skip to main content


Title: Reliability and accelerated testing of 14nm FinFET ring oscillators
Accelerated lifetime tests are necessary for reliability evaluation of circuits and systems, but the parameters for choosing the test conditions are often unknown. Furthermore, reliability testing is generally performed on test structures that have different properties than actual circuits and systems, which may create inconsistencies in how circuits and systems work in reality. To combat this problem, we use ring oscillators, which are similar to circuits, based on the 14nm FinFET node as the circuit vehicle to extract wearout data. We explore the effects of testing time, sample size, and number of stages on the ability to detect failures for various test conditions, focusing on front-end time dependent dielectric breakdown, which is one of the most dominant wearout mechanisms.  more » « less
Award ID(s):
1700914
NSF-PAR ID:
10197912
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Design of Circuits and Integrated Systems
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Advanced FinFET SRAMs undergo reliability degradation due to various front-end and back-end wearout mechanisms. The design of reliable SRAMs benefits from accurate wearout models that are calibrated by accelerated test. With respect to testing, the accelerated conditions which can help separate the dominant wearout mechanisms related to circuit failure is crucial for model calibration and reliability prediction. In this paper, the estimation of optimal accelerated test regions for a 14nm FinFET SRAM under various wearout mechanisms is presented. The dominant regions for specific mechanisms are compared and analyzed for effective testing. It is observed that for our SRAM example circuit only bias temperature instability (BTI) and middle-of-line time-dependent dielectric breakdown (MTDDB) have test regions where their failures can be isolated, while the other mechanisms can’t be extracted individually due to acceptable regions’ overlap. Meanwhile, the SRAM cell activity distribution has a small influence on test regions and selectivity. 
    more » « less
  2. This paper proposes a methodology to find optimal accelerated test regions for lifetime parameter estimation for not only the traditional reliability concern, frontend-of-line dielectric breakdown (FEOL TDDB), but also the newly emerging wearout mechanism, middle-of-line time dependent dielectric breakdown (MOL TDDB) in 14nm FinFET technology. The framework to find the optimal test regions is introduced; the error estimating methodology is discussed in detail. Three digital circuits are presented for evaluation and comparison. The optimal test regions depend on the circuit size as well as the types of standard cells in the circuits. To ensure accurate lifetime parameter estimation, both error from sampling and error from selectivity should be considered at the same time. As a general guideline, higher estimation accuracy will be achieved by testing gate TDDB lifetime parameters at higher voltages, while testing middle-of-line TDDB at higher temperatures. 
    more » « less
  3. Because data from a variety of wearout mechanisms is confounded in circuits, we apply machine learning techniques to detect the parameters of competing failure mechanisms in ring oscillators, which more closely mimic circuit behavior than test structures. This is the first known application using data analysis to distinguish competing wearout mechanisms in circuit-level data. To quickly and efficiently analyze failure data, we propose to use maximum likelihood estimation to separately determine the parameters of each underlying distribution by only observing the time-to-failure of samples. The quasi-Newton method 
    more » « less
  4. Neuromorphic computing systems execute machine learning tasks designed with spiking neural networks. These systems are embracing non-volatile memory to implement high-density and low-energy synaptic storage. Elevated voltages and currents needed to operate non-volatile memories cause aging of CMOS-based transistors in each neuron and synapse circuit in the hardware, drifting the transistor’s parameters from their nominal values. If these circuits are used continuously for too long, the parameter drifts cannot be reversed, resulting in permanent degradation of circuit performance over time, eventually leading to hardware faults. Aggressive device scaling increases power density and temperature, which further accelerates the aging, challenging the reliable operation of neuromorphic systems. Existing reliability-oriented techniques periodically de-stress all neuron and synapse circuits in the hardware at fixed intervals, assuming worst-case operating conditions, without actually tracking their aging at run-time. To de-stress these circuits, normal operation must be interrupted, which introduces latency in spike generation and propagation, impacting the inter-spike interval and hence, performance (e.g., accuracy). We observe that in contrast to long-term aging, which permanently damages the hardware, short-term aging in scaled CMOS transistors is mostly due to bias temperature instability. The latter is heavily workload-dependent and, more importantly, partially reversible. We propose a new architectural technique to mitigate the aging-related reliability problems in neuromorphic systems by designing an intelligent run-time manager (NCRTM), which dynamically de-stresses neuron and synapse circuits in response to the short-term aging in their CMOS transistors during the execution of machine learning workloads, with the objective of meeting a reliability target. NCRTM de-stresses these circuits only when it is absolutely necessary to do so, otherwise reducing the performance impact by scheduling de-stress operations off the critical path. We evaluate NCRTM with state-of-the-art machine learning workloads on a neuromorphic hardware. Our results demonstrate that NCRTM significantly improves the reliability of neuromorphic hardware, with marginal impact on performance. 
    more » « less
  5. null (Ed.)
    To accurately determine the reliability of SRAMs, we propose a method to estimate the wearout parameters of FEOL TDDB using on-line data collected during operations. Errors in estimating lifetime model parameters are determined as a function of time, which are based on the available failure sample size. Systematic errors are also computed due to uncertainty in estimation of temperature and supply voltage during operations, as well as uncertainty in process parameters and use conditions. 
    more » « less