Dr. Dennis (Scooter) Ball of Vanderbilt University, and Dr. Jeffrey Black of Sandia National Laboratories, will give a presentation on the multi-tooled approaches to modeling and simulation of single event transients in CMOS technologies. The course will cover the modeling tools being used: Technology Computer Aided Design (TCAD), radiation transport, circuit simulation, and fault injection/emulation. An example will be presented demonstrating how each of the tools are used to reproduce ground-based accelerator experimental data and then used to predict environmental rates. Further examples will be provided showing how single event modeling is done in emerging technologies like FinFETsand Gate All Around (GAA) FETs and what we have learned from modeling results.

Flash memories are ubiquitous in all digital systems today and are also attractive as non-volatile storage in space, because their capacity is unmatched by any rad-hard memory. In the first part of the short course, Prof. Marta Bagatin of the University of Padova will cover the effects of ionizing radiation in Flash memories. After reviewing the operation principles of these devices, total ionizing dose and single event effects will be illustrated. Possible short- and long-term phenomena following radiation exposure both in the memory cells and in the peripheral circuitry will be covered. The course will be supported by experimental data, analysis, and simulations, highlighting the impact of the technological evolution in a journey from traditional, planar devices to the most novel 3D integrated architectures. In the second half of this short course, Dr. T. Patrick Xiao of Sandia National Labs will review the current understanding of the effects of ionizing radiation and displacement damage on emerging non-volatile memories such as charge-trapping memory, magnetic memory, resistive memory, phase change memory, and electrochemical memory. The physics of radiation effects in emerging memories is often fundamentally distinct from effects on commercial NAND flash and other CMOS electronics. For non-charge-based memories, this leads to very high intrinsic radiation tolerance, making them attractive for space applications.

Furthermore, as the end of Moore’s law has slowed progress in computer processors, there has been significant research momentum to repurpose these emerging memories as low-power analog computational devices. However, using a traditionally binary memory as an analog memory can greatly increase radiation sensitivity. This course will also review the operation and design principles of analog in-memory computing systems and how the accuracy of analog computing is affected by radiation effects.

In the first part of this course, prepared in collaboration with Prof. Mike Wirthlin and Prof. Jeff Goeders, both of BYU will discuss radiation effects in modern, complex systems, with a focus on FPGAs, GPUs, and SoCs. These types of commercial off-the-shelf (COTS) devices are increasingly being used in space and other radiation environments due to their high performance, low cost, and rich features. However, these devices are subject to various single-event effects (SEEs) and total ionizing dose (TID) effects, which can manifest as silent data corruption, system crashes and hangs, and permanent damage. The complexity of these devices makes it difficult to model and predict their radiation response, as the devices typically contain substantial internal state and complex interactions between components.

Radiation testing of such systems requires substantial effort, time and expertise in developing appropriate test frameworks. This course will discuss both the radiation effects in these systems, as well as different strategies and lessons learned in how to effectively test and characterize these devices. In the second half of this course, Prof. Fernando Fernandes from INRIA, with the support of Prof. Paolo Rech from the University of Trento, will discuss the reliability challenges of adopting emerging post-Von Neumann architectures in safety-critical applications and space missions. While existing hardware architectures offer high computational performance, the memory bottleneck remains a limiting factor for energy efficiency and scalability since, in most applications, including AI, most of the power is wasted on data movement. New technologies such as processing in-memory and neuromorphic computing are emerging as alternatives to existing ones. Unfortunately, characterizing radiation effects and fault models on emerging hardware is challenging as conventional evaluation methods are not suitable for the task. The challenges include not only fault identification and correction but also designing radiation tests and qualification methodologies for emerging architectures.