spefit 3.0.0

Fitting of Single Photoelectron Spectra.

spefit

Optimised framework for the fitting of Single Photoelectron Spectra (SPE) in order to characterize the properties of photomultipliers which influence the measured illumination response.

• Supported Python versions: 3.9+
• Supported platforms: Linux, OSX
• Source: https://gitlab.cta-observatory.org/cta-consortium/aswg/tools/spefit
• License: BSD-3-Clause
• Citation: pending

Package Features

• Basic numpy API
• Runtime-selectable Probability Density Functions (PDFs), optimised using numba
• PDFs for the SPE spectra of both Photomultiplier Tubes and Silicon Photomultipliers
• Configuration of PDFs for the case where no pedestal peak exists (e.g. dark counting)
• Estimation of SPE parameters for improved initial fit values
• Runtime-selectable minimization cost definitions, optimised using numba
• Simultaneous fitting of multiple datasets (e.g. containing different average illuminations) for better parameter constraining
• Minimization provided by iminuit - Python frontend to the MINUIT2 C++ library
• Calculation of parameter errors and resulting p-value
• Extendable to allow the inclusion of any additional SPE description and minimization cost definitions
• Compatible with other minimization routines
• Convenience class provided for the parallel processing of cameras containing multiple photomultiplier pixels

Currently Implemented:

SPE formula:

• PMT Single Gaussian
• SiPM Gentile
• SiPM Modified Poisson

Minimization Cost functions:

• Unbinned Negative Log-likelihood
• Binned Negative Log-likelihood
• Least Squares

Installation

The package is available via pypi:

pip install spefit

Alternatively, this package is installable via pip directly from this repository (via https or ssh), e.g.:

pip install 'git+https://gitlab.cta-observatory.org/cta-consortium/aswg/tools/spefit.git#egg=spefit'

Also see the Makefile for development commands.

Optional Dependencies

Utilisation of Intel's short vector math library (SVML) for improved performance provided by numba:

conda install -c numba icc_rt

Usage

With a numpy array of size (n_events) called charge_array, containing the measured charges from the low illumination of a photomultiplier, the parameters of the SPE spectra can be extracted with:

from spefit.dataset import Dataset
from spefit.pdf import PMTSingleGaussian
from spefit.cost import BinnedNLL
from spefit.fitter import minimize_with_iminuit

charges = [Dataset(charge_array, n_bins=100, range_=(-3, 6))]
pdf = PMTSingleGaussian()
cost = BinnedNLL(pdf, charges)
values, errors = minimize_with_iminuit(cost)

Jupyter notebook demonstrations are provided in tutorials.