i-RASE: intelligent radiation sensor readout systems for real-time photon-by-photon radiation detection

Intelligent Radiation Sensor Readout Systems (i-RASE) is a European Innovation Council (EIC) funded project that advances the integration of physics-informed artificial neural networks (ANNs) directly into the readout chain of semiconductor radiation detectors for real-time, photon-by-photon signal reconstruction. The demonstrator application is designed to combine a 3D CdZnTe drift-strip detector, a custom low-noise front-end electronics, high-rate waveform digitization, and FPGA-based ANN inference to estimate photon energy, interaction position, and timing while reducing reliance on conventional pulse-shape analysis. The developments presented in this work demonstrate the performance of the individual building blocks of the i-RASE system. An ANN trained on synthetic data derived from detector models and experimental characterization achieves a spatial resolution of 0.5 mm full width at half maximum at 662 keV, is limited primarily by electronic noise, and exceeds the performance of conventional reconstruction methods. A custom eight-channel readout ASIC has now been implemented in silicon, and first characterization measurements are in progress. It integrates charge-sensitive preamplifiers with pulsed-reset feedback, leakage-current compensation, support for both signal polarities, wide dynamic range, and high count-rate capability. The digital readout architecture employs synchronized cyclic buffers and neural-state machines to perform efficient event detection and provide waveform data for FPGA-based ANN processing. The presented detector, electronics, and real-time processing architecture together demonstrate the feasibility of ANN-assisted radiation sensing and provide a foundation for future compact, low-power systems for space instrumentation, medical imaging, nondestructive testing, and environmental radiation monitoring.