Isabella D. Jung and Stefan Mohrdiek
The development of uncooled pump sources for long-haul WDM has also given designers new options for metro amplifiers. Eliminating a Peltier cooler enables smaller, low-cost diode packages with much less power consumption.
Erbium-doped fiber amplifiers (EDFAs) are well-established in long-haul terrestrial and undersea networks, the performance of which depends critically on the pump laser. In particular, the development of 980-nm pump laser diodes have been optimized to fulfill stringent requirements for the submarine environment for reliability, efficiency, low power consumption, and wavelength stabilization over the entire operating and temperature range.
As optical networks move closer to the end user, amplifiers must not only compensate for distance loss but for signal losses due to the increasing number of nodes, add/drop channels, routers, and switches. For such metro systems, designers are increasingly concerned with the cost and power consumption of EDFAs. One important step toward reduced costs is eliminating the Peltier cooler in the pump source, enabling smaller laser-diode packages and significantly less power consumption.
Undersea systems require pumps with uncooled operation at lower temperature ranges with emphasis on ultrahigh reliability, whereas metro systems need uncooled operation over extreme temperature and current ranges. The design targets for metro pump lasers evoke "déjà vu" for designers because they have dealt with the same issues in 980-nm pump lasers for undersea applications. Indeed, the development of EDFA pump lasers suitable for the undersea market is the ideal precursor to the application of these lasers in the metro market. Three key properties determine the useful operating regime: thermal management, reliability, and wavelength stabilization.
Without a thermoelectric cooler, high efficiency at elevated temperatures is the key to keeping the power consumption low while still providing the optical amplifier with the necessary pump power. Semiconductor pump lasers at 980 nm are typically based on aluminum gallium arsenide/indium gallium arsenide (AlGaAs/InGaAs) material systems, which can display high temperature insensitivity with properly designed ridge-waveguide structures.
Some pump lasers can obtain a high characteristic temperature, To of 130 to 150 K.1 In contrast, 1480-nm pump lasers based on indium gallium arsenide phosphide (InGaAsP) are more likely to have a To of 50 to 100 K, and suffer from a large increase in laser threshold current and decrease of efficiency with increasing temperature.2 Hence, uncooled operation of 1480-nm pump lasers at high temperatures is inherently inefficient.
The ridge waveguide design used in some uncooled 980-nm pump chips greatly improves high-temperature properties (see Fig. 1). This design, along with 10 years of improvements in design and materials, results in improved slope efficiency, and wavelength-stabilized fiber powers of more than 100 mW even at 120°C. The wallplug power efficiency for 100-mW ex-fiber power amounts to 25% at 70°C and 33% at 0°C.
The excellent thermal behavior of 980-nm pump lasers is of great advantage in uncooled submarine systems for the lower temperature range as well as in uncooled metro systems with typically higher-temperature environments.
The thermal properties of 980-nm pump lasers have continually improved with the need for low power consumption and higher output powers. The need for stringent reliability testing in undersea applications has also driven development.3 Lasers with such reliability requirements were qualified in mid-2001 with less than 100 FIT for more than 400 mW optical power at 25°C based on more than 2.5 million device hours. Sophisticated qualification tests covering a wide range of conditions are performed to model the reliability and evaluate the failure rate at every operation level. Stress is applied with case temperatures exceeding 120°C and currents of more than 1 A to ensure proper scaling from accelerated conditions to the device lifetime.
The failure rate versus ex-fiber output power can be calculated using this reliability model and assuming conservative laser-package design rules with moderate coupling efficiency (see Fig. 2). Currently, an uncooled wavelength-stabilized 500-FIT device delivers around 120 mW of ex-fiber power. The next generation will provide 200 mW. The reliability of a module in operation 95% of the time at 45°C and 5% of the time at 75°C will not differ significantly from the reliability of continuous operation at 45°C.
To date, more than 400,000 of our terrestrial pump lasers have been deployed, of which field data are available for 202,000 lasers. The total accumulated device time is greater than 1370 million hours and 99 devices have failed since 1995. These data enable comparison of the reliability data estimated from accelerated life tests with real-world reliability. The field failure rate is 74 FIT with a 60% confidence level. The failure rate is still decreasing and does not saturate as field hours accumulate (see Fig. 3).
The pump modules' field reliability compares favorably to the acceleration model and the sub-100 FIT derived for undersea application. Thus, the field data provide confidence that uncooled 980-nm pump lasers easily meet the reliability demands for uncooled metro modules.
In addition to thermal properties and reliability, wavelength stabilization over the temperature operating range is a critical design factor in uncooled 980-nm pump laser modules. Wavelength stabilization pins the pump power to the absorption bandwidth of EDFAs for efficient pumping, independent of the operating temperature and condition. Further, it enables EDFA designers to combine several pump sources within a single amplifier, resulting in more pump power.
An external fiber Bragg grating (FBG) locks the wavelength to achieve wavelength stabilization. Wavelength locking occurs over a range of drive current and temperature at the fixed FBG wavelength. For temperatures exceeding 135 K, the natural wavelength shift of a laser diode is more than 40 nm. An adequate FBG design can reduce this shift to 1 nm (see Fig. 4).
Uncooled modules at even higher powers of around 200 mW would allow system manufacturers to use them for long-haul systems, instead of more costly and more power-hungry cooled devices. The next-generation chip will support 200-mW uncooled modules with even increased reliability (see Figs. 2 and 5). Rollover power greater than 1 W and outstanding power-conversion efficiency even at high temperatures are the goals for satisfying the metro and submarine markets.
- B. Schmidt et al., Proc. OFC 2001, paper WC1-1-3.
- E. Kapon, Semiconductor I, Academic Press (1999).
- H. U. Pfeiffer et al., Proc. OFC 2002, paper ThN4.
Isabella D. Jung is product line manager and Stefan Mohrdiek is manager of advanced applications at Nortel Networks Optical Components GmbH, Binzstrasse 17, CH-8045 Zürich, Switzerland. Isabella Jung can be reached at email@example.com.