Title :
A robust method to estimate Power and Delay for Digital Integrated Circuits
Author :
Aghababa, Hossein ; Forouzandeh, Behjat ; Dehghan, Houman ; Afzali-Kusha, Ali
Author_Institution :
Sch. of Electr. & Comput. Eng., Univ. of Tehran, Tehran, Iran
Abstract :
Advancements in nano-scale Integrated Circuits manufacturing technology has resulted in variability of performance metrics. The performance parameters such as Power and Delay are no longer represented deterministically. As a result, circuit designers and manufacturers need to make use of statistical analysis to estimate performance of Integrated Circuits. In this paper we present a new methodology to increase the accuracy of estimation compared to prior methods. We introduce Bayesian analysis as a powerful mathematical and statistical approach to incorporate the prior observations in calculating the Probability Density Function (PDF) of performance parameters like Power and Delay. We apply this technique on a few Digital Gates and compare the results with previous methods. We also introduce Bayesian analysis as a powerful method to update the PDF of performance parameters. Finally, we demonstrate how this statistical approach could supersede the approaches established on Frequentist analysis so as to achieve a more accurate estimation on Power and Delay for Digital Integrated Circuits.
Keywords :
belief networks; delays; integrated circuit modelling; integrated logic circuits; logic gates; statistical analysis; statistical distributions; Bayesian analysis; delay; digital gates; digital integrated circuits; frequentist analysis; nanoscale integrated circuits; performance parameters; power; probability density function; statistical approach; Bayesian methods; Delay estimation; Digital integrated circuits; Integrated circuit manufacture; Integrated circuit technology; Measurement; Performance analysis; Probability density function; Robustness; Statistical analysis; Bayesian estimation; circuit performance; power and delay estimation;
Conference_Titel :
NORCHIP, 2009
Conference_Location :
Trondheim
Print_ISBN :
978-1-4244-4310-9
Electronic_ISBN :
978-1-4244-4311-6
DOI :
10.1109/NORCHP.2009.5397842