1. Introduction

Two fundamental limitations of long unrepeatered span submarine transmission systems are fiber attenuation and Stimulated Brillouin Scattering (SBS) nonlinear impairments [1]. The optical signal-to-noise ratio (OSNR) of the received signal is largely determined by the optical channel power at the end of the span entering the optical pre-amplifier. The received power is mainly determined by the span loss through fiber attenuation and the channel launch power. To achieve long single-span distances, system architects may resort to high launch powers, alternative modulation formats, remotely pumped amplifiers, Raman amplifiers at one or both ends of the span, and strong FEC [2–4]. However, it is desirable to simplify systems as much as possible to minimize cost, and this aim can be furthered by using fiber with lower attenuation [5] and launching channel powers as high as possible. The limiting factor for launch power level is often SBS, a single channel optical nonlinear impairment. Several different approaches have been demonstrated to increase the SBS threshold to allow higher channel launch powers including frequency dithering [6], phase modulation [7], heterogeneous fiber span construction [8,9], and the use of alternative modulation formats such as duobinary and DPSK [10,11]. All of these techniques and approaches can be quite effective but may add at least some complexity to the transmitter, receiver, or span design.

On the other hand, it has been shown that system performance can be improved via SBS suppression by taking advantage of the phase modulation that is induced by cross-phase modulation (XPM) in wavelength-division multiplexing (WDM) systems [12]. Another technique demonstrated to reduce SBS effects was to use XPM generated with a low bit rate supervisory channel [13]. In this work, we explore further the system performance improvements obtainable from XPM effects in high-power unrepeatered WDM systems at 10.7 Gb/s. We investigate simple long unrepeatered systems at 10.7 Gb/s using non-return-to-zero (NRZ) signals and amplified only with erbium-doped fiber amplifiers (EDFAs). We experimentally compare system performance for single channel transmission and WDM transmission with different channel counts and channel spacings. Finally, we also briefly look at the effects of SBS suppression from XPM in a system that already has an increased SBS threshold from a heterogeneous span design. We find that BER improvements of close to 5 orders of magnitude are possible in a WDM system compared to the same system with single channel transmission, and we demonstrate WDM transmission over 323 km with NRZ signals and EDFA-only amplification, with >5 dB system margin over the FEC threshold.

2. Experimental configuration

The general experimental system set-up used for all transmission system measurements is shown in Fig. 1. Up to 8 wavelengths from DFB semiconductor lasers centered around 1550 nm were modulated with the NRZ format with a Mach-Zehnder modulator (MZM) at a bit rate of 10.7 Gb/s. No frequency dithering or phase modulation was applied in the transmitter. The data stream was a pseudorandom bit sequence (PRBS) of length 2³¹-1. For some WDM transmission tests, a piece of standard single-mode fiber of length 10 km was placed after the modulator and prior to the launch amplifier to de-correlate the bit streams in the different WDM channels. This de-correlating fiber length was sufficient for the channel spacing values of 100 GHz and 200 GHz that were investigated. The channels were then amplified with a high output power EDFA with a maximum possible total power of 33 dBm. The total launch power was controlled with a variable optical attenuator (VOA) at the input to the transmission fiber span. The optical fiber used in all transmission experiments was Corning^® Vascade^® EX1000, a low-attenuation G.654-compliant fiber designed for unrepeatered submarine systems [5]. The measured average attenuation of the fiber under test was 0.165 dB/km. The dispersion at 1550 nm is about 18 ps/(nm-km). The SBS threshold of this fiber is nominally the same as that of standard single-mode fiber. Span lengths of 304 km and 323 km were studied here.

 

Fig. 1. Transmission system experimental set-up for unrepeatered system testing. PG: pattern generator, PM: power monitor, OSA: optical spectrum analyzer, Rx: receiver.

Download Full Size | PDF

At the output end of the span, the channel(s) passed through 3 EDFAs with 2 stages of optical dispersion compensation fiber (DCF) in between the amplifiers. The residual dispersion was optimized with the appropriate length of DCF to minimize the received signal BER. A red-pass filter was employed in the first DCF stage to eliminate excessive build-up of amplified spontaneous emission (ASE) in the blue end of the C-band spectrum. A tunable optical filter of bandwidth 0.25 nm selected a channel for detection in an optical receiver, which was a PIN photoreceiver with clock and data recovery. The recovered data and clock signals were passed to a bit error rate tester (BERT) for measurement of the signal BER.

Besides transmission experiments, measurements of the SBS-generated reflected Stokes power were also made to help understand the transmission results. The set-up is the commonly used configuration with a circulator and power monitor to measure reflected power as a function of launch power. The set-up is shown in Fig. 2 and was used for both a single channel and multiple channels.

 

Fig. 2. Experimental set-up used to measure reflected power from SBS non-linear impairment.

