Content area
Abstract
Abstract
One of the main tracking detectors of the forthcoming ALICE Experiment at the LHC is a cylindrical Time Projection Chamber (TPC) with an expected data volume of about 75 MByte per event. This data volume, in combination with the presumed maximum bandwidth of 1.2 GByte/s to the mass storage system, would limit the maximum event rate to 20 Hz. In order to achieve higher event rates, online data processing has to be applied. This implies either the detection and read-out of only those events which contain interesting physical signatures or an efficient compression of the data by modeling techniques. In order to cope with the anticipated data rate, massive parallel computing power is required. It will be provided in form of a clustered farm of SMP-nodes, based on off-the-shelf PCs, which are connected with a high bandwidth low overhead network. This High-Level Trigger (HLT) will be able to process a data rate of 25 GByte/s online. The front-end electronics of the individual sub-detectors is connected to the HLT via an optical link and a custom PCI card which is mounted in the clustered PCs. The PCI card is equipped with an FPGA necessary for the implementation of the PCI-bus protocol. Therefore, this FPGA can also be used to assist the host processor with first-level processing. The first-level processing done on the FPGA includes conventional cluster-finding for low multiplicity events and local track finding based on the Hough Transformation of the raw data for high multiplicity events.
PACS: 07.05.-t Computers in experimental physics – 07.05.Hd Data acquisition: hardware and software – 29.85.+c Computer data analysis
Details
1 Kirchhoff Institute for Physics, University of Heidelberg, Germany
2 Department of Physics, University of Bergen, Norway
3 Bergen College, Norway
4 Institute of Nuclear Physics, University of Frankfurt, Germany
5 Department of Physics, University of Oslo, Norway