Title :
Computer based methods for production tuning of microwave components
Author_Institution :
MIA-COM, Lowell Semicond. Oper., MA, USA
Abstract :
Three multivariate discrimination methods are applied as computer-based tuning algorithms during the manufacture of more than 9000 C-band power amplifiers. The three methods are parametric, nearest neighbor, and Monte Carlo discrimination. The nearest neighbor and Monte Carlo algorithms are found to be the most useful for tuning microwave components. Monte Carlo discrimination is judged to be the best method because it is nonparametric and handles specification changes gracefully. Application of the nearest neighbor and Monte Carlo methods improved the rate of successfully reclaiming initially failing parts from 51% to 82% as compared with manual tuning techniques.<>
Keywords :
Monte Carlo methods; microwave amplifiers; power amplifiers; production engineering computing; solid-state microwave circuits; C-band power amplifiers; Monte Carlo discrimination; computer-based tuning algorithms; initially failing parts; microwave components; multivariate discrimination methods; nearest neighbor; parametric discrimination; production tuning; specification changes; Computer aided manufacturing; Covariance matrix; Gaussian distribution; Iterative algorithms; Microwave theory and techniques; Monte Carlo methods; Nearest neighbor searches; Power amplifiers; Production; Semiconductor device manufacture;
Conference_Titel :
Microwave Symposium Digest, 1993., IEEE MTT-S International
Conference_Location :
Atlanta, GA, USA
Print_ISBN :
0-7803-1209-0
DOI :
10.1109/MWSYM.1993.276927