A robust fault detection algorithm for the improvement of OTDR sensitivity

In the paper, an algorithm is presented for the detection of jump discontinuities in exponential signals buried in high level additive noise. It is employed as a post-processor in an optical time domain reflectometer (OTDR) to locate the position of discontinuities along the fiber. The proposed technique is based on a generalized likelihood ratio test applied to the innovation sequence resulting from Kalman filtering. A rigorous and thorough statistical analysis of the algorithm is reported, showing the achievable resolution together with the probability of error in the detection of faults.