This paper describes a C++ idiom for replacing types. This idiom allows effective and elegant implementations of patterns such as virtual object construction, message dispatching and tables. It also improves compile-time checks and simultaneously allows rows to be added dynamically to tables.SmartEnums have been developed within a project of numerical software for microsystem simulation. Copyright © 1999 John Wiley & Sons, Ltd.

We present an automatic method to produce compact equivalent circuit models of spatially inhomogeneous resistors. Local variations in space of the resistivity due to physical interactions such as magnetic fields or mechanical stress are automatically included. The equivalent circuit model is computed using symbolic algebra, such that the functional relation between the resistivity and the fields interacting with it is included in the circuit design model. Modeling is based on the discretization of the sensor geometry with a mesh of elements and vertex nodes together with the current continuity equation using the box integration method. The resistivity is described by the tensor field of electrical conductivity and depends on the physical interactions to be modeled. The element internode conductivity is mapped to a set of lumped conductances and transconductances (voltage controlled current sources) between the nodes of the discretization mesh. These conductances and transconductances are translated into an equivalent circuit net list. Optionally, the electrical network representing the sensor is simplified before translation by symbolic linear algebra. Thus, equivalent circuit models consisting of many simple elements can be generated as well as models with only a few, algebraically complicated elements. The method is demonstrated using the public domain circuit simulator SPICE3 for the example of a magnetic Hall sensor, with and without the piezoresistive effect.

Accurate electrostatic simulations are required for the analysis of micro electromechanical systems (MEMS) and interconnects in very large scale integration (VLSI) design. Typical simulations involve complex three-dimensional (3-D) geometries together with various dielectric materials, conductors, and boundary conditions. The boundary element method is well suited for such computations. For highly accurate solutions, the meshing of the geometry becomes increasingly important. A scheme is presented which allows generating an optimal mesh automatically based on a coarse initial discretization, e.g., a CAD model. An error indicator derived from boundary integral equations monitors the solution accuracy in each boundary element. H-type or p-type mesh refinement is applied to areas which contribute strongly to the overall error. The method applies to both two-dimensional (2-D) and 3-D simulations containing elements of various orders and shapes. The generated refined meshes result in significantly higher solution accuracy for a given simulation size.