Open Access
Add to BibSonomyAbstract
This paper presents a novel and simple fiber monitoring system based on multi-wavelength transmission-reflection analysis for long-reach time and wavelength division multiplexing passive optical networks. For the first time, the full localization functionality of long-reach passive optical networks is possible with the proposed monitoring scheme, including supporting fault detection, identification, and localization in both feeder and distribution fiber segments. By measuring the transmitted and reflected/backscattered optical powers launched by an unmodulated continuous-wave optical source, the proposed solution is able to supervise the network with good spatial accuracy, a high detection speed and a low impact on data traffic. Both the theoretical analysis and experimental validation show that the proposed scheme is capable of providing an accurate fault monitoring functionality for long-reach time and wavelength division multiplexing passive optical networks.
© 2016 Optical Society of America
1. Introduction
Long-Reach Passive Optical Networks (LR-PONs) extend the distance between the Central Office (CO) and the end users from the traditional range of 20km to 100km and beyond, leading to the consolidation of metropolitan and access networks [1,2]. This in turn simplifies the network architecture by reducing the number of COs in the field, potentially bringing both infrastructure and maintenance cost savings [3]. Among different topologies for LR-PONs, the “ring and spur” approach (see Fig. 1), is currently subject to extensive research [1,4,5]. In such LR-PONs, a cable ring is employed as the feeder section, followed by several sub-trees as distribution segments. In this structure, feeder fibers for different LR-PONs share the same cable ring leading to low infrastructure costs while offering high reliability [4]. Meanwhile, with the emergence of the 5th Generation (5G) mobile services and Internet of Things (IoT) paradigm, data traffic is growing incredibly fast. Under these circumstances, Time and Wavelength Division Multiplexing (TWDM) technology has been selected by the Full Service Access Network (FSAN) group as a primary solution for Next Generation Passive Optical Network 2 (NG-PON2) [6], providing high capacity and supporting multi-service environments [7]. In TWDM-PON, multiple 10G-TDM PONs are stacked via multiple wavelength channels (e.g. 4 or 8) leading to a bandwidth increase [6]. Such TWDM technology can be easily employed in ring-and-spur LR-PON (which is referred to as LR-TWDM-PON), achieving both high capacity and large coverage.
Due to the high capacity and large coverage of LR-TWDM-PON, it has become of key importance to guarantee its reliability. A failure on the physical layer can cause low power reception and hence introduce service interruption, which may affect the operators’ revenue. In particular, business users and mobile backhaul/fronthaul systems are quite “quality of service” oriented. These “important customers” for the network providers are typically willing to pay more to avoid long service interruptions. In this regard, an effective and fast system for fault detection and localization is required to minimize the interruption time and to improve the system’s reliability. This is especially the case for the distribution fiber segment, where protection is not yet provided in ring-and-spur LR-PONs.
An effective solution for monitoring the feeder ring of the “ring-and-spur” LR-PONs was recently proposed in [8,9]. The proposed technique is based on the use of a “dark fiber” and consists of the use of two-wavelength Bi-Directional Transmission Reflection Analysis (2λ-BD-TRA) as the localization technique to detect, identify and localize any major faults (e.g., fiber cuttings, macro bending and connector mismatching) that impact all the fibers within the same cable. The newly developed 2λ-BD-TRA approach, only requiring the measurement of the power levels of transmitted and backscattered monitoring signals, outperforms many other monitoring methods (e.g., Optical Time Domain Reflectometry OTDR) with respect to measurement time, system complexity and dynamic range. However, it should be noted that more than 80% of PON failures occur in the distribution segment close to the user end [10]. Some research, such as [11], have presented some solutions for feeder section monitoring in ring-and-spur LR-PONs without proposing distribution section coverage. In [12], optical reflectors were proposed to monitor the whole network. Nevertheless, by using optical reflectors the fault localization functionality is not possible as the monitoring module is only able to detect reflected signals from them. To address these problems, in [13] a novel fault supervision system was suggested to carry out the fault localization functionality, covering both feeder and distribution sections in LR-PON. Nevertheless, only a basic principle was reported in [13]. The proposed technique was unable to cover all fault types which may occur in LR-PON, making it impractical for real deployment.
This paper further extends the previous work presented in [13] by introducing a complete mathematic model of the operation principle for the proposed full monitoring scheme and provides a comprehensive performance evaluation to cover various possible optical events, including macro bending, connector mismatching and fiber breaks, that may interrupt service. Furthermore, three critical issues that may affect the performance of the proposed full monitoring scheme are also investigated in this paper. These are (1) how the return loss of optical events impact the localization accuracy; (2) whether data transparent wavelengths can be used for monitoring, in order to minimize the impact of monitoring signals on data channels; and (3) whether an additional powermeter can be eliminated or a low-cost one can be used, at the user end, to improve the cost efficiency of the overall system design. For the second issue, fault localization accuracy will be investigated using data transparent wavelengths, which gives valuable insights into the possibility of using transmission transparent wavelengths for localizing various optical events. Regarding the last issue, different options of powermeters were tested which revealed that it is possible to remove the additional powermeter at the user end while keeping the monitoring accuracy at an acceptable level. This implies that the proposed full monitoring solution can be implemented with low cost.
