- Email: [email protected]
Self-match based on polling scheme for passive optical network monitoring
Optics Communications 417 (2018) 19–23
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
Self-match based on polling scheme for passive optical network monitoring Xuan Zhang a, *, Hao Guo b , Xinhong Jia a , Qinghua Liao c a
College of Physics and Electronic Engineering, Sichuan Normal University, Chengdu, Sichuan, 610101, China State Key Laboratory of Electronic Thin Films and Integrated Devices, College of Optoelectronic Information, University of Electronic Science and Technology of China, Chengdu, Sichuan, 610054, China c College of Science, Nanchang University, Nanchang, Jiangxi, 330031, China b
a r t i c l e
i n f o
Keywords: Passive optical network (PON) Link monitoring Self-match Polling False alarm rate (FAR)
a b s t r a c t We propose a self-match based on polling scheme for passive optical network monitoring. Each end-user is equipped with an optical matcher that exploits only the specific length patchcord and two different fiber Bragg gratings with 100% reflectivity. The simple and low-cost scheme can greatly simplify the final recognition processing of the network link status and reduce the sensitivity of the photodetector. We analyze the timedomain relation between reflected pulses and establish the calculation model to evaluate the false alarm rate. The feasibility of the proposed scheme and the validity of the time-domain relation analysis are experimentally demonstrated.
1. Introduction Optical access network (OAN) is capable of delivering high-capacity data, consuming less energy and saving capital expenditure (CapEx), which can remove the bottleneck of the telecommunication development [1,2]. Passive optical networks (PONs) are considered one of the most promising candidates for OANs due to the high transmission rate and low-cost infrastructure [3,4]. With the rapid development of PONs, management and maintenance issues are emerging. As an effective solution, PON monitoring can reduce provisioning time, improve quality of service (QoS), attract more clients and reduce maintenance costs [5,6]. Therefore, a simple but effective monitoring scheme is essential for continuously developing PONs. Optical time domain reflectometry (OTDR) is widely used in the point-to-point (P2P) fiber link test, which can provide a plot of distance versus signal level in a fiber. However, it may be ineffective in pointto-multipoint (P2MP) networks because all returned signals coming from different branches add up together [6,7]. Some modified OTDR techniques, such as reference reflector and tunable OTDR (i.e., RROTDR and T-OTDR), are limited by the network size and expensive OTDR equipment, respectively [8]. Note that the PON market is highly sensitive to cost. Thus, the PON monitoring system must be costeffective, especially at the user terminals because the components cannot be shared. Some PON monitoring schemes aiming at improving the network size and reducing the cost have been proposed [9–12]. The periodic coding (PC) scheme uses a unique code to distinguish between different drop fibers (DFs) [13]. The optical encoder installed at each *
DF end is constructed by two fiber Bragg gratings (FBGs) with different reflectivity and appropriate length patchcord. The PC scheme presented in Ref. [13] is a typical technique with simple and low-cost design. However, the insertion loss can be 4.2 dB for PC encoders with two identical center reflected wavelengths due to partial reflection (i.e., 38% reflectivity) of the first FBG. In addition, the infinite-length periodic sequence with poor correlation characteristics increases the difficultly in the recognition process, i.e., localization of the encoders at the center office (CO). The reduced complexity maximum likelihood sequence estimation (RC-MLSE) algorithm used in the PC scheme is difficult to apply in practical PON systems [14]. Brillouin OTDR (BOTDR) based monitoring technique exploits the Brillouin frequency shift (BFS) to distinguish the backscattered signals from each DF. Due to the special dopant concentration for each DF, this technique requires a dramatic change in current existing PON infrastructure [15]. Obviously, it drastically increases the CapEx and is not welcomed in the cost-sensitive PON market. The monitoring technique based on the optical frequency domain reflectometer (OFDR) uses a frequency-modulated continuouswave as the probe signal. The interference signal with a unique beat frequency created by each interferometer (IF) unit is used to distinguish each other [16]. Each IF unit including a coupler, a FBG, a mirror and the patchcord with appropriate fiber length may be complex and not conducive to reduce the cost. For the self-injection locked reflective semiconductor optical amplifier (SL-RSOA) based monitoring technique, an upstream transmitter utilizing SL-RSOA can generate both upstream data and probe signals [17]. However, this technique requires a protocol
Corresponding author. E-mail address: [email protected] (X. Zhang).
https://doi.org/10.1016/j.optcom.2018.02.035 Received 30 August 2017; Received in revised form 13 February 2018; Accepted 13 February 2018 0030-4018/© 2018 Elsevier B.V. All rights reserved.
X. Zhang et al.
Optics Communications 417 (2018) 19–23
extension to avoid the interference between the probe data and the upstream data, which increases the complexity of the PON system. In this paper, we propose a self-match based on polling scheme for PON monitoring. The status of all DF links is identified one by one using a specific probe signal, which is so-called polling. Undeniably, this design examines only one DF at a time, which creates a delay between consecutive probes, especially for large network users. The patchcord length of each optical matcher (OM) is separately configured to match with the pulse interval of the corresponding probe signal. As a direct consequence, only one OM at a time makes the pulse interval of the corresponding pulse signal to become zero, which is referred to as self-match. In a PON with non-equidistance, the status of each DF link needs only be identified by the presence of the corresponding overlapped pulse. Here, non-equidistance means that all optical network units (ONUs) are distributed randomly. That is, the distance between the optical line terminal (OLT) and all ONUs is different. Compared with the PC encoders, the proposed scheme incurs almost no insertion loss because two FBGs with different center reflected wavelengths in each OM are 100% reflectivity. The recognition process in the proposed scheme mainly involves amplitude detection, which is much simpler than the auto-correlation detection used in the PC scheme. In a PON with non-equidistance, the overlapped pulse is easy to distinguish from other pulses. Especially, due to the round trip of the probe signal, the loss budget may be huge in a PON with a high splitting ratio. The power loss can be compensated by the overlapped pulse to some extent, which is conductive to reduce the sensitivity of the photodetector (PD). Obviously, the proposed scheme is simpler and cheaper when compared with the above three techniques (i.e., BOTDR, OFDR and SL-RSOA based monitoring techniques).
