Improvement of the smoothing procedure via preliminary logarithmic transformation of the x-ray spectrum

Spectrum processing is a critical step in energy-dispersive X-ray fluorescence analysis, with very strong influence on the final performance of the method. Spectrum processing includes two major procedures: (i) preliminary processing (which deals mainly with reduction of random and systematic noise); and (ii) evaluation of peak parameters (most often peak area, and in some cases also peak position, amplitude and width). Preliminary processing solves two main problems: (i) search and correction of erroneous channel contents (outlier values); and (ii) random noise reduction, Several procedures for search and correction of outlier values have been proposed for Raman, gamma and X-ray spectra. Two approaches for reduction of random noise are employed. The first is based on filtering in the frequency domain by multiplying the Fourier transform of the spectrum with an appropriate transfer function. The smoothed spectrum is obtained after inverse Fourier transform. The second approach is based on the assumption that each small region of the spectrum can be approximated by a polynomial function with a sufficient accuracy. By means of the least-squares method one ran construct a digital non-recursive filter for suppression of the high frequency random noise. This approach has been known and used for a long time, but it became very popular after the publication of Savitsky and Golay and later corrections. 20 refs., 4 figs