We show what is required of a good model, with emphasis from a process development point of view. The merits of good testchips, physical correctness of the model, and realistic skews (best/worst case) are investigated. Prediction of future technologies with some level of confidence is also looked at. A close working relationship between process integration, model engineering, and circuit design is considered. Advantages of models with well-behaved, smooth, glitchless characteristics and well-behaved derivatives are covered. Parameter correlations, parameter redundancy, and model documention round out the basic model discussions. Extra features are also covered, including noise models, high frequency characterization, diode models, parasitic bipolar models, etc. And finally, the need for models to evolve with manufacturing improvements is highlighted.
Bookmark or share