The AMB Cut

The AMB cut is coded as a SNOMAN hard-wired cut, `flt_amb_cut.for.' The parameters for the cut are found in filter.dat:

*DO   FLTH  10   -i(30I 10F 4I 4F -I)                  #.  FLTH: AMB Cut
#.
20041201        0 20380517 03331900   6*0   #.   Standard Database Header
20041201        0 20380517 03331900  16*0
#.
#.   The following are constants, not cut parameters, and should not
#. be adjusted without a re-calibration of AMB.
#.
59.5                         #. Pedestal for AMB Integral Measurement
#.
9.04                         #. Pedestal for AMB Peak Measurement
#.
1.20                         #. Mean of Integral/NHIT distribution
#.
10.94                        #. Slope for scaling of RMS of Integral/NHIT
                             #.  distribution as a function of NHIT.
#.
0.097                        #. Offset for scaling of RMS of Integral/NHIT
                             #.  distribution as a function of NHIT.
#.
0.054                        #. Mean of Peak/NHIT distribution
#.
0.973                        #. Slope for scaling of RMS of Peak/NHIT
                             #.  distribution as a function of NHIT.
#.
0.0043                       #. Offset for scaling of RMS of Peak/NHIT
                             #.  distribution as a function of NHIT.
#.
#.  Cut parameters---can be adjusted for different cut results.
#.

#.				Symmetric Cut values
3.7                          #. Cut value (in units of RMS) for Integral:
                             #.  How many sigma away from mean should we cut?

3.7                          #. Cut value (in units of RMS) for Peak:
                             #.  How many sigma away from mean should we cut?

20                           #. Minimum NHIT above which cut is applied.

1                            #. Cut mode:  1=OR of Integral and Peak Cut
                             #.            2=AND of Integral and Peak Cut
1                            #. Special handling for high NHIT events:
                             #.            1=If NHIT is such that we expect
                             #.               to be near or above the rail of
                             #.               INT, ignore INT test result
                             #.               if INT is in fact railed (ie,
                             #.               only use PEAK). If the same is
                             #.               true of PEAK also, pass
                             #.               event.
                             #.             2=Treat all NHITs the same.
#.
#.  Asymmetric Cut Parameter Values
#.
2                            #. Symmetric or Asymmetric cut mode
                             #. 1 = use asymmetric cut values
                             #. 2 = use symmetric cut values

4.0                         #. Upper cut value (in units of RMS) for Integral:
                            #.  How many sigma away from mean should we cut?

4.0                         #. Upper cut value (in units of RMS) for Peak:
                            #.  How many sigma away from mean should we cut?

3.5                         #. Lower cut value (in units of RMS) for Integral:
                            #.  How many sigma away from mean should we cut?

3.5                         #. Lower cut value (in units of RMS) for Peak:
                            #.  How many sigma away from mean should we cut?

The first eight words are calibration constants that are derived from ${^16}{\rm N}$ data: pedestals for both the Integral and the Peak measurements derived from the extrapolation of the Integral vs. NHIT and the Peak vs. NHIT curve down to NHIT=0; the mean of the Integral/NHIT distribution around which the cut is placed; a slope and offset parameter which give the RMS ($\sigma$) as a function of NHIT for the Integral measurement; and finally the mean and fit parameters for the Peak measurement. These are not intended to be changed unless the AMB is re-calibrated.

The next ten words are the parameters which influence the cut. The first two are the cuts in terms number of $\sigma$ away from the mean Integral/NHIT and Peak/NHIT. The next is just the minimum NHIT to consider in placing the cut (events with NHIT below this number will be passed). The next word, the `cut mode' determines whether the results of the Integral/NHIT and Peak/NHIT cut will be ANDed or ORed ($1=$OR). The fifth parameter effects the handling of high NHIT events, because of the railing of either the Integral or the Peak measurement. When enabled, the `special handling' of high NHIT events means that for cases where the NHIT is high enough that we expect either the Integral or the Peak to be railed if they are more than $N\sigma$ above the mean ($N$ is the number of $\sigma$ to use as the cut defined in the earlier words), then that measurement is disregarded if, indeed, the measurement is railed. As an example, take an NHIT=250 Flasher event. The constants for the RMS of the integral tell us that we expect $\sigma$ to be $\sigma = 0.064 + 7.8/250 = 0.090$. The mean of the Integral/NHIT distribution is 0.82, and so $3\sigma$ above the mean is an Integral/NHIT$=1.11$. At an NHIT of 250, this implies that for an event to fail the cut, it would have a pedestal-subtracted Integral of $250*1.11 = 276$. However, the rail of the Integral measurement is at 255, and so can never be as high as $276+65.0({\rm pedestal}) = 341$. In this case, then, we would ignore the Integral measurement, and cut only on the Peak. It is recommended that the `special handling' of the high NHIT events should always be enabled, unless one is testing out variations on the cuts.

Applying the AMB cut asymmetrically is also possible. The sixth parameter sets the "symmetric or asymmetric cut mode', which is defaulted to use the symmetric cut. Usually, the symmetric cut mode should be used unless one is specifically studying effects of different $\sigma$ cut values. The last four parameters set the cut sigma values for the asymmetric cut. For both the integral and peak, events greater than the upper $\sigma$ away from the mean Integral/NHIT and Peak/NHIT or less than the lower $\sigma$ away from the mean Integral/NHIT and Peak/NHIT are cut.

To use the cut, one can put a call flt($amb_cut) in the progammable event loop. The cut returns `true' if the event is a good event.