2. Operating principle
TRA is a simple and efficient monitoring technique, which was originally proposed for sensing applications [14] using a single-mode fiber in a point-to-point configuration. Briefly, TRA is based on the unique relationship between the powers transmitted (PT) and backscattered (PB) by the fiber for a given loss location [14]. When a Continuous Wave (CW) light with a launched power P0 is injected into one end of the optical fiber, a transmitted power PT can be measured at the other end. At the same time, a reflected/backscattered power PB, which is mainly due to Rayleigh backscattering distributed all along the fiber, can be measured at the fiber input. If a single event generating a local loss appears in the fiber, the relationship between PT and PB only depends on the event location. The event can, therefore, be localized by measuring PB and PT. However, this TRA approach is only suitable for non-reflective events (e.g., macro bending). In order to apply the TRA technique to PON monitoring, a novel two-wavelength TRA (2λ-TRA) technique (as shown in Fig. 2) was proposed in [15]. One of the two Super Luminescent Diodes (SLD1 at λ1 and SLD2 at λ2) launches CW light into the test fiber. Then the transmitted power (PT1 at λ1 and PT2 at λ2) and the integrated reflected/Rayleigh backscattered power (PB1 at λ1 and PB2 at λ2) are measured by both powermeters. The localization process of a reflective event is based on the unique relationship between the backscattered (PB1 and PB2) and transmitted (PT1 and PT2) powers for a given event location zp and a given Return Loss (RL). In [15] it was shown that the 2λ-TRA scheme was able to localize a reflective event (e.g., fiber break) with high accuracy (localization error of a few meters). In the proposed LR-PON full monitoring scheme, the nλ-TRA technique, where n is equal to 1 or 2, was applied as the monitoring technique thanks to its good performance in fault localization.
For TWDM-PON, the legacy (i.e., purely power splitter (SPL) based) Optical Distribution Network (ODN) is the primary option. Nevertheless, wavelength filters (e.g. arrayed waveguide grating AWG) may also be deployed in the field [16]. As the TWDM-LR-PON depicted in Fig. 3 shows a 1xN SPL (or a 1xN Arrayed Waveguide Grating (AWG)) and several 1 × M SPLs are used as the first stage and second stage ODNs respectively. Accordingly, this monitoring scheme is designed to be compatible with the legacy ODN as well as with the ODN schemes that include AWG. To carry out a full monitoring functionality, the nλ-TRA unit is implemented at the Remote Node (RN) covering both the Feeder Fiber (FF) ring and the Distribution Fiber (DF) segment by means of a 1 × (N + 1) Optical Switch (OS). After receiving a fault alarm triggered by the network layer, indicating whether the failure has occurred in the feeder section or in a certain drop link, the 1 × (N + 1) OS switches the monitoring signals either to the feeder ring or to the corresponding drop port which contains the faulty branch. Implementing a monitoring system at the RN involves active devices in the field. However, it should be noted that in LR-PONs, the RNs are often kept active due to the installation of amplifiers for reach extension [10], and therefore the power supply for the monitoring system can be easily provided.
Fig. 3 Schematic diagram of the proposed full monitoring scheme (OS: optical switch, ONU: optical network unit, CO: central office, FF: feeder fiber, RN: remote node, DF: distribution fiber, SPL: splitter, SLD: super luminescent diode, AWG: arrayed waveguide grating), PBi is measured at the RN while PTi is measured at the ONU side.
The feeder ring monitoring scheme has already been depicted in [8,9]. In this article the main focus is on the monitoring procedure of the distribution section. As shown in Fig. 3, the 1st stage splitting point is bypassed by the monitoring signals thanks to the OS. This splitting point can use either an SPL or an AWG to be compatible with different types of ODNs for TWDM-PONs. The TRA monitoring signal is launched at the 2nd stage splitting points (i.e., 1xM SPL as shown in Fig. 3) and therefore simultaneously interrogates M distribution fibers after the 1xM SPL. The size of the 1st stage splitting point determines the number of monitoring ports. In case the 1st stage and 2nd stage splitting points are not at the same RN, the proposed monitoring scheme can be implemented at the 1st stage splitting point, still covering the distribution segment. Besides, given that the feeder ring has a built-in protection feature, in case a fiber cut occurs in the feeder segment, the fault monitoring information can still be transmitted to the CO through the protection path.
When the 1 × (N + 1) OS has been positioned with regards to the faulty branch, the TRA based fault localization is then triggered. A CW light emitted by SLD1 (at λ1) is launched into the M distribution fibers through the 1 × M SPL. Let us consider M = 4. The fiber lengths of the four DFs are given by L1, L2, L3 and L4, respectively. The transmitted power PT1j (j = 1 to 4) is measured at the corresponding optical network unit (ONU). In case a fiber break occurs in a dedicated distribution fiber, the solution proposed in [17] can be used to transmit PT information to the TRA unit. The sum of the Rayleigh-backscattered/reflected powers all along the four fibers from one monitoring port is denoted as PB1 (at λ1), which is measured by the powermeter in the TRA unit (see Fig. 3). By switching the source from SLD1 to SLD2, the corresponding transmitted power (PT2j at jth DF) and backscattered power (PB2) can be obtained at wavelength λ2. P0i is the output power of the light source (i = 1@λ1, i = 2@λ2). The Insertion Loss (IL) of an event occurring in a fiber is typically wavelength-sensitive (e.g., macro bending). Therefore, ILi is the insertion loss at wavelength i (λi). However, this is not the case for the Return Loss (RL) according to the previous experimental verifications [15]. It is, therefore, assumed that RL is wavelength-insensitive.
