Correlated Bayesian Model Fusion

Abstract

Tunable circuit has emerged as a promising methodology to address the grand challenge posed by process variations. Efficient high-dimensional performance modeling of tunable analog/RF circuits is an important yet challenging task. In this paper, we propose a novel performance modeling approach for tunable circuits, referred to as Correlated Bayesian Model Fusion (C-BMF). The key idea is to encode the correlation information for both model template and coefficient magnitude among different knob configurations by using a unified prior distribution. The prior distribution is then combined with a few simulation samples via Bayesian inference to efficiently determine the unknown model coefficients. Two circuit examples designed in a commercial 32nm SOI CMOS process demonstrate that C-BMF achieves more than 2× cost reduction over the traditional state-of-the-art modeling technique without surrendering any accuracy.