Database monitoring

The quickest way to asses the quality of calibrations is to examine quanitities sensitive to calibration parameters as a function of run number. Quantities found useful for assessing the quality of calibration for the e1 and g1 running periods are enumerated below.

single particle yields:

Examining the normalized yeild of various pariticle types as a function of run number can quickly identify problems with the timing of the apparatus. Shown in Fig. 4 is the single particle yield for g1a pass 0 and pass 1. Several features changed between pass 0 and pass 1, for example, $\pi^+$ are more common in pass 1 compared to pass 0 due to improvments in the time-to-distance function calibration. In addition, particle yields are more uniform over the running period. Most notably, the precipitous drop in proton yield seen from run 11800 to 11950 in pass0 is gone in pass 1, and the spike in the yields for kaons and unknowns near run 11750 seen in pass 0 is gone in pass 1. These changes reflect improvements in the time-of-flight and tagger calibration. The remaining variations in particle yield seen for the pass 1 data are due primarily to changes in running condition (30% field runs near 11750, and no-field runs near 11825) and normalization. The single particle database entries are derived from the pid_mon program, a list of quantities is available in Table 5. In addition to the standard particle yields, pid_mon also calculates a number corresponding to the average quality of the tracks for each particle type, this number is proportional to the Chi-squared per degree of freedom determined from time based tracking. When looked at versus run number, a change in the quality factor can indicate a problem in the tracking calibrations. For example, Fig. 5 shows that the quality of the electron and proton tracks for runs 16152 and 16154 are much higher than the rest of the run period. Since this effect appears for both the electron and proton particle types, the logical source of the error points to the tracking calibrations. In this case, scan_map revealed that the first calibration in the DC_DOCA.map for this run period had a time stamp associated with run 16155. Thus, runs 16152 and 16154 were using tracking calibrations based on a previous run period. This problem was easily fixed by moving the calibration from run 16155 to run 16152.

 

Figure 4: Particles identified by type per million triggers for the g1 period, ``pass 0'' results compared to ``pass 1.''  

 

Figure 5: Quality of particle tracks from pid_mon for step 3 of the E1B running period using prod-1-9. 

ratio of time-based to hit-based tracks, hits per time based track:

Sometimes tracking calibrations can be so poor that tracks are simply not reconstructed. When this occurs one will typically notice that the number of hits per time based track will be low in some sectors and the yields for number of time based tracks per sector will vary from sector to sector because of differing levels of calibration in various regions of the detector. Improvements in the overall level of tracking calibration can be seen in Figure 4 in the increase in yield for $\pi^+$ events from pass 0 (approximately 225K events per million triggers) to pass 1 (325K events per million triggers).

residuals:

For a a more detailed understanding of the resolution provided by the tracking it is neccessary to observe the residuals as a function of run number. Shown in Figure 6 is the RMS of the residuals for various superlayers in sector 1, for an early iteration of the calibration and the final production pass. Note that for pass 0 the residuals for superlayer 1 are the largest (solid circles) this was noticed to be unusual because it was expected that superlayer 1 would be the smallest due to its cell size. Upon closer inspection an anomaly was found in the trk_mon histograms leading to a discovery of a bug in the time based tracking software. You will note that for pass 1 of the data, superlayer 1 has the smallest residuals as expected, in addition there is now seen a clear gradation in resolution as a function of superlayer (superlayer 1 is smaller than superlayer 2 etc...).

 

Figure 6: Top Figure: residuals in cm by superlayer for g1a pass 0, early tracking calibration before bug fix, Bottom Figure: residuals in cm by superlayer g1a pass 1, after bug fix and refined calibration procedure  

dead wires:

Observing the number of dead wires as a function of run-number is most useful for final run selection for data analysis.

yields for exclusive processes:

Examining these quantitites provides similiar information to a single particle plot but can additionally provide useful information regarding the quality of the normalization of various runs.

Most of these quantities will vary with running period and are sensitive to changes in trigger and magnetic field, however, within a running period, these quanitites should be uniform.