No abstract

No abstract

No abstract

International audienceReflectometry-based methods are the standard choice for fault detection techniques in wire networks. While effective when dealing with simple networks and relatively hard faults, their results can be of more difficult interpretation if a network presents more than two branches. In this paper we propose the use of an alternative technique based on a coherent multi-port characterization of a network under test. The data thus collected are used to define excitation signals that will be focusing over the position of a fault, following a method already successfully applied in geophysical prospection techniques and non-destructive testing, namely the DORT method, based on the synthesis of time-reversed signals. It is shown that a direct transposition of this technique to wire networks is not possible, due to the guided nature of wave propagation in wire networks, leading to the impossibility of assuming a dominant direction of propagation, as opposed to the case of propagation in open media. A differential version of the DORT method is introduced, enabling an accurate identification of the original position of faults. Numerical and experimental results are presented to demonstrate the feasibility of this approach

Decomposition of the time reversal operator (DORT) was recently applied to the problem of detection and location of soft faults in wire networks and proved effectual when dealing with a single fault, even in the case of complex network configurations. In this paper, the case of location of multiple faults is addressed, first proving that the standard DORT formulation does not allow to take a clear decision about the individual position of each fault. An alternative version of the DORT, based on an updating procedure, is presented and demonstrated to enable accurate and selective location of multiple soft faults. The proposed procedure is also shown to allow estimating the reflection coefficient of each fault, thus giving access to their severity.

It has been proposed [J. Derosny, Ph.D. Thesis, Université Paris VI, 2000] that the performance of time reversal at recreating a coherent pulse in a strongly reverberating medium is directly proportional to the number of resonant modes M actively taking part at the transmission of energy. This idea is here tested against experimental results, showing that as soon as losses are taken into account, the quality of the focused pulse is a sublinear function of M , leading to a saturation phenomenon that was previously unacknowledged. This is here proven to be caused by mutual coupling between lossy resonant modes, thanks to a statistical modal description of the transmission of signals through the medium. Closed-form relationships are proposed for the first two moments of the pulse signal-to-noise ratio, linking them to the occupied bandwidth, the number of active modes and the degree of resonance of the medium. These formulae, supported by experimental and numerical results, prove that the performance of time reversal can be affected by a strong statistical dispersion. The proposed analysis also predicts that time reversal is a self-averaging process when applied to a reverberating medium, thus allowing the use of models developed in an ensemble-average framework.

International audienceAlthough the number of significant modes is intuitive, this concept has never been clearly defined, and this, mainly because of the unbound number of modes involved in modal overlap. In the present paper, we show that, for a perfect stirring process, the effect of modal overlap can be modeled as an equivalent filtering formulation. By introducing the statistical-bandwidth concept we show that the electromagnetic field statistics due to an infinite number of modes can be summarized by a finite number of significant modes. The case of the electric-energy density in an mode-stirred reverberation chamber (MSRC) has been considered and a new expression of its variability has been established. The good agreement found between the new expression and experimental and simulation results support the several concepts introduced in this paper

This paper introduces models of the time-domain echoes generated by faults in transmission lines excited by test signals, e.g., as in applications of time-domain reflectometry. Faults here considered include local modifications of the propagation characteristics of a transmission line. It is shown that the response of faults are strongly dispersive in nature, which implies that the peak of their echo is far from providing an accurate measure of the severity of the fault, as it heavily depends on the frequency content of the test signal, as well as on the length of the fault. It is argued that fault detection in transmission lines is an ill-posed problem that requires a priori knowledge on the fault itself. These results are important for applications of time-domain reflectometry methods, particularly for early-warning monitoring of potentially critical faults from their onset, since it is shown that echoes from faults tested at relatively low frequencies can lead to underestimate their actual severity.