Simple Colorless Long-Reach WDM-PON with Rayleigh Backscattering Noise Mitigation Employing Remodulated Orthogonal Coding

Zhisheng Li , Qiang Dou, Lei Wang, Peijun Yang, Chaobing Zhou

School of Computer Science, National University of Defense Technology, Changsha,410073, China

* The corresponding author, email:lizhsh_123@163.com

I. INTRODUCTION

The full-duplex transmission on single-feeder fiber based on the reflective semiconductor optical amplifier (RSOA) has become the main solution for wavelength-division-multiplexing passive optical networks (WDMPON), in which the colorless optical network unit (ONU) and mitigation of optical beat interference (OBI) noise caused by Rayleigh backscattering are critical technologies. One method uses an extra centralized continuous wave (CW) light source in optical line terminal (OLT) which is modulated at ONU as the seed of upstream (US) data, but it will increase the cost and introduce more OBI noises. In the other method, the downstream (DS) signal is used as the seed source after removing the modulated signal by gain-saturation RSOA.Then the US data is remodulated at the same wavelength of DS by RSOA. However the gain-saturated RSOA is not very suitable in long reach and high speed PON for its limit in injecting power, extinction ratio(ER) and remodulation bandwidth [1-4].

The signal’s transmission performance in a full-duplex fiber is limited by OBI noise resulting from Rayleigh backscattering and reflection of discrete components[5][6]. From ref [7-9], the Rayleigh backscattering is spontaneous. The backscattered close-in noise spectrum is concentrated at low frequency from 5MHz to 200MHz and grows linearly with light linewidth, input power and fiber length. As the spectra of backscattered light of CW sources, DS signals and modulated US signals have a wide spectral overlap, so the OBI noise is severe. The most effective method to mitigate the OBI noise is to reduce the spectral overlap between the US signal and the interferer noise[10]. Using coding method to reshape the power spectrum of signal and to separate the US and DS spectra such as Manchester [11] and 8b10b [12] is a good method,because it needs no extra optical devices and is easy to be realized. Recently Qi Guo and An Vu Tran proposed a kind of effective coding method called level coding in which {0} is replaced by {1} or {-1} and {1} is replaced by {0}. As this DC-balanced correlative level codes can shift the US’s lower frequency components to higher frequency and can reduce spectral overlap along with the signal’s low frequency power, it can reduce OBI noise effectively[13-16].

The authors proposed a new method to mitigate the OBI noise caused by Rayleigh backscattering in full-duplex WDM-PON system.

Different from the proposed level coding method, we employ a pair of orthogonal Walsh codes sequence composed by {1, 1} and {1,-1}, respectively, in which {1, 1} is used to encode the DS signal while another {1,-1} is intended to carry the US signal. Both of them are generated and modulated at same wavelength at OLT and are transmitted to ONU. In ONU, the raw US is directly modulated on the coded DS signal without erasing the DS. As the coded DS and remodulated US are mutual orthogonal, this will not only result in coding gain but also mitigate the beat noise and white noise more effectively by convolution integral for US signal received at OLT. Thanks to the proposed orthogonal codes, the spectrum of coded US is moved near to the frequency of US seed. It can reduce the spectral overlap between US signal and DS signal, which mitigates the OBI noise. When the DS and US have the same bit rate, our method can reduce the spectral overlap just like level coding, but when bit rate of US is smaller than that of DS,our method has less spectral overlap to reduce the OBI more effectively. Besides, we can get a coding gain by correlation algorithm in the uplink. It means long reach without amplifier or high splitter ratio can be realized.

The paper is arranged as followings: the second part is the principle describes, the third is the experiment setup and results discussion,at last is the conclusion.

II. PRINCIPLES

2.1 Configuration of our proposed WDM-PON employing RSOA

Figure 1 illustrates the configuration of the proposed US remodulation scheme in a full-duplex WDM-PON. At OLT, a pair of orthogonal code sequences composed by {1,1} and {1,-1}, respectively are generated with same code chip rate (Data Rate × Code Length= Chip Rate) for DS code element and US seed, respectively. For example, the DS sequence can be coded by the simplest first order Walsh code {1, 1}, which has no effects on the raw data rate of DS. While the US seed is used the second Walsh code {1,-1}. Then the coded DS signal and the second order Walsh code{1,-1} used as US seed are mixed and optical modulated. This means the coded DS signal is a four-level signal consisting {-1, 0, 1, 2} four levels randomly.

Fig. 1 The principles of US remodulation scheme in the full-duplex WDM-PON

After de-multiplexing (DEMUX) in remote node, signals in different wavelengths are assigned to corresponding ONUs and are divided into two branches in each ONU. One branch is detected by a PD and then filtered by a LPF to recover the DS signal. The other branch is used as the seed sources for US remodulation.Because the coded DS and the other orthogonal code are mutually orthogonal, the coded DS does not require to be erased. So a common RSOA without gain-saturation function can be used here to amplify the light and modulate the US signal.

