New pump lasers power Raman amplifiers

John Connolly

The primary source of pump power in the world of erbium-doped fiber amplifiers (EDFAs) has been the 980-nm laser. Historically, this pump source provided higher power, lower noise, and improved optical-to-electrical efficiency compared to its counterpart at 1480 nm. Recently, however, significantly higher optical-power levels have been achieved at 1400 to 1480 nm.1 Higher power combined with lower manufacturing costs has resulted in a major increase in the demand for 1480-nm pump lasers in EDFAs. These pump benefits are also being exploited by Raman amplifier designers.

Raman gain, on the other hand, is available over the entire fiber transmission band and is limited only by the availability of optical pump sources at the required wavelengths. The Raman gain effect occurs in all optical fibers, which permits any optical network, including both existing and new links, to be considered for broadband Raman systems. This advantage will play a pivotal role in ultralong-haul, long-haul, and metro system applications. Thus, very broadband gain amplifier systems are possible through the use of multiple Raman pump sources and gain flatteners.2

RAMAN AMPLIFICATION
The Raman gain process in optical fiber is the transfer of power from one optical beam to another through an exchange of energy. Unlike EDFA amplifier systems, the energy transfer occurs along the entire optical link in the direction opposite to the traveling optical signal.

Unfortunately, Raman amplifiers also have some limitations associated with their use in the optical network. The Raman gain effect in silica-based fibers is polarization dependent or, in other words, the gain of the wavelength spectrum is only available if both the pump and signal beams have the same polarization direction. Further, the Raman gain effect provides significantly lower gain if the polarization of the two light beams are orthogonal to one another.

In the optical network, where the gain level must be maintained, it is critical to avoid polarization-dependent loss brought about by different polarization signals as they propagate through the conventional single-mode fiber. To compensate for this loss, dual polarization-dependent pump sources are commonly used in most amplifier systems.

Even though the advantages of Raman amplification, such as low noise and wide spectral range, have been known for more than a decade, deployment of Raman amplifiers was limited by the availability of efficient and compact pump modules in the C-, S-, and L-bands.

HIGH-POWER PUMP MODULES
Only recently, high power (>200 mW) pump modules based on indium gallium arsenide/indium phosphide (InGaAsP/InP) semiconductor diode lasers have become available. Increasing output levels are needed to provide the necessary gain margin for DWDM applications as well as provide additional flexibility for amplifier designers. As additional frequency-dependent signal wavelengths are added to the network, increasing pump energy is required to provide the necessary gain for increased bandwidth. Therefore, higher energy or power is a desirable feature for the discrete pump source.

When the optical power cannot be achieved by use of a single-pump module, additional pump modules are combined to increase the total available power. These additional modules add to the Raman amplifier system complexity and the subsequent cost, so there is a very strong motivation to increase the output power from a single pump source. Higher-power pump sources also provide other desirable characteristics associated with pump modules, such as improved linearity, stability, and reliability.

The majority of InGaAsP/InP laser devices are grown by low-pressure metal organic chemical vapor deposition and contain a strained InGaAsP quantum well active region. Lateral-mode confinement is normally provided by a buried heterostructure or a ridge waveguide structure. Single-mode operation for the laser is maintained by accurate control of the lateral-index step and active waveguide width. Normally, the emitting and rear facets of the devices are coated with low- and high-reflective coatings, respectively. The device beam divergence is optimized for both high output power and coupling to a single-mode fiber (see Fig. 1).

Output power for the device is measured on a curve showing chip output power at about 1 W while the fiber launch power is greater than 700 mW at an operating current of about 3.0 to 3.5 A (see Fig. 2). The emission wavelength for this device was designed for 1420 nm to support optical pumping in the C-band. The threshold current is approximately 75 mA and the maximum operating voltage is about 2.75 V. The peak efficiency is greater than 0.40 A/W.

Higher-power pump modules are the key to reducing cost in Raman amplifier systems since individual pumps provide gain over a narrower spectrum than EDFA systems. This feature may present an opportunity in the metro DWDM market as a low-cost amplifier for localized optical loss. Coupled with polarization dependency, this Raman feature leads to an increase in the number of optical pumps required for amplification. Thus, for Raman amplifiers to be competitive, pump-module power must be increased, while at the same time the module cost ($/mW) must be reduced.

