Electrical Measurement of Cell-to-Cell Variation of Critical Charge in SRAM and Sensitivity to Single-Event Upsets by Low-Energy Protons
James M. Cannon, T. D. Loveless, Rafael Botella Estrada, Ryan Boggs, S. P. Lawrence, Gabriel Santos, Michael W. McCurdy, Andrew L. Sternberg, Donald R. Reising, Thomas Finzell, Ann Cannon
Abstract
The distribution of single-event upsets (SEUs) in commercial off-the-shelf (COTS) static random access memory (SRAM) has generally been thought to be uniform in a device. However, process-induced variation within a device gives rise to variation within the cell-level electrical characteristic known as the data retention voltage ( V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> DR</sub> ). Furthermore, V <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> DR</sub> has been shown to directly influence the critical charge ( Q <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> C</sub> ) necessary for stored data to upset. Low-energy proton irradiation of COTS SRAMs exhibits a preference toward upsetting cells with lower Q <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> C</sub> . An electrical procedure is presented to map relative cell critical charge values without knowledge of the underlying circuitry, allowing customized device usage. Given cell-level Q <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> C</sub> knowledge, a device can have its cells “screened” such that those with low Q <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> C</sub> are not used to store data. This work presents the results such a screening process has on the SEU per bit cross section.