Fig. 1. Principle of the self-match based on polling scheme in a PON.
all OMs. It is also worth noting that the high number of components included in a coding device and their assembly increase the manufacturing, installation, and inventory cost. To ensure that only one completely overlapped pulse is generated by the corresponding OM at a time, the relation between the length 𝑙𝑘 of the patchcord and the pulse interval 𝜏𝑘 of the probe signal must satisfy: 𝜏𝑘 = 2𝑙𝑘 𝑛𝑔 ∕𝑐, where 𝑐 is the speed of light in a vacuum and 𝑛𝑔 is the effective group index. Here, we also give some brief comparison and analysis between the OM and a single FBG at all DF terminations [18]. It seems more effective or cheaper to configure a single FBG at all DF terminations, however, this design may increase the overall cost of the PON monitoring system. That is, a simple structure of the DF terminal may increase the complexity of other modules of the PON monitoring system, i.e., the transmitter and receiver modules. Note that any PON monitoring system must carefully consider the cost factor because the PON market is costsensitive. A single FBG at all DF terminations used in the prior presented PON monitoring schemes includes two cases: (a) All FBGs at all DF terminations are identical (i.e., the same center reflected wavelength and bandwidth), for example, RR-OTDR based monitoring technique. Note that this technique requires an OTDR. OTDR is a precision instrument that integrates transmitter and receiver; however, it is usually expensive. For the RR-OTDR based monitoring technique, dead zones created by a high Fresnel reflection in the field often mask the faults. (b) All FBGs at all DF terminations are different (i.e., different center reflected wavelengths), for example, T-OTDR based monitoring technique. Different FBGs that cannot be shared by users are detrimental to mass production and result in high costs. In addition, the limited spectrum of very expensive T-OTDR is difficult to use in the high capacity PON monitoring. For the proposed scheme, two FBGs with 100% reflectivity and different center reflected wavelengths are configured for all DF terminations. This design may be conductive to mass production. In addition, the simple transmitter and receiver modules can also effectively reduce the overall cost of the corresponding PON monitoring system. Due to the simple signal characteristics created by each OM, the proposed scheme only checks the amplitude of the overlapped pulse in the final recognition process. Hence, the PON’s electronic impulse response can be faster than a pre-recorded healthy response used in the most of prior works (i.e., the PC scheme). The status of all DFs can be directly identified by the threshold. For many prior works, the return signals need to be first converted into the recognizable codes and then compared with the pre-recorded healthy codes to identify the status of each DF. During this process, it may involve a lot of computing due to the complicated network recognition algorithm (i.e., RC-MLSE used in the PC scheme). For all OMs, FBGs with the same center reflected wavelengths are located on the same side of the patchcord. When the wavelength of the first subpulse transmitted from the detecting source is 𝜆1 (𝜆2 ), the FBG with the center reflected wavelength of 𝜆1 (𝜆2 ) must be physically closer to the ONU. In addition, the bandwidth of all FBGs should be much
2. Self-match based polling scheme 2.1. Principle of operation Fig. 1 illustrates the principle of the self-match based on polling scheme for centralized monitoring in a PON. The U-band (1625– 1675 nm) detecting source contains a laser array and is directly modulated by a pulse generator (PG). Each probe signal consists of a unique pulse interval and two subpulses with center wavelengths of 𝜆1 and 𝜆2 . The first probe signal with pulse interval of 𝜏1 passes through a circulator (CIR) and is coupled with the traffic data via a wavelength division multiplexer (WDM). Then, the probe and data signals are transmitted from the feeder fiber (FF) down to the network. The probe signal splits into n subsequences by the power splitter/combiner (PSC) at the remote node (RN), dispatched through the DFs to all ONUs. Each DF link is terminated by an OM that generates a corresponding reflected signal. The total received optical signal from different DF links is converted to an electrical signal by the PD located physically close to the OLT. Finally, the processing unit checks the overlapped pulse in the total received signal and identifies the corresponding DF link status. The control unit gives the detecting source a triggering signal after the processing unit completes the identification. Next, the second probe signal with the pulse interval of 𝜏2 is injected into the network and the processing unit identify the corresponding DF link status, and so on, until the last probe signal with the pulse interval of 𝜏𝑛 is injected into the network for the identification of the last DF link status. 2.2. Optical matcher design As shown in Fig. 1, the OM installed in the front of each ONU contains a pair of FBGs and the required length patchcord. The center reflected wavelengths and reflectivity of these two FBGs are 𝜆1 , 𝜆2 and 100%, respectively. The patchcord of 𝑙𝑘 (𝑘 = 1, 2, … , 𝑛) is used to connect these two FBGs. Note that 𝑙𝑘 needs to be unique to respectively match with the pulse interval of the probe signal. Due to the high reflectivity of FBGs, the insertion loss may be rarely introduced for 20
X. Zhang et al.
Optics Communications 417 (2018) 19–23
(a) When the reflected subpulse 𝜆1 appears after the reflected subpulse 𝜆2 , as illustrated in case I of Fig. 2, the position 𝑃𝑖′ of the reflected subpulse 𝜆1 in the time domain can be written as: 𝑃𝑖′ = 𝑃𝑖 − (𝑙𝑚 − 2𝑙𝑘 ).
(1)
(b) When the reflected subpulse 𝜆1 appears before the reflected subpulse 𝜆2 , as illustrated in case II of Fig. 2, the position 𝑃𝑖′ of the reflected subpulse 𝜆1 in the time domain can be written as: 𝑃𝑖′ = 𝑃𝑖 + (2𝑙𝑘 − 𝑙𝑚 ).
Fig. 2. Schematic diagram of the reflected pulses in the time-domain.
(2)
Combined with the above two cases, the position of the reflected subpulse 𝜆1 corresponding to the 𝑖th DF link in the time domain can be uniformly expressed as:
wider than the linewidth of the probe signals, which ensures that the overlapped signals generated by all OMs can be fully reflected back to the receiver. Note that all OMs used in the proposed scheme are usually fabricated on an extra fiber instead of the entire DF link from the RN to the ONU. Then, these two can be connected by the fiber connector (i.e., FC/PC, FC/UPC, FC/APC, etc.). That is, the connection between these two is active rather than fixed. Therefore, it is easy and convenient for the PON monitoring system to be replaced, added, or upgraded.
𝑃𝑖′ = 𝑃𝑖 + 2𝑙𝑘 − 𝑙𝑚 .
(3)
Similarly, the position of the reflected subpulse 𝜆1 corresponding to the 𝑗th DF link in the time domain can be expressed as: 𝑃𝑗′ = 𝑃𝑗 + 2𝑙𝑘 − 𝑙𝑚 .
(4)
In a PON with non-equidistance, when the reflected pulses generated | | | | by any two different OMs satisfy |𝑃𝑖 − 𝑃𝑗′ | = 𝛥𝐿 or |𝑃𝑖′ − 𝑃𝑗 | = 𝛥𝐿 (For | | | | 𝜆1 ,𝜆2 simplicity, the two are unified as 𝑃𝑖,𝑗 = 𝛥𝐿), the reflected pulses corresponding to 𝜆1 and 𝜆2 returned from two different DF links may generate an overlapped pulse in the FF. According to the above analysis, 𝛥𝐿 can be defined as the position difference between two reflected pulses returned from different DF links in the time domain. For example, 𝛥𝐿 = 0 shows that the reflected pulses returned from two different DF links completely overlap in the time domain.
