Livetime

In Hall A we typically refer to the "livetime" as the fraction of the time the experiment is sensitive to events of interest, and "deadtime" as then (1-livetime).




When an event of interest occurs and leaves a sufficiently large signal in our detector, it still might not be readout due to the front-end electronics or the DAQ system being "dead" at that time. Here I discuss two types of deadtime in our experiment:

• Electronic (or front-end) deadtime
• DAQ deadtime

They are important for the asymmetry measurement since they are a function of the trigger rate and could, in principle, be different for the two helicity states.



The electronic or "front-end" deadtime deals with the effective masking of signals/triggers within the front-end electronics (PMTs, discriminators, etc.) before reaching the ADC/TDC/Trigger Supervisor. If a PMT fires such that the discriminator puts out a pulse 50ns long, and another signal comes within that 50ns it will be "hidden".

Let us take the "true" input trigger rate as m, and number of triggers in a period T is k, and the time the electronics are dead for each event is d. Then the relation between the real trigger count and the observed trigger count is (for the non-updating case):

m T = k + m (k d)

where m (k d) accounts for the number of triggers missed due to the electronics being "dead". Then the fractional "deadtime" D is:

D = ( m (k d) ) / (m T) =  (k/T) d .

Since (k/T) is the observed rate, we see that, for relatively small deadtime-fractions and in the non-updating case, the fractional deadtime is linear in the observed rate.

For GEn, the electronic dead-time was measured by using a pulser to send an analog signal to the BigHand? and BigBite trigger electronics to generate an artificial T3 trigger; these were the T7 events. A separate copy of the T7 signal was sent to the trigger supervisor and the 1877 TDC which recorded the relative timing of all input trigger signals. By looking at the time difference between the T7 and T3 signals in the TDC, one can look for the corresponding artificial T3 that should be present with each T7 and very tightly correlated in time, and therefore measure the fraction of the time the T3 is missing. This fraction is the "front-end" deadtime, D.

This is the simpler deadtime to measure and understand, especially since we ran in an unbuffered mode. When the Trigger Supervisor (TS) receives a trigger that passes the prescale-factor cuts, it will instruct the other ROCs to digitize the present event and send the data onto the CODA event builder. During this digitization process, the TS and ROCs send and hold a BUSY signal such that the TS does not accept another trigger until all the ROCs are ready again. This period of being BUSY was ~300-500 micro-seconds per event, during which the DAQ and TS were "dead" to new triggers — hence the designation as the DAQ deadtime.

The DAQ deadtime is rather straight-forward to measure, since scalers are used to count each trigger for each helicity state irrespective of the TS state. DAQ livetime for T3 (our production trigger) is then just

LT_DAQ_T3 = prescale_T3 * N_T3_datafile / N_T3_scaler

where prescale_T3 is the prescale factor for the trigger, N_T3_datafile is the number of events with that specific trigger in the CODA datafile (determined by using D.bit? or D.evtypebits), and N_T3_scaler is the number of times that specific trigger was sent to the TS as readout from the scalers.

How this is accomplished electronically is that an exact copy of the inputs sent to the trigger supervisor is also sent to a Lecroy 1877 TDC with a COMMON-STOP from the BB trigger. For the standard GEn analysis, these TDC channels are read into and stored as the D.bit1 through D.bit9 variables in the ROOT-file.


W.R. Leo, Techniques for Nuclear and Particle Physics Experiments: A How-to Approach, Second Edition. (Springer-Verlag, Berlin, Heidelberg 1994)