Raman Amplifiers for Telecommunications

548 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 8, NO. 3, MAY/JUNE 2002

Raman Amplifiers for TelecommunicationsMohammed N. Islam

Invited Paper

Abstract—Raman amplifiers are being deployed in almostevery new long-haul and ultralong-haul fiber-optic transmissionsystems, making them one of the first widely commercializednonlinear optical devices in telecommunications. This paperreviews some of the technical reasons behind the wide-spreadacceptance of Raman technology. Distributed Raman amplifiersimprove the noise figure and reduce the nonlinear penalty of fibersystems, allowing for longer amplifier spans, higher bit rates,closer channel spacing, and operation near the zero-dispersionwavelength. Lumped or discrete Raman amplifiers are primarilyused to increase the capacity of fiber-optic networks, opening upnew wavelength windows for wavelength-division multiplexingsuch as the 1300 nm, 1400 nm, or short-wavelength-band. Asan example, using a cascade of -band lumped amplifiers, a20-channel, OC-192 system is shown that propagates over 867 kmof standard, single-mode fiber. Raman amplifiers provide a simplesingle platform for long-haul and ultralong-haul amplifier needsand, therefore, should see a wide range of deployment in the nextfew years.

Index Terms—Optical amplifiers, optical fiber amplifiers, opticalfiber communications, Raman lasers, nonlinear optics.

I. INTRODUCTION

I N THE EARLY 1970s, Stolen and Ippen [1] demonstratedRaman amplification in optical fibers. However, throughout

the 1970s and the first half of the 1980s, Raman amplifiersremained primarily laboratory curiosities. In the mid-1980s,many research papers elucidated the promise of Raman ampli-fiers, but much of that work was overtaken by erbium-dopedfiber amplifiers (EDFAs) by the late 1980s [2]. However, in themid- to late 1990s, there was a resurged interest in Raman am-plification. By the early part of 2000s, almost every long-haul(typically defined 300 to 800 km) or ultralong-haul (typ-ically defined above 800 km) fiber-optic transmission systemuses Raman amplification. There are some fundamental andtechnological reasons for the interest in Raman amplifiers thatthis paper will explore.

The first section of this paper reviews the physics of Ramanamplification in optical fibers. The second section of the paperfocuses on a distributed Raman amplifiers (DRAs). We showthat a DRA can improve the signal to noise ratio and nonlinearpenalty of amplifiers. Then, the third section discusses discreteor lumped Raman amplifiers. We show that discrete Ramanamplifiers are often used to open up new wavelength bands in

Manuscript received January 31, 2002; revised March 29, 2002.The author is with Xtera Communications, Inc., Allen, TX 75013 USA, on

leave-of-absence from the Department of Electrical Engineering and ComputerScience, University of Michigan, Ann Arbor, MI 48109 USA.

Publisher Item Identifier S 1077-260X(02)05901-4.

fiber systems. Finally, we described all Raman transmissionsystem experiments in the short wavelength-band to illustratesome of the performance achieved in lumped or discrete Ramanamplifiers.

II. RAMAN AMPLIFICATION IN OPTICAL FIBERS

Raman gain arises from the transfer of power from one op-tical beam to another that is downshifted in frequency by the en-ergy of an optical phonon . The Raman gain spectrum in fusedsilica fibers is illustrated in Fig. 1(a) [1]. The gain bandwidth isover 40 THz wide, with the dominant peak near 13.2 THz. Thegain band shifts with the pump spectrum, and the peak value ofthe gain coefficient is inversely proportional to the pump wave-length. In the telecommunications bands are around 1500 nm13.2 THz corresponds to approximately 100 nm.

Fig. 1(b) illustrates the polarization dependence of Ramangain [18]. The copolarized gain is almost an order of magni-tude larger than the orthogonal polarization gain near the peakof the Raman curve. Nonetheless, a polarization-independentRaman amplifier can be made by using polarization diversitypumping to avoid polarization dependent loss. Furthermore, themixture of modes in a nonpolarization-maintaining fiber helpsto scramble the polarization dependence.

Raman amplifiers have some fundamental advantages. First,Raman gain exists in every fiber, which provides a cost-effectivemeans of upgrading from the terminal ends. Second, the gain isnonresonant, which means that gain is available over the entiretransparency region of the fiber ranging from approximately 0.3to 2 m. A third advantage of Raman amplifiers is that the gainspectrum can be tailored by adjusting the pump wavelengths.For instance, multiple pump lines can be used to increase the op-tical bandwidth, and the pump distribution determines the gainflatness. Another advantage of Raman amplification is that it isa relatively broad-band amplifier with a bandwidth5 THz, andthe gain is reasonably flat over a wide wavelength range.

However, a number of challenges for Raman amplifiers pre-vented their earlier adoption. First, compared to the EDFAs,Raman amplifiers have relatively poor pumping efficiency atlower signal powers. Although a disadvantage, this lack of pumpefficiency also makes gain clamping easier in Raman amplifiers.Second, Raman amplifiers require a longer gain fiber. However,this disadvantage can be mitigated by combining gain and thedispersion compensation in a single fiber. A third disadvantageof Raman amplifiers is a fast response time, which gives rise tonew sources of noise, as further discussed below. Finally, thereare concerns of nonlinear penalty in the amplifier for the WDMsignal channels.

1077-260X/02$17.00 © 2002 IEEE

ISLAM: RAMAN AMPLIFIERS FOR TELECOMMUNICATIONS 549

(a)

(b)

Fig. 1. (a) Raman gain curve in fused silica fiber for copolarized pump andsignal beams. Inset shows an energy level diagram representative of the Ramanprocess, which takes a higher energy pump photon and splits it into a lowerenergy signal photon and a phonon. (b) Normalized Raman gain coefficient forcopolarized and orthogonally polarized pump and signal beams.

Despite these challenges, there has been a revived interest inRaman amplifiers. Several technological advances over the lastfew years have made Raman amplifiers feasible and practical. Itis interesting to note that the physics of Raman has not changed,but rather it is new technologies that have enabled Raman am-plifiers to come of age. The first key development has been theavailability of higher Raman gain fibers with relatively low loss.As an example, there is more than a tenfold increase in gain effi-ciency in commercial dispersion compensating fiber comparedto standard single-mode fiber (SMF). Moreover, new Ramangain fibers continue to be introduced commercially with dif-ferent dispersion profiles and dispersion slopes.

A second key development for Raman amplifiers has been theavailability of high pump power laser diodes or cladding-pumpfiber lasers. Cladding-pump fiber lasers are available withoutput powers 10 W in a SMF. Commercial laser diodes areavailable with more than 300-mW output powers, and they willsoon to be upgraded to 400 mW [19]. In addition, research onhigh-power laser diodes shows the availability of output powersin excess of one watt in SMFs in the near future [3].

