Open Access
Add to BibSonomyAbstract
This paper presents a novel technique capable of Rayleigh backscattering (RB) mitigation and chromatic dispersion (CD) compensation for wavelength-division-multiplexed passive optical network (WDM-PON). The reduction of the interference caused by RB and CD in the uplink based on reflective electro-absorption modulator (REAM) is realized by the proposed correlative level (CL) coding. We investigate the RB-induced interferometric crosstalk for different fiber lengths. 10 Gb/s and 20 Gb/s transmissions over 70 km and 35 km fiber are demonstrated using the CL codes of dicode and modified duobinary (MD), respectively. Significant improvement in system resilience to backscattered seed light is verified for both dicode and MD coding. MD-coded signal also exhibits considerable robustness against the effects of CD.
©2012 Optical Society of America
1. Introduction
In order to meet the rapidly growing bandwidth demand, fiber to the home (FTTH) is being widely deployed over the world, and its related technologies are being extensively studied. Traditional FTTH architectures usually support asymmetric data rate in up- and downlink, assuming less traffic from the user-end to the central office. Conversely, the recent fast developing applications such as personal video, high-definition teleconferencing and various peer-to-peer services, stimulate a need for more upstream bandwidth. Wavelength-division-multiplexed passive optical network (WDM-PON) is generally considered as a promising technology for the future PON system that should support symmetrical high capacity in up- and downlinks. In WDM-PON, colorless optical network unit (ONU) and centralized light source (CLS) are commonly adopted to reduce the cost and raise the deployment feasibility [1–3]. Numerous efforts have been made on employing reflective semiconductor optical amplifier (RSOA) as the upstream transmitter in WDM-PON with the seed light delivered from the optical line terminal (OLT) [4–6]. However, RSOA has the drawback of small modulation bandwidth which is only 1 to 2 GHz. Recently, the integrated semiconductor optical amplifier with reflective electro-absorption modulator (SOA-REAM) has been reported as a more attractive upstream transmitter, because of its low chirp parameter and large modulation bandwidth [7–9]. Nevertheless, in high-speed single-feeder WDM-PONs based on SOA-REAM, cost-effective techniques are required to combat fiber impairments including Rayleigh backscattering (RB) and chromatic dispersion (CD). For such systems, the bottleneck is in the uplink faced with double path losses and reflection noise. Previously, spectrum-shaping codes such as Manchester [10] and 8b10b [11] have been applied to mitigate the reflection-induced crosstalk, but these techniques suffer from the required extra bandwidth and reduced spectral efficiency. Subcarrier multiplexing (SCM) [12] has also been investigated for crosstalk reduction in WDM-PON, but needs high-bandwidth transceivers. Moreover, CD as the other key issue of the system incurs inter-symbol interference (ISI), limiting the transmission capacity and distance. Apparently, signals with smaller bandwidth suffer less from the CD’s impacts. Duobinary modulation [13] and 4-ary pulse amplitude modulation (PAM) [14] that can shrink the signal bandwidth by around 50% have been exploited to alleviate ISI in RSOA-based WDM-PON. However, to generate duobinary signal by adjusting the RSOA’s response in [13] has tight requirement on the device uniformity, and multilevel signaling like 4-PAM [14] has an intrinsic receiver sensitivity penalty.
Previously we proposed the use of correlative level (CL) coding, which is one type of spectrum-shaping techniques, to resolve the issues of RB and CD in the uplink of REAM-based WDM-PON [15–17]. In this paper, we expand on these findings by comprehensively investigating the characteristics of RB and CD in the upstream channel. The crosstalk and ISI in the uplink are mathematically evaluated and theoretically studied. Based on the analysis, the scheme of CL coding is devised and presented. The proposed technique makes use of two particular CL codes: dicode and modified duobinary (MD). Because the two signals have spectral null at DC and up-shifts the signal spectrum compared with the non-return-to-zero (NRZ) signal, the beat noise between the reflected continuous-wave (cw) seed light and the upstream modulated signal can be greatly suppressed. The experiments at the rate of 10 Gb/s confirm that the dicode modulation allows 3 dB higher cw launched power and around 4.3 dB more tolerance to RB noise than the conventional modulation in REAM-based uplink. Moreover, 10 Gb/s 3-level dicode transmission is demonstrated for extended reach up to 70 km which is 15 km longer than the binary one. The results of the 20 Gb/s experiments verify that the 3-level MD uplink achieves the reach of 35 km, which is 13 km longer compared with the uncoded one. In addition, we evaluate the performance of 20 Gb/s binary, dicode and MD uplinks in the presence of cw-induced RB, and demonstrates the significant superiority of dicode and MD codes.
