Highs and lows of radiation testing

The presentation concerns a practical approach for dealing with difficulties associated to real time testing in a natural environment of microelectronic devices. This activity has several redeeming characteristics that make it very interesting for reliability engineers that are concerned with the Single Event Effects (SEEs) affecting the functioning of their devices in a natural working environment. The paramount advantage that the circuit behavior observed during the experiment will be identical to the effects that may appear during the useful lifetime of the product. Thus, the test results will closely match real- life scenario from both quantitative (event rate) and qualitative (types of events) perspectives. However, some difficulties are still present and may strongly impact the interest of this testing method. The first difficulty consists in obtaining a significant number of events in order to be able to compute statistical data with an adequate order of accuracy. This requirement has repercussions on the hardware test setAcircnot up and test protocol but also on the mathematical methods that will be used for result analysis. An appropriate test set-up for real-time testing should consist in a high number of same-type DUTs (Devices Under Test). This way, the event rate of a single device will be accelerated by the number of devices under test. This will have an obvious impact on the preparation of the experiment. The hardware board, tester, power supplies, and other instruments will have to be designed to support a large number of devices. To further accelerate the apparition frequency of the errors, the devices should operate in a natural working environment with a higher particle flux. Since the neutron flux increases with the altitude following a well known equation, we should place the test set-up (DUTs, tester, power supply, support instrumentation) in a high-altitude test facility. This facility could be either mobile (atmospheric balloon, dirigibles, a- irplanes, satellites, etc) or fixed (ground-level). In this presentation we only present our tests in mountain research stations such as Jungfraujoch, Swiss and Plateau de Bure, France, since these facilities benefits from all amenities (stable power main, internet access, lodging, accurate neutron flux monitoring) while still offering a good acceleration (15-20 times more neutrons compared to sea-level testing). In any case, the practical duration of real-time tests is quite long (several months). The second difficulty is related to the origin of the SEEs. Since internal radioactive impurities will produce alpha particles that may also cause SEEs, we will need to discriminate the contribution of external and internal perturbations. Accelerated radiation testing doesn´t present this problems, since the test times are very short (few hours), making the relative contributions of alpha particles very low. However, since real-time testing is performed during several months, the alpha particles cannot be ignored. Thus, we should be able to measure/estimate the contribution of the alpha particles to the Soft Error Rate of the circuit. Two approaches are possible. The first one consists in measuring the error count of the devices placed in a shielded environment (such as a cave or an underground tunnel) where the contribution of neutrons can be minimized, or estimating the alpha SER using the alpha emissivity of the package (measured through alpha-counting) and sensitivity results from dedicated alpha testing on exposed silicon dies. The third difficulty consists in correctly analyzing the test results. The low event rate requires adequate mathematical methods. Other possible external perturbations or set-up instability may cause additional errors that could mix with SEE-induced errors. Lastly, since the experiment will run during a long period of time, the test set-up will have to be controlled and monitored remotely. Additionally, the neutron atmospheric flux should b