[Paper Review] FADC Pulse Reconstruction Using a Digital Filter for the MAGIC Telescope

H. Bartko, M. Gaug+2 more|arXiv (Cornell University)|Jun 20, 2005

Astronomy and Astrophysical Research3 citations

TL;DR

This paper presents a digital filtering method for reconstructing charge and arrival time from FADC samples in the MAGIC Cherenkov telescope, using a known pulse shape and noise autocorrelation to minimize noise contributions. The technique achieves sub-ns timing resolution (down to 200 ps for large signals) and superior resolution over alternative algorithms, enabling lower energy thresholds and improved gamma-hadron separation.

ABSTRACT

Presently, the MAGIC telescope uses a 300 MHz FADC system to sample the transmitted and shaped signals from the captured Cherenkov light of air showers. We describe a method of Digital Filtering of the FADC samples to extract the charge and the arrival time of the signal: Since the pulse shape is dominated by the electronic pulse shaper, a numerical fit can be applied to the FADC samples taking the noise autocorrelation into account. The achievable performance of the digital filter is presented and compared to other signal reconstruction algorithms.

Motivation & Objective

• To improve signal reconstruction accuracy in the MAGIC Cherenkov telescope by minimizing noise contributions in FADC samples.
• To achieve high timing and charge resolution for low-amplitude signals, especially below 100 GeV.
• To reduce the energy threshold of gamma-ray analysis by enhancing signal-to-noise ratio through optimal filtering.
• To enable better discrimination between gamma-induced showers and background events (e.g., night sky background) using precise timing information.
• To validate the performance of digital filtering against alternative reconstruction methods like spline interpolation and sliding window techniques.

Proposed method

• The method uses a linearized model of the signal as a function of amplitude E and time shift τ, expressed as yi = E·gi − Eτ·ġi + O(τ²) + bi.
• It applies a χ² minimization with respect to the known noise autocorrelation matrix B to estimate E and Eτ, leading to optimal weight vectors w_amp and w_time.
• The optimal weights are derived analytically from the pulse shape g(t), its derivative ġ(t), and the inverse noise autocorrelation matrix B⁻¹.
• The algorithm accounts for pulse stretching (6 ns) to ensure constant signal shape and noise properties, satisfying the required assumptions for optimal filtering.
• Iterative refinement is applied to reduce O(τ²) errors by updating the pulse shape and weight functions based on the estimated τ.
• The method is implemented using FADC samples at 300 MS/s, with separate high and low gain branches to extend dynamic range.

Experimental results

Research questions

• RQ1 Can digital filtering achieve better charge and timing resolution than conventional methods like sliding window or spline interpolation in FADC-based Cherenkov telescopes?
• RQ2 What is the achievable timing resolution for low-charge signals (e.g., 10 photo-electrons) using this filtering approach?
• RQ3 How does the digital filter reduce noise contributions to amplitude and time reconstruction errors?
• RQ4 To what extent can this method lower the energy threshold of gamma-ray analysis in the MAGIC telescope?
• RQ5 How well does the method preserve resolution despite small fluctuations in pulse shape and noise behavior?

Key findings

• For 10 photo-electrons, the digital filter achieves a timing resolution of 700 ps, significantly improving sensitivity to low-energy showers.
• For large signals, the timing resolution reaches 200 ps, demonstrating sub-nanosecond precision in arrival time reconstruction.
• The digital filter outperforms both cubic spline interpolation and sliding window methods in both charge and timing resolution across all pulse heights.
• The method reduces noise contributions to reconstruction errors by optimally weighting FADC samples based on the known pulse shape and noise autocorrelation.
• The algorithm enables lower image cleaning thresholds and reduced energy thresholds in gamma-ray analysis due to improved signal-to-noise performance.
• The technique is robust under realistic conditions, including minor pulse shape variations and noise behavior fluctuations, due to pulse stretching and stable assumptions.