In-flight upset - 154 km west of Learmonth, WA, 7 October 2008,

3.6.6 Single event effectsBackgroundThere is a constant stream of high-energy galactic and occasional bursts of solarradiation interacting with the Earth’s upper atmosphere. That interaction creates acascade of secondary particles, and some of those particles (particularly neutrons)can affect aircraft avionics systems. A single event effect (SEE) is the response ofan electronic component to the impact of a single particle. Unlike EMI, whichaffects interfaces between chips, SEE affects a chip directly.The most common type of SEE, a single event upset (SEU), occurs when theparticle deposits an electric charge inside a memory cell that is sufficient to changethe cell’s logic state (known as a ‘bit flip’). A range of other types of SEE can alsooccur. Although some SEEs can result in permanent damage to components, mostare soft faults. 151High-density integrated circuits, such as memory and CPU chips, can beparticularly susceptible to SEEs due to their relatively large number of memorylocations and the reduced ‘feature’ size (that is, functional parts in newer chipsgenerally become smaller and more sensitive). SEEs have been confirmed in space-,air- and even ground-based computer systems, and have been shown to generatesoft errors in a wide range of different aircraft systems.The probability of an SEE occurring to a particular component at any particularpoint in time is dependent on a range of factors, including the number, energy, andtype of particles in the area of the component’s operation and the component’ssensitivity to SEE. Other relevant factors include the direction and location of theparticle strike, and the time in the device’s program cycle at which the strike occurs.Most SEEs have no adverse effects on a system’s performance. For example,changing the logic state of a bit within a data word would change the value of thatword, but only until the word was refreshed with a new value.More adverse effects occur when the cells that are affected include program data;that is, information critical to the successful execution of the core software, such assoftware instructions, CPU register information or program pointers. For example,software might ‘jump’ to the wrong location in a program, or even to a locationoutside of program memory and treat the data there as valid program instructions.Most of this critical information is stored and used internally within the chip. As thesuccessful execution of software depends heavily on program ‘states’ (that is, thetasks being performed at any given point in time), such a disruption can havewide-ranging and complex outcomes. The most common type of adverse effect inthese cases results in the system ‘hanging’, although many other adverse states arealso possible and may not be predictable in practice.System manufacturers use a variety of techniques to reduce the effects ofunintended changes of logic states, whether they are due to an SEE or other source.These include hardware and software design features such as redundancy,monitoring and partitioning. A specific technique to reduce the effects of SEE is151In the SEE field, a ‘soft fault’ occurs when the logic state of a cell is changed to an invalid value,and a ‘firm fault’ refers to a failure that can only be reset by rebooting or cycling the power to thesystem. In this report, both of these phenomena are termed soft faults.- 143 -