Design, Statistical Analysis & Detection Probability
Optical Fiber Fault Detection must be Fast, Distributed and Pervasive
To materially improve the reliability of the Access Network, fault detection must be fast, distributed and pervasive. The detection of a fault needs to be fast so that the operation of the network, particularly an Ethernet based one, is minimally impacted. A fault should be detected, located and reported to the Host Switch within a fraction of a second. If the fault is momentary, then the link should instantly recover. If not,then there ought to be the capability to initiate a series of diagnostic and/or repair options.
Such responsiveness to a service disruption may translate directly into improved Service Level Agreements (SLAs) and the competitive advantages they may offer the Communications Service Provider (CSP).
Having a sophisticated piece of equipment monitoring a small number of optical fibers (thereby “consuming” these fibers, making them unavailable for revenue generation), is not very efficient in the extremely high link density environment found at the “edge” of the Access Network. We know that the link distances are relatively short (under 40 Km) and that the performance requirements of an OTDR to operate in this segment are modest.
A large number of moderate performance monitors, and their associated OTDRs, distributed throughout the Access Network, makes more sense than fewer, higher performance (and much more costly) systems.
In order to have the highest level of effectiveness, these distributed moderate performance monitors and OTDRs need to be virtually everywhere. Near 100% fault detection may be achieved when large numbers of (or preferably all) optical links incorporate this performance monitoring.
Optical fiber fault detection is based upon the reflected signal from the point of fault. Fiber faults and intermittent connections present optical reflections of varying intensities. The reflection intensity of a fiber break has a known statistical distribution. The distribution of the return signal from a fiber break/cut has been empirically determined.
To achieve a detection probability of greater than 95%, the sensitiviy of the detection circuit must be better than 51 dB.
Since the Fiber Fault Histogram indicates that the detection probability of an optical fiber fault follows a Normal Distribution, modeling to predict outcomes is straightforward. A series of curves for the likely optical fiber cable deployments encountered in the Access portion of the Communications Network may be developed.
For any combination of Micro-OTDR™ Dynamic Range (DR) values, Optical Fiber Spans to be monitored, Location of Micro-OTDR™ SFPs and number of Optical Fibers to be monitored, a Chart of Detection Probability for Random Faults (random location and random reflectance or ORL) may be developed. These Charts offer guidelines to the network architect for the deployment of Micro-OTDR™ SFPs into a network.