In-Service Location of Multiple Fiber Faults in WDM/SCMPONs with Low-Frequency Stepwise Sweep and l1 Regularization G. C. Amaral1, J. D. Garcia1, B. Fanzeres1, P. J. Urban2, and J. P. von der Weid1 1
Center for Telecommunications Studies, Pontifical Catholic University of Rio de Janeiro (PUC-Rio), Rua MarquΓͺs de SΓ£o Vicente, 225. 2 Ericsson Research, Ericsson AB, Stockholm, Sweden.
[email protected]
Abstract: We present a monitoring technique that can be directly integrated in the transceiver for WDM/SCM-PON applications. It is based on the detection of the backscattered signal from a baseband tone and interpretation with the LASSO operator for multiple fault detection. OCIS codes: (120.4825) Optical Time Domain Reflectometry, (120.5700) Reflection.
1. Introduction Centralization in Radio Access Network (RAN), i.e. pooling/hoteling Digital Units (DU) and Radio Units (RU) in the Central Office, requires efficient Mobile Fronthaul (MFH), and SCM technology is considered as a potential candidate [1-3]. The emerging topologies of Radio over Fiber (RoF) consider that the optical signal is transmitted towards a suitable resilient optical-electrical (OE) converter from which the analog radio frequency signal is then directed to an antenna [1]. In such systems, the optical distribution link usually would not exceed a 15 kilometers reach [1], which does not demand high dynamic range monitoring techniques such as standard Optical Time Domain Reflectometry (OTDR) or Optical Frequency Domain Reflectometry (OFDR). Therefore, a monitoring technique that is low-cost, in-service, compatible with the WDM/SCM-PON architecture, and embedded in the transmitter unit is highly desirable. In this work, we propose an architecture for in-service, centralized link supervision using a stepwise frequency sweep of a base band tone in an WDM/SCM-PON framework. The method for multiple fault detection is based on the Low-Frequency Stepwise Sweep method of [4] and on the Least Absolute Shrinkage and Selection Operator (LASSO) [5,6]. According to [4], the magnitude and phase profile in the frequency domain resulting from a step frequency OFDR measurement of an optical fiber is associated to the fiberβs characteristics. The developed mathematical model permits one to associate fiber faults to cardinal sinusoidal functions in the frequency domain profile with high accuracy. Nevertheless, multiple faults are associated to a sum of cardinal sinusoids and, without an adequate signal processing method, are hardly identified. Using a high-dimension search algorithm we can accurately identify multiple faults in a Low-Frequency Stepwise Sweep fiber profile overcoming the need of an as-built or previous knowledge of the fiber [4]. Our results show that the technique is capable of multiple fault detection and localization in an AWG-based WDM/SCM-PON link (since power splitters would be prejudicial given the low achieved dynamic range). The impact on SCM data transmission is negligible so in-service monitoring is feasible with a low operating cost [4]. Comparison with the results from a standard OTDR show a discrepancy of less than one-hundred meters in a multiple fault scenario with a 6.2 dB dynamic range. 2. LASSO As shown in [4], it is possible to measure an optical fiber transfer function relative to a frequency tone sweep by measuring the backscattered signal at the respective optical sub-carrier frequency. For a given set I of fault position indexes and a set πΌπ of reflective event indexes (πΌπ β πΌ), the fiber profile in the frequency domain, in a multiple fault context, can be described as the following linear combination of cardinal sinusoidal functions and sinusoidal functions [4]: π(πΎ) = βπβπΌ π΄π π πππ(πΎπ₯π )π ππΎπ₯π + βπβπΌπ π΅π π ππΎπ₯π ; where πΎ is the low-frequency modulating tone complex wave vector and π₯π are the indexed fiber positions. The magnitude of the backscattered signal can, therefore, be expressed as a linear combination of cardinal sinusoidal functions dependent on the modulation frequency with its phase proportional to the fiber distance along with a term that characterizes reflective events. The period and amplitude of these sinc functions are defined by the position of the loss. Therefore, to associate the backscattered signal to eventual fiber faults, one should be able to identify which sinc functions compose the recovered signal; a simple task when the fiber conditions are known a priori and/or when the linear combination has few components. However, when multiple faults are present and no knowledge of the fiber is available, an adequate algorithm to identify fiber faults in this high-dimension framework must be employed. The proposed methodology assumes available a set of traversed lengths as candidates for fiber position faults. Formally, let πΆ be this set of candidates. Then, in the first step of the proposed algorithm, a least-square with L1 norm penalization is performed to identify the candidates
within πΆ that best fit the acquired signal to the previously presented mathematical model. More specifically, we solve the minimization problem: minβπ β βπβπΆ (π΄π (πΎπ₯π )π ππΎπ₯π + π΅π π ππΎπ₯π )β + π βπβπΆ (|π΄π | + |π΅π |), where π is an a priori defined π΄π ,π΅π
penalty parameter. The idea, known in technical literature as Least Absolute Shrinkage and Selection Operator (LASSO) [5], is to perform a standard least-square fit (first term of the optimization problem) but penalize the choices of some βunnecessaryβ coefficients (second term of the optimization problem). Methodologically, LASSO overcomes the highdimensional problem inherent to a high discretization needed to construct a credible set of candidate fault positions by exploiting the problem's sparsity. As a result of solving the presented problem, only a small subset of coefficients within πΆ takes values different from zero, indicating the points in the fiber where the faults are (potentially) located, thus estimating the βtrueβ' set of positions πΌ. Several efficient algorithms have been proposed in technical literature to solve the LASSO, and we chose the Coordinate Descent [6] for our tests. 3. Experimental Results To validate the henceforth-dubbed SincLASSO-based monitoring method, the setup described in Fig. 1 was assembled. Thelow-frequency monitoring signal was generated and processed in the Lock-In Amplifier after detection in a high-gain low-band photodiode. The data channel that simulates the live monitoring is a 64-QAM LTE 40 MHz channel, which is combined to the monitoring signal and sent to the laser through a power combiner.
