Improved Receiver Characterization and Source Selection Technique for Microwave-Frequency Noise Measurement

Abstract

An accurate noise measurement is essential to a proper characterization of a noisy device. In the 1950s, the IRE first proposed the classical noise parameters for characterizing a noisy linear two-port network, and subsequently a measurement and extraction procedure. Since then, the task of accurately characterizing the parameters has always been challenging due to the sensitive nature of the noise parameters. This is especially so for an on-wafer device noise measurement, as opposed to that of a packaged device, due to various factors such as the lower noise level and losses in the signal path. Combined with the downscaling of the MOSFET technology in recent decades, which also improved the device's noise performance, they make noise measurement and characterization become even more difficult. A typical noise measurement starts with the calibration or characterization of the measurement system. This step is as important as the measurement itself in terms of the ultimate accuracy of the results. This thesis presents a noise receiver characterization method which improves upon existing methods by accurately taking into account the different reflection coefficient of the noise source between its hot and cold states. The improvement allows more precise determination of the receiver characterizations. Numerous studies have investigated the effect of the choice of the source terminations on the noise measurement results. These studies often provided contradicting suggestions on the selection techniques. In the thesis, a selection technique is proposed that allows quick determination of desirable terminations. Analyses using real measurement data on a 65 nm n-type MOSFET show that the proposed technique is able to provide terminations that yield noise parameters with smaller uncertainties relative to other terminations.