Single Event Upsets (SEU)
Single event upsets are unwanted effects in the latch state or memory cell of semiconductor devices caused by radiation.
Introduction
Single event upsets (SEU) are caused by ionizing radiation strikes in storage elements, such as configuration memory cells, user memory, and registers. In terrestrial applications, the main ionizing radiation sources of concern are alpha particles emitted from radioactive impurities in materials, high-energy neutrons produced by the interaction of cosmic rays with the earth's atmosphere and thermal neutrons that in most cases are thermalized high-energy neutrons but can also be produced in man-made equipment. Studies conducted over the last 20 years have led to high purity package materials minimizing SEU effects caused by alpha particle radiation. Unavoidable atmospheric neutrons remain the primary cause for SEU effects today. Soft errors are random and happen according to a probability related to energy levels, flux, and cell susceptibility.
Intel has been studying the effects of SEUs on its devices for many process generations and has built up extensive experience in both the reduction of soft-error rates through SEU-optimized physical layout and process technology, and in soft-error mitigation techniques. Intel introduced the industry's first automatic cyclical redundancy check (CRC) and removed the extra logic and complexity requirements common to other error checking solutions. Intel® device families are all tested for SEU behavior and performance using facilities such as Los Alamos Weapons Neutron Research (WNR) using standard test procedures defined by JEDEC's JESD-89 spec.
SEU testing of Intel® FPGAs at Los Alamos Neutron Science Center (LANSCE) has revealed the following results:
- No SEU errors have been observed in hard CRC circuit and I/O registers for all products, other than Stratix 10.
- There is a Mean Time Between Functional Interrupt (MTBFI) of hundreds of years, even for very large, high-density FPGAs.
Intel® Stratix® series, Arria® GX series, and Cyclone® series of FPGA families feature built-in dedicated hard circuitry to continually and automatically check CRC at no extra cost. For products manufactured on 28 nm process technology and subsequent process nodes, Intel has implemented CRAM upset bit correction (scrubbing) in addition to enhanced CRAM bit upset detection and correction. You can easily set up the CRC checker through the Intel® Quartus® Prime design software.
For more information regarding other mitigation techniques and for further details about SEU testing of Intel FPGA devices, please contact your local Intel sales representative or distributor.
Documentation
Documentation categorized by product lifecycle stages.