OPERA: optimization with ellipsoidal uncertainty for robust analog IC design

As the d esig n-m anu factu ring interface becom es increasing ly com plicated with IC technolog y scaling , the correspond ing process variability poses g reat challeng es for nanoscale analog / R F d esig n. D esig n optim ization based on the enu m eration of process corners has been wid ely u sed , bu t can su ff er from ineffi ciency and overd esig n. In this paper we propose to form u late the analog and R F d esig n with variability problem as a special type of robu st optim ization problem , nam ely robu st g eom etric prog ram m ing . T he statistical variations in both the process param eters and d esig n variables are captu red by a pre-specifi ed confi d ence ellipsoid . U sing su ch optim iza tion w ith ellipsoid a l u n certa in ty approach, robu st d esig n can be obtained with g u aranteed yield bou nd and lower d esig n cost, and m ost im portantly, the problem size g rows lin ea rly with nu m ber of u ncertain param eters. N u m erical ex am ples d em onstrate the ef...

• Type: Conference
• Year: 2005
• Where: DAC
• Authors: Yang Xu, Kan-Lin Hsiung, Xin Li, Ivan Nausieda, Stephen P. Boyd, Lawrence T. Pileggi
• Keywords: DAC 2005, Design Automation, Optim Iza Tion, Optim Ization, Robu St Optim