2.3. Analysis of time-domain relation between reflected pulses Usually, only one overlapped pulse generated by the OM may appear in the total received signal. However, due to the random geographical distribution of users, the similarly overlapped pulse may be also generated by the combined signal from different DF links. For example, the overlapped pulse with the same wavelength is generated when two equidistant DF links exist in a PON. Note that this scenario is entirely possible in theory. Here, we do not deliberately emphasize this scenario and start the analysis from non-equidistant case. In a PON with non-equidistance, the reflected pulse with wavelength of 𝜆1 (or 𝜆2 ) returned from the 𝑖th DF link may overlap with the reflected pulse with wavelength of 𝜆2 (or 𝜆1 ) returned from the 𝑗th DF link. Consequently, it can interfere with the overlapped pulse generated by the OM due to the same signal characteristics. As shown in Fig. 2, the pulse interval of the probe signal is defined as 𝜏𝑚 , corresponding to 𝑙𝑚 in spatial length (𝜏𝑚 = 𝑙𝑚 𝑛𝑔 ∕𝑐). The relative position of the reflected pulses between 𝜆1 and 𝜆2 in the time domain varies with the pulse interval and patchcord length. Note that the fiber length of all DF links are fixed when the network laying is completed. That is, the physical location of the FBG2 with the center reflected wavelength of 𝜆2 is fixed (Assume that all OMs keep the same distance from ONUs and the FBG2 is further away from the ONU). Then, the position of the FBG2 in the time domain can be selected as the reference point. The physical location of the FBG1 with the center reflected wavelength of 𝜆1 depends on the patchcord length of the OM. Assuming the patchcord length is 𝑙𝑘 , we can discuss it from two aspects: (1) when 𝑙𝑚 is greater than 2𝑙𝑘 , the first incident subpulse 𝜆1 has been reflected back by the FBG1 and repasses through the FBG2 as the later incident subpulse 𝜆2 has not yet reached the FBG2 . During this time, the relative position of these two subpulses continues to change until the later incident subpulse 𝜆2 is reflected back by the FBG2 . The relatively fixed position of these two subpulses is 𝑙𝑚 − 2𝑙𝑘 ; (2) when 𝑙𝑚 is less than or equal to 2𝑙𝑘 , the later incident subpulse 𝜆2 has been reflected back by the FBG2 and repasses through the FBG1 as the first incident subpulse 𝜆1 has not yet reached the FBG1 . The relatively fixed position of these two subpulses is 2𝑙𝑘 − 𝑙𝑚 . Specially, these two subpulses overlap when 𝑙𝑚 is equal to 2𝑙𝑛 . Consider a geographical distribution of users over a round coverage region with area of 𝑀 km2 . The pulse widths of the probe signal are 𝑇0 . The position of the 𝑖th and 𝑗th DF links corresponding to the FBG2 in the time domain are 𝑃𝑖 and 𝑃𝑗 , respectively. Based on the previous analysis, the reflected subpulse 𝜆1 may appear before or after the reflected subpulse 𝜆2 . Therefore, we can do further discussion:
2.4. False alarm rate False alarm rate (FAR) is defined as the probability of declaring a DF faulty when it is healthy. For the proposed scheme, each DF link status is identified by the presence of the overlapped pulse generated by the corresponding OM. Due to the random user geographical distribution, other reflected pulses returned from two different DF links may overlap together when all the reflected pulses are combined in the FF. Thus, the overlapped pulse signal generated by different DF links and the OM, respectively, may be the same in amplitude. Mathematically, the number of overlapped pulses determines the probability of misjudgment. Based on the above analysis of the time-domain relation, the FAR can be calculated as: 𝐹 𝐴𝑅 = 1 − 1+
∑𝑛−1 ∑𝑛 𝑖=1
1
𝑗=𝑖+1
𝜆 ,𝜆2
𝑁(𝑃𝑖,𝑗1
.
(5)
= 𝛥𝐿)
In Fig. 3, the FAR versus different coverage area of a PON with 64 and 128 users are shown with the solid and dotted lines, respectively. In order to better simulate the real situation, Monte Carlo simulations (105 times iteration) with randomly located users is used to calculate the FAR in the PON. That is, all DF lengths are unknown before the application of the PON monitoring system. This is why we do not deliberately emphasize two or more identical DF lengths existing in a PON. The FAR gradually decreases as the coverage area increases, which results from the probability of the corresponding reflected pulses overlapped in the time domain. In the calculation, the pulse width 𝑇0 and the step between any two adjacent pulse intervals are 1 ns. In a PON with 64 users, the patchcord length of the OMs are from 0.1 m to 6.4 m with the step of 0.1 m. When the round coverage area is 𝑀 = 1 km2 , the FAR with 𝛥𝐿 < 𝑐𝑇0 ∕𝑛𝑔 , 𝛥𝐿 ≤ 0.5𝑐𝑇0 ∕𝑛𝑔 and 𝛥𝐿 ≈ 0 are 0.5854,0.4151 and 0.0015, respectively. By contrast, the FAR with the round coverage area of 𝑀 = 10 km2 reduces to 0.3094, 0.1836 and 4.7118 × 10−4 . In a PON with 128 users, the maximum patchcord length is 12.8 m. For the same 𝛥𝐿 as above, the FAR with coverage area of 1 km2 and 10 km2 are 0.8507, 0.7405, 0.0057 and 0.6442, 0.4756, 0.0018, respectively. Note 21
X. Zhang et al.
Optics Communications 417 (2018) 19–23
Fig. 4. Experimental setup for the monitoring of a network with 4 DFs.
dense wavelength division multiplexers (DWDMs). In this experiment, the patchcord with different fiber lengths are respectively connected to two channels of the laser array for generating the desired probe signals. The probe signals with two different wavelengths transmitted from the detecting source are directly modulated into rectangular pulses. The pulse width and repetition rate of the probe signals are 1 μs and 1 kHz, respectively. The probe signal is sent into a 20 km FF via a CIR and then dropped to each DF link through a 1 × 4 PSC. Four OMs are located at the end of the DF link and have a patchcord length of 𝑙1 = 100 m, 𝑙2 = 150 m, 𝑙3 = 200 m and 𝑙4 = 250 m, respectively. According to the analysis of Section 2.2, the patchcord length used to generate the probe signals corresponding to above four OMs are 200 m, 300 m, 400 m and 500 m. The center reflected wavelengths of FBGs with 3-dB bandwidth of 0.3 nm are 𝜆1 = 1550.9 nm and 𝜆2 = 1549.3 nm, respectively. The measured traces of the total received signal are captured in the oscilloscope (OSC). The experimental results are illustrated in Fig. 5. Fig. 5(a), (b), (c) and (d) show the traces of the total received signal corresponding to DF1 , DF2 , DF3 and DF4 in the healthy case, respectively. Only one overlapped pulse with larger amplitude appears in each total received signal. Obviously, it can provide a great convenience in the recognition processing of the network link status. Compared to other reflected pulses in the total received signal, the overlapped pulse is more easily identified due to the larger amplitude. Moreover, it may reduce the sensitivity of the PD because only the overlapped pulse provides the useful information for the final recognition processing. In Fig. 5(e) and (f), we simulate the break in FF and DF2 , respectively. When a break occurs in FF, no reflected pulse signal appears, as shown in Fig. 5(e). In Fig. 5(f), the overlapped pulse is absent when the DF2 has a break. In Fig. 5(a), (b), (c) and (d), the reflected pulses are marked by the corresponding wavelengths in order to better illustrate the relation between reflected pulses in the time-domain. The position difference between two adjacent reflected pulses generated by the same OM agrees with the previous theory analysis.
