Jocelyn Lauzon, Pierre-Yves Cortes, and Francis Généreux
Switching technologies such as MEMS are aimed at current switching requirements, but fast, wavelength-selective optical switches fulfill the demand for tunable OADMs and optical crossconnects in next-generation WDM networks. A lithium-niobate-based tunable optoelectronic-frequency filter has the potential to fill this void.
Optical switching is on its way to fulfilling its promise by providing all-optical restoration, reconfiguration, and protection to optical communication networks. The next step for optical switching is to provide fast (on the order of picoseconds) and wavelength-selective (better than 100 GHz) performance to enable the creation of all-optical crossconnects or tunable optical add-drop multiplexers (OADMs). Alternatives are currently under development to meet this demand.
To increase the capacity of communication networks, optical-electrical-optical (OEO) conversions along the transmission links must be minimized, thus driving a convergence toward all-optical networks. Wavelength-division multiplexing is at the center of this convergence since wavelength channels can be used—instead of electronically encoded headers—to do some of the necessary switching and routing in hubs along the network. By using wavelength channels efficiently, switching and routing functions can be performed without converting the optical signals. These functions would then be data-rate and protocol transparent.
Some advocates suggest these functions can be performed using passive optical components, with fixed wavelength channels for fixed communication connections. In most cases, however, passive optical networks do not offer enough flexibility and do not use the optical bandwidth efficiently enough to be a viable choice—even more so in the network periphery, as it gets closer to the end users.
The most recognizable technologies recently introduced into the market are microelectromechanical systems (MEMS) and liquid-crystal switches. These switches offer millisecond response times that are adequate for reconfiguration, restoration, and protection functions in optical systems. Even though these response times are sufficient, there is no doubt that faster switches would be advantageous for these applications.
For applications such as tunable OADMs and optical crossconnects, optical switches with faster response times and wavelength selectivity become mandatory. It is unlikely that MEMS and liquid-crystal technologies will evolve to meet either of these requirements. Wavelength channels must be demultiplexed into spatial channels in multiple switching stages to obtain such selectivity, creating a more cumbersome component.
Moreover, MEMS and liquid-crystal switches are bulk optics components, using free-space collimated beams rather than waveguides. Free-space switches do not fix the lightpath as successfully as waveguide switches, and the accumulation of conditions that push the limits of individual tolerances of components is a significant challenge to overcome. Thus, for fast, wavelength-selective optical switches, other technologies must be considered.
The technology roadmaps for high-speed tunable OADMs and optical crossconnects indicate that optical switches currently being developed are on different paths (see Fig. 1). Optical switches with millisecond response times are associated with a roadmap ending with the optical communication network functions of all-optical reconfiguration, restoration, and protection. These functions are now becoming available and are creating large amounts of interest—and justifiably so. Thinking ahead, though, the paths toward tunable OADMs and optical crossconnects have to be undertaken right away to reach these goals in a not-so-distant future.
All of the optical switches that are being developed should have one common capability: integration. Taken in a loose sense, integration means a number of things: miniaturization, built-in units, and multi-functioning abilities. These components should be small for scalability, power consumption, and space consumption reasons. They should also be adapted to an automated manufacturing process for obvious cost-reduction reasons. An automated manufacturing process can be implemented more easily if the fabrication of a component involves starting with a single substrate on which features are built-in progressively through multiple successive treatments.
Integration also leads to adding different photonic functions on the same substrate— combining optical switching and wavelength conversion, for example. Such integrated optical switches should be based on a waveguide configuration: waveguide switches reliably stabilize the light path with respect to the switching elements and input/output ports, even during shock, vibrations and other environmental influences.1
The evolution of increased integration should ultimately lead to the feasibility of photonic chips that might become as popular as currently available electronic chips. Presently, the most promising path toward such photonic chips is through the development of photonic-bandgap structure components, which might eventually assimilate the developments made in creating wavelength-selective fast optical switches for tunable OADM and optical crossconnect applications.2
The current developments leading to fast optical switches are associated with two different types of materials: semiconductor and electro-optic. The semiconductor material is well-known since it is associated with laser-diode sources of all kinds. Because of its large refractive index, this material has the disadvantage of being associated with large optical-fiber coupling losses. However, this same characteristic offers the possibility of higher integration levels since the waveguides on the semiconductor substrate can be made closer together, with larger curvatures.
As for electro-optic materials, the best-known of this group is lithium niobate (LiNbO3), from which most high-speed modulators are made. Other materials such as doped optical polymers offer electro-optic characteristics and should also be considered to make fast optical switches. The refractive indexes of these electro-optic materials are smaller than for semiconductor materials; they offer smaller optical-fiber coupling losses, but they also provide a lower integration level. Electro-optic material might thus be considered a better choice for smaller-port-count switches and semiconductor material more suitable for higher-port-count switches.