The remodulated US signal is coupled into fiber after an optical circulator for uplink transmission. At OLT, an optical circulator is used to separate the US signal. For the receiving of US signals, we just need to carry out the convolution with the local code storing in OLT. By using correlation algorithm, the coding gain can be achieved, which can be adjusted by changing the date rate of US.

Fig. 2 An example coding and coding gain for DS (5Gb/s) US(1.25Gb/s)

Fig. 3 The spectrum analysis for level code in ref [14], uncoded NRA, our proposed orthogonal coded DS and our proposed remodulated US

As shown in figure 2, it is an example of the coding sketch. Here, we adopt orthogonal Walsh code with the code element of {1,-1} for US coding in our system. The Walsh codes with other code element such as {1, 1,-1,-1} and{1,-1,-1, 1} can also be used for US seed and remodulation, but they will spread the spectrum of US more severely which will expand the US bandwidth. So we employ the simplest orthogonal code, the first order {1, 1} and second order{1,-1} Walsh code to code the raw DS and US,respectively. Therefore in figure 2 the pair of orthogonal code’s chip rate is 10Gchips/s. It is a twice spreading-spectrum orthogonal code.In this case, the US seed is absolutely separate form raw DS in frequency, which has no effect on DS detecting and can reduce the spectral overlap of coded DS and US. Furthermore,code {1,-1} is just like level code which can upshift the US spectrum to higher spectrum to reduce the spectral overlap between the coded DS and US. It can suppress the OBI noise. The code length of US in our system is 8 chips, so we can get 9dB coding gain in theory. With multilevel modulation and SNR degradation,the practical coding gain should be less than the theoretical value of 9dB.

2.2 Analysis of spectral overlap

The level code[14] has much less baseband components compared with the uncoded NRZ signal shown in figure 3(a). The measured spectrum of our proposed orthogonal-coded NRZ DS signal is displayed in figure 3(b) and the spectra of remodulated US signal at different data rates are shown in figure 3(c). Here we explore the spectra of DS 5Gb/s, 1.25Gb/s US and 2.5Gb/s US, respectively.

Viewing of figure 3 (b) and (c), the spectral overlap between DS and US expands with US bit rate. For example, for a 5Gb/s DS and 1.25Gb/s remodulated US, only about 10%spectra are overlapped. For a 5Gb/s DS and 2.5Gb/s remodulaed US, about 40% spectra are overlapped. For a symmetrical bit rate, it is the same as the level code [14]. The proposed method can reduce the overlap further than that of level code[14].By this way, the beat noise can be degraded more effectively.

The improvement of system performance by means of our proposed method is mainly in four aspects. Firstly, with the help of a pair of orthogonal codes, we can move the spectrum of US to higher frequency. It can reduce the OBI more effectively than by using the level code in reference [14]. Secondly, we can get a coding gain by correlation algorithm in the uplink. It means long reach without amplifier[18] or high splitter ratio can be realized. The third, no extra CW light is needed for US seed light which can reduce cost and OBI caused by CW light.Finally, we do not need gain-saturation RSOA to erase the DS. In the λ-reused system, the extinction ratio(ER) is a critical parameter since it significantly affects the performance of both up- and down-link. In our proposed system,since erasing DS signal is not desired in the ONU and with help of orthogonality between DS and US code, our method has better performance at the lower extinction ratio modulator.

III. EXPERIMENTS AND ANALYSIS

3.1 Transmission performance

Figure 4 shows the experimental setup of our proposed remodulated US WDM-PON system. At OLT, a 1550nm wavelength DFB laser and a Mach-Zehnder modulator (MZM) are used for optical modulation. The DS signal is generated by an Arbitrary waveform generator(AWG) working at 10GS/s sampling rate with 8bit resolution. A pseudorandom bit sequence of length 217-1 is generated and coded by the first order Walsh code for DS. After being mixed with the US seed code and normalizing,the digital signal is converted to analog signal by a DAC and then fed into a MZM. The Extinction Ratio (ER) of the output signal of MZM is about 14dB. Because of limited instruments, the raw bit rate of DS signal is 5Gb/s and an EDFA and a MZM are used instead of the RSOA in this experiment. However, we will test our power margin compared with the normal on-off keying (OOK) signal for DS and US with the same experiment equipments.The “normal” OOK signal means the light is modulated by uncoded NRZ electrical signal.Through an optical circulator, the modulated DS signal is fed into the fiber for 0km, 20km and 50km transmission, respectively, with 3dBm launch power. A variable optical attenuator (VOA) is used to emulate the optical power splitter and to adjust the power budget.At the ONU, the received signal is filtered by the tunable optical band-pass filter (OBPF),and is split into two branches. The DS is directly detected by 10GHz PD with a 4GHz bandwidth electrical low pass filter (LPF). The US raw is NRZ of 1.25Gb/s and remodulated on the amplified DS signal without erasing DS. The remodulated US signal is fed into the link through the optical circulator and the TOF. For the reception of US signal, the OLT detects the US signal using a 10GHz PD and then samples at 50GS/s. Then it will be operated by correlation process. The VOA in OLT is adjusted to change the received US light power from -27dBm to -19dBm.

