Simulation of random telegraph Noise with 2-stage equivalent circuit

Accurate Prediction of Random Telegraph Noise Effects in SRAMs and DRAMs

“…The state-of-the-art simulation-driven strategy (excluding SAMURAI and MUSTARD) is the one proposed by Ye et al [15], which uses a 2-stage comparator circuit to generate stationary RTN under constant bias, starting from an independent white noise source for each trap. However, this method cannot truly reproduce non-stationary RTN in circuits such as SRAMs and DRAMs.…”

“…9. Schematic proposed by Ye et al [15] for producing traces that approximate stationary (constant bias) RTN. shows that the bit is properly written (i.e., Q and Q settle to their desired values at the end of the write).…”

“…However, such results do not apply to circuits such as SRAMs and DRAMs, which are subject to large, rapid, and non-periodic bias swings. For such applications, Ye et al [15] recently proposed a 2-stage comparator topology, driven by white noise, for generating RTN. However, this applies only to stationary RTN at constant bias; it cannot perform non-stationary analysis.…”

“…3) We validate SAMURAI against analytical expressions known for stationary RTN (Section V), by showing that the statistical properties of SAMURAI-generated RTN 3 SAMURAI stands for SRAM analysis via Markov uniformization with RTN awareness incorporated, and was presented [16] at DATE 2011. This paper elaborates on [16], providing additional results and comparisons against a previously published method [15]. 4 MUSTARD stands for Markov uniformization based simulation of trap activity for RTN aware design, and was presented [17] at DAC 2011.…”

“…By contrast, SAMURAI and MUSTARD do not make such overly pessimistic predictions. 5) We compare SAMURAI with the 2-stage method recently proposed by Ye et al [15], from both an accuracy and an efficiency perspective (Section VII). We demonstrate that SAMURAI is not only more accurate, but also significantly more efficient than 2-stage RTN generation.…”