First, let us define PB0i as the reference power backscattered/reflected (measured when there is no fault) by the four DFs. Considering parameters like directivity of the circulator, ILi and RL of all the devices in the field, PB0i can be expressed as:
where DIR [dB] is the directivity of the circulator (e.g., power fraction directly transmitted from port 1 to port 3 in Fig. 3) and RLOS [dB] is the return loss of the 1 × (N + 1) OS. RLfilter1(2) [dB] refers to the return loss of the WDM filter that is implemented before the 1st stage splitting point (before the dedicated 2nd stage of the splitting points). RLendj [dB] (j = 1, 2, 3, 4) is the return losses from the end of the jth DF. Ti (i = 1@λ1, i = 2@λ2) is the fiber transmission coefficient depending on the fiber attenuation coefficient αi [dB/km] and can be expressed as Ti(x) = exp(-αix). Since αi varies with the wavelength, different wavelengths will lead to distinct transmission coefficients. RAYi is the Rayleigh backscattered power coefficient which is expressed as:in which αsi [dB/km] is the Rayleigh scattering coefficient, proportional to 1/λi4, and Si is the capture coefficient (dimensionless, also wavelength dependent) [18] (i = 1@λ1, i = 2@λ2). Iik is (i = 1@λ1, i = 2@λ2, and k = 1, 2…5) the coefficient corresponding to the insertion losses of the devices at different wavelengths. They can be expressed as: where ILcir_i1 (i = 1@λ1, i = 2@λ2) is the insertion loss of the circulator from port 1 to port 2 and ILcir_i2 is the insertion loss of the circulator from port 2 to port 3. ILOSi (i = 1@λ1, i = 2@λ2) is the insertion loss of the 1 × N optical switch, ILfilter1(2)i refers to the insertion loss of the WDM filter that is implemented before the 1st stage splitting point (before the dedicated 2nd stage of splitting points). ILSPLi is the insertion loss of the 2nd stage splitter. The latter four insertion losses are divided by 5 instead of 10 due to the double passage of the light.PT0ij (i = 1@λ1, i = 2@λ2) is defined as the reference transmitted power measured at the end of the jth DF (with length Li). In the example shown in Fig. 3, an event with a RL [dB] and an ILi [dB] occurs in the first DF at a distance of zp far from the corresponding second splitting point. As it is rare to have more than one faulty DF for a monitoring port [19], a single faulty DF case was considered. Given that each time the 1 × (N + 1) OS switches the monitoring signal to one of the monitoring ports, the proposed scheme is capable of supporting the detection of multiple event (N + 1 simultaneous faults, one per monitoring port) in the overall system.
In the presence of an event, PBi and PTi1 are the measured backscattered/ reflected and transmitter powers when the event occurs. Since PTij = PT0ij·10^ (-ILi/10), ILi can be calculated by measuring PT0i1 and PTi1. PBi can be expressed as:
Ri (i = 1@λ1, i = 2@λ2,) is defined as the ratio of backscattered powers with and without events. As described in [14], Ri can be used to calculate the fault location zp and is expressed as:αi, αsi, L1 to L4 and other parameters (e.g., directivity of the circulator, RLend1-4, ILs and RLs of the circulator, optical switch, filters and splitter) are known parameters. ILi, PBi0 and PBi, can be obtained through measurements. The problem of localizing the event (i.e., determining zp) can finally be addressed by solving two equations (Eq. (9) for i = 1@λ1 and i = 2@λ2) with two unknown variables (zp and RL).Let us note that the events occurring in a fiber can be classified into non-reflective and reflective events. The first type includes events such as macro bending, while fiber breaks and connector mismatching are counted in the second type. In order to achieve high localization accuracy, the number of monitoring wavelengths must be adapted to the event type, namely 1λ-TRA for non-reflective events and 2λ-TRA for reflective events [20]. A detailed fault identification procedure for a nλ-TRA (n equals to 1 or 2) based monitoring scheme is described in [20]. In short, for a fiber break with an infinite IL, both PT1 and PT2 are close to zero (in the linear scale). Then, for all the other events, by comparing ΔPT1 (PT01-PT1@λ1) with ΔPT2 (PT02-PT2@λ2, λ1>λ2), it can be distinguished whether the event is a non-reflective event (e.g., macro bending) or a reflective event with a finite IL (e.g., connector mismatching, fiber break with finite IL). Note that 1λ-TRA for non-reflective events monitoring simplifies Eq. (9) as a unique equation (i equals to either 1 or 2) with only one unknown variable (i.e. zp) since RL is known to be infinite.
