1. Introduction

System designs that can take advantage of longer amplifier spans may offer cost savings through reduced numbers of amplifiers and amplifier huts in the field. However, in order to maintain this cost advantage it is necessary to still attain sufficient reach lengths such that more frequent signal regeneration is not necessary compared to conventional 80-100 km span systems. One approach that has been shown to enable longer span lengths by increasing the optical signal-to-noise ratio (OSNR) is the use of hybrid Raman/EDFA amplification [1,2]. Another means to increasing system OSNR is through the choice, and span design, of the optical transmission fiber. In particular, higher system OSNR can be enabled by either lowering fiber loss or allowing higher channel launch powers through reduced fiber nonlinearity, or both. In principle, while either fiber characteristic can be effective for increasing OSNR, the combination of both is ideal, and therefore it is desirable to lower the fiber attenuation as much as possible while making the fiber effective area as large as possible [3–5].

Recent reports have demonstrated 43 Gb/s transmission systems with long amplifier spans that were based on hybrid Raman/EDFA amplification and ultra-low loss G.652-compliant fiber [6,7]. In [6], 1000 km transmission was shown with 200 km spans and backward pumped Raman amplifiers. In [7], transmission over 1000 km was shown with 250 km spans that were both forward and backward Raman pumped. In both cases, the optical fiber used was ultra-low loss with intrinsic fiber attenuation < 0.17 dB/km but with nominal effective area compliant with the G.652 standard. For both of these systems, 40 channels were transmitted with 100 GHz channel spacing.

In this paper, we extend the previous work by demonstrating ultra-long haul (ULH) transmission of 112 Gb/s polarization-multiplexed quadrature phase-shift keying (PM-QPSK) signals over a system with 200 km spans and hybrid Raman/EDFA amplification with backward Raman pumping. We employ a simple hybrid fiber span configuration that combines an optical fiber with both ultra-low loss (0.162 dB/km) and very large effective area (134 µm²) with an ultra-low loss G.652-compliant fiber, in equal lengths of 100 km each. The 50/50 span design facilitates increased OSNR compared to a single fiber system while keeping the configuration simple. We demonstrate 6000 km transmission for 8 channels and 5400 km transmission for 32 channels, each with 50 GHz channel spacing.

2. Experimental set-up

The basic experimental configuration is shown in Fig. 1 . Experiments were performed first with 8 channels and then with 32 channels, with both systems on a 50 GHz grid. Odd and even DFB lasers with nominal linewidth of a few MHz were combined and then modulated together with a QPSK modulator driven by two de-correlated 2¹⁵-1 PRBS patterns with a symbol rate of 28 Gbaud. The output from the QPSK modulator was optically polarization multiplexed to produce the 112 Gb/s PM-QPSK signals. The channels were launched into a re-circulating loop with a nominally flat spectrum. An EDFA at the beginning of the loop in power control mode determined the launch power into the first span. The loop was comprised of 3 spans of length 200 km. Each span was constructed from 100 km of a prototype ultra-low loss very large effective area fiber spliced to 100 km of Corning^® SMF-28^® ULL fiber, an ultra-low loss G.652-compliant fiber, as illustrated at the bottom of Fig. 1. The average effective area of the first fiber was 134 µm², allowing higher channel launch power, while that of the second fiber was 85 µm². The average total loss of each 200 km fiber span including splices and connectors was 33 dB (0.165 dB/km). The span loss was compensated with backward pumped Raman and single-stage Erbium doped amplifiers. The hybrid fiber span design promotes higher system OSNR than could be achieved by the use of a homogeneous span using either individual fiber. This is because the very large effective area fiber at the front end allows larger launch power into the span, while the smaller effective area fiber at the back end allows significantly higher Raman gain, with no real trade-offs to be made since both fibers are ultra-low loss. The maximum possible total launch power into each span was 19 dBm, as limited by the EDFAs. A loop synchronous polarization scrambler (LSPS) was used to mitigate possible loop polarization artifacts. For the 8 channel system, we placed a 6 nm wide bandpass filter centered around the channels after the polarization scrambler at the end of the loop to filter ASE noise and flatten the channel spectrum. For the 32 channel experiments, the bandpass filter was replaced with a dynamic gain equalizer (DGE). All chromatic dispersion was compensated digitally in the digital coherent receiver.

 

Fig. 1 Experimental set-up of transmission system with re-circulating loop and illustration of span configuration.

Download Full Size | PDF

At the receiver end, a tunable optical filter selected a channel for measurement. The measurement channel was detected in a polarization- and phase-diverse digital coherent receiver with a free-running local oscillator with 100 kHz nominal linewidth. The four signals from the balanced photodetectors were digitized by analog-to-digital converters operating at 50 Gsamples/s using a real-time sampling oscilloscope with 20 GHz electrical bandwidth. The sampled waveforms were processed off-line in a computer, with digital signal processing steps including (i) quadrature imbalance compensation [8], (ii) up-sampling to 56 Gsamples/s and dispersion compensation using a frequency-domain equalizer, (iii) digital square and filter clock recovery [9], (iv) polarization recovery and equalization using an adaptive butterfly structure with filter coefficients determined using the constant modulus algorithm [10,11], (v) carrier frequency offset using a spectral domain algorithm [12], (vi) phase recovery using a pre-decision algorithm [13], and (vii) bit decisions. The bit error rate (BER) was measured for each 28 Gb/s tributary signal by direct error counting.

