Xiang Zhou 1 , Eduardo F. Mateo 2 and Guifang Li 2 - AT&T

161kB Size 5 Downloads 6 Views

OCIS codes: (060.1660) Coherent communications; (060.2330) Fiber optics .... For example, to add 40Gb/s or 100G/s PM-QPSK into the legacy 10G/s platform ...
Fiber Nonlinearity Management – from Carrier Perspective

Xiang Zhou1, Eduardo F. Mateo 2and Guifang Li2

1) AT&T Labs-Research, 200 Laurel Ave South, Middletown, NJ 07748, USA, [email protected] 2) CREOL, The College of Optics and Photonics, University of Central Florida; 4000 Central Florida Blvd, Orlando FL, 32816 (USA).

Abstract: Technologies toward fiber nonlinearity management for spectrally-efficient high-speed optical transmission systems have been discussed from carrier perspective. While optimal electrical dispersion management remains an effective method, new ultra-low- nonlinearity fiber might be needed in the future. ©2011 Optical Society of America OCIS codes: (060.1660) Coherent communications; (060.2330) Fiber optics communications.

Introduction The demand for higher capacity and the advent of digital coherent detection technology have spurred worldwide industry interest in various multi-level coherent modulation formats in order to achieve more spectrally efficient optical transmission at higher per channel data rate (100Gb/s and beyond). While these multilevel coherent modulation formats bring us improved spectral efficiency (SE), generally their tolerance toward both noise and fiber nonlinear impairments decreases because the Euclidean distance decreases [1]. To extend the transmission reach, aggressive management of noise and fiber nonlinear impairments thus becomes critically important to make these spectrally efficient high-speed systems really cost- effective. In this paper we review recent progress as well as the faced challenges on fiber nonlinearity management from carrier perspective. We show that DSP enabled electrical dispersion compensation not only bring us significant operational advantage, it also improves the system’s nonlinear tolerance as compared to the traditional method using inline optical dispersion compensation. On the other hand, although various digital nonlinear compensation methods have been proposed, their performance in the realistic territorial long-haul (LH) WDM system requiring dynamic wavelength routing is still quite limited. To realize ultra-high SE transmission, new ultra-low-nonlinearity fiber might be needed in the future.

Dispersion management Through optimal dispersion management to increase the system’s tolerance toward fiber nonlinearity is well known and has been applied in both 10Gb/s and 40Gb/s commercial WDM system. For 10Gb/s OOK system, a small amount of pre-compensation is typically introduced to lower the in- fiber peak power while the intra-channel pulse overlapping is minimized by introducing periodically inline optical dispersion compensation. Because 100% inline optical dispersion compensation significantly increases inter- channel nonlinear effects, an inline dispersion compensation ratio slightly lower than 100% (around 95%) usually gives the best overall nonlinear tolerance (against both the intra- and inter-channel nonlinear effects). This dispersion map also works for 40Gb/s systems using single-polarization based modulation formats such as OOK, duobinary as well as DPSK, although the optimal pre-dispersion compensation ratio is different. For 40 and 100Gb/s systems using coherent polarization-multiplexed (PM) QPSK, however, using inline optical dispersion compensation does not help in reducing both intra- and inter-channel nonlinear effects [2-4]. On the contrary, it is found that using full electrical dispersion compensation can achieve better nonlinear tolerance. The reason is partly due to the fact that using periodically inline optical dispersion compensation will enhance impairment caused by inter-channel nonlinear cross-polarization modulation (XPolM) [5,6] that cause negligible penalty in the single-polarization based 10Gb/s and 40Gb/s systems. As a numerical example, in Fig. 1a and b we show simulated results on the impact of different dispersion management methods on the nonlinear tolerance of PM-QPSK modulation format. In Fig. 1a we study the effectiveness of inline optical dispersion compensation for both 42.8Gb/s and 112Gb/s PM-QPSK WDM system (50GHz-spaced, 7 channels, 16(100km reach, EDFA-only) using standard single-mode fiber (SSMF). To isolate major fiber nonlinear effects no ASE noise is added in this simulation and the maximum phase deviation normalized to the ideal decision threshold is used to measure the nonlinear impairment. For this study no pre-transmission dispersion compensation is used and all the residual dispersion is compensate at the digital coherent receiver. From Fig. 1a one can see that the performance continues to improve as the amount of inline optical dispersion compensation decreases and the best performance is achieved without using inline optical dispersion compensation. Note that removing inline optical dispersion compensation not only improves the nonlinear tolerance, it also brings several additional advantages such as reduced transmission latency, simplified inline optical amplifier (two stages to one stage) as well as the resulted noise performance improvement. So using full electrical dispersion compensation is very attractive from carrier perspective. Because the amount of DSP processing required for full electrical dispersion compensation is significant for 100Gb/s and beyond system [7], it may be very challenging to implement full receiver-side electrical dispersion compensation for some ultra-LH systems due to the available CMOS capability. One effective method is to split dispersion compensation into both the transmitter and the receiver. In fact we have found that using both pre- and post-transmission dispersion compensation gives slightly better nonlinear tolerance than using only post-transmission dispersion compensation when the system is operating in the strong nonlinear region. We have found that there exists multiple optimal pre- compensation values and pre-compensating about half of the total dispersion is among the optimal points. In Fig. 1b we show the simulated maximum phase deviation versus launch power for a 112Gb/s PM-QPSK WDM system (1600km, SSMF, EDFA only) using three different dispersion management methods, the traditional method with a pre-compensation ratio=0.38 and inline compensation ratio=0.95, the recently introduced full receiver-side electrical dispersion compensation method and the proposed symmetric pre- and post-transmission dispersion compensation method. The performance advantage of using pure electrical dispersion compensation methods are clearly seen from this figure. [pic] [pic] a) (b) Fig.1. Simulated results on the effectiveness of different dispersion management methods toward fiber nonlinear tolerance for PM-QPSK system

