Tunable lasers will impact optical nets

he optical networking industry experienced exponential growth from 1996 to 2000, during which time a record amount of inventory was created by bullish organizations embracing the advanced capabilities of dense wavelength division multiplexing (DWDM). Because of DWDM, the number of wavelengths supported on an optical fiber grew from one to 200 in just five years, which meant that the supply chain had to ramp up to support 200 times the number of wavelengths.

At a time when the industry was growing very rapidly, DWDM technology created difficulties in forecasting and supplying lasers with the correct wavelength, which meant missed revenue opportunities up and down the supply chain. As a consequence, tunable lasers have emerged as promising solutions to reduce inventory and costs. In addition, tunable lasers enable new dynamic network architectures that are expected to radically decrease the cost of optical networking.

There are many optical networking applications for tunable lasers. These range from the immediate to those four or five years away. At Agility, we focus on applications expected to achieve wide deployment inside the next three years.

To date, deployment of optical add/drop multiplexers has been somewhat hampered, either by the lack of flexibility offered by commercial solutions, or by the cost premium. There are many ways that a reconfigurable optical add/drop multiplexer (R-OADM) can be implemented–from a tunable filter approach to a back-to-back multiplexer with a switching fabric between. In fact, some of the most promising R-OADM solutions use this secondary approach with a semiconductor optical amplifier-based switching fabric.

In either case the application demands tunable lasers. An R-OADM offers the ability to change the capacity being dropped and inserted at a remote site without the need for an expensive bandwidth manager. This results in major savings for the network provider. In this application, tunable lasers allow the resolution of blocking scenarios without the need for a full complement of transponders. These savings means reduced price sensitivity for the tunable laser component.

Photonic, or all-optical, cross-connects have been of great interest for some time but the technologies, such as microelectromechanical systems (MEMS), are just starting to come to fruition. The application of these photonic devices is varied.

A coarse-bandwidth management function can be implemented using a photonic switch surrounded by transponders, which overcome wavelength blocking and provide Sonet-like performance monitoring. Implementing the same function using tunable transponders can result in a 50 percent saving on transponder cards.

The tunable transponders can be removed from around the switch and moved to the edge of the core network, thereby enabling further reductions in transponder costs. In this case the wavelength selected at the edge of the network transitions the network optically, including the photonic switch (or bandwidth manager). This is really starting to approach the dream of an all-optical network.

If the photonic switch resides at a point in the network where two metro rings meet, tunable lasers can be used to enable traffic to transfer between rings without crossing an electrical switch. This can also be implemented using fixed-wavelength lasers, but one standby laser is required for every wavelength on the fiber. Clearly, tunable lasers save on transponder costs.

In these applications the tunable laser is enabling new functionality, which drastically decreases the capital and operating costs of the network.Consequently there is room to explore specifications and pricing on a customer-by-customer basis.

One key feature of a solid-state tunable laser is that it can be switched from one wavelength to another in a few nanoseconds. This enables wavelength switching at the packet level, which can then be used to route data traffic based on color through a passive router such as an N x N arrayed waveguide grating (AWG). Changing the wavelength of light at any input port will cause the data to emerge on a different port.

Several key tunable laser technologies are vying for supremacy in this market space--each has strong points and weaknesses to be overcome. Here are some of those technologies:

• The external cavity laser (ECL) is capable of very high spectral purity, high output power, and wide tuning range. These features make the emerging ultra-long-haul applications a natural target for this technology. Tuning is achieved by physically moving the mirror and grating in concert. While this shows much promise for controllability, there are some major challenges to overcome in qualifying the product and proving telecom-level reliability. The laser is fairly complex to manufacture and is therefore unlikely to achieve the cost points necessary to address the metropolitan or regular long-haul applications.
  
• Vertical cavity lasers (VCSEL) usually are used with a MEMS mirror to provide the tuning. There are two main types of tunable VCSEL: electrically pumped and optically pumped.

The electrically pumped architecture offers wide tunability at relatively low power levels. Using direct modulation, this approach can be well suited to metro-access solutions where low power and short-dispersion limited reach can be accommodated. This technology has the potential for low- cost manufacturing; the lasers chips can be tested on wafer, so only "known good die" are assembled, and coupling efficiency can be quite high due the circular nature of the beam. The power capabilities largely preclude this technology from addressing the long-haul applications.

The optically pumped variant of the tunable VCSEL offers slightly greater power and is targeted more toward long-haul applications. To achieve the power necessary for long-haul applications, it is necessary to use an additional semiconductor optical amplifier (SOA) chip to boost the output power; this adds cost and complexity to the manufacturing process. It is also possible to modulate the pump laser to address the metro-access market but the multichip assembly approach does not lend itself well to addressing the cost requirements for this space.

