Inverse Simulation - An Error Analysis

This paper analyzes the numerical errors of two inverse simulation algorithms: differentiation inverse method and integration inverse method. The differential inverse method uses numerical differentiation as a key step, and the step size used for numerical differentiation directly impacts on the simulation results. This paper proposes a global error measure to evaluate the accuracy of the inverse simulation, instead of relying on the conventional local truncation errors. While the integration inverse method avoids numerical differentiation, it is pointed out that there is another numerical instability problem the integration inverse method may encounter when there are uncontrolled variables in the simulation. Examples are provided to facilitate the discussion.