Normalisation

Figure: Normalisation: the top spectrum is the normalised version of the lower spectrum. Note that the flux values in the uncorrected spectrum have been scaled and shifted for this plot.

If you are not flux calibrating your data then you will probably want to remove the low-frequency instrument profile from the data. This process is known as normalisation or blaze correction. The figure above shows a plot of a spectrum and the same spectrum after normalisation. The continuum is more or less flat in the corrected spectrum. Normalisation can be useful when you want to look at absorption line profiles, model their shapes, or determine their widths.

There are several approaches to normalising the data. One method is to fit a polynomial or other curve to the spectrum and then divide it by the curve. This often works; however, it may well be necessary to manually select which parts of the spectrum to fit as strong spectral features will lead to a poor fit to the continuum. Another method is to `draw' points of the continuum on to a plot of the spectrum and fit a curve to these points. This is usually an entirely manual process.

Another method is to fit a curve to the extracted `spectrum' of a flat field and use that for normalisation. As for the object spectra, this method will only work if the flat-field is devoid of strong spectral features (which it should be, otherwise it isn't much use for flat-fielding).