Density and Guard Band in Migration Scenarios to Coherent Ultra-Dense WDM

Density and Guard Band in Migration Scenarios to Coherent Ultra-Dense WDM Jacklyn D. Reis jacklyn@ua.pt Darlene M. Neves darlene@ua.pt António L. Teixeira Nokia Siemens Networks teixeira@ua.pt Abstract In this work the maximum density is numerically investigated giving the compromise of physical impairments and spectral efficiency on coherent ultra-dense wavelength division multiplexing passive optical networks (UDWDM-PON). The required guard band to legacy 1.25-2.5 Gb/s intensity modulation direct detection (IMDD) systems specified closely to Gigabitcapable PON equipments is also analyzed in order to mitigate linear and nonlinear crosstalk. Using 3 GHz frequency grid guaranteed the best performance of the system (32x1.25 Gb/s quadrature phase-shift keying with coherent detection) with respect to the 10% Error Vector Magnitude limit; 12.5 GHz guard band to legacy IMDD system was sufficient to minimize strong linear and nonlinear crosstalk (mostly four-wave mixing) on both systems. Keywords-passive optical networks; ultra-dense WDM; coherent optical communcations; advanced modulation formats. I. INTRODUCTION The path to on-going smooth migration in existing populated wavelength spectrum by different Passive Optical Networks (PON) technologies such as Video overlay, G-PON (Gigabit), E-PON (Ethernet), XG-PON (10Gigabit) [1,2] could pass through adoption of Ultra-Dense Wavelength Division Multiplexing (UDWDM) technologies. Those systems seek to deliver, up to 1000 wavelengths per subscribers separated by a few Gigahertz broadband connections at 1-10Gb/s through a single optical fiber [3]. They rely on employing high-order modulation formats such as M-ary PSK (phase-shift keying) or M-ary QAM (quadrature amplitude modulation) along with coherent detection to provide high receiver sensitivity and wavelength selectivity [4]. As such, the system has improved spectral efficiency and transmission capabilities as to comply with network coverage and capacity, upgradeability bandwidth lack and user density requirements of broadband optical networks. For instance, using quadrature phase-shift keying (QPSK) over UDWDM provides three key functionalities: (i) high wavelength selectivity alleviates the need for costly ultradense filter technology; (ii) better receiver sensitivity enables a system with higher splitting ratios and extended reach (power budget); (iii) digital signal processing eases the implementation of transmission impairments equalization as well as forward error correction (FEC) [5-8]. UDWDM requires that several channels spaced at only a few GHz are transmitted through tens of km of a single optical fiber. Therefore, the occurrence of very strong fiber nonlinear crosstalk such as inter-channel four-wave mixing (FWM) and cross-phase modulation (XPM) may impair the overall system performance if the transmitter parameters are not properly optimized [9,10]. Recently, we have shown, by recurring to Volterra series, that inter-channel FWM is, among the compared, the most effective nonlinear degradation occurring on UDWDM systems (32 1.25 Gb/s-QPSK) with very small channel spacing (3 GHz), even when standard single-mode fiber (SSMF) is employed [11,12]. In addition, transmitting 32 QPSK channels spaced by 3 GHz provides maximum FWM impact, indicating that the number of channels/subscribers can be further increased to as high as one thousand with total data rate of Terabit/s per fiber [13]. Employing only phase-modulated channels impose minimal XPM impact. The same is not true when those signals copropagate with intensity modulated direct detection (IMDD) channels [14]. Thus a frequency guard band is required in order to avoid strong interference induced by IMDD channels through XPM. Therefore, the deployment of coherent ultradense WDM-PON systems should take into account the following aspects: (i) using the minimum channel spacing to achieve the best trade off between spectral efficiency and performance limited by fiber nonlinearities; (ii) finding the minimal guard band to already deployed G/E-PON systems in order to maximize the overall performance of both coherent UDWDM-PON and legacy systems. This paper is organized as follows: we first investigate some of the limits on the system performance of coherent UDWDM-PON technologies regarding different channel spacing that provides the best compromise of spectral efficiency and overall system performance. Secondly, when the coherent UDWDM-PON coexists with legacy technologies, the required guard band is estimated in order to mitigate the impact of fiber nonlinear and linear crosstalk on both systems. II. COHERENT ULTRA-DENSE WDM-PON The UDWDM-PON scenario, depicted in Fig. 1 is divided into 2 different architectures: (i) coherent (Fig. 1.