2. Analysis of RB and CD in WDM-PON
The remotely-seeded upstream transmission in single-feeder WDM-PON suffers from the excessive in-band crosstalk owing to RB and discrete reflections. External reflections from the passive elements can be minimized by using angled connectors and splices. RB, however, is inevitable as it is the intrinsic property of the optical fiber, but the RB-induced degradation can be reduced by changing the power spectral density (PSD) of the signal to be more separated from the RB interferers. Hence, we aims to alter the signal’s PSD using signal processing technique. Figure 1 illustrates the model of the uplink in a single-feeder CLS WDM-PON based on SOA-REAM, showing two major types of interference resulted from RB: RBcw (backscattered cw light) and RBsig (backscattered upstream signal). RBcw travels in the opposite direction of cw light and arrives at the OLT’s receiver together with the upstream signal. RBsig travels in the same direction with the cw light, and gets reamplified, remodulated and reflected at the ONU, then heads back to the OLT. Since both RBcw and RBsig fall within the signal bandwidth, they become crosstalk and incur beat noise in the signal. The mean power of RBcw and RBsig, denoted as PRBcw and PRBsig, can be mathematically expressed by [18, 19]:
where Pc is the launched carrier power from the OLT, S is the fiber recapture coefficient, α is the attenuation coefficient of the fiber, αc is the coefficient of the total insertion loss of components, αs is the fiber scattering coefficient, GONU is the gain provided by the ONU, and L is the fiber length. From Eqs. (1) and (2), the power ratio between the two types of RB interferers at the upstream receiver can be derived aswhere Lp is the single path loss from the ONU to the OLT before the demultiplexer. It is clear from Eq. (3) that the composition of Rayleigh noise greatly depends on the ONU gain and the path loss. The gain of the SOA-REAM is mainly determined by the input optical power when its operating parameters are stable. We measured the gain of the SOA-REAM against its input power under the condition of 80 mA bias for the SOA and −1.6V bias for the REAM, shown in Fig. 2 . Because of the saturation effect, the ONU gain decreases as the input power increases, and becomes negative after the injection power grows above 3 dBm, owing to the insertion loss of the REAM. The gain function is almost linear in the region of the input power from −5 to 5 dBm. Then, through polynomial fitting, the interpolated gain profile of the SOA-REAM is mathematically expressed bywhere Pin is the input power to the ONU, and equals toThe interpolated ONU gain curve via Eq. (4), also given in Fig. 2, is perfectly aligned with the measured results. It can be seen from Eq. (2) that the RBsig depends greatly on the GONU to compensate the link loss. Shown by Fig. 3 , longer reach causes lower Pin and higher GONU. However, the increase in GONU is not sufficient to balance the fiber loss from the distance increase, especially after ONU reaches saturation. Therefore, according to Eq. (3), PRBsig should be less than PRBcw, and the difference increases with the fiber length. Assuming PC is fixed to 3 dBm, the fiber has α of 0.25 dB/km, the insertion loss of the AWG at the remote node is 6 dB, we can attain the path loss from the ONU to the OLT, and the power ratio between RBcw and RBsig as a function of L and PC is plotted in Figs. 3(a) and 3(b), respectively. It is observed that RBcw surpasses RBsig when PC is over −1 dBm for 25 km fiber length and becomes more dominant at longer distances. Therefore, the objective of this work is to mitigate the degradation caused by only RBcw.Besides Rayleigh noise, fiber dispersion is another key factor degrading the performance of the system operating at high data rate of ≥10 Gb/s. The normalized frequency response of a standard single mode fiber (SSMF) with transmitter’s chirp parameter k and fiber length L can be obtained as [20]:
where c is the light speed, λ0 is the carrier wavelength and D(λ0) is the fiber dispersion parameter. For simplification, we only consider the impact of CD in deriving the fiber transfer function. The REAM chirp parameter k is low and measured to be 0.3 using the method described in [20]. CD, together with the frequency chirp, incurs frequency dips on the fiber channel according to Eq. (6). We plot the channel response expressed by Eq. (6) in Fig. 4 for fiber length L from 15 to 70 km. It is obvious that as CD accumulates the dips on the channel transfer function move closer to the baseband. When L is less than 30 km, the frequency dip is outside the 0-to-10-GHz window, hence CD hardly incurs ISI for the 10 Gb/s transmission. Even when L reaches 70 km, the frequency null closest to the direct current (DC) is still above 7 GHz. Hence the influence of CD on 10 Gb/s uplink with reach up to 70 km is not crucial. However, the CD of only 35 km fiber limits the channel bandwidth to around 10 GHz which is only half of the required bandwidth for 20 Gb/s uplink. Therefore, CD is a vital problem for the transmission at high speed like 20 Gb/s.3. Principle of correlative level (CL) coding
Since RBcw is the major RB interferer in the REAM-based WDM-PON, this paper only focuses on the RBcw suppression. For this purpose, we first derive the PSD of RBcw, mathematically expressed by the following [21]:
where Ib is the RBcw intensity and Δω denotes the laser linewidth. Equation (7) indicates that the bandwidth of RBcw is twice the Δω. Hence the cw interferer that occupies very narrow spectrum can be largely removed by a high-pass filter (HPF) if the signal energy is moved away from DC. This idea can be simply realized by processing the binary message with a DC-balanced code prior to modulation. A group of CL codes that can enable zero DC content have a typical coding transfer function as [22, 23]where D is 1-bit delay. When n equals to 0 and 1, the CL codes are called dicode and MD, respectively, whose spectra are plotted in Fig. 5(a) and compared with that of binary signal. Figures 5(b) and 5(c) illustrate the procedures of these two CL codes as simple delay-and-subtract lines. Both dicode and MD signals have three levels and can be simply converted back into binary signal by a full-wave rectifier, since the top and bottom levels of these two signals correspond to binary “0” and the middle level corresponds to binary “1”.
Fig. 5 (a) Spectra of 20 Gb/s binary, dicode and MD signals, and diagrams of (b) dicode and (c) MD coding.
Apparently seen from Fig. 5(a), dicode reforms the power distribution of the binary signal by shifting the peak furthest away from DC among all three signals. Compared with mBnB and Manchester codes that also achieve DC null, the advantage of dicode code is zero overhead. MD signal has similar performance with duobinary signal which has found many applications in optical systems owing to its robustness to fiber impairments such as CD [24, 25]. MD differs from duobinary in only the additional (1-D) term that forces the DC content to be null. As shown in Fig. 5(a), MD signal has about half the bandwidth of the binary signal. Therefore, besides RB suppression, MD signal is also expected to be tolerant against CD.
4. Experiments and results
The experimental setup to evaluate the upstream performance by dicode and MD coded signals is depicted in Fig. 6 . Seed light at 1550 nm is generated by a DFB laser and launched into a length of SSMF. The output seed power is set to 3 dBm, which is sufficient to cover the link loss for maintaining the performance of SOA-REAM and low enough to avoid the severe degradation induced by Brillouin backscattering. The cw light injected into the SOA-REAM employed at the ONU is modulated by a 215 −1 pseudo-random binary sequence (PRBS) in the formats of NRZ, dicode and MD. The SOA-REAM is biased at 75 mA and −1.5 V with 3 Vpp electrical input. Figures 7 (a) and 7(b) show that the central wavelength of the SOA is 1554 nm and the REAM 3-dB modulation bandwidth is around 20 GHz, respectively. At the OLT side, 10% of the signal is tapped off for monitoring the received optical signal to RB ratio (OSRR). Then the remaining optical power is detected by an 18 GHz photodiode (PD) followed by two types of decision circuit depending on whether the signal is coded. The decision circuit for uncoded binary signal comprises a low-pass filter (LPF) with the bandwidth matched to bit rate to eliminate the out-of-band noise. The Dicode and MD circuits that have the same design use an HPF with cut-off frequency of 200 MHz to filter out the reflection noise without causing serious degradation on the signal. They also consist of a data-rate LPF same with the binary circuit and a full-wave rectifier for decoding the signal. Due to the lack of analog rectifier, signals are captured by a 15 GHz storage oscilloscope at 2 samples/bit with a length of 4 × 106 bits for digital filtering, rectification and bit error rate (BER) calculation. Thus, the receivers for the uncoded and coded signals are the same except the post-detection signal processing.
