Where's the 'holy grail' of tunable lasers?

Th 96465

Trends in Tunability

A 'one size fits all' tunable-laser product is far less likely than an application-specific design.

DR. DANIEL CHU, Blue Sky ResearchTh 96465

Tunable lasers promise to provide tomorrow's telecommunications networks with new tools for wavelength management, improved network efficiency, and development of next-generation optical networks. Today though, the technology is immature and its volume manufacturability and long-term reliability have not been demonstrated.

Several variations of distributed-feedback (DFB), distributed Bragg reflector (DBR), vertical-cavity surface-emitting laser (VCSEL), and external-cavity laser (ECL) designs have been developed for DWDM applications-each with its own pros and cons-but no single technology has yet proved superior. The picture is further complicated because different tunable-laser applications have different requirements.

For example, the biggest current market is for tunable lasers to be used as spares for fixed wavelength sources. These spares require broad tuning range, but switching speed is not an issue. Conversely, switching speed can be critical to developing markets such as dynamic wavelength provisioning and routing.

Tunable lasers, regardless of their specific architecture, contain three basic elements: a source diode with an active gain section and resonant cavity; a tuning mechanism for varying and selecting wavelength; and a means of stabilizing the output wavelength. Source diodes are generally a variation of the Fabry Perot (FP) diode, with the notable exception being VCSELs. Tuning mechanisms can be temperature-, current-, or mechanically controlled, including micro-electromechanical systems (MEMS). Wavelength stabilization is provided by some type of wavelength locker or etalon operating in a feedback control loop.

Fabry Perot laser diodes
FP laser diodes are the workhorses of the photonics world. A typical FP diode is a multiple epitaxial-layer semiconductor chip containing the active layer (or layers) and some type of internal waveguide structure. The epitaxial structure is built of compound semiconductor materials, typically indium phosphide (InP) or indium gallium arsenide phosphide (InGaAsP) for C-band (1525-1565-nm) diodes. Th 96466

Figure 1. In a distributed-feedback design (a), the diffraction grating is integrated into the active section of the diode, while in a distributed Bragg reflector design (b), the grating is contained in a separate section coupled with the active section-though in both designs they are manufactured as monolithic chips.

The resonant cavity is formed within the waveguide by controlling the reflectivity of the cleaved reflecting facets at either end. These diodes are typically about 250 microns wide by 500-1,000 microns long by 100 microns thick, with the resonant cavity along the long axis of the diode. The output wavelength is a function of the gain material, whose refractive index controls the speed of the photons and the geometry of the resonant cavity.

In practice, FP diodes generate several closely spaced wavelengths over a few nanometers, and the output is very sensitive to slight changes in temperature and input current. With DWDM wavelengths spaced 0.8 nm apart or less, an FP diode requires refinements to provide the control necessary for optical networks.

Distributed Bragg reflectors
Tunable lasers based on DBR and the related DFB are among the most common designs. Each uses an FP gain section with an added diffraction grating. The diffraction grating (often referred to as a Bragg grating or Bragg reflector) provides feedback to the optical-signal oscillation and acts to select a single mode of oscillation, or wavelength, based on the pitch of the grating. In a DFB design, the diffraction grating is integrated into the active section of the diode, while in a DBR design, the grating is contained in a separate section coupled with the active section (see Figure 1)-though in both designs, they are manufactured as monolithic chips.

The wavelengths of DFBs and DBRs are tuned by altering the refractive index of the cavity semiconductor material, which changes the lasing conditions as photons travel between reflecting surfaces. The refractive index can be varied either by temperature control or by controlling the electrical current passing through. Th 96467

Figure 2. Some vendors are marketing sampled grating distributed Bragg reflectors, which extend the tuning range by adding a second diffraction grating situated at the opposite end of the cavity.

DFBs offer good power output (~10 mW), but their tuning range is limited to 2-5 nm. Tuning range can be expanded by combining multiple DFB diodes into an array on a single chip and merging their output, as is done by Fujitsu (Richardson, TX), but these are much more complicated to construct and control, and output power is somewhat reduced. To improve on their basic performance, several variations of DBRs have been developed.

