Veto, Trigger and Data Acquisition

Veto, Trigger and DAQ


The main task of the trigger is to select all neutrino induced events in the target and to reject background from cosmic rays, muons from the beam, and neutrino interactions outside the target. In addition, trigger modes for calibration and alignment of the detector are needed. The expected neutrino event rate is 0.75 events per spill of 10^13 protons on target for an effective mass of 2400 kg.

The neutrino interaction trigger is using information from two hodoscope modules, one close to the target region, T1, and one just in front of the calorimeter, T2 (see figure 2). T1 is made of two layers of 15 thin scintillator strips of 10 cm width and 160 cm length, oriented horizontally. The two layers are staggered and cover an area of 160*160 cm2. The thickness is minimised to avoid loss of tracking precision due to multiple scattering. For the same reason the planes are mounted as close as possible to the last fibre tracker of the target region. Trigger plane T2 is made of two layers of 20 strips of 10 cm width and 200 cm length, oriented horizontally.

A veto system (V) is positioned 2 m in front of the target region and consists of two planes of twenty 20 cm wide scintillators oriented vertically, and staggered by one-half width. They form a plane of 4*3.2 m2 surface area. An additional plane of 32 scintillators vetoes events from the concrete floor.

The neutrino interaction trigger is defined by a combination of pulses from the hodoscope strips T1 and T2 consistent with a track with tan(theta) < 0.25 with respect to the neutrino beam. A veto is formed by any combination of a veto counter hit and a hit in T1, with precise timing to avoid vetoes due to back scattering. All scintillator strips of T1 and V are viewed at both ends by photo multipliers to fulfil the timing requirements.

The size of the veto counter has been chosen to fully cover the area defined by the angular acceptance of the hodoscope trigger.

The muon flux incident on the detector located in the BEBC hall has been measured during neutrino beam operation and was found to be N_mu ~ 500/m2 for one cycle of 2*10^13 protons. At this point of the hall the shielding is weak and has been reinforced by adding iron in front of the detector (Section IV.3). This is reducing the muon flux by a factor ~ 10. A space of about 10 m is kept between the shielding and the detector to minimise background from neutrino induced particles. The number of incident muons must be kept as low as possible to minimise background in the emulsion stacks.

After removal of beam-related incoming muon tracks, the main background is created by neutrino interactions in the surrounding material such as supports and the floor. These events can satisfy the trigger conditions if several tracks hit the hodoscopes, thereby faking one small angle track. This background is kept low by minimising heavy support structures close to the fiducial volume. Cosmic ray background is expected to be small.

Additional signals from the calorimeter and the muon spectrometer will be used to build independent pattern triggers for evaluation of the trigger efficiency, and to record neutrino interactions for beam flux monitoring.

A 4x4 m2 streamer tube wall has been put directly upstream of the veto plane (figure 17) to define, together with a range requirement in the muon spectrometer, high momentum muon tracks for alignment and calibration.

Data Acquisition

The event rate in this experiment is low, on average less than one event per neutrino burst. Each event contains a large amount of raw data (~ 20 Mbyte fibre pixels, and ~ 10 kbytes of calorimeter, spectrometer and streamer tube data). This has to be reduced to a few hundred kbytes in a very early stage before the data is combined into one single event, to avoid the need for large bandwidth data buses. This requirement calls for simple pattern processing at the level of the conversion. A second level software trigger is not needed.

The requirement to read-out up to three or four events per burst sets the upper limit of a few hundred millisecond read-out time per event. The data volume produced is doubled by calibration and alignment triggers taken between the neutrino bursts. It is expected that data tapes will be written with a frequency of one every few hours.

The data acquisition is based on a VME-bus hardware with CPU's running the OS-9 operating system (figure 14). The heart of the system is formed by a root crate containing the event builder CPU, and a dual 3480 cartridge unit connected via a SCSI bus. Its task is to assemble events from the converted data in the other crates, data recording on cartridges and to send events to the UNIX cluster for sample data analysis for quality monitoring. The event data flow will be channelled through a VIC vertical bus. The video signals from the fibre trackers are digitised in the cameras and processed with dedicated CPU's. Each CPU serves two cameras. Interfaces to CAMAC for ADC and TDC modules and trigger logic, etc., are foreseen. The system is organised hierarchically with at least one VME crate and processor serving each detector sub-system.

All ADC and TDC channels are equipped with multiple event buffers and have fast conversion times less than 20 usec. Events are stored in these buffers during the neutrino burst, and are read-out between and after them.

The data quality will be monitored with UNIX work stations connected through ethernet lines to all VME crates. They can perform sample data analysis, on-line calibration and the display of histograms prepared on the OS-9 system. Permanently updated information will be available on screens connected to the workstations.

Dead time

Trigger dead time is introduced by activity in the veto counter, by the strobe signal, by the duration of the conversion of valid events, and by the overflow of read-out buffers. The latter source can be removed by choosing sufficient buffer depth, or by overlaying events in the limited buffers of the CCD's. The expected dead time contributions are summarised in Table 3 (the effective spill length is 4 ms).
                          Table 3
       Contribution to event losses by dead time
Source	                     Number        Length   Dead time
                             per burst              (fraction)
veto                           400          60 ns    6  x10-3  
pretrigger   V*T1*T2            10         600 ns    1.5x10-3
strobes from other detectors   100          60 ns    1.5x10-3
MCP gate + delays                0.75       30 us    5.6x10-3
CCD conversion                   0.75      120 us   22.5x10-3
ADC/TDC conversion other events  5          20 us   25  x10-3
TOTAL			                              62x10-3
We expect a total loss of events due to dead time of 6%. In addition, if we limit the maximum number of events to two, the dead time is increased by another 6.5%. This can be avoided by overlaying the second and the third event in one CCD frame.

Because of the high redundancy of the trigger system, and of the simplicity of the requirements, a trigger efficiency of 98% can be estimated for events passing the off-line selection criteria. The total detector efficiency is eps_det = 92%.

Back to CHORUS home page.
JV / 17 Jan 1994