Fig. 3. FAR versus different coverage area for 64 and 128 users.
that the round coverage area of 1 km2 and 10 km2 corresponding to the maximum fiber length 𝐿max of the DF links are 564 m and 1784 m, respectively. That is, all DF lengths are randomly distributed in the range between 0 to 𝐿max . In the above calculation, the FAR is closely related to 𝛥𝐿. For the same coverage area, smaller 𝛥𝐿 contributes to lower FAR. Recall that 𝛥𝐿 denotes the position difference between two reflected pulses in the time domain. 𝛥𝐿 is easily quantified in the final recognition processing when the probe signal is a rectangular pulse. For example, after a overlapped pulse waveform is acquired by the data acquisition card (DAQ), the consecutive data points number corresponding to the peak power of completely overlapped pulse (𝛥𝐿 ≈ 0) are more than that of partially overlapped pulse (𝛥𝐿 < 𝑐𝑇0 ∕𝑛𝑔 and 𝛥𝐿 ≤ 0.5𝑐𝑇0 ∕𝑛𝑔 ). That is, the pulse width corresponding to the peak power of the overlapped pulse generated by the OM differs from that generated by two different DF links. Consequently, the FAR can fall into a very low range in most cases. 2.5. Loss budget analysis In this section, we investigate the loss budget of the proposed scheme for future capacity expansion and/or longer reach PONs. A 0.3 dB/km of fiber loss is considered for U-band signals. The probe signal is injected into the network via a circulator with total insertion loss of 𝛼𝐶 . Due to the round trip of the probe signal through the network, the total insertion losses of passive components are doubled. Taking into account the other losses 𝛼𝐿 (i.e., splicing and connectors, etc.), the total loss budget in the proposed scheme can be written as (in decibels): − 𝛼𝑇 = 20 log10 𝑁 + 0.6𝐿 + 𝛼𝐿 + 𝛼𝐶
4. Conclusion We have proposed and experimentally demonstrated a self-match based on polling scheme for PON monitoring. Each network link status is identified by the presence or absence of the overlapped pulse, which greatly simplifies the final recognition processing. The analysis of the time-domain relation between reflected pulses is experimentally verified. We also establish the calculation model to evaluate the FAR. The calculation results show that the FAR can be at a very low level by quantifying the number of consecutive data points. The loss budget of the proposed scheme for future capacity expansion and/or longer reach PONs is investigated. The proposed scheme with simple and low-cost design offers a promising solution for monitoring of expanding PONs.
(6)
where 𝐿 is the fiber reach, i.e., sum of fiber feeder (FF) and longest DF. 𝑁 is the number of users in a PON. Logarithmic term is used to calculate the insertion loss of the splitter in the RN. Note that a FF of 20 km would be typical in a PON. 𝛼𝐿 and 𝛼𝐶 are set to 5 dB and 3 dB, respectively. For example, the total loss budget is roughly 63.2 dB in a PON with 128 users and the longest DF of 2 km. 3. Experimental demonstration The feasibility of the proposed scheme is experimentally demonstrated in a simplified 1 × 4 PON with non-equidistance. The experimental setup is illustrated in Fig. 4. Due to the equipment availability, the detecting source is constructed by a 6-channel (6-CH) laser array with synchronous output, the patchcord with fiber length of 𝑙𝑖 and two
Acknowledgment This work was supported by the National Natural Science Foundation of China (grant number 61367006). 22
X. Zhang et al.
Optics Communications 417 (2018) 19–23
Fig. 5. Experimental results in different cases: (a) DF1 , (b) DF2 , (c) DF3 and (d) DF4 are in the healthy case; (e) a break of FF and (f) a break of DF2 are in the faulty case.
References
[9] X. Zhou, F.D. Zhang, X.H. Sun, IEEE Photon. Technol. Lett. 25 (2013) 795. [10] X. Zhang, F.J. Lu, S. Chen, X.Q. Zhao, M. Zhu, X.H. Sun, Opt. Express 24 (2016) 4351. [11] M.A. Esmail, Habib Fathallah, IEEE Commun. Lett. 15 (2011) 677. [12] K. Yüksel, M. Wuilpart, V. Moeyaert, P. Mégret, J. Opt. Commun. Netw. 2 (2010) 463. [13] H. Fathallah, M.M. Rad, L.A. Rusch, IEEE Photon. Technol. Lett. 20 (2008) 2039. [14] M.M. Rad, H.A. Fathallah, S. LaRochelle, L.A. Rusch, J. Opt. Commun. Netw. 3 (2011) 77. [15] D. Iida, N. Honda, H. Izumita, F. Ito, J. Lightwave Technol. I 25 (2007) 1290. [16] K. Yuksel, M. Wuilpart, V. Moeyaert, P. Megret, J. Opt. Commun. Netw. 2 (2010) 463. [17] M. Thollabandi, H. Bang, K. Shim, S. Hann, C. Park, Opt. Fiber. Technol. 15 (2009) 451. [18] K. Fouli, L.R. Chen, M. Maier, Comput. Commun. 34 (2011) 1011.
[1] R. Davey, J. Kani, F. Bourgart, K. McCammon, IEEE Commun. Mag. 44 (2006) 50. [2] G.K. Chang, A. Chowdhury, Z.S. Jia, H.C. Chien, M.F. Huang, J.J. Yu, G. Ellinas, J. Opt. Commun. Netw. 1 (2009) C35. [3] M.C. Yuang, D.Z. Hsu, P.L. Tien, H.Y. Chen, C.C. Wei, S.H. Chen, J. Chen, J. Lightwave Technol. 31 (2013) 2809. [4] D. Nesset, J. Opt. Commun. Netw. 9 (2017) A71. [5] K. Yuksel, V. Moeyaert, M. Wuilpart, P. Megret, Optical layer monitoring in Passive Optical Networks (PONs): A review, in: 10th Anniversary International Conference on Transparent Optical Networks, 2008, pp. 92–98. [6] M.A. Esmail, H. Fathallah, IEEE Commun. Surv. Tut. 15 (2013) 943. [7] R. Sánchez, J.A. Hernández, D. Larrabeiti, IEEE Commun. Mag. 52 (2014) S7. [8] M.M. Rad, K. Fouli, H.A. Fathallah, L.A. Rusch, M. Maier, IEEE Commun. Mag. 49 (2011) S45.