3. Experimental results

Initial transmission system experiments were conducted with 8 DWDM channels ranging from 1546.92 nm to 1549.72 nm. We first alternately varied Raman pump power and thus Raman gain, and launch power per channel to determine appropriate settings for each. Results obtained for average Raman gain vs. total Raman pump power are shown in Fig. 2(a) . The Raman gain data in this figure were inferred from the total span losses and the reported EDFA gains. In the figure, the data points for pump powers ≤ 728 mW correspond to the use of 2 pump wavelengths at 1443 nm and 1461 nm, while for higher total pump powers, additional pumps at 1427 nm were added. The measured Q value for the 1548.51 nm channel after 3600 km transmission as a function of average Raman gain is given in Fig. 2(b), showing that the optimal level of Raman gain is approximately between 20 and 25 dB. The rollover in Q for increasing Raman gain may partly be limited by double Rayleigh backscattering effects [14], but is also likely attributable to a corresponding increase in the noise figure of the EDFAs for smaller gains (as Raman gain increases). Based on this data, we chose to use the 2 pump wavelength configuration with total average power of 728 mW and average gain of 20.8 dB for the 8 channel system.

 

Fig. 2 8 channel transmission results. (a) Average Raman gain as a function of total Raman pump power, (b) Q vs. average Raman gain for 1548.51 nm channel.

Download Full Size | PDF

The optimal channel launch power was determined for the 8 channel system by varying the total power into each span at 2 different transmission distances: 3000 km and 3600 km. The measured BER results are given in Fig. 3(a) –they show that the optimal launch power was about 4.5 dBm per channel at both distances. The data in Fig. 3(b) shows the measured BER values as a function of the corresponding measured OSNR values compared to data taken in a back-to-back condition. This figure shows the transmission system data clearly approaching the back-to-back results in the linear regime, suggesting that any double Rayleigh backscattering penalty is fairly small.

 

Fig. 3 8 channel transmission results for 1548.51 nm channel. (a) BER vs. channel power, (b) BER vs. OSNR for transmission over 3000 km and 3600 km systems, and back-to-back.

Download Full Size | PDF

Setting the total launch power into each span at 13.5 dBm (4. 5 dBm per channel), we next measured the BER and OSNR of the central 1548.51 nm channel as a function of distance. The results are shown in Fig. 4(a) in which the measured BER has been converted into 20log(Q) values. The results show that the 8 channel system has a maximum reach of 6000 km defined by the assumed FEC threshold of 8.5 dBQ. All 8 channels were measured at the 6000 km distance verifying Q performance above the threshold, as shown in Fig. 4(b).

 

Fig. 4 8 channel transmission results. (a) Q and OSNR vs. transmission distance for the 1548.51 nm channel, (b) Q and OSNR for all 8 channels at 6000 km.

Download Full Size | PDF

For the 32 channel transmission system, the channel plan ranged from 1542.14 nm to 1554.53 nm, again with 50 GHz channel spacing. We added the third Raman pump wavelength at 1427 nm to help broaden the Raman gain spectrum. For this system, the total average Raman pump power used was about 870 mW, providing an average measured ON/OFF gain across the channel plan and over all 3 spans of about 21.6 dB. The Raman gain ripple across the channel spectrum was about 1 dB. The average Raman gain spectrum is shown in Fig. 5 . For the 32 channel system, a dynamic gain equalizer (Finisar WaveShaper) was substituted for the passive filter of the 8 channel system. The DGE equalized the overall gain shape at the end of each 600 km loop circulation and allowed a flat spectrum to be obtained after long transmission distances.

 

Fig. 5 Average Raman gain profile for 32 channel system.

Download Full Size | PDF

Measurements of BER of the central channel at 1548.91 nm in the 32 channel system were performed as a function of channel power. Given the maximum available EDFA output power of 19 dBm, this permitted a maximum channel power of 4 dBm into each span. The results of the measurements given in Fig. 6(a) suggest that 4 dBm may be slightly sub-optimal, which agrees with the 8 channel results in which we found an optimal channel power of 4.5 dBm. Setting the channel power at the maximum 4 dBm, we then measured the Q value of the central channel as a function of distance. The results are given in Fig. 6(b) and compared with the previous 8 channel results. The 32 and 8 channel results agree well, with the very small difference likely being primarily due to the 0.5 dB lower channel power for the 32 channel system.

 

Fig. 6 32 channel transmission results for central channel. (a) BER vs. channel power,(b) Q vs. distance.

Download Full Size | PDF

The maximum reach of the 32 channel system indicated in Fig. 6(b) was 5400 km. We measured all 32 channels at this distance. The Q and OSNR results are given in Fig. 7(a) , verifying that all channels have Q values exceeding the FEC threshold. The DWDM spectrum is given in Fig. 7(b) after 5400 km transmission.

 

Fig. 7 32 channel system results after 5400 km transmission. (a) OSNR and Q vs. channel wavelength, (b) received spectrum.

Download Full Size | PDF

4. Summary and conclusions

We have experimentally demonstrated ultra-long haul 112 Gb/s PM-QPSK transmission over links designed with 200 km hybrid fiber spans. The 2 fiber types used in each span both have ultra-low loss and the leading fiber in each span has very large effective area. The span design maximizes overall system OSNR by allowing both higher launch powers and high Raman gain. We achieved 6000 km reach with an 8 channel system and 5400 km reach with a 32 channel system. The maximum channel count demonstrated was limited only by the EDFA total output power and operation at or near the optimal channel power level.