OLT-OTDR Dataset

This dataset collection includes OLT-OTDR measurements designed specifically for fault classification and localization in passive optical networks (PONs). All datasets share the same data structure and were collected from three experimental testbeds located at Fraunhofer HHI (Berlin), ADTRAN Networks SE (Munich), and ADTRAN Networks SE (Meiningen). Each testbed was built to support a variety of faults, covering both feeder fibers and the optical distribution network (ODN). The dataset includes six fault types: feeder fiber tap, feeder fiber cut, distribution fiber tap, distribution fiber cut, unplugged optical network unit (ONU), and powered-off ONU. Each sample captures measurements from both the OLT and OTDR, providing complementary information for accurate fault classification and localization. From the OLT we collect the received optical power values for each ONU, while from the OTDR we collect deep PON analysis traces (DPA). DPA traces are obtained by processing OTDR traces using a proprietary algorithm from ADTRAN, as described in [1]. In total, the dataset comprises more than 2000 samples, with class distribution summarized in Table 1. For the Berlin testbed, augmented data is also provided to facilitate training of AI models. All the data samples have been collected experimentally by repeatedly introducing each of the described faults into the network. For emulating fiber tapping we used a device known as fiber tester in the Meiningen testbed. A fiber tester bends the fiber and measures the intensity of the escaping light. In our scenario a ~3dB attenuation was recorded in Meiningen. As a fiber tester was not available in Berlin and Munich, we just emulated the tapping using a 3dB attenuation as suggested by the experiments in Meiningen.