Multiple sensitive volume based soft error rate estimation with machine learning

Nanometer integrated circuits are getting increasingly vulnerable to soft errors and making the soft error rate (SER) estimation an important challenge. In this study, a novel approach is proposed for SER estimation of combinational circuits based on vulnerability analysis. The authors introduce a concept called probabilistic vulnerability window (PVW) which is an inference of necessary conditions for a single event transient (SET) to cause observable errors in the circuit. A proposed computational framework calculates PVWs for all circuit gates in a backward‐traversing algorithm enabling the circuit designers for an accurate and efficient SER estimation. Experimental results show that the proposed approach is 2× faster than the traditional SER estimation methods and keep its efficiency when it is applied for estimating the SER considering various different SET widths while runtime of traditional estimation methods increases in such cases. In addition, results verify the accuracy (average difference of 0.02) and speedup (about four orders of magnitude) of the proposed method when compared with the Monte Carlo‐based fault injection simulation on ISCAS'85 benchmark circuits.

Fast and accurate Soft Error Rate (SER) estimation is an important system design aspect. With the continued scaling of SRAM bit-cell dimensions and cell pitches, accurate modeling of the Multi-Bit Upset (MBU) component is becoming increasingly critical. This paper introduces MBU-Calc, a compact model for MBU SER estimation. The tool leverages Multi-Cell Upset (MCU) probabilities obtained through direct test-chip measurements. The main improvement with respect to prior published work [2] is the introduction of so-called Line Filling (LF) factors which enable more accurate projection and bucketing of silent data corruption (SDC) versus detected unrecoverable errors (DUE). The accuracy of the introduced model is demonstrated against measured MBU results.

Recent advanced high-speed communication systems, such as optical systems, require highest reliability at lowest possible power consumption. Thus, Pass Transistor Logic (PTL) is gaining lots of interest in these communication systems due to its power saving potential compared to traditional CMOS logic. However, due to the non-conventional logic structure, its susceptibility to radiation-induced soft errors is different from CMOS circuitry. Due to the unique generation and propagation of Single Event Transients (SETs) in PTL, different approaches for PTL soft error rate (SER) estimation are required. In this paper we propose a machine learning (ML) approach for SET propagation in PTL logic. Multi-layer feed-forward neural network together with support vector classifier (SVC) are used to build the SET pulse width and pulse amplitude models. Bayesian optimization using Gaussian Processes is utilized to tune the hyperparameters of neural network. The experimental results on full adder (FA), which is the key component in many large cirucits such as ALU, and comparison with Monte Carlo (MC) spectre simulations confirm the accuracy and speed of the proposed method.

Soft error rate (SER) estimation is becoming more and more important since nanometer digital integrated circuits are getting increasingly vulnerable to soft errors. In this paper, a novel approach is proposed for soft error rate analysis of digital combinational circuit considering all masking factors. We introduce a concept called Probabilistic Vulnerability Window (PVW) which is an inference of the necessary conditions for a Single Event Transient (SET) to cause observable errors in the circuit. A computation model is proposed to calculate PVW's for all circuit gate outputs. Using the computation model, the proposed method estimates the soft error rate of the circuit by computing the probabilistic vulnerability windows in a backward approach. Experimental results show that the proposed method increases the SER computation speed by 1000X, with less than 10% accuracy loss when compared to the Monte-Carlo based fault injection methods. The results also show than the proposed approach keeps its efficiency when it is applied for estimating the soft error rate considering various SET's with different initial widths while the runtime of traditional SER estimation methods increases rapidly in such cases.

Efficient algorithms to accurately compute derating factors of digital circuits

Soft error simulations utilizing PHYSERD are presented. The comparison with the experimental data for 28 nm SRAM demonstrates that PHYSERD provides the high-accuracy estimation of soft error rate when combined with circuit simulation, while PHYSERD alone tends to overestimate soft error rate. We also analyze the contributions of transistors constituting SRAM and of secondary ions to soft error rate. We find that circuit simulation has a significant effect on these contributions and is essential to the prediction of soft error rate.

While circuit density and power efficiency increase with each major advance in IC technology, reliability with respect to soft errors tends to decrease. Current solutions to this problem such as TMR require high area and power overhead. In this work, soft-error reliability is improved with minimal area overhead by careful, localized circuit restructuring. The key idea is to increase logic masking of errors by taking advantage of conditions already present in the circuit, such as observability don't-cares. We describe two circuit modification techniques to improve reliability: don't-care-based resynthesis and local rewriting. A key feature of these techniques is fast, on-the-fly estimation of soft error rate (SER) using our reliability evaluator AnSER. This tool is compared against prior SER evaluators and found to run orders of magnitude faster. We show empirically that our reliability-driven synthesis methods can reduce SER by 29-40% with only 5-13% area overhead.

