Publications (1)0 Total impact
[show abstract] [hide abstract]
ABSTRACT: Memory systems used in space applications suffer from radiation-induced errors, either temporary upsets (soft errors) or permanent defects (hard errors or stuck-at errors). Scrubbing is a method to protect memory contents by periodically decoding the stored data to correct those soft and stuck-at errors then rewriting the corrected data back into memory. However, defective cells will remain and accumulate over time. Conventional coding disregards defective cells, however this may be inefficient for memory protection in space. In this study, alternative coding schemes for scrubbing are investigated, where the channel model depends on the cell states, defective or not, and the encoder uses channel state information (CSI) or side information. At every scrubbing, the error correcting code (ECC) decoder provides partial CSI back to the encoder and the encoder uses the CSI to improve the performance of memory systems with scrubbing. Information theoretic limits of the channel with partial CSI are investigated and several coding schemes are introduced to mitigate the effects of defective cells, particularly those caused by stuck-at defects. In addition, coding schemes with partial CSI are concatenated with binary Bose-Chaudhuri-Hocquenghem (BCH) codes to protect memory contents from both soft and stuck-at errors in space radiation environments. Numerical simulation results show that scrubbing with partial CSI improves reliability over the state-agnostic approaches.