RF instrument accuracy - Some effects of measurement uncertainty

Abstract

When measuring the RF noise figure on-wafer on ultra low-noise GaAs microwave devices, the magnitude of the noise figures measured become comparable to the instrumentation accuracy. It becomes necessary to determine the discriminating capability of the instrumentation, and take this into account. For the GaAs device discussed, the instrumentation accuracy is determined by running a repeat measurement on a statistically large sample, and calculating the limits inside of which the measurement system is not capable of discriminating. These measurements are done on an RF autoprober, which allows the noise figure of each of the test devices to be characterized repetitively at multiple frequencies without operator intervention. The results and their implication with respect to conclusions that can be drawn when searching for process performance improvements which are comparable to the discriminating power of the instrumentation are discussed. Some of the alternate statistical techniques applied to attempt to discriminate these improvements are discussed.<>