Raman amplification is emerging as a crucial technology for high-speed, long-distance fiberoptic transmission. Distributed Raman amplification can supplement erbium-doped fiber amplifiers (EDFAs) in two vital ways: by increasing total gain and by flattening gain across the transmission spectrum for dense wavelength-division multiplexing. "Raman amplification is mandatory to achieve the necessary optical power budget" for transmission at 40 Gbit/s, said Gerhard Elze, chief technology officer of Alcatel Optronics, at the executive forum conducted by the Optical Society of America before the Optical Fiber Communications conference (OFC, March 2001; Anaheim, CA). Few at OFC seemed inclined to disagree.
Wide interest in Raman amplification was evident at OFC, from many exhibitors on the show floor touting Raman capabilities to post-deadline papers that depended on Raman amplification for record-setting transmission demonstrations. System developers first turned to Raman amplification to balance gain across the erbium-amplifier band without wasting power in gain-equalizing filters. Now developers are extending Raman technology in many ways, seeking more gain across a broader range of wavelengths, as well as new pump sources and pumping schemes.
Raman physics
Raman amplification is based on Raman scattering, which occurs when an atom absorbs a photon and then releases a photon with different energy, using the difference in energy to shift its vibrational energy state. The process can either raise or lower the atom's vibrational energy, but the effect is stronger if it converts some of the input photon energy to vibrational energy (called a phonon), so the atom releases a photon with lower energy than the input light. Raman amplifiers depend on Raman shifts to longer wavelengths and thus lower photon energies. The vibrational transitions are not sharp, and span a range of energies that depends on the material.
Stimulated Raman scattering can occur when a strong pump beam and a weaker signal beam at a longer wavelength pass through the material simultaneously. The strong pump beam excites many atoms, and photons at the signal wavelength can stimulate them to emit their energy as an added photon at the signal wavelength. This stimulated Raman scattering amplifies the input signal in the same way that stimulated emission by erbium atoms amplifies signals in an EDFA. Some crucial differences shape the functional distinctions between Raman and erbium amplifiers.
Raman gain is offset from the pump wavelength by the frequency shift arising from vibrational excitation, which is characteristic of the material; thus, changing the pump wavelength directly shifts the amplified wavelength. The amount of Raman scattering and its spectrum depend on the material. For silica, the peak Raman gain is 13 THz lower than the pump frequency, corresponding to a wavelength about 100 nm longer than the pump near 1550 nm. This means that unlike EDFAs, Raman amplifiers can be tailor-made to operate at specific wavelengths, including parts of the spectrum where other amplifiers are not available.
Stimulated Raman scattering is a nonlinear process, so its cross-section is fairly weak, and depends strongly on optical power density. This means that high pump powers are needed over long lengths of single-mode fiber to produce reasonable gain. Although this raises the threshold for Raman amplification, it also controls undesired Raman noise that could impair performance of fiber systems.
The scattering elements in Raman amplifiers are the atoms in the glass itself, not a dopant present only in low concentrations, such as erbium in EDFAs. So far, most Raman amplifiers have been based on scattering from the silica body of the fiber, although silica has lower Raman gain than germanium, phosphate, and borate glass. Doping fibers with other glasses can increase Raman amplification, but it also increases attenuation. Progress has been made in reducing loss in high-germanium fibers to about 0.5 dB/km near 1550 nm, said E. M. Dianov of the Fiber Optics Research Center (General Physics Institute; Moscow, Russia) in an invited OFC paper. The lowest loss in high-phosphate fibers is about 1 dB/km.
Types of Raman amplification
So far most development has concentrated on distributed Raman amplifiers, in which the gain medium is a long stretch of conventional single-mode fiber used for signal transmission. In the usual arrangement, the amplification occurs in the final length of fiber before the receiver or EDFA. The signal and pump beam travel in opposite directions, with the pump coupled into the fiber at the receiver end. A coupler directs the pump light into the transmitting fiber, while diverting signals arriving through the fiber to the receiver or EDFA. Stimulated Raman scattering transfers energy from the strong pump beam to the weak signal beam passing in the other direction. The gain per unit length is much smaller than in EDFAs, but the pumping is spread through kilometers of fiber, so gain readily reaches several decibels. Distributed Raman amplification is inherently lower in noise than lumped EDFAs. However, noise arising from double Rayleigh scattering currently limits total distributed Raman gain to about 20 dB.
