Mandatory:

• The input timeseries may have values in either counts per bin (of integer data type int16 or int32) or counts s (real32). If lccorr finds counts per bin in the input TS it converts them to (real32) counts s in the output; the same goes for errors, if present. If an error column is not present, lccorr calculates Poissonian values and stores them in the output.

• If a background TS is supplied via the withbkgtsset and bkgtsset parameters, lccorr converts these values similarly to the above and stores them in columns BACKV and BACKE in the output file. If the source TS already contains columns of these names, the values these contain are copied to the output only if no separate background file is supplied to lccorr. If a separate background file is supplied, then the contents of BACKV and BACKE columns in the source TS are discarded. Columns of these names in the background TS are always discarded.

• If a background TS is supplied, but this has a different binning scheme to the source TS, lccorr rebins the background TS to the same binning scheme as the source TS. Bins which are outside the duration of the original background time series are filled with nulls; sections of the original background time series which extend beyond the duration of the source time series are discarded.

• The background TS values are multiplied by the ratio of the source to the background collection area (see subsubsection 5.5.3 for details on how these areas are calculated).

• Each event is initially assigned the central time of the readout frame in which it occurs. But the bin sizes of time series are in general not integer multiples of the CCD readout period. This has potential to cause aliasing or Moiré noise, illustrated as follows. Say we have the following (graphically illustrated) situation:

  Frames boundaries along the time axis:
  
    |---------|---------|---------|---------|---------|---------|---------|
  
  Number of events in each frame, aligned with the time of frame centre:
  
         6         6         6         6         6         6         6
  
  Numbered bins:
  
        |--------- 1 ---------|--------- 2 ---------|--------- 3 ---------| etc
  
  Events in bins: 18                    12                    12
  
  but should be:  13.2                  13.2                  13.2
  

  I have neglected the usual natural noise in the signal in order to illustrate the problem better. lccorr corrects for this problem. In the case that the collection region lies entirely on one ccd/node, the correction is done by applying the following formula:

                                                    bin_duration
  true_rate(bin) = measured_rate(bin) x ------------------------------------- .
                                         (no. frames in bin) x frameDuration
  

  The correction is more complicated when multiple ccds are involved: in that case the total frame time in the bin for each ccd must be weighted by the fraction of events recorded from that ccd.

  The MOS event-processing chain contains a facility to dither the event times, which prevents the occurrence of binning noise. This is recorded in the keyword RAND_TIM in the EXPOSUnn extensions of the event list. If lccorr finds that this keyword is set, that ccd/node is not included in the binning noise correction. Such dithering is not yet done in PN.