Research And Implementation Of Soft Error Sensitivity Analysis Methodlogy For SystemC

With the continuous advancing in VLSI manufacturing technology, the soft error issue due to particle radiation and circuit internal noise has caused more and more serious impact on system reliability. The soft error sensitivity analysis in circuit design stage can be used to evaluate the circuit reliability, decrease the design iterations and save development costs. With the increasing complexity of circuit design and the development of IP reuse technology in modern IC design, SystemC language is widely used in high level system modeling and verification of electronic system level design because of the superior hardware and software collaborative design ability.This paper proposed a method to analyze soft error sensitivity for SystemC. This method realizes the random fault injection and the system monitoring by the system simulator during the system running. The soft error sensitivity can be calculated based on the statistical data.Firstly, the high level system modeling in SystemC based on the equivalence of SystemC and Verilog languages can effectively improve the simulation efficiency. Verify the system function by the SystemC/Verilog hybrid simulation method with Modelsim simulator. Secondly, the overall consideration of fault injection points and the introduction of module area influence factor can improve the soft error rate analysis accuracy. Thirdly, the compression of fault sample space through the stratified sampling strategy also can be used to speed up the simulation.With the fault injection method and the simulation speedup strategy, the fault injection platform is built on the SystemC model which is the minus system based on OR1200 processor. This platform is used for about 300,000 fault injection experiments based on the stratified sampling strategy. The analysis shows that the register file module is the most sensitive to soft error, the overall system soft error sensitivity is about 0.0063. The effectiveness and stability of this method is proved through the module soft error sensitivity analysis based on the experimental statistical data. In the existing research, the soft error sensitivity is calculated without area factor. Thus, the analysis method proposed in this paper has better precision.