Exploring algorithmic error tolerrance of circuit simulators

• Main Author: Lim, Hui Hui.
• Other Authors: School of Computer Engineering
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: http://hdl.handle.net/10356/55040
• Institution: Nanyang Technological University

Integrated Circuit (IC) technology is getting more advanced and the number of transistors required for an Integrated Circuit (IC) is increases by a factor of 2 every 18 months. Not only that, the size of the transistor is getting smaller, from few micrometers to current tens nanometers, therefore the increment in transistor complexity and decrement in size of transistor and IC increase the probability of voltage glitches occurrence. This will cause errors in the circuitry since the transistor not to behave as per specification due to these frequent voltage glitches occurrence. The impact to the circuitry can be shown by introducing error inserting. This can be done by simulating circuit benchmarks using two different error insertion algorithms – error insertion by inserting constant error percentage which is to find the approximate range of the error tolerance and bit-flipping algorithm is to find out the actual impact on the result. For simplication, we are focusing on insert error into the matrices. This project aims to ascertain how much error tolerance that a simulator called “Simulation Program with Integrated Circuit Emphasis” (SPICE) can be tolerated. We use SPICE as a case study to analyase the impact of the hardware errors on software behavior. It is shown that error injected into the decomposed matrix is far more dangerous than matrix state before matrix decompostion. It also shows that the maximum error tolerance is less than probability value of 10*e^{−7}. Though the tolerance value is low, in fact, in reality, error does not occur that frequent. Therefore, the simulator is able to tolerate most errors in reality.