Title:Machine Learning Peeling and Loss Modelling of Time-Domain Reflectometry

Abstract:A fundamental pursuit of microwave metrology is the determination of the characteristic impedance profile of microwave systems. Among other methods, this can be practically achieved by means of time-domain reflectometry (TDR) that measures the reflections from a device due to an applied stimulus. Conventional TDR allows for the measurement of systems comprising a single impedance. However, real systems typically feature impedance variations that obscure the determination of all impedances subsequent to the first one. This problem has been studied previously and is generally known as scattering inversion or, in the context of microwave metrology, time-domain "peeling". In this article, we demonstrate the implementation of a space-time efficient peeling algorithm that corrects for the effect of prior impedance mismatch in a nonuniform lossless transmission line, regardless of the nature of the stimulus. We generalize TDR measurement analysis by introducing two tools: A stochastic machine learning clustering tool and an arbitrary lossy transmission line modeling tool. The former mitigates many of the imperfections typically plaguing TDR measurements (except for dispersion) and allows for an efficient processing of large datasets; the latter allows for a complete transmission line characterization including both conductor and dielectric loss.