Sensitivity analysis and empirical optimization of cross-domain coupling on RFICs using polynomial chaos expansion

This work aims at investigating the variability of internal coupling mechanisms between different domains of a Radio Frequency Integrated Circuit (RFIC). In doing so, an Equivalent Circuit Model (ECM) is used in order to model the various coupling paths. First, the elements of the ECM are assumed to be uniformly distributed and their exact values therefore not fully known. Afterwards, the impact of theses uncertainties on the cross-domain transfer function, i.e. the coupling from the digital domain - the aggressor - into the other more sensitive domains - the victims - is analyzed using Polynomial Chaos Expansion (PCE) in combination with a sampling-based technique known as Stochastic Testing (ST). This approach intends to account for not accurately known configurations of the RFIC and package. In addition, the PCE representation allows for conducting sensitivity analysis through the derivation of conditional variances, i.e. Sobol indices, and sensitivities. PCE in combination with ST offers an efficient way to analyze the coupling effects. Finally, an improved version of the original design is proposed using an empirical optimization approach based on the techniques described above.