Speeds up convergences and reduces ringing by excluding sharp features in the bolometer
time streams from the filtering performed by the FLT model FLT.CLIP
"flt.filt_edge_largescale" etc. Any such sharp features are
replaced by a linear function joining the data values before and after the sharp feature. Sharp
features are found by first smoothing the time stream using a simple box kernel with a width
corresponding to the largest features passed by the FLT filter. The RMS of the residuals between
the smoothed and original time stream is then found. Any samples that have residual
greater than " flt.clip" times the RMS are set bad. Additional samples equivalent to 0.25
of the filter size are also set bad on either side of a clipped sample. If any such samples
are found, the process is repeated (i.e. the remaining data is smoothed, the RMS residual
recalculated and the check for outliers repeated). Once no further outliers are found, the bad data
values are replaced by values that interpolate the adjoining good values using a linear
function.
The value supplied for this parameter can optionally be a comma-separated list of two clipping levels enclosed in parentheses. In this case the first clipping level is used as described above. The second clipping level should be smaller than the first. It is used in much the same way as the first clipping level, except that points that would be clipped using the second clipping level are only clipped if at least 3 of the 5 nearest neighbours (including the point itself) are above the second clipping level.
This parameter is used to prevent the ringing that would otherwise be caused by the action of the
filter on the sharp features. It should usually be seen as an alternative to the masking controlled via
the " flt.zero_" set of parameters. Note, in some cases it may slow down convergence due to the fact
that each bolometer is masked independently of the sky, thus reducing the consistency between each
bolometer and the sky map.
Note, the masking described above can be restricted to selected iterations by setting a value for the
parameter "flt.zero_niter" . [0.0]