An alternative design is a discrete Raman amplifier that is not part of the transmission line. Discrete amplifiers could use higher-loss fiber with higher Raman gain, a stretch of fiber separate from the main transmission line designed to have higher Raman gain per unit length. The technology is not well-developed, but it has some attractions because it could compensate for loss in components other than transmission fibers, and might provide amplifiers in bands where none have been available. At OFC, T. Tsuzaki of Sumitomo Electric's Yokohama Research Laboratories (Yokohama, Japan) reported a discrete Raman amplifier with gain of more than 12 dB between 1642 and 1672 nm.
Stimulated Raman scattering also can be used as the basis of a laser. The most attractive application in optical networking is shifting wavelengths of strong laser sources to the bands needed to pump optical amplifiers. A nested cascade of oscillators can shift the 1100-nm output of a high-power, ytterbium-doped fiber laser in steps to the 1400-nm band needed for Raman pumping. In an OFC post-deadline paper, M. D. Mermelstein of Lucent Technologies' Specialty Fiber Devices (Somerset, NJ) described a cascaded laser with output at 1427, 1455, and 1480 nm.
Hybrid Raman amplifiers and gain flattening
The practical limits on Raman gain have helped focus most immediate interest on hybrid Raman/erbium amplifiers in which a distributed Raman stage enhances total gain and increases spectral uniformity without adding noise. Gain flattening is important and relatively straightforward. Gain in EDFAs peaks at shorter wavelengths, while Raman gain peaks at longer wavelengths. With a suitable pump source, distributed Raman amplification in the transmitting fiber can deliver signals that have higher powers at longer wavelengths in the erbium band, offsetting the lower gain in EDFAs. This levels the overall gain without sacrificing power in gain-flattening filters.
Further enhancements in gain flattening come from pumping the Raman stage at multiple wavelengths. Superposition of these gain peaks produces uniform gain across a broader range, although interactions between the pump and signal beams appears to limit the usable range.
The extra gain Raman amplifiers can produce with little noise penalty is expected to be critical for pushing data rates to 40 Gbit/s per optical channel. Noise and nonlinear effects such as four-wave mixing limit total power available from EDFAs. Distributed Raman amplification can avoid these effects, with the distributed nature of the process keeping powers low enough to control nonlinear effects. That's crucial because 40-Gbit/s systems will require average powers 6 dB higher than present 10-Gbit/s systems. Receiver sensitivity depends on the number of photons per bit, and multiplying the bit rate by four divides the bit interval by the same amount—so the same number of photons must be delivered in one-fourth the time.
Several 40-Gbit/s transmission experiments described at the OFC post-deadline session relied on some distributed Raman amplification to achieve record performance, including groups that squeezed 10 Tbit/s through a single fiber. The NEC group, which set a record by transmitting 10.92 Tbit/s, used distributed Raman amplification to complement a thulium-doped fiber amplifier, so they could pack 85 channels into the S-band at wavelengths of 1467.81 to 1508.01 nm. That points to another important role for Raman amplification: helping to open bands where good doped-fiber amplifiers are not available. Other experiments showed that distributed Raman amplification could stretch spacing between EDFAs. In another OFC post-deadline paper, Takayuki Miyakawa and a group from KDD R&D Laboratories (Saitama, Japan) transmitted 64 channels, operating at 40 Gbit/s with 50-GHz spacing, a distance of 230 km without erbium amplifiers.
Promise and challenges
The flexibility of Raman amplification gives the bright promise of using Raman amplifiers to fill in holes where other good optical amplifiers are not available. Commercial deployment of the technology is just starting, but the signs appear encouraging.
Yet important challenges remain. Although distributed Raman amplification can be used throughout the low-loss fiber transmission window, it can't be used across that entire range in the same fiber because pump wavelengths would overlay signal wavelengths. The Raman pumps needed to amplify the L-band at 1570 to 1620 nm would sit near 1480 nm in the S-band. The power demands of pump lasers seem forbidding to developers of transoceanic submarine cables who have only limited power available from their land terminals. Some observers warn that the high pump powers needed for Raman amplification raise serious concerns about eye safety in an operating environment. These and other challenges mean that 40-Gbit/s transmission systems will not come easily, but there is general optimism that they will be available eventually.
