Flat Fielding

A flat-field frame is needed to correct for the pixel-to-pixel sensitivity variations in the detector (and the small-scale variations in the throughput of the instrument optics). When imaging, an evenly illuminated scene is used as the flat field. For spectroscopy we would want to use a light source which is `white', i.e., the same brightness at all wavelengths. In practice this is not possible and the lamp used for the flat field has some wavelength-dependent variation in brightness. (The effect becomes greater as dispersion increases.) To remove the response of the lamp, a low-order curve is fitted to the response - assuming the spectrum runs with dispersion parallel to the X-axis, we collapse the flat-field frame by summing columns, producing a `spectrum' for the lamp, then fit a curve to this spectrum. The figure below shows how one of these `spectra' might appear. The spectrum has a simple, continuous shape with small-scale noise superimposed. Having fitted a curve, we divide the flat-field by it so as to leave the small-scale sensitivity variations only.

Figure: A section through a typical flat field for spectroscopy (left) and the same flat field after the large-scale curve has been divided out (right). Produced using ECHOMOP.