M. Kolander

Realization of a second level neural network trigger for the H1 experiment at HERA

Abstract Since 1996 the H1 experiment is fully equipped with two independent fast pattern recognition systems operating as second level triggers (L2). The decision time is 20 μs. One of the two is the neural network trigger. It runs an array of presently ten VME-boards with CNAPS 1064 chips (20 MHz, 128 Mcps) by Adaptive Solutions. The input trigger data from the detector components arrive in various formats on a 8 × 16 bit wide 10 MHz bus. Before usable as 8-bit input values to the CNAPS they are preprocessed by several bit-manipulating algorithms and arithmetic functions implemented on XILINX 4008 field programmable gate arrays (FPGA). The startup strategy for the new system is to concentrate on photoproduction channels or low multiplicity final states which so far could only be efficiently triggered with unacceptable high rates.

A fast high resolution track trigger for the H1 experiment

After 2001 the upgraded ep collider HERA will provide an about five times higher luminosity for the two experiments H1 and ZEUS. In order to cope with the expected higher event rates the H1 collaboration is building a track based trigger system, the Fast Track Trigger (FTT). It will be integrated in the first three levels (L1–L3) of the H1 trigger scheme to provide higher selectivity for events with charged particles. The FTT will allow to reconstruct 3-dimensional tracks in the central drift chamber down to 100 MeV/c within the L2 latency of ∼23 μs. To reach the necessary momentum resolution of ∼5% (at 1 GeV/c) sophisticated reconstruction algorithms have to be implemented using high density Field Programmable Gate Arrays (FPGA) and their embedded Content Addressable Memories (CAM). The final track parameter optimization will be done using non-iterative fits implemented in DSPs. While at the first trigger level rough track information will be provided, at L2 tracks with high resolution are available to form trigger decisions on topological and other track based criteria like multiplicities and momenta. At the third trigger level a farm of commercial processor boards will be used to compute physics quantities such as invariant masses. Keywords— Trigger, Fast Track Trigger, Track Trigger, FPGA, Content Addressable Memory, CAM, DSP, H1 Collaboration, HERA ColliderAfter 2001, the upgraded ep collider HERA will provide an about five times higher luminosity for the two experiments H1 and ZEUS. To cope with the expected higher event rates, the H1 collaboration is building a track-based trigger system, the Fast Track Trigger (FTT). It will be integrated in the first three levels (L1-L3) of the H1 trigger scheme to provide higher selectivity for events with charged particles. The FTT will allow reconstruction of three-dimensional tracks in the central drift chamber down to 100 MeV/c within the L2 latency of /spl sim/23 /spl mu/s. To reach the necessary momentum resolution of /spl sim/5% (at 1 GeV/c), sophisticated reconstruction algorithms have to be implemented using high-density field-programmable gate arrays and their embedded content addressable memories. The final track parameter optimization will be done using noniterative fits implemented in digital signal processors. While at the first trigger level rough track information will be provided, at L2 tracks with high resolution are available to form trigger decisions on topological and other track-based criteria like multiplicities and momenta. At the third trigger level, a farm of commercial processor boards will be used to compute physics quantities such as invariant masses.

Performance of the H1 fast track trigger - operation and commissioning results

The H1 experiment at the electron-proton collider HERA has built a new fast track trigger to increase the selectivity for exclusive final states, especially those with heavy quarks, and to cope with the higher background rates after the HERA luminosity upgrade. Hits measured in the central jet chamber of H1 are combined to track segments by performing 5 middot 1012 mask comparisons per second using content addressable memories (CAMs). These segments are collected and transmitted via 5 Gbit/s LVDS links to custom made multipurpose processing boards, where they get linked and reconstructed to three dimensional tracks within 20 mus. On the third level resonances are identified in 100 mus by a farm of PowerPC boards. Since 2005 the FTT level one has replaced the existing drift chamber trigger and is the major track trigger of the H1 experiment. In order to further increase the selectivity the second level has started to operate. First analyses of the data show that also the second level fulfills the design specifications

A Purity Monitoring System for the H1 Liquid Argon Calorimeter

The ionization probes used for monitoring the liquid argon purity in the H1 calorimeter are described and results of their operation in tests at CERN and during the period 1992 to the end of 1998 at HERA are given. The high sensitivity of the charge measurements leads to refined charge collection models, and to the observation of a variation of the ionization yield of our electron sources with temperature.The ionization probes used for monitoring the liquid argon purity in the H1 calorimeter are described and results of their operation in tests at CERN and during the period 1992 to the end of 1998 at HERA are given. The high sensitivity of the charge measurements leads to refined charge collection models, and to the observation of a variation of the ionization yield of our electron sources with temperature.

THE ARGUS VERTEX TRIGGER

Abstract A fast second level trigger has been developed for the ARGUS experiment which recognizes tracks originating from the interaction region. The processor compares the hits in the ARGUS micro vertex drift chamber to 245 760 masks stored in random access memories. The masks which are fully defined in three dimensions are able to reject tracks originating in the wall of the narrow beampipe of 10.5 mm radius.

Artificial Neural Networks as a Level-2 Trigger for the H1 Experiment: Status of the Hardware Implementation

Triggering at the HERA ep collider is challenging because of the high bunch crossing rate and an expected large background. In the H1 experiment, a trigger decision is made in four steps (level 1–4), stepwise decreasing the event rate and allowing for more sophisticated trigger decisions. The time available for L2 is about 20 μs. We have proposed to use an artificial neural network (ANN) for the L2 trigger based on the CNAPS-1064 chip available from Adaptive Solutions, (Oregon, USA). The intrinsic parallelism of the ANN algorithm together with the dedicated hardware offers fast processing of the trigger informations. The trigger system uses up to 10 decision units, each consisting of a Pattern Recognition Module (PRM) and a Data Distribution Board (DDB). A DDB receives the L2 data stream and generates the network inputs used by the algorithms on the PRM. A PRM is a commercial VME board carrying the CNAPS processors.

First Results from the Third Level of the H1 Fast Track Trigger

To make the best possible use of the higher luminosity provided by the upgraded HERA collider, the H1 collaboration has built the Fast Track Trigger (FTT). It is integrated in the first three levels (L1-L3) of the H1 trigger scheme and provides enhanced selectivity for events with charged particles. The FTT allows the reconstruction of tracks in the central drift chambers down to 100 MeV. Within the 2.3 mus latency of the first trigger level coarse two dimensional track information in the plane transverse to the beam is provided. At the second trigger level (20 mus latency), high resolution, three dimensional tracks are reconstructed. Trigger decisions are derived from track momenta, multiplicities and topologies. At the third trigger level a farm of commercial PowerPC boards allows a partial event reconstruction. Within the L3 latency of 100 mus exclusive final states (e.g. D*,J/psi) are identified using track based invariant mass calculations. In addition an on-line particle identification of electrons and muons with additional information from other subdetectors is performed. First results obtained from the third level, which is fully operational since 2006, are presented.