SummaryAn analytical stage‐based harmonic distortion (HD) analysis method for multistage operational amplifiers (Op Amps) is developed in this work. This work contributes two fundamental methods that make the analytical HD prediction possible at the circuit level. Firstly, we propose that the traditionally used first order small‐signal transistor quantities gm (transconductance) and go (output conductance) in the gm/ID design methodology for bulk complementary metal‐oxide‐semiconductor (CMOS) technology can be extended to the higher order quantities gm(k) and go(k) ( ). With proper normalization, these quantities become neutral to the device dimensions and operation currents, hence can be precharacterized by sweeping simulations and used as lookup tables. Secondly, we further develop analytical nonlinearity expressions for a set of commonly used amplifier stages, represented as the functions of the nonlinearity parameters gm(k) and go(k) of the transistors that form a stage circuit. A combination of these two fundamental methods on hierarchical nonlinearity modeling enables us to apply the existing analytical HD estimation methods for the stage‐form macromodels to predict the circuit‐level HD behavior, overcoming the need of running repeated simulations under device resizing and rebiasing. The proposed harmonic distortion analysis method has been validated by application to real multistage amplifiers, achieving HD prediction results in excellent agreement to fully transistor‐level circuit simulation results but with substantial speedup.

Behavioral modeling of analog integrated cir- cuits has many applications. When used for sizing, ac- curacy of compact models can be used for directly eval- uating part of circuit performance targets, helping to quickly forming initial sizing for further iterations. In this paper we elaborate on modeling improvement of the current buffer (CB) compensation recently received attention in multi-stage operational amplifier (op amp) designs. Limitation in conventional models is pointed out and improvement is proposed. We demonstrate via cases that the improved macromodels can leverage the efficiency of sizing by incorporating the gm/ID method.

No abstract

No abstract