Abstract

It is shown that a suitable curve fitting procedure can reduce the standard deviation of intensity measurements made with a scanning diffractometer except when the intensity is large in relation to the background level. The procedure has been implemented with an Arndt-Phillips linear diffractometer, on-line to a Ferranti Argus 312 computer. The shape of the profile which is fitted to the observed profile is learnt and continually revised. In typical cases the standard deviation of the intensity, as determined from counting statistics alone, is about 0.4-0.6 times that given by alternative methods. In addition to profile fitting, a straightforward peak-minus-background treatment is always done, and the two results are compared. This enables a large measure of internal consistency checking to be done, so that certain types of error condition can be detected. Criteria are also developed whereby the better of the two results may be chosen in each case, having regard to such parameters as the peak to background ratio and the closeness of fit accomplished in profile fitting.