"is zero, causing a single mean COM model to be used for all sub-arrays. In such cases, if the common-mode signals estimated from the individual sub-arrays show significantly different structure, subtracting a mean COM model will leave large residuals in the bolometer time streams. If a FLT model is being used, the high pass filter will remove much of this residual signal, but any high frequency component will remain, often causing strong
"in the map.
The COM model can identify and flag time slices where the individual common-mode signals appear to be
significantly different to the mean common-mode signal. The individual common-mode signals are first
high-pass filtered using the same filter as the FLT model. At each time slice, the mean and standard deviation of
the remaining high frequency common-modes are found. The RMS of these standard deviations, taken over the
whole time stream, is then found. All bolometers within individual time slices for which the standard deviation
" times the RMS value are flagged in all sub-arrays, causing them to be excluded
from the FLT model and the map. Note, this flagging process is performed from scratch on each
iteration - that is, samples flagged on a previous iteration are tested again on each subsequent
iteration and may be
" if the differences between individual common-mode signal become
Setting this parameter to 3 is a good starting point. Setting it to zero (the default) causes this consistency filter to
be skipped. See also parameter
" .