By Ezekiel Kruglick
A case study shows the conversion of a dynamic optical add/drop multiplexer from a switch–based architecture to a four–plane crossconnect architecture. The results of this integration are reduced cost, increased reliability, and increased flexibility.
One of the most common module types currently being deployed is the optical add/drop multiplexer (OADM), which adds and drops certain select wavelengths from a given node. Reconfigurable OADMs (R–OADMs) allow further cost reduction and flexibility over that of OADMs in the face of uncertain service projections. Reconfigurable OADMs also decrease congestion and allow more rapid provisioning.1
The advantages of integrated R–OADMs over those of discrete R–OADMS provide the motivation for this case study. The conversion of 1 × 2 array–based R–OADM switches to integrated four–plane crossconnects can increase flexibility, add trouble–shooting features, reduce cost, ease manufacturing, and increase reliability. The switch–based R–OADM is used as a reference because it offers testing and flexibility advantages over "broadcast" and "select" architectures.2 The switch–based module can also be applied to small wavelength counts on multiple fibers, which is not feasible with broadcast and select, as well as many wavelengths on a single fiber.
A typical 1 × 2–based R–OADM module can be designed for different applications—for example, a bidirectional line–switched ring (BLSR) application (see Fig. 1). In this example, the incoming signal is shown with up to 12 channels, of which up to four channels can be dropped.
The 1 × 2 architecture is designed to respond correctly to fail conditions (see Fig. 2). In case of a link failure on one side (for example, both directions out on the East link), the node can perform a "far–side reflect" to avoid the link while still allowing both East and West transceivers full transmit/receive access to the data. In case of a local node failure the optical switches automatically fail to a "reflect" state to redirect traffic around the ring. With minimum reconfiguration, this can be designed so that failures pass the signal straight through the node. Software–commanded bidirectional or unidirectional "pass–through" states can be implemented in case of transceiver failure.3
Four–plane optical crossconnects based on two–dimensional microelectromechanical systems (MEMS) are widely used (see Fig. 3).4 These devices allow the inputs at plane 1 to be arbitrarily connected to any output at plane 2 or allowed through to plane 3. Any port on plane 2 not blocked receives the appropriate signal from plane 4 automatically. The system architect can implement testing and protection functions using the third and fourth plane.
One–way four–plane crossconnects can be used to implement the 1 × 2 R–OADM architecture. Each path uses a single 16 × 16 crossconnect identical to that shown in Fig. 3 (West–East will be used for detail). The first four inputs on plane 1 are the add channels from the East side transmitter; the shaded section serves the purpose of the 12 × 4 and 1 × 2 switches that connect any transmitted signal to any outgoing port. In case of failure or diagnostic need the add signals are connected through plane 3 to the monitor. The incoming channels come into the 12 bottom ports on plane 1 and are either dropped to one of the first four ports on plane 2 (which are hooked to the West receiver), or are connected to the last 12 ports on plane 2 (which are hooked to the output). This constitutes the bulk of the functionality of Fig. 1.
The tinted sections of the crossconnects give the arbitrary channel drop to arbitrary receiver function of the optional tinted 12 × 4s in the integrated design. The crossconnect also adds shuffling functionality so that if the incoming 12 channels started, for example, on several different fibers, they can be swapped among outgoing fibers. This is notably different from a broadcast and select architecture, which cannot perform R–OADM on multiple fibers.
Plane 4 receives test signals in the first four ports and equalization signals in the last 12 ports. The test signals serve to test any drop receiver that is not currently receiving data. The equalization signals can fill unused channels to maintain constant power through amplifiers in the spans or to similarly enable testing.
Plane 3 is the default output location for plane 1 in case the crossconnect board should fail. Since this plane is terminated at a protection switch during standard operation it can also be considered the "termination plane" and used to dispose of power equalization channels from previous spans so that the channel can be reused. Any channels sent into this termination plane are dissipated.