A third technological development important for Ramanamplifiers has been the availability of all fiber components toreplace bulk optics. For example, gratings, specialty couplers,

wavelength-division multiplexers (WDMs), etc., are now avail-able for splicing easily into all fiber configurations. In addition,fiber-pigtailed bulk-optic couplers are also now commerciallyavailable with full high-power reliability qualification mainlybased on epoxy-free technologies. A spliced-together fiberamplifier configuration is much more resistant to environmentaldisturbances than its bulk-optical counterpart. Thus, an upto 100-fold increase in pump power combined with up to aten-fold increase in Raman gain coefficient leads to a renewedinterest in Raman amplification in an all-fiber configuration.

A. Sources of Noise in Raman Amplifiers

There are four primary sources of noise in Raman ampli-fiers. The first is double Rayleigh scattering (DRS), which cor-responds to two scattering events (one backward and the otherforward) due to the microscopic glass composition nonunifor-mity. Amplified spontaneous emission (ASE) traveling in thebackward direction will be reflected in the forward direction byDRS and experience gain due to stimulated Raman scattering.This in addition to ASE experiencing multiple reflections, willdegrade the signal-to-noise ratio (SNR). Furthermore, multipathinterference of the signal from DRS will also lower the SNR.DRS is proportional to the length of the fiber and the gain in thefiber, so it is particularly important in Raman amplifiers due tothe long length of fiber, where lengths of several kilometers aretypically required. From a practical viewpoint, DRS limits thegain per stage to approximately 10 to 15 dB. Higher gain am-plifiers can be obtained through the use of isolators between themultiple stages of amplification. For example, a 30-dB discreteRaman amplifier has been demonstrated with two stages of am-plification and a noise figure less than 5.5 dB [4].

The second source of noise arises from the short upper-statelifetime of Raman amplification, as short as 3 to 6 fs. This vir-tually instantaneous gain can lead to a coupling of pump fluctu-ations to the signal. The usual way of avoiding this deleteriouscoupling is to make the pump and signal counterpropagating,which has the effect of introducing an effective upper-state life-time equal to the transit time through the fiber. If copropagatingpumps and signals are to be used, then the pump lasers must bevery quiet. That is, they must have a very low so-called rela-tive intensity noise (RIN). For example, copropagating pumpsuse Fabry–Pérot laser diodes instead of grating-stabilized laserdiodes [5].

A third primary source of noise in Raman amplifiers is theusual ASE. As is typical for reasonable signal power levels,signal–ASE beating noise dominates over ASE–ASE beatingnoise. Fortunately, Raman amplifiers have inherently low noisefrom signal–ASE beating because a Raman system always actsas a fully inverted system. For example, the ASE power spectraldensity can be written as

and the noise figure as

where is the upper state population and is the lower statepopulation. For Raman amplifiers, the term isalways equal to one, whereas in EDFAs it is usually greater than


one [6]. In an EDFA, this term is only equal to one for an ampli-fier fully inverted through the entire length of gain fiber. On theother hand, since Raman amplifiers use long fiber lengths, thesmall fraction of passive loss of the gain fiber needs to be addedto the noise figure. Nonetheless, discrete Raman amplifiers withnoise figures as low as 4.2 dB have been reported [7].

Finally, a fourth source of noise arises from the phonon-stim-ulated optical noise created when wavelength signals being am-plified reside spectrally close to pump the wavelengths used foramplification. In other words, at room or elevated temperaturesthere is a population of thermally induced phonons in the glassfiber that can spontaneously experience gain from the pumps,thereby creating additional noise for signals close to the pumpwavelengths. It has been shown that this effect can lead to anincrease in noise figure of up to 3 dB for signals near the pumpwavelength [8], [20].

B. Raman for DWDM Long-Haul Systems

Raman amplifiers provide a simple, single platform forlong-haul and ultralong-haul amplifier needs. Raman ampli-fiers are broad-band and wavelength agnostic. For example,gain bandwidths of up 100 nm have been demonstrated [9],[10], [21]–[25], and the 100 nm can fall anywhere in thetransparency window between roughly 1300 and 1650 nmtypically used in fiber-optic communications. Moreover,recently a broad-band Raman amplifier with a bandwidthof 136 nm has been demonstrated using a pump-and-signalwavelength-interleaved scheme, where stop bands surround thepump wavelengths that are within the signal band [26]. Ramanamplifiers can be distributed, lumped or discrete, or hybrid [25],[27]–[32], [37]. Thus, either Raman amplifiers can serve onlyas low-noise preamplifiers for rare-earth-doped fiber ampli-fiers, or they can meet the full amplifier needs in “all-Raman”systems. Also, in Raman amplifiers, the amplification anddispersion compensation can be combined in the same fiberlength [20], [33]–[35], [38]. The dispersion compensating fiberends up having net gain rather than loss, which leads to a widersystem margin and the ability to insert other elements such asopticalADD–DROPmultiplexers into the system.

In the dense WDM long-haul and ultralong-haul fiber-opticsystem markets, Raman amplifiers should win out as the dom-inant form of amplifier because of their simplicity and flexi-bility. As an example, suppose that a 100 nm wideband system,which overlaps the -, -, and -bands, is to be developed.If the system were to mimic today’s systems, then band com-biners and band splitters would be used around three discreteamplifiers [Fig. 2(a)]. Also, an additional triband DRA wouldbe used for low-noise preamplification. Each of these amplifierswould require their own pump lasers, control circuitry, moni-toring system, and gain equalization. In addition, a separate dis-persion compensation fiber would be typically used in each dis-crete amplifier. Moreover, because wideband filters cannot haveabrupt spectral profiles, several nanometers of guard band wave-lengths need to be reserved around each band [Fig. 2(c)]. Also,a system margin of typically 3 dB or more has to be reserved forthe additional insertion loss from the band couplers.

In comparison, consider an “all-Raman” solution for the100-nm wideband system. The amplifier configuration wouldsimplify to something like Fig. 2(b), where a broad-band DRA

is followed by a broad-band discrete Raman amplifier. TheDRA in this case is no more complicated than above, but thediscrete amplifier is considerably simpler. Fewer pumps arerequired than for the three-amplifier case, only one monitoringsystem is required, and the single dispersion-compensatingfiber (DCF) can be combined with the gain fiber in the discreteRaman amplifier. More significantly, the band splitter andcombiner are not required, which removes the requirement forguard bands and allocation of 3 dB or more system margin forthe coupler losses. The price to be paid for the simplificationin Fig. 2(b) is that 100 nm, wide-band gain equalization anddispersion compensating fibers are required, as opposed tothe typically 35-nm-wide devices in use today. For example,this means that dispersion slope compensation, as well asdispersion magnitude compensation are required. However, thetechnologies for these wide-band devices are already becomingavailable commercially.

