We considered several algorithms for recovering data from frequency-switched spectra. Naive linear deconvolution, a possibility for some but not all switching intervals, is rendered useless by the noise and baseline imperfections of real-world data (Sect. 2.1). The usual shift and subtract technique (2.2) treats baseline imperfections most benignly (though not necessarily well) and results in the lowest possible noise levels; when the data permit, this is the algorithm of choice. The bootstrap or EKH techniques discussed here (Sects. 2.3 and 2.4) can (i) recover line profiles for which the switching interval is too small for the shift and subtract algorithm to function; (ii) account as precisely as possible for fractional channel switching intervals; and (iii) check that broader-lined signals have not been lost. Though sensitive to baseline imperfections, they are capable of repairing spectra which might otherwise appear badly damaged or useless.
Acknowledgements
The National Radio Astronomy Observatory is a facility of the National Science Foundation, operated by AUI, Inc. under a cooperative agreement. I thank Darrel Emerson for pointing out the equivalence of the bootstrap and dual-beam restoration algorithms. I also thank the referee, Glyn Haslam, for his bemused, approving remarks.