Year: 2018
Type: article
Abstract: For several decades, computers have helped analog designers with circuit simulation and evaluation. To further simplify and speed-up designer’s work, novel methods are being introduced that help to fine-tune numerical parameters to meet the performance criteria. With a lack of capable engineers, a shortage of specific knowledge or time to design an analog building block, software for fully automated synthesis of both topology and parameters is becoming crucial. Most research in this field is based on circuit modifications according to evolutionary principles of surviving of the fittest. One of the challenges of the design of appropriate software is a representation of a circuit topology that will allow topology modifications with the smallest possible computational effort. Many existing solutions suffer either from the uncontrolled growth of the size of the circuit (so-called bloat) or from the limitation of the topology structure to a set of predefined blocks. In this paper, we discuss an analog circuit topology representation in a form of a binary upper-triangular matrix that is both bloat safe and offers a large solution space. We describe the basic structure of the matrix, the redundancy phenomena of logical elements, and the translation of the matrix representation to a regular SPICE netlist. We use an evolutionary algorithm to evolve the topology matrix and a classical parameter optimization algorithm to tune the circuit parameters. Based on a high-level circuit definition and a fixed building-block bank, our topology representation technique showed success in a fully automatic synthesis of passive circuits. We demonstrate the ability to automatically discover a passive high-pass filter topology.
Authors: Žiga Rojec, Jernej Olenšek, Iztok Fajfar
Institution: University of Ljubljana
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $215
OpenAlex ID: W2920003549
