Privacy and Security Notice
Beam Trip Events Cut Analysis
Beam Trip Events Cut
Analysis
Preliminaries
The beam ramps up in a series of very short steps (tens of
microseconds) upward toward the desired current. Anytime the beam
becomes stable on a very small scale of a few nC per mps quartet
(1 mps =
1 imps = 1/30 s) the data acquisition system starts to count events.
Because the beam can become stable so quickly, events can
often be counted during this non-useful beam ramp-up period.
Plots
of detector yield show periods of high yield events that
correspond to the periods of stable low-current beam during these
beam-recovery periods. Figure 1 shows a plot of beam
current (charge_BCM1, in nC) over time and octant 1 CED9 direct yield
(NAY_OCT1_CED9_eD--basically the raw singles rate, in counts/nC) over
time for a section of run 31315 when the beam
was ramping up after a beam trip. Several of these short stable
low-current high-yield periods are apparent in these plots.
Figure 1

Once the beam has stabilized the data acquisition system uses the beam
trip event cut (tripeventcut) to determine how many events to
ignore before starting to take data. By increasing this cut
we can make all these short discarded periods long enough to overlap
and eliminate all of this non-useful ramp-up data. Figure 2 shows
how successively increasing this cut successively eliminates more and
more of this low-current data. The final plot shows that a cut of
250 eliminates all the low-current data. (Because all the low
current events had been eliminated, root automatically changed the
vertical scale for the final plot.)
Figure 2

Figure 3 shows the corresponding octant 1 CED9 yield plots.
Again, the final plot shows that a cut of 250 eliminates all the
low-current high-yield data. (Again, because all the high-yield
events had been eliminated, root automatically changed the vertical
scale for the final plot.)
Figure 3

Part I
Several collaborators have noticed that the
charge asymmetries have tails in addition to the Gaussian shape we
expect. We decided
to see whether changing the beam
trip event cut
would eliminate these tails. I first looked at runs with many
beam trips and no prolonged periods of steady
low-current beam. Figure 4 shows the beam current for the whole
of run 31315. At this time scale we see no prolonged periods of
stable low-current beam during the recoveries after beam trips.
Figure 4

I selected a few runs (31315, 31345 and
31346) with quick beam recovery at this scale and reanalyzed these
runs with the g0analyzer using several different trip event cuts for
each. Then I
looked at
the direct yield for octant 1
CED 9 versus
time. Figure 5, yield plots for all of run
31315, shows that a
tripeventcut = 250 or more eliminates all of the high-yield events
immediately following the beam trips.
Figure 5

Figure 6, plots of the octant 1 elastic
electron
charge asymmetry for the same
cuts, confirms that when the high-yield beam-recovery trips are
eliminated the tails also disappear from the asymmetry histograms and
the RMS stabilizes--again, after a cut of
250. (The mean is also stable at 0 within its error bars of
RMS/(N)^1/2.)
Figure 6

Runs 31345 and 31346 show similar results. A cut of 200 removes
the high-yield events and charge asymmetry tails for run 31345 (figure
7)...
Figure 7

...while a cut of 250 removes
the high-yield events and charge asymmetry tails for run 31346 (figure
8).
Figure 8

Because I'd been looking at runs close in number (31300s) I also looked
at a couple runs with many beam trips that were more spread out in run
number and, therefore, in time. Analysis of runs 31469 and 31551
also showed that a cut of 250 events eliminates the
unwanted events after a beam trip for these runs as well. We've
decided to set
tripeventcut=300 to be safe.
Part II
Setting the tripeventcut to 300 eliminates the unwanted low-current
high-yield events during relatively quick beam-recovery periods, but
during some runs, on the way to recovery, the beam settles into more
prolonged periods of stable low current. Figure 9 shows the beam
current for all of run 31316. Even at this large time scale we
can see a few periods during which the beam current stabilizes at a low
level before continuing up to the desired 60uA.
Figure 9

Figure 10 shows yield and asymmetry plots for run 31316. The
upper row of plots
shows many high-yield events from low
steady-current periods counted by the
data acquisition system even after cuts of 250 and 300 have eliminate
the initial low-current high-yield events immediately after a beam
trip. (For some reason the re-analyzed run contains only the first half
of the run data. But we can see by looking at the time axis that
the high-yield events in the upper plots of figure 10 correspond
to the prolonged low-current periods from the middle of figure
9.) The lower row of plots shows clearly that these steady
low-current events create tails in the asymmetry plots.
Figure 10

So, as a next step, I
varied the minimum cut on q_targ to eliminate all events at currents
lower than this setting.
(The default minimum had been 20nC--1800nC corresponds to 60uA.)
I found a few runs (31313, 31316, 31330, 31338 and 31340) with
prolonged stable low-current periods and reanalyzed them using minimum
q_targ settings of 1000, 1200, 1400 and 1600. Figure 11 shows
yield
plots for run 31316. (Plots to the right have higher tripeventcut
while lower plots have higher minimum
q_targ.) These plots show minimal change with different minimum
cuts on q_targ, and even with different trip event cuts once a minimum
q_targ cut has been defined.
Figure 11

Figure 12 shows that the corresponding elastic
electron asymmetries also do not have tails, regardless of the cuts.
Figure 12

All of the runs (31313, 31316, 31330, 31338 and 31340) with
prolonged stable low-current periods that I reanalyzed showed a similar
lack
of response to changes in either cut. The plots with tripeventcut
= 200 show that setting a minimum on q_targ can eliminate some of the
same events that the trip event cut alone eliminates. With the
minimum q_targ cut set at 1000 I reanalyzed runs 31316 and 31315 with
lower trip event cuts. The top row of plots in figure 13 shows
that a trip event cut of 150 eliminates the undesired high-yield
low-current events in this case for run 31316.
Figure 13

While the yield plots in figure 14, for run 31315, do not change
significantly with the different cuts, the bottom row
shows that a trip event cut as low as 100 returns the RMS values of the
asymmetry plots to the "good"
value (1.88e+4) shown in figure 6.
Figure 14
I didn't reanalyze runs with a minimum q_targ cut of less than
1000. I
could go further if we need it, but for now at least we know that the
minimum q_targ cut can be pretty loose. Of course we expect that
the allowable looseness of this cut could depend on the run. For
example, if the beam settled for a while at 50uA on its way up to 60uA,
the minimum q_targ cut would have to be fairly tight to eliminate these
events. But generally it looks like a trip event cut of 300 and a
minimum q_targ cut of at least 1000 is very safe.