1 June 2016 in Microelectronic Engineering
10 August 2016 in IEEE Transactions on Electron Devices
ADAdrie Zegers-Van Duijnhoven
Adrie Zegers-Van Duijnhoven
In this paper, we develop a statistical model for random telegraph noise (RTN) related low-frequency noise (LFN). With our proposed model, one can calculate the expected value and the variability of the noise as a function of bias and device parameters. We clarify why the variability of RTN/LFN does not follow a 1/√area dependence. The model explains the effect of the halo implanted regions on the LFN statistics and the large variability of long channel devices found under certain bias conditions, which can be as large as that of short channel devices. We show that the LFN of n-channel and p-channel MOSFETs can be described by the same mechanism. From our results, we derive that the trap density of the p-channel device is a strongly varying function of the Fermi level, whereas for the n-channel the trap density can be considered constant. We validate the model through numerous experimental results from different CMOS nodes, down to 40 nm. We also demonstrate that the noise, and its variability, found in our measurements can be modeled using reasonable physical quantities.
21 April 2011 in IEEE Transactions on Nuclear Science
EFEvaldo Carlos Pereira Fonseca
Evaldo Carlos Pereira Fonseca
We exposed a flash-based FPGA to radiation to measure variations in current, temperature, propagation-delay and duty-cycle in logic circuits. Propagation-delay degradations vary from 400% to 1100% before functional failure, according to circuit and logical mapping. Electrical simulations are carried out to study the difference of behavior in the degradation of different logic mappings.
1 January 2018 in Microelectronics Reliability
1 November 2017 in Solid-State Electronics
8 December 2020 in IEEE Transactions on Electron Devices
With the downscaling of device dimensions, the variability of metal-oxide-semiconductor field-effect transistor (MOSFET) electrical behavior is produced by factors other than variations in physical dimensions and doping profiles, which are there since device fabrication and remain static over time. Besides these time-zero variability factors, factors that lead to performance variability from one instant in time to the other start playing a significant role. Random telegraph noise (RTN) is among these relevant time-dependent variability sources, causing the threshold voltage (VT) of a transistor to change from one instant in time to the other. In this work, we extend the knowledge of the time-dependent random variability induced by RTN, by providing a statistical model for RTN-induced threshold voltage variance over time. The area scaling of threshold voltage variance is detailed and discussed, supporting designers in transistor sizing toward a more reliable design. Not only the threshold voltage variance expected in a single transistor is modeled but also its variability among transistors that by design should behave the same way. It is shown that with device size downscaling, the variability of RTN among devices increases faster than its mean value. The relationship between the time domain (RTN) and frequency domain (low-frequency noise) is studied. The modeling provides equations to extract trap amplitude contribution and trap densities from the moments of a measured low-frequency noise distribution or from the moments of a measured VT variation distribution, without having to characterize individual step heights. Besides analytical modeling, Monte Carlo simulations are run, illustrating the RTN-induced time-dependent random VT variation and model applicability.