The rapid development of higher-power pump modules will accelerate the cost-reduction curve for the modules, while at the same time increase the deployment of Raman-based amplifiers. In a recent technology report, RHK (South San Francisco, CA) predicts that the combined 1480- and 14xx-nm pump module production will surpass 980-nm module production in 2001.3 At the Optical Fiber Communications Conference 2001, several of the pump vendors announced new pump modules with output powers of 400 and 500 mW for release later this year.

HIGH POWER YIELDS REWARDS
In addition to output power, other key parameters for module operation include linearity, slope efficiency, spectral range, spectral width, and tracking ratio.

One of the benefits of higher laser-chip power is improved power linearity in modules. In the 14xx module, thermal effects significantly limit the fiber power linearity leading to a condition called thermal rollover. Thermal rollover occurs when increases in operating current no longer produce additional output power. This rollover condition occurs independent of laser mode and waveguide losses and is strictly associated with the overall thermal properties of the laser chip and package. Thus, thermal rollover can occur while the single-mode behavior of the chip remains unchanged.

Although thermal saturation does not alter chip-mode stability, it does significantly impact power linearity; in other words, slope efficiency is reduced with increasing output power. Thermal rollover effects observed 50 to 100 mW above the recommended module performance level may lead to reduced performance between beginning of life (BOL) and end of life (EOL) module characteristics. The module performance change between BOL and EOL is about 10% to 20%.

Since the available gain from a Raman amplifier system is narrower than the gain from an EDFA amplifier, optical-pump sources must traverse the entire communication spectrum. Further, the transfer of power by the optical-phonon process results in approximately a 100-nm offset from the pump frequency. Thus, the frequency range of pump modules for C- and L-band communication systems are approximately 1400 to 1450 nm and 1450 to 1500 nm, respectively. For each pump range, between four and eight pump modules are used to compensate for the polarization dependency, narrow spectral gain, and lower overall system efficiency.

To meet the current needs of Raman amplifier users of C- and L-band systems, pump manufacturers must be able to provide very high-power pump modules from 1400 to 1500 nm (see Fig. 3). The spectral width for all devices showed similar behavior of 8 to 12 nm at the full-width half-maximum (FWHM) value.

The stability of the pump wavelength is a key parameter for Raman amplifiers. In typical pump modules, an external fiber Bragg grating (FBG) and closed-loop control of the laser chip temperature are used to maintain spectral stability. The FBG can be thought of as a spectral filter at the output of the module. The overall spectral properties of FBGs are less sensitive to temperature, thereby providing improved long-term stability as well as improved aging effects. The reflectivity and bandwidth of the grating is normally selected to match wavelength, power level, stability, and spectral width for the module.

The narrow spectral width afforded by the FBG supports the need for wavelength combining in the Raman amplifier so that uniform signal gain can be provided across the entire communication band (see Fig. 4). The module emission wavelength stays locked to the FBG frequency over its entire power range. The FWHM of the spectral output is less than 1 nm as compared with 8 to 12 nm for modules without FBG feedback.

Another key parameter in Raman pump modules is the tracking between the internal pin detector and the fiber power. This relationship is of particular importance as modules move to higher powers. Most amplifier designers would prefer tracking accuracies of about 1%. In conventional pump modules, a two-lens design is employed where the first lens is used to collect the light from the laser chip while the second lens is used to focus the light into the single-mode optical fiber.

Alignment tolerances are normally below a micrometer or less. Accurate tracking is difficult to achieve when using a module design that incorporates a non-temperature-controlled second lens position. The most popular location for this lens is the front wall of the package where it can be subject to the overall module environment. Recent approaches to module design utilize a common platform that is temperature controlled. All optical components are positioned on the platform to prevent misalignment between optical elements that can lead to increased tracking inaccuracies.

In the future, higher-power pump modules will be a welcome relief to amplifier designers since the additional power can be used to relax system requirements on other system elements that introduce optical loss such as pump combiners and gain-flattening filters.

John Connolly is a founder and vice president of engineering at Princeton Lightwave, 2601 Route 130 S, Cranbury, NJ 08512. He can be contacted at 609-925-8100 or through www.princetonlightwave.com

ACKNOWLEDGMENT
The author wishes to thank his colleagues, D. Garbuzov, M. Maiorov, V. Khalfin, D. Hayes, A. Komissarov, R. Menna, and P. York for their contributions to this article.

REFERENCES

1. D. Garbuzov et al., Optical Fiber Conference, PD-18 (March 2001).
2. M. Islam et al., WDM Solutions, 53 (March 2001).
3. RHK, Technology Brief (November 2000).