3. Experimental validations
Experiments were performed to validate the capability of the proposed system. The experimental setup is depicted in Fig. 3, where a 1 × 4 splitter is implemented (i.e., M = 4). The lengths of the four fibers connected to the investigated splitter were 6.60km, 3.60km, 5.23km and 2.22km respectively (measured by an OTDR). The SLDs used in the experiments operated at 1564.6nm (SLD1) and 1307.5nm (SLD2) with a line-width of 57.9nm and 80.4 nm, respectively. The input power (P0) was 12.79mW for SLD1 and 10.73mW for SLD2. The broad linewidth of the SLDs in the experiments (ΔνP is around 4.8 × 106 MHz @1550nm) was much wider than the Brillouin gain spectrum (ΔνB is around 100MHz @1550nm [21]). Thus, the Brillouin gain of the SLDs was reduced by a factor of 1 + ΔνP/ΔνB (which is around 4.8 × 104 here) [21] compared with the monochromatic case. The Brillouin backscattering effect could therefore be neglected.
For the purpose of comparison, the event localization was also measured by a commercial OTDR presenting a peak power of around 10mW and a spatial resolution of 10m (i.e., a pulse duration of 100ns was selected in order to get a sufficient dynamic range to overcome the splitting loss). In the experiments, different types of events were tested, including fiber breaks (with an infinite IL), connector mismatching and macro bendings.
3.1 Fiber break with infinite IL
2λ-TRA is applied in the case of an infinite-IL fiber break, since it is a reflective event. By measuring PBi0, PBi, PT0i1, PTi1 and implementing them into Eq. (9) under two different wavelengths, the fault location zp can be calculated. In the experiments, an-infinite IL fiber break was introduced into the first drop fiber at three different locations. The experimental results are presented in Table 1. Good agreements between the OTDR measurements and the proposed 2λ-TRA scheme were achieved (a detailed localization accuracy discussion will be given in Chapter 4).
Table 1. Comparison between the measured zp by OTDR and the 2λ-TRA solution.
3.2 Reflective events with finite IL
In this experiment, a mismatched FC/PC connector was applied to simulate a finite-IL reflective event, which was introduced into the third drop fiber at two different locations. A misconnected FC/PC connector was characterized with a low-RL (e.g. 15dB), which could introduce a large localization error as will be specified later in section 4.1. In this experiment, the measurement of a misconnected FC/PC connector was therefore applied as a “worst-case” to verify the localization capability of finite-IL reflective events. 2λ-TRA measurement was applied here as the monitoring technique. The corresponding experimental results are depicted in Table 2, according to which, the localization difference (δzp) of the mismatched FC/PC connectors was quite large, especially the 100m difference when zp = 0km. As mentioned above, it was mainly due to the small RL of a mismatched FC/PC connector. Detailed studies will be given in section 4.1.
Table 2. Experimental results for connector mismatching.
3.3 Fiber macro bending
For non-reflective events (like macro bending), 1λ-TRA is applied. The SLD1 (@λ1 = 1550nm) is used as the monitoring light source. Here, a fiber bending was introduced in the second drop fiber at three different locations. As will be specified later in section 4.2, the localization accuracy almost kept constant as a function of the macro bending insertion loss. In the experiment, a fiber bending with an 11.2dB-IL was used as a general case for experimental verification. The experimental results for the TRA based solution are presented in Table 3 and are compared with the OTDR measurements. A maximum localization difference (δzp) of 20m was observed at the location of 3.6km, which demonstrates the good accuracy of localizing the bends in a distributed fiber.
Table 3. Comparison between the measured zp by OTDR and the 1λ-TRA solution.
4. Localization accuracy analysis
In this section, the localization accuracy of the proposed technique will be investigated. Different optical events (fiber breaks, macro bendings and other reflective events with finite IL) will be discussed. As presented in [15,22], the inaccuracy of the two powermeters is the primary factor that limits the localization accuracy for the TRA based monitoring solution. In order to take the inaccuracy of the powermeters into consideration, two measurement relative uncertainty coefficients ζ1 and ζ2, were introduced, which correspond to the maximum measurement accuracy of the two powermeters, respectively. The measured powers can therefore be considered to lie inside the ranges PBi ± PBi·ζ1 and PTi ± PTi·ζ2 (here PT and PB are the real values). The performed simulations take into account a uniform distribution of the measured powers within the aforementioned ranges, where 10,000 samples are considered. Based on this, the expected localization error (i.e., the difference between the mean value of the simulated 10,000 localization results and the actual location) and the corresponding standard deviations (STD) can be calculated as a function of RL and IL. Furthermore, three main issues, and their possible impacts on localization accuracy, will be discussed: (1) the effect of return loss, (2) using data transparent wavelengths for monitoring and (3) the choice of the powermeter used at ONUs. In the following studies, the total length of the fiber link and event location zp were set to 5km and 2km, respectively. For a relatively short fiber (<10km), localization errors almost keep constant when zp changes. Therefore setting zp = 2km can represent a general case.
