Methodology for analysis of noise susceptibility and design space exploration of integrated current sensors for detection of transient faults

Abstract

The semiconductor industry is being developed since the decade of 1960 in a remarkable fashion, by being able to follow the prediction that Gordon Moore stated at that time: the spatial density of transistors, the most fundamental components of microchips, would double at each technology node. Such law only began to be inaccurate at the date of thiswriting, and lead to an outstanding enhancement of functionality of integrated circuits. Despite the revolutions in human daily life the technology scaling produces, many are the threats to chips functionality in contemporary technologies. Numerous physical phenomena and very high spatial transistor density contribute to disrupt a circuits operation, so the development of detection and counteraction systems is of utmost importance. Since noise levels increase also accompany technology scaling, such systems are also endangered by those undesired signals.In this work, a system that senses potentially error-causing currents, called the modular Bulk Built-In Current Sensor (mBBICS), is submitted to noise in several simulations, and its behavior is studied. Two types of noise are considered in the study: switching noise and device noise. Switching noise is generated by switching circuits and propagated through the chips Silicon substrate, possibly affecting the mBBICS, and is directly related to the device density in the chip. Device noise takes place due to quantum effects inside the transistors channel. Hereby, switching noise was verified to be a threat to an mBBICS designed in a 90nm technology, whereas device noise is not high enough to pose a threat. However, with technology advancements, there is the possibility that device noise turns into an issue. In order to contribute to this possibility, a few parameters of the mBBICS were adjusted and its behavior in the presence of device noise, verified. More specifically, the circuits robustness to noise was analyzed, which was measured by how a sensitivity increase in the sensor leads also to vulnerability to noise. Nevertheless an increase in sensitivity necessarily causes an increase in susceptibility to noise, the performed methodology lead to optimal dimensional and electrical parameters, that highly favor sensitivity, whilepresenting little penalty in noise susceptibility. This empowers a designer to drive his or her circuit development based on the parameters design space that better leads to a robust behavior.