Mohammed Islam and Mark Nietubyc
Optical amplifiers based on lumped Raman technology have many of the advantages of EDFAs and thulium-doped fiber amplifiers, while extending the range of amplification into the shorter-wavelength S band.
Much of the current discussion in communications revolves around network capacity. The capacity of today's optical infrastructure—which is a function of the makeup (TDM rate and channel count) of transport wavelengths within the C band (1525 to 1565 nm)—will not suffice for the long term. Additional capacity can come from the L (1570 to 1620 nm) or S (1480 to 1520 nm) bands; the question is not which band will be used first, but when the capacity will be made available.
The emergence of the erbium-doped fiber amplifier (EDFA) relaxed the dependence of the optical domain upon the electrical domain for efficient and economical long-haul transport of DWDM signals. Optically multiplexed signals that once required expensive demultiplexing, electrical regeneration, and remultiplexing after every fiber span could be entirely optically amplified in its entirety. The all-optical network was made possible by the advent of EDFAs. However, EDFAs fall short on delivering benefits over the entire usable bandwidth available in today's optical fiber infrastructure. Specifically, EDFAs do not allow the exploitation of the capacity already existing at other, lower wavelengths, such as those contained in the S band (see Fig. 1).
AMPLIFYING IN THE S BAND
Three technologies have been extensively studied in hopes of bringing the same cost-effective means of optical amplification to the S band that EDFAs provide for the C band. Semiconductor optical amplifiers (SOAs), thulium-doped fiber amplifiers (TDFAs) using either fluoride or multicomponent silicate as the dopant host fiber, and lumped (discrete) Raman amplifiers (LRAs) have all been proposed as ways to open up transmission through optical amplification for wavelengths shorter than the traditional 1530 to 1560 nm range of EDFAs. These three methodologies differ in their capabilities.
Determining which of the three types of optical amplifiers holds the most promise for opening up the S band requires a close look at their capabilities, including their principles of operation, implementation challenges, and possible applications. It should be noted that, while all of these technologies have benefits to offer, only one provides for cost-effective long-haul transmission over additional wavelength bands.
SEMICONDUCTOR OPTICAL AMPLIFIERS
The principles of operation of the SOA differ from the other two amplifier types in that an optical pumping source is not required. Optical amplification can be achieved in an SOA through the application of an electrical field to the input optical-transmission signal traveling through an SOA waveguide structure. In the electric field within the SOA, an absorption or stimulated emission of a photon can occur. The absorption of a photon results in the generation of an electron-hole pair. Stimulated emission is brought about by a photon that initiates the recombination of an electron-hole pair. To obtain amplification of an optical signal, the stimulated emission must be greater than the absorption of photons.
The characteristics of the gain, the center wavelength, and the gain bandwidth offered by an SOA are dependent upon the type of semiconductor material used and the structure of the device. For optical transmission applications, attention has focused on SOAs fabricated from indium phosphide, indium gallium arsenide, and indium gallium arsenide phosphide. The 3-dB gain bandwidth of a single device is typically 20 to 30 nm, although operation has been reported up to 40 to 60 nm away from the gain peak wavelength. With current technology, maximum gains made available through the use of SOAs have been limited to 15 dB. The physical properties of the semiconductor material permit the construction of SOAs that can operate across a wide range of wavelengths from 1300 up to 1600 nm.
The very aspects that make SOAs so attractive as optical amplifiers also pose the greatest challenges to their eventual use in long-haul DWDM applications. This paradox arises because the nonlinear nature of the amplification mechanisms involved in an SOA are applied to a linear application of cascading optical amplifiers along a number of consecutive fiber spans. Challenges exist because SOAs suffer from low-gain and high-noise figures, as well as high-bit-rate and multiplex signal-handling sensitivities.
The rapid response time of the SOA brings about a number of difficulties when attempts are made to achieve amplification of high-bit-rate DWDM signals within a fiber. With recovery times on the same order as some of today's TDM bit rates, current SOAs only effectively handle signals below 10 Gbit/s before saturation, although 20-Gbit/s operation using lower powers to remain below saturation has been reported.1 However, these lower power levels are inadequate for typical standard span lengths of 80 to 100 km.
