On the Possibility to Speed Up the Analysis of Signals from Vector Network Analyzers in the Time Domain

Abstract

The aim of the study is to enhance the response of the measurement system for analyzing microwave frequency signals in the time domain based on vector network analyzers. The measurement cycle consists of two stages: obtaining frequency samples (sweeping) and performing calculations for transition to the time domain. The time taken to perform the first stage can be shortened only by degrading the dynamic range and/or reducing the number of frequency samples analyzed. The time taken to perform the second stage can be reduced by organizing the computational process into a few parallel threads and using fast algorithms with taking into account the features of the source data. A computing task can be organized into parallel threads provided that these threads are independent from one another. However, an attempt to do so encounters difficulties in using iterative fast algorithms for calculating the inverse Fourier transform. In addition, these algorithms impose certain limitations on the transform length. Nonetheless, an effective solution to the problem does exist. It is shown analytically that a several-fold decrease (in proportion to the number of computing device cores) of the computation time is possible, even taking into account the costs of setting up parallel computing (splitting the signal into disjoint sections to run independent computing threads and uniting the calculation results over each section). The theoretical conclusions have been confirmed by an experiment, which demonstrates a several-fold reduction in the computation time with increasing the number of independent computation threads. Moreover, since the experiment is based only on dividing the task into several computation threads, additional shortening of the calculation time is expected in using fast Fourier transform computing algorithms.