In optical amplification, most next-generation erbium-doped fiber amplifiers (EDFAs) no longer boast higher bit rates and more wavelengths. Rather, new EDFAs are cheaper, smaller, and consume less power without sacrificing the performance or reliability of their predecessors.
Advances in 980-nm pump lasers such as fiber-Bragg-grating (FBG) stabilization enable such EDFAs.1 Of particular importance are ultra-high-power 500-mW 980-nm pumps in cooled-butterfly packages, and 200-mW uncooled 980-nm pumps in miniDIL (dual in-line) packages. As 980-nm pump-module power has increased by approximately 50% every 18 months, ultra-high-power 980-nm pumps are increasingly cost-effective replacements for multiple older-generation output-stage 980- and 1480-nm pumps. MiniDIL uncooled 980-nm pumps lower the bar for cost, power consumption, and size. In both cases, the replacement of legacy solutions pushes the requirements of 980-nm spectral performance, which can be met with polarization-maintaining fiber (PMF).
The key to effective FBG wavelength stabilization is maintaining adequate optical feedback into the laser-diode cavity. A Fabry-Perot laser diode is a TE-polarized emitter, so only the TE-polarized component of the light reflected at the FBG influences the diode. Any birefringence in the roundtrip optical path between the laser diode and FBG reduces the proportion of reflected light providing the necessary feedback. If the birefringence produces a polarization rotation near 90° (or 270°, and so on), insufficient optical feedback decouples the laser diode from the FBG, potentially impacting EDFA noise performance. Thus, minimizing birefringence across all operating conditions is the key to improving 980-nm pump-module spectral quality.
In single-mode pigtails, deformation of the cylindrical core is the primary source of externally induced birefringence. Deformation arises from bends and twists introduced along the fiber during spooling (or "fiber lay"), and from any radial pressure exerted along the pigtail. Since birefringence cannot be eliminated, conventional 980-nm designs are guard-banded by high FBG reflectivity to maintain acceptable side-mode suppression ratio (SMSR) if only a small fraction of the feedback is TE-polarized (often because customers over-specify SMSR).
The high intrinsic birefringence of polarization-maintaining fiber makes it insensitive to small perturbations. Thus, at similar FBG strength, a PMF-pigtailed 980-nm pump module can maintain excellent SMSR across a wider dynamic power and temperature range, providing both improved manufacturing yields and wider operating ranges for cooled and uncooled pumps.
The benefit of PMF pigtails for FBG-stabilized 980-nm pumps has long been understood, but the high cost of PMF limited their use to undersea applications. Lately, PMF pricing has fallen to a level viable for broad use. Because polarization need not be maintained beyond the FBG for conventional 980-nm applications, standard splicing equipment may be retained.
The drive to smaller, less-power-hungry EDFAs will spur rapid adoption of uncooled pumps. Eliminating the bulky thermo-electric cooler (TEC) can reduce 980-nm pump-module power consumption by 75% while permitting smaller, cheaper miniDil packaging. MiniDils are well suited to emerging, cost-sensitive narrowband EDFA architectures that do not require the highest power pumps. Under a multisource agreement, the miniDil platform is successfully standardized.
Recent chip advances allow the extension of uncooled 980-nm pump-module operation to greater than 200 mW with excellent reliability. For high-power miniDils to gain acceptance in wideband EDFA applications, PMF pigtails are needed to realize spectral quality across comparably wide temperature and power dynamic ranges as TEC-cooled pumps (see Fig. 1). Data from uncooled PMF-pigtailed miniDIL technology, representing mean SMSR averaged from 72 discrete uncooled pump modules, show the capability of PMF-pigtailed modules to maintain excellent SMSR over 20- to 240-mW fiber-coupled power across the full -5°C to 75°C standard telecom temperature range.
Uncooled 980-nm pump lasers have an additional test burden. Because external temperature variations affect the laser-diode band gap, spectral quality must be tested rigorously across the full rated temperature and power ranges, whereas TEC-cooled 980-nm pumps are only spot tested. Because PMF-pigtailed 980-nm performance is fiber-lay independent, EDFA assemblers can be confident of factory-measured performance. On the other hand, uncooled pump lasers without PMF should be conservatively guard-banded to ensure the required spectral performance is satisfied regardless of fiber lay.
The chief reliability challenge for uncooled 980-nm pumps is minimizing failure in the upper end of the -5°C to 75°C temperature range. Understanding both laser diode and miniDIL package reliability requires extensive multicell testing of the product across a range of temperature and operating currents to develop accurate activation-energies and power-acceleration models.