The biggest argument made against Raman amplifiers hasbeen the perception of their poor efficiency compared to EDFAs,the industry’s workhorse. However, as the total signal powerwithin the transmission fiber increases as the number of chan-nels and the bit rate of the channels increases, Raman ampli-fiers become increasingly attractive. In fact, the gain obtainedfrom Raman is greater at the higher input pump powers necessi-tated by the power levels of future systems [Fig. 3(a)]. Even thescheme found in most deployed EDFAs, where a 980-nm firststage pump is used in conjunction with a 1480-nm second stagepump, are beginning to exhibit the “leveling off” of their gainprofile as they reach saturation and fail to provide the same ad-ditional gain for similar increases in pump power. Early WDMsystems had less than 32 channels, or power levels under 100mW, a region where Raman is much less efficient than EDFAs.In the 1999–2001 time frame, systems ranged from 64 to asmany as 160 channels, with powers below about 200 mW; thehigh end of the range is where Raman begins to become com-petitive with EDFAs. Starting in 2002, systems will be availablewith 240 channels and more, and the signal output powers willbe in the range well above 200 mW. In this new generation ofsystems, Raman amplifiers are superior in terms of pump powerefficiency compared even to 1480-nm pumped EDFAs.

Plotting along alternate axes, Fig. 3(b) is a comparison of thepower conversion efficiency (PCE) between a 1480-nm pumpEDFA and a Raman amplifier assuming a signal input power of20 mW. The PCE is defined as 100(output signal powerinput signal power)/(pump power). Here, it can be seen that atlower power levels, EDFAs exhibit a PCE of over 30%, whereasRaman remains below 20%. However, as the power levels re-quired by network capacity increases reach cause pump powerlevels into the higher region, Raman becomes more efficient.This can be seen as Raman’s PCE crosses over that of EDFAsat around 0.45 W (or about 26 dBm), a level already surpassedin current distributed Raman deployments.

A couple of interesting points to note include comparing thehighest level PCE, where both the Raman and EDFA “leveloff.” At these points, the 60% Raman PCE versus the40%EDFA PCE results in a more (cost) efficient amplification ofthe higher capacity systems. Raman’s gain can be further en-hanced through an optimization of the gain fiber employed andthe pumping strategy.


(a)

(b)

(c)

Fig. 2. (a) A typical 3-band amplifier configuration using multiple discrete amplifiers along with a triband DRA. (b) An “all-Raman” amplifier usinga widebandDRA followed by a single, wideband lumped, or discrete Raman amplifier. (c) The systems are not “equivalent” as an “all-Raman” system offers more bandwidththrough a single band without penalties from the obligatory “guard bands.”

The details of the assumptions in Fig. 3 are as follows. Theinput is one signal at 1529.5 nm with an input signal power of20 mW. For the Raman amplifier curve, a Raman pump wave-length of 1433 nm was chosen, so that the signal falls on the gainpeak at 13.2 THz away from the pump wavelength. The fiber pa-rameters used correspond to Lucent’s WBDK dispersion com-pensating fiber. The loss of the fiber at the signal wavelength is

0.472 dB/km, the length of the gain fiber is 5.64 km, and theRaman gain coefficient is 13.5 dB/(W km). For the EDFAcurves, it is assumed that the amplifier is forward pumped byeither a 980- or 1480-nm pump. The passive loss in the EDFAis assumed to be 0.15 dB/m, and the length of the coil (20 m) isoptimized to give maximum signal output power when pumpedwith 750 mW of 1480-nm pump for 20 mW of input power.


(a)

(b)

Fig. 3. Raman amplifiers surpass the power conversion efficiency of EDFAs at the higher power levels brought about by future high capacity long reach systems.A Raman amplifier is compared with a 1480-nm pumped EDFA and a 980-nm pumped EDFA. The signal input power is assumed to be 20 mW. (a) Signal outputpower versus launched pump power. (b) Power conversion efficiency versus launched pump power.

III. D ISTRIBUTED RAMAN AMPLIFIERS

A DRA is an amplifier where the pump power extends intothe transmission line fiber. As shown in Fig. 4, the DRA utilizesthe transmission fiber in the network as the Raman gain mediumto obtain amplification. Typically, high-powered counterpropa-gating Raman pumps are deployed in conjunction with discreteamplifiers, such as EDFAs.

The power of using a DRA to improve the SNR and reducethe nonlinear penalty is illustrated in Fig. 5 [11], [21]. In thisfigure, the signal power in decibels is plotted versus distance

for a periodically amplified system. The saw tooth like figurescorrespond to lumped amplification, while the curved figurescorrespond to the use of a DRA assisting a lumped amplifier.Using a DRA reduces the overall excursion that the signal levelexperiences. At the top signal level, a DRA does not require ashigh a signal level. Consequently, nonlinear effects are reduced.At the bottom, the signal does not dip down as low when a DRAis used. Consequently, the signal to noise ratio remains higherwith the use of the DRA.

One way to think about the DRA is to consider it as can-celing out the loss for a fraction of the fiber length. Therefore,


Fig. 4. DRAs utilize the transmission fiber in the network as the Raman gainmedium to obtain amplification. Typically, high power, counterpropagatingRaman pumps are deployed in conjunction with inline discrete amplifiers suchas EDFAs.

Fig. 5. Signal power in a periodically amplified transmission system.Comparison is made between a purely lumped amplified system (“originaltransmission system”) and a system using DRA to assist in the amplification.Also shown is the pump profile for counterpropagating pump. DRA permitsthe transmission of the original signal to occur below the level of nonlineareffects since signal amplification along roughly the second half of the spankeeps the transmission above the system noise floor.

the SNR improvement can be thought of in terms of an equiva-lent amplifier spacing. Assuming that the original amplifier sitesare spaced by 80 km, using an optimally pumped DRA givesthe performance as if the discrete amplifiers were spaced be-tween 35 to 38 km. Thus, DRAs give an equivalent performanceof much closer spaced amplifiers, giving a very attractive eco-nomic incentive for their use.

There are many examples from the literature that show the ad-vantages of using DRAs in transmission systems. DRAs can beused to achieve longer amplifier spans or higher bit rates. Thesetake advantage of the improvement in SNR. Alternately, DRAscan be used to achieve closer channel spacing or operation nearthe zero dispersion wavelength. These improvements take ad-vantage of the reduction in nonlinear penalty by using DRAs.

