Etude et caractérisation des circuits micro-ondes linéaires bruyants

Abstract

In this thesis, a system that completely characterizes the noise temperature of linear microwave circuit over the 100 MHz to 2.9 GHz band is presented. The system is based on a "long" transmission line and frequency variation. It allows measurements through the classical y-factor method or the more recent z-factor method. When applying the y-factor method, the system uses a broadband mismatched noise source and an auto-calibration feature. The noise parameters values extraction from the noise measurement was done using two methods. The first one is based on an optimization routine, while the second one uses a novel set of simple and accurate formulas. This new extraction method, based on the frequency variation noise measurement principle and used with a simple hardware, can be a practical tool for workers in this field. To check the accuracy, we have used passive 2-ports designed to exhibit, versus frequency, a behavior close to a transistor. The passive device is measured with its noise parameters computed from the S-parameters. The results obtained by both methods are compared together and with respect to the values derived from the S-parameters. When measuring the noise by the y-factor method, mismatched noise sources is required. In this thesis, two methods are presented for calibrating broadband highly mismatched noise sources. The first one is a reformulation and extension of the procedure used for very accurately calibrating matched noise sources. It uses isolators and works whatever the source reflection factor. With the second method, the excess noise ratio of the noise source is numerically optimized until the measured reverse noise wave of an attenuator equals the theoretical value derived from its S-parameters. Both methods appear to give very consistent results.