused channels, utilizing four 8 x 8 matrix optical switches as the optical switching fabric.
DYNAMIC SWITCH CONFIGURATION
Because the switch is bidirectional, any of the ports to the switch can be used for the four channels designated to the switch (see Fig. 3, left). Four of the channels on Port A can be designated as the inputs and four of the channels on Port B can be designated as outputs for express traffic. The switch can be configured with a simple matrix command that activates the mirrors required to connect the input ports to the desired output ports.
To drop a channel, any one of the input ports can be selected and the mirror moved out of the beam's path (see Fig. 3, right). If the dropped traffic needs to go to a specific destination, any one of the targeted drop channels can be coupled. So system operators and network management software can have dynamic control of the final destination for the designated wavelength.
In similar fashion, any one of the add ports can be dynamically selected and routed to any one of the target output channels. In this switch configuration, the through channels are routed to the desired output ports along with an add channel. The add function can be dynamically used for new traffic, groomed traffic, or for 3R (reamplify, reshape, retime) restoration for a dropped wavelength.
This configuration would be set with a command from the host software that activates the mirrors required to connect these fibers. Again, the system operator and node software controls the provisioning and addition of traffic for any of the specifically designated dynamic channels.
It is often the case that only 30% of traffic is dropped at a node. However, a designer may not want to preselect and bypass the wavelengths for designated express traffic. The enabling technology of the matrix switch allows for dynamic selection of the optical layer routing to meet any need. The fundamental advantage of the matrix switch design enables system designers to use different channel groupings, tunable lasers, and wide band-pass transceivers for added efficiency.
Instead of a 4 + 4 configuration, the switches can be set up in a 6 + 2 configuration, with six wavelengths routed through the switch and two channels available for add or drop functions (see Fig. 4). The two add channels can be supplied with tunable lasers, set to cover the wavelengths routed to the switch. In similar fashion, a wide-band transceiver can be used to allow any wavelength to be dropped.
A 32-channel DWDM system configuration with 24 used wavelengths would still use four 8 x 8 optical matrix switches, as in the example in Fig. 2, but all the active channels would pass through the optical switch fabric.
In operation, the system operator and network management software have ultimate control of the optical layer configuration. The routing of the wavelengths can be changed dynamically, providing for fast provisioning and new and profitable customer services. Because these switches are wavelength and protocol independent, the node can be entirely reconfigured anytime in the future using the same switching core and avoiding time-consuming upgrades that require replacing or rearranging racks and devices.
John Barthel and Tom Chuh are product marketing managers at Onix Microsystems, 4138 Lakeside Drive, Richmond, CA 94806. They can be reached at 510-669-2020 or by e-mail at email@example.com and firstname.lastname@example.org.
MEMS switches offer high-volume capabilities
Microelectromechanical systems-based N x N matrix optical switches offer fundamental advantages over other technologies. Based on silicon, MEMS brings the same high-volume batch production capabilities found in the semiconductor industry, giving the opportunity to develop high-volume, lower-cost products. MEMS also assist in miniaturizing the size of the switch fabric.
The switches are designed as compact components, which helps minimize the space needed for the board and network box. Also, MEMS-based switches operative independent of wavelength or protocol. Because they use simple free-space optics, the signals do not interfere with each other and the connections between channels are unaffected whether the information transferred is 10-Gigabit Ethernet, OC-192, or ATM signals.
MEMS switches are also completely bidirectional and thus can be used in many different system configurations. MEMS-based optical switches decouple the optical layer from the rest of the communication stack.
MEMS have been developed and studied for many commercial and industrial applications, including air-bag crash sensors, video-projector elements, and pressure sensors. Although the application to fiberoptic switching is fairly recent, any concerns over reliability and ruggedness will be addressed as customer field trials continue to move forward and switch designs are verified to the Telcordia GR-1073 and GR-1221 requirements.