23
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
Self-match based on polling scheme for passive optical network monitoring Xuan Zhang a, *, Hao Guo b , Xinhong Jia a , Qinghua Liao c a
College of Physics and Electronic Engineering, Sichuan Normal University, Chengdu, Sichuan, 610101, China State Key Laboratory of Electronic Thin Films and Integrated Devices, College of Optoelectronic Information, University of Electronic Science and Technology of China, Chengdu, Sichuan, 610054, China c College of Science, Nanchang University, Nanchang, Jiangxi, 330031, China b
a r t i c l e
i n f o
Keywords: Passive optical network (PON) Link monitoring Self-match Polling False alarm rate (FAR)
a b s t r a c t We propose a self-match based on polling scheme for passive optical network monitoring. Each end-user is equipped with an optical matcher that exploits only the specific length patchcord and two different fiber Bragg gratings with 100% reflectivity. The simple and low-cost scheme can greatly simplify the final recognition processing of the network link status and reduce the sensitivity of the photodetector. We analyze the timedomain relation between reflected pulses and establish the calculation model to evaluate the false alarm rate. The feasibility of the proposed scheme and the validity of the time-domain relation analysis are experimentally demonstrated.
1. Introduction Optical access network (OAN) is capable of delivering high-capacity data, consuming less energy and saving capital expenditure (CapEx), which can remove the bottleneck of the telecommunication development [1,2]. Passive optical networks (PONs) are considered one of the most promising candidates for OANs due to the high transmission rate and low-cost infrastructure [3,4]. With the rapid development of PONs, management and maintenance issues are emerging. As an effective solution, PON monitoring can reduce provisioning time, improve quality of service (QoS), attract more clients and reduce maintenance costs [5,6]. Therefore, a simple but effective monitoring scheme is essential for continuously developing PONs. Optical time domain reflectometry (OTDR) is widely used in the point-to-point (P2P) fiber link test, which can provide a plot of distance versus signal level in a fiber. However, it may be ineffective in pointto-multipoint (P2MP) networks because all returned signals coming from different branches add up together [6,7]. Some modified OTDR techniques, such as reference reflector and tunable OTDR (i.e., RROTDR and T-OTDR), are limited by the network size and expensive OTDR equipment, respectively [8]. Note that the PON market is highly sensitive to cost. Thus, the PON monitoring system must be costeffective, especially at the user terminals because the components cannot be shared. Some PON monitoring schemes aiming at improving the network size and reducing the cost have been proposed [9–12]. The periodic coding (PC) scheme uses a unique code to distinguish between different drop fibers (DFs) [13]. The optical encoder installed at each *
DF end is constructed by two fiber Bragg gratings (FBGs) with different reflectivity and appropriate length patchcord. The PC scheme presented in Ref. [13] is a typical technique with simple and low-cost design. However, the insertion loss can be 4.2 dB for PC encoders with two identical center reflected wavelengths due to partial reflection (i.e., 38% reflectivity) of the first FBG. In addition, the infinite-length periodic sequence with poor correlation characteristics increases the difficultly in the recognition process, i.e., localization of the encoders at the center office (CO). The reduced complexity maximum likelihood sequence estimation (RC-MLSE) algorithm used in the PC scheme is difficult to apply in practical PON systems [14]. Brillouin OTDR (BOTDR) based monitoring technique exploits the Brillouin frequency shift (BFS) to distinguish the backscattered signals from each DF. Due to the special dopant concentration for each DF, this technique requires a dramatic change in current existing PON infrastructure [15]. Obviously, it drastically increases the CapEx and is not welcomed in the cost-sensitive PON market. The monitoring technique based on the optical frequency domain reflectometer (OFDR) uses a frequency-modulated continuouswave as the probe signal. The interference signal with a unique beat frequency created by each interferometer (IF) unit is used to distinguish each other [16]. Each IF unit including a coupler, a FBG, a mirror and the patchcord with appropriate fiber length may be complex and not conducive to reduce the cost. For the self-injection locked reflective semiconductor optical amplifier (SL-RSOA) based monitoring technique, an upstream transmitter utilizing SL-RSOA can generate both upstream data and probe signals [17]. However, this technique requires a protocol
Corresponding author. E-mail address: [email protected] (X. Zhang).
https://doi.org/10.1016/j.optcom.2018.02.035 Received 30 August 2017; Received in revised form 13 February 2018; Accepted 13 February 2018 0030-4018/© 2018 Elsevier B.V. All rights reserved.
X. Zhang et al.
Optics Communications 417 (2018) 19–23
extension to avoid the interference between the probe data and the upstream data, which increases the complexity of the PON system. In this paper, we propose a self-match based on polling scheme for PON monitoring. The status of all DF links is identified one by one using a specific probe signal, which is so-called polling. Undeniably, this design examines only one DF at a time, which creates a delay between consecutive probes, especially for large network users. The patchcord length of each optical matcher (OM) is separately configured to match with the pulse interval of the corresponding probe signal. As a direct consequence, only one OM at a time makes the pulse interval of the corresponding pulse signal to become zero, which is referred to as self-match. In a PON with non-equidistance, the status of each DF link needs only be identified by the presence of the corresponding overlapped pulse. Here, non-equidistance means that all optical network units (ONUs) are distributed randomly. That is, the distance between the optical line terminal (OLT) and all ONUs is different. Compared with the PC encoders, the proposed scheme incurs almost no insertion loss because two FBGs with different center reflected wavelengths in each OM are 100% reflectivity. The recognition process in the proposed scheme mainly involves amplitude detection, which is much simpler than the auto-correlation detection used in the PC scheme. In a PON with non-equidistance, the overlapped pulse is easy to distinguish from other pulses. Especially, due to the round trip of the probe signal, the loss budget may be huge in a PON with a high splitting ratio. The power loss can be compensated by the overlapped pulse to some extent, which is conductive to reduce the sensitivity of the photodetector (PD). Obviously, the proposed scheme is simpler and cheaper when compared with the above three techniques (i.e., BOTDR, OFDR and SL-RSOA based monitoring techniques).