Several experiments have shown that longer amplifiers spanscan be used with the DRA. Terahara and coworkers employeda dual-band DRA technique to the-band and -bands forthe first time to long-haul WDM transmission systems [12]. Inparticular, they transmit 1.28-Tb/s WDM signals over a 840-kmstandard SMF with a 140-km repeater spacing. In comparison,the typical amplifier spacing in a link is about 80 km. Thedual-band DRA effectively improves the OSNR by 3.7 dB inthe -band and by 4.3 dB in the-bands. Garrett,et al., alsodid a field demonstration of DRA with 3.8 dB-improvementfor 5 120-km transmission compared with just using discreteEDFAs [13]. They demonstrate a DRA in an 8 10-Gb/sWDM experiment with five 120-km spans incorporating 40 kmof installed, older vintage fibers. At the optimum pump power

level of 650 mW, the average channel-value was increasedby 3.8 dB for a 5 120-km system, and the number of 120-kmspans with an error rate of 10 was increased from three spansto seven spans.

As an example of higher bit rates achieved using DRAs,Nielsenet al., experimentally show a 3.28 Tb/s (8240-Gb/sNRZ) record aggregate capacity transmitted over 3100 km ofnonzero-dispersion shifted fiber [14]. The system incorporatesfor the first time dual - and -bands transmission and DRA inaddition to the 40-Gb/s NRZ line rate. The 3.28-Tb/s capacityis comprised of forty 100-GHz spaced WDM channels in the

-band and forty-two 100-GHz spaced WDM channels inthe -bands. Bit-error rates of less than 10 were obtainedwithout forward-error-correction for all of the channels.

Several papers have also demonstrated the reduction innonlinear penalty because of the lower signal levels requiredwith DRAs. As an example of a hero experiment, Suzuki,et al., demonstrated 25-GHz-spaced 1-Tb/s (10010 Gb/s)DWDM transmission with the high spectral efficiency of 0.4b/s/Hz in the -band over a 320 km (4 80 km) DSF [15].By employing DRA and polarization interleave multiplexing,four-wave mixing (FWM) is suppressed in the-band overdispersion-shifted fiber (DSF).

A. Issues in Using DRAs

As can be seen if from the above examples, there are somekey advantages of using DRAs. First is the improvement in noisefigure over discrete amplifiers. Thus, lower signal powers can beused, a higher loss can be tolerated, or a longer transmission dis-tance can be used between regenerators. A second advantage isa more uniform gain along the length of the fiber. This gives riseto better SNR performance and reduced nonlinear penalty. It isalso better for higher bit rates and soliton transmission. Finally,when used in conjunction with an EDFA, the complexity of typ-ical inline amplifiers can be placed in the lumped EDFA. Inother words, the DRA is just a low noise preamplifier. The gainequalization, gain level correction,ADD–DROPmultiplexers, anddispersion compensation can all be placed at the midstage in theEDFA.

There are also challenges in using DRAs. A first challenge isthat the effective length is less than 40 km of fiber in a typicalDRA. For any nonlinear effect, the effective length is set by theattenuation of the pump. For pump wavelengths that are around1450 nm, the penetration of the pump is less than 40 km intothe fiber. A second challenge of using DRAs is the high pumppowers propagating in the transmission fiber. For example, theoptimal noise figure requires about 580 mW for a DSF, or ap-proximately 1.28 W in a standard SMF [16]. At these powerlevels, connectors are highly vulnerable to damage.

In addition, there are issues associated with using DRAs infield deployment. For example, there is a higher sensitivity tospurious reflections. Moreover, there can be sensitivity to envi-ronmental and mechanical changes under the ground.

Another potential issue in DRAs is the double Rayleigh scat-tering penalty. As an example, a comparison is made of a DRAand discrete Raman amplifier. If the same fiber is used in bothamplifiers, then the DRA will have a higher DRS penalty. How-ever, if the DRA is made in a typical transmission fiber while


(a)

(b)

Fig. 6. (a) Effective Rayleigh reflection coefficient versus distributed Ramanfiber length as calculated for a dispersion-shifted fiber. A discrete Ramanamplifier is typically under 15 km, while a DRA is at least 40 km long. Curvesare shown for no pump power (lowest curve), 400, 600, and 800 mW (highestcurve) of pump power. (b) Effective Rayleigh reflection coefficient versuspump-on/pump-off gain for a DRA implemented in standard single-mode fiber(SMF-28) and a discrete amplifier implemented in dispersion compensatingfiber (DK fiber). The length of the DK-40 is 6.5 km, while the length of theDK-80 is 13 km.

the discrete Raman amplifier is implemented in a DCF, then thepenalty in the DRA may be less. In other words, the comparisonof DRS penalty for a DRA versus a discrete Raman amplifierdepends on the specific details of the amplifiers. First, a DRAis compared with a DRA both of which are implemented in thesame fiber. For example, Fig. 6(a) plots the effective Rayleighreflection coefficient in decibel verses the fiber length in kilome-ters for a DSF fiber. Different curves are plotted for no Ramanpumps, 400-, 600-, and 800-mW pump power. A typical dis-crete Raman amplifier will be less than 15 km in length, whilea DRA will be at least 40 km in length. Therefore, a DRA hasa double Rayleigh scattering penalty as much as 25 dB higherthan a discrete amplifier that is implemented in the same fibertype for the same pump power.

On the other hand, Fig. 6(b) compares the DRS penalty for aDRA in a standard single-mode fiber (SMF-28) and a discreteamplifier in DCF fiber (DK-40 is 6.5 km long, while DK-80is 13 km long). In particular, the effective Rayleigh reflectioncoefficient in decibels is plotted versus pump-on/pump-off gainin decibels. Because of the higher double Rayleigh scattering

Fig. 7. Lumped or discrete Raman amplifiers have the pump power confined tothe gain fiber in the amplifier box. Counterpropagating pump power is confinedwithin the unit via isolators.

coefficient in the DCF, the DRS penalty is higher for a given gainlevel even though the fiber length is shorter. The simulationsassume a pump at 1451 nm, where the loss is 0.29 dB/km forthe standard SMF and 0.63 dB/km for the DCF. The signal isassumed to be at 1550 nm, where the fiber loss is 0.25 dB/km inthe standard fiber and 0.49 dB/km in the DCF. The Raman gaincoefficient between the simulated wavelengths is 1.62 dB/(Wkm) for the standard fiber and 13.5 dB/(Wkm) in the DCF.The Rayleigh scattering coefficient at 1550 nm is42.3 dB/kmfor the standard fiber and33.8 dB/km for the DCF.

In summary, the power and double Rayleigh scattering limitsmay set some fundamental limits on system applications ofDRAs. There are indeed many practical issues related to havingan amplifier “stretched across the country.” Despite all of thesechallenges, DRAs are being used in almost every long-haul andultralong-haul transmission systems because of the SNR andnonlinear benefits.