(a)) aims identifying the best channel spacing (density); (ii) hybrid (Fig. 1.(b)) aims analyzing the required guard band to the IMDD channel. This work was partially supported by the European Union within the EURO- FOS project, a Network of Excellence funded by the EU 7th ICT-Framework Programme. J.D. Reis and D.M. Neves also acknowledge their PhD grant from FCT - Fundação para a Ciência e a Tecnologia. 978-1-4244-9268-8/11/$26.00 2011 IEEE

The coherent QPSK channels are then coupled with the IMDD channel (15 db of extinction ratio) that is filtered at transmitter and receiver sides using an optical filter with 5 GHz-bandwidth. Figure 1. UDWDM-PON scenarios. (a) Coherent: 32 QPSK channels with different densities; (b) Hybrid: 31 QPSK channels + 1 IMDD with different guard band. SSMF: standard single-mode fiber; LPF: low-pass filter; ADC: analog-to-digital converter; DSP: digital signal processing. A. Coherent system In the coherent scenario (Fig. 1.(a)) the transmission model specifications are outlined as follows: the 32 continuous-wave (CW) laser sources (without phase noise) are phase modulated using an IQ (in phase quadrature) modulator fed with two 625 Mb/s-NRZ (non-return to zero) electrical signals with 2 9 bitlong sequence shaped by a 5 th order Bessel filter with electrical bandwidth of 1.25 GHz. As such, the optical QPSK signal operates at 625 Mbaud (1.25 Gb/s). The resulting independently encoded QPSK channels are then multiplexed via an ideal lossless AWG (arrayed waveguide grating) modeled as a 2 nd order super-gaussian filter with 3 dbbandwidth equal to 2.5 GHz. B. Hybrid system In the hybrid scenario, an IMDD channel replaces the 32 nd QPSK channel. In this case, the CW laser source is intensity modulated through a Mach-Zehnder modulator using NRZ electrical signals with 2 10 and 2 11 bit-long sequences corresponding to 1.25 Gb/s and 2.5 Gb/s, respectively. The resulting extinction ratio was kept constant around 15 db. This channel follows closely the specs for GPON transmitter maximum and minimum average powers for all categories [15]: A (-3dBm/2dBm@1.25 Gb/s; 0dBm/4dBm@2.5 Gb/s); B (-2dBm/3dBm@1.25 Gb/s; 5dBm/9dBm@2.5 Gb/s); C (2dBm/7dBm@1.25 Gb/s; 3dBm/7dBm@2.5 Gb/s). C. Transmission and reception The total optical signal is transmitted over 25 km, 60 km and 100 km of SSMF in the coherent scenario with the following physical parameters: the reference optical frequency positioned between channels 16 th and 17 th is 193.4 THz (1550 nm); fiber attenuation = 0.20 db/km; chromatic dispersion = 16.5 ps/(nm.km); dispersion slope = 0.07 ps 2 /nm.km; nonlinear parameter = 1.35 (W.km) -1. The WDM optical signal propagation through the fiber is modeled using the symmetric version of the Split-Step Fourier method with very high temporal and spatial resolution. At the coherent receiver depicted in the inset of Fig. 1.(a) intradyne detection [16] is performed in which the optical phase is recovered digitally and the local oscillators are tuned to the corresponding WDM channel as follows: QPSK channels under test in the Coherent scenario: 1 st, 4 th, 8 th, 16 th, 17 th, 20 th, 24 th, 28 th, 32 nd ; QPSK channels under test in the Hybrid scenario: 1 st, 4 th, 8 th, 16 th, 17 th, 20 th, 24 th, 28 th, 31 st. Firstly, the optical signal from the 1:32 splitter is mixed along with the 0 dbm-local oscillator (without phase noise) through 2x4-90º optical hybrid. Secondly, the resulting optical signal is optical-to-electrical down-converted by 2 balanced photo detectors. Thirdly, the electrical signal is filtered using a 5 th order low-pass Bessel filter with 3dB-bandwidth equal to 0.7 baudrate. Then, the analog