In this thesis a sampling time domain measurement system primarily intended for measurements on radio frequency power amplifiers is discussed. The need for such a measurement system is established. Impairments due to non-ideal measurement instruments are discussed as well as methods to compensate for these impairments. Techniques to improve upon the raw measurement performance of the measurement instruments with regard to bandwidth, dynamic range, linear and nonlinear distortion are discussed. A method to simultaneously find the phase and amplitude ripple of a vector signal generator and a vector signal analyzer is presented. The method is verified with extensive measurements.
Two techniques, frequency stitching and Zhu's generalized sampling theorem, to extend the effective measurement bandwidth of the measurement system is discussed and evaluated with measurements. They are both found to be able to extend the effective bandwidth for measurements of output signals of nonlinear power amplifiers with more than five times.
The measurement system is used for sampled input - output measurements of power amplifiers and the obtained data are fitted to different behavioral power amplifier models including memory. Some different behavioral models are evaluated and compared for different kinds of power amplifiers. A neural network model and extensions to the well-known parallel Hammerstein model are specifically discussed. The parallel Hammerstein model are also used together with frequency stitching and Zhu's generalized sampling theorem.
A general hardware and software structure of a versatile measurement system based on virtual instruments for measurements on power amplifiers is discussed in some detail. Special attention is given to the software architecture and to the concepts of hardware and software reusability.
An automated, fast, accurate and production-friendly method for two-tone power and frequency sweep measurements, including measurement of the phase of the intermodulation products in addition to the amplitude, is also presented.