Real time data processing of the TileCal calorimeter of the ATLAS detector

The TileCal detector is an iron-scintillating fiber sampling calorimeter of the ATLAS experiment, one of the experiments at the Large Hadron Collider (LHC) accelerator at CERN (European Laboratory for Particle Physics), scheduled to start in 2007. At a Level 1 trigger rate of 100 kHz, the read out system of the TileCal has to be able to process and format the data of 10,000 read-out channels (112.91 Gbps of input bandwidth) in 2 mus to avoid dead time. Furthermore, to reduce the amount of data and to meet the maximum output bandwidth of the Level 2 Trigger of 53.7 Gbps, real time computation of the deposited energy applying optimal filtering techniques and collision time computation must be performed for each channel. This is accomplished with a read out driver system of which 32 9U VME boards are the main components. Each board consists of 8 G-links receivers and 4 S-Link transmitters, and 8 digital signal processors (Texas Instruments TMS320C6414@600 MHz DSP), totalizing 68992 MIPS of processing power. These boards are presently at the production stage. We report on the overall design of such system, using the DSP for energy and collision time computation, data formatting, and identification of low transverse momentum muons in real time with high efficiency