signal is down-sampled at symbol rate (625 MSamples/s or one sample per symbol) considering an analog-to-digital converter with 8 bits resolution to avoid any quantization error. Thus, the recovered symbols are normalized to 1 (average constellation energy) and the phase sync is performed as follows: 6.25% of transmitted symbols (32/512) are assigned as pilot symbols with phases φ i ; The pilot symbols with φ i are known at receiver side; The phase difference of the received symbols φ i and pilot symbols φ i at each time instant is calculated as Δφ i =φ i -φ i ; The expected value of the phase difference is estimated over the 32 instants of time: Δφ = E{Δφ i }; Δφ is then feed back to the actual symbol as exp(j(φ i - Δφ)) being j = -1. It is worth emphasizing that digital chromatic dispersion compensation was not performed since the temporal effect of the total accumulated dispersion has negligible effect on 625/1250 Mbaud signals in typical access network links with fiber lengths up to 100 km. To measure the system performance we calculated the root mean squared values (RMS) of the Error Vector Magnitude (EVM) between the received symbols s i =a i +jb i and the ideal transmitted symbols

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings. si=ai+jbi as: EVMrms(%) = ( si-s i 2/ si 2) 100. In each transmission, EVM is estimated over 512 transmitted symbols per QPSK channel. The resulting EVM is averaged over 32 transmissions with a total of 16384 simulated QPSK symbols or 32768 bits. The IMDD channel was analyzed in terms of Q factor evaluated in the optimal sampling time of the received electrical eye diagram. In this case, 32768 bits were used for Q factor estimation at 1.25 Gb/s whereas 65536 bits were at 2.5 Gb/s. III. RESULTS AND DISCUSSION The results in Fig. 2.(a) outline the performance of the coherent scenario in terms of the worst EVM among all channels under test as a function of the channel spacing for input powers per channel (average optical power) -6 dbm and 3 dbm in blue solid line and red dash line, respectively. We evaluated the performance after transmission over 25 km (circles), 60 km (squares) and 100 km (diamonds) of fiber. We point out that both fiber spans have very similar performance for channel spacing higher than 2 GHz indicating that coherent UDWDM-PON can have extended reach as high as 100 km Figure 2. (a) Maximum EVM versus channel spacing: blue solid line 6dBm per WDM channel; red dash line -3dBm per WDM channel. Insets show received constellations; (b) EVM among tested channels at 3.125 GHz. with splitting ratios as high as 1:32. The system spectral efficiency, which is measured in terms of the aggregated bit rate (32 1.25 Gb/s=40 Gb/s) divided by the total occupied bandwidth (number of channels times channel spacing) decreases as the channel spacing increases. Therefore, lower channel spacing gives better spectral efficiency, e.g. at 1.5625 GHz channel spacing gives spectral efficiency of 0.8 b/s/hz. On the other hand, operating at 1.5625 GHz the system is severally impaired by inter-channel FWM crosstalk resulting EVM around 15 % even for -6 dbm per channel case. The solution is operating at channel spacing higher than 3.125 GHz (spectral efficiency=0.40 b/s/hz) that assures EVM bellow the 10 % limit [17]. This limit corresponds to a system BER<10-9 if the signal is affected by Gaussian amplitude and phase noises. Note that EVM around 31 % corresponds to a system BER=10-3 (typical FEC limit) [18]. The insets in Fig. 2.(a) show the received constellation whereas Fig. 2.(b) highlights the EVM distribution among all tested channels at 3.125 GHz frequency grid. After identifying the best density for the coherent scenario, Fig. 3.(a) shows the performance of the hybrid scenario, after transmission through 25 km of fiber in terms of EVM at the 31st QPSK channel as a function of the guard band to the 32nd IMDD channel operated at 1.25 Gb/s. The inset in Fig. 3.(a) Figure 3. EVM of the received 31st QPSK versus guard band to the IMDD channel with different GPON transmitter powers: (a) 1.25 Gb/s; (b) 2.5 Gb/s. Insets show the EVM distribution among tested channels at 3.125 GHz guard band. Green dash-dot line: reference curve without the IMDD channel.

