Sampling Issues in the Design of Experiments for the Undergraduate Engineering Laboratory

An extremely important aspect of the proper design of an experiment is specification of the sample size, sample rate, and duration of test.When sampling real signal data from the wide variety of transducers currently available, the presence of noise, generated from many sources, usually makes it necessary to sample the associated input signal numerous times in order to determine accurate statistical information; typically mean and standard deviation.From these statistics, and the associated sample size, it is possible to arrive at a reasonable estimate of the confidence interval for the sampled signal mean.Furthermore, in some applications, for example in acquiring statistical information about turbulence intensity, the focus may also be on the standard deviation itself.This paper provides a simple intuitive, but quantitative, procedure that students can follow in the undergraduate engineering laboratory to ascertain reasonable estimates of the required sampling information.The procedure has the advantage of not requiring complex background in sampling theory.In particular, it does not require detailed theoretical understanding of auto-correlation or cross-correlation statistical concepts that the typical undergraduate student has not yet acquired, which makes it readily applicable to the introductory undergraduate engineering laboratory.The technique is illustrated using simulations of time-series data generated by LabVIEW, as well as from real sampled signals.The paper also addresses the confidence interval for the standard deviation, which is frequently not given specific attention in the undergraduate laboratory experience.