Evaluating System-Level Radiation Hardness With Minimum Order Distributions

A framework for incorporating part-level total dose-induced failure distributions into system-level failure prediction with order statistics is presented. Minimum order distributions can describe system failure from a collection of parts when the parts are independent and identical, the part failure distribution is known, and a single part failure results in system failure. Confidence-bounded part failure distributions from total ionizing dose (TID) testing are used to estimate failure from a series connection of parts. Monte Carlo (MC) trials are used to confirm the confidence of the result. In addition, the formulation is modified to consider systems with nonidentical parts and systems with part dependencies. All cases are analyzed, and an example dependency between components is considered. Implications for board-level testing are explored. While breaking a board into components can improve the amount of constraint provided by a single board test, the number of part populations on the board and part dependencies significantly affect how much the failure probability is constrained.