An improved manufacturing yield measure for gold bumping processes with very low nonconformities

Chia-Huang Wu, Wen Lea Pearn, Y. T. Tai*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

5 Scopus citations

Abstract

Nowadays, thinner and slimmer smartphones have been widely applied in daily life. In most portable devices, liquid-crystal display driver integrated circuit (IC) or new touch display driver IC are critical components, in which the gold bumping process is used as an essential interconnection technology and usually outsources to out-sourced assembly and testing factories. In semiconductor packaging manufacturing, due to popular high-definition requirements of display devices, gold bumping processes require very low fraction of nonconformities. For this reason, accurately evaluating the manufacturing yield is very important to guarantee the quality requirements being met. To evaluate the yield of gold bumping processes, the most popular yield index C-{\textrm {pk}} is widely employed in outsourcing contract arrangements. However, the typical existing nature estimator is a biased estimator and more conservative. In this paper, we provide an unbiased estimator to evaluate the gold bumping manufacturing yield more accurately. Based on the new estimator, we further propose a weighted parametric bootstrap method to obtain a reliable lower confidence bound that is closer to the true manufacturing yield. The new yield assessment can avoid underestimating the gold bumping manufacturing yield and can make more reliable decisions.

Original languageEnglish
Article number8624547
Pages (from-to)991-997
Number of pages7
JournalIEEE Transactions on Components, Packaging and Manufacturing Technology
Volume9
Issue number5
DOIs
StatePublished - 1 May 2019

Keywords

  • Gold bumping
  • Lower confidence bound (lcb)
  • Manufacturing yield
  • Unbiased estimator

Fingerprint

Dive into the research topics of 'An improved manufacturing yield measure for gold bumping processes with very low nonconformities'. Together they form a unique fingerprint.

Cite this