Agility Communications (Santa Barbara, CA) and Marconi (Warrendale, PA) are marketing sampled grating distributed Bragg reflectors (SGDBRs), which extend the tuning range by adding a second diffraction grating situated at the opposite end of the cavity (see Figure 2). Each grating has a slightly different pitch, and the output wavelength is tuned by varying the current to the grating cavities. That in turn varies the refractive index of each, selecting for matching resonant frequencies and producing the output wavelength.

SGDBRs offer a wider tuning range, up to the C-band, but they are generally limited to lower power outputs-on the order of 2 mW. Higher powers-up to 10 mW-can be achieved by incorporating an additional semiconductor-optical-amplifier (SOA) section in the chip, but that again increases the cost and complexity.

Another version of the DBR is ADC's (Minneapolis) grating-assisted co-directional coupler with sampled reflector (GCSR), which contains four sections (gain, Bragg reflector, coupling, and phase correction) and is tuned using three currents. The current-controlled waveguide coupler acts as a coarse tuner to deliver a narrow range of wavelengths from the Bragg reflector (current-controlled) to the phase correction section (also current-controlled), which acts as a fine-tuning section (see Figure 2). Like most other multisectional diode lasers, power output is sacrificed and is low at 2 mW. Power output can be increased, at the expense of tuning range, by eliminating the coarse tuning section. The typical linewidth of DFB and DBR lasers can range from 5 to 20 MHz, which can cause dispersion problems.

External-cavity lasers
ECLs use a resonant cavity external to the active semiconductor section, typically a simple FP gain chip. A relatively large cavity containing a mechanically tunable (as opposed to current- or temperature-controlled) mirror/diffraction grating configuration is used to provide tuning. ECLs are generally based on a Littrow cavity or Littman-Metcalf cavity design. In both designs, one facet of the diode is antireflection-coated and the laser output is directed through a collimating lens to the cavity on one side of the diode.

The Littrow cavity is the simpler of the two designs and uses a diffraction grating that, by diffracting the beam back on itself toward the active section, also acts as one mirror of the cavity. Tuning is performed by mechanically rotating the grating, which changes its effective pitch. The Littman-Metcalf design uses both diffraction grating and mirror, whereby the diffraction grating diffracts the light to the mirror, which reflects the beam back to the grating and the active section. Tuning is performed by rotating the mirror, which varies the effective cavity length.Th 96468

Figure 3. An example of the new wave of external-cavity lasers, this device uses a non-mechanical, non-grating, electro-optical tuning mechanism in a standard butterfly package.

ECLs potentially offer better performance in terms of higher output power, wider tuning range, and narrow linewidth but have traditionally been too bulky and expensive to find practical application in optical networks. Their mechanical designs are also subject to hysteresis and simple wear and tear that compromises the long-term reliability required for telecommunications applications. However, new tuning-element designs are allowing ECL designs to incorporate their advantages of high power and wide tuning range in configurations suiting optical-network applications.

New Focus (San Jose, CA) has developed an ECL based on a mechanical tuning element for optical-networking applications. Details of the tuning mechanism are proprietary, but it reportedly offers 20-mW output over the entire C-band in a compact package sized for networking applications, with mechanical reliability meeting telecom standards. In another variation on the ECL, iolon (San Jose, CA) has applied MEMS technology to rotate a micro-mirror in a miniature Littman-Metcalf configuration. The iolon ECL provides 14-mW output power over the C-band, with 15-msec switch speed, in a standard butterfly package.

Blue Sky Research (Milpitas, CA) has also developed a compact ECL, using a non-mechanical, non-grating electro-optical tuning mechanism in a standard butterfly package (see Figure 3). Microsecond tuning speeds have been demonstrated, with output power of 20 mW over the entire C-band.

VCSELs are an entirely different approach to tunable lasers-one with exciting prospects but also significant challenges. VCSELs are epitaxially grown semiconductors, as are DBRs and ECLs, but the resonant cavity is perpendicular to the semiconductor layers rather than transverse. The cavity is formed by containing a quantum well gain section between a pair of mirrors (see Figure 4). The gain section is extremely thin-on the order of tens of nanometers.

VCSELs are small and allow high manufacturing density-up to 20,000 VCSELs can be grown on a single 3.5-inch wafer. Also, because the laser output of a VCSEL is emitted from its surface, and they are manufactured nearly to completion on the wafer, they can be fully tested while still on the wafer, greatly increasing manufacturing yields and lowering costs. Tunable VCSELs incorporate a movable MEMS mirror structure at the top of the resonant cavity, varying its length to tune the output wavelength. The tunable range can be over the entire C-band, with switching speeds of <10 msec.