IV. DISCRETERAMAN AMPLIFIERS

Discrete Raman amplifiers refer to a lumped element thatis inserted into the transmission line to provide gain. Unlike aDRA, all of the pump power is confined to the lumped element.For example, Fig. 7 shows the typical setup for a lumpedRaman amplifier. In this particular case, counterpropagatingpump power is confined within the unit by the use of isolatorssurrounding the amplifier. Compared with Fig. 4, no pumppower enters the transmission line.

The primary use for discrete Raman amplifiers is to open newwavelength bands in fused silica fibers. For example, differentwavelength bands are illustrated in Fig. 8. EDFAs operate inthe -band, which stretches from 1530 to 1565 nm, and the

-band, which stretches from about 1565 to 1625 nm. Thereis also the -band, which stretches from roughly 1480 to 1530nm, which has at least as low loss as the EDFA bands. In ad-dition, the band extends from approximately 1430 to 1480nm. Earlier transmission systems were deployed in the 1310-nmband, which can stretch between from 1280 and 1340 nm. Thereis also a 1400-nm band, which is only useful in new fibers thatuse special drying techniques to reduce the water peak absorp-tion around 1390 nm. Thus, Raman amplifiers can be used toopen up wavelengths between about 1280 and 1530 nm, a wave-length range that is inaccessible by EDFAs.


Fig. 8. Several windows exist within the low-loss valley of typical fusedsilica fibers. The solid curve corresponds to standard SMF, while the dottedcurve corresponds to fiber where additional drying steps are used to reduce thewater absorption peak near 1390 nm. EDFAs provide gain in theC-band, andextended-band EDFAs can provide gain in theL-band. Raman amplifiers arethe only fused-silica-based technology for opening up the S+ andS-bands.

Fig. 9. Ring cavity analog Raman amplifier for gain in the 1310-nm band. Thecounterpropagating amplifier corresponds to the lower part of the ring, while theupper part of the ring corresponds to a Raman oscillator and wavelength shifter.

Some of the earliest work on discrete Raman amplifiers wasdone for the 1310-nm band for cable television applications.A schematic illustration of the Raman ring amplifier designfor analog applications is depicted in Fig. 9 [7]. Light from aneodymium-doped cladding-pumped fiber laser at 1060 nm isinjected into the upper half of the amplifier to pump two cas-caded Raman lasers that lase between gratings at 1115 and 1175nm. The interstage isolator is used to reduce the effects of mul-tipath interference due to DRS within the amplifier. The use ofstrictly counterpropagating pump geometry prevents pump fluc-tuations from coupling to the signals. Output powers in excess of

23 dBm were generated while maintaining very good analogsignal performance (carrier-to-noise ratio 49 dB, com-posite second-order 60 dBc, and a composite triplebeat 68 dBc).

Despite this early work on amplifiers for the 1310-nm band,not much commercial activity has transpired since. There area number of fundamental reasons that keep the 1310 nm frombeing interesting for long-haul applications. First, the loss istoo high. For example, the typical loss in the 1310-nm windowis 0.35 dB instead of 0.2 dB in the 1550-nm window. Thehigher loss translates into shorter distance between regenerators.

Second, the 1310 nm is of primary interest for standard SMF, butit is very difficult to use WDM near the zero-dispersion wave-length of fibers. Hence, not too many channels can be openedup in the 1310-nm window. Third, the 1310-nm window is rel-atively narrow compared to the wide 1550-nm valley. At bestan additional 20 nm of bandwidth can be used in the 1310-nmwindow, and the loss rises very rapidly on the short wavelengthside.

Another window where research has been done on dis-crete Raman amplifiers is the 1400-nm band. Srivastava andcoworkers have demonstrated transmission of four 10-Gb/schannels at 1400 nm and sixteen 2.5-Gb/s channels in the1550-nm window in so-called AllWave fiber, which has lowloss and moderate chromatic dispersion in the 1400-nm region[17]. Unfortunately, this amplification window is only availablefor new installations that use the premium-cost fiber.

A. Opening Up the -band Using Optical Amplifiers

Of the new bands to be opened up, perhaps the most importantis the -band [4], [36]. The -band has comparable or betterattenuation characteristics in standard SMFs than the-band,and the -band has far less sensitivity to attenuation caused bybending during cabling and installation than the-band. In ad-dition, the -band has better dispersion characteristics in stan-dard SMF than - and -band. For example, the-band hasapproximately 30% less dispersion than the-bands. Semicon-ductor optical amplifiers, thulium-doped fiber amplifiers in ei-ther fluoride or multicomponent silicate glasses, and lumped,or discrete, Raman amplifiers (LRAs) have all been proposedas key enablers for opening up the-band. However, Ramanamplifiers are the only fused-silica-based amplifier solution forthe -band.

As an example of the application of-band LRA, considerthe following experiment of the first -band long-haul WDMtransmission using a cascade of dispersion compensating LRAs.Twenty NRZ channels, spanning the entire-band, were trans-mitted over ten spans of SSMF, each achievingwithout forward error correction.

In particular, Puc and coworkers demonstrate the cascadeof 11 rack-mounted -band dispersion-compensating lumpedRaman amplifiers (SLRAs) to transmit 20-band channelsmodulated at 10.67 Gb/s over 867 km of standard SMF [4]. Themargins accumulated in this demonstration show the capabilityfor such a system to achieve 80-channel transmission over10 25-dB SSMF with standard out-of-band forward errorcorrection (OOB-FEC) and presently available SLRAs.

Fig. 10 shows a photograph of a standard 7-ft rack containingsix SLRAs. Each SLRA is a two-stage amplifier containing again-flattening filter (GFF) with a midstage access [Fig. 11(a)].The pump module corresponds to four laser diodes wavelengthand polarization multiplexed together [Fig. 11(b)]. The pumpwavelengths are selected to achieve sufficiently flat gain acrossthe -band. The net pump power used in this configuration isapproximately 600 mW into the fiber with about 300 mW ateach pump wavelength.

Fig. 12 shows the typical gain and noise performance of theSLRAs. It can be seen that high values of gain (up to 30 dB)


Fig. 10. Photograph of sixS-band lumped Raman amplifiers (SLRAs)mounted in a seven foot rack. (BCM= band coupler module. APM=administrative processing module. OSA= optical spectrum analyzer).

(a)

(b)

Fig. 11. Block diagram of a two-stageS-band Raman amplifier. (a) Twostages of gain fiber with counterpropagating pumping and midstage access. (b)Polarization diversity pumping scheme using two lasers at each wavelengththat are polarization combined. The two pump wavelengths are then combinedusing a wavelength-division multiplexer (WDM).

can be achieved with a very small gain ripple (1 dB), in ad-dition to noise figure values on the order of 5.5 dB across the

-band (1493–1523 nm). The SLRA gain fiber has a high neg-ative dispersion in the -band, as well as negative dispersionslope, providing coarse dispersion and dispersion slope com-pensation throughout the band. Each SLRA compensates forabout 75 km of SSMF.