Commercially available 980-nm pump laser chips are governed by reliability models incorporating an Arrhenius factor that depends on the diode junction temperature. As a rule of thumb, the chip failure rate triples with each 20°C increase in operating temperature. A viable uncooled 980-nm pump module must demonstrate a vanishingly small chip failure rate at operating power and 25°C. A detailed understanding of the chip's reliability performance scaled by both temperature and operating current is required. Multicell testing has proven that chips can operate uncooled at 50% of the rated operating power of the same chip in a conventional TEC-cooled module, or greater than 200-mW of fiber-coupled power.
Optical alignment technologies developed for a 25°C TEC-cooled environment may be unproven at higher temperatures. To accurately model reliability under typical operating conditions in the 40°C to 75°C ambient temperature range, millions of device-test hours are being generated at temperatures ranging from 25°C to 85°C. The resulting analytical model will enable EDFA manufacturers to accurately specify beginning-of-life output power and EDFA FIT rates for the ambient temperatures specific to their application. Long the norm for undersea 980-nm qualification, multicell test rigor is required for uncooled terrestrial pumps, which cannot rely on a long deployment history to mitigate the need for up-front analytical understanding of product reliability.
Over the past three years, the increase of fiber-coupled power available in a single butterfly package at 980 nm has far outstripped that at 1480 nm, enabling 980-nm pumps to achieve $/dBm cost parity. At similar cost, 980-nm pumps are vastly preferred because of their reduced power consumption and lower noise figure (even in the output stage). Comparison of optical output power (L-I) and power consumption curves vs. drive current for a 980-nm pump module under worst-case and typical (75°C and 40°C) operation demonstrates that total power consumption is approximately half that of a 1480-nm pump module (see Fig. 2).
For full adoption, ultra-high-power 980-nm pump modules must match the wide dynamic range of Fabry-Perot 1480-nm lasers. In particular, output-stage pumps are often required to operate above threshold current when only slight amplification is required. While traditional 980-nm technology is stretched to a 15-dB-power dynamic range (12 to 350 mW), ultra-high-power 980-nm pumps with PMF-pigtails are capable of more than 20 dB.
Key spectral data from 30-odd discrete PMF-pigtailed 980-nm pump modules show that low-frequency noise (chiefly caused by shifting longitudinal laser modes) and the spectral purity of the fiber-coupled output power impacts EDFA performance (see table). The data indicates an optimized PMF-pigtailed FBG maintains low-noise spectrally pure operation across 20 dB of power dynamic range all the way down to 5 mW of output power.
The availability of PMF-enabled 980-nm pump modules with ever-higher output power and versatility will impact future EDFAs, such as, for example, three-stage, dispersion-compensated, gain-flattened EDFA architecture (see Fig. 3). Previously, such designs incorporated cooled 980-nm pumps in the first stage, 980- or 1480-nm pumps in the second stage, and exclusively 1480 nm in the output stage. We expect the transition to all-980-nm EDFA architectures to be apparent in 2003.
In 2003, EDFA development will focus on low-cost miniDIL packages in the preamplification stages to replace their cooled cousins, as well as harnessing 980-nm pumps for output-stage pumping through polarization and wavelength combining. As a result, EDFAs produced in 2004 will have the lowest possible premplification-stage costs, and will rely on multiple polarization-multiplexed copropagating 980-nm pumps for low-noise output-stage power. The 2004 generation will be the first to fully capitalize on the ability of 980-nm pump technology to provide the lowest possible power consumption.
During 2004, new chip technology will increase the rated power of both cooled and uncooled 980-nm pumps by another 50%. In 2005, EDFAs will incorporate 300-mW miniDILs and 750-mW cooled 980-nm pumps. Fewer pumps will provide the same dBm output power. Also, 300-mW miniDIL technology will become attractive for low-power output-stage applications, meaning 2005 could witness the first fully uncooled wideband EDFA deployments.
Jay Skidmore is a development manager for coolerless pumps and reliability, Ed Wolak is development manager for high-power pumps, and Toby Strite is business development manager at JDS Uniphase. Toby Strite can be reached at firstname.lastname@example.org.
The authors thank their colleagues Don Hargreaves, Nina Morozova, Michael Staskus, Dhrupad Trivedi, Ken Wang, Vince Wong, and Lei Xu for their contributions to this article.
- Q. Wang et al., WDM Solutions (July 2001).