The small resonant cavity and high-reflectivity mirrors result in increased efficiency and enable direct modulation. That eliminates the need for an external modulator and lowers operating currents to eliminate the need for thermo-electric cooling. Th 96469

Figure 4. In a vertical-cavity surface-emitting laser (VCSEL), the resonant cavity is perpendicular to the semiconductor layers rather than transverse. The cavity is formed by containing an extremely thin quantum well gain section between a pair of mirrors.

Another consequence of the small cavity is lower output power, typically on the order of 1-2 mW. However, a directly modulated 2-mW VCSEL is comparable to a 10-mW externally modulated laser, since external modulators typically consume a large fraction of a laser's continuous-wave output.

While VCSELs dominate the 850-nm LAN and SAN markets, material and manufacturing difficulties have hindered their entry into C-band applications. Because the resonant cavity in a VCSEL is so small, very-high-reflectivity mirrors are required to get good output power.

The AlGaAs/GaAs system of 850-nm VCSELs makes good mirrors but can't be used for a 1550-nm C-band gain section. The InGaAsP/InP used in C-band DBRs and ECLs makes a good gain section but makes poor mirrors. Unfortunately, GaAs/AlGaAs mirrors cannot be directly grown with an InGaAsP/InP active section due to the requirement for lattice matching, which restricts the materials that can be epitaxially grown together to those with similar spacing between atoms.

VCSEL developers are applying several very different approaches to extend wavelengths to the C-band. Epitaxial techniques use new materials such as InGaAs or InGaAsN for gain sections by varying their relative proportions to allow them to be grown on GaAs/AlGaAs substrates. That generally involves including some type of "strained layer compensation region" to accommodate the lattice mismatch.

Bandwidth9's (Duluth, GA) directly modulated tunable VCSEL is based on this approach, with the addition of a non-lattice-matched top mirror incorporated in a MEMS tuning structure. Bandwidth9's VCSEL covers the C-band with about 0.5-mW output (and up to 1 mW with an integrated optical amplifier).

Other methods include wafer bonding, which overcomes lattice-matching problems by fabricating the gain section and mirrors independently and bonding them together under heat and pressure. While that eliminates the defects resulting from lattice mismatching, new problems are introduced from imperfections, air voids, etc., in the bonding process, and the multistep manufacturing process increases costs. Hybrid approaches such as that used by Nortel Networks (Brampton, Ontario) use an additional optical pump laser and external MEMS mirror design to get 20-mW power at 1550 nm.

Viable choices
The market now offers several viable tunable-laser architectures, but it is too soon to tell if a given technology will take the lead. It's unrealistic to expect tunable lasers to evolve into a "one size fits all" technology. It's more likely that designs will be developed to be application-specific.

For example, sparing applications may be satisfied with millisecond, second, or even slower switching speeds, while dynamic switching architectures may require microsecond or higher speeds. Low power may be perfectly acceptable for short-distance applications. Cost, of course, is perhaps as important as any other feature, but for many tunable lasers, cost has not been announced or even determined. For sparing applications, cost must compete with fixed-wavelength diodes-an acceptable premium of 20% over a fixed transmitter cost is sometimes cited.

So the market may bear a cost of about $2,000 for a tunable laser that replaces a fixed laser diode with a wavelength locker. Dynamically switched DWDM networks are in their infancy, and it is not clear precisely what performance specifications will be required, let alone what will constitute acceptable component cost.

Trade-show speculation suggests that pricing on the order of two to three times that of sparing transmitters may be acceptable, depending on specific features. At this point, however, neither the network nor tunable-laser technology is mature enough to make a prediction. So only time will tell.

Dr. Daniel Chu is vice president of product-line marketing at Blue Sky Research (Milpitas, CA). He can be reached via the company's Website, www.blueskyresearch.com.

LIGHTWAVE Editorial Mission
Lightwave is a monthly international publication focusing on fiber optics and optoelectronics, the technologies driving the growth, convergence, and im proved performance of telephony, computer communications, and video. Lightwave provides technology news as well as applications and product information for corporate and technical managers and staff engineers. Lightwave's editors emphasize analysis and interpretation in their reports on the technological impact of fiber-optic components, systems, and networks in these markets.

More in Transmission