Improvement of fault identification performance using neural networks in passive double star optical networks

Summary form only given. Passive double star (PDS) optical networks are expected to be used to construct low cost access networks for broadband services. We have already proposed a testing method with a dichroic reflective optical (DRO) filter for PDS networks, which identifies faults between an optical line and transmission equipment on the subscriber side. When the reflections are completely separated, we can identify faults with the conventional method described above. However, when the reflections from the filters are superimposed, this becomes difficult and the identification resolution is greatly degraded. This paper proposes a novel software method using neural networks (NN) to overcome this problem.