With the fast gain response affecting the entire length of the signal within the waveguide of an SOA, changes in the magnitudes of the electric field or input signals impart fluctuations to the resulting output signals. Fluctuations lead to interchannel and intersymbol interference, along with high noise figures around 8 to 10 dB. This high noise figure is further exacerbated by the high coupling losses that are incurred when connecting an SOA to transmission fiber.
A challenge also exists in bringing together the different modes of propagation brought about by the difference in the geometries between the fibers and SOA waveguide (see Fig. 2). Problems with mechanical alignment can lead to coupling losses for SOAs on the order of 3 dB for each end (input and output fiber), further affecting the already-high noise figure.
FIGURE 3. Thulium-doped fiber amplifier uses an upconversion pump with a counterpropagating pump unit to obtain a gain respons near 1460 nm. An upconversion pump is used to overpopulate one quasistable level (E2), then another pump is used to populate the other quasistable level (E3) in order to release photons.
The nonlinear operation of SOAs is becoming better understood through more research. As such, the SOA will likely find applications in the world of optical networks—not as a linear amplifier for long-haul DWDM, but as a device that exploits the very nature of its operation. Current research has focused on the application of SOAs as modulators, optical switches, and wavelength converters. The size of the device and its high level of integration make it attractive for applications that are highly cost-sensitive. Thus, much recent work has concentrated on applications within the metropolitan optical network that require a lower level of amplification to overcome losses from multiple encounters with optical add/drop nodes—and, in the future, optical switches.
THULIUM-DOPED FIBER AMPLIFIERS
A TDFA operates in a manner similar to that of the EDFA: optical amplification of a signal is accomplished through energy conversion from pump signals through a length of specially doped fiber. Differences between the two types of devices lie in the type of gain fiber used, dopants employed, and pump configurations. An energy-level schematic depicts the optical amplification process in a TDFA, utilizing either a fluoride or a multicomponent silicate-fiber gain medium (see Fig. 3). Two pumps of either different or identical wavelengths achieve optical amplification.2, 3
Using this pumping scheme, together with either a fluoride or fluorozirconate or multicomponent silicate such as antimony silicate, a fiber doped with high levels of thulium shows a gain response centered around 1460 nm. Gain bandwidths of +20 dB have approached 35 nm in width. Maximum gains of 31 dB have been reported. Unlike the SOA, operation is limited to a range of approximately 65 nm.
The fiber hosting the dopants poses the main challenge to the wide acceptance and commercial deployment of TDFAs. These fibers are more brittle than silica-based fibers; thus, greater care must be exercised throughout the manufacturing, deployment, and maintenance processes. Keeping the same level of product reliability as exists with current silica-based technologies—for example, EDFAs—will pose a significant challenge, as currently accepted practices must change.
Being of different material composition than fiber deployed into the network's transport infrastructure, TDFAs that must be connected into a network pose another challenge. Because fusion splicing is not an available option, less-reliable and lossier mechanical splicing or epoxy pigtailing must be used. This compromise results in a loss of more than 0.3 dB per mechanical splice versus less than 0.05 dB per fusion splice. The exotic composition of these fibers will also reveal itself in higher component costs. Unless some significant material science or processing breakthroughs occur, reliability and cost challenges will inhibit the widespread deployment of TDFAs.
LUMPED RAMAN AMPLIFIERS
Raman gain in optical fibers is the transfer of power from one optical beam to another through the energy of an optical phonon. An optical phonon arises as a beam of light couples with the vibrational modes of the medium (see Fig. 4). In this instance, the optical fiber is the medium. The transfer of power is downshifted in frequency and occurs rapidly. Since this phenomenon is nonresonant, gain is made available at any wavelength.
Pumped with a laser source of 1500 nm, the Raman gain spectrum of silica fiber is over 40 THz wide, with a dominant peak occurring near the +13.2 THz (or, approximately 100 nm) offset.4 The gain presented by the Raman effect in fused silica glass is polarization-dependent. That is, the gain of the spectrum shown only occurs if both the signal and pump beams are of the same polarization. The Raman effect presents much lower gain if the polarizations of the two light beams are orthogonal to each other. To ensure that gain is maintained and to avoid polarization-dependent loss brought about by the different polarizations of the transmission wavelengths as they propagate through nonpolarization-maintaining fiber, polarization diverse pump sources are used.
