This issue is a duplicate of kazewong#226.

farr opened on Jun 11

There should be two checks in post-processing that make sure that the data windowing and (for time-domain waveforms) waveform windowing occur when the signal in the data is well below the minimum frequency used to compute the likelihood.

For the data windowing check, I am imagining using the waveform model to compute the times at which the waveform (or mode(s) within the waveform) reach f_min, the minimum frequency contributing to the likelihood. As long as the bulk (>90%, >99%, maybe?) of these times are later than the time at which the data window reaches 1 in every IFO, then the check passes; otherwise it fails, and the user is warned to use a longer data segment (so that the data window corrupts earlier times in the signal).

For the waveform windowing check, I am imagining using the waveform model to compute for each sample the frequency at which the waveform begins (if that is after the start of the analysis segment) or the frequency at which the waveform enters the analysis segment; these frequencies should also be below the minimum frequency contributing to the likelihood (maybe not very much---19 Hz versus 20 Hz should be fine).