Selecting which runs to skim and calibrate

Before step 2, described in section 5, it is necessary to make a selection of runs which will be used for this calibration pass. The E1B run period will be presented as an example. The E1B run period occupied about 5 weeks of actual beam time, during which roughly 500 production runs were taken. It is neither practical nor necessary to calibrate each run individually. Instead, a subset of the production runs is chosen for calibration.

The process of choosing which runs to calibrate began with a review of the progress files for the E1B run period. All Hall B shift workers use a progress file to record the runs taken during their shifts, and those files are stored in the clasrun directory on the clon cluster. Most run periods have a specified number of triggers which should be accumulated for each run. The E1B run period used a target of ten million triggers per run, which normally corresponded to about 2 hours of data taking. However, some of the production runs taken during E1B contain much less than ten million triggers. There are many reasons runs were terminated at less than ten million triggers, and those reasons will not be discussed here. If a run does contain at least ten million triggers, it suggests the run conditions were smooth and uneventful, which are ideal conditions for deriving useful calibration constants. In addition, the runs that contain the most triggers are the runs that will dominate the scientific papers published by Hall B, so it is important for these runs to be well calibrated. The progress file were therefore perused to determine which runs possess at least ten million triggers.

The next criterium used to select which runs to calibrate is the elapsed time between the calibrations. The first time the E1B runs were calibrated, the elapsed time between the runs chosen for calibration was typically between eight and twelve hours, so either two or three runs per day of data taking were selected. The results of those calibrations determine if it is necessary to recalibrate in smaller time intervals. The Hall B online web page at http://db6.jlab.org/clas/html/clas-online/runinfo.html is consulted to determine the elasped times between the runs.

Finally, the operator logbook is reviewed to see if there were any anomalous conditions during any of the runs. There were hardly any problems with the detector hardware during the E1B run period. Unstable beam conditions were probably the most common potential problem for some of the runs. As much as possible, only runs which had stable beam conditions were selected for calibration.

It may sound as if a very small fraction of the runs are being calibrated, but actually the runs selected for calibration contain about 25 percent of all the data accumulated during the E1B run period, and the vast majority of the remaining 75 percent was taken within eight hours of one of the calibrated runs. Keep in mind that ten million triggers corresponds to roughly two hours of data taking, and three runs each containing at least ten million triggers corresponds to at least six hours of data taking. During the E1B run period, the average amount of data taken per day was probably about 14 hours.