The TSP processor is designed to shift the phototube times in
events in the same way that the TAC droop does. This provides a way
of testing the systematic effects of this droop on (for example)
reconstruction algorithms. TSP can be applied either to Monte Carlo tube
times or to data tube times (if one wants, for example, to make high rate
data look like low rate data, since it only modifies the
times held in the PMT bank. At this time, no charge droop is simulated.
The TSP time shifts are based upon a `time-since-last-hit' (TSLH), in the same way that the droop in the detector is correlated with this. However, both for Monte Carlo data, the `time-since-last-hit' is not well-defined because we only simulate one signal type at a time. TSP thus takes as user input (through the titles bank TTSP in tsp_com.dat) both a total event rate and a total average NHIT, from which it calculates the channel hit rate (ignoring different rates on different channels for now). This channel hit rate is used to produce an exponential distribution of TSLH's for each channel in an event, ignoring what the Monte Carlo TSLH actually is.
With this `fake' TSLH, TSP then calculates what the expected droop would be for a given channel (making the look channel earlier than it really is) and applies it to the time in the PMT bank. The channel-by-channel shifts are obtained through fits to pedestal data taken with different delay times between the events.