EVOLUTIONARY SEARCH FOR OPTIMIZED LNA COMPONENTS GEOMETRY

Abstract

In this paper, an optimized design procedure based on genetic algorithm (GA) for automatic synthesis of dual-band concurrent fully integrated low-noise amplifiers (LNA) targeted to 802.16d @ 3.5 GHz and 802.11b, g @ 2.4 GHz standards is discussed. The algorithm delivers the circuit elements geometry, rather than their values, and bias levels to secure the best level of LNA gain, input matching, output matching and power consumption. Working on the components geometry level aims at considering the elements parasitic effects. The basic cascode and a current reuse folded cascode LNA's are tried. GA as an optimization engine is programmed in MATLAB and performance evaluation in 0.18 μm RF CMOS TSMC technology is ceded to HSPICE. Results indicate that the automated scheme well computes the desired circuit in an acceptable time span; otherwise, it may be explored by either tremendous manual trial and error or astronomical cycles of an exhaustive search. This is not accomplished without imposing certain approximate search space constraints.