An accurate model for soft error rate estimation considering dynamic voltage and frequency scaling effects

A simple methodology is necessary to characterize the SEU behavior of large quantities of RAM cell types, latches, flip-flops, other logic cells, I/O cells, etc. Such a methodology, called "Boxes", is being used for the Honeywell S150 radiation-hard 0.15 mum partially-depleted SOI process. Using physics-based equations, this paper shows how to break up each critical transistor into several "boxes", each with its own dimensions and critical charge, for the purpose of calculating soft error rate (SER). The Boxes methodology also allows for calculation of SER due to an ion that must simultaneously strike two separated sensitive volumes in order to cause an upset. This can be the dominant upset mechanism for many types of cells such as certain hardened SRAM's and other cells that obtain radiation hardness via extra transistors (such as triple modular redundancy, the DICE latch, etc.). Boxes also predicts upsets that can occur when an ion strike pulls a circuit node below ground or above the positive power supply. The boxes methodology was applied to a 6 T non-hardened SRAM, a hardened SRAM, and a D-type flip-flop. The theoretical predictions correlated well with experimental vertical ion strike data

This paper presents a new 8T (8-transistor) SRAM cell layout mitigating multiple-bit upset (MBU) in a divided wordline structure. Because bitlines along unselected columns are not activated, the divided wordline structure eliminates a half-select problem and achieves low-power operation, which is often preferred for low-power / low-voltage applications. However, the conventional 8T SRAM with the divided wordline structure engenders MBUs because all bits in the same word are physically adjoining. Consequently, error correction coding (ECC) techniques are difficult to apply. This paper presents a new 8T cell layout pattern that separates internal latches in SRAM cells using both an n-well and a p-substrate. We investigated an SEU cross section of nMOS that is 3.5–4.5 times higher than that of pMOS. Using an iRoC TFIT simulator, we confirmed that the proposed 8T cell has better neutron-induced MBU tolerance. The MBU in the proposed 8T SRAM is improved by 90.70% and the MBU soft error rate (SER) is decreased to 3.46 FIT at 0.9 V when ECC is implemented. Additionally, we conducted Synopsys 3-D TCAD simulation, which indicates that the LET threshold (LETth) in single-event upset (SEU) is also improved by 66.47% in the proposed 8T SRAM by a common-mode effect.

Accelerated soft error testing with proton beam was performed for 65 nm CMOS latches for the first time. The soft-error rate (SER) dependence on the physical layout was clarified. SER has dependence on the size of the diffusion regions since critical charge and charge collection is strong function of them. By optimizing it, SER can be reduced by 70%. The scaling trend of SER was also investigated. It is shown that SER degradation due to scaling can be suppressed by the moderate reduction of the supply voltage.

A soft error rate analysis model for MOS dynamic RAM's is presented. The soft error rate can be quantitatively calculated by using a solution of the equations for diffusion and collection of alpha-particle-induced excess electrons and by combining a statistical treatment of alpha particle energy, incidence angles, and incidence positions with the noise charge calculation. The model is then applied to analyze a soft error experiment on 64-kbit dynamic RAM's. It is shown that soft error characteristics with regard to signal charge (critical charge), as well as alpha energy and incidence angle dependencies, can be definitely determined. The model can also be used to predict the location of soft errors in MOS dynamic RAM's.

Near threshold computing (NTC) has significant role in reducing the energy consumption of modern very large scale integrated circuits designs. However, NTC designs suffer from functional failures and performance loss. Understanding the characteristics of the functional failures and variability effects is of decisive importance in order to mitigate them, and get the utmost NTC benefits. This paper presents a comprehensive cross-layer reliability analysis framework to assess the effect of soft error, aging, and process variation in the operation of near threshold voltage caches. The objective is to quantify the reliability of different SRAM designs, evaluate voltage scaling potential of caches, and to find a reliability-performance optimal cache organization for an NTC microprocessor. In this paper, the soft error rate (SER) and static noise margin (SNM) of 6T and 8T SRAM cells and their dependencies on aging and process variation are investigated by considering device, circuit, and architecture level analysis. Their experimental results show that in NTC, process variation and aging-induced SNM degradation is $2.5{\boldsymbol \times }$ higher than in the super threshold domain while SER is $8{\boldsymbol \times }$ higher. At NTC, the use of 8T instead of 6T SRAM cells can reduce the system-level SNM and SER by 14% and 22%, respectively. Besides, we observe that we can find the right balance between performance and reliability by using an appropriate cache organization at NTC which is different from the super threshold.

Alpha-particle induced soft error rates (SER) due to naturally occurring radioisotopes in process metals are a serious reliability issue for deep-submicron CMOS technologies where continued scaling reduces the critical charge required to upset sensitive drains, in, for example, the off MOSET of an SRAM cell. With increased scaling the subsequent diffusion of charge from an ion strike covers a larger number of cells leading to increased multiple bit upset (MBU) rates in highly efficient designs. The amount of charge shared amongst neighboring nodes is critical in determining the overall SER. Charge sharing proceeds via two mechanisms: diffusion of carriers from the alpha strike to charge collecting nodes which depends on minority carrier diffusivity and the induction of charge on capacitive coupled nodes (V. Ferlet-Cavrois et al., 2003) as described by Ramo's theorem (S. Ramo, 1939). Results indicate the first mechanism to be dominant in terms of the quantity of charge collected which can lead to SER. Picosecond lasers have been widely used in simulating the energy-loss of MeV ions to better understand radiation phenomena by providing high-fidelity spatio-temporal information on charge collection processes (D. McMorrow et al., 1992), (J.S. Melinger et al., 1994). In this paper we describe the newly completed JPL picosecond laser microprobe for investigating SER and use it to investigate the temperature dependence of charge collection/sharing in a TSMC 0.18 mum CMOS "ring-like" test structure specifically designed for this task.

Characterizing, modeling, and analyzing soft error propagation in asynchronous and synchronous digital circuits