With 3mm cells, and 4.8% voxel occupancy, and a pad-distribution function with width(sigma)=70% of a cell width, 29% of hits are "touched" by noise.
Shown on the left is the FADC responce of one cell for this noise level. On the right is the event diplay, showing only the noise hits that overlap in time.
A pathology in the time pattern recognition that affected even zero noise was discovered during code development prior to the 2007 meeting.
The pathology caused spurious hits.
This was resolved by requiring that new time hits can not be defined within a smoothing width (a tuning parameter).
Double hits, obseved in the upper analysis of an FADC response, are eliminated in the lower analysis of the same response.