Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Signal Processing and Machine Learning Algorithms for Precise Timing with PICOSEC Micromegas Detectors

A. Kallitsopoulou, I. Maniatis, I. Manthos, T. Papaevangelou, L. Sohl, A. Tsiamis, S. E. Tzamarias

TL;DR

This work tackles online, sub-20 ps timing for PICOSEC-Micromegas detectors to mitigate pile-up in high-rate experiments. It develops an emulation model to reproduce multi-photoelectron PICOSEC pulses from single-photoelectron waveforms and uses it to train an Artificial Neural Network for timing, benchmarking against traditional CFD and a constant-threshold technique. The empirical results show an achievable timing resolution of about $18.3 \,\pm\, 0.6$ ps with CFD and $18.5 \,\pm\, 0.6$ ps with the ANN under similar conditions, with the emulation enabling scalable training. The findings demonstrate that high-precision timing is feasible online and can be integrated with low-cost front-end electronics, offering pathway toward real-time triggering and event reconstruction in future high-rate experiments.

Abstract

High particle rates in current and future experiments make pile-up phenomena a critical issue for extracting useful information. In this context, timing can be important as the 4$^{\mathrm{th}}$ dimension parameter for triggering or event reconstruction. The PICOSEC-Micromegas detector has been shown to offer precise timing of the order of tens of\,ps. In this work, novel signal processing algorithms are being developed and evaluated to demonstrate the technology's ability for online precise timing. We propose, an algorithm based on Artificial Neural Networks (ANN). This algorithm uses a model to train the ANN. The performance of the different algorithms is evaluated using experimental data, resulting in a timing resolution of 18.3 $\pm$ 0.6\,ps, comparable to the standard analysis based on the Constant Fraction Discrimination technique. Additionally, an alternative algorithm using the charge of the pulse exceeding a threshold as a parameter to correct for systematic effects is reported.

Signal Processing and Machine Learning Algorithms for Precise Timing with PICOSEC Micromegas Detectors

TL;DR

This work tackles online, sub-20 ps timing for PICOSEC-Micromegas detectors to mitigate pile-up in high-rate experiments. It develops an emulation model to reproduce multi-photoelectron PICOSEC pulses from single-photoelectron waveforms and uses it to train an Artificial Neural Network for timing, benchmarking against traditional CFD and a constant-threshold technique. The empirical results show an achievable timing resolution of about ps with CFD and ps with the ANN under similar conditions, with the emulation enabling scalable training. The findings demonstrate that high-precision timing is feasible online and can be integrated with low-cost front-end electronics, offering pathway toward real-time triggering and event reconstruction in future high-rate experiments.

Abstract

High particle rates in current and future experiments make pile-up phenomena a critical issue for extracting useful information. In this context, timing can be important as the 4 dimension parameter for triggering or event reconstruction. The PICOSEC-Micromegas detector has been shown to offer precise timing of the order of tens of\,ps. In this work, novel signal processing algorithms are being developed and evaluated to demonstrate the technology's ability for online precise timing. We propose, an algorithm based on Artificial Neural Networks (ANN). This algorithm uses a model to train the ANN. The performance of the different algorithms is evaluated using experimental data, resulting in a timing resolution of 18.3 0.6\,ps, comparable to the standard analysis based on the Constant Fraction Discrimination technique. Additionally, an alternative algorithm using the charge of the pulse exceeding a threshold as a parameter to correct for systematic effects is reported.

Paper Structure

This paper contains 12 sections, 4 equations, 20 figures.

Table of Contents

  1. Introduction
  2. Alternative Timing Technique with Constant Threshold
  3. An Emulation Model for PICOSEC-MM Pulses
  4. Characterization of Single-Photoelectron Waveforms Properties
  5. Synchronizing and Summing Single Photo-electron Waveforms
  6. Validation of the Emulation Model
  7. Comparison of the Emulation Model with Experimental Data (EXP-set)
  8. Evaluating the Timing Properties of the Emulation Model
  9. Using the Emulation Model to Train an Artificial Neural Network
  10. Validation of the Artificial Neural Network's Performance
  11. Validation of the ANN’s Unbiased Performance
  12. Conclusions

Figures (20)

  • Figure 1: (a) A graphical representation of a PICOSEC MicroMegas detector BORTFELDT2018317. (b) The typical pulse of a PICOSEC-MM Detector is a two-component signal. It comprises a fast e-peak and an extended ion tail of the slow-moving ions.
  • Figure 2: Definition of the charge above relevant threshold, $Q_{up}$.
  • Figure 3: (a) SAT as a function of Charge above Threshold for the lowest threshold at 100 mV, fitted with the corresponding power law curve. (b) Resolution as a function of e-peak amplitude after time walk correction, using both Constant Fraction Discrimination (red points) and multi-charge over threshold timing techniques (black points), for the same amplitude threshold.
  • Figure 4: SAT as a function of Charge above Threshold for the three other applied thresholds, fitted with the corresponding power law curve. Threshold at: (a) 200 mV, (b) 400 mV and (c) 600 mV.
  • Figure 5: Time resolution as a function of e-peak amplitude after time walk correction, using both Constant Fraction Discrimination (red) and multi-Charge over threshold timing techniques (black), for threshold at (a) 200 mV, (b) 400 mV, (c) 600 mV.
  • ...and 15 more figures