Tunable lasers enable new optical networks to meet changing demands

The ability of one laser to tune across multiple channels offers both flexibility and economy for DWDM applications.

Robert Plastow Altitun AB

The explosive growth of dense wavelength-division multiplexing (DWDM) over the past three years has resulted in a proliferation of wavelengths used in optical transmission systems. The single-wavelength transmitters of a few years ago have been replaced by individual transmitters specified at 40, 80, or as many as 200 wavelengths, in addition to separate specifications for transmission distance and connector type.

This huge growth in component count has provided network operators with both significant challenges and opportunities. The challenges arise from the need to back up, inventory, and effectively manage the proliferation of wavelengths now being used. The opportunities come from the fact that each of these wavelengths is potentially another revenue stream, to be exploited where it is most needed in the network.

A new class of commercially available components--widely tunable semiconductor lasers--provides the means to both manage the existing backup and inventory-control problem and to enable flexible future networks where individual optical channels can be routed through the network to meet changing customer demand. These components, which can switch wavelengths in less than 20 nsec, eventually will additionally be used to route not only optical channels, but also individual data cells or packets, vastly increasing the throughput of data networks and switches.

The key to this capability is a semiconductor laser that can be tuned by current injection into the single laser chip to emit light at any wavelength in the erbium band. A tuning range from 1528 to 1565 nm is commercially available, and the technology has demonstrated tuning greater than 130 nm. This single component can replace the multiple-wavelength lasers currently used, first as a backup and then as a direct substitute.

Tunable laser technology

Any laser requires wavelength-selective feedback to operate. The key to making a tunable laser is to incorporate a sufficiently selective and wavelength-tunable element in the laser cavity. In the case of semiconductor lasers, the gain peak and refractive index both move with temperature, giving a limited wavelength-tuning capability. A simple Fabry-Perot laser will tune with temperature at around 0.5 nm/C, the movement of the peak optical gain. In doing so, it will lase at several wavelengths simultaneously, making it unsuitable for long-distance telecommunications applications. A distributed-feedback (DFB) laser provides feedback at only one wavelength and consequently tunes with the refractive index at a lower rate of around 0.08 nm/C.

The DFB is the laser of choice for existing WDM systems, and provides limited tunability. Up to 5 nm has been demonstrated, using a 60°C temperature excursion. This "fry it and freeze it" approach is generally undesirable due to variations in launch power, high cooler-power consumption, and reliability considerations of both the laser and cooler.

The use of an external wavelength-selective element (typically a diffraction grating) with an antireflection-coated Fabry-Perot laser is the traditional method of obtaining broadly tunable laboratory sources in the 1550-nm band. But the mechanical nature of this approach makes it inherently slow and unsuitable for the reliability demands of telecommunications systems. An extension of this approach using micromechanical structures with surface-emitting lasers is presently at the research stage.

The standard solution to the requirement for a single-chip tunable laser is the distributed Bragg reflector laser (DBR). In this structure, an integrated Bragg reflector performs the wavelength selection. This is a wavelength-selective reflector similar to that used in a DFB laser, but placed in its own passive section of the laser chip. By current injection in this Bragg section, the laser wavelength can be tuned. Unfortunately, the tuning range of these lasers is again in the order of 5 to 10 nm, limited by the refractive index change caused by current injection. This limited range has proven insufficient to make them attractive sources for DWDM.

Wider tuning range

To achieve wider tuning, a mechanism is needed that can give wider wavelength change relative to the refractive index change. Two approaches have been tried. The first is to use a co-directional coupler. In this structure, light is coupled between two waveguides. This coupling is highly wavelength-dependent, and the selected wavelength depends on a difference between two refractive indices, enabling wider tuning for any given index change.1,2 Unfortunately, it is difficult to make these coupler filters sufficiently narrow to guarantee single-wavelength operation with the side-mode suppression required for telecommunications systems.

An alternative mechanism is the Vernier effect using two Bragg reflectors.3,4 In these lasers, two grating sections are used. Each grating is modified to reflect a comb of regularly spaced wavelengths. Lasing will occur where two reflection peaks overlap, which again can give large jumps in wavelength for small changes in refractive index but at a penalty of complex control and the requirement that light be extracted from the device through the grating.

A combination of the two approaches is the grating coupler sampled reflector laser (GCSR--see Fig. 1). The laser is a 4-section semiconductor chip, with currents for optical power and coarse, medium, and fine wavelength tuning. The first section is a gain section. Current through this section determines optical power. Behind this section is a vertical coupler section that acts as a coarse wavelength tuning element, tuning over greater than 100 nm. This section selects one of the wavelengths reflected back from the rear Bragg grating section. Medium tuning is accomplished by varying the current to this Bragg section, which itself can be tuned over greater than 4 nm. Finally, a fine-tuning section adjusts the optical phase to give the good side-mode suppression required. This single laser has the function of 10 or more DBR lasers, and can give complete wavelength coverage over more than 60 nm and a total tuning range exceeding 100 nm.5,6

The full laser structure is shown in more detail in Figure 2. It is grown by standard processes. The multiple growth steps and the use of both active and passive waveguide sections make it approximately equivalent in production complexity to an electroabsorption modulator/DFB combination.

This type of laser is longer than a standard DFB but is otherwise optically and mechanically similar, allowing it to be packaged in an industry-standard butterfly package, with isolator and cooler, using conventional laser welding techniques.

Tuning results on a typical module are illustrated in Figure 3. Coarse tuning is performed by the current in the coupler section, yielding a wavelength "terrace" that is accessible by the other tuning currents. The Bragg-section current tunes the laser along this terrace to the desired wavelength. Fine tuning is accomplished by the phase section current, to ensure high side-mode suppression. The straightforward tuning principle allows a relatively simple and practical wavelength-control system. The photo on page 77 shows a commercial module mounted on a microprocessor-based control board. Any ITU wavelength channel can be obtained.

