Kelvin Bui
Search the website
 
   
  Papers/Reports  
 
Timing Analysis Inaccuracies Due to Multiple Input Switching and PVT Variations Rohan Patel, Kelvin Bui. Timing Analysis Inaccuracies Due to Multiple Input Switching and PVT Variations.
Department of Electrical and Computer Engineering, UCSD, La Jolla, CA, March 07
Abstract:
Statistical Timing Analysis (STA) tools such as Synopsys PrimeTime are optimized for speed and often linearize nonlinear problems or completely ignore aspects of complete timing analysis. As a result, PrimeTime neglects multiple input switching(MIS) which has been shown to create significant variations. At times, MIS can underestimate the delay by over 20% in some cases as demonstrated in referenced papers. In a timing analysis tool, the most critical path choice is made at a high level so each singular gate may or may not be a limiting factor. As a result, we cannot assume that each gate is responsible for the most critical timing. In this study, we will vary the relative delay between the input signals to a single gate on a critical path to see how the gate delay varies as a function of the delay between two input signals. We will also analyze voltage and temperature variations to understand the deviation from linear k-factor PVT simulations (180nm and before) and BC-WC guardbanding which has been in use since the 130nm technology node.
Image Processing Using the StreamIt Architecture Kelvin Bui, Jeff Moguillansky. Image Processing Using the StreamIt Architecture.
Department of Computer Science and Engineering, UCSD, La Jolla, CA, March 07
Autonomous Computing for Embedded Systems Kelvin Bui. Autonomous Computing for Embedded Systems.
Department of Computer Science and Engineering, UCSD, La Jolla, CA, Nov 06
Abstract:
The growing complexity of device infrastructures is outpacing the amount of labor we have to handle and manage that device’s interaction with the rest of its environment. Autonomic computing is a term first used by IBM in 2001 to describe a system that is primarily self-sufficient. The primary goal of self-sufficiency is to increase the efficiency of managing large complex systems without the intervention of human action. With conventional techniques, skilled technicians manually tweak the system for better improvements, and teams are formed to look for small hidden problems whenever they show up. This become unmanageable when the system is large and complex enough that no one can fully understand the system as a whole. We will explore and review the world of Autonomic Computing and see how it attempts to solve this alleviate this problem.
 
 
Download Resume (doc|pdf)