IceTop Gain Precision with DOM-Cal

August 2006

The following analysis examines the measurement precision in DOM-Cal's gain vs. HV calibration using three back-to-back DOM-Cal runs taken with IceTop on August 21, 2006.

Summary

• The precision at gains of 5E6 and 1E7 is quite good (RMS of ~2%, with no outliers beyond about 5%).
  
• The precision gets significantly worse for low-gain settings (5E4 for low-gain IceTop DOMs), as the fit is extrapolated. The RMS of the gain distribution is approximately 8%, with some outliers. This suggests that much of the "gain variation" that Serap sees in her analysis is DOM-Cal measurement error, not physical gain shift.
  
• The low-gain spread is reasonably distributed around the combined fit from the three runs, suggesting the measurement precision can be improved in a 1/sqrt(N) fashion by simply iterating the analysis a few times. This was done offline in this test but could be implemented in DOM-Cal as a run-time option for IceTop.
  
• As a byproduct of the combined-run analysis, updated HV and discriminator settings for the DOMs have been calculated.

The following plots show the fractional gain shift for all IceTop DOMs for three back-to-back DOM-Cal measurements. The gain shift for each DOM is calculated for each of the three runs, relative to the combined fit using the data from all three runs together.

The distribution of the gains at 5E4 is not quite Gaussian -- it's a bit more peaked and with longer tails. But roughly, we should be able to improve the RMS to under 5% if we iterate the HV/gain analysis, say, four times. This will increase the full DOM-Cal run time by 50% or so (say, an extra half-hour).