Digital compensation Because digital coherent detection allows linear mapping of optical field to the electrical field and fiber nonlinear effects due to signal-to- signal nonlinear interaction are deterministic processes, using advanced digital signal processing to compensate or mitigate fiber nonlinear effects thus becomes feasible. If all the WDM signals can be detected at the receiver, theoretically all the impairments caused by signal-signal nonlinear interaction can be effectively compensated by using digital backward propagation (DBP) [8-10]. For this ideal case, the ultimate nonlinear tolerance is only limited by the signal-noise nonlinear interaction that is found to be much smaller than signal-to-signal nonlinear effects for 100Gb/s and beyond system [6]. But such a condition generally cannot be met in a realistic terrestrial WDM network requiring dynamic wavelength routing capability. For such a dynamic photonic network, the wavelengths originating from the same terminal may go to different locations and therefore it is highly unlikely to receive all the wavelengths having nonlinear interaction during transmission at the receiver. For such networks DBP can only be used to compensate intra- channel nonlinear effect, i.e. the self-phase modulation (SPM). But our recent numerical investigations have revealed that, the DBP method is not very effective in compensating intra-channel nonlinear effects in a WDM environment [11]. The reason is because DBP is very sensitive to the initial condition (i.e. the waveform). For WDM transmission, even if the intra-channel nonlinear effects is dominant over inter-channel nonlinear effects, the inter-channel nonlinear effects will modify the pulse evolution process through dispersion during transmission and thus result in different waveform at the receiver (as compared to the single channel transmission). Although such a waveform difference may not result in big BER difference, its impact on the DBP can be big as is manifested in Fig. 2a, where we show a simulated result on the efficiency of DBP method for both 1( 800Gbs and 3(800Gbs 200GHz-spaced PM-16QAM orthogonal frequency division multiplexing (OFDM) systems (10(100km, SSMF, EDFA only). Here DBP1 represents dispersion compensation only while DBP2 denotes dispersion plus SPM compensation. For the case using only dispersion compensation, one can see that the Q performance is very close for the single channel system and the WDM system, implying the dominance of intra-channel nonlinear effects over inter-channel nonlinear effects. For the single channel system, SPM compensation can improve the Q factor by more than 8dB, but only 2 dB performance improvement is observed for the WDM system. Under identical spectral efficiency of 4b/s/Hz, the efficiency of DBP2 is even poorer for lower per channel data rate as can be seen in Fig. 2b and c, where we show the results for 100GHz-spaced 6(400Gbs and 50GHz-spaced 12(200Gbs PM-16QAM- OFDM system, respectively. With 50GHz-spaced 12(200Gbs PM-16QAM-OFDM, the Q improvement achieved by SPM compensation is very small. Although OFDM is used for this study, similar results can be expected for single carrier based modulation formats due to the same nature of DBP method as is reported in [12]. Because the required amount of DSP process for DBP based nonlinear compensation is very intense (about one order of magnitude more than dispersion-only compensation even with coarse step method [13]) and the achievable performance improvement is quite limited, the usefulness of this technology in the territorial LH network is quite questionable. But using simpler digital methods to address fiber nonlinear issues under some special conditions can be very useful in the real systems. For example, to add 40Gb/s or 100G/s PM-QPSK into the legacy 10G/s platform with installed inline optical dispersion compensation, the penalty caused by cross-phase- modulation (XPM) from the neighboring OOK channels can be partly mitigated by using advanced digital phase recovery algorithm [14], and the enhanced XPolM may also be mitigated by using advanced DSP algorithms [15,16]. |[pic] |[pic] |[pic] |