depicts the EVM distribution among all tested channels at 3.125 GHz guard band. The IMDD channel has different powers compatible with the specifications for GPON equipments for ONU transmitters at upstream direction (Optical Network Unit ONU to Optical Line Termination OLT) [15]. The EVM curves point out that guard bands higher than 12.5 GHz is sufficient to minimize the FWM impact on the system. This results from the fact that the IMDD contribution to FWM components starts to be negligible since the phase matching condition stop being satisfied for higher guard bands. Additionally, for guard bands higher than 12.5 GHz the extra impact of fiber nonlinearities on the system is contributed essentially by XPM induced from the IMDD channel up to 200 GHz. The extra XPM penalty can be further enhanced for longer links (60 or 100 km) and/or IMDD channel counts. The XPM effect is emphasized for higher power categories of GPON transmitters, i.e. class C with max laser power of 7 dbm. For other classes where the power is bellow 3 dbm the XPM effect is negligible since the EVM converges to the reference system without the IMDD channel identified by the green line (EVM=10.3 %). It is worth pointing out that crosstalk between the IMDD and 31 st QPSK channels is very strong when the guard band is equal to only 1.5625 GHz. In this case, the EVM accounts for FWM, XPM and linear crosstalk. Fig. 3.(b) shows the performance of the hybrid scenario when the IMDD channel operates at 2.5 Gb/s instead. The inset highlights the distribution of the EVM among all tested channels when the guard band is 3.125 GHz. The IMDD power in this case is compatible with GPON classes for OLT transmitters at downstream direction (OLT to ONU) [15]. As discussed before, 12.5 GHz is sufficient to reduce the FWM impact on the system although those classes require much higher power compared to the 1.25 Gb/s-IMDD specifications. The overlapping in power occurs for the maximum power of class C at 1.25 Gb/s and 2.5 Gb/s. As such, this overlapping allows analyzing the XPM dependence on the IMDD bit rate. In the regime where XPM is dominant (guard band > 12.5 GHz) the black dash-lines from Fig. 3.(a)-(b) show that the XPM impact is slightly reduced when the IMDD channel operates at 2.5 Gb/s, e.g. EVM 1.25Gb/s@200GHz = 12.3 % and EVM 2.5Gb/s@200GHz = 11.3 %. To identify the performance of the IMDD channel impaired by linear and nonlinear crosstalk induced by coherent QPSK channels on the hybrid scenario, Fig. 4.(a)-(b) analyze the Q factor of the IMDD channel as a function of guard band. Classes with lower transmitter powers are more affected by crosstalk (lower Q) since the signal-to-interference ratio is reduced. Guard bands higher than 6.5 GHz assure error free operation (Q>15.5 db) without recurring to FEC for both 1.25 Gb/s and 2.5 Gb/s systems. On the other hand, the system may operate at 3.125 GHz guard band if FEC is employed. The Q factor is limited to around 20 db when the IMDD channel operates at 2.5 Gb/s. This comes from that fact that the 5 GHzfilter used at transmitter and receiver sides imposes more ISI (inter-symbol interference) for 2.5 Gb/s-signal than 1.25 Gb/ssignal as shown in the eye diagrams inserted in Fig. 4.(a)-(b). Figure 4. Q factor of the IMDD channel as a function of guard band at (a) 1.25 Gb/s and (b) 2.5 Gb/s with different GPON transmitter powers. The orange dash-dot line theoretically corresponds to a system BER=10-3 (FEC limit); Insets show the electrical eye diagram. IV. CONCLUSIONS We numerically investigated the system performance of coherent Ultra-Dense WDM based optical access networks with different requirements in terms of density and guard band to IMDD channel. As far as channel spacing is concerned, operating at 3 GHz gave the best compromise in terms of the overall system performance and spectral efficiency of the coherent-qpsk UDWDM scenario. When the coherent QPSK channels are coexisting with legacy G/EPON deployments, the required guard band of 12.5 GHz showed to be sufficient to guarantee minimum FWM impact on the overall system performance measured at both IMDD and QPSK channels. In this case the remaining EVM penalty with respect to the reference system (without IMDD) essentially was related to the induced nonlinear phase noise through XPM. In summary, UDWDM technology is a promising solution to enable broadband passive optical networks. Using coherent detection along with complex modulation formats increases spectral efficiency, wavelength selectivity as well as