The setup block diagram is shown in Fig. 13. Eleven disper-sion-compensating SLRAs were used to transmit 20 channelsin the -band (between 1493.36 and 1521.77 nm) nominallyspaced by 200 GHz and modulated at 10.67 Gb/s over ten spansof SSMF, for a total length of 867 km. Each span contained on

Fig. 12. Gain (upper curves) and noise figure (lower curve) versus wavelengthfor (a)�14-dBm total input power and (b)�8-dBm total input power. The pumppower is varied between 18 and 27 dBm, and the gain ripple remains<1 dBover the entire dynamic range. Also, the noise figure is�5.5 dB over the entireS-band.

average six connectorized joints and additional loss elements tobring the average span loss to 21 dB.

The average amplifier output power was only14 dBm, re-sulting in a launched power per channel of1 dBm, whilst eachSLRA was capable of at least19-dBm output power. All thechannels were launched in parallel polarization, modulated witha pseudorandom binary sequence (PRBS) NRZ pattern at10.67 Gb/s through an external LiNbOmodulator. The receiverhad an adjustable decision threshold and a clock extractor, al-lowing all the BER readings to be made at a maximum likeli-hood setting.

The dispersion map is sketched in Fig. 14. The line consistedof sections of SSMF, with a dispersion of about 15 ps/nmkmat 1510 nm. A section of DCF was inserted between the twostages of the fifth SLRA. The zero-dispersion wavelength of thedispersion map was at 1508 nm. The crosses and the legend inFig. 14 show the measured residual dispersion at the center andtwo extreme wavelengths.

The received spectrum is shown in Fig. 15. The averagereceived OSNR (0.1-nm bandwidth) was about 20.7 dB. Atthis nominal level, all the channels operated atwithout any error correction at 10.67-Gb/s line rate. In addition,Puc, et al., measured BER versus OSNR curves, by raisingthe noise floor of the spectrum whilst maintaining the peaksignal power constant. The results are summarized in Fig. 16.The most dispersive channel was #20. Its received optical eyediagram is shown as an inset in Fig. 16. Almost no dispersionpenalty was observed.


Fig. 13. Block diagram of the experimental setup. Eleven dispersion-compensating SLRAs were used to transmit 20 channels in theS-band (between 1493.36and 1521.77 nm) nominally spaced by 200 GHz and modulated at 10.67 Gb/s over ten spans of SSMF, for a total length of 867 km.

Fig. 14. Sketch of the dispersion map. The crosses and the legend show themeasured residual dispersion at the center and the two extreme wavelengths.

Fig. 15. Received spectrum after the last amplifier stage. The average OSNR(0.1-nm bandwidth) is�20.7 dB.

Similarly the channels did not suffer from FWM effect. Evenwhen five channels were spaced by 50 GHz, no FWM intermod-ulation product was observed down to 28 dB below the channellevel (Fig. 17). At the nominal launched channel power level( 1 dBm), the self-phase modulation (SPM) appeared to be thelargest impairment aside from amplified spontaneous emission.The estimated SPM penalty was less than 1 dB for any of the

Fig. 16. Bit error rate (BER) performance versus OSNR, which is measured byraising the noise floor of the spectrum while maintaining the peak signal powerconstant. The inset shows the eye diagram for the most dispersive channel (#20).

transmitted channels. No cross-phase modulation penalty wasobserved.

For an 80-channel system using the full power level of theSLRAs ( 19 dBm), the power per channel would be 0 dBm.Furthermore, if a standard OOB-FEC such as Reed–Solomon255/239 is employed, one can expect that a reduction of theOSNR by 5 dB would result in a similar BER performance as ifno FEC were used. Therefore, the present demonstration trans-lates into a capability of transmitting 80 channels at 10.67 Gb/swith OOB-FEC (OC-192 line rate) over ten spans of 25 dB.

In summary, a novel, -band long-haul WDM transmissionscheme using dispersion-compensating lumped Raman am-plifiers, has been successfully tested for the first time. Thisten-span transmission demonstration confirms the viability


Fig. 17. Output spectrum for five 10.67-Gb/s channels that are spaced by 50 GHz. The left curve is with all five channels on, and the right curve is with the centerchannel off. No FWM intermodulation product is observed down to 28 dB below the channel level.

of SLRAs as a key enabling technology for an expansion ofoptical networks into the S/S-band regions, and potentially toother wavelength windows.

V. SUMMARY

There has been a revived interest in Raman amplification dueto the availability of high pump powers and improvements insmall core size fibers. Two general categories of Raman ampli-fiers are DRAs and lumped or discrete Raman amplifiers. DRAsimprove the noise figure and reduce the nonlinear penalty of theamplifier, allowing for longer amplifiers spans, higher bit rates,closer channel spacings, and operation near the zero-dispersionwavelength. DRAs are already becoming commonplace in mostlong-haul networks.

Discrete Raman amplifiers are primarily used to increase thecapacity of fiber-optic networks. Amplifiers have been demon-strated in the 1310-nm band, but this wavelength range suffersfrom higher loss and more nonlinear penalty in standard SMFs.For fibers with less water absorption, Raman amplifiers can alsobe used in the 1400-nm band. The capacity of most fibers canalmost be doubled by opening up the-band, which is inacces-sible by EDFAs. As an example, we described a novel-band,long-haul WDM transmission, which uses a cascade of disper-sion compensating -band lumped Raman amplifiers. Trans-mission is shown for 20 OC-192 NRZ channels over 867 kmof standard fiber.

Raman amplifiers provide a simple single platform forlong-haul and ultralong-haul amplifier needs. Raman am-plifiers are broad-band and wavelength agnostic. Ramanamplifiers can be distributed, lumped or discrete, or hybrid.Also, in Raman amplifiers the amplification and dispersioncompensation can be combined in the same fiber length. Forhigh channel count systems, as will be deployed in the nextfew years, Raman amplifiers’ efficiency actually exceedseven 1480-nm pumped-band EDFAs. Consequently, Raman

amplifiers should see a wide range of deployment in the nextfew years.

REFERENCES

[1] R. H. Stolen and E. P. Ippen, “Raman gain in glass optical waveguides,”Appl. Phys. Lett., vol. 22, no. 6, 1973.

[2] G. P. Agrawal,Nonlinear Fiber Optics, 2nd ed. New York: Academic,1995.

[3] A. Mathur, M. Ziari, and V. Dominic, “Record 1 watt fiber-cou-pled-power 1480 nm diode laser pump for Raman and erbium dopedfiber amplification,” inOptical Fiber Communication Conf. OSA Tech.Dig. Washington, DC, 2000.

