Abstract:
A system, method, and computer program product for facilitating model binning in circuit simulators. Embodiments enable specification of models spanning binning dimensions, such as device width and length, in a model group via inheritable model bins. New simulator modeling syntax and semantics eliminate much of the redundancy and parsing overhead from model parameter specifications in foundry process design kits. Indirect and optional inheritance is also enabled, allowing for fine grain and coarse grain grids in the same model group.
Abstract:
A system, method, and computer program product for extending device model parameter specification flexibility when using a subcircuit wrapper. Embodiments facilitate device modeling by allowing a modeling engineer to eliminate the explicit specification of a large set of wrapped device instance parameters as parameters to the subcircuit wrapper itself. A circuit designer may now use the subcircuit wrapper to specify an instance of the subcircuit without having to explicitly provide values for all such parameters. The simulator program's built-in device model calculates its default parameter values, which are often the result of complex expressions involving the other parameters, resulting in more accurate simulations. Subcircuit wrappers no longer need to be explicitly regenerated when a new version of the wrapped device model becomes available for the simulator (e.g., one that supports additional instance parameters that were not present on the earlier version when the subcircuit wrapper was created).
Abstract:
A system, method, and computer program product for automatically approximating conventional Monte Carlo statistical device model evaluation for circuit simulation with drastic speed improvements, while preserving significant accuracy. Embodiments enable quick inspection of the effects of process mismatch variations on single devices and even large circuits compared to standard computationally prohibitive Monte Carlo analysis. Statistical device model variation is calculated as if all such variation is due to changes in threshold voltage, even though other physical phenomena are known to contribute. Threshold voltage variation is modeled as a function of statistical variation, device size, and working bias condition. Circuit simulation is faster when the full internal device model parameter set is not rebuilt for every Monte Carlo analysis iteration. Embodiments are compatible with both conventional SPICE and newer Fast SPICE simulations. Circuit designers may capture design sensitivity to manufacture process changes more easily with simplified statistical models.