Trends in pump laser diode markets and technology
Previously, 980- and 1480-nm technology battled for supremacy. Advances in technology should widen the application of both types of modules, particularly 980-nm devices.
Toby Strite and Gerlas van den Hoven
Uniphase Corp.
The development of erbium-doped fiber amplifiers (EDFAs), more than any other technology, has enabled the proliferation of high-bandwidth data networking. EDFAs coherently amplify 1550-nm signals through the conversion of either 980- or 1480-nm pump laser light. Because the process is all optical, many wavelengths can be amplified simultaneously with no delay and minimal electronics. Service providers benefit from a transparent network that transmits data point to point without the need for extensive data management in between. The pump lasers powering EDFAs must be highly reliable but at the same time are pushed to their physical limits to provide maximum amplification power.
An important milestone in optical networking occurred in 1993 when MCI announced a successful fiber link between Sacramento, CA, and Chicago, using EDFAs powered by 980-nm pump lasers. Previously, only 1480-nm pump lasers based on the mature InGaAsP material system were in service, but these systems were limited in both bandwidth and range. Other service providers quickly followed suit, and 980-nm pump demand burgeoned, soon surpassing 1480-nm devices in market share. The availability of reliable, high-power 980-nm pump lasers has played a major role in enabling the commercialization of high-bandwidth wavelength-division multiplexing (WDM) systems.
The current shift to dense WDM (DWDM) systems capable of transmitting hundreds of gigabits per second over 64 or more 2.5-Gbit/sec channels or up to 40 channels of 10 Gbits/sec each has radically affected EDFA design and the specifications of their pump lasers. DWDM has also changed the status quo between pump laser wavelengths, which has produced a resurgence of 1480-nm pump laser demand. Today, both technologies are essential to supply the performance required in current and future optical communications links. As optical networking architectures continue to evolve, further impact in the pump laser marketplace is foreseeable.
980- and 1480-nm pump lasers
The pump laser market is especially interesting because of the availability of complementary 980- and 1480-nm alternatives. Single lateral-mode laser light at either wavelength can efficiently amplify 1550-nm light in an erbium-doped fiber. However, the semiconductor materials used to produce each wavelength and the physics of the erbium atom introduce technical tradeoffs.
Global forces also exert an influence. The 980-nm wavelength lasers are almost exclusively manufactured in the United States and Europe, whereas Japanese companies dominate the production of 1480-nm lasers. Not surprisingly, therefore, Japanese EDFA designs tend to have higher 1480-nm content than the rest of the world, which employs predominantly 980-nm pumps. Fluctuations in the exchange rate between the yen and dollar also influence wavelength design choices.
All telecommunications-qualified 980-nm lasers are based on the ternary AlGaAs and InGaAs alloys of the AlGaInAs material system. The excellent lattice match, refractive index contrast, dopability, and thermal conductivity of AlGaAs give designers freedom to optimize the vertical structure, while a single pseudomorphic InGaAs quantum well active region produces high gain, good electrical confinement, and therefore a low-threshold current and high quantum efficiency.
Molecular beam epitaxy and metalorganic vapor phase epitaxy (MOVPE) are mature production techniques for growing high-quality AlGaAs lasers. Wet etching is used to form a ridge waveguide for lateral confinement of the optical field. The ridge is embedded in an insulating dielectric (e.g., SiNx) for lateral current confinement. An alternative approach for lateral optical and electrical confinement is ion implantation followed by annealing.
A key part of any reliable AlGaAs laser technology is the mirror passivation. If the cleaved mirror is not properly protected, degradation of the mirror related to aluminum oxidation eventually leads to catastrophic optical mirror damage (COMD), a failure mechanism caused by absorption at the facet leading to thermal runaway. But patented E2 mirror passivation techniques completely suppress COMD, proving once widespread claims of inherent unreliability of Al-containing lasers to be totally without merit.
1480-nm pump lasers are based on quaternary and ter nary alloys of the InGaAsP material system, which includes all long-wavelength (1200- to 1400-nm) telecommunications lasers. Laser heterostructures are deposited by MOVPE onto InP substrates. Multiple (three to six) quantum wells separated by barrier layers form the active region, which is surrounded in the vertical structure by optical and electrical confinement layers. Because of the different atomic radii of In/Ga, and As/P, all layers must be lattice-matched to the InP substrate or pseudomorphically strained. Lateral optical confinement is achieved by a wet chemical ridge etch, while lateral current confinement is realized by regrowth around the ridge by either liquid phase epitaxy (LPE) or MOVPE. With LPE, a sandwich structure of p- and n-doped layers forms a reverse-biased junction to block lateral current flow. Using MOVPE, an insulating InP:Fe layer serves the same purpose. LPE forms near-perfect interfaces between the regrown and original layers that can improve chip reliability, while MOVPE regrowth is a higher throughput process with widespread commercial availability.
The fabrication of both 980- and 1480-nm pumps require special skills, but in different steps of the device processing. Fabrication of 980-nm lasers does not require expensive epitaxial regrowth, but the AlGaAs mirror passivation adds significant cost and complexity. As a result, only a handful of companies have developed a Bellcore-qualified 980-nm pump laser manufacturing process, and even fewer can realize greater than 200-mW output power with high reliability.
