Sensitivity analysis of cable crosstalk to uncertain parameters using stochastic reduced order models

Abstract

This paper presents the sensitivity analysis of cable crosstalk against different uncertain variables using the stochastic reduced order model (SROM) method. A ranking of these uncertain variables is produced based on their impact levels on the crosstalk. This study shows that the complexity of the statistical problem can be reduced by ignoring the weak variables without affecting the accuracy. The statistical result is obtained with the SROM method, and compared with that of the conventional Monte Carlo (MC) method. It is found that the SROM method can reduce the computational cost required by the converged MC by at least two orders of magnitude. Also the SROM method is non-intrusive and straightforward to implement. Therefore, it could be an ideal solution to obtain the statistical solution in randomness-embedded electromagnetic compatibility (EMC) problems.