The VERITAS Trigger System
A. Weinstein1 for the VERITAS Collaboration2
CFD x 499
(1 per pixel)
Pattern Trigger
Shower Delay
( 1 PDM channel
per trigger signal)
Array Trigger Coincidence Logic(SAT Board)
FIFO Buffer
Compensating Delay
( 1 PDM channel
per trigger signal)
Event
Decision
Event
Information
Harvester Process
Serialized Event Information
L3 Trigger
FADC Modules
(1 channel per pixel)
(x4 : 1 per telescope)
Readout
instructions
Assembled telescope data
L3 Trigger signal
64μs circular buffer
Telescope Data
Acquisition System
L1
L2
L3
BUSY level
Event Logic
Inhibitor
Abstract: The VERITAS gamma-ray observatory, situated in southern Arizona, is an array of four 12m-diameter imaging Cherenkov telescopes, each with a499-pixel photomultiplier-tube camera. The instrument is designed to detect astrophysical gamma rays at energies above 100 GeV. At the low end of theVERITAS energy range, fluctuations in the night-sky background light and single muons from cosmic-ray showers constitute significant backgrounds. VERITASemploys a three-level trigger system to reduce the rate of these background events.
The Level Two (L2, Pattern)Trigger system acts on the relativetiming and distribution of L1 triggerswithin a given telescope camera,preferentially selecting compactCherenkov light images and reducingthe rate of triggers due to NSBfluctuations. It divides the telescopecamera into overlapping patches of19 pixels, and fires if a programmednumber of L1 triggers within a patchoverlap. The standard pixelcoincidence requirement is threeadjacent pixels within a patch; therequired overlap time betweenadjacent CFD signals is ~6 ns. EachL2 system (four in all) produces asingle logical signal, the L2 trigger,which is sent to the array triggersystem (L3).
Figure 1: A block diagram of the VERITAS Trigger System operation, including its interface with thedata acquisition systems.
The Level One (L1) Trigger systemacts at the single pixel level. Itconsists of custom-built ConstantFraction Discriminators (CFDs), onefor each photomultplier-tube (PMT)pixel in a telescope camera. The CFDtriggers (produces an output pulse) ifthe sum of the voltages from theoriginal PMT pulse and a time-delayed copy crosses a threshold.The VERITAS CFDs are equipped witha rate feed-back (RFB) loop, whichautomatically increases the effectivethreshold when the noise level (andthus CFD trigger rate) rises.
The full VERITAS array of four telescopes was completed
in Spring 2007. Most of the preliminary studiescharacterizing trigger performance, however, weredone during the commissioning process, with a three-telescope subset of the array (T1-T3).
Figure 2 illustrates the effectiveness of the array triggerin suppressing events due to night-sky background,well below the operating threshold where a singletelescope trigger would be dominated by NSB. The 50mV operating threshold for the L1 trigger (whichcorresponds to 4-5 photoelectrons) was conservativelychosen to ensure both stable operation andreasonable dead time performance for a wide range ofweather conditions and array configurations.
Figure 3 shows the array trigger rate as a function ofcoincidence window width. The rate is stable forwindow sizes between 25-100ns, which is consistentwith the observed spread in L2 trigger arrival times aftershower delays have been applied.
The dead time of the array is determined, to first order,by the array trigger rate and the average telescopereadout time (~400μs). The observed correlation ofarray trigger rate and fractional array dead time isshown in Figure 4. The array dead time ranges from~6-7% for array trigger rates around 150Hz and 10-11%in the vicinity of 230Hz.
The Level Three (L3, Array)Trigger system requiressimultaneous observation of anair-shower event in multipletelescopes. This requirementsignificantly reduces the rate ofbackground events, particularlythose due to single muons.
L3 uses the relative timing of theL2 trigger signals to identify ashower event. First, the systemuses programmable “shower”delays to compensate for thedifferences in the arrival times ofthe Cherenkov light front at thedifferent telescopes, as well as thedifferences in L2 signalpropagation times. The corecoincidence logic continuallymonitors the delayed L2 signalsfor a pattern that lies within acoincidence window. The widthof the coincidence windowcompensates for the residualvariation in L2 signal arrival timesdue to the width and curvature ofthe Cherenkov wavefront,variation in L2 response withrespect to image size, and timingjitter in the various electronicscomponents. Both the allowedpatterns of L2 signals (typicallyimplemented as a simplemultiplicity requirement) and thecoincidence window width areprogrammable.
Trigger System Performance:
Figure 5: L2 and L3 rates for a typical three-telescope run (right)and a run taken on the same night, under partial moonlight(left). It is clear the L3 rate is stable with respect to significantfluctuations in the pattern trigger rates. This, and the fact that thearray dead-time is not influenced by L2 trigger rates, allows forstable running under a variety of conditions, including partialmoonlight.
Figure 2: The dependence of L2 and L3 triggerrates on L1 (CFD) threshold and array triggermultiplicity, for a three-telescope array with a 50nscoincidence window. This data was taken whilepointing at a dark patch near zenith, undermoderate weather conditions.
Figure 3: Array trigger rate (for a three-telescopearray with 2/3 multiplicity requirement) as a func-tion of coincidence window width.
Figure 4: Fractional array dead time vs. array triggerrate for a representative sample of good runs.
Custom-built 500MS/s flash-ADC (FADC) modules (one FADC channel perpixel) continuously digitize the PMT pulses with a memory buffer depth of 64μs.The telescope data acquisition systems read out a portion of this buffer (24samples) for every channel as directed by the array trigger system.Compensating delays applied to the L3 trigger signals ensure that every L3trigger signal is received at the telescopes at a fixed time relative to the startof the PMT pulse, allowing the data acquisition system to “look back” to theappropriate starting point in the buffer before reading out. Each dataacquisition system also raises an ECL level (BUSY) while occupied with bufferreadout; the L3 system inhibits the coincidence logic as long as one of theselevels is raised.
(x4 : 1 per telescope)
2. See G. Maier, “VERITAS: Status and Latest Results”, for a complete listing of the VERITAS collaborationand ICRC contributions.
Single-tel. operating threshold
(6-7 photoelectrons)
(Standard operation)
This research is supported by grants from the U.S. Department of Energy, theU.S. National Science Foundation, and the Smithsonian Institution, by NSERCin Canada, by PPARC in the UK, and by Science Foundation Ireland.