4.1 Experiments on data rate of 10 Gb/s
At the data rate of 10 Gb/s, dicode rather than MD coding is used in the uplink, since CD is a less serious problem than RB, and dicode format has signal energy shifted further from the baseband than MD format. For studies on enhanced RB tolerance by dicode modulation, the fiber length is fixed to 20 km, and the launched power from the OLT is tuned to vary the received OSRR. Figure 8(a) gives the obtained BERs as a function of OSRR swept from 13 dB to 21.5 dB for binary and dicode modulation. At BER of 10−4, dicode has around 4.3 dB more in Rayleigh noise tolerance over NRZ, and this improvement tends to become larger at lower BER levels. With the help of 7% forward error correction (FEC), the minimal OSRR to recover error-free dicode signal is about 16.5 dB. In order to examine the capability of the proposed technique in terms of reach extension, we also measured the BERs for different transmission distances. The obtained results of both dicode and NRZ formats are shown in Fig. 8(b). The BERs of the 10 Gb/s uplink over 70 km bidirectional fiber can be achieved below the FEC limit of 3.8 × 10−3. It is shown that the maximal reach of the uplink utilizing dicode modulation is about 15 km longer than the one adopting binary modulation.
4.2 Experiments on data rate of 20 Gb/s
When the bit rate increases to 20 Gb/s, binary, dicode and MD signals are transmitted from the ONU in the experiments to compare their robustness against both fiber dispersion and RB noise. The system performance is evaluated by varying the fiber length from 5 to 35 km. Figure 9 plots the BERs versus transmission distance, showing that error-free 20 Gb/s transmission over 35 km fiber can be enabled by MD coding with the help of 7% FEC. At BER of 5 × 10−4, the MD transmission achieves about 8 km and 13 km longer reach than the dicode and the binary transmission, respectively. Therefore, MD coding is a better choice for bandwidth-limited channel because it has the smallest bandwidth among these three signals.
Since Brillioun scattering can be triggered at high injection power and falls around the frequency of 11 GHz which is within the bandwidth of the received signal for 20 Gb/s uplink, the method of tuning launch power to adjust the OSRR in 10 Gb/s experiment cannot be applied here. As a consequence, we evaluate the capability of the proposed techniques in mitigating only the RBcw by carrying out the experiment with the setup displayed in Fig. 10 . The cw light at 1550 nm provided by a DFB laser is split into two paths: 20% of power is injected into the SOA-REAM that is modulated by the upstream data at 20 Gb/s, while 80% of power is used to generate the RBcw from a 50 km SSMF. The two paths are coupled together before reception that works in the same way as the transmission experiment. Figure 11 displays the measured BER results as a function of the crosstalk-to-signal ratio (CSR) at the receiver. Eye diagrams of binary, dicode, and MD signals are also given in Fig. 11 near their respective BER results in dashed circles. The BER curves of the uplink employing dicode or MD coding appear to be much more stable than the conventional binary transmission at different CSR values. Dicode signal exhibits the best reflection tolerance which is 7 dB higher than the binary signal at BER of 10−6, because of its low power distribution near DC. The resilience to RBcw of MD transmission appears to be slightly worse than the dicode format, but much better than the standard NRZ format. Although MD signal is less effective in crosstalk reduction, it outperforms dicode signal at the presence of accumulated ISI caused by CD.
