This thesis presents methods and tools for the calibration and operation for the CALICE Analogue Hadronic Calorimeter (AHCAL), a highly-granular, steel-scintillator calorimeter designed for use in Particle Flow in a future precision lepton collider experiment requiring excellent jet energy resolution. This calorimetry method relies on highly-granular calorimeters, excellent detector calibration and sophisticated clustering algorithms to resolve energy deposits from different particles. The AHCAL has around 22,000 readout channels, utilising silicon photomultipliers (SiPMs) to read scintillation light, and is unique for its capacity to measure both energy and a timestamp with up to 100 ps timing resolution.
In this thesis, software tools and algorithms are developed to calibrate and operate the AHCAL and SiPMs. Firstly, a flexible Monte Carlo program called LightSimtastic is presented which simulates the response of SiPMs in the linear regime in which saturation effects can be ignored. Inputs to the program are the mean number and time distribution of Geiger discharges from photons, and the dark-count rate. Then, another software tool for the characterisation of SiPM spectra called PeakOTron is introduced. This program fits the entire charge spectra, including the intervals in-between the photoelectron peaks, which allows determining, in addition to the mean number of detected photons, gain, gain spread, prompt cross-talk, pedestal, and electronics noise, the dark-count rate as well as the probability and time constant of after-pulses from charge spectra. The starting values of the fit parameters are extracted from the charge spectra, and the program provides a good description of both simulation and experimental data. Thirdly, a neural network model for software compensation developed for the AHCAL is presented, using spatial and temporal event information from the AHCAL and energy information, which is expected to improve sensitivity to shower development and the neutron fraction of the hadron shower. The method produced a linear detector response in compensating both simulation and experimental hadron shower data. It outperformed a published control method in terms of resolution for every particle energy studied. Lastly, neural network models for shower separation are applied to separating charged and synthetic neutral hadron shower events using the AHCAL detector. The AHCAL is demonstrated to be a highly effective Particle Flow Calorimeter, with > 90 % of events being reconstructed in the calorimeter resolution for most particle energy combinations that improve significantly in the most challenging cases using timing information as an additional clustering dimension.