Defining Analog Standard Cell Libraries for Mixed-Signal Computing Enabled through Educational Directions

This work presents a mixed-signal cell library built through multiple generations of educational experiences. Digital standard cell libraries are ubiquitous for commercial and academic IC design. Analog standard cell libraries are rare within system-level analog design. Educational efforts in analog system design provide a path for developing these standard cell libraries. The effort included determining the essential blocks for this standard cell library based on previous explorations, as well as constraints to make the layout efficient. Analog and mixed-signal standard cells enable compilation of a high-level description of analog and mixed signal designs to full IC layout. A standard cell library definition allows components designed in any process to have broad utilization. A high-level to layout compilation empowers rapid movement between different IC processes. Digital standard cell design is ubiquitous for commercial and academic IC design (Fig. 1). Very few individuals do custom digital IC circuit or layout design unless they are building, or are in the process of building, digital standard cells. This intellectual structure even enables research into automated generation of digital standard cells [1]. Digital standard cells enable tool abstraction for compiling from higher-level representations (e.g. [2]). Analog and resulting mixed-signal design abstractions are in a very different place (Fig. 1). Major companies that have a wide array of digital standard cells (e.g. ARM) do not have analog standard cells. Companies offering larger blocks have struggled in commercial spaces (e.g. [3]). Previous research efforts into analog standard cells are very few [4], [5], [6]. One can generate an initial list of generic standard cells, such as amplifiers, comparators, DACs and ADCs, references, bias currents, oscillators, and filters [4], [5], [7], topics one would find in a classic analog circuit design textbook (e.g. [8]). A few blocks (e.g. Op-amps, capacitors, and switches) could build a filter [4]. One could make straight-forward generalizations from digital standard cells [6]. The typical view is that analog standard cells require considerably larger number of design parameters for each cell than digital cells, resulting in far larger difficulties in developing analog standard cell libraries. Interest in larger analog computing systems was initially fueled by neurally inspired computing and machine learning systems in the 1980s e.g. [9] ), starting soon after the digital VLSI approach had become an accepted practice (e.g. [10] ). These topics have regained some of that previous momentum. Research efforts in analog computing typically start with building a set of initial blocks to be used in later design. These efforts are excellent educational efforts by teams of High Level Concept