Detector exclusion
A dead detector contributes nothing but dilutes a group’s asymmetry normalisation; a hot one (HV breakdown, light leak, noisy PMT) injects counts uncorrelated with the muons and drags the asymmetry toward zero with a spurious time dependence. Either way the cure is the same: drop the detector from the analysis without touching the data.
The grouping dialog’s Exclude Detectors field takes a WiMDA-style list —
1,5,10-15 — of 1-based detector ids; the detector layout editor offers
the same thing visually via its Exclude mode toggle (click a detector in
the schematic to strike it out). Excluded detectors are removed from every
group sum at reduction time: the raw histograms are untouched, no reload is
needed, the exclusion is recorded with the grouping in the project, and it
applies consistently to the asymmetry, alpha estimation, per-group fitting
and Fourier inputs (which all consume the same group sums).
Identifying a bad detector
The per-detector counts tell the story: a dead detector shows totals far below its neighbours (often zero); a hot one shows an excess, and its spectrum on a log scale deviates from the straight muon-lifetime line — flat noise at all times rather than the exponential. The detector spread reported by Find t0 is another quick flag: a detector whose prompt peak sits far from the consensus has a timing problem even if its rate looks normal.
When to use this. Exclude rather than correct whenever a detector’s problem is not a well-modelled inefficiency: deadtime correction can fix rate-dependent counting loss, but nothing recovers a detector that lost HV halfway through a run or counts light-leak noise. The statistical cost of dropping one detector from a 32-or-more-detector group is a percent-level increase in error bars; the systematic cost of keeping a bad one can be a distorted asymmetry. After excluding, re-estimate alpha — the group balance changes with the group membership.
WiMDA zeroes excluded detectors when the file is read (a reload per change); applying the exclusion at grouping time gives the same reduction with the raw data intact (study record: divergence D10).
References
Pratt, Physica B 289–290, 710 (2000).