transmission capabilities. The coexistence with legacy deployments such as GPON is feasible as demonstrated in this paper. Future studies include employing higher order modulation formats such as quadrature amplitude modulation with constellation orders as high as 16, 64 and 256 to achieve 10 Gb/s per user/wavelength. REFERENCES [1] J. Prat (Ed.), Next-Generation FTTH Passive Optical Networks: Research Towards Unlimited Bandwidth Access. Dordrecht, Netherlands: Springer, 2008. [2] F. J. Effenberger, The XG-PON System: Cost Effective 10Gb/s Access, J. Lightw. Technol., vol. 29, no. 4, Feb. 2011. [3] S. Smolorz, E. Gottwald, H. Rohde, D. Smith, and A. Poustie Demonstration of a Coherent UDWDM-PON with Real-Time Processing Proc OFC/NFOEC 2011, Los Angeles CA, USA, pp. PDPD4. [4] E. Ip, A. Lau, D. Barros, and J. M. Kahn, Coherent detection in optical fiber systems, Opt. Expr., vol. 16, no. 2, pp. 753 791, 2008. [5] E. Ip and J. M. Kahn, Compensation of dispersion and nonlinear impairments using digital back propagation, J. Lightw. Technol., vol. 26, no. 20, pp. 3416 3425, 2008. [6] S.J. Savory, Digital filters for coherent optical receivers, Opt. Express, vol.16, no. 2, pp. 804-817, Jan. 2008. [7] X. Li, X. Chen, G. Goldfarb, E. Mateo, I. Kim, F.Yaman, and G. Li, Electronic post-compensation of WDM transmission impairments using coherent detection and digital signal processing, Opt. Expr., vol. 16, no. 2, pp. 880 888, 2008. [8] E. Mateo, L. Zhu, and G. Li, "Impact of XPM and FWM on the digital implementation of impairment compensation for WDM transmission using backward propagation," Opt. Express, vol. 16, no. 20, pp. 16124-16137, Sep. 2008. [9] R.-J. Essiambre, G. Kramer, P.J. Winzer, G.J. Foschini, and B. Goebel, "Capacity limits of optical fiber networks," J. Lightw. Technol., vol. 28, no. 4, pp. 662-701, Feb. 2010. [10] V. Curri, P. Poggiolini, A. Carena and F. Forghieri, Dispersion Compensation and Mitigation of Nonlinear Effects in 111-Gb/s WDM Coherent PM-QPSK Systems, IEEE Photon. Technol. Lett., vol. 20, no. 17, pp. 1473-1475, Sep. 2008. [11] J.D. Reis, P. M. A Monteiro and A. T. Teixeira, Fiber Nonlinear Impact on Hybrid Ultra-Dense WDM based Optical Networks, Proc ICTON 2010, Munich, Germany, Vol. 12, pp. 1-4, July, 2010. [12] J.D. Reis and A.L. Teixeira, "Impact Of Nonlinear Effects Distortion On Hybrid Ultra-Dense WDM Based Networks", in Proc. OSA Access Networks and In-house Communications, Karlsruhe, Germany, Jun. 21-24, 2010, paper ATuB5. [13] J.D. Reis and A.L. Teixeira, Architectural Optimization of Coherent Ultra-Dense WDM based Optical Access Networks Proc OFC/NFOEC 2011, Los Angeles CA, USA, pp. OTuB7. [14] A. Bononi, M. Bertolini, P. Serena, G. Bellotti, Cross-Phase Modulation Induced by OOK Channels on Higher-Rate DQPSK and Coherent QPSK Channels, IEEE J. Lightwave Technol., vol. 27, pp. 3974-3983, 2009. [15] ITU-T G.984.2, SG 15, Gigabit-capable Passive Optical Networks (GPON); Physical Media Dependent Layer Specification, Mar. 2003. [16] P.J. Winzer, A. H. Gnauck, C. R. Doerr, M. Magarini, and L. L. Buhl, Spectrally Efficient Long-Haul Optical Networking Using 112-Gb/s Polarization Multiplexed 16-QAM J. Lightw. Technol., vol. 28, no. 4, pp. 547-556, Feb. 2010. [17] 3GPP TS 45.005, V8.6.0, Radio Access Network: Radio transmission and reception, Sep. 2009. [18] R.A. Shafik, Md.S. Rahman, A.R. Islam, On the Extended Relationships Among EVM, BER and SNR as Performance Metrics, ICECE 2006, Dhaka, Bangladesh, 19-21, December 2006.