Download Full Size | PDF

3. Experimental results and discussion

We first look at transmission results obtained for a 304 km span of Vascade EX1000, for a single channel and 8 WDM channels with channel spacing of 100 GHz. The NRZ signals were undithered, as was true for all experiments. For the WDM experiments, all channels were modulated with the same MZM, so they were co-polarized at the modulator egress, but no polarization control was used anywhere else in the system. To create two different conditions with regard to the generation and magnitude of XPM amongst the WDM channels, we conducted experiments with and without the 10 km piece of bit sequence de-correlating fiber prior to the launch amplifier. The de-correlating fiber provided sufficient dispersion to shift in time each WDM channel from its neighbors by more than 1 bit period, effectively decorrelating each PRBS pattern from the others. It is also likely to randomize the polarization states of the WDM channels before the transmission span launch, although the actual states are dependent on the fiber and configuration. This creates more realistic conditions resembling field systems with WDM channels encoded with independent data patterns. On the other hand, removal of this de-correlating fiber means that all WDM channels are aligned in time as well as being co-polarized at the launch into the fiber span, providing a worst case scenario for the creation of non-linear impairments such as XPM. While this condition was created artificially, it was useful to understand the influence of XPM in SBS suppression and received signal quality. The results from three transmission experiments are shown in Fig. 3, comparing the measured BER vs. channel launch power for a single channel, and the two 8-channel systems with and without the de-correlation fiber. The BER values measured in the WDM experiments were for the same 1550.12 nm channel transmitted in the single channel experiment.

 

Fig. 3. Experimental results for transmission tests over 304 km fiber span. WDM systems had 8 channels with 100 GHz channel spacing.

Download Full Size | PDF

As observed in the figure, there is a dramatic decrease in the minimum BER for the measured channel in the WDM system configurations compared to the single channel system. Furthermore, the WDM system with channel bit sequences aligned at launch shows the best performance and in fact produced error-free transmission results at channel launch power levels greater than 14 dBm. The WDM system with de-correlated bit sequences at launch still shows a minimum BER value more than 4 orders of magnitude lower than the single channel. Since we know that the single channel system performance is limited by an SBS impairment, these results clearly show that the onset of XPM as the channel power increases in the WDM systems significantly suppresses the SBS impairment and improves overall signal quality.

To verify that the improved signal quality obtained in the WDM system configurations was due to SBS suppression, we measured the reflected power of the three different cases in the set-up shown in Fig. 2. The results of these measurements are shown in Fig. 4, in which the reflected power per channel is plotted against the launch power per channel. The launch power levels of the eight WDM channels were all equal, so these quantities were easily derived from the total reflected and launch powers measured in the experiment.

 

Fig. 4. Reflected power results as obtained from experimental set-up in Fig. 2.

Download Full Size | PDF

The data in Fig. 4 shows significant reduction of reflected power attributable to SBS in the WDM system configurations for channel launch power levels higher than approximately 10 dBm. Larger levels of XPM as channel power is increased clearly induce phase modulation effects that can significantly mitigate the SBS impairment in comparison to the single-channel case. In fact, for both WDM systems, we observe that the SBS-induced reflected power actually decreases at sufficiently high channel power levels above approximately 14 dBm. At a channel launch power level of about 15 dBm, the de-correlated WDM system shows a decrease in SBS-induced reflected power of about 8 dB compared to the single channel, while the difference between the other WDM system with initially aligned bit streams and the single channel is approximately 20 dB. This data helps to understand the differences in measured BER shown in Fig. 3.

For the 304 km span of Vascade EX1000 fiber, we also examined the difference in performance for WDM systems with different channel counts. In particular, we compared the performance of the 8-channel WDM system with a 4-channel WDM system, both with 100 GHz channel spacing and using the de-correlating fiber prior to launching into the span. Results from these two experiments, along with the single-channel results, are shown in Fig. 5, again in the form of measured BER for the 1550 nm channel as a function of channel launch power. The performance of the 4-channel WDM system at the optimal launch power is only marginally better than the single-channel system and substantially underperforms the 8-channel system. The level of SBS suppression induced by the 4-channel system is significantly smaller than that produced by the system with 8 WDM channels.

 

Fig. 5. Experimental results for transmission tests over 304 km fiber span, for a single-channel system, and WDM systems with 4 and 8 channels. WDM channel spacing was 100 GHz and channel bit sequences were de-correlated prior to launch.

Download Full Size | PDF

We next look at the transmission performance and compare the WDM system results for experiments with 100 GHz channel spacing and 200 GHz channel spacing. The same piece of de-correlating fiber was used for both cases (10 km of standard single-mode fiber). We performed experiments for both the 304 km span and a slightly longer span with length 323 km, both with the Vascade EX1000 fiber as the transmission medium. The WDM results and the corresponding single-channel results for both span lengths are shown in Fig. 6. For both span lengths, we observe that the WDM system performance as measured for the 1550.12 nm channel is significantly better with 100 GHz channel spacing than with 200 GHz channel spacing. This is attributable to the greater degree of XPM phase modulation generated with the tighter channel spacing that leads to greater SBS suppression. So for these systems, we can obtain significantly better system performance with 100 GHz WDM systems than for a more widely spaced channel configuration or for a single channel alone. For both span lengths, the 100 GHz system has optimal channel power of about 16 dBm, limited largely by XPM impairments and partly by SBS.