[4] A. B. Puc, M. W. Chbat, J. D. Henrie, N. A. Weaver, H. Kim,A. Kaminski, A. Rahman, and H. A. Fevrier, “Long-haul WDMNRZ transmission at 10.7 Gb/s in S-band using cascade of lumpedRaman amplifiers,” inOptical Fiber Communication Conf. OSA Tech.Dig. Washington, DC, 2001.

[5] V. Dominic, E. Mao, J. Zhang, B. Fidric, S. Sanders, and D.Mehuys, “Distributed Raman amplification with co-propagatingpump light,” in Optical Amplifiers and Their Applications OSA Tech.Dig. Washington, DC, 2001.

[6] E. Desurvire,Erbium-Doped Fiber Amplifiers Principles and Applica-tions. New York: Wiley, 1994.

[7] A. J. Stentz, T. Nielsen, S. G. Grubb, T. A. Strasser, and J. R. Pedrazzani,“Raman ring amplifier at 1.3�m with analog-grade performance and anoutput power of 23 dBm,” inOptical Fiber Communication Conf. OSATech. Dig. Washington, DC, 1996.

[8] C. R. S. Fludger and V. Henderek, “Fundamental noise limits in broad-band Raman amplifiers,” inOptical Fiber Communication Conf. OSATech. Dig. Washington, DC, 2001.

[9] H. Kidorf, K. Rottwitt, M. Nissov, M. Ma, and E. Rabarijaona, “Pumpinteractions in 100 nm bandwidth Raman amplifier,”IEEE Photon.Technol. Lett., vol. 11, May 1999.

[10] S. Namiki and Y. Emori, “Broadband Raman amplifiers design andpractice,” in Optical Amplifiers and Their Applications OSA Tech.Dig. Washington, DC, 2000.

[11] M. Nissov, C. R. Davidson, K. Rottwitt, R. Menges, P. C. Corbett, D.Innis, and N. S. Bergano, “100 Gb/s (10� 10 Gb/s) WDM transmissionover 7200 km using distributed Raman amplification,” inEuropean Con-ference Optical Commun. 1997, Postdeadline Paper, pp. 9–12.

[12] T. Terahara, T. Hoshida, J. Kumasako, and H. Onaka, “128� 10.66Gbit/s transmission over 840-km standard SMF with 140-km opticalrepeater spacing (30.4-dB loss) employing dual-band distributedRaman amplification,” inOptical Fiber Communication Conf. OSATech. Dig. Washington, DC, 2000.


[13] L. D. Garrett, M. Eiselt, and R. W. Tkach, “Field demonstration of dis-tributed Ramn amplification with 3.8 dB Q-improvement for 5� 120km transmission,” inOptical Fiber Communication Conf. OSA Tech.Dig. Washington, DC, 2000.

[14] T. N. Nielsen, A. J. Stents, K. Rottwitt, D. S. Vengsarkar, Z. J. Chen, P.B. Hansen, J. H. Park, K. S. Feder, S. Cabot, S. Stulz, D. W. Peckham,L. Hsu, C. K. Kan, A. F. Judy, S. Y. Park, L. E. Nelson, and L. Grüner-Nielsen, “3.28-Tb/s transmission over 3� 100 km of nonzero-disper-sion fiber using dual C- and L-band distributed Raman amplification,”IEEE Photon. Technol. Lett., vol. 12, Aug. 2000.

[15] H. Suzuki, J. Kani, H. Masuda, N. Takachio, K. Iwatsuki, Y. Tada, andM. Sumida, “25 GHz-spaced, 1 Tb/s (100� 10 Gb/s) super dense-WDMtransmission in the C-band over a dispersion-shifted fiber cable em-ploying distributed Raman amplification,” inEur. Conf. Optical Com-munication 1999, Postdeadline Paper PD 30-31.

[16] P. B. Hansen, L. Eskildsen, A. J. Stentz, T. A. Strasser, J. Judkins, J.J. DeMarco, R. Pedrazzani, and D. J. DiGiovanni, “Rayleigh scatteringlimitations in distributed Raman pre-amplifiers,”IEEE Photon. Technol.Lett., vol. 10, pp. 159–161, Jan. 1998.

[17] A. K. Srivastava, D. L. Tzeng, A. J. Stentz, J. E. Johnson, M. L. Pearsall,O. Mizuhara, T. A. Strasser, K. F. Dreyer, J. W. Sulhoff, L. Zhang, P. D.Yates, J. R. Pedrazzani, A. M. Sergent, R. E. Tench, J. M. Freund, T.V. Nguyen, H. Manzar, Y. Sun, C. Wolf, M. M. Choy, R. B. Kummer,D. Kalish, and A. R. Chraplyvy, “High-speed WDM transmission in allwave fiber in both the 1.4-�m and 1.55-�m Bands,” inOSA Tech. Dig.,Optical Amplifiers and Their ApplicationsWashington, DC, 1998.

[18] R. H. Stolen, “Issues in Raman gain measurements,” inTech. Dig. Symp.Optical Fiber Measurements, NIST Special Publication 953 (NationalInstitute of Standards and Technology), Gaithersburg, MD, 2000, pp.139–142.

[19] J. Yoshida, N. Tsukiji, A. Naki, T. Fukushima, and A. Kasukawa,“Highly reliable high power 1480 nm pump lasers,” inTesting, Relia-bility, and Application of Optoelectronic Devices. Bellingham, WA:SPIE, Society of Photo-Optical Instrumentation Engineers, 2001, Proc.SPIE Volume 4285, pp. 146–158.

[20] S. A. Lewis, S. V. Chernikov, and J. R. Taylor, “Temperature-dependentgain and noise in fiber Raman amplifiers,”Opt. Lett., vol. 24, no. 24, pp.1823–1825, 1999.

[21] S. Namiki and Y. Emori, “Ultrabroad-band Raman amplifiers pumpedand gain-equalized by wavelength-division-multiplexed high-powerlaser diodes,”IEEE J. Select. Topics Quantum Electron., vol. 7,Jan.–Feb. 2001.

[22] Y. Emori, Y. Akasaka, and S. Namiki, “100 nm bandwidthflat-gain Raman amplifiers pumped and gain-equalized by12-wavelegnth-channel WDM laser diode unit,”Electron. Lett.,vol. 35, pp. 1355–1356, 1999.

[23] K. Rottwitt and H. D. Kidorf, “A 92 nm bandwidth Raman amplifier,”in Proc. Optical Fiber Commun. Conf., 1998, Postdeadline Paper PD6.

[24] Y. Emori and S. Namiki, “100 nm bandwidth flat gain Raman ampli-fiers pumped and gain-equalized by 12-wavelength-channel WDM highpower laser diodes,” inProc. Optical Fiber Commun. Conf., 1999, Post-deadline Paper PD19.

