On the determination of noise parameters of low‐noise transistor devices

Abstract

In this paper, we discuss the experimental determination of noise parameters for high-electron-mobility transistor (HEMT) devices and analyze the problems encountered with conventional extraction techniques. HEMT devices are distinguished by a very small minimum noise figure of only a fraction of a dB, and this minimum noise is accomplished with a source reflection coefficient very close to the edge of the Smith chart. These properties impede a direct experimental determination of the optimum noise impedance by means of a tuner, and it has to be extrapolated from measurements of a large number of tuner states. Common linear regression techniques are found to yield erroneous results and even may fail completely to determine the noise parameters. We develop criteria to check the uncertainty of the result from the linear regression and show that, in general, the result needs to be examined for consistency. The occasional failure of extraction is traced back to a consistency violation in the conditions the elements of the correlation matrix need to satisfy. A new iterative technique is suggested for the extraction of noise parameters that warrants consistency. Moreover, a correction of the receiver gain is found to be necessary to obtain a good reproduction of the measured power patterns by the extracted noise parameters. This fact was attributed to thermal drift of the receiver