SIGNAL-TO-NOISE RATIO ASSESSMENT AND MEASUREMENT IN SPECTROSCOPIES WITH PARTICULAR REFERENCE TO AUGER AND X-RAY PHOTOELECTRON SPECTROSCOPIES

An analysis is made of some of the problems associated with the measurement of signal-to-noise ratios in electron spectroscopy. We develop here the assessment of the signal-to-noise ratio for counting systems, in particular where the statistics are Poissonian. In order to measure the noise statistics accurately it is shown that at least 135 measures of the counts at a given setting are required. These may help to confirm or disprove the Poissonian nature of the count. At high counting rates the counting system dead time causes deviations from the Poissonian character. However, if the counting system gives Poissonian statistics, simple equations may be used to give the signal-to-noise values from spectral measurements. In systems with more than one detector the allocation of the data from the different detectors to different spectral energy channels may provide some effective smoothing so that the noise appearing in the spectrum may be up to 25% too low compared with a Poissonian estimate. In this case the true signal-to-noise ratio may only be observed from the detector outputs directly, prior to data reduction by the computer. In practical situations the dominance of the Poissonian counting statistics should be exhibited over most of the effective measurement range. However, when searching at the limits of spatial resolution in AES (very low beam currents) or at the highest sensitivity (small signals on a large background) in AES or XPS, other noise sources become important and should be considered.