4.1 Effect of the event return loss
In previous studies [15] it has been found that RL may greatly influence the localization accuracy in the TRA based technique for reflective events. Accordingly, this issue will be taken into account in this section. The two wavelengths used in the simulation study were λ1: 1550nm and λ2: 1310nm which are in line with the experiment. The related parameters of the two wavelengths are listed in Table 4, in which, α was obtained by the OTDR measurements and S· αs was calculated from the obtained TRA measurements [15]. The powermeters used in the lab have the maximum measurement errors of 0.1%, and therefore, ζ1 and ζ2 were set at 0.001.
Table 4. Parameters of different wavelengths.
First, let us consider the case of an infinite-IL fiber break. Using Eq. (9), the expected localization errors and STDs, as a function of the event RL, can be calculated (shown in Fig. 4, pink curves). In Fig. 4, it clearly shows that in most cases the proposed monitoring solution provided a high accuracy (less than 5m expected localization error and STD when RL>35dB). However, it can be seen that when the RL of an event approached 30dB, the expected localization error showed a peak, implying very low accuracy.
Fig. 4 Expected localization errors and STDs as a function of RL when a fiber break occurs in a distribution fiber.
As depicted in Eq. (9), for a fixed zp and a given monitoring wavelength combination, the relationship between R1 and R2 (where Ri is the ratio of backscattered powers with and without events) is dependent on the event return loss (RL). As a result, by analyzing the relationship between R1 and R2 under different values of RLs (see Fig. 5), it can be explained why the localization error is significant when the RL is near 30dB with 1550nm and 1310nm used as the monitoring wavelengths. In Fig. 5, each curve represents the dependency between R1 and R2 when an event is introduced at various locations along a 5 km-long fiber for a given RL. For example, the blue curve located in the upper right corner depicts the relationship between R1 and R2 when the RL is equal to 20 dB. The introduced points on the curves correspond to different zp (the distance between two adjacent points along one curve is 1km). Note that due to the uncertainty of the powermeters, a rectangle can be used to represent the measurement uncertainty of R1 and R2. In other words, the measurement of R1 and R2 are distributed inside the ranges of the rectangle (the rectangles in Fig. 5 are symbolic). In this regard, the interaction length of each curve with the measurement variation space leads to a variation range for zp. As depicted in Fig. 5, when the RL is getting close to 30 dB, the corresponding curve for different event locations is squeezed and has a large overlap with the measurement variation space. Therefore, a large localization error can be expected.
Fig. 5 R1 and R2 for different RL values when an event is introduced at various locations along a 5km-long drop fiber.
Note that localization error is also wavelength dependent and its peak can be varied by using different wavelength combinations for the 2λ-TRA technique, as shown in Fig. 4. The location of peak values for the tested monitoring wavelength combinations are close to 30dB, but are not exactly the same under any two wavelength combinations.
Nevertheless, according to [15], the measured RL of the fiber breaks are always higher than 33 dB and the typical value is around 46 dB. Thanks to the high RL values of the fiber breaks, the localization inaccuracy peak value at 30dB RL will not appear in reality, and therefore, does not influence the monitoring of fiber breaks.
Considering other reflective events with finite IL (IL = 10dB is considered here), the expected localization error and the corresponding STDs, calculated as a function of RL, have almost the same behavior as presented in Fig. 4. Regarding FC/PC connectors, an RL of around 40dB corresponds to a good connection. When it is misconnected, a strong Fresnel reflection occurs between the two flat surfaces, resulting in a much smaller RL, e.g. 15dB. As shown in Fig. 4, when RL is less than 27dB, the expected localization error is around 80 meters, which is in line with the experimental results of a mismatched FC/PC connector (as depicted in Table 2). However, it should be noted that, in practice, the position of a connector is typically known in advance. Therefore, a relatively low localization accuracy is sufficient to identify the misconnected FC/PC connector. For other reflective events (e.g. FC/APC connector, a fiber break with finite IL) whose RLs are normally higher than that of an infinite-IL fiber break [23], the proposed monitoring solution can provide a high localization accuracy (the localization errors are less than 5m).
4.2 Data transparency issue
It should be noted that since the distribution section monitoring is carried out in the same fiber in which the data are transmitted, it is necessary to avoid using the wavelengths dedicated to the data signals for in-service fault supervision, in order to reduce the impact on data transmission. In this regard, it is suggested to use the wavelengths 1260/1650nm instead of 1310/1550nm. Considering that these wavelengths are over 50 nm away from the data channels, the impact of cross-talk on the transmission performance of the data channels is negligible. The related parameters are listed in Table 5 [18]. Similar to the previous study, the expected localization errors and STDs were calculated as a function of the event RL under 1260nm/1650nm. For the purpose of comparison, the results of 1260nm/1650nm are also depicted in Fig. 4 (blue curves).
Table 5. Parameters of different wavelengths.
As shown in Fig. 4, the 1260nm/1650nm combination presents a larger expected localization error than 1310nm/1550nm when the RL approaches to 30dB. In the other cases, both of them provide almost the same localization error. For RL>35dB and RL<27dB, the expected localization errors are 5m and 80m respectively, while the STDs for both cases are less than 5m. As mentioned above, the peak value of a localization error at 30dB-RL does not influence the monitoring of fiber breaks and connector mismatching. In this regard, data transparent wavelengths (e.g. 1260nm and 1650nm) can be used as the monitoring wavelengths for the proposed monitoring solution with good localization accuracy.