The buried heterostructure 1480-nm laser design involves epitaxial regrowth, and the lower InGaAsP thermal conductivity requires junction-side-down mounting for proper device characterization. Each of these factors complicates manufacture and adds cost. On the other hand, InGaAsP laser mirrors require no processing beyond standard reflectivity modification because the Al-free facet is not a critical failure mode. Bellcore-qualified 1480-nm chips are more widely available, with at least 10 suppliers actively vying for market share.
The table lists the relevant performance parameters associated with 980-nm versus 1480-nm pumping. Because it lies close to the 1550-nm signal band and is a two-level pumping process, 1480-nm excitation introduces more amplification noise than three-level 980-nm pumping. Amplification noise is especially harmful to DWDM systems having narrow 100-GHz (or 50-GHz) channel spacing and to long-haul systems that cascade many EDFAs, so 980-nm lasers are strongly preferred for small-signal amplification. On the other hand, 1480-nm lasers provide higher optical conversion efficiency due to the single photon conversion process of erbium amplification. Converting a 980-nm photon (1.265 eV) to 1550 nm is approximately 50% less power-efficient than using a 1480-nm photon (0.838 eV) for the same job.
The lower optical efficiency of 980-nm pumping is offset at the systems level by better device efficiency due to the properties of AlGaInAs, which are more conducive to high-power laser operation than those of InGaAsP. Figure 1 shows the optical output power versus drive current for a typical 1480-nm pump laser chip. Thermal rollover effects, already apparent at 100-mA drive current, limit the overall device quantum efficiency at 200-mW output power to 0.36 W/A, compared to the greater than 0.9 W/A typically obtained in 980-nm lasers. As a result, a 16-dBm EDFA powered by a 100-mW, 980-nm pump consumes less than half the electrical power of the same design powered by a 65-mW, 1480-nm device, while dissipating less heat.
If pump lasers producing the same EDFA output power are available at a comparable price, EDFA designers generally prefer 980 nm because of its lower noise and reduced thermal and electrical demands. In 1997, when this condition was largely true, we estimate that 980-nm pump lasers enjoyed a 2:1 unit volume advantage over 1480 nm.
In 1998, two factors raised the market share of 1480-nm pump lasers. DWDM systems require more gain and EDFA output power than their predecessors to amplify and manage additional channels. The demand for high pump power was largely filled by 140-mW, 1480-nm lasers because the equivalent 200-mW, 980-nm pumps were not commercially available. Additionally, the decline of the yen provided a pricing advantage to the Japanese producers of 1480-nm pumps. In 1999, the available power will continue to favor 1480 nm, although the performance gap will narrow somewhat as 200-mW, 980-nm pump modules become available.
A crucial issue to all telecommunications system operators is reliability, and pump lasers, as a critical active component, undergo particularly close scrutiny. The general perception has been that 1480-nm lasers are more reliable than 980-nm devices. InGaAsP lasers tend not to fail suddenly, but degrade slowly and predictably over time. Early on, AlGaAs lasers failed frequently, suddenly, and unpredictably due to COMD and dark-line defect propagation. However, the suppression of dark lines through better crystal growth and the development of mirror passivation procedures have eliminated the major sudden-fail mechanisms.
Meanwhile, adequate field data have been accumulated with 980-nm pump lasers to develop a convincing reliability picture. Nortel Networks has monitored a group of 80- to 120-mW modules over greater than 190 million field hours. Their data now show 980-nm pump module reliability to be 110 FIT with a 60% confidence level, whereas their announced value stood at 180 FIT only eight months before. Extended life testing (see Fig. 2) has allowed Uniphase to lift its estimate of the mean time to failure (MTTF) for 980-nm lasers above two million hours, whereas one year ago we could only claim 1 million hours MTTF. Both results suggest that as 980-nm pump lasers continue to accumulate a track record, the reliability gap will narrow further or disappear altogether.
Also shown in the table is the available erbium-pumping band, a characteristic that has important implications. The absorption spectrum versus pump wavelength around 980 nm is strongly peaked. As a result, 980-nm pumps are increasingly wavelength-stabilized using fiber Bragg gratings (FBGs) to eliminate gain variations caused by pump-to-pump wavelength and spectral differences. Wavelength variation in 1480-nm pumps is less of a concern, since erbium absorption increases monotonically with pump wavelength, but much more slowly than at 980 nm. The broader absorption band also enables more 1480-nm pump lasers to be usefully multiplexed into a single fiber to realize the highest possible EDFA output power.
DWDM`s effects on design
Because pump lasers are an EDFA component, pump requirements are dictated by amplifier designers who must constantly improve their EDFAs to support higher-bandwidth system architectures. Today, DWDM is the major force influencing EDFA design and the performance of the diode lasers that pump them. Increased channel count necessitates proportionately higher total pump laser power. For example, if 64 wavelengths are used in place of 16, EDFAs must have four times higher output to continue launching the several milliwatts per channel required to reach the next EDFA.
