Augusto Nascetti | Searcharxiv

arXiv2026

A Hydrogenated Amorphous Silicon Photodiode indirect flexible device for radiation flux measurements at low intensities

The objective of the Photo-HASPIDE experiment is the construction and test of an indirect a-Si:H (Hydrogenated Amorphous Silicon) photo-detector plus scintillator device on a flexible substrate for the detection and measurement of particles fluxes (X-rays, electrons and protons) and for dosimetric measurements. The idea behind this experimental project lies in the utilization of Hydrogenated Amorphous Silicon (a-Si:H) as photodiode material; owing to its notable attributes of radiation hardness, light detection capability and mechanical flexibility. After the implementation of the HASPIDE experiment, which explored direct radiation detection using a-Si:H devices on a polyimide (PI) substrate, we aim to delve into indirect detection by employing these devices in conjunction with flexible and rad-hard scintillators like polysiloxane. The indirect detector design holds promise for improved responsiveness to low radiation fluxes compared to direct detection methods. The indirect a-Si:H detector should be composed of arrays of small (about 5 x 5 mm2) scintillator crystals read by a-Si:H photodiodes. Through optimization of the scintillator and detector thicknesses, we expect to achieve a better performance for low minimum detectable fluxes compared to direct detection methodologies. This new detector will find application in in-vivo dosimetry during radiotherapy or hadron-therapy and also, due to its expected fast response, in FLASH therapy. Another important application will be also in Solar Physics using these devices to measure particle fluxes in solar energetic particle events.

physics.ins-det↗

arXiv2025

Efficient FPGA-accelerated Convolutional Neural Networks for Cloud Detection on CubeSats

We present the implementation of four FPGA-accelerated convolutional neural network (CNN) models for onboard cloud detection in resource-constrained CubeSat missions, leveraging Xilinx's Vitis AI (VAI) framework and Deep Learning Processing Unit (DPU), a programmable engine with pre-implemented, parameterizable IP cores optimized for deep neural networks, on a Zynq UltraScale+ MPSoC. This study explores both pixel-wise (Pixel-Net and Patch-Net) and image-wise (U-Net and Scene-Net) models to benchmark trade-offs in accuracy, latency, and model complexity. Applying channel pruning, we achieved substantial reductions in model parameters (up to 98.6%) and floating-point operations (up to 90.7%) with minimal accuracy loss. Furthermore, the VAI tool was used to quantize the models to 8-bit precision, ensuring optimized hardware performance with negligible impact on accuracy. All models retained high accuracy post-FPGA integration, with a cumulative maximum accuracy drop of only 0.6% after quantization and pruning. The image-wise Scene-Net and U-Net models demonstrated strong real-time inference capabilities, achieving frame rates per second of 57.14 and 37.45, respectively, with power consumption of around 2.5 W, surpassing state-of-the-art onboard cloud detection solutions. Our approach underscores the potential of DPU-based hardware accelerators to expand the processing capabilities of small satellites, enabling efficient and flexible onboard CNN-based applications.

eess.SP↗