## Abstract

In this paper, we propose a methodology for adaptive modeling of analog/RF circuits. This modeling technique is specifically geared towards evaluating the response of a faulty circuit in terms of its specifications and/or measurements. The goal of this modeling approach is to compute important test metrics, such as fail probability, fault coverage, and/or yield coverage of a given measurement under process variations. Once the models for the faulty and fault-free circuit are generated, we can simply use Monte-Carlo sampling (as opposed to Monte-Carlo simulations) to compute these statistical parameters with high accuracy. We use the error budget that is defined in terms of computing the statistical metrics and the position of the threshold(s) to decide how precisely we need to extract the necessary models. Experiments on LNA and Mixer confirm that the proposed techniques can reduce the number of necessary simulations by factor of 7 respectively, in the computation of the fail probability.

Original language | English (US) |
---|---|

Pages (from-to) | 465-476 |

Number of pages | 12 |

Journal | Journal of Electronic Testing: Theory and Applications (JETTA) |

Volume | 27 |

Issue number | 4 |

DOIs | |

State | Published - Aug 2011 |

## Keywords

- Analog/RF circuits
- Fault modeling
- Robust regression
- Simulation reduction
- Test statistics computation

## ASJC Scopus subject areas

- Electrical and Electronic Engineering