With the question of a reference time for Radphi resolved, the problem remains of how to efficiently remove the charged component from the sample. Noting the tight timing correlation between the cpv and the tagger, it is suggested that the cpv might be used to remove tagger hits associated with charged particles before they are considered for coincidence with the recoil detectors. Used offline in a similar way in which the upv was used online, a hit in the cpv simply gates off the tagger for a short period. Only tagger hits that do not appear near cpv hits can contribute toward the accidental-subtracted yields described above. In this way the coincidence with the tagger can be used to perform a double function, both to remove beam photons outside the tagging energy range and to remove all interactions with forward-going charged particles. In such an analysis one gives up the ability to decide on an event-by-event basis which are neutral/tagged and which are not and in return obtains an efficient and unbiased method for estimating the spectra that would be obtained if one were able to separate them.
The statistical errors are minimized when one choses as narrow a coincidence window as possible, which means in turn that a recoil time as precise as possible should be obtained for comparison with the tagger. Events with a single pixel in the bsd were selected and the average of the times for the three counters associated to the pixel was taken. When compared with the tagger, the coincidence peak shifted slightly later with increasing pixel index, which corresponds to the position of the pixel moving from upstream to downstream. For pixel rings 0-6 the slope was measured to be 0.5 ns/pixel. The dependence is linear within measurement errors. This dependence is taken out by forming a software ``recoil time'' from which the above z-dependent offset is subtracted from the pixel time average. The coincidence peak for pixel ring 7 was much weaker and did not reflect the trend of the others. It is left uncorrected.
The spectrum of the recoil time is shown in Fig. 3. The open histogram is obtained in reference to all tagger hits, while the shaded histogram refers to tagger hits with a cp veto. A 6 ns window around the cpv-tagger coincidence peak was used in forming the cp veto. The reduction factor of about 0.6 in the accidentals level between the two histograms in the figure reveals the cost of the veto. The coincidence peak is reduced by a much greater factor than 0.6 which confirms the prior assumption that most of the triggers in the Radphi data sample come from cp events. This statement will be quantified in the following sections.
 |
The neutral-tagged histogram in Fig. 3 exhibits several features that are important to the understanding of how the gated tagger coincidence analysis works. The sawtooth structure seen across the spectrum is a consequence of the 2 ns pulse period of the accelerator and is expected in the difference spectrum between any two observables whose spread from time-of-flight is of order 1 ns or less. The prominence of the spikes in Fig. 3 indicate that resolution of the recoil time is of order 1 ns, which is at the level predicted on the basis of Monte Carlo for diffractive photoproduction from a nuclear target [1].
The tapering off of the accidental spectrum outside the interval [-40,40] ns is a consequence of the selection of tagging hits only within the window [-50,50] ns and the fact that the trigger defines the recoil time to lie within a 20 ns region around t=0. The region [-30,40] ns in Fig. 3 can be considered as a free-running correlation between the recoil and tagging counters with a given that something had to happen in a 20 ns coincidence window near t=0 in order for the event to pass the level-1 trigger. This would be expected to give rise to a small enhancement in the continuum over a 20 ns region containing the coincidence peak. This enhancement that comes from the level-1 trigger requirement is seen in Fig. 3 between the limits of approximately -5 and +15 ns. Unfortunately the boundaries of this region are very diffuse, evidence of the significant differences between the timing of individual bsd and tagging signals that formed the trigger. That the level of accidentals is roughly the same inside and outside the hardware trigger window is a consequence of running the tagger in at high rates.
This 20% enhancement is also seen in the shaded histogram in Fig. 3 but is partially obscured by the depression in the accidentals that appears around the coincidence peak. This depression occurs because, when the cp veto is applied to the tagger hits, it has a side effect of also suppressing a fraction of the recoils, even though whatever caused the recoil and cpv hits might be in random association with the tagger. This depression is a physical effect because it appears only when the cp veto is applied offline. The width of the depression corresponds to the 6 ns width of the cp veto cut.
 |
Correct subtraction of accidentals in Fig. 3 requires the selection of non-overlapping windows for coincidences and accidentals, both within the range of the 20 ns hardware coincidence. As a compromise, the choice was made of the region [-1,5] ns for coincidences and [6,11] ns for accidentals. A small renormalization of the accidental weights can be used to account for the partial clipping of events in the accidentals window by the level-1 trigger. These two windows are shown in Fig. 4. The dashed histogram shows a similar depression in the accidentals window to the one discussed above that appears in the coincidences window when the cp veto is applied. Accidentals or coincidences, only tagger hits without nearby cpv hits are counted when forming a time reference for registering the recoil time. The two depressions are correctly canceled out in the subtraction.
ns window around
the event trigger, while the shaded histogram refers only to those
tagger hits that have no cpv hit within the range [-1,5] ns in
Fig.