Charles Duvall
As data becomes the dominant traffic over carrier networks, end-user demands are becoming more dynamic. Consequently, it is crucial that technologies are developed to build dynamic optical metro area networks that are responsive to changing traffic patterns.
Standardization efforts, such as Optical Internetworking Forum, Internet Engineering Task Force, and Optical Domain Service Interconnect, which integrate data layers with a static wavelength optical layer, are among the first steps toward dynamic bandwidth or bandwidth-on-demand. Ultimately, metro dense wavelength-division multiplexing (DWDM) systems that are based on cost-effective 1550-nm tunable laser sources will be capable of providing wavelength-on-demand at the optical layer because these systems will be able to set up and tear down protocol-transparent wavelength services as required. Economical, long-wavelength, directly modulated, tunable vertical-cavity surface-emitting lasers (VCSELs) can play a key role in this transformation.
Metro system requirements
Traffic patterns in metropolitan areas are distinct from long-haul patterns because of the uneven flow of data traffic and the need to accommodate multiple protocols. As a result, equipment must meet different requirements. Typical long-haul DWDM systems are high capacity point-to-point transport systems with SONET/SDH network elements providing aggregation and protection. On the other hand, metro WDM systems are designed to collect traffic from a variety of physical locations within a metro area and aggregate many protocols into one transport system. Physical protection mechanisms, much like SONET/SDH, are integrated into the system.
Metro WDM systems based on fixed wavelength add/drop nodes work well with a limited number of wavelengths and with limited numbers of system reconfigurations. As service, traffic, and customer requirements change, system reconfigurations are required. However, many capital and operational cost are associated with reconfiguring existing deployments of static systems. This logistical problem, commonly called "churn," is currently solved, in part, by physically changing cards at the node and, if logistically feasible, reusing decommissioned line cards elsewhere. At least one industry study estimated that total costs associated with churn could amount to 30% of carrier operational costs.
In addition, to limit downtime due to failures, carriers use spare components to establish infrastructure redundancy. In the case of laser sources, this sparing methodology requires a redundant line card with fixed wavelength source and filters for each working wavelength at each add/drop node to meet SONET/SDH protection metrics. This scheme can easily equate to very large line card inventories and thousands of dollars per working wavelength for redundancy.
The logistics of ensuring that the correct fixed wavelength card from inventory is delivered to the correct add/drop node within the specified time adds operational costs-which could be avoided by using broadly tunable sources. Additional logistics problems could be reduced by remotely configuring a broadly tunable laser to a specific International Telecommunications Union (ITU) wavelength.
Narrowly tunable sources
Three technologies are used to create tunable laser sources-mechanical tuning, temperature tuning, and current tuning. All tune by changing the length of the optical cavity within the laser.
The classic, narrowly tunable laser is the distributed Bragg reflector (DBR) laser. In this structure, an integrated Bragg grating reflector performs the wavelength selection. The laser wavelength is tunable by current injection. This tuning method changes the refractive index (and thus the output wavelength) of the laser cavity. The limitation is the tuning range: generally on the order of 5 to 10 nm. The advantage is there are no moving parts.
A second type of narrowly tunable source is the distributed feedback (DFB) laser. The DFB is an edge-emitting laser and it is tuned by changing the temperature maintained by the thermoelectric (TE) cooler. This laser closely resembles lasers used as fixed wavelength transmission sources (which are temperature-tuned to reach their desired fixed wavelength).
Several commercial versions of tunable DFB lasers are available, with an electro-absorptive (integrated, on-the-die) modulator and an integrated Fabry-Perot wavelength stabilizer. Benefits of these lasers include compactness, cost-effectiveness, and high power output for long transmission distances. The primary limitation is they have a very narrow tuning range-when used as a backup for a fixed wavelength laser, generally one tunable DFB laser will act as a spare for up to four fixed-wavelength lasers.
Broadly tunable sources
While DBR and DFB tunable lasers have narrow tuning ranges, one version of a broadly tunable laser builds on the methodology for tuning DBR lasers. This laser-the grating-coupled sampled reflector (GCSR) laser-starts with a modulated Bragg reflector, which is coarsely tuned across a range of output peaks using a codirectional coupler. Each of these peaks can then be fine-tuned by changing the refractive index (and thus the output wavelength) of the laser cavity. Both levels of tuning are done by current injection.
In essence, the GCSR laser has the function of approximately 10 DBR lasers, with the tuning ranges interleaved so that the transmitter provides complete wavelength coverage over the desired band. This method overcomes the tuning limitation of the DBR laser but adds complexity, which results in manufacturing and characterization difficulties.
The tuning method using a GCSR laser is described as quasi-continuous, which means that any wavelength can be accessed by supplying the appropriate combination of tuning currents, but there is no mechanism to continuously sweep across the entire wavelength range.
