Adaptive circuit design methodology and test applied to millimeter-wave circuits

Abstract

A theoretical framework is proposed that allows the prediction of RF and mm-wave circuit performance with on-chip sensors through the use of indirect sensing with Bayesian model fusion (BMF) instead of through costly and difficult direct integrated measurement. Because the simpler sensors required by indirect sensing, along with the circuit actuators, can be efficiently integrated, a large number of adaptive circuit loops can be envisioned, which will allow transceivers to adapt to process variability and external changes such that the energy spent by bit transmitted is minimized. This methodology also enables significant convergence between the code used for circuit verification and that use for test. Finally, we described the design example of a 60-GHz LNA that can be self-healed using indirect NF sensing and adaptive biasing. Monte Carlo simulations show that the LNA average power consumption can be improved by 25% with adaptive biasing while achieving the same 100% yield as that achieved with the optimum fixed bias.