A 3-D simulation of substrate currents is crucial to analyze parasitic coupling effects due to minority carrier injection in smart power ICs. In this paper, a substrate parasitic extraction methodology is introduced by dividing the IC layout into elementary elements to solve the continuity equation for minority carriers in the volume based on the finite-difference method. A substrate parasitic network is derived from the mesh generated through the existing mixed-signal design flow. The induced substrate model is included in circuit simulators such as SPICE to predict the effects of substrate couplings during the design phase. Furthermore, this analysis enables optimization of layout with minimal parasitic effects. By linking the substrate model to the active components, the couplings between the integrated circuit with the substrate parasitic currents can be analyzed during circuit simulations. Simulations and measurements on an high voltage driver reveal consistent results and therefore confirm the validity of the method. Therefore, the approach developed herein is effective to predict parasitic couplings due the injection of minority carriers.

In smart power IC technology, low and high voltage circuits are integrated on the same substrate. The commutation of the high voltage circuits can induce substrate parasitic currents which can severely disturb the operation of the low voltage circuits. The parasitic currents due to minority carriers in the high voltage technology can be significantly high. However, the minority carrier propagation into the substrate is not considered in most of existing circuit simulators. In this paper, a novel computer-aided design tool for substrate parasitic extraction is proposed. A simple circuit with an injecting and a collecting N-wells over a P-substrate is studied. With the distance between the wells varying, the lateral bipolar effect is illustrated. The spectre simulation results of extracted substrate equivalent circuit are compared to a TCAD simulation results. The comparison shows an acceptable relevant error. However, the simulation time was reduced by approximately 1400 times with respect to the TCAD.