Measuring the uniqueness and variety of analog circuit design features Cristian Ferent, Alex Doboli à Department of Electrical and Computer Engineering, State University of New York at Stony Brook, Stony Brook, NY 11794-2350, USA article info Article history: Received 9 February 2010 Received in revised form 19 June 2010 Accepted 22 June 2010 Keywords: Analog circuit Synthesis Computer-aided design (CAD) Uniqueness and variety metrics Design features abstract Analog circuit design activity is currently a less formalized process, in which the main source for innovation is the designer’s ability to produce new designs by combining basic devices, sub-circuits, and ideas from similar solutions. There are few systematic methods that can fuse and transform the useful features of the existing designs into new solutions. Moreover, most automated circuit synthesis tools are still limited to routine tasks, like transistor sizing and layout design. Developing new design techniques that can combine the existing design features requires metrics that describe the uniqueness and variety of the features. This paper evaluates for analog circuits two such general-purpose metrics proposed in [1,2]. Three case studies are discussed on using the metrics to characterize the design features of current mirrors, transconductors, and operational amplifiers. The two metrics and the presented study is useful in producing an overall characterization of analog circuit features. This can help in enhancing the circuit design process, training of young designers, and developing new automated synthesis tools that can explore more solution space regions that are likely to include novel design features. & 2010 Elsevier B.V. All rights reserved. 1. Introduction Analog circuit design is considered to be by and large an art [3– 5]. This is arguably due to the circuit design activity being a less structured process that is not fully formalized as an algorithmic sequence of steps. Designers often rely on similarities with previous designs (experience), on analogies with solutions in other engineering domains, and even on inspiration drawn from biology and anatomy [6]. However, the main vehicle in creating novel circuit solutions is still a designer’s talent to combine the defining features of basic devices and building blocks (e.g., sub- circuits) into new design solutions. While the characteristics of the actual building blocks are well understood, the innovating process of creatively combining them is not. Designing analog circuits with new features is currently a slow, expensive, and error-prone activity, accessible mainly to a small group of experts. There are no systematic methods to help and improve the process of analog design through fusion and transformation of existing circuit designs. Also, educating de- signers is tedious and often spans 10–15 years. Without a theory on how to efficiently combine existing circuit structures into new solutions, it is hard to train engineers on developing efficient designs for new requirements. Furthermore, computer-aided design (CAD) tools are continuously challenged to keep pace with state-of-the-art circuit design. Still, most tools are limited to routine tasks, like transistor sizing and constructing layouts [3,4]. With few exceptions, tools cannot produce new circuit topologies or exploit the innovating features of existing, manually crafted designs. Novel methods are needed to help understand, emulate, and enhance the process of innovation in circuit design. An intriguing approach towards developing a theory on creative circuit design is to formalize a computational model that captures the process of feature combination and transformation in circuit design. This model would be based on the main cognitive steps of design innovation, such as expanding and contracting the active conceptual space, imposing a new context on a solution, similarities, and deconceptualization [7–9]. As in other engineering domains, like mechanical engineering design [1,2], the core of the model would include a set of metrics that can accurately express the uniqueness and variety of the design features of the circuits implementing various specifications. There is currently few work on metrics that describe the uniqueness and variety of design features. Research in design science has recently proposed new metrics used mainly in mechanical engineering [1]. The metrics characterize both the frequency and variety of the features of an individual design as well as a set of designs. Enhancements to the original metrics have been subsequently proposed to increase the scope of the metrics [2]. Alternatively, the new features of a design also depend to a significant degree on the characteristics of the design flow. Many engineering systems are designed through multiple iterations that involve parameter optimization and technological changes [10]. Design science research focuses on determining the Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/vlsi INTEGRATION, the VLSI journal 0167-9260/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.vlsi.2010.06.003 à Corresponding author. Tel.: + 1 631 632 1611; fax: + 1 631 632 8494. E-mail addresses: cferent@ece.sunysb.edu (C. Ferent), adoboli@ece.sunysb.edu (A. Doboli). INTEGRATION, the VLSI journal 44 (2011) 39–50