Optical-switch technologies crossconnect at OSA meeting

The market focus on optical networking has shifted from capacity expansion to cost-effective network management, according to Mark Krol from Corning Incorporated (Corning, NY). About 40% of the cost of optical networking lies in unnecessary optical-to-electrical-to-optical (OEO) conversions.

Krol and other speakers at a joint symposium on optical crossconnects, held in June (Monterey, CA) by the Optical Society of America, generally agreed that the solution to this problem, based on available technology, lies in separating network traffic into an OEO-free transport layer and an OEO-dependent service layer (see figure). The differences among speakers had to do with exactly how to do that, and the various solutions tended to follow the lines of the component technologies chosen by various equipment vendors.

At a typical optical-network node in the United States, about 80% of the traffic can be routed with transparent optical switches, while the substantial cost of optical-transducer components can be limited to the remaining 20% of signal traffic that requires some sort of opaque provisioning, packet switching, or signal processing, according to John Bowers, chief technical officer at Calient Networks (Santa Barbara, CA). "Photonic switches save costs by eliminating many transponders," he said.

Competing technologies

Among the various and competing switch-fabric technologies for photonic switching, presenters from Calient and Lucent Technologies (Holmdel, NJ) discussed microelectromechanical-systems (MEMS) technology for three-dimensional (3-D) switch fabrics in which optical signals are routed in free space between a two-dimensional (2-D) array of input ports and a 2-D array of output ports by mirrors with two degrees of rotational freedom.

Primary advantages of 3-D MEMS over 2-D switch fabrics are the ability to scale up in port count with capacity expansion, relatively low path-dependent loss, and the availability of semiconductor methods for high-volume manufacturing, Bowers said in describing the commercially available Calient product based on bulk electrostatic MEMS technology. In order to have a large number of ports in a single stage, a switch fabric must use 3-D MEMS, he said, which also provides a highly parallel architecture in the high-port-count environment to compensate for the speed advantages of 2-D switch fabrics.

While acknowledging the advantages of 3-D MEMS switch-fabric technology for high-port count applications, Krol emphasized that a broad range of switch-fabric technologies will be needed in any comprehensive solution. He separated the range of factors to be considered in designing optical crossconnect modules into five categories: fiber management, collimators, mirror arrays, servo controls, and packaging. Krol described development efforts at Corning in several areas; in addressing all of the factor categories in the current market environment, he focused primarily on a wavelength-selective crossconnect technology that is only optically transparent at a designated wavelength, but that can be cascaded to build broadband switching modules.

Johnathan Lacey, from Agilent Technologies (Santa Clara, CA), emphasized current market demand and the need for multiple-technology solutions even more strongly in his talk. He argued that the overwhelming demand is primarily for relatively simple switching applications for which 2-D switch fabrics are sufficient. Due to factors such as a lack of cross-vendor interfacing standards for all-optical crossconnects, the continued need for electronics for signal regeneration, monitoring, and wavelength conversion, as well as continued technology improvement in electronics, Lacey projected a future optical-network architecture in which "islands of transparency" are likely to be interconnected by opaque OEO nodes. He also described Agilent's 2-D and modular technology for optical switching, in which light-transiting fluid waveguides can be redirected at node points through a process of total internal reflection by bubbles similar to those used in Agilent's ink-jet printer technology.

Liquid crystal technology

Liquid-crystal optical switches were described at the meeting by Jung-Chih Chao of Chorum Technologies (Richardson, TX), which announced the issuance in May of four patents related to optical-networking technology and components. Inherent advantages of liquid-crystal optical switches include solid-state reliability, low power consumption, and low insertion loss, Chao said, while traditional limitations have included high temperature-dependent losses and relatively low switching speeds. The use of passive compensation and newer liquid-crystal materials without heaters or temperature controllers has dropped temperature-dependent loss below 0.2 dB for a -5°C to +70°C temperature range and also lowered switching times from 170 ms to less than 1 ms, he said.

David Hunter from the University of Strathclyde (Glasgow, Scotland) concluded the invited presentations by describing testbed research for optical-packet switching. Research collaborators include the University of Essex (Colchester, England), Ilatron Ltd. (Colchester, England), the University of Strathclyde, and the University of Bristol (Bristol, England). One of the issues that the test project, dubbed WASPNET, had to confront was the lack of a memory technology for buffering optical data, analogous to random-access memory (RAM) for buffering electronic data. The researchers managed to work around this, however, with a technique for capacity sharing across wavelengths that reduced the requirement for optical buffering. The method, called scattered wavelength path (SCWP), "is essentially a multiplexing scheme for optical packets carried over WDM," he said. "In SCWP, each optical-packet path is not allocated the same wavelength over its entire path but dynamically converted to a suitable wavelength in each network link."

In a panel discussion after the main presentations, Daniel Blumenthal from the University of California-Santa Barbara spoke on behalf of Calient and replied to Lacey's comment about lack of cross-vendor standardization by saying that generalized mulitprotocol label-switching allows cross-communication between all-optical crossconnect products from different vendors. Krol said that in Corning's investigation of optical-switching fabrics, "the jury is still out" as to whether electrostatic or electromagnetic MEMS will work best. Blumenthal, however, argued the advantages of the electrostatic route chosen by Calient. Paul Yuhas discussed the experience so far at OMM (San Diego, CA) in developing, testing and beginning to provide a range of optical switches from 3-D MEMS to modular 2-D components. Yuhas said that although users might be expected to prefer the flexible growth potential of modular switch fabrics, some have decided not to go modular because they did not want to incur the risk of making adjustments on a working switch fabric.

In terms of vendor agreements, Lacey mentioned that the Agilent switching fabric is being developed and demonstrated in cooperation with Alcatel as well as other vendors, who preferred not to be mentioned. Representatives of Ciena and Tellium in the audience said that their companies are working with the Corning switch fabric.

Overall, many of the panelists seemed to agree with an assertion by Blumenthal that MEMS optical-switching technology is driving toward the very-large-scale-integration electronics model of development. "Photonics sits back where electronics was in the 1950s and integration is absolutely key," Blumenthal said.