The external-cavity laser has moved to the forefront of DWDM tunable-laser technology through the use of micro-electromechanical systems.
CINDANA TURKATTE, iolon Inc.
Today's data-hungry telecommunications environment relies increasingly on the ability to transfer large amounts of data over great distances in very little time at very small cost. The full potential of the Internet cannot be realized until network bandwidth can be optimized to handle the intense demands of the data-rich, selectively provisioned applications demanded by the Internet economy.
DWDM has emerged as an answer to the bandwidth bottleneck. Until recently, each wavelength in a wavelength-multiplexed system required its own fixed-wavelength source laser to produce a signal at the desired frequency. Add to that the fixed-wavelength lasers needed for each switch and add/drop multiplexer, plus the lasers needed for backup and sparing, and the telecom industry was facing huge inventory, complexity, and cost problems to fully implement DWDM.
Consequently, tunable lasers have become a quasi "Holy Grail" of optical networking, with the promise of supporting flexible wavelength-routing capabilities and realizing the all-optical network. If a single tunable laser can replace an array of fixed-wavelength lasers, then the optical network will realize step function increases in flexibility and capacity, with massive reductions in network complexity, wavelength contention, and investment in inventory.
Tunable lasers have been under development for over a decade, but in the past, tunable-laser technology did not meet the basic optical performance parameters, such as output power and wavelength spectral performance, necessary for telecom applications. Recent advances in tunable-laser technologies have allowed tunable lasers to approach the performance levels of their fixed-wavelength cousins. There are several new technologies that allow a single tunable laser to provide a multitude of wavelengths, which previously would have required many multitudes of fixed-wavelength lasers.
There are several distinct applications within DWDM that have varying requirements for tunable lasers. There are three main dependencies affecting performance requirements for the transport markets: distance (access versus metro versus long haul), channel spacing (100, 50, and 25 GHz), and data rate (2.5, 10, and 40 Gbits/sec). In addition to the transport applications, tunable lasers will also enable switching and optical add/drop multiplexing applications that will drive a different set of performance requirements largely dependent on output power, tuning time, and tuning range.
Given the variety of applications with differing needs, it has become clear that no tunable-laser technology is optimized for all of these DWDM requirements. The choice of laser technology will be determined by many factors, including output power, line width, relative intensity noise (RIN), tuning range, tuning time, and stability. It is important to understand the relative strengths and weaknesses of the different tunable-laser technologies against these requirements when considering what type of laser to implement for a given optical-network application.
Tunable lasers can be produced using a variety of laser structures, each with its own advantages and disadvantages. Five of the main basic laser structures are distributed feedback (DFB), distributed Bragg reflector (DBR), sampled-grating DBR (SGDBR), vertical-cavity surface-emitting laser (VCSEL), and external-cavity lasers (ECL).
DFB lasers are simple structures that work by using an internal grating to change the wavelength of operation (see Figure 1). They are tuned to Inter national Telecommunication Union (ITU) grid wavelengths by changing the temperature of the medium, either through drive-current changes or a temperature-controlled heat-sink, which changes the refractive index of the internal waveguide. With improvements in thermo-electric coolers, the temperature can be precisely controlled to produce a stable, well-defined output wavelength with an acceptable line width. In general, DFB lasers are well behaved and characterized and have proven reliable.
While DFBs offer some manufacturing, performance, and operational advantages, they have the disadvantages of low output power and very limited tuning range. Their effective tuning range tends to be limited to about 5 nm because, as the tuning temperature increases, the efficiency and output power of the device decreases.
To extend the tuning range, these devices can be integrated into ensemble or side-by-side laser arrays of several lasers on one chip coupled into a single output. One laser at a time is driven to select a wavelength. There are, however, some limitations even to this design. This solution is not continuously tunable, the combining mechanism is optically inefficient, and the chip size leads to yield issues.
Cascade or inline laser arrays overcome the coupling loss by allowing light to pass transparently through other laser sections on the device (see Figure 2). Their main challenge is achieving mode stability for each of the laser sections.
DBR lasers consist of two or more sections with at least one active region as well as a passive region (see Figure 3). The passive region contains a grating, and each end of the laser cavity has a reflective surface. The DBR uses current changes to the passive region to change the refractive index and thereby tune the laser frequency. The DBR differs from the DFB in that the active region and grating region in the DBR are separated, while in the DFB, they are combined.
DBRs have some appealing advantages. One is that the tuning time is very fast. Also, like the DFB lasers, DBR lasers are relatively simple and well understood. The continuous tuning range has been improved significantly to about 40 nm; however, designs can be limited by current saturation.
