Ensemble Averaging of Repeatable Noisy Signals

Abstract

All signals produced by instruments contain various systematic and non-systematic components which are not produced by the phenomenon of interest (e.g., random noise generated by electronic components, background absorption in spectrophotometers, etc.). Collectively, all components which are extraneous and which the scientist would prefer to have entirely eliminated from the signal are called “noise.” Clearly, noise may be either random or systematic. Various methods may be employed to minimize or eliminate noise, but all depend upon a detailed understanding of the source of the noise and its relationship to the signal of interest. In many cases, careful hardware design can minimize noise via judicious use of electronic filters, good wiring practices, and the like. It is, however, often the case that due to experimental limitations it is necessary to accept a low signal-to-noise ratio from an instrument. In this situation, provided the noise has certain characteristics, it may become possible to significantly improve the signal-to-noise ratio by using a combination of appropriate data-sampling techniques and mathematical analysis. In this experiment, the use of such methods to eliminate random noise will be explored.