5. Conclusions
In this paper, we have investigated the different causes of the RB-induced crosstalk and the power of RBcw and RBsig to analyze their impacts on the system performance. Based on our findings about the dominance of RBcw, we present a cost-effective technique using spectrum-shaping CL code to mitigate Rayleigh noise and fiber dispersion in the uplink of REAM-based WDM-PON. Dicode as a CL code is applied in 10 Gb/s REAM-based upstream transmission and achieves good enhancement in tolerance to RB-accumulated power. Another CL code called MD is employed for 20 Gb/s uplink and effectively overcomes the impacts of both RB and CD. At 10 Gb/s, the minimal allowable OSRR at the BER of 10−4 is reduced to 16.5 dB by dicode coding, which is 4.3 dB better than the uncoded transmission. Experiments also demonstrate that 10 Gb/s and 20 Gb/s uplink can reach up to 70 km and 35 km using dicode and MD signaling, respectively. Although MD signal is less robust to RB noise than dicode signal, it achieves about 8 km and 13 km longer reach than dicode and binary at the uplink rate of 20 Gb/s. The proposed CL coding is a potential technique to upgrade the capacity and reach of the uplink in WDM-PON at the cost of small additional complexity of signal processing.
References and links
1. P. Healey, P. Townsend, C. Ford, L. Johnston, P. Townley, I. Lealman, L. Rivers, S. Perrin, and R. Moore, “Spectral slicing WDM-PON using wavelength-seeded reflective SOAs,” Electron. Lett. 37(19), 1181–1182 (2001). [CrossRef]
2. K. M. Choi, J. S. Baik, and C. H. Lee, “Broad-band light source using mutually injected Fabry-Perot laser diodes for WDM-PON,” IEEE Photon. Technol. Lett. 17(12), 2529–2531 (2005). [CrossRef]
3. E. K. MacHale, G. Talli, P. D. Townsend, A. Borghesani, I. Lealman, D. G. Moodie, and D. W. Smith, “Extended-reach PON employing 10Gbits/s integrated reflective EAM-SOA,” Proc. European Conference on Optical Communication (ECOC), paper Th.2.F.1 (2008).
4. S. J. Park, G. Y. Kim, T. Park, E. H. Choi, S. H. Oh, Y. S. Baek, K. R. Oh, Y. J. Park, J. U. Shin, and H. K. Sung, “WDM-PON system based on the laser light injected reflective semiconductor optical amplifier,” Proc. European Conference on Optical Communication (ECOC), paper We3.3.6 (2005).
5. C. Arellano, C. Bock, J. Prat, and K. D. Langer, “RSOA-based Optical Network Units for WDM-PON,” Proc. Optical Fiber Communication (OFC) Conference, paper OTuC1 (2006).
6. W. Lee, M. Y. Park, S. H. Cho, J. H. Lee, C. Kim, G. Jeong, and B. W. Kim, “Bidirectional WDM-PON based on gain-saturated reflective semiconductor optical amplifiers,” IEEE Photon. Technol. Lett. 17(11), 2460–2462 (2005). [CrossRef]
7. Q. Guo and A. V. Tran, “Demonstration of 40-Gb/s WDM-PON system using SOA-REAM and equalization,” IEEE Photon. Technol. Lett. 24(11), 951–953 (2012). [CrossRef]
8. H. S. Kim, D. C. Kim, K. S. Kim, B. S. Choi, and O. K. Kwon, “10.7 Gb/s reflective electroabsorption modulator monolithically integrated with semiconductor optical amplifier for colorless WDM-PON,” Opt. Express 18(22), 23324–23330 (2010). [CrossRef] [PubMed]
9. G. Girault, L. Bramerie, O. Vaudel, S. Lobo, P. Besnard, M. Joindot, J.-C. Simon, C. Kazmierski, N. Dupuis, A. Garreau, Z. Belfqih, and P. Chanclou, “10 Gbit/s PON demonstration using a REAM-SOA in a bidirectional fiber configuration up to 25 km SMF,” Proc. European Conference on Optical Communication (ECOC), paper P.6.08 (2008).
