Toward automated reasoning for analog IC design by symbolic computation - A survey

Analog integrated circuit (IC) design highly depends on reasoning, which distinguishes itself from other areas of IC design. Most of its innovation arises from qualitative reasoning by a pencil and paper. Innovation on the circuit structure needs quick analytical justification. Circuit-level reduced-scale modeling is a popular reasoning means. Circuit simulation tools can only serve partial justification on a design, while design insight still has to be acquired via manual analysis. A basic question has been in existence for many decades: how can we automate the analog IC design process? Many analog synthesis tools proposed decades ago could not make it to this date in the design practice. In this survey the major reason is attributed to the black-box style of the tool design. Human designer's creativity is shielded away from the tool operation while the formal design knowledge hardcoded in those tools remains at a very primitive level. By analyzing the defects of those existing tools, this survey advocates an open tool development philosophy whose major goal is to support human-machine interaction. On the one side a design automation tool is mainly aimed at providing aid for tasks that require analytical deduction while on the other side designers are expected to exercise their creativity based on the machine-generated results. Such human-machine co-working style is believed to be a more feasible solution to analog IC design automation based on the currently available computation technology. In this survey the art of symbolic computation is promoted to be the enabling technology for computer-aided analytical generation. The symbolic computation technology today can support topological and analytical reasoning that is the most demanding need in the analog IC design practice. This survey further calls for more research on the formal methods that are applicable to design knowledge representation, human-machine interaction, and design inference. Some preliminary research results are reviewed and future research directions are pointed out.