ONOFF

     ONOFF

     This task produces tables of counts or fluxes, whichever is available,
     from ON-OFF scans. Select whether you require sky noise removal or not.
     The following is carried out:

       - compute the mean signal <Sj> and its variance for all subscan j.
       - remove a baseline from the raw signal by least-square fitting to a
         straight line a weighted sequence of means. Each mean value is the
         mean value of subsequent ON and OFF phases.
       - on sky noise removal it removes the high frequency noise in excess.
         In this case the task computes the weighted means of the ON and of
         the OFF phases for the whole scan.
         - it then removes from a channel k the signal difference between
           the baseline corrected signal and, depending on subscan, either
           the mean ON or OFF signal averaged over all other channels except
           the central one.
	 - it then removes the excess high frequency noise from all the
           channels and computes the mean noise corrected signal <Sj> and
           its standard deviation sj.
       - depending on the COUNTS_PER_JY conversion factor it compute either
         counts or fluxes and the errors W by averaging all weighted diffe-
         rences between the ON and OFF phases.
       - finally it converts all the beam offsets from the horizontal to the
         equatorial frame of reference for the epoch corresponding to the
         source coordinates.
       - the results are written to files with the extension .onf

     Avoid repetitive noise removal. If the fluctuations in two adjacent
     channels i and j are correlated, we expect to find on average the
     variance in both channels increased by the covariance between the two
     channels. Once the correlated term is removed, we expect the fluctua-
     tions to have faded away leaving uncorrelated noise with similar va-
     riance in channel i and j. Suppose the noise removal was very effec-
     tive and no scrap of correlated noise is left, then every further
     attempt to reduce the fluctuations with ONOFF will fail. Indeed, as the
     fluctuations will be independent further removal will tend to increase
     on average the variance of the j-th channel by a fraction M, where M
     is the number of channels from which the task derives the correlated
     noise contribution.

     Be aware that the data can be perturbed by low frequency noise that
     can create virtual positive or negative sources. Moreover, the low
     skynoise variation can hardly be estimated (in general, not at all)
     and, so far, is not taken into account in the standard deviation of
     the results.