A second type of broadly tunable laser is a dual-sampled grating design. In this system, two DBR gratings, each with a slightly different pattern of gratings, can be tuned by current injection against each other. Changing the current adjusts the alignment of the modes, causing lasing at different, and precisely controllable, wavelengths. As in the GCSR laser, tuning is quasi-continuous: all of the ITU grid wavelengths can be programmed, but there is no mechanism to scan across the entire band.
A mechanically tuned VCSEL has a very broad and continuous tuning range, which allows one device to act as a spare for any fixed wavelength laser within either the C (1530 to 1560 nm) or the L (1570 to 1620 nm) bands. It is directly modulated, which eliminates the additional cost of an external modulator. Although direct modulation introduces chirp onto the signal, within the distance limits of metro applications the addition of chirp will probably not introduce a problem.
Since the VCSEL is made using semiconductor manufacturing procedures, the laser can be inexpensively manufactuctured in volume. Also, because the device is surface emitting, it can be tested earlier in the manufacturing process, another source of cost savings. Finally, one version can be electrostatically tuned using a cantilevered microelectromechanical system (MEMS) mirror, which is a highly repeatable and predictable method of tuning.
Tunable VCSELs
The basic principle of operation of the tunable VCSEL is very similar to the operation of a tunable micromechanical Fabry-Perot (FP) cavity. An FP cavity is formed between two aluminum-gallium-arsenide (AlGaAs) DBR mirrors, with the upper DBR mirror suspended 1.2 µm above the bottom DBR by a cantilever. Selective etching of a sacrificial layer forms the cantilever, which in this case is fabricated from gallium arsenide (GaAs) (see Fig. 1). The cantilever remains straight and free of stress or strain after the selective removal of the sacrificial layer.
The resonance wavelengths are determined by the cavity thickness, which is the sum of the air gap thickness between the two DBR mirrors and the effective lengths (energy penetration depths) of the mirrors. One major operational difference between the VCSEL structure and the FP cavity is the air gap is no longer at the center of the cavity. This difference exists because the center of the cavity must coincide with the active region to maximize the overlap of gain and optical intensity (see Fig. 2).
The laser is tuned by moving the cantilever using an electrostatic force, which shortens or lengthens the air gap. The total tuning range is governed by the wavelength difference resulting from maximum deflection or the minimum free spectra range at any point of tuning, whichever is smaller. As the cavity is extremely short (~3 µm), the free spectra range-or the wavelength separation between two adjacent Fabry-Perot modes-can be very large. A typical tuning range is up to 40 nm, covering either the C or L band for telecommunication applications.
VCSELs can be fabricated so the light emits from either the bottom of the structure or from the top. While a bottom-emitting VCSEL is typically easier to fabricate, a top-emitting MEMs tunable structure is more desirable for optical coupling, packaging, and integration.
Toward advanced metro sources
Future WDM systems, optimized to meet the demands of metro area traffic, can use VCSEL-based line card sources. These subassemblies, consisting of a directly modulated tunable VCSEL, wavelength locker, detectors, filters, and appropriate electronics, will provide an economic spares solution for the near term, which will significantly reduce carrier inventory and logistical costs.
Since metro area systems traverse much shorter distances (less than 100 km) than long-haul systems, the chirp induced on the signal by direct modulation will not be a factor. Furthermore, the reduced footprint and reduced costs of this solution will make it highly desirable for metro applications.
More advanced applications may also benefit from tunable VCSELs. For example, optical cross-connects (OXCs), either with electrical or all-optical switch fabrics, are emerging to handle high-density optical terminations typically found at the interconnection points between long-haul and metro WDM systems. OXCs are used to groom and reroute traffic across optical tributaries. They also provide protection and restoration of the signal.
These functions can be distributed throughout the metro WDM transport network when tunable sources can be used to control the wavelength selection or wavelength interchanging. Wavelength tuning capabilities will increase the degree of connectivity and number of logical connections, and will support a variety of protection and restoration methods on physical ring deployments. A combination of these currently distinct network elements could ease the migration to the efficient implementation of metro WDM systems.
Charles Duvall is senior director of applications at Bandwidth9, 46410 Fremont Blvd., Fremont, CA 94538; he can be reached at 678-482-4021, ext. 40, or [email protected].FIGURE 2. The active region of this tunable VCSEL is a multiquantum well structure that is electrically pumped (V1). The bottom mirror is flat and the top mirror has a cantilever structure. Tuning is achieved by applying a small voltage (V2) to the top mirror, which causes the cantilever to move up and down. Changing the length of the cavity, which in turn changes the wavelength of operation. Direct modulation is achieved by varying the pumping current through the quantum well (V1/I1).