Binary time domain reflectometry: a simpler and more efficient way of diagnosing defects in wired networks

摘要

Today, the most efficient methods for wired networks diagnosis are based on reflectometry: they can detect hard (i.e. open/short circuits) and soft defects (i.e. insulation damage, hot spots or rust), either permanent or intermittent, and provide their location. This is an important information for maintenance operators who can focus repair in a shorter time. The possibility of monitoring the wires health using specifically designed reflectometry methods offers interesting opportunities for preventive maintenance. Electronic implementations of such methods are mostly based on the use of programmable logic digital systems, such as FPGA or microcontrollers, both for probe signals generation and measured signals analysis. But the cables have a pure analog behavior, therefore these systems are connected to additional costly devices such as digital to analog and analog to digital converters. These components have a direct impact on the systems performances: a higher sampling rate means a better defects location accuracy, but implies a higher cost. A higher resolution enables to detect lower signature defects but requires more memory and computing power for data processing. This paper presents a new purely binary reflectometry method using a simpler electronic architecture and quicker signal analysis, and showing better performances than standard methods (in terms of defect location accuracy, processing speed and memory requirement) at a lower cost. Equivalent detection performances are shown, and the new simplified electronic architecture enables to take advantage of all the digital resources, such as a higher clock frequency than that of most available converters, thus naturally improving the location accuracy. As the conversion components are not required anymore, the consumption and cost of the system is drastically reduced. Performance comparison is presented and an innovative process for the reflectogram computation is introduced, which enables a drastic reduction of the computing power needs and a quicker reflectogram update.