A drawback in using DBRs as tunable lasers is that it is difficult to control the optical-path length between the two reflectors at each end of the cavity, resulting in broad line width and wavelength instability.
SGDBR tunable lasers use grating reflectors at either end of the cavity to produce a spectral comb response. The back and front sampled gratings have slightly different pitch so the resulting spectral combs have slightly different mode spacing. Tuning to a specific wavelength is achieved by controlling the current in the two grating sections so as to align the two combs at the chosen wavelength. The laser, therefore, "hops" between wavelengths. An additional contact is normally required to adjust the phase so an integral number of half-wavelengths exists. If this phase adjustment is not included, then mode stability can suffer and noise can increase.
While these types of lasers offer a wide tuning range, they suffer from low output power and broad line width. Furthermore, they have complex drive requirements when compared to other laser designs. They have more electrodes, and accurate wavelength selection requires matching of numerous input currents with the appropriate electrodes.
Sampled gratings with grating-assisted co-directional coupler filters add a filter to select one of the sampled-grating peaks, making it easier and cleaner to select a wavelength but at the cost of additional manufacturing complexity. SGDBRs also suffer from low output power, which can potentially be recovered with a semiconductor optical amplifier (SOA). However, SOAs are noisy devices and can cause other perturbations that will prevent them from meeting requirements of RIN and line width for 10-Gbit/sec narrow-wavelength-spacing extended-reach applications.
Because SGDBR devices are grown on indium phosphide wafers, the ability of the SGDBR laser designs to meet all system noise, power, and tuning range requirements are limited to the physical characteristics that can be achieved by a single semiconductor system. In short, with all monolithic growth approaches, tradeoffs must be made between the material gain characteristics, electro-optical coefficients, and DBR current-tuning efficiencies. That is further complicated by the intricate fabrication process of the SGDBR device.
The VCSEL consists of a gain layer surrounded by mirrors on the top and bottom (see Figure 4). The cavity produces light that is emitted from the top surface of the laser structure, rather than the edge like conventional diode lasers.
VCSELs offer a few key advantages. One is that the emitted laser beam is circular and therefore much easier and less expensive to couple to a fiber. VCSELs demonstrate narrow line widths, show low power consumption, and can offer continuous tunability without mode hops. They also have some manufacturing advantages in that the devices can be tested at wafer level prior to dicing and packaging, which could lead to reduced manufacturing costs.
The main disadvantage of the VCSEL, however, is its limited output power. This limitation is fundamental to the VCSEL design, since there is a constraint to maintain a single spatial mode of operation with a very small active region.
ECLs are straightforward and well-understood laser designs. The external-cavity approach alters the beam wavelength by mechanically adjusting the laser cavity, rather than through current or temperature changes applied to a semiconductor material.
The grating-based ECL shown in Figure 5 is designed in the Littman-Metcalf cavity configuration. This laser consists of a separately fabricated gain medium (a simple Fabry-Perot laser diode) and an external cavity formed of separately fabricated optical structures (a diffraction grating and mirror) integrated at an assembly step. Wavelength tuning is achieved by applying a voltage to the micro-electromechanical system (MEMS) actuator, which rotates the mirror to allow a particular diffracted wavelength to couple back into the laser diode. The gain bandwidth of the diode, the grating dispersion, and the external-cavity-mode structure combine to determine the actual wavelength of the laser output.
ECLs with continuous tuning have been traditionally used in optical test and measurement equipment since they provide high power, large tuning range, and narrow line widths with high stability and low noise. Furthermore, they provide continuous tuning through the entire spectrum of the gain medium, where other common laser technologies (like DBRs) exhibit mode hops between stable points in the spectrum. However, ECLs were generally too large, costly, and sensitive to shock and other environmental influences to be used in telecommunications components.
Recent technological advances, however, have brought ECLs to the forefront of optical-networking component technology. In particular, the application of MEMS to optical-component designs produces high-performance micro-optics that readily fit on standard transmitter cards and can be manufactured at competitive costs in the optical-networking industry.
One key breakthrough in the development of MEMS-based ECLs is the use of deep reactive ion etching (DRIE) techniques to fabricate the MEMS actuators. DRIE techniques allow the cost-effective and reliable production of rigid mechanical drive structures that provide suitable force for high-speed and high-precision movement of optical elements over large linear and angular deflections. Further more, a low-cost, precision servo-control system can provide real-time error compensation, making the MEMS actuators quite accurate and insensitive to effects from shock, vibration, temperature changes, or creep.