[25] C. R. S. Fludger, V. Handerek, and R. J. Mears, “Ultra-wide bandwidthRaman amplifiers,” inOptical Fiber Communication Conf. TechnicalDig., 2002, Paper TuJ3, pp. 60–62.

[26] T. Naito, T. Tanaka, K. Torii, N. Shimojoh, H. Nakamoto, and M.Suyama, “A broadband distributed Raman amplifier for bandwidthsbeyond 100 nm,” inOptical Fiber Communication Conf. TechnicalDig., 2002, Paper TuR1, pp. 116–117.

[27] H. Masuda, S. Kawai, K. I. Suzuki, and K. Aida, “75-nm 3-dB gain-bandoptical amplification with erbium doped fluoride fiber amplifiers anddistributed Raman amplifiers in 9� 2.5Gb/s WDM transmission ex-periment,” inProc. Eur. Conf. Optical Communications, vol. 5, 1997,Postdeadline Paper, p. 74.

[28] H. Takara, H. Masuda, K. Mori, K. Sato, Y. Inoue, T. Ohara, A. Mori,M. Mohtuku, Y. Miyamoto, and S. Kawanishi, “Ultra-wideband tellu-rite/silica fiber Raman amplifier and supercontinuum lightwave sourcefor 124-nm seamless bandwidth DWDM transmission,” inOptical FiberCommunication Conf., 2002, Paper FB1.

[29] A. H. Gnauck, G. Raybon, D. Grosz, S. Hunsche, A. Kung, A. Marhe-lyuk, D. Maywar, M. Movassaghi, X. Liu, C. Xu, X. Wei, and D. M.Gill, “2.5Tb/s (64� 42.7 Gb/s) transmission Over 40� 100 km NZDSFusing RZ-DPSK format and all-Raman-amplied spans,” inOptical FiberCommunication Conf., 2002, Paper FC2.

[30] F. Liu, J. Bennike, S. Dey, C. Rasmussen, B. Mikkelsen, P. Mamyshev,D. Gapontsev, and V. Ivshin, “1.6 Tbit/s (40� 42.7 Gbit/s) transmis-sion over 3600 km UltraWave™ fiber with all-Raman amplified 100 kmterrestrial spans using ETDM transmitter and receiver,” inOptical FiberCommunication Conf., 2002, Paper FC7.

[31] B. Zhu, L. Leng, L. E. Nelson, L. Grüner-Nielsen, Y. Qian, J. Bromage,S. Stulz, S. Kado, Y. Emori, S. Namiki, P. Gaarde, A. Judy, B. Palsdottir,and R. L. Lingle, Jr., “3.2 Tb/s (80� 42.7Gb/s) transmission over 20�100km of nonzero dispersion fiber with simultaneous C+ L-band dis-persion compensation,” inOptical Fiber Communication Conf., 2002,Paper FC8.

[32] L. Grüner-Nielsen, Y. Qian, B. Pálsdóttir, P. B. Gaarde, S. Dyrbøl, and T.Verg, “Module for simultaneous C+ L-band dispersion compensationand Raman amplification,” inTech. Dig., Optical Fiber CommunicationConf., 2002, Paper TuJ6, pp. 65–66.

[33] P. B. Hansen, G. Jacobovitz-Veselka, L. Grüner-Nielsen, and A. J.Stentz, “Raman amplification for loss compensation in dispersioncompensating fiber modules,”Electron. Lett., vol. 34, pp. 1136–1137,1998.

[34] Y. Emori, Y. Akasaka, and S. Namiki, “Less than 4.7 dB noise figurebroadband in-line EDFA with a Raman amplified-1300 ps/nm DCFpumped by multi-channel WDM laser diodes,” inProc. OpticalAmplifiers and Their Applications, 1998, Postdeadline Paper PD3.

[35] , “Broadband losses DCF using Raman amplification pumpedby multichannel WDM laser diodes,”Electron. Lett., vol. 34, pp.2145–2146, 1998.

[36] J. Bromage, H. J. Thiele, and L. E. Nelson, “Raman amplification in theS-band,” inOptical Fiber Communication Conf. Tech. Dig., 2002, PaperThB3, pp. 383–385.

[37] H. Masuda, S. Kawai, and K.-I. Suzuki, “Optical SNR enhanced ampli-fication in long-distance recirculating-loop WDM transmission experi-ment using 1580nm band hybrid amplifier,”Electron. Lett., vol. 35, pp.411–412, 1999.

[38] S. A. E. Lewis, S. V. Chrenikov, and J. R. Taylor, “Broadband high-gaindispersion compensating Raman amplifier,” inElectron. Lett., vol. 36,2000, pp. 1355–1356.

Mohammed N. Islam received the B.S., M.S., andSc.D. degrees in electrical engineering from theMassachusetts Institute of Technology, Cambridge,MA, in 1981, 1983 and 1985, respectively.

From 1985 to 1992, he was a member of thetechnical staff in the photonic switching departmentand then the advanced photonics department atAT&T Bell Laboratories, Holmdel, NJ. In 1992, hejoined the EECS department at the University ofMichigan, Ann Arbor, MI, where he is currently afull tenured professor. In 1999, he went on sabbatical

leave at Stanford University, and during the 2000 and 2001 academic yearswas on leave-of-absence at Xtera Communications, Inc. He was a Fannieand John Hertz Fellow from 1981 to 1985, and in 1992 we was awarded theOSA Adolf Lomb Medal for pioneering contributions to nonlinear opticalphenomena and all-optical switching in optical fibers. He also received theUniversity of Michigan research excellence award in 1997 and became aFellow of the Optical Society of America in 1998. He has published over 100papers in refereed journals and holds over 75 patents awarded or pending.He has also authored one book, edited a second book, and written numerousbook chapters. In addition, he has founded several start-up companies basedon his research at the University of Michigan. AccuPhotonics, Inc. was startedin 1994, which designed and manufactured fiber-optic probes for biomedicalimaging. AccuPhotonics was acquired by Seaflower Associates in 1997.Then, in 1998, he founded Xtera Communications, Inc., which provideslong-haul fiber-optic systems based on its unique all-Raman amplificationtechnology. He remains the Chief Technology Officer of Xtera. In 2000, healso founded Optical Regen, Inc., and Cheetah Optics, Inc. Optical Regendesigned all-optical regeneration equipment for long-haul fiber-optic systems,while Cheetah Optics designed tunable filters andADD–DROPmultiplexers fordynamic fiber-optic networks. Both new companies have been acquired andmerged into Xtera Communications.

Date post:	22-Jan-2023
Category:	Documents
Upload:	uinsgd
View:	0 times
Download:	0 times

Raman Amplifiers for Telecommunications

Documents