Figure 5 shows the measurement of BER versus receiver sensitivity of remodulated coded US and uncoded NRZ in a back to back(B2B) system. In figure 5, at the BER of 10-3,remodulated coded signal has about 4dB lower power penalty than NRZ signal. With development of light power, the multi-level signal will have better performance, the power margin increases to 5.40dB at BER of 10-7. For the code length of US is 8 chips, we should get 9dB coding gain in theory. However, as our remodulated seed signal is a four-level signal, with multilevel modulation and SNR degradation, the coding gain is smaller than the expected value. In the B2B system, the OBI noise is mainly from reflection of discrete components. The coding gain triggers 4～6dB power margin in our proposed system. The eye diagrams in the inset of figure 5 are the remodulated US coded (upper) and NRZ signal(lower) at the power of -24dBm.

Fig. 4 The experimental setup of remodulated DS WDM-PON system

The BER measurement results of remodulated US at 20km and 50km transmission are shown in figure 6 (a) and (b) respectively.As shown in figure 6 (a) and (b), the power margin by using remodulated US increases to 4.71dB at 20km at BER of 10-3, 5.20dB at 50km at BER of 10-3. At BER of 10-7,the power margin is improved to 7.34dB at 20km and 9.60dB at 50km, respectively. Our proposed method can suppress the Rayleigh noise, so the power margin at 20km or 50km is larger than that in B2B system.

Fig. 5 The BER versus receiver sensitivity power in B2B system

Fig. 6 The BER versus US receiver sensitivity at 20km (a) and 50 km (b)

With the length of fiber, the power margin of remodulated US becomes larger. This is because the OBI noise caused by Rayleigh backscattering increases with fiber length,resulting in the performance of uncoded NRZ signal degrading rapidly. However, the ability of our proposed orthogonal coded signal is more superior to mitigate Rayleigh noise, due to much less spectral overlap between US and DS signal or CW light.

Figure 7 is the power margin versus BER in our remodulated US system. With the length of fiber, the power margin increases accordingly, demonstrating the effectiveness of OBI suppression in our proposed method.

3.2 Uplink crosstalk tolerance

To evaluate the resilience to crosstalk caused by Rayleigh backscattering and reflection of discrete components in the uplink, we use the experimental setup shown in figure 8. The lower branch simulates the reflected DS signal as the crosstalk to US signal. [14][15] The experiment condition is same as above-mentioned system.The received power is fixed to -36dBm for US and the BERs are measured as a function of the CSR (Clutter-to-Signal Ratios).

As shown in Fig.9, we can get BER of 10-7 for uncoded NRZ signal, but the CSR lower than -25dB is hard to be realized for experimental instruments limitation. So the curves of remodulated US signal is gotten by using the straight line extrapolation. As shown in figure 9(a), the remodulated orthogonal coded US signal has a 5.78dB larger crosstalk tolerance than NRZ signal at BER of 10-3and a 22.71dB larger tolerance at BER of 10-7in B2B system. The improvement of crosstalk tolerance increases with BER reducing, demonstrating that our method has better performance when CSR is lower. From figure 9, we can discover that the improvement of crosstalk tolerance increases with the fiber length expanding, compared with NRZ signal. It demonstrates that our system has much better transmission performance than uncoded NRZ signal in uplink for suppression of the OBI noise and coding gain.

IV. CONCLUSION

We proposed a new method to mitigate the OBI noise caused by Rayleigh backscattering in full-duplex WDM-PON system. By combining a pair of orthogonal codes and with correlation receiver for uplink receiving, it can effectively enhance the system resilience to the OBI noise. Without using extra centralized CW light for US seed lightwave at OLT and without erasing the DS at ONU, the proposed scheme is a simple method which can be used in the legacy WDM-PON. Our method does not require more optical devices for DS receiver at ONU along with low ER for US. By using correlation receiving method to achieve code gain and suppress white noise in US receiving, the method is suitable for long-reach and high splitter ratio PON system. The experiments show a total 4～9dB power margin at 1.25Gb/s US, 5Gb/s DS system at 20km-70km long-reach fiber transmission.

