Majority voting algorithm based on self-test
Jihong Zhu · Journal of Tsinghua University(Science and Technology) · 2005
A majority voting algorithm based on self-test was developed to resolve the no-output problem in cases of disagreement in the majority voting algorithm and to more effectively cope with transient errors. This algorithm records transient fault information through self-test codes inserted separately into the task codes for use as auxiliary voting information. A self-test sequence was developed for detecting random access memory (RAM) and arithmetic logical unit (ALU) errors. A simple code interpreter was created to simulate the effects of detect density, error duration and single modular dependability. Tests show that, in the best case, the method gives 3% more correct results than the extended history based majority voting algorithm and 15% more than the majority-median voting algorithm.