We studied the coincidence between the loudest MF events seen in the Dark Fringe (DF) and glitches in Pr_B2_3f_d1_ACp detected by MF.

Plots and results are here:

http://wwwcascina.virgo.infn.it/DataAnalysis/Burst/GLITCHES/coincidence.html

Some comments:

• No optimization of the coincidence parameters has been done (actually the objective was to produce a trigger lists without all these "z" events)
• A nice thing is that it seems that Pr_B2_3f_d1_ACp trigs before Pr_B1_ACp, but actually not always with the same delay (there is kind of 50 ms "jitter" depending on the segment). I suspect that is due to timing estimation accuracy of these oscillating events ...
• Parameters used for all segments:
• Coincidence window: +/-100 ms
• Pr_B2_3f_d1_ACp SNR threshold: 15
• DF SNR threshold: 5
• It seems that the correlation between Pr_B2_3f_d1_ACp and "z" events is high.
• The trigger rate is clearly different from one segment to another!

Effect on the events distribution (rough test):

Some veto lists have been generated with:

• threshold SNR>15
• Time window around the peak time of +/- half of the duration of the events estimated by MF (those events are rather large, can be as large as half a second, and hence the peak time difference between DF and Pr_B2_3f_d1_ACp can be larger than 100ms for instance ... ). Question: is that correct to do that? Is it really needed? 

Then the veto is applied on DF triggers. Here is a plot of the results. All (except very few) the loudest events are vetoed. The loudest is not killed while it should be. Actually there are 3 gliches in less than 2 seconds. I guess there is a problem of timing accuracy.