ACKNOWLEDGEMENTS

This work is supported by Memory access mode Adaptive Perception Intelligent Storage Architecture and Key Technologies (under granted: 61402501), and School of Computer Science, National University of Defense Technology. At the same time, I am also very grateful to the teachers and students who helped me in this research.

Fig. 7 The power margin caused by our proposed method versus BER

Fig. 8 Rayleigh backscattering and reflection crosstalk tolerance our proposed system

Fig. 9 The BER versus CSR at B2B(a), 20km(b), 50km(c), 70km(d)

[1] Bo Zhang, Chinlon Lin, Li Huo, Zhaoxin Wang,and Chun-Kit Chan; “A Simple High-Speed WDM PON Utilizing a Centralized Super continuum Broadband Light Source for Colorless ONUs” , OTuC6. OFC 2006.

[2] Bock C., Prat J., Langer K.-D; “RSOA-based Optical Network Units for WDM-PON”, OTuCl. OFC 2006

[3] Paola Parolari, Lucia Marazzi,Marco Brunero; et al; “10-Gb/s Operation of a colorless self-seeded transmitter over more than 70km of SSMF;IEEE Journal of lightwave technology”; Vol.26,NO.5: 599-602. (2014)

[4] S.A.Gebrewold, L.Marazzietal; “Colorless self-seeded fiber cavity laser transmitter for WDM-PON”; CLEO (2014)

[5] Hanwu Hu and Hannan Anis; “Degradation of bi-directional single fiber transmission in WDMPON due to beat noise”; Journal of lightwave technology Vol.26, No.8: 870-881 (2008)

[6] U.H.Hong, K.Y.Cho, Y.Takushima et at; “Effects of Rayleigh backscattering in long-reach RSOABased WDM PON”; Conference on (OFC/NFOEC) (2010)

[7] Olukayode Okusaga, James P.Cahill, Andrew Docherty; et al; “Spontaneous inelastic Rayleigh scattering in optical fibers”; Optics Letters; Vol.38, No.4: 549-552.(2013)

[8] Peter Gysel and Roland Karl Staubli; “Spectral properties of Rayleigh backscattered light from sing-mode firbers caused a modulated probe signal”; Journal of lightwave technology; Vol.8,No.12: 1792-1798 (1990)

[9] Hanlin Feng, Jia Ge, Shilin Xiao et al; “Suppression of Rayleigh backscattering noise using cascaded-SOA microwave photonics filter for 10Gb/s loop-back WDM PON”; Optics Express;Vol.22, Nol.10: 11770-11775 (2014)

[10] C.W.Chow, G.Talli,P.D.Townsend; “Rayleigh noise reduction in 10Gb/S DWDM-PONs by wavelength detuning and phase-modulation-induced spectral broadening”; IEEEE Photonics Technology Letters, Vol.19, No.6. 423-425 (2007)

[11] S.Y.Kim, S.B.Jun, Y.Takushima, E.S.Son and Y.C.Chung; ”Enhanced performance of RSOA-based WDM PON by using Manchester coding”,J.Opt.Netw.6(6),624-630 (2007)

[12] M.Presi, A.Chiuchiarelli, R.Proieui, p.Choudhury,G.Contestabile and E.Ciaramella; “Single feeder bidirectional WDM PON with enhanced resilience to Rayleigh-backscattering”, Pro.Optical fiber communication (OFC)conference, paer OThG2 (2010)

[13] Andrea Chiuchiarelli, Marco Presi, Roberto Proietti; et al; “Enhancing Resilience to Rayleigh crosstalk by means of line coding and electrical filtering; IEEE Photonics technology letters”, Vol.22, No.2. 85-87 (2010)

[14] Qi Guo; An Vu Tran; “Performance Enhancement in RSOA-Based WDM passive optical networks using level coding”; IEEE Journal of lightwave technology, Vol. 31, No.1: 67-73 (2013)

[15] Q.Guo and A.V.Tran, “Mitigation of Rayleigh noise and dispersion in REAM-based WDM –PON using spectrum-shaping codes,” Opt. Express 20(26), B452-B461 (2012)

[16] Gupta M.K., Dali P.P. and Singh G.; “A novel approach to enhance the performance of ring based WDM PON”; International conference on signal propagation and computer technology”;(ICSPCT): 329-331 (2014)

[17] Zhengxuan Li, Lilin Yi and Weisheng Hu; “Symmetric 40Gb/s TWDM-PON with 51-dB loss budget by using a single SOA as preamplifier,booster and format converter in ONU”; Optics Express; Vol.22,No.20, (2014)

[18] Gyaneshwar C., Gupta. “Over 100km Bidirectional, Multi-channels COF-PON without Optical Amplifier”, 2006 OSA/OFC (2006)