Trial Configurations and Results
A field trial using Verizon’s metro SSMF optical cable running around North Dallas, Texas was conducted using an All-Raman ULH system, 40Gbps transponders and Juniper T640 routing platforms. Two ULH trial distances were configured. The first configuration carried 2×40Gbps channels (CS-RZ and RZ-DPSK) and 68×10Gbps NRZ channels over 2560 Km, the transmission limit for 40Gbps CS-RZ. The second configuration was set at 3040 km; the 40Gbps CS-RZ channel was replaced by a 10Gbps NRZ channel. The 40Gbps RZ-DPSK was error-free after E-FEC and exhibited a 3.5 dBQ margin. The Verizon’s cable used in this trial, installed on year 19xx, contains 432 SSMF optical fibers of 80 km in length, with revenue baring traffic on some of the fibers. 32 of these fibers were used to assemble the 2560 km system; then the distance was extended to 3040 km by adding 6 more spans to the link configuration. Each fiber goes through 8 ODFs, adding up to a total of 576 and 684 connectors for the two configurations respectively.
The average span loss, as reported by the EMS, was 20.6 dB, including all connector losses.
The architecture of the All-Raman wideband amplifier is schematically shown in fig. 2 left. Three stages of amplification are used: a distributed stage and 2 discrete stages constituted of negative dispersion fiber, which also helps to partially compensate the span positive dispersion. One of the biggest challenges in the design of a wide band amplifier is to keep the noise figure low across the entire signal spectrum. The optimized choice of pump wavelengths and pump levels for these amplifiers results in an equivalent noise figure less than 1 dB [4].
 |
 |
Fig. 2. Left: All-Raman amplifier structure. Right: 3040 Km link accumulated dispersion for a 10Gbps channel.
The 100 nm amplifier bandwidth is divided in four sub-bands of 60 channels each. The link’s residual dispersion is compensated at the terminal’s sub-band amplifiers with the proper amount of DCF that makes each sub-band channel error-free with an adequate margin. Channel aggregation of 50 GHz-spaced 10Gbps channels is obtained by two 30 channels x 100 GHz Mux/Demux per sub-band. At the drop side, an interleaver, besides demultiplexing odd and even channels, is used to limit the signal bandwidth reaching the receivers. Due to the larger spectrum of the 40Gbps channels transmitted in sub-band 3, the interleaver was removed in this band only. Fig. 2 right shows the accumulated dispersion at 1550.75 nm for the 3040 Km link. Residual dispersion of the 40Gbps transponders needed a finer tuning compared to the 10Gbps transponders.
 |
Fig. 3. Optical spectra at 2560 Km. Right picture shows the signal fluctuation due to PDL-PDG after 10 days.
The standard tolerance window (±50 ps/nm) of the 40 Gbs channels is extended to 500 ps/nm with the help of a tunable grating dispersion compensator incorporated into the transponders, which also introduces an offset in dispersion of about 400 ps/nm. In order to center the residual dispersion in this window, an extra CD compensation was added at the launch side to each transponder; for the 2040 Km link the additional CDC amounted to -400 ps/nm and to -900 ps/nm for the 3040 Km link.
The two 40Gbps transponders were at 1554.14 nm (CS-RZ) and 1555.75 nm (RZ-DPSK). 68x10Gbps transponders were also present; the resulting spectrum at the output of the 2560 Km link is shown in Figure 3 left. The launched power and distributed Raman gain have been optimized for this particular configuration to minimize the ripple: the 10Gbps channels power was set at -3.5 dBm; the measured ripple at the output of the 32nd (2560 km) amplifier was less than 8 dB. The 40Gbps channels powers were optimized individually to minimize the BER and both resulted in -2 dBm. The amount of signal fluctuation at the receiver side (fig. 3 right), due to PDL-PDG of the link, is 1.3 dB after 10 days; the corresponding OSNR fluctuation is less than 1 dB. In figure 4 left, a plot of BER vs. time is shown for the 40Gbps channels and the two 10Gbps channels interleaved with them (refer to fig. 3 right).
On the right, the Q values calculated from pre-FEC BER are plotted. Average Q of all 10Gbps channels is 14.7 dB; the corresponding average OSNR is 17.2 dB. The lowest Q was recorded for the channel at the beginning of band 1 (1609.63 nm) where the dispersion was not compensated accurately and the ripple was high. Measured Q is 13.8 dB and OSNR is 18.7 dB for the RZ-DPSK channel; Q is 10.8 dB and OSNR is 17.2 dB for the CS-RZ channel. The 10Gbps NRZ channels use a standard Reed-Solomon FEC [RS(255,239)] and thus need a minimum input Q of 11.5 dB for error-free (10-15 BER) operation; the E-FEC adopted in the 40Gbps channels needs a minimum Q of 9.5 dB for the same output BER.
| Previous Page: Introduction | Next Page: Trial Configurations and Results (continued) |