CERN Accelerating science

ATLAS Slides
Report number	ATL-DAQ-SLIDE-2015-460
Title	FTK AMchip05: an Associative Memory Chip Prototype for Track Reconstruction at Hadron Collider Experiments
Author(s)	Crescioli, Francesco (Laboratoire de Physique Nucleaire et de Hautes Energies (LPNHE), UPMC, Universite Paris-Diderot, CNRS/IN2P3) ; Annovi, Alberto (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Beretta, Matteo (INFN Laboratori Nazionali di Frascati) ; Beccherle, Roberto (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Giannetti, Paola (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Citraro, Saverio (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Liberali, Valentino (INFN Milano and Universita' di Milano, Dipartimento di Fisica) ; Luciano, Pierluigi (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Rossi, Enrico (INFN Pisa and Universita' di Pisa, Dipartimento di Fisica) ; Stabile, Alberto (INFN Milano and Universita' di Milano, Dipartimento di Fisica) ; Mirzaei, Mohammad Ali (Centre National de la Recherche Scientifique (FR)) ; Piadyk, Yurii (Taras Shevchenko National University of Kyiv)
Corporate author(s)	The ATLAS collaboration
Submitted to	European Physical Society Conference on High Energy Physics 2015, Vienna, Austria, 22 - 29 Jul 2015
Submitted by	[email protected] on 19 Aug 2015
Subject category	Particle Physics - Experiment
Accelerator/Facility, Experiment	CERN LHC ; ATLAS
Free keywords	FTK ; TDAQ ; Level2 ; Associative Memory
Abstract	The ATLAS FTK track trigger project is an upcoming upgrade for the ATLAS trigger system currently under installation. A first reduced-coverage FTK is expected to take data by the end of 2015, while full detector coverage will be reached in 2016 and more processing power will be added in 2017-2018. The ATLAS FTK is a dedicated supercomputing processor based on FPGAs and a custom ASIC: the Associative Memory chip (AMchip). The AMchip is the core processor in charge of the real-time pattern recognition stage of the FTK algorithm. It is based on Content Addressable Memory elements connected by advanced computation logic that adds the unique feature to look for correlated data patterns (e.g. tracks made of hits) within a sparse data set (e.g. hits of an event) processing data at the full I/O speed. The AMchip05 is the latest AMchip prototype before the final FTK production. It is functionally identical to the upcoming production chip with the only difference of pattern capacity. We will show the AMchip05 architecture, the design implementation and the in depth performance analysis under different working condition (power supply voltage and operating frequency). We will discuss the impact on physics outreach enabled by the new features added with respect to the previous AMchip generation, in particular the possibility to have variable resolution patterns. We will also address the impact on LHC Phase-2 tracking applications of the current chip and the foreseen developments.