For both semiconductor and electro-optic materials, there are many different configurations enabling the use of these substrates as fast, wavelength-selective optical switches. To name just a few, there are the phasar/Mach-Zehnder interferometer combination and the Bragg grating configurations.3, 4 Another promising configuration is the tunable optoelectronic-frequency filter (TOFF).5
TUNABLE WAVELENGTH FILTER
In a preferred embodiment of the TOFF configuration, three waveguides are inserted in a LiNbO3 substrate (see Fig. 2). The three waveguides have different geometries, and thus different propagation constants—different enough, in fact, so that even though they are in close proximity, the phase-mismatch forbids any evanescent wave coupling between them. Coupling only occurs when a periodic perturbation of the refractive index of the material compensates for the differences between the waveguides to create phase-matching over a given spectral range. The periodic perturbation is created by a well-known technology called PPLN for periodically-poled lithium niobate.6
Preferably, the period of the perturbation is chosen to create forward-propagating coupling. The periodic poling of LiNbO3 does not create the needed refractive-index grating directly. It only exists once an electrical voltage is applied to the poled region. The stronger the electric field is, the stronger the amplitude of the refractive index grating, and thus the greater the coupling efficiency will be. If the grating is longer and the difference between the two side-by-side waveguides is larger, then the spectral response will be more selective.
To obtain wavelength tunability on the resulting spectral response, the average refractive index of one of the two waveguides involved in the coupling region must be modified by locally changing the applied voltage. Either independent electrodes or a unique one can be used to create both the refractive-index grating from the periodically poled region and the tuning average-index change on the unperturbed region of the coupler.
Adding a third waveguide after the first coupling region creates a possibility of spectral slicing when the spectral responses of the two coupling regions do not overlap perfectly. This way, not only can the center wavelength of the filter be tuned, but its spectral response bandwidth can also be tuned by adjusting the overlap of the spectral regions of the different coupling regions. The device offers three possible outputs having three different and controllable spectral responses. This design can be upgraded to a larger number of outputs or of colinear waveguides, each having its own tunable coupling region.
The targeted specifications for the TOFF, which is presently under development, include adjustable bandwidth filtering from >2000 GHz to <100 GHz; covering at least the 1530- to 1560-nm spectral region; required tuning voltage not exceeding 20 V; and response time <100 ps. The configuration should also be adapted to minimize polarization-mode dispersion.
If a mature electro-optic material with photosensitive properties could be found, the periodic refractive index grating that creates the phase-matching in the TOFF could be permanently imprinted on the substrate by simple illumination. In this case, there would be no need to use an electric field over the grating region, and the strength of the coupling efficiency would not be electrically tunable.
The TOFF configuration is just one example of what the future holds for wavelength-selective high-speed optical switches. The intrinsic flexibility of this configuration allows coupling efficiency, bandwidth, and wavelength. However, many more similar configurations are presently being developed throughout the world.
Optical crossconnects, the centerpiece of optical communication hubs, should offer a wavelength-conversion function allowing network managers full access to all the information capacity offered by WDM systems, in addition to multiple tunable OADM functions. Lithium niobate also offers the wavelength-conversion function through its nonlinear coefficient. Again, a periodically-poled perturbation can create phase-matching between the incident wavelength and the wavelength to which the signal should be converted, possibly through a difference frequency mixer using an external source.7
If the wavelength-selective fast optical switch is made out of LiNbO3, like the TOFF, it is quite straightforward to envision how the optical switching and the wavelength-conversion functions can be integrated within a single component (see Fig. 3, top). Such an optical crossconnect can be used within a long-haul WDM network (see Fig. 3, bottom).
There is still a lot of work to be done before these next-generation optical switches become commercially available. The motivation behind such a development is very important, so many R&D groups are focusing attention on this challenge.
For this massive worldwide effort to occur, the optical communications industry should make sure that efforts are not concentrated on the qualification of the current optical switches that provide reconfiguration, restoration, and protection, and that the larger picture is not blurred by the immediate excitement. Long-term development should co-exist with short-term efforts. The signals being sent by the industry on this matter seem to be reassuring.
Fast, wavelength-selective optical switches are needed to fulfill the demand for tunable OADM and optical crossconnects in the next generation of WDM networks. These components will not only serve long-haul systems, but should also prove to be very useful in metro and even access networks if designed properly.
The authors would like to thank Chiara Meneghini, project leader, integrated optics; Georges Bladenberger, project leader, PPLN; and Claude Paré, project leader, numerical simulations, at INO for their fruitful discussions.
- M. Zdeblick, Laser Focus World, 139 (March 2001).
- M. D. B. Charlton et al., Mat. Sci. and Eng. B 49, 155 (1997).
- C.G. Herben et al., IEEE Phot. Tech. Lett. 10, 678 (May 1998).
- T. Cahall, Lightwave, 70 (December 2000).
- J. Lauzon and P.-Y. Cortès, "Tunable Optoelectronic Frequency Filter," Canadian and US patent filed on November 30, 2000, assignee INO.
- M. Yamada and K. Kanagawa, "Method of controlling the domain of a nonlinear ferroelectric optics substrate," US patent #5193023 (March 1993).
- M.-S. Chou et al., OFC 2000, postdeadline paper FB1-1, 16 (March 2000).
Jocelyn Lauzon is director of photonics and guided-wave optics, Pierre-Yves Cortès is head of optical communications, and Francis Généreux is a scholar at INO, 2740 Einstein street, Ste-Foy, Que., Canada, G1P 4S4. Jocelyn Lauzon can be reached at 418-657-7006, or firstname.lastname@example.org.