10. S. Y. Kim, S. B. Jun, Y. Takushima, E. S. Son, and Y. C. Chung, “Enhanced performance of RSOA-based. WDMPON by using Manchester coding,” J. Opt. Netw. 6(6), 624–630 (2007). [CrossRef]
11. M. Presi, A. Chiuchiarelli, R. Proietti, P. Choudhury, G. Contestabile, and E. Ciaramella, “Single Feeder Bidirectional WDM-PON with Enhanced Resilience to Rayleigh-Backscattering,” Proc. Optical Fiber Communication (OFC) Conference, paper OThG2 (2010).
12. K. Y. Cho, A. Murakami, Y. J. Lee, A. Agata, Y. Takushima, and Y. C. Chung, “Demonstration of RSOA-based WDM PON operating at symmetric rate of 1.25 Gb/s with high reflection tolerance,” Proc. Optical Fiber Communication (OFC) Conference, paper OTuH4 (2008).
13. M. Omella, V. Polo, J. Lazaro, B. Schrenk, and J. Prat, “10 Gb/s RSOA transmission by direct duobinary modulation,” Proc. European Conference on Optical Communication (ECOC), paper Tu.3.E.4 (2008).
14. K. Y. Cho, Y. Takushima, and Y. C. Chung, “Enhanced chromatic dispersion tolerance of 11 Gbit/s RSOA-based WDM PON using 4-ary PAM signal,” Electron. Lett. 46(22), 1510–1511 (2010). [CrossRef]
15. Q. Guo and A. V. Tran, “Mitigation of Rayleigh Noise and Dispersion in REAM-based WDM-PON using Partial-Response Signaling” Proc. European Conference on Optical Communication (ECOC), paper We.2.B.4 (2012).
16. Q. Guo and A. V. Tran, “Level coding technique for a wavelength-division-multiplexed optical access system using a remodulation scheme,” Opt. Lett. 37(19), 4137–4139 (2012). [CrossRef] [PubMed]
17. Q. Guo and A. Tran, “40 Gb/s Operation of SOA-REAM in Single-Feeder WDM-PON [Invited],” J. Opt. Commun. Netw. 4(11), B77–B84 (2012). [CrossRef]
18. J. Xu, M. Li, and L. K. Chen, “Rayleigh Noise Reduction in 10-Gb/s Carrier-Distributed WDM-PONs Using In-Band Optical Filtering,” J. Lightwave Technol. 29(24), 3632–3639 (2011). [CrossRef]
19. C. Arellano, K. Langer, and J. Prat, “Reflections and multiple Rayleigh backscattering in WDM single-fiber loopback access networks,” J. Lightwave Technol. 27(1), 12–18 (2009). [CrossRef]
20. F. Devaux, Y. Sorel, and J. F. Kerdiles, “Simple measurement of fiber dispersion and of chirp parameter of intensity modulated light emitter,” J. Lightwave Technol. 11(12), 1937–1940 (1993). [CrossRef]
21. R. K. Staubli and P. Gysel, “Statistical Properties of Single-Mode Fiber Rayleigh Backscattered Intensity and Resulting Detector Current,” IEEE Trans. Commun. 40(6), 1091–1097 (1992). [CrossRef]
22. P. Kabal and S. Pasupathy, “Partial response signaling,” IEEE. Trans. Commun. COM 23(9), 921–934 (1975). [CrossRef]
23. A. Lender, “Correlative digital communication techniques,” IEEE Trans. Commun. Technol. COM 12(4), 128–135 (1964). [CrossRef]
24. X. Gu and L. C. Blank, “10 Gb/s unrepeatered three-level optical transmission over 100 km of standard fiber,” Electron. Lett. 29(25), 2209–2211 (1993). [CrossRef]
25. K. Yonenaga and S. Kuwano, “Dispersion-tolerant optical transmission system using duobinary transmitter and binary receiver,” J. Lightwave Technol. 15(8), 1530–1537 (1997). [CrossRef]
