The final stage is to merge the scrunched orders from possibly several images into a single long spectrum. This is done using the `echmerge' program. If the data is from a region of the field where the full free spectral range is obtained then adjacent orders will overlap and the contributions from the overlapping orders are weighted with estimates of their inverse variances on the assumption of purely Poisson noise. Where one contribution is below a given fraction of the other then it is ignored completely. The variance estimates are based on median filtered versions of the input orders and the median filter box size is controlled by the `box' parameter.
The input images must all have the same X size, units and values and can be one-dimensional or two-dimensional. The output image is always one-dimensional and can be the same as either of the input files, in which case no new image is created. The second input image can be specified as blank in which case it is not required or used.
ICL> echmerge image=objs image1='' box=7 cutoff=4 output=objl
FIGARO A general data reduction system