Conclusion

Abstract

Over the years, analog CAD has progressed from SPICE simulators, to schematic and layout editors, to recent nominal global optimizers and local variation-aware optimizers. This book has proposed and demonstrated a roadmap for the next several steps of tool evolution (Fig. 11.1). The roadmap starts with global variation-aware optimization, then progresses to trustworthy structural synthesis, and then adds variation-awareness and novelty to the structural synthesis.

Along the way, this book has also demonstrated techniques to use automation for knowledge extraction so that the designer can maintain and build his insight. Techniques include the extraction of template-free symbolic performance models, topology decision trees, and nonlinear impact analysis of topology vs. sizing/biasing variables. The techniques developed apply not only to analog CAD, but to a whole host of other AI problems related to the robust design of complex systems.

If one takes care of the means, the end will take care of itself.

– Ghandi