Regarding non-reflective events that may occur in a fiber, such as macro bending, 1λ-TRA was used in the proposed monitoring technique due to the infinite RL value. Through simulations, the expected localization errors and STDs could be calculated as a function of the event IL under two different wavelengths (i.e. 1550nm and 1650nm). Related results are depicted in Fig. 6. It can be observed that the expected localization error of 1650nm is only 3 meters larger than that of 1550nm, while the STD values at the two wavelengths are quite similar. Therefore, “data-transparent” wavelengths (e.g. 1650nm) can be used for monitoring non-reflective events with good localization accuracy.
Fig. 6 Expected localization errors and STDs as a function of IL when a fiber bending occurs at the distributed fiber.
4.3 Choice of the powermeters used at the ONUs
As mentioned in Section 2, PT must be measured at the ONU side (i.e., user end), which means that a powermeter needs to be implemented at the ONU. ONU is a user equipment and hence quite cost-sensitive. As the powermeter cost depends greatly on the level of measurement accuracy (related to the aforementioned parameter ζ2), in this section the impacts of ζ2 on the localization accuracy will be investigated. Powermeters with three different measurement accuracy levels: 0.1%, 5% and 50% will be considered. These inaccuracy levels represent a powermeter with high precision, a low-cost commercial powermeter and an embedded powermeter at the ONU defined in PON standard (e.g. G984.2), respectively. In GPON standard G984.2, the ONU should be capable of measuring received powers with ± 3 dB accuracy, corresponding to accuracy level of 50%. Considering that this feature is also included in the NG-PON standard, the ONU itself can provide the PT values with a 50% accuracy level without an extra powermeter dedicated to monitoring. Given that the cost of PB measurement powermeter can be shared by the whole central office, it is possible to afford a high precision powermeter. Therefore, in the following study, the inaccuracy of PB measuring powermeter is still set at 0.1%. In this study, two data-transparent wavelengths (1650nm and 1260nm as discussed in the previous section) were applied and both reflective and non-reflective events were considered.
1. Reflective events (2λ-TRA)
As the IL directly affects the PT measurement, reflective events with various values of IL (from 3dB to 30dB) were considered in this study, and the RL was set at 45dB. Expected localization errors and the corresponding STDs versus the event IL are depicted in Fig. 7, where the blue, pink and green curves represent the results employing powermeters with measurement accuracies of 0.1%, 5% and 50%, respectively. From Fig. 7, it can be seen as expected that the localization error increases with the measurement inaccuracy, especially when the insertion loss of events is lower than 5dB. For instance, the expected localization errors are −350m, −3m and −2m for a measurement accuracy of 50%, 5% and 0.1% at 3dB IL respectively, while the STD values are 1.2 km, 51m and 1m respectively. When the event IL is higher than 5 dB, both 0.1% and 5% measurement accuracies provide a low localization error (approximately 10m expected localization error and 5m STD). If the event IL is higher than 10 dB (this includes fiber breaks), all the three cases can present low localization errors (the expected localization error and STD are less than 10m and 5m respectively).
Fig. 7 Expected localization errors and STDs under three different accuracy levels (0.1%, 5% and 50%) of powermeters as a function of IL when a reflective event occurs in a distribution fiber.
According to these results, the proposed monitoring scheme is able to accurately localize an event that has an IL of higher than 5 dB (10dB) with a 5% (50%) PT measurement inaccuracy. In order to verify the feasibility under various RL values (only 45dB has been considered so far), localization errors were further calculated as a function of event RLs with different PT measurement powermeter precision levels (0.1%, 5% and 50%) and different event ILs (5dB, 10dB and 15dB). In Fig. 8(a), the IL is set to 5dB and the measurement inaccuracies of 0.1% and 5% are taken into account. In Figs. 8(b) and 8(c), events with a 10dB and 15dB ILs are considered with powermeters characterized by the three measurement accuracy levels (0.1%, 5%, 50%). From Figs. 8(a)-8(c) it can be observed that in each IL case, all the relative measurement accuracy cases present similar behavior. In particular, when the RL is less than 27dB, all the cases show no significant localization error differences. It is also worth noting that with a higher IL, the detrimental effect of the measurement accuracy becomes less significant. For example, a 5% inaccuracy introduces certain localization errors (e.g. higher than 10m STD with an RL of more than 30dB) when the IL is 5 dB, while in the case of 15 dB IL, a 50% measurement accuracy gives almost the same localization performance as a high precision powermeter (e.g. STD becomes less than 5m with an RL of more than 30dB). The three IL levels shown in Figs. 8(a)-8(c) do not necessarily refer to practical values, nevertheless, the operability ranges of the three considered types of powermeters can be defined accordingly.
Fig. 8 Expected localization errors and STDs under different accuracy levels of powermeters as a function of RL when a reflective event occurs in a distribution fiber. (a) IL = 5dB, (b) IL = 10dB, (c) IL = 15dB
2. Non-reflective events (1λ-TRA)
For non-reflective events (e.g. macro bending), expected localization errors and STDs versus the event IL are depicted in Fig. 9, which presents a very similar trend to the one observed for reflective events. When the bending IL is higher than 5 dB, 0.1% and 5% measurement inaccuracies provide a low localization error (approximately 25m expected localization accuracy and less than 10m STD). If the bending IL is higher than 10dB, low localization errors can be obtained in all three cases.