 

Fig. 6. Experimental results for transmission tests comparing single-channel with 8-channel 100 GHz and 200 GHz WDM systems. (a) 304 km span length, (b) 323 km span length.

Download Full Size | PDF

The channel Q values as calculated from the measured BER values for all 8 WDM channels are shown in Fig. 7 for the 323 km span. The performance is consistent among all 8 channels in both the 100 GHz and 200 GHz systems. The overall system Q margin over the FEC threshold is greater than 5 dB for the 100 GHz WDM system, using undithered NRZ modulation and only EDFA amplification.

 

Fig. 7. Experimentally measured Q values for 8 WDM channels in 100 GHz and 200 GHz systems over 323 km span.

Download Full Size | PDF

Finally, we also examined the ability of XPM induced in WDM systems to enhance overall system performance via SBS suppression in a system with an already intrinsically higher SBS threshold. An example of such a system is a heterogeneous fiber span comprised of a short piece of high-SBS threshold G.652 fiber [14] at the span entrance followed by a long piece of the Vascade EX1000 fiber, a span design which has been previously demonstrated to enable significantly higher SBS thresholds than a homogeneous span of the Vascade EX1000 fiber alone [9]. For the purposes of these experiments, we constructed a heterogeneous span with 20 km of the high-SBS threshold G.652-compliant fiber followed by 304 km of the Vascade EX1000 fiber, for a total span length of 324 km. Transmission experiments with a single channel and WDM systems with 100 GHz and 200 GHz channel spacing were performed with this composite fiber span to compare to the previous experiments. The results of the experiments are shown in Fig. 8.

Comparing the results of the higher SBS threshold fiber span in Fig. 8 with those from the same length (323 km) homogeneous span in Fig. 6(b), we observe several significant differences. First, for single channel transmission that is limited by SBS impairments, the optimal channel launch power is approximately 4 dB higher (16 dBm compared to 12 dBm) for the heterogeneous span. This is direct evidence of the ~4 dB higher SBS threshold of the heterogeneous span. We also note that the relative system performance between 100 GHz and 200 GHz WDM systems is opposite for the two fiber span types. While we found that the homogeneous 323 km span showed significantly better transmission performance with 100 GHz channel spacing, the higher SBS threshold heterogeneous span shows better system performance for a 200 GHz WDM system. This is because for 100 GHz channel spacing, the main limiting impairment for both spans is the XPM itself. Indeed, we see that the BER vs. launch power results for this channel spacing are nearly the same for the two fiber spans as the optimal channel launch power is primarily limited by XPM. For the heterogeneous span, there is only about a 1 dBm channel power difference between the single channel SBS limitation and the 100 GHz-XPM limitation, and this is why the 100 GHz system provides relatively small system benefit. However, since the heterogeneous span has a higher SBS threshold, the transmission can be improved by using a 200 GHz channel spacing because the XPM impairment is smaller in this case, although the XPM effect still provides a phase modulation that further mitigates the SBS impairment and allows an optimal launch power out to approximately 18 dBm or higher. On the other hand, we observed in Fig. 6(b) that a 200 GHz WDM system over the 323 km homogeneous span did not introduce sufficient XPM-induced phase modulation to overcome the SBS limitation to any significant degree.

 

Fig. 8. Experimental results for transmission tests comparing single-channel with 100 GHz and 200 GHz WDM systems over a 324 km heterogeneous fiber span comprised of 20 km of high-SBS threshold G.652-compliant fiber and 304 km of Vascade EX1000 fiber.

Download Full Size | PDF

4. Summary and conclusion

We have experimentally investigated the magnitude of XPM-induced SBS impairment mitigation in long unrepeatered spans at 10.7 Gb/s, using undithered NRZ signals and EDFA-only amplification schemes. We found that for homogeneous spans of a low-attenuation G.654-compliant fiber, 100 GHz 8-channel WDM systems can provide more than 4 orders of magnitude BER improvement in comparison to a single-channel system. We showed this was directly attributable to SBS mitigation with reflected power measurements indicating significant Stokes power reduction from the XPM phase modulation effect. We demonstrated WDM transmission over 323 km using the simple NRZ/EDFA system configuration with more than 5 dB Q margin over the FEC threshold. We also explored the system behavior with 4 WDM channels, and then for 8-channel WDM systems with 100 GHz and 200 GHz channel spacing. For the homogeneous span configuration, the 8-channel 100 GHz system had the best system performance at both 304 km and 323 km span lengths due to the significant SBS mitigation benefit conferred by the XPM. Finally, we also investigated a heterogeneous span configuration with an intrinsically higher SBS threshold by about 4 dB. In this case, better performance was observed with the 200 GHz WDM system since the XPM limiting channel power was higher than allowed for 100 GHz spacing, while the XPM-induced phase modulation effect still provided relief from the SBS impairment and allowed the higher channel power to be launched.