A constraint driven technique for MOS amplifier design

Abstract

The performance of a single stage amplifier when expressed as a function of designable and technology related parameters can be improved by adjusting the component values and transistor dimensions. In the process to optimize the performance, the corresponding parameters can be expressed in terms of objective function and constraints so as to graphically form a polytope type feasible region. Thereafter, simplex method and interior point based method are used to find an optimal value by traversing through each of the corner point and interior of the polytope formed. Consequently, the amplifier design problem can be realized as a special form of optimization problem called Non Linear Programming (NLP), for which efficient global optimization methods have been developed. The present method yields completely an automated synthesis of single stage amplifiers directly from the specifications. In this paper, the proposed method is first described showing in detail the formulation of the design problem as NLP for a specific amplifier circuit and then applied to a wide variety of amplifier architectures. The optimal trade-off curves related to performance measures such as small signal gain(Av), unity gain frequency(UGF), slew rate(SR) and power are derived so as to observe the corresponding dependencies.