00329 Îf4634893f39e8884db8751e17aa760

332

McCarville & Montgomery

0.136772. This becomes a potential yield loss of 13.6772% due to the gage error.

A Linear Model Approach

Equations (9) and (14) are not "user friendly" models to optimize. ManuaÅ‚ manipulation of the formulas to find a minimum yield loss while achieving a targeted defect level goal would be difficult. There are several Computer software packages that could perform the computations; however, the computation time may be extensive, especially when several gages are in use. Simplified approximations of the functions are recommended so as to reduce computations as well as increase the ease of graphical methods for optimization. The approach we recommend is to generate Information about the defect levels and gage losses using a designed experiment and then optimize the system using a response surface modeling approach, approximations that would likely fit the complex model very nicely over the region of interest. From this linear model, an optimized operating region can easily be established.

Model Building

We will illustrate the response surface optimization approach for a two-gage system such as often found in the semiconductor industry. The following example utilizes two gages in series at the finaÅ‚ test operation prior to shipping to the customer. Two gages are used so as to reduce the level of defects being shipped to the customer due to gage error. The guard bands were originally chosen arbitrarily and adjusted when the customers reÄ…uested corrective action due to receiving reject units that passed at finaÅ‚ test.

The distribution of the product characteristic was determined using several months of finaÅ‚ test data and the gage distributions were approximated by performing gage capability studies as previously discussed. The distributions are as follows:

	Product	Gagel	Gage2
Distribution	Normal	Normal	Normal
Mean	500	GB1	GB2
StDev	100	30	20

The product specification limit is 650 and the corporate Ä…uality level of allowable defects passed to the customer is 100 ppm.