a) (b) (c) Figure 2: Q-factors versus channel powers for each WDM system and BP scheme. Q-factors are the result of averaging over WDM channels and polarization tributaries. BP1 represents dispersion compensation, BP2 includes SPM compensation, BP3 includes XPM compensation and BP4 includes XPolM. Finally, BP5 represents full impairment compensation including FWM.

Ultra-low-nonlinearity fiber Another more long-term method to address the nonlinear issue is to use large area fiber. Significant progress has been made in this direction. For example, effective area as large as 150(m2 has been reported for ultra- long-haul transmission [17]. To further increase the effective area with single mode operation may be difficult by using the traditional design method because the bending loss increases as the core diameter goes up. Recently a new concept of few-mode fiber has been proposed [18]. The basic idea of this method is to design a fiber in such a way that fiber itself can support multiple modes but the modal coupling is made (by special control of the index profile) to be less likely to happen in the normal operational environment within one span. Thus the fiber will behave like a single-mode fiber if only one mode is excited at the transmitter. Between spans, periodic spatial mode filtering can be realized by splicing a section of SSMF. This method may have the potential to further increase the effective area but more researches are required to understand if it is possible to realize this potential, especially the tolerance toward bending loss. Apart from increasing the effective area, recently it has been found that increasing fiber dispersion also improves the fiber nonlinear tolerance [3]. But increasing dispersion also increases the DSP complexity, a trade-off between DSP complexity and system performance has to be considered for the design of next generation of ultra-low nonlinearity fiber. Conclusion Recent technological advancement in fiber nonlinearity management has been discussed. It is shown that full electrical dispersion compensation gives better nonlinear tolerance than the traditional methods using inline optical dispersion compensation for the new generation of coherent systems. Unless under some special conditions, the much-expected digital nonlinear compensation techniques can only provide very limited performance gain in the realistic terrestrial network. Thus new ultra-low-nonlinearity fiber might be needed in the future References

[1] X. Zhou et al, JLT, 2009, pp. 3641-3653 [2] A. Carena1 et al, OFC/NFOEC’08, . JThA57. [3] Vittorio Currim et al, PTL, 2010, pp.1446 [4] D.V. D. Borne, et al, ECOC’09, Paper 3.4.1 [5] C.Xie, OFC’10, paper OWE1 [6] A. Bononi et al, ECOC’10, paper Th.10.E.1. [7] B. Spinnler, IEEE/ LEOS summer topical, pp. 227-228. [8] X. Li et al., Opt. Express, 16, 880 (2008). [9] G. Goldfarb, et al., Photon. Technol. Lett. 20, 1887 (2008). [10] F. Yaman, et al., Photonics Journal, IEEE, 1,144 (2009). [11] E. Mateo et al, accepted for IEEE J. Quatum electronics. [12] Seb J. Savory et a, PTL 2010, pp.673-675. [13] E. Ip et al, J. Lightw. Technol., 2008, pp.3416–3425 [14] M. Kuschnerov et al, JLT 2009, pp. 3614-3622. [15] Lei Li et al, OFC2010, paper OWE3 [16] Y. Cai et al, OFC2010, paper OTuE1 [17] J. X. Cai et al, OFC2010, PDPB10 [18] F. Yaman et al, Optics Express 2010, No.12, pp. 13250