Study On Random Fault Injection Technology Based On Perl And Verilog-A

With the reduction of feature size,the silicon-based CMOS integrated circuit technology has entered deep submicron and even nano-age,the single event effects has intensified and has replaced the total dose effect as the uppermost radiation effect,single event effects have become the most pressing problem in space IC design.At present,for the simulation of single event transients in single event effects,many years of research have been carried out,and some achievements have been achieved in device-level simulation techniques,however,circuit-level simulation for large-scale circuits still lacks effective technical means.Establishing a circuit-level model to evaluate the anti-radiation ability and sensitive factors of a large-scale circuit,obtain effective reinforcement techniques and experimental methods to obtain effective reinforcement techniques and experimental methods is very important.In this paper,the commonly used circuit-level single-particle transient pulse model is introduced first,and after the analysis and summarization of each model,a standard double-index current source is selected.Secondly,based on the SMIC 130 nm silicon-based CMOS process,using TCAD software,an NMOS device-level model was established to obtain the total charge and pulse current collected by the NMOS drain with different linear energy transfer.According to the above simulation results,the related parameters of the double exponential current source model are obtained by fitting with MATLAB software,and the analytical expressions of the parameters of the LET and double exponential current source models are established,and the total charge and pulse current obtained by the expression and the TCAD device level are established.Comparing the simulation results,the maximum error does not exceed 10%.For large-scale circuit simulation,since the peak current and pulse width of the single-particle pulse are random when the particle striking the circuit,this article using the Verilog-A behavior description language to further improve the fitted double-exponential current source,and randomize their current peaks and pulse widths.A wide range of random processing is performed.Statistical results show that the peak current and pulse width of the current source follow uniform random distribution.Subsequently,the critical path of the RISC processor is extracted,and the faulty injection is performed on the sensitive nodes of the critical path,the simulation results show that the critical path is the path most affected by the single event effect.For the improved double exponential current source model,this paper uses the SRAM six-cell unit to verify its circuit level on the Cadence spectre platform.Compare the simulation result with the LET flipping threshold in the research literature,which is consistent with the results in the literature.A RISC processor is designed and the Perl description language is used to realize the extraction and fault injection of sensitive nodes in the logic circuit,inject the improved double exponential current source into the circuit net-list,and the analysis shows that the single-particle effect manifests in the circuit as single-particle transient and single event upset,the analysis shows that there are two kinds of transient pulse current propagation in the circuit: one case is transient pulse current because of logic masking,latching window masking two masking effects does not lead to circuit function error,and the other is the pulsed current is captured by the clock and causes circuit malfunctions.Subsequently,the critical path of the RISC processor is extracted,and the faulty injection is performed on the sensitive nodes of the critical path,the simulation results show that the critical path is the path most affected by the single-particle effect.Finally,randomly inject faults into the RISC processor,randomly selecting nodes and occurrence time by using Perl scripts,statistics show the error rate of RISC processors under different LETs,provides evidence for evaluating its anti-single particle capabilities.