Fig. 1. Principle of the self-match based on polling scheme in a PON.
all OMs. It is also worth noting that the high number of components included in a coding device and their assembly increase the manufacturing, installation, and inventory cost. To ensure that only one completely overlapped pulse is generated by the corresponding OM at a time, the relation between the length 𝑙𝑘 of the patchcord and the pulse interval 𝜏𝑘 of the probe signal must satisfy: 𝜏𝑘 = 2𝑙𝑘 𝑛𝑔 ∕𝑐, where 𝑐 is the speed of light in a vacuum and 𝑛𝑔 is the effective group index. Here, we also give some brief comparison and analysis between the OM and a single FBG at all DF terminations [18]. It seems more effective or cheaper to configure a single FBG at all DF terminations, however, this design may increase the overall cost of the PON monitoring system. That is, a simple structure of the DF terminal may increase the complexity of other modules of the PON monitoring system, i.e., the transmitter and receiver modules. Note that any PON monitoring system must carefully consider the cost factor because the PON market is costsensitive. A single FBG at all DF terminations used in the prior presented PON monitoring schemes includes two cases: (a) All FBGs at all DF terminations are identical (i.e., the same center reflected wavelength and bandwidth), for example, RR-OTDR based monitoring technique. Note that this technique requires an OTDR. OTDR is a precision instrument that integrates transmitter and receiver; however, it is usually expensive. For the RR-OTDR based monitoring technique, dead zones created by a high Fresnel reflection in the field often mask the faults. (b) All FBGs at all DF terminations are different (i.e., different center reflected wavelengths), for example, T-OTDR based monitoring technique. Different FBGs that cannot be shared by users are detrimental to mass production and result in high costs. In addition, the limited spectrum of very expensive T-OTDR is difficult to use in the high capacity PON monitoring. For the proposed scheme, two FBGs with 100% reflectivity and different center reflected wavelengths are configured for all DF terminations. This design may be conductive to mass production. In addition, the simple transmitter and receiver modules can also effectively reduce the overall cost of the corresponding PON monitoring system. Due to the simple signal characteristics created by each OM, the proposed scheme only checks the amplitude of the overlapped pulse in the final recognition process. Hence, the PON’s electronic impulse response can be faster than a pre-recorded healthy response used in the most of prior works (i.e., the PC scheme). The status of all DFs can be directly identified by the threshold. For many prior works, the return signals need to be first converted into the recognizable codes and then compared with the pre-recorded healthy codes to identify the status of each DF. During this process, it may involve a lot of computing due to the complicated network recognition algorithm (i.e., RC-MLSE used in the PC scheme). For all OMs, FBGs with the same center reflected wavelengths are located on the same side of the patchcord. When the wavelength of the first subpulse transmitted from the detecting source is 𝜆1 (𝜆2 ), the FBG with the center reflected wavelength of 𝜆1 (𝜆2 ) must be physically closer to the ONU. In addition, the bandwidth of all FBGs should be much
2. Self-match based polling scheme 2.1. Principle of operation Fig. 1 illustrates the principle of the self-match based on polling scheme for centralized monitoring in a PON. The U-band (1625– 1675 nm) detecting source contains a laser array and is directly modulated by a pulse generator (PG). Each probe signal consists of a unique pulse interval and two subpulses with center wavelengths of 𝜆1 and 𝜆2 . The first probe signal with pulse interval of 𝜏1 passes through a circulator (CIR) and is coupled with the traffic data via a wavelength division multiplexer (WDM). Then, the probe and data signals are transmitted from the feeder fiber (FF) down to the network. The probe signal splits into n subsequences by the power splitter/combiner (PSC) at the remote node (RN), dispatched through the DFs to all ONUs. Each DF link is terminated by an OM that generates a corresponding reflected signal. The total received optical signal from different DF links is converted to an electrical signal by the PD located physically close to the OLT. Finally, the processing unit checks the overlapped pulse in the total received signal and identifies the corresponding DF link status. The control unit gives the detecting source a triggering signal after the processing unit completes the identification. Next, the second probe signal with the pulse interval of 𝜏2 is injected into the network and the processing unit identify the corresponding DF link status, and so on, until the last probe signal with the pulse interval of 𝜏𝑛 is injected into the network for the identification of the last DF link status. 2.2. Optical matcher design As shown in Fig. 1, the OM installed in the front of each ONU contains a pair of FBGs and the required length patchcord. The center reflected wavelengths and reflectivity of these two FBGs are 𝜆1 , 𝜆2 and 100%, respectively. The patchcord of 𝑙𝑘 (𝑘 = 1, 2, … , 𝑛) is used to connect these two FBGs. Note that 𝑙𝑘 needs to be unique to respectively match with the pulse interval of the probe signal. Due to the high reflectivity of FBGs, the insertion loss may be rarely introduced for 20
X. Zhang et al.
Optics Communications 417 (2018) 19–23
(a) When the reflected subpulse 𝜆1 appears after the reflected subpulse 𝜆2 , as illustrated in case I of Fig. 2, the position 𝑃𝑖′ of the reflected subpulse 𝜆1 in the time domain can be written as: 𝑃𝑖′ = 𝑃𝑖 − (𝑙𝑚 − 2𝑙𝑘 ).
(1)
(b) When the reflected subpulse 𝜆1 appears before the reflected subpulse 𝜆2 , as illustrated in case II of Fig. 2, the position 𝑃𝑖′ of the reflected subpulse 𝜆1 in the time domain can be written as: 𝑃𝑖′ = 𝑃𝑖 + (2𝑙𝑘 − 𝑙𝑚 ).
Fig. 2. Schematic diagram of the reflected pulses in the time-domain.
(2)
Combined with the above two cases, the position of the reflected subpulse 𝜆1 corresponding to the 𝑖th DF link in the time domain can be uniformly expressed as:
wider than the linewidth of the probe signals, which ensures that the overlapped signals generated by all OMs can be fully reflected back to the receiver. Note that all OMs used in the proposed scheme are usually fabricated on an extra fiber instead of the entire DF link from the RN to the ONU. Then, these two can be connected by the fiber connector (i.e., FC/PC, FC/UPC, FC/APC, etc.). That is, the connection between these two is active rather than fixed. Therefore, it is easy and convenient for the PON monitoring system to be replaced, added, or upgraded.
𝑃𝑖′ = 𝑃𝑖 + 2𝑙𝑘 − 𝑙𝑚 .
(3)
Similarly, the position of the reflected subpulse 𝜆1 corresponding to the 𝑗th DF link in the time domain can be expressed as: 𝑃𝑗′ = 𝑃𝑗 + 2𝑙𝑘 − 𝑙𝑚 .
(4)
In a PON with non-equidistance, when the reflected pulses generated | | | | by any two different OMs satisfy |𝑃𝑖 − 𝑃𝑗′ | = 𝛥𝐿 or |𝑃𝑖′ − 𝑃𝑗 | = 𝛥𝐿 (For | | | | 𝜆1 ,𝜆2 simplicity, the two are unified as 𝑃𝑖,𝑗 = 𝛥𝐿), the reflected pulses corresponding to 𝜆1 and 𝜆2 returned from two different DF links may generate an overlapped pulse in the FF. According to the above analysis, 𝛥𝐿 can be defined as the position difference between two reflected pulses returned from different DF links in the time domain. For example, 𝛥𝐿 = 0 shows that the reflected pulses returned from two different DF links completely overlap in the time domain.
