Analysis of radiation induced errors in transistors in memory elements

Author

Deekshitha Masani, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2020

Degree Name

Master of Science

Department

Electrical and Computer Engineering

First Advisor

Themistoklis, Haniotakis

Abstract

From the first integrated circuit which has 16-transistor chip built by Heiman and Steven Hofstein in 1962 to the latest 39.54 billion MOSFET’s using 7nm FinFET technology as of 2019 the scaling of transistors is still challenging. The scaling always needs to satisfy the minimal power constraint, minimal area constraint and high speed as possible. As of 2020, the worlds smallest transistor is 1nm long build by a team at Lawrence Berkeley National Laboratory. Looking at the latest trends of 14nm, 7nm technologies present where a single die holds more than a billion transistors on it. Thinking of it, it is more challenging for dyeing a 1nm technology. The scaling keeps going on and if silicon does not satisfy the requirement, they switch to carbon nanotubes and molybdenum disulfide or some newer materials. The transistor sizing is reducing but the pressure of radiation effects on transistor is in quench of more and more efficient circuits to tolerate errors. The radiation errors which are of higher voltage are capable of hitting a node and flipping its value. However, it is not possible to have a perfect material to satisfy no error requirement for a circuit. But it is possible to maintain the value before causing the error and retain the value even after occurrence of the error. In the advanced technologies due to transistor scaling multiple simultaneous radiation induced errors are the issue. Different latch designs are proposed to fix this problem. Using the CMOS 90nm technology different latch designs are proposed which will recover the value even after the error strikes the latch. Initially the errors are generally Single event upsets (SEUs) which when the high radiation particle strikes only one transistor. Since the era of scaling, the multiple simultaneous radiation errors are common. The general errors are Double Node Upset (DNU) which occurs when the high radiation particle strikes the two transistors due to replacing one transistor by more than one after scaling. Existing designs of SEUs and DNUs accurately determine the error rates in a circuit. However, with reference to the dissertation of Dr. Adam Watkins, proposed HRDNUT latch in the paper “Analysis and mitigation of multiple radiation induced errors in modern circuits”, the circuits can retain its error value in 2.13ps. Two circuits are introduced to increase the speed in retaining the error value after the high energy particle strikes the node. Upon the evaluation of the past designs how the error is introduced inside the circuit is not clear. Some designs used a pass gate to actually introduce the error logic value but not in terms of voltage. The current thesis introduces a method to introduce error with reduced power and delay overhead compared to the previous circuits. Introducing the error in the circuits from the literature survey and comparing the delay and power with and without introducing the error is shown. Introducing the errors in the two new circuits are also shown and compared with when no errors are injected.