Vector Coding Method for Symmetry Detection in the Analog Circuits

摘要

Symmetry constraints in the analog circuit layout have significant impacts on analog system performance. In this paper, an efficient framework is proposed to detect symmetry pairs of analog devices from the circuit netlist. Firstly, we extract the specialized feature vectors of a MOSFET device from an analog circuit netlist, and then obtain the coded vector for the device connection relationship between two MOSFET devices. According to the coded vectors, the basic building blocks of the analog circuits can be classified and the basic symmetry constraints of the circuit is recognized. Furthermore, more symmetry constraints of devices are detected through the combination of the basic building block recognition and feature representation. We executed the experiments on the open-source analog dataset of MAGICAL and our tape-out analog circuit. The experimental results showed that the proposed framework with the vector coding method achieved the promising performance in terms of the accuracy on symmetry constraint detection and building block recognition.