2.3. Analysis of time-domain relation between reflected pulses Usually, only one overlapped pulse generated by the OM may appear in the total received signal. However, due to the random geographical distribution of users, the similarly overlapped pulse may be also generated by the combined signal from different DF links. For example, the overlapped pulse with the same wavelength is generated when two equidistant DF links exist in a PON. Note that this scenario is entirely possible in theory. Here, we do not deliberately emphasize this scenario and start the analysis from non-equidistant case. In a PON with non-equidistance, the reflected pulse with wavelength of 𝜆1 (or 𝜆2 ) returned from the 𝑖th DF link may overlap with the reflected pulse with wavelength of 𝜆2 (or 𝜆1 ) returned from the 𝑗th DF link. Consequently, it can interfere with the overlapped pulse generated by the OM due to the same signal characteristics. As shown in Fig. 2, the pulse interval of the probe signal is defined as 𝜏𝑚 , corresponding to 𝑙𝑚 in spatial length (𝜏𝑚 = 𝑙𝑚 𝑛𝑔 ∕𝑐). The relative position of the reflected pulses between 𝜆1 and 𝜆2 in the time domain varies with the pulse interval and patchcord length. Note that the fiber length of all DF links are fixed when the network laying is completed. That is, the physical location of the FBG2 with the center reflected wavelength of 𝜆2 is fixed (Assume that all OMs keep the same distance from ONUs and the FBG2 is further away from the ONU). Then, the position of the FBG2 in the time domain can be selected as the reference point. The physical location of the FBG1 with the center reflected wavelength of 𝜆1 depends on the patchcord length of the OM. Assuming the patchcord length is 𝑙𝑘 , we can discuss it from two aspects: (1) when 𝑙𝑚 is greater than 2𝑙𝑘 , the first incident subpulse 𝜆1 has been reflected back by the FBG1 and repasses through the FBG2 as the later incident subpulse 𝜆2 has not yet reached the FBG2 . During this time, the relative position of these two subpulses continues to change until the later incident subpulse 𝜆2 is reflected back by the FBG2 . The relatively fixed position of these two subpulses is 𝑙𝑚 − 2𝑙𝑘 ; (2) when 𝑙𝑚 is less than or equal to 2𝑙𝑘 , the later incident subpulse 𝜆2 has been reflected back by the FBG2 and repasses through the FBG1 as the first incident subpulse 𝜆1 has not yet reached the FBG1 . The relatively fixed position of these two subpulses is 2𝑙𝑘 − 𝑙𝑚 . Specially, these two subpulses overlap when 𝑙𝑚 is equal to 2𝑙𝑛 . Consider a geographical distribution of users over a round coverage region with area of 𝑀 km2 . The pulse widths of the probe signal are 𝑇0 . The position of the 𝑖th and 𝑗th DF links corresponding to the FBG2 in the time domain are 𝑃𝑖 and 𝑃𝑗 , respectively. Based on the previous analysis, the reflected subpulse 𝜆1 may appear before or after the reflected subpulse 𝜆2 . Therefore, we can do further discussion:
2.4. False alarm rate False alarm rate (FAR) is defined as the probability of declaring a DF faulty when it is healthy. For the proposed scheme, each DF link status is identified by the presence of the overlapped pulse generated by the corresponding OM. Due to the random user geographical distribution, other reflected pulses returned from two different DF links may overlap together when all the reflected pulses are combined in the FF. Thus, the overlapped pulse signal generated by different DF links and the OM, respectively, may be the same in amplitude. Mathematically, the number of overlapped pulses determines the probability of misjudgment. Based on the above analysis of the time-domain relation, the FAR can be calculated as: 𝐹 𝐴𝑅 = 1 − 1+
∑𝑛−1 ∑𝑛 𝑖=1
1
𝑗=𝑖+1
𝜆 ,𝜆2
𝑁(𝑃𝑖,𝑗1
.
(5)
= 𝛥𝐿)
In Fig. 3, the FAR versus different coverage area of a PON with 64 and 128 users are shown with the solid and dotted lines, respectively. In order to better simulate the real situation, Monte Carlo simulations (105 times iteration) with randomly located users is used to calculate the FAR in the PON. That is, all DF lengths are unknown before the application of the PON monitoring system. This is why we do not deliberately emphasize two or more identical DF lengths existing in a PON. The FAR gradually decreases as the coverage area increases, which results from the probability of the corresponding reflected pulses overlapped in the time domain. In the calculation, the pulse width 𝑇0 and the step between any two adjacent pulse intervals are 1 ns. In a PON with 64 users, the patchcord length of the OMs are from 0.1 m to 6.4 m with the step of 0.1 m. When the round coverage area is 𝑀 = 1 km2 , the FAR with 𝛥𝐿 < 𝑐𝑇0 ∕𝑛𝑔 , 𝛥𝐿 ≤ 0.5𝑐𝑇0 ∕𝑛𝑔 and 𝛥𝐿 ≈ 0 are 0.5854,0.4151 and 0.0015, respectively. By contrast, the FAR with the round coverage area of 𝑀 = 10 km2 reduces to 0.3094, 0.1836 and 4.7118 × 10−4 . In a PON with 128 users, the maximum patchcord length is 12.8 m. For the same 𝛥𝐿 as above, the FAR with coverage area of 1 km2 and 10 km2 are 0.8507, 0.7405, 0.0057 and 0.6442, 0.4756, 0.0018, respectively. Note 21
X. Zhang et al.
Optics Communications 417 (2018) 19–23
Fig. 4. Experimental setup for the monitoring of a network with 4 DFs.
dense wavelength division multiplexers (DWDMs). In this experiment, the patchcord with different fiber lengths are respectively connected to two channels of the laser array for generating the desired probe signals. The probe signals with two different wavelengths transmitted from the detecting source are directly modulated into rectangular pulses. The pulse width and repetition rate of the probe signals are 1 μs and 1 kHz, respectively. The probe signal is sent into a 20 km FF via a CIR and then dropped to each DF link through a 1 × 4 PSC. Four OMs are located at the end of the DF link and have a patchcord length of 𝑙1 = 100 m, 𝑙2 = 150 m, 𝑙3 = 200 m and 𝑙4 = 250 m, respectively. According to the analysis of Section 2.2, the patchcord length used to generate the probe signals corresponding to above four OMs are 200 m, 300 m, 400 m and 500 m. The center reflected wavelengths of FBGs with 3-dB bandwidth of 0.3 nm are 𝜆1 = 1550.9 nm and 𝜆2 = 1549.3 nm, respectively. The measured traces of the total received signal are captured in the oscilloscope (OSC). The experimental results are illustrated in Fig. 5. Fig. 5(a), (b), (c) and (d) show the traces of the total received signal corresponding to DF1 , DF2 , DF3 and DF4 in the healthy case, respectively. Only one overlapped pulse with larger amplitude appears in each total received signal. Obviously, it can provide a great convenience in the recognition processing of the network link status. Compared to other reflected pulses in the total received signal, the overlapped pulse is more easily identified due to the larger amplitude. Moreover, it may reduce the sensitivity of the PD because only the overlapped pulse provides the useful information for the final recognition processing. In Fig. 5(e) and (f), we simulate the break in FF and DF2 , respectively. When a break occurs in FF, no reflected pulse signal appears, as shown in Fig. 5(e). In Fig. 5(f), the overlapped pulse is absent when the DF2 has a break. In Fig. 5(a), (b), (c) and (d), the reflected pulses are marked by the corresponding wavelengths in order to better illustrate the relation between reflected pulses in the time-domain. The position difference between two adjacent reflected pulses generated by the same OM agrees with the previous theory analysis.
