Spring00 smartgallery

Page 1

Defect Reduction F

E

A T

U

R

E

S

SmartGallery Technology for ADC Enhances Fab Productivity by Phil March, Verlyn Fischer, KLA-Tencor Corporation

The benefits of Automatic Defect Classification (ADC) are considerable: speed, reproducibility and reliability that do not fluctuate when the production shift changes.1 When defects are classified consistently, accurately and quickly, defectivity can be effectively monitored by type — a more valuable strategy than monitoring total defect count.2 However, as product life cycles continue to shorten throughout the industry, and as foundries require the ability to implement simple modifications of defectivity control strategies to respond to their rapid changes in product mix, minimizing the time and expertise required to set up the ADC system has gained importance.

Recently, KLA-Tencor introduced a new ADC system, IMPACT XP™, which incorporates an innovative, visually intuitive user interface that provides faster computation and significant improvements to ease of use. Prototype investigations were conducted at two large semiconductor manufacturing sites — one in the U.S. and one in Japan. At both these sites IMPACT XP was shown to reduce ADC setup time by a factor of three. The reduction in set-up time together with simplification of the set-up process, lowers the barriers to implementation of ADC. Opportunities for improving ADC

New requirements for the ADC system include not only faster set-up of the system, but also a reliable and accurate set-up process that results in production-worthy classifiers. Where do the opportunities for speeding up the ADC set-up time lie? Thorough training of the ADC system can require upwards of 1000 manually classified images per layer. A scheme for classifying the defects evolves as the defects are imaged, through a combination of automatic algorithms and input from the user. The user selects a set of representative defect images that are used for training the ADC system

and building a knowledge base of recognizable defects. After training the system the classifiers are verified, then optimized iteratively with reference to the calculated confidence levels for the classes. A team from KLA-Tencor examined each of these steps and focused on the classifier creation area, where 80 percent of the improvement was made. Upgrades to the ADC manager hardware accounted for the remaining 20 percent of the reduction time.3 New tool speeds ADC set-up

IMPACT XP was designed to meet the challenges of faster and easier ADC set-up. A key component of IMPACT XP is SmartGallery™, an innovative technology that allows users to look at images of all defect types from a given layer at the same time before defining the classification scheme (Figure 1). It’s intuitive user interface with drag-and-drop capability allows the user to view and compare multiple defect images simultaneously. This allows quick ADC classifier set-up and facilitates verification and optimization of classifiers through real-time feedback on classifier performance. SmartGallery combines the ability to view and manipulate a gallery of images with verification tools and real-time feedback on the confidence level of a given class. To simplify optimization of the classifiers, separate galleries are incorporated for the working image set and the training image set, with full sorting capability. Seeing all the relevant data in one place, simultaneously, strongly enhances the ability of the user to guide the Spring 2000

Yield Management Solutions

13


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.
Spring00 smartgallery by KLA Corporation - Issuu