Both of these VCSEL approaches rely on small mechanical movement to provide the tuning function. The same reliability and qualification challenges outlined for the external cavity lasers also apply here.

• A grating-assisted co-directional coupler with rear sampled reflector (GCSR) tunable laser was one of the first to be widely sampled for telecom applications. It offers wide tunability and is an indium phosphide (InP) solid-state laser, so it should be a simpler platform on which to prove telecom-level reliability. This architecture is somewhat limited in output power and there is little scope to address this through monolithic integration of an SOA. This approach is being targeted toward metro-core type applications, though it could also be used in long-haul applications with an additional optical amplifier.
• The sampled grating distributed Bragg reflector (SG-DBR) tunable laser is probably the most flexible and extendable tunable laser architecture available today. It is widely tunable and can be monolithically integrated with an SOA for high power, long-haul applications, or an electro-absorption modulator (EAM) for metro-core applications. The performance of these lasers can rival the best of the approaches outlined previously, while the solid-state architecture promises telecom reliability and low cost.

Simultaneously meeting the desired specifications for these applications would require a tunable laser with high output power (more than 10 mW), wide tuning range (more than 32 nm), rapid wavelength tuning (less than 10 ns) direct or integrated modulation ( 2.5 Gbytes/s), high reliability (> 20 year mean-time to failure), and high volume production. The widely tunable, SG-DBR laser is the only tunable laser technology that has been shown to deliver on each of these areas.

The SG-DBR is a very elegant solution that addresses the need for wide-tuning-range lasers. A widely tunable DWDM laser presents an inherently difficult problem because conceptually it must be made more sensitive than is required to cover a narrower tuning range; at the same time, very precise wavelength selection and high stability is required.

The SG-DBR overcomes this conflict by using a tuning mechanism with two degrees of freedom. Instead of using one knob with really fine control required to select the channel (the more channels, the finer the control required), two knobs with coarser control are used in tandem (much like a row-and-column address in a matrix). In this way the tuning sensitivity required to achieve 100 50-GHz channels (40 nm) in an SG-DBR is actually less than that of a conventional DBR that can obtain only 20 50-GHz channels (8 nm).

Figure 1 shows a schematic of the SG-DBR, which is made up of a front mirror, back mirror, gain and phase section. The mirrors are what gives the laser its name, "sampled grating". A conventional DBR mirror is formed by a continuous grating having a narrow reflectivity spectrum that is used to select the desired ITU channel. This mirror can be tuned electrically a maximum of about 8-12 nm (limited by the properties of InP used to make the laser).

A sampled grating is a modification of the continuous grating in that grating teeth are periodically removed along the length of the grating (as depicted in Fig. 1), i.e. the grating has been "sampled". By sampling the grating, multiple reflection peaks ("aliases" of the original peak) are formed. These peaks are spaced apart in wavelength at a period inversely proportional to the period of the sampling. The front and back mirrors of the laser are sampled at different periods such that only one of their multiple reflection peaks can coincide at a time, as shown in Fig. 2a. (This is known as the Vernier effect). In this way the desired ITU channel can be selected by tuning the two mirrors (all of the reflection peaks move together) such that the closest reflection peak of each mirror is aligned at the desired channel and lasing occurs.

In addition to the wide tunability, the SG-DBR platform offers many other advantages. Given that the laser has both mirrors present (none are formed by a cleaved interface) on the wafer, the lasers can be wafer tested, offering many of the advantages touted by a VCSEL platform. Furthermore, the structure lends itself to further monolithic integration with other optical elements. Figure 3 shows an SG-DBR integrated with an SOA and EAM.

The benefit of this integration is that it is a very cost-effective way to achieve higher levels of performance and functionality. Figure 4 shows the performance of an SG-DBR/SOA, illustrating the uniform, high power (more than 10 mW) and spectral purity (SMSR more than 40 dB) over the entire C-band.

In addition to the SOA, it is possible to also integrate an electro-absorption modulator, allowing extended-reach tunable operation in a very compact and low- cost footprint. One of the reasons that this is such a good fit is that the EAM is essentially the same as the tuning sections (mirrors and phase); only it is operated in reverse bias. Thus, optimized, wideband, EAM performance in reverse bias is simultaneously achieved with highly efficient mirror tuning in forward bias.

Thus, the SG-DBR widely tunable laser can serve multiple market segments while addressing the concerns of reliability and cost evolution through a solid-state, monolithically integrated architecture and automated assembly process.