Fig. 3. FAR versus different coverage area for 64 and 128 users.
that the round coverage area of 1 km2 and 10 km2 corresponding to the maximum fiber length 𝐿max of the DF links are 564 m and 1784 m, respectively. That is, all DF lengths are randomly distributed in the range between 0 to 𝐿max . In the above calculation, the FAR is closely related to 𝛥𝐿. For the same coverage area, smaller 𝛥𝐿 contributes to lower FAR. Recall that 𝛥𝐿 denotes the position difference between two reflected pulses in the time domain. 𝛥𝐿 is easily quantified in the final recognition processing when the probe signal is a rectangular pulse. For example, after a overlapped pulse waveform is acquired by the data acquisition card (DAQ), the consecutive data points number corresponding to the peak power of completely overlapped pulse (𝛥𝐿 ≈ 0) are more than that of partially overlapped pulse (𝛥𝐿 < 𝑐𝑇0 ∕𝑛𝑔 and 𝛥𝐿 ≤ 0.5𝑐𝑇0 ∕𝑛𝑔 ). That is, the pulse width corresponding to the peak power of the overlapped pulse generated by the OM differs from that generated by two different DF links. Consequently, the FAR can fall into a very low range in most cases. 2.5. Loss budget analysis In this section, we investigate the loss budget of the proposed scheme for future capacity expansion and/or longer reach PONs. A 0.3 dB/km of fiber loss is considered for U-band signals. The probe signal is injected into the network via a circulator with total insertion loss of 𝛼𝐶 . Due to the round trip of the probe signal through the network, the total insertion losses of passive components are doubled. Taking into account the other losses 𝛼𝐿 (i.e., splicing and connectors, etc.), the total loss budget in the proposed scheme can be written as (in decibels): − 𝛼𝑇 = 20 log10 𝑁 + 0.6𝐿 + 𝛼𝐿 + 𝛼𝐶
4. Conclusion We have proposed and experimentally demonstrated a self-match based on polling scheme for PON monitoring. Each network link status is identified by the presence or absence of the overlapped pulse, which greatly simplifies the final recognition processing. The analysis of the time-domain relation between reflected pulses is experimentally verified. We also establish the calculation model to evaluate the FAR. The calculation results show that the FAR can be at a very low level by quantifying the number of consecutive data points. The loss budget of the proposed scheme for future capacity expansion and/or longer reach PONs is investigated. The proposed scheme with simple and low-cost design offers a promising solution for monitoring of expanding PONs.
(6)
where 𝐿 is the fiber reach, i.e., sum of fiber feeder (FF) and longest DF. 𝑁 is the number of users in a PON. Logarithmic term is used to calculate the insertion loss of the splitter in the RN. Note that a FF of 20 km would be typical in a PON. 𝛼𝐿 and 𝛼𝐶 are set to 5 dB and 3 dB, respectively. For example, the total loss budget is roughly 63.2 dB in a PON with 128 users and the longest DF of 2 km. 3. Experimental demonstration The feasibility of the proposed scheme is experimentally demonstrated in a simplified 1 × 4 PON with non-equidistance. The experimental setup is illustrated in Fig. 4. Due to the equipment availability, the detecting source is constructed by a 6-channel (6-CH) laser array with synchronous output, the patchcord with fiber length of 𝑙𝑖 and two
Acknowledgment This work was supported by the National Natural Science Foundation of China (grant number 61367006). 22
X. Zhang et al.
Optics Communications 417 (2018) 19–23
Fig. 5. Experimental results in different cases: (a) DF1 , (b) DF2 , (c) DF3 and (d) DF4 are in the healthy case; (e) a break of FF and (f) a break of DF2 are in the faulty case.
References
[9] X. Zhou, F.D. Zhang, X.H. Sun, IEEE Photon. Technol. Lett. 25 (2013) 795. [10] X. Zhang, F.J. Lu, S. Chen, X.Q. Zhao, M. Zhu, X.H. Sun, Opt. Express 24 (2016) 4351. [11] M.A. Esmail, Habib Fathallah, IEEE Commun. Lett. 15 (2011) 677. [12] K. Yüksel, M. Wuilpart, V. Moeyaert, P. Mégret, J. Opt. Commun. Netw. 2 (2010) 463. [13] H. Fathallah, M.M. Rad, L.A. Rusch, IEEE Photon. Technol. Lett. 20 (2008) 2039. [14] M.M. Rad, H.A. Fathallah, S. LaRochelle, L.A. Rusch, J. Opt. Commun. Netw. 3 (2011) 77. [15] D. Iida, N. Honda, H. Izumita, F. Ito, J. Lightwave Technol. I 25 (2007) 1290. [16] K. Yuksel, M. Wuilpart, V. Moeyaert, P. Megret, J. Opt. Commun. Netw. 2 (2010) 463. [17] M. Thollabandi, H. Bang, K. Shim, S. Hann, C. Park, Opt. Fiber. Technol. 15 (2009) 451. [18] K. Fouli, L.R. Chen, M. Maier, Comput. Commun. 34 (2011) 1011.
[1] R. Davey, J. Kani, F. Bourgart, K. McCammon, IEEE Commun. Mag. 44 (2006) 50. [2] G.K. Chang, A. Chowdhury, Z.S. Jia, H.C. Chien, M.F. Huang, J.J. Yu, G. Ellinas, J. Opt. Commun. Netw. 1 (2009) C35. [3] M.C. Yuang, D.Z. Hsu, P.L. Tien, H.Y. Chen, C.C. Wei, S.H. Chen, J. Chen, J. Lightwave Technol. 31 (2013) 2809. [4] D. Nesset, J. Opt. Commun. Netw. 9 (2017) A71. [5] K. Yuksel, V. Moeyaert, M. Wuilpart, P. Megret, Optical layer monitoring in Passive Optical Networks (PONs): A review, in: 10th Anniversary International Conference on Transparent Optical Networks, 2008, pp. 92–98. [6] M.A. Esmail, H. Fathallah, IEEE Commun. Surv. Tut. 15 (2013) 943. [7] R. Sánchez, J.A. Hernández, D. Larrabeiti, IEEE Commun. Mag. 52 (2014) S7. [8] M.M. Rad, K. Fouli, H.A. Fathallah, L.A. Rusch, M. Maier, IEEE Commun. Mag. 49 (2011) S45.
23