Fault resilient FPGA design for 28 nm ZYNQ system-on-chip based radiation monitoring system at CERN

CERN's new generation of radiation monitoring devices for radiation protection, CROME (CERN RadiatiOn Monitoring Electronics) uses a fully reconfigurable 28 nmXilinx Zynq SoC (System on Chip) for high-end embedded calculations, communication and data storage. In order to meet stringent safety requirements applicable forradiation protection instrumentation, CROME uses the FPGA section of the SoC for all safety critical functions. Whereas the SoC's dual core ARM processor is runningan embedded operating system which is used both for communication with the CERN supervisory system and for data management. Due to the use of an embeddedLinux OS without a virtualisation layer, the functional reliability of the SoC's FPGA section is considered much greater than the ARM processing system which can besubject to software crashes due to data corruption. This assumption had a central role in CROME's calculation architecture. In order to confirm the assumption andtherefore the reliability and robustness of our design, random faults have been voluntarily induced in the SoC by exposing it to ionising radiation of sufficient energyin the CHARM facility, creating Single Event Upsets (SEU). CROME is thefirst known radiation monitoring system using the FPGA section of a SoC in a safetyapplication. This paper presents the characterisation results of the system in the CHARM facility during a testing campaign of 6 months under an average dose rate of≥0.1 Gy/day. The tests have provided valuable information on the suitability of this architecture for similar applications.