Co-simulation of Circuit/Circuit type Solvers for EMC Applications Using a New Relaxation Method

This paper studies a new co-simulation scheme for coupled problems that is based on the principle of the adapted transmission line with an application to circuit-circuit type coupled simulations. Co-simulation between two or more circuit solvers is usually based on waveform relaxation methods and it results in an iterative process whose convergence depends on the type of interface variables (IVs) on the interface between the sub-circuits. The convergence of the IVs of this new method is tested in two examples of circuit simulations: a functional simulation of a Buck converter and an EMC simulation.

I. Introduction

The increasing number of static conversion devices (power electronics) in automobile systems and their growing impact on the on-board electronics and on the vehicle's electrical network involve severe constraints in terms of reliability and safety. Today, car and aircraft manufacturers are faced with two major concerns: On the one hand, they must ensure compliance with standards limiting parasitic emissions and the exposure of people to electromagnetic fields (EM) radiated inside the vehicle, and on the other hand, they must guarantee the proper functioning of all equipments in a normally polluted EM environment. The challenges in the vehicles for the future move beyond these aspects through the vulnerability to EM aggressions of the autonomous driving systems. Indeed, the cables and the inputs/outputs are inputs for disturbances aimed to dysfunction the guidance or to facilitate EM spy on the vehicle status. It is therefore essential to evaluate the levels of EM disturbances of these systems from the design phase. Although regular efforts to develop predictive models of electromagnetic disturbances [1]–[5], the measurement remains essential for complex embedded systems in the automotive industry. In addition, despite the improvement of system-oriented modeling and the computing power available over the last twenty years, the direct use of existing simulation tools (MATLAB, PSIM, SABER, CST Studio) for the analysis of a large system, considering all the strong interactions between subsystems, is an almost impossible mission. Several reasons can be given: •

In the automotive domain, where different industrial partners are involved, a multi-level analysis leads to multi-scale problems that are difficult to access by only one of the partners.

•

A global simulation is very costly in terms of time and storage memory.

•

The global numerical simulation of the system can only be carried out if all the models of the subsystems are known and available (different simulation platforms, evolution of the models in the development phases, confidentiality).