Extract templates for 2025-A data

Description

We decided to explore the idea of reconstructing data from 2025a run (e^- beam session) exploiting template fitting for WB-based detectors.

The advantage of this approach is that we do not need to introduce an analytical parameterization of the signal shape -- possibly different cell-by-cell --, but we rely on data to extract a description of the expected signal shape.

As discussed with @lmarsica and @rdusaev, in the construction of the template function of each cell we will not assume that a single template function can be used for all pulse amplitudes, simply scaling it by a constant factor. Rather, we will analyze data by categorizing with respect to the pulse amplitude A, creating for each amplitude interval (to be defined) a specific template function f_A.

Events filtering

The creation of templates require to select input events that are as clean as possible. These are ideally characterized by the presence of a single, well-identified signal pulse in the center of the readout window, without any pile-up effect. This selection can be based either on the current cell being considered for template creation, but also on signals from other cells, exploiting the physical correlation. For example, scintillator counters S_0, S_1, S_2, S_3 can be exploited: these are read by very fast, low-noise PMTs, and thus we can isolate though them events with (almost) no pileup.

Template preparation

• Waveform reconstruction: we start using the p348-reco algorithms to determine, for each pulse, the amplitude A and the time t. The algorithm to reconstruct the pulse time is based on a constant-fraction like approach.

  • For ECAL, HCAL, what about using the v1 fit algorithm based on the n-pole function?

• Waveform alignment: before merging together different pulses from different events, these have to be aligned in time. We use the pulse time t measured before.

• Waveform normalization: although the amplitude of the pulse is used to categorize the pulses, still in each category (or "bin") there will be pulses with slightly different amplitudes. These are normalized so that the maximum amplitude is 1.

Template output

For the moment, let's adopt a ROOT-based approach, the template is a TH3D, where the z-axis is the pulse amplitude (in ADC), the x-axis is the time (in ns), the y-axis is the template (aka, the normalized signal value).