Fig. 9 Expected localization errors and STDs under three different powermeter accuracy levels (0.1%, 5% and 50%) as a function of IL when a non-reflective event occurs in a distribution fiber.
3. Discussion
Based on the above results, we can see that the proposed monitoring scheme is capable of accurately localizing an event with a high IL (>10dB) without installing any extra powermeter at the ONU side. Regarding an event with an IL of higher than 5 dB, a relatively low-cost powermeter with 5% accuracy level can accurately localize it. Let us note that thanks to the high link optical power budget of NG-PONs (29-35 dB [24]), a 5 dB-IL event may be tolerated, which does not require any action. Given this fact, a cost-effective powermeter is sufficient for accurate monitoring. Network operators can therefore easily apply a “pay-as-you-need” scheme. For example, in most cases (e.g. Fiber To The Home), an extra PT measurement powermeter is not necessary to be implemented at the ONU. In that case, only the events with a high IL (>10dB), which have a high impact on the connection, need to be accurately localized. Regarding business users and mobile back/front-haul systems, who are willing to pay more for higher service quality, network operators can install an extra powermeter at the ONU side to accurately detect and localize a small-IL event.
5. Further discussion and conclusion
In this paper, a novel TRA-based monitoring solution has been proposed, which, for the first time, carries out a full fault localization functionality in LR-TWDM-PONs, covering the monitoring in both the feeder and the distribution sections. Its capability of localizing events in distribution fibers was both theoretically analyzed and experimentally verified. The comparison between OTDR and the proposed TRA based technique is summarized in Table 6. Although the OTDR technique has the technical potential to carry out full monitoring for LR-TWDM-PON, the TRA based solution outperforms OTDR with respect to measurement time, light source complexity and dynamic range. The proposed TRA based scheme has a better or similar spatial resolution compared to OTDR (with a 100ns pulsewidth) in most cases, except in the presence of a small-RL event (e.g. mismatched FC/PC connector) where a relatively low resolution (80m) was obtained. In the proposed TRA based monitoring scheme, increasing the input power improves the dynamic range without affecting the spatial resolution (the Brillouin effect can be neglected by using a broadband light source), meaning that a large dynamic range (>60dB) and a high spatial resolution can be achieved simultaneously. Besides, the proposed scheme supports a “pay-as-you-need” approach, offering network operators high flexibility when selecting proper monitoring configurations. Let us note that the powermeter for PT measurement can also be implemented at the demarcation point which is the border of the operator’s and the customer’s network segments [17]. Whether the event occurs at ODN can be identified by comparing the measured PT in the presence of an event with the reference value PT0 (with no event). If PT = PT0, the event occurs at the customer side. If PT<PT0, the event occurs at the ODN.
Table 6. Comparison between OTDR and the proposed TRA based monitoring scheme for LR-PON.
For the case that more than one fault occur in the same monitoring port of the distribution section, the proposed TRA-based monitoring scheme can be extended to be capable of detecting two simultaneous fiber cuts, which are the most critical events to be monitored. There are two cases: (1) the two faults occur in the same monitoring port but in two different distribution fibers and (2) the two events occur in the same drop fiber. For case (1), a solution consists in using two additional monitoring wavelengths. Their implementation will provide two extra equations required to determine the two extra unknowns introduced by the second event. For case (2), the proposed monitoring scheme will be able to detect the event closer to the RN. The second cut can be localized by applying the same procedure after the repair of the first cut.
Acknowledgments
The authors would like to thank the F.R.S.-FNRS (F.R.I.A), the European Community's Seventh Framework Programme ICT-DISCUS, Göran Gustafssons Foundation, the National Natural Science Foundation of China (Grant No. 61550110240), and the Interuniversity Attraction Poles program of the Belgian Science Policy Office (grant IAP P7-35 photonics@be) for their financial support.
References
1. H. Song, B. W. Kim, and B. Mukherjee, “Long-reach optical access networks: a survey of research challenges, demonstrations, and bandwidth assignment mechanisms,” IEEE Comm. Surv. and Tutor. 12(1), 112–123 (2010). [CrossRef]
2. X. Zhang, Z. Li, J. Li, C. Yu, A. P. Lau, and C. Lu, “Low-cost coherent receiver for long-reach optical access network using single-ended detection,” Opt. Lett. 39(18), 5248–5250 (2014). [CrossRef] [PubMed]
3. C. M. Machuca, J. Chen, and L. Wosinska, “Efficient protection in TDM PONs,” IEEE Commun. Mag. 50(8), 110–117 (2012). [CrossRef]
4. E. Wong, “Next-generation broadband access networks and technologies,” J. Lightwave Technol. 30(4), 597–608 (2012). [CrossRef]
5. M. Ruffini, L. Wosinska, M. Achouche, J. Chen, N. Doran, F. Farjady, J. Montalvo, P. Ossieur, B. O’Sullivan, N. Parsons, T. Pfeiffer, X. Qiu, C. Raack, H. Rohde, M. Schiano, P. Townsend, R. Wessaly, X. Yin, and D. B. Payne, “DISCUS: An end-to-end solution for ubiquitous broadband optical access,” IEEE Commun. Mag. 52(2), S24–S32 (2014). [CrossRef]
6. Y. Luo, X. Zhou, F. Effenberger, X. Yan, G. Peng, Y. Qian, and Y. Ma, “Time- and Wavelength-Division Multiplexed Passive Optical Network (TWDM-PON) for next-generation PON stage 2 (NG-PON2),” J. Lightwave Technol. 31(4), 587–593 (2013). [CrossRef]
7. G. Talli and P. D. Townsend, “Hybrid DWDM-TDM long-reach PON for next-generation optical access,” J. Lightwave Technol. 24(7), 2827–2834 (2006). [CrossRef]
8. M. Cen, J. Chen, P. Mégret, V. Moeyaert, and M. Wuilpart, “Fast and simple fault monitoring for long-reach passive optical networks,” in Proceedings of the European Conference on Optical Communications (IEEE, 2014), paper P.7.15. [CrossRef]
9. M. Cen, J. Chen, V. Moeyaert, P. Mégret, and M. Wuilpart, “Efficient monitoring for ring-based long-reach passive optical networks,” in Optical Fiber Communication Conference, (Optical Society of America, 2015), paper Th4H.4. [CrossRef]
10. M. A. Esmail and H. Fathallah, “Physical layer monitoring techniques for TDM-passive optical networks: a survey,” IEEE Comm. Surv. and Tutor. 15(2), 943–958 (2013). [CrossRef]
11. H. Fathallah and M. Esmail, “Performance evaluation of special optical coding techniques appropriate for physical layer monitoring of access and metro optical networks,” Photonic Netw. Commun. 30(2), 223–233 (2015). [CrossRef]
12. K. Fouli, L. Chen, and M. Maier, “Time-, wavelength-, and code-domain optical reflection monitoring for next-generation access-metro networks,” Comput. Commun. 34(8), 1011–1021 (2011). [CrossRef]
13. M. Cen, J. Chen, V. Moeyaert, P. Mégret, and M. Wuilpart, “A full monitoring scheme for long-reach TWDM PONs,” in Asia Communications and Photonics Conference (Optical Society of America, 2015), paper AM2E.1. [CrossRef]
14. V. V. Spirin, F. J. Mendieta, S. V. Miridonov, M. G. Shlyagin, A. A. Chtcherbakov, and P. L. Swart, “Localization of a loss-inducing perturbation with variable accuracy along a test fiber using transmission-reflection analysis,” IEEE Photonics Technol. Lett. 16(2), 569–571 (2004). [CrossRef]
15. M. Cen, J. Chen, V. Moeyaert, P. Mégret, and M. Wuilpart, “Multi-wavelength Transmission-Reflection Analysis for fiber monitoring,” Opt. Express 22(25), 31248–31262 (2014). [CrossRef] [PubMed]
16. J. S. Wey, D. Nesset, M. Valvo, K. Grobe, H. Roberts, Y. Luo, and J. Smith, “Physical layer aspects of NG-PON2 standards—part 1: optical link design [Invited],” J. Opt. Commun. Netw. 8(1), 33–42 (2016). [CrossRef]
17. M. Roeger, B. Hiba, J. Hehmann, M. Straub, H. Schmuck, M. Hedrich, T. Pfeiffer, C. Koos, J. Leuthold, and W. Freude, “In-service monitoring of PON access networks with powerline independent devices,” J. Opt. Commun. Netw. 6(11), 1018–1027 (2014). [CrossRef]
18. Luc Thévenaz and M. Wuilpart, Advanced Fiber Optics: Concepts and Technology, 1st ed. (EPFL Press, 2011), Chap. 8.
19. P. Urban, G. V. Llosera, E. Medeiros, and S. Dahlfort, “Fiber plant manager: An OTDR- and OTM-Based PON monitoring system,” IEEE Commun. Mag. 51(2), S9–S15 (2013).
20. M. Cen, J. Chen, P. Mégret, V. Moeyaert, and M. Wuilpart, “Advanced fault monitoring scheme for ring based long-reach optical access networks,” J. Lightwave Technol. (submitted).
21. G. P. Agrawal, Nonlinear Fiber Optics, 5th ed. (Academic Press, 2012), Chap. 9.
22. M. Cen, V. Moeyaert, P. Mégret, and M. Wuilpart, “Localization and quantification of reflective events along an optical fiber using a bi-directional TRA technique,” Opt. Express 22(8), 9839–9853 (2014). [CrossRef] [PubMed]
23. W. Lee, S. I. Myong, J. C. Lee, and S. Lee, “Identification method of non-reflective faults based on index distribution of optical fibers,” Opt. Express 22(1), 325–337 (2014). [CrossRef] [PubMed]
24. F. Effenberg, “NG-PON2-A system vendor’s perspective,” inProceedings of the European Conference on Optical Communications (IEEE, 2014), paper WS5.
