DIGITAL SMOOTHING OF INFRARED SPECTRA

Abstract

Using our mini-computer (PDP-12) on line with our spectrometer, we have investigated the effects on spectral lines and on noise of digital least squares smoothing by convolution. We use the algorithm made popular by Savitzky and $Golay.^{1}$ The most important ``parameter’’ is the length of the smoothing range. For single lines we can show that for maximum discrimination against noise with minimum effect on line height, half-width, and shape, one should smooth only once with a smoothing range equal to the full-width at half-height. To smooth over a range longer than 25 sample points we use the general equation for the coefficients of the convolution, as given by Whittaker and $Robinson.^{2}$ We are now investigating the effects of smoothing on overlapping lines.