DWDM has also added to the functionality demanded of EDFAs. New systems that exceed 64 channels, even if channel spacing is halved to 100 GHz (0.8 nm), use more of the 1550-nm low-dispersion window of silica fiber. As a result, EDFAs require gain flattening to compensate for the wavelength dependence of the erbium gain in the 1550-nm band. Long-haul DWDM systems can cascade many EDFAs. If EDFA gain flatness across the entire band is not tightly controlled, the channel power distribution can eventually vary widely. This variance introduces deleterious nonlinear effects in higher-power channels and starves other channels of power. Also, different wavelengths tend to propagate at slightly different speeds in the optical fiber, an effect called dispersion. Dispersion must be closely managed at 10-Gbit/sec modulation rates because of the smaller time window between successive bits compared to 2.5-Gbit/sec transmission.
Both gain flattening and dispersion compensation are increasingly accomplished by specialized FBGs integrated into the EDFA. However, each grating introduces loss, adding further to the total pump power requirements.
To combine high output power with added EDFA functionality in DWDM systems, up to three high-power pump stages are used (see Fig. 3). In the first stage, the dissipated (approximately -30 dBm) incoming signal is pre-amplified by counter-propagating 980-nm pump lasers whose low noise figure make them the clear choice for small-signal amplification. The FBGs fix the pump wavelengths and narrow the spectra, making pumps more interchangeable and, if necessary, easy to multiplex. Increasingly, 973- to 977-nm lasers are chosen because the per-channel gain distribution (i.e., gain tilting) is less sensitive to pump power variations at these wavelengths. Despite reduced optical conversion efficiency at shorter wavelengths, the resulting relaxation of other component specifications makes this tradeoff attractive for high-performance DWDM systems. To compensate for the short pump wavelength and the lossy FBG elements to come, first-stage 980-nm pump laser modules require greater than 150 mW of linear power, significantly more than was previously needed.
Next, a specialty FBG in conjunction with a variable attenuator (VAT) equalizes the channel powers before a chirped FBG compensates dispersion. The FBG and VAT losses necessitate a second high-power pump stage (either 980 or 1480 nm, depending on the EDFA specifics) before another gain-flattening FBG pre-adjusts the channel-power distribution in anticipation of the booster-amplification stage. In the booster, as many multiplexed FBG-stabilized pumps as necessary provide pump power adequate for launching 23 to 26 dBm of total signal into the network. The larger effective pumping power at 1480 nm presently makes this wavelength the preferred choice in the booster stage, since noise generation can be accommodated by providing an adequate input signal power. In general, the choice of pump wavelength in the booster (and sometimes the intermediate stage) is determined by dBm output per dollar, a figure of merit that currently favors 1480-nm lasers.
As a result of increased channel counts and the transition to OC-192 (10 Gbits/sec), the roughly 100-km typical amplifier spacing in long-haul DWDM networks is predicted to decrease over the next few years. Added EDFA functionality for DWDM introduces large internal losses that must be compensated by higher-power pumps and additional pump stages. More amplifiers per kilometer and more pumps per EDFA have created a dramatic surge in pump laser unit demand in 1998 that should extend into 1999. In response, both 980- and 1480-nm pump manufacturers are redesigning their chips for higher power and rapidly adding capacity.
Other developments
The introduction of 980-nm pump laser technology into submarine networks portends another major market shift in 1999. Submarine networks are long-haul in nature, necessitating that many EDFAs be cascaded between landfalls. Previous 8-channel, 2.5-Gbit/sec systems could tolerate the noise generation of 1480-nm pumps; however, with submarine designers moving to 16 10-Gbit/sec channels, the improved noise performance of 980-nm pump lasers is now indispensable.
Submarine systems also require extremely high reliability due to the inaccessibility of key components once deployed. In the past, network designers have exclusively chosen 1480-nm pumps for undersea applications because of their excellent reliability track record. In December, Uniphase, in conjunction with Lucent, announced the first submarine-qualified 980-nm pump laser module, which is being deployed by Tyco Submarine Systems. The higher bandwidth resulting from undersea 980-nm pumps is expected to drive a rapid transition from 1480 to 980 nm in the submarine market, much like what occurred five years before in terrestrial systems.
This year will also bring Bellcore-qualified multimode AlGaInAs pump lasers for next-generation fiber amplifiers capable of greater than 30-dBm output power. Demand for high-power (greater than or equal to 1W) multimode pump modules is driven by the need to reduce the total number of pump modules required for high-power EDFAs. Typically, 70% to 80% of pump laser cost is in the butterfly package, so considerable cost can be spared using fewer modules, especially considering the additional cost of wavelength combiners for pump laser multiplexing.
Many manufacturers are readying Yb/Er cladding-pumped fiber amplifier and Raman-shifted amplifier technologies for commercial production in 2000. Driven by the availability of reliable, high-power multimode pump lasers, these new amplifier technologies will enable further DWDM capacity expansion at ever lower cost. u
Toby Strite is product manager for pump lasers at Uniphase Laser Enterprise. Gerlas van den Hoven is product line manager for amplification products at Uniphase Netherlands. They can be reached at [email protected] [email protected], respectively.