A discussion is presented of the implementation of thermal analysis in SPICE that analyzes the circuit by computing stability temperatures for each device at operating point in DC analysis. Conditions for thermal instability or thermal runway are discussed along with a critique of the current temperature models for some of the devices in SPICE. Thermal instability analysis, also implemented in SPICE, points out the specific devices in the circuit that lead to thermal instability.>

Statistical analysis simulates circuit performance variations caused by tolerance variations and other circuit production factors. The procedure of statistical circuit simulation is illustrated with the help of a simple circuit example. Measurements are made on a sample of this circuit. These measured results are compared with the simulation results from worst-case and statistical analyses without and with model parameter correlations for the devices used in the circuit. The necessity for properly including the model parameter correlations is evident from this comparison.

A four-terminal model is formulated for the depletion-mode MOSFET using simple charge-voltage relationships. Different regions of operation are taken into account according to the surface conditions such as accumulation, depletion, and inversion. This simple formulation is then modified to account for the second-order effects such as mobility reduction, drift-velocity saturation, and channel length modulation. Simple expressions are used to express the model parameters in terms of device dimensions. A charge-based capacitance model is used to compute transient currents. The depletion-mode model is implemented in the circuit simulation program HP-SPICE and the simulation results are discussed.

An avalanche generation model is developed and implemented in a circuit simulation program SLIC. The model provides accuracy, as well as simplicity in parameter extraction. Computer programs are developed for automated measurement and parameter determination of model coefficients. Comparison of measured and simulated results shows good agreement.