Application requirements

The first application for tunable lasers is in backup of standard fixed-wavelength systems. Typically, a Synchronous Optical Network/Synchronous Digital Hierarchy-based WDM system will react to a fault by rapidly reconfiguring the data path through the network. A favored approach is a ring network, where traffic is redirected around the ring. This approach takes care of the immediate emergency but leaves the network vulnerable to further failure. It is imperative to replace the failed component as soon as possible to restore full network integrity.

With large numbers of wavelengths, it becomes extremely expensive to stock spare transmitter cards of each wavelength at the transmit site and problematic to ensure timely delivery from either a central inventory point or the manufacturer. This problem of inventory location extends beyond just the transmitter-card manufacturer to the component manufacturer.

By providing a tunable laser card on-site, which can be set to any wavelength, this problem is avoided. The tunable card can replace any failed fixed-wavelength card, and will continue operating until a replacement fixed-wavelength transmitter is delivered from the manufacturer. This capability minimizes inventory costs for the operator, network-equipment manufacturer, and component manufacturer, while providing the required rapid replacement of the failed component.

A second application is in flexible add/drop multiplexing. Most system providers deliver systems with fixed-wavelength optical add/drop multiplexers (OADM) in several nodes down a transmission link. This way, a basic level of optical networking may be achieved, and many network operators worldwide have already implemented OADMs.

Networks that span both North America and Europe are being established by several operators mainly based on WDM ring architectures that incorporate OADMs. The attraction of this approach is that optical signals can be routed where required in the network without the need to electrically demultiplex, wavelengths can be re-used, and different traffic formats can, in principle, co-exist. The drawback of fixed OADMs is that they are not configurable to accommodate changes in traffic demand. The use of a tunable laser allows rapid network reconfiguration, efficient bandwidth reuse, and wavelength protection.

These advantages also require development of the network-management system, so this application will follow the use of tunable lasers in simple backup. Eventually, full optical crossconnects (OXCs) will be required, switching any wavelength on any input fiber to any wavelength on any output fiber. By modulating the output of a tunable laser, either directly or through an external modulator, with the signal from an optical receiver, any input wavelength can be converted to any output wavelength. This conversion capability is a key functional requirement for OXCs.

A particularly interesting application is the direct replacement of fixed-wavelength lasers. Direct replacement would provide the significant inventory savings that apply in the backup case, with an additional significant simplification of equipment operation and maintenance.

Network operators would ideally like to see a universal card, whose wavelength is determined solely by its position in the equipment rack. This type of card would eliminate the possibility of human error both in installation and repair, as no manual wavelength selection or setting would be required. Inventory savings again would occur at all points in the supply chain, down to the laser manufacturer. Since this application would be a one-for-one replacement of existing lasers, the cost and volume targets would be more challenging than in the backup case. Thus, this application will be served as production volumes of tunable lasers ramp up to become equivalent to fixed-wavelength versions.

Switching capability is key

Future applications depend on the fast switching capability of semiconductor tunable lasers. Both the gain current and tuning currents can be modulated. Current devices are intended for external modulation of optical power and relatively slow wavelength switching. Using external lithium niobate modulators, 10 Gbits/sec over 80 km has been demonstrated, and using existing commercial control boards, the wavelength switching times are around 10 microsec. The laser itself shows a small signal response of 4 GHz and has been directly modulated at 1.2 Gbits/sec over 30 km. Future development may allow a low-cost, directly modulated OC-48 (2.5-Gbit/sec) tunable laser for use in metropolitan and access networks. Similarly, the wavelength tuning speed of the chip is very fast at less than 20 nsec (see Fig. 4).7 This capability may allow its future use in packet and cell networks and switches, where path routing is accomplished by fast wavelength switching of each cell, either in a large core data switch or as part of a network. This speed allows terabit data switching, but requires larger volumes and lower cost than is currently the case in the fixed-wavelength replacement market. u

References

1. M-C Amann, S. Illek, Tunable laser diodes utilising transverse tuning scheme, IEEE J. Lightwave Tech. vol. 11, no. 7, p. 1168 (July 1993).

2. Kim, Alferness et al., Broadly tunable vertical coupler filter tensile-strained InGaAs/InGaAsP multiple quantum well laser, Appl. Phys. Lett. vol. 64, p. 264 (May 1994).

3. V. Jayaraman et al., theory, design and performance of extended tuning range semiconductior lasers with sampled gratings, IEEE, J. Quantum Electr., vol. 29, p.1824 (1993).

4. H. Ishii et al., Quasi-continuous wavelength tuning in Super Structure Grating (SSG) DBR lasers, IEEE J. Quant. Electr., vol. 32 (3), p. 433 (March 1996).

5. P-J Rigole et al., Wavelength coverage over 67 nm with a GCSR laser. Tuning characteristics and switching speed, 15th IEEE Semiconductor Laser Conf., Haifa, p. 125 (Oct. 1996).

6. P-J Rigole et al., 114 nm wavelength tuning range of a vertical grating assisted codirectional coupler laser with a super structure grating distributed Bragg reflector, IEEE Photonics Tech. Lett. vol. 7, no. 7, p. 697 (July 1995).

7. P-J Rigole et al., Fast wavelength switching in a widely tunable GCSR laser using a pulse pre-distortion technique, Proc. Conf. on Opt. Fib. Comm. (OFC `97), Dallas (Feb. 97).

Robert Plastow is director of sales and marketing in the U.K. office of Altitun AB (Stockholm).