Fault Injection Model of SRAM Memory Circuit Based on Hybrid Modeling

As one of the important parts of the space system, the reliability of the SRAM storage circuit always affects the safety of the spacecraft. The harsh environment full of radiation and extreme temperature in space brings radiation effect and aging effect to the circuit, which causes accidental damage and life reduction of the device, and in serious cases, circuit error and system failure. In this paper, from the perspective of fault model and fault injection, the hybrid modeling method of device level and circuit level is adopted to establish the fault model. At the same time, the sensitive area and the randomness of the fault are considered, and the switching factor is introduced to comprehensively consider the fault injection to improve the accuracy of the model. A device-level Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET) simulation model is established in Technology Computer-Aided Design (TCAD), and the process calibration of the established MOSFET model is carried out through the process library file, to better approximate the actual model. Combined with the established MOSFET device model, the failure mechanism of single-particle effect in radiation effect and negative bias temperature instability effect in aging effect was studied. Based on the physical mechanism of the device level, the fault is mapped to the circuit level and the corresponding hybrid model is established. Combined with the advantages of the Veriloga language, the current source, and voltage source models are established, and the fault sensitive area and the randomness of fault occurrence are considered, and the switching factor is introduced into the fault model. Then, the established fault model is injected into the SRAM storage circuit in the form of fault devices for fault simulation, and the power current data is collected and compared with the fault model without a switching factor. Taking the measured data as a reference, the data errors of the two are compared. Experimental results show that the proposed fault model fits the actual data better and has higher accuracy.