The unshaded 4 × 4 portion of the crossconnect can be used to connect each transmitter to the opposite–side receiver for testing. When a transceiver is added during field upgrade, the signal will immediately be routed (plane 1 to plane 3) to the same–side testing port. The signal can then be redirected to the opposite–side receiver for send and receive testing on demand. This could be done, for example, during periods when data demand is low and a given transceiver is not needed. This continuous checkup capability is an inherent advantage of the integrated design and encourages condition reporting and monitoring. Many of these failure and testing functions are not available to the discrete implementation (and there are no inbuilt testing paths at all within a broadcast– and select–based node).
If a full crossconnect board fails, the incoming signals are reflected, so the transmit and receive channels are hooked into the test configuration. A span can also be subjected to line test/equalization transmission by setting the failure state to route the signal generator to the unused span.
In our case study, the failure–in–time (FIT) characteristic was used to conduct a failure analysis comparison. The 16 × 16 crossconnect was assigned the FIT value of 800 based on a published milestone for widely available Telcordia–qualified MEMS–based crossconnects plus packaging and driver.5 Even if the electromechanical devices are given exceptional values of FIT, the comparison between the FIT for 110 discrete switches and two parts with FIT 800 indicates that the integrated solution is more reliable under all reasonable criteria. Under the simplest FIT calculation using FIT of 200 for each electromechanical 1 × 2 (and assuming the optional 12 × 4s are not used), the change in architecture to an integrated four–plane approach improves reliability by more than a factor of 13. This calculation leaves out the reduced control electronics and reduced splicing, which will further favor the integrated approach.
The integrated circuit industry used to talk of the "tyranny of large numbers" problem that would occur when building complex nonintegrated circuits using hundreds of discrete components. Issues of inventory¸ manufacturing¸ interconnection¸ and testing all become vastly more complicated as system sizes increase. Gains are dramatic and easily quantified in cost, capability, and reliability.
At current prices, the integrated four–plane solution is cheaper in component cost and carries vastly reduced splicing and volume overhead. The manufacturability is dramatically improved because there are a small number of ribbon splices and only a few parts to support. Physical board design is dramatically simpler.
The cost of parts is impacted primarily by the simple conversion of 110 parts to only two. A comparison using approximate current prices (as best available) indicates cost–of–goods savings of more than 65%. Since the rate of cost decrease for the integrated parts is steeper than that of the simpler switches, these savings promise to get better. Additional savings are also realized in dramatically reduced manufacturing steps (such as turn rates and labor costs) and vastly reduced complexity. The reduced number of components also reduces the support electronics and board space needed. The integrated crossconnects typically require little more than a serial digital command line and a voltage reference.
The increased capability of this particular example is in the unglamorous but always important category of troubleshooting and diagnostics. The integrated approach allows advanced local diagnostic tools without requiring a technician. The capability for internal bit–rate testing with the switches in the loop ensures that the system will not suffer outages due to misconfiguration and reduces capital expense for the end user. The automatic matrix crossconnections also simplify software requirements. These factors are especially important in light of industry data indicating that the average carrier will see more outages for procedural errors and software issues than hardware failures.6
Improved reliability translates directly into operational expense savings for the end user.
The smaller size and power consumption of the integrated architecture also are attractive. As vendors switch to integrated elements for modules and systems, we should see decreasing cost, increasing reliability, and increasing capability in optical systems—much as the shift to integrated circuits impacted electronic systems.
Ezekiel Kruglick is a principal MEMS designer at OMM, 9410 Carroll Park Dr., San Diego, CA 92121. He can be reached at firstname.lastname@example.org.
1. J. M. Simmons, E. L. Goldstein, A. A. M. Saleh; J. Lightwave Tech. 17(1) (1999).
2. J. Bayne, M. Sharma, Lightwave (2001).
3. GR–1230–CORE BLSR SONET standard satisfying R6–110 and R6–111 requirements.
4. Fernandez et al., Lightwave (2000).
5. P. DeDobbelaere et al., Opt. Fiber Comm. Conf. (2002).
6. FCC publication, Quality of service of the local operating companies aggregated to the holding company level (1995 to 2001).