Soft error propagations and effects analysis on CAN controller

Abstract

This paper presents the evaluation of soft error effects and propagations on a Controller Area Network (CAN) controller using SINJECT fault injection tool. To do this, three main sub-modules of the core were selected for fault injection targets. The experimental results of fault injection sites show that the Register module is more fault tolerant than the other parts, since only 0.3% of the injected faults led to a system failure. On the other hand, the Bit Timing module is the most vulnerable part in the CAN controller, since the failure rate for this module is about 93%. The Bit Stream Processor is somehow in the middle, because from all injected faults in the Bit Stream Processor, 6.2% of them lead to system failure, 70.1% of them were detected and recovered during simulations, and 23.7% of them were remained latent in the controller.