The Raman effect occurs in every optical fiber and, as such, is available as an enhancement on every new or existing link of an optical network. With the gain essentially defined by the pump wavelength, applications over the entire transparent band from 300 to 2000 nm are possible. Broadband amplifiers can be tailored to the wavelength band through the use of multiple pumps and gain flattening (see Fig. 5).
Multiwavelength polarization-diverse pumping units combined with multistage configurations allow gains approaching 20 dB to be achieved. Gain bandwidths can be tailored through the use of additional wavelengths and are not limited by the material composition and doping. Using silica fiber as the gain medium, Raman amplification does not suffer from the same implementation (deployment) challenges as SOAs and TDFAs. Amplifiers can be fusion-spliced to the same type of fibers now used within the transportation infrastructure.
Although comparable in performance with TDFAs, the biggest challenge that Raman amplifiers have had to overcome is their relatively poor efficiency when measured against EDFAs. When EDFA performance was compared to that of some early-generation Raman amplifiers, Raman appeared to be a less-than-adequate alternative. Erbium-doped fiber amplifiers pumped with 980- or 1480-nm light can achieve fiber amplifier gains of greater than 40 or approximately 30 dB, respectively, using input pump powers of less than 10 mW. Conversely, to achieve Raman amplifier gain on par with that of EDFAs, more than 1 W of 1500-nm input pump power was required. These results occurred prior to some of the latest advances in both fiber and laser technology.
A better understanding of the Raman gain efficiency of optical fibers has led to the identification of dispersion-compensating fiber (DCF) as a leading candidate for a gain medium. Increases of more than a factor of ten in Raman gain efficiencies are being seen in commercially available DCFs. The newest high-power laser diodes now have greater than 1 W of output power.5 New diode-array-cladding-pumped lasers are reaching output power levels on the order of 10 W and greater. The difficulties imposed by the poor efficiency of Raman amplification are fading with these increases in fiber Raman gain efficiency and laser-pump output powers.
Noise issues within Raman amplifiers brought about by double Rayleigh scattering and the extremely fast response of the Raman effect itself also are being addressed. Double Rayleigh scattering can be controlled through the use of improved isolation between the multiple stages of an LRA. The extremely fast response of the amplification process can be averaged out through the use of counterpropagating pumping; thus, noise figures are now decreasing toward 5 dB. Packaging challenges for Raman amplifiers still exist, as the lengths required to achieve appreciable gain can easily reach into the tens of kilometers. These lengths decrease, however, as the Raman gain efficiencies of the fiber increase.
Raman amplification has enjoyed resurgence as a result of distributed Raman amplifiers in ultralong-haul DWDM. Improvements in technology have brought a discrete version close to commercial deployment, which will allow the tapping of additional capacity through an expansion into the S band and beyond. These improvements boost laser pump powers and improve Raman gain in small-core fibers, both of which increase the efficiency of Raman amplifiers.
Lumped Raman amplifier deployment should begin shortly in an expansion of capacity on optical fibers. A favorable direction for expansion on standard single-mode fiber is towards the shorter wavelengths (1480 to 1530 nm) of the S band. Higher-power laser pumps in concert with higher Raman gain efficiencies now obtainable fibers have led to enhancements in amplifier performance. Although the great lengths of gain fiber required to achieve high Raman gain pose a packaging challenge, using DCF as the gain fiber permits the combination of amplifier with a dispersion compensation unit into a single device. Raman amplifiers are the only silica-based technology that allows for this expansion while providing performance and reliability benchmarks on par with those EDFAs.
- L. H. Spiekman et al., ECOC 2000, Munich, Germany (September 2000).
- B. N. Samson et al., OAA 2000, PD-6 (July 2000).
- S. Aozasa et al., IEEE Phot. Tech. Lett.12, 10 (October 2000).
- R. H. Stolen et al., Jour. Opt. Soc. America 1, 4 (August 1984).
- A. Mathur et al., OFC 2000, PD-6 (March 2000).
Mohammed Islam is a founder and chief technical officer and Mark Nietubyc is technical marketing manager of Xtera Communications, 500 W. Bethany Road, Allen, TX 75013. Both authors can be contacted at 972-649-5000.