GPU Based Nearline Analysis System for the Nab Experiment

ORAL

Abstract

The Nab neutron decay correlation experiment will utilize a pair of 127-pixel silicon detectors to measure the electron-neutrino correlation coefficient from the 2D energy distribution of protons and electrons. For each decay event, analog signals from ~30 pixels are digitized at a rate of 250 MHz resulting in an approximately 50 MB/s data rate. First-order timing and energy results are extracted by the data acquisition firmware using traditional filtering techniques, but in order to reach the precision goals of the experiment, a new technique is required. A GPU-based weighted least squares fitting routine has been developed that fits template waveforms, while subtracting baseline noise features, at an excess of 100 MB/s per GPU. This system will be used as part of the data acquisition process as well as during offline analysis on GPU clusters.

Authors

  • David Mathews

    University of Kentucky