Fig. 1: Experimental setup for the Low-Frequency Stepwise Sweep monitoring mehtod.
Controlled tests were conducted using four different fiber links, which were also characterized by a standard OTDR device. In these links, several faults β either reflective or non-reflective - were induced at different positions. The results are presented in the form of an event list (Tables 1 to 3) since the SincLASSO-based method does not retrieve an OTDRlike trace. The links were assembled using fibers with different lengths as follows. Link 1: 1.14 + 2.13 + 1.58 + 3.36 kilometers. Link 2: 1.58 + 2.13 + 1.14 kilometers. Link 3: 3.36 + 2.13 + 1.14 kilometers. OTDR [m] 1147 3330 5118 8247
Link 1 SincLASSO [m] 1227 3346 5033 8257
Error [m] 80 16 85 30
OTDR [m] 1587 3756 4904
Link 2 SincLASSO [m] 1551 3736 4900
Table 1: Link 1 Results. Table 2: Link 2 Results.
Error [m] 36 20 4
OTDR [m] 3361 5539 6676
Link 3 SincLASSO [m] 3392 5573 6679
Error [m] 31 34 3
Table 3: Link 3 Results.
A more intuitive image of the results is presented in Fig. 2, with both the selections of the SincLASSO-based method and the standard OTDR trace for links 1 and 3 in the same image. We observe that the fault intensity retrieved by the SincLASSObased method is not yet calibrated to match the actual loss observed in the OTDR measurement. However, the technique successfully identified multiple reflective and non-reflective events with accuracy better than 100 meters. It should be pointed out that the low accuracy achieved by the method can be improved with different signal processing heuristic approaches and given a higher amplitude of the monitoring tone. Nevertheless, the results are promising, especially due to the simplicity and low cost of the monitoring method and the fact that it is readily embeddable in the transmission system.
(a)
(b)
Fig. 2: Comparison of the selections from a standard OTDR and the SincLASSO-based method for links (a) 1 and (b) 3.
The possibility of in-service network monitoring has been assessed by measuring the resulting Error Vector Magnitude (EVMrms) for different data signal amplitude under two conditions: with; and without the monitoring tone. The tone amplitude was kept at one eighth of the laser's full modulation depth in order to simulate an eight-channel SCM network. The result indicates that the monitoring signal deteriorates the quality of the communication, even if by a very small factor. Under these conditions, however, the required EVMrms for LTE communication is respected with a large margin [7]. It can be observed that, above a certain value of the data signal's amplitude, the quality of communication is deteriorated. This behavior is associated to the laser's non-linear response region, achieved when the modulation signal violates the laser's full modulation depth at around 1.5 [Vpp], or 30 [mA].
Fig. 3: Error Vector Magnitude rms of LTE 64-QAM communication when the monitoring tone is turned on and off.
4. Conclusions We proposed and experimentally demonstrated the validity of the SincLASSO-based method for WDM/SCM-PON monitoring which relates the backscattered signal from a stepwise frequency-swept base band tone with the position of faults in a fiber optical link. Using a high-dimensional statistics filter to elect the best candidates, which compose the received signal and to relate the former to fault positions has proven its effectiveness. The laboratory tests were performed using 64-QAM LTE 40 MHz channel and showed negligible impacts over data transmission, thus characterizing the technique as an in-service monitoring solution for short-reach WDM/SCM-PON links. 5. References [1] Optical Access Seamless Evolution, βSurvey of next-generation optical access system concepts,β OASE White Paper. [2] ITU-T Series G: Transmission Systems and Media, Digital Systems and Networks - Radio-over-fibre (RoF) technologies and their applications. [3] L. Giorgi et al, "Subcarrier Multiplexing RF Plans for Analog Radio over Fiber in Heterogeneous Networks", IEEE/OSA Journal of Lightwave Technology, vol. 34, pp. 3859-3866, 2016. [4] G. C. Amaral et al., βA Low-Frequency Tone Sweep Method for in-Service Fault Location in Sub-Carrier Multiplexed Optical Fiber Networks,β arXiv preprint, arXiv:1609.02933,2016 [5] R. Tibshirani, βRegression shrinkage and selection via the lasso,β Journalof the Royal Statistical Society.Series B (Methodological), pp. 67β288,1996. [6] J. P. von der Weid et al., βAdaptive Filter for Automatic Identification of Multiple Faults in a Noisy OTDR Profile,β Journal of Lightwave Technology, vol. 32, pp. 3418-3424, 2016. [7] 3GPP, "3GPP ts 36.521-1 v. 11.2.0, LTE; radio transmission and reception; part 1: Conformance testing tech. spec. group radio access network, rel. 11," 2013.