DRIE MEMS actuators, externally fabricated optics, and high-precision servo-control systems are ideal building blocks for the creation of new optical-networking components. MEMS actuators can be readily combined with externally fabricated optics (diffraction-limited lenses, high-reflectivity mirrors, wavelength-selective coatings) using precision servo-control systems to rapidly develop solutions for critical telecom-component requirements. For instance, these building blocks could be repackaged to produce tunable receivers, polarization controllers, optical monitors, variable attenuators, optical switches, and tunable filters.
The DRIE MEMS ECL performs very well. Figure 6 shows a representative output spectrum of a MEMS ECL locked to the 100-GHz ITU wavelength grid with about 10-mW constant output power across a 13-nm tuning range. This design is capable of providing high-power output (products soon will be available at 20 mW) and can continuously tune across a 40-nm tuning range. Furthermore, the device exhibits narrow line width, low RIN, and excellent side-mode suppression-all while meeting the market requirements for small-module footprint.
There are several reasons why the MEMS ECL performs so well. Using MEMS to reduce the ECL geometry to the micro-optics level maintains the high-performance characteristics of traditional ECLs, while enabling excellent laser-frequency stability over ambient temperature, because the entire ECL fits on a thermo-electric cooler. Further more, the MEMS ECL solution easily meets the market form-factor requirements of small-module footprint and standard pinout configurations. By fabricating the MEMS actuator, precision optics, and laser-gain diode separately, each can be optimized without compromising between the different component functionalities.
Using servo-control of actuators allows most assembly steps to be performed passively for pick-and-place automated volume manufacturing, reducing the manufacturing costs, time-to-market, and time-to-volume. Further more, it provides in-use active optical alignment, allowing the system to self-correct in the event of environmental changes, shock, vibration, or creep.
ECLs do have some potential disadvantages. MEMS-based ECLs are potentially susceptible to shock and vibration. However, proper actuator designs and servo-controls can sufficiently mitigate these and other external influences. Also, it remains to be seen whether the MEMS-based ECLs can be manufactured at competitive costs to be viable for the telecommunications markets.
Table 3 outlines the relative advantages and disadvantages of the laser technologies discussed. When considered against the requirements of various optical-networking applications, we see that different tunable-laser technologies are more suitable for different applications. It is clear that the ECL, enabled with DRIE MEMS technologies, is a strong candidate for many of these applications.
The comparison is largely limited to the technical performance requirements for DWDM networking. It does not consider cost, availability, and reliability, which will be critical factors in deciding which laser technologies are employed in which applications, if at all. It remains to be seen which technologies will make it to market and survive.
Cindana Turkatte is vice president of marketing at iolon Inc. (San Jose, CA). She wishes to thank Dr. Jill Berger, manager of optical design; Dr. John Clark, CEO; Dr. John Grade, manager of micromachining design; Dr. Hal Jerman, director of micromachining design; Eric Selvik, product-line manager; and Dr. Yongwei Zhang, manager of optical integration, all of iolon, for their contributions to this article.
- M. Littman and H. Metcalf, "Spectrally narrow pulsed dye laser without beam expander," Applied Optics 17, 2224-2227 (1978).
- D. Vakhshoori, P. Tayebati, Chih-Cheng Lu, M. Azimi, P. Wang, Jiang-Huai Zhou, and E. Canoglu, "2 mW CW single mode operation of a tunable 1550 nm vertical cavity surface emitting laser with 50 nm tuning range," Electronics Letters 35, pp.1-2 (1999).
- B. Mason, G.A. Fish, S.P. DenBaars, and L.A. Coldren, "Widely tunable sampled grating DBR laser with integrated electroabsorption modulator," Photonics Technology Letters 11, pp. 638-640 (1999).
- J.D. Grade, "MEMS electrostatic actuators for optical switching applications," OFC 2001 technical paper.
- J.D. Berger, "Widely tunable external cavity diode laser based on a MEMS electrostatic rotary actuator," OFC 2001 technical paper.
- J. Hong, H. Kim, and T. Makino, "Enhanced wavelength tuning range in two-section complex coupled DFB lasers by alternating gain and loss coupling," IEEE Journal of Lightwave Technology 16, pp. 1323-1328 (1998).
- G. Gilder and C. Burger, "The Tunable Telecosm," Gilder Technology Report, December 2000, Vol. V, No. 12.
- RHK Telecommunications Industry Analysis, "The Role of Tunable Lasers in the Emerging Optical Network Infrastructure," Optical Components, September 2000.