Skip to Main Content
The increasing number of safety-critical commercial applications has generated a need for components with high levels of reliability. As CMOS process sizes continue to shrink, the reliability of ICs is negatively affected since they become more sensitive to transient faults. New circuit designs must take this fact into consideration, and incorporate adequate protection against the effects of transient faults. This paper presents a novel method for protecting the pipelined execution unit of an embedded processor. It is based on a self-configured architecture with hybrid redundancy that can mask single and multiple errors, which can occur on storage elements due to transient or permanent faults. This concept can be easily applied to any processing architecture of this nature with a high safety integrity level. Results from error-injection experiments are also reported that show that this design can maintain a non-interrupted and failure-free operation under single and double errors with a probability that exceeds 99.4%.
Date of Conference: 2-4 Nov. 2005