9.5.2 Use of Raman Amplification

The nonlinear effects can also be controlled to some extent by using a distributed amplification scheme. This issue has been discussed in Section 6.6 in the context of single-channel systems. As was seen there, the use of Raman amplification allows one to obtain the same value of the Q factor at much lower values of the launch powers by lowering the level of accumulated noise. As one may expect, a lower launch power also helps in reducing the interchannel XPM effects in WDM systems. Indeed, the use of Raman amplification has become quite common for modern WDM systems [114]-[119].

There is a second reason why Raman amplification may help in reducing the XPM-induced timing jitter. Its use helps to make peak-power variations along the link much less drastic than those occurring when lumped amplifiers are used. As a result, the function p(z) appearing in Eq. (9.4.10) becomes nearly constant and can be pulled out of the integral. If the collision of two pulses belonging to neighboring channels is complete in a fiber section with constant values of the dispersion and nonlinear parameter, the XPM-induced frequency shift vanishes, and no temporal shift is produced by such collisions.

Several experiments have shown that distributed Raman amplification can be used to transmit a large number of channels with a relatively small channel spacing. Such systems are referred to as superdense or ultradense WDM systems. In a 2000 experiment, 100 channels, each operating at 10 Gb/s and spaced by only 25 GHz from the neighboring channels, were transmitted over 320 km of dispersion-shifted fiber [114]. Figure 9.20 shows the experimental setup schematically. Link losses were compensated every 80 km through distributed Raman amplification using a backward pumping configuration. It was necessary to employ the technique of forward error correction (FEC) to ensure a BER of <10^-9. Because of a 25-GHz channel spacing, the entire WDM signal could fit in the wavelength range of 1,540 to 1,560 nm. The zero-dispersion wavelength of the dispersion-shifted fiber was near 1,550 nm, indicating that the dispersion was relatively low for all channels. Indeed, it was not necessary to compensate dispersion over the 320-km length.

Figure 9.20: Schematic of experimental setup used to transmit 100 channels over 320 km with 25-GHz channel spacing. A Raman-pumping unit (RPU) compensates fiber losses every 80 km. FEC: forward error correction; LN: lithium niobate modulator, PC: polarization controller; OTF: optical tunable filter; Rx: optical receiver. (After Ref. [114]; ©2000 IEEE.)

The reason why the 100-channel system was not limited by interchannel crosstalk in spite of its 25-GHz channel spacing is related to a relatively low input power (only -16 dBm/channel) that was launched into the fiber link. The use of Raman amplification lowered the noise level enough that the system could be operated at such low input power levels without much penalty from the intrachannel as well as interchannel nonlinear effects. Indeed, the system performance after 320 km was limited by accumulated noise rather than the nonlinear effects. The transmission distance could be increased to beyond 1,000 km by employing the bidirectional pumping scheme that is known to lower the noise level even further. Figure 9.21 shows the optical spectrum recorded after 1,040 km together with the the BER and the Q factors (without FEC) measured as a function of channel wavelength [114]. The power launched into each channel was -18 dBm. In the absence of FEC, the BER varied in the range of 10^-4 to 10^-8. However, the BER for all channels could be reduced to below 10^-12 with the use of FEC. In a similar experiment, 100 channels with 25-GHz spacing were transmitted in the L band using a wavelength range of 1,570.9 to 1,591.6 nm [115]. A distinct advantage of the Raman amplification is that link losses can be compensated in the S, C, or L band by simply changing the pump wavelengths.

Figure 9.21: (a) Optical spectrum and (b) measured Q factors after 320 km for the 100-channel WDM system. The dotted line shows the Q value required to maintain a BER below 10^-12 with FEC. (After Ref. [114]; ©2000 IEEE.)

The performance of Raman-amplified WDM systems has improved considerably in recent years [118]. In a 2001 experiment, 128 channels, each operating at 10 Gb/s, were transmitted over 4,000 km of reduced-slope dispersion-shifted fiber (RS-DSF) using a 100-km-long recirculating loop. An average power of -5 dBm was launched into each channel. The channel spacing was 50 GHz in this experiment. Even when standard fibers with a relatively large dispersion were employed, the WDM system could be operated over 3,200 km. Such performance is difficult to realize with lumped amplifiers. Several experiments have used Raman amplification for WDM systems in which each channel operates at 40 Gb/s. In one 64-channel experiment, the WDM signal with a capacity of 2.56 Tb/s could be transmitted over 1,600 km of RS-DSF. It was necessary to optimize the launch power as the system performance degraded considerably when power was increased from -2 to -1 dBm. Figure 9.22 shows the system margin and the penalty measured as a function of channel wavelength at these two power levels. The system margin is defined as the difference between the received and required optical SNR (both measured on a dB scale). The penalty is defined as the increase in required optical SNR compared with the value needed in the absence of fiber link (both measured on a dB scale). The penalty is below 3 dB for all channels at a power level of -2 dBm/channel but increases to beyond 6 dB for some channels when this power is increased by 1 dB.

Figure 9.22: Measured (a) system margin and (b) power penalties after 1,600 km as a function of channel wavelengths for two values of launched power/channel for a 64-channel WDM system with 2.56-Tb/s capacity. (After Ref. [118]; ©2004 IEEE.)