Bonn 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 86: DAQ, trigger and electronics IV

T 86.8: Vortrag

Donnerstag, 2. April 2020, 18:15–18:30, L-3.015

Development of Digital Signal Processing for the ATLAS Liquid-Argon Calorimeters with Artificial Neural Networks using Field Programmable Gate Arrays — •Nick Fritzsche, Anne-Sophie Berthold, Rainer Hentges, Philipp Horn, and Arno Straessner — Institut für Kern- und Teilchenphysik, Dresden, Germany

The upgrade plans for the Large Hadron Collider result in more challenging requirements for the data readout of the Liquid-Argon calorimeters of the ATLAS detector. The energy deposits of particles that are formed in high-energy proton-proton collisions create electrical pulses in the detector. These undergo a digital signal processing for selecting and preparing the signals in real-time for data acquisition and trigger. In signal processing artificial neural network algorithms can be used for fast, precise and resource-saving trigger decisions and energy reconstruction. A general implementation for feed-forward networks in FPGA hardware is introduced, which is freely configurable regarding neuron number and network depth. It is capable of processing parallel as well as time-lagged inputs. Applications as optimal filters, dense and time-lagged feed-forward neural networks are presented. A simulation and an implementation for FPGAs is considered optimizing the circuit with respect to resource usage and signal delay. An efficient use of digital signal processors is realized by time division multiplexing in order to use one network for